>> Study Design, Content, and Administration 1992 PRE-POST STUDY DESIGN The 1992 National Election Study entailed both a pre-election interview and a post-election re-interview. Approximately half of the 1992 cases are comprised of empaneled respondents who were first interviewed in the 1990 National Election Study and later in the 1991 Political Consequences of War Study. The other half of the cases are a freshly drawn cross-section sample. (Details of the sample design are given in "Sample Design of the 1992 Pre- and Post-Election Study", below.) The panel component of the study design provides an opportunity to trace how the changing fortunes of the Bush presidency, from the high levels of approval at the start of the Gulf War, through the decline after the onset of a recession, affected voting in the November 1992 presidential election. It also permits analysts to investigate the origins of the Clinton and Perot coalitions as well as changes in the public's political preferences over the two years preceding the 1992 election. Altogether, 2485 citizens were interviewed in the 9 weeks prior to the November 3, 1992 election. [Note: The original study Staff release of the 1992 National Election Study in April, 1993 contained 2,487 cases. See the note on "A Note on Deletion of Cases", below, for further information about the two cases deleted from this edition of the collection.] To permit analysis of the impact of the unfolding election campaign, a random half of the sample was released to the field on September 1 and the other half on October 1st. 1359 of the pre-election interviews were conducted with panel respondents; 1126 with cross-section respondents. In the weeks following the election, 2255 pre-election respondents were reinterviewed; 1250 panel, 1005 cross-section. Further details of the administration of the surveys are given in "Study Administration", below. The two components of the study -- the panel and the new cross-section -- were designed to be easily used together to create a combined nationally representative sample of the American electorate. Several case weights are provided with this data set. V3008 (which incorporates sampling, nonresponse and post-stratification adjustments) should be used when analyzing the combined sample (the panel and the new cross-section respondents). V3009 (which incorporates sampling, nonresponse and post-stratification adjustments) should be used when analyzing the panel respondents alone. V7000 (which corrects for panel attrition and the aging of the panel respondents, but does not incorporate sampling, nonresponse and post- stratification adjustments) should be used when comparing either the panel respondents or the combined panel and new cross-section respondents to previous (unweighted) National Election Studies data collections. See "Sample Design of the 1992 Pre- and Post-Election Study", below, and the documentation for V3008, V3009, and V7000, for further information. STUDY CONTENT; SUBSTANTIVE THEMES The content for the 1992 Election Study reflects its double duty, both as the traditional presidential election year time-series data collection and as a panel study. The substantive themes represented in the 1992 questionnaires include: * interest in the political campaigns; concern about the outcome; and attentiveness to the media's coverage of the campaign * information about politics * evaluation of the presidential candidates and placement of presidential candidates on various issue dimensions * partisanship and evaluations of the political parties * knowledge of, contact with, and evaluation of House candidates (including questions on how their Representative voted on the Persian Gulf War resolution and whether he/she was implicated in the House banking scandal) ; opinions on term limitations * political participation: turnout in the Presidential primaries and in the November general election; other forms of electoral campaign activity * vote choice for President, the U.S. House, and the U.S. Senate, including second choice for President * personal and national economic well-being, with particular attention to the impact of the recession * positions on social welfare issues including: social security; government health insurance; federal budget priorities, and the role of the government in the provision of jobs and good standard of living * positions on social issues including: abortion, the death penalty; prayer in the schools; the rights of homosexuals; sexual harassment and women's rights * racial and ethnic stereotypes; opinions on school integration and affirmative action; attitudes towards immigrants (particularly Hispanics and Asians); opinions on immigration policy and bilingual education * opinions about the nation's most important problem and the most important issues discussed during the local congressional campaign * political predispositions: moral traditionalism; patriotism; political efficacy; egalitarianism; individualism; trust in government; racial prejudice; and feminist consciousness * social altruism and social connectedness * assessments of U.S. involvement in the Persian Gulf War and of U.S. foreign policy goals * feeling thermometers on a wide range of political figures and political groups; affinity with various social groups * detailed demographic information and measures of religious affiliation and religiosity Congressional Ballot Cards, Candidate Lists, and Candidate Numbers In the usual NES Post-Election survey, and for 1992, in the Pre-Election survey as well, respondents are asked several questions about their particular Congresspersons and Senators. Interviewers pre-edited questionnaires to fill in the names appropriate for the state and congressional district in which the respondent was living (or was living during the pre-election interview). Each candidate and Senator is assigned a unique number that reflects his or her incumbency status and party. (See Candidate Number Codes and Lists). Particular questions in the survey require the insertion by the interviewer during pre-editing of the names of candidates. See, for example, post- election question B1, which includes feeling thermometers for the various candidates. The Candidate Lists used by the interviewers, which show which candidates are associated with which congressional district and with which numbers they are tagged, can be found in Appendix F. Asking questions about incumbent candidates is somewhat more problematic in a year when redistricting occurred, and for the Pre-Election survey there is the additional complication that a number of states held their Congressional primaries after the Pre-Election field work had started. Further details can be found at the documentation for Pre-Election questions J10-J11. Handling of Congressional Incumbency Where Redistricting has Occurred Throughout, whenever the word "incumbent" is used, its referent is a representative who was a member of the 102nd Congress; i.e., the Congress in session prior to the November 1992 General Election. Due to redistricting as a result of the 1990 U.S. Census, any given incumbent's district for the 103rd Congress may consist of a fairly different geographical area from the area covered by the district prior to the boundary changes. Therefore, prior to 1992, the "incumbent" may or may not have been the representative for the particular piece of geography (the sample segment or census tract) in which the respondent lives. For each sample segment, we have included in the dataset its 1992 congressional district number, v3019, and its congressional district number in 1990, v3020. By comparing the two, it can be determined whether the "incumbent" in question was actually the respondent's incumbent prior to the 1992 general election. "Lagged" Measures Obtained from 1990 and 1991 Interviews Slightly more than half of the respondents in the 1992 study were also interviewed in 1990 and 1991. Therefore, all of the variables associated with the 1990 Post-Election Study (ICPSR 9548) and the 1991 Political Consequences of War Study (ICPSR 9673) are available for use as "lagged" measures in the current release of this collection. STUDY ADMINISTRATION Pre-election Study Release of Sample To permit analysis of the impact of the unfolding election campaign and to minimize the relationship between interviews taken late in the campaign period and the difficulty of obtaining an interview, NES divided the Pre-Election study sample into two random parts. Administration of the first random half occurred between September 1 and September 30; the second half between October 1 and October 31st, with the first two days of November as "cleanup." The two part division applied to both panel and cross-section samples. Note that the study period began before Labor Day, the traditional start of the Election Studies (and Presidential campaigns). The combination of a late date for Labor Day (Sept. 7) and an early date for Election Day (Nov. 3rd) would have shortened our standard field period by about a week, which would have reduced the overall response rate. Sample "Replicates" To more closely tailor the field effort to the actual sample performance during this study, both parts of the sample (panel and cross-section) were randomly subdivided into five replicates, each of which is a proper, random subsample of the NES sample. Replicates 1 and 2 were considered the "base sample," certain to be released, with three replicates being held in reserve to be released for fieldwork October 1, 1992, if it was decided they were needed. Replicates 4 and 5 were released at that time. Survey Modes: Design and Implementation One of the administrative problems in fielding a panel study is that respondents have had an intervening period of time in which to relocate, perhaps at some remove from areas where field staff is maintained. Additionally, some of the SRC sample primary areas were replaced between 1990 and 1992, and therefore potentially some of the 1990 Election Study respondents lived in areas where SRC interviewers were no longer on staff. We estimated that between 50 and 125 respondents might have moved to areas in which SRC did not have interviewers, or might be living in their 1990 residence, in a place where SRC no longer maintained interviewing capability. (As it turned out, the total number of panel respondents that we interviewed who were "out of range" for either of these two reasons was 43.) It was our intention to interview as many panel respondents as possible, but we did not want to incur the additional costs associated with interviewer travel. Therefore, we prepared a truncated version of both Pre- and Post-Election Survey questionnaires, (the "Short-Form") to be administered over the telephone to those panel respondents who had moved out of range. Interviews, both in the Pre- and in the Post Election surveys, were also administered over the telephone to many respondents, both panel and cross-section, who did not meet the "panel out-of-range" criteria for telephone interviewing. The mis-implementation of the design also entailed the inappropriate use of the full-length questionnaire. Table 7, below, sums up the situation. In total, 86 percent of the interviews (91 percent before the election and 81 percent of those conducted after the election) were administered as mandated by the study design: face-to-face with the full length questionnaires or by phone for those panel respondents who moved out of range. A NOTE ON DELETION OF CASES In putting together the panel file, study staff examined with particular attention the work of one interviewer and decided that his entire production for 1990 was suspect. Two panel reinterviews in 1992 were thus based on 1990 interviews which were very likely faked in whole or very large part. The decision was made to eliminate these interviews from the 1992 dataset (and also from the panel file). Consequently, the total N for the ICPSR release of these data is 2485 as compared with a N of 2487 in the Study Staff release of the 1992 Cross-Section data. The tables found in this introduction were produced using the original Study Staff release of the data and reflect the original N of 2487. Table 7: Mode and Form Administration in the 1992 Pre-/Post Election Studies Panel Respondents Mode Questionnaire Pre-Election Post-Election Face-to-face(A) Full 1155 84.8% 951 76.%1 Phone(B) Short 149 11.0% 186 14.9% Phone Full 57 4.2% 113 9.0% Subtotal 1361 100.0% 1250 100.0% Cross Section Respondents Mode Questionnaire Pre-Election Post-Election Face-to-face(C) Full 1053 93.6% 830 82.6% Phone (D) Short 5 .4% 4 .4% Phone Full 68 6.0% 171 17.0% Subtotal 1126 100.0% 1005 100.0% Total Respondents Mode Questionnaire Pre-Election Post-Election Face-to-face Full 2208 88.8% 1781 79.0% Phone Short 154 6.2% 190 8.4% Phone Full 125 5.0% 284 12.6% Total 2487 100.0% 2255 100.0% A. The 1155 Pre-election respondents in this category include 16 Panel interviews taken F-T-F using the Spanish version of the questionnaire. B. The Pre-election respondents in this category include 1 Spanish language panel interview, taken by phone. C. The pre-election total includes 4 Spanish version questionnaires taken F-T-F. D. The 5 cases in the Pre-election category consist of 1 F-T-F and 3 Phone short-form, plus 1 Spanish language cross-section case. SURVEY FORMS: DESIGN AND IMPLEMENTATION There were two [5] forms of both the Pre- and the Post- Election Study questionnaire: a short form, to be administered over the phone to panel respondents who were "out of range," as described above, and a standard, or full-length questionnaire to be administered to everyone else. The questions on the short-form were a subset of those on the full length questionnaires whose 70 minutes in length was thought to be unacceptably long for a telephone interview. 50 minutes worth of content was selected for the short form, both Pre- and Post-Election Surveys. The criteria for inclusion were that the questions were "core," i.e., questions part of the NES time-series, as opposed to recently piloted or topical items, or that they related to the focus of the 1991 Political Consequences of War Study. We decided not to repeat most of the demographics items for the approximately 100 panel respondents we expected would be interviewed with the short form, relying instead on their responses in the 1990 survey. Additionally, some congressional content was deleted, because of the difficulty in assigning respondents over the phone to the newly drawn congressional districts. Because we estimated the number of cases affected to be few and randomly scattered across the country, we did not design the instrument for the telephone. Except for the income question, we made no adjustments to the questionnaire for the difference in mode. In general, interviewers were expected to read response options to the respondent and to repeat them as necessary until they were clear to the respondent. All interviews with a short form questionnaire, except for Spanish language, and including "legitimate" or "out-of-range" panel respondent interviews, have been designated as partial interviews, in the result code variables for the Pre- and Post-Election Studies (v3033 and v5012). EVALUATION OF PROBLEMS IN STUDY IMPLEMENTATION The problems mentioned above did not become fully evident until coding was virtually completed, in the last week of February. At its March 1 meeting, the NES Board of Overseers, to whom these problems were reported, instructed the Principal Investigators to assess the significance of these problems with respect to data quality. This work was carried out by the Principal Investigators and members of the Study Staff in consultation with Board members, SRC methodologists and Center for Political Studies personnel as appropriate. The findings are available in NES Technical Report No. 43, available from NES Project Staff. As the Technical Report documents in detail, the inappropriate use of the telephone and the short-form questionnaire thankfully had only a negligible impact on the quality of the 1992 data. When the short-form questionnaire was used, it of course generated missing data on those items that appeared on the full-length questionnaire but not on the short-form. But this resulted in a very slight increase (less than .05 percentage points) in the standard errors of the affected variables. The pattern of missing data (from use of the short-form questionnaire) is unrelated to the demographic or political characteristics of respondents. Instead, interviewers turned to the short form when it appeared they would have difficulty securing an interview for other reasons having to do with the field administration of the study. The same holds for use of phone instead of face-to-face interviewing. Respondents interviewed over the phone are politically indistinguishable from those interviewed face-to-face. Attributes of the study administration, not attributes of the individual respondents, are associated with the propensity of interviewers to conduct some of their interviews over the phone. Finally, although some survey questions perform differently across the two modes of interviewing, the distribution of responses and the relationship among variables are substantively the same among phone and face-to-face respondents. RESPONSE RATES The Pre-Election study response rate for the cross section sample was 74.0%. Recalculating the response rate to eliminate 4 short-form, cross-section interviews (partials) results in a response rate of 73.7%[6]. For the panel sample, the response (or reinterview) rate is 77.7% when partials, or short form interviews, are included, but drops to 69.2% when they are excluded. Post-Election reinterview rates are 91.8% for the panel, including partials, and 85.0% excluding the partial or short-form interviews. The cross-section Post-Election reinterview rate was 89.3% including 4 partials; 88.9% excluding them. These calculations do not differentiate between face-to-face and telephone modes of interviewing. INTERVIEW COMPLETION RATE Table 8 lays out the number of interviews taken for each week elapsing after the Nov. 3 General Election. In 1992, 25.8% of the interviews were completed in the first two weeks after the election; 53.1% in the first four weeks. For comparison, in 1988, 55% of the interviews were taken in the first two weeks after the election, and 82% in the first four weeks. Table 8: Number of and Cumulative Percent of Interviews Taken in the Post-Election Study by Week of Interview DATES NUMBER OF CUMULATIVE CUMULATIVE INTERVIEWS NUMBER OF PERCENT OF INTERVIEWS INTERVIEWS Nov. 4-Nov.10 237 237 10.5% Nov.11-Nov.17 344 581 25.8 Nov.18-Nov.24 372 953 42.3 Nov.25-Dec. 1 245 1198 53.1 Dec. 2- Dec. 8 348 1546 68.6 Dec. 9-Dec.15 278 1824 80.9 Dec.16-Dec.22 175 1999 88.7 Dec.23-Dec.29 86 2085 92.5 Dec.30-Jan. 5 125 2210 98.0 Jan. 6-Jan.13 45 2255 100.0% VARIABLES SUPPRESSED FOR REASONS OF CONFIDENTIALITY Starting with the 1986 Election Study, NES has released occupation code variables in somewhat less detail than in years past. This dataset includes a two-digit code with 71 categories corresponding to Census Bureau occupational groupings. Those who need the full occupation code for their research should contact the NES project staff for information about the conditions under which access may be provided. Similarly, the National Election Studies have not included information for census tracts or minor civil divisions since 1978. Permission to use the more detailed geographic information for scholarly research may be obtained from the Board of Overseers. More information about this is available from NES project staff. Coding of the new religious denomination variable is in some cases based on an alphabetic "other, please specify" variable. This variable is restricted for reasons of confidentiality, but access may be provided to legitimate scholars under established NES procedures. OPEN-ENDED MATERIALS Traditionally, the National Election Studies have contained several minutes of open-ended responses (for example, the candidate likes and dislikes). These questions are put into Master Codes by the SRC coding section. Other scholars have developed alternative or supplemental coding schemes for the questions (for example, the levels of conceptualization, released as ICPSR 8151). The Board of Overseers wishes to encourage these efforts but in ways which respect the NES and SRC obligation to protect the privacy and anonymity of respondents. Circumstances under which individuals may have access to transcribed versions of these questions have been worked out and those interested should contact the NES project staff for further details. Table 1: Field Administration Information Response Rate: 71.4% Length of Interview: 78.0 min No. of Respondents: 2000 - Table 2: Number and Cumulative Percent of Interviews in Two-Week Intervals from Election Day, 1990 Nov. 07-Nov. 17 836 42% Nov. 18-Dec. 01 594 72% Dec. 02-Dec. 22 413 92% Dec. 23-Jan. 05 106 97% Jan. 06-Jan. 26 51 100% NOTES {There are no notes [1] - [4]} [5] There were actually three forms of both questionnaires, since they were translated in Spanish. The Spanish language questionnaires are also "short-form" since only core items were translated. They are not, however, treated as "short-form" for "partials" for the purpose of this discussion. [6] The denominator for the calculations in this paragraph are as given in Tables 14 and 15 this Introduction. Information about the numerators appears in Table 7. [7] Text prepared by the Sampling Section of the Survey Research Center, Institute for Social Research, University of Michigan, March, 1993. [8] While the Panel segments were selected from the 1980 STF1B file, most of the Cross-section segments were selected from the nearly equivalent 1990 Census file (PL94-171 file on CD ROM) which contains the block-level 1990 Census housing unit (HU) data. At the time of selection the 1990 STF1B file was not available. Therefore, the PL94-171 file was used, which had "total HU's" (rather than "occupied HU's") per block; for these Cross-section segments, linkage was designed to achieve a minimum measure of 72 TOTAL HU's per SSU. Also, since in 1990 all areas had been divided into Census Tracts and blocks, no Enumeration Districts were involved as SSU's. In other respects the second stage selection was the same for both sets of area segments. [9] See Note 3. [10] The 1986 NES was the most recent NES sample using the two-thirds National Sample. Response rate in 1986 was .701 and occupancy eligibility rate was .835. [11] Based on field experience in 1986 NES study. [12] About 55% of the base sample was assigned to the first release, September 1, 1992. [13] Released to field October 1, 1992. [14] All "reserve" replicates were to have coversheets sent to the field October 1, 1992, in sealed envelopes which were not to be opened by the interviewers until notified of their "release". As it happened, it was decided to release Replicates 4 and 5 on October 1, 1992. Replicate 3 was never released. (However, a few cases from Replicate 3 were released by mistake; these cases can be identified by using variables 3023 and 3024.) [15] An overall Panel response rate of 75% was assumed. Based on recontact response to the 1991 Persian Gulf Study: 1385 cases at 87% response rate = 1205 cases, and 615 cases at 50% response rate = 308 cases. Therefore, Overall: 1513/2000 = .756 [16] See Note 12. [17] Based on 1986 NES field experience using the two-thirds National Sample (.835). [18] No provision of update growth was applied in early estimates. Since the updating process was applied to the cross-section component of the 1992 NES Sample, and since it typically produces about 3% increase in sample lines over the count selected from the National Sample system, the update inflation factor was set at 1.03 for the cross-section component. [19] One percent of the sample was lost due to subsampling in three locked and two dangerous areas. [20] An overall Panel response rate of 75% was assumed, based on previous recontact experience (response to the 1991 Persian Gulf Study): 1385 cases at 87% response rate = 1205 cases, and 615 cases at 50% response rate = 308 cases. Overall: 1513/2000 = .756 [21] This figure was left without applying the usual growth factor for updating to the cross-section component of the sample, since this was the table presented (see Table 11) in the original planning for the study. The equivalent figure for the actually released Replicates 1,2,4 and 5) was taken with the growth factor of 1.03 applied to the cross-section component only. [22] In constructing the analysis weight, a maximum of three eligible adults was allowed. [23] For cross-sectional analysis of the 1992 NES data the combined cross-section and panel data must be used. Cross- section component data cannot be used alone. [24] The design effects from the 1988 NES are expected to be similar to those for the 1992 NES. Sampling errors for the 1992 NES have not yet been run. [25] The standard error of a percentage is a symmetric function with its maximum centered at p=50%; i.e., the standard error of p=40% and p=60% estimates are equal. >> Study Design, Content, and Administration 1993 PILOT SURVEY CONTENT AND OBJECTIVES Overview The 1993 Pilot Study is the second of a projected three wave study. The 1993 wave was in the field approximately one year after the first wave of the study which is the 1992 Pre- and Post-election study, from which the 1005 cross-section respondents were selected for reinterview in 1993. We anticipate that respondents will be interviewed for a third time as part of the 1994 Election Study. The three-wave study is designed to exploit the special circumstances of the 1992-94 elections: a minority president who is struggling to forge a majority coalition in the face of a strong third-party challenge, and the replacement in 1992 of fully one-quarter of the House of Representatives. Each presents an unique opportunity which we propose to seize through projects that are directed at understanding how electoral coalitions form (and decay) and how new members of the House secure their districts. Additionally, the Pilot Study fulfills its role as the vehicle for testing and developing new instrumentation for the 1994 National Election Study. The Clinton Coalition The 1994 elections present both a substantial opportunity and risk to the Democratic Party. The stakes are high: the party needs to consolidate the gains of 1992 and build a majority coalition. In some ways, the Clinton Administration began this political task from a position of extraordinary weakness. Although Bill Clinton captured a clear majority of the electoral votes, he entered the White House without a clear mandate, winning just a shade over 43 percent of the popular vote. Indeed, early interpretations of the 1992 election have emphasized less that Clinton won the Presidency and more that Bush lost it. At the same time, whether in possession of a popular mandate or not, Clinton came to Washington with significant legislative initiatives in mind. He introduced major proposals on taxes and spending. He appears determined to grapple with health care, not to tinker with it but to reform it fundamentally. Clinton's election has of course meant the return of unified government to the national scene, though early readings suggest that Republican unity in the Senate and Democratic defections from Clinton's proposals may undermine the promises of unified control. Still, there is the prospect of real change: major proposals, passed into law, with the consequences broadly felt throughout the country. From the perspective of coalition maintenance, this is a special political moment, one portentous for the future electoral success not only of the Democratic and Republican Parties but for third party challenges as well (a point we take up immediately below). We want to assess how all this consequential and high-profile political churning intrudes upon Clinton's capacity to hold together and expand his political coalition over the first critical years of his administration. How have each of Clinton's major policy initiatives added or subtracted support from his political coalition? The 1993 Pilot Study re-asks a number of items from the 1992 Study, and adds others, to give as complete a picture as possible of how Clinton is faring with the coalition which elected him. These items are: Evaluation of economy (V7238-7260) Approval ratings of several aspects of Clinton's performance in office (V7101-7120) Thermometer ratings of Bill and Hillary Clinton (V7130-7138) Who would R vote for if the election were held today (V7161) Liberal-conservative placement of Clinton (V7209- 7216) Traits and affects batteries (V7226-7230, V7267- 7270) Opinion on NAFTA (V7261-7266) Opinion on budget deficit (V7315-7323) From a slightly different angle, the 1992-1994 study, of which the 1993 Pilot Study is the middle piece, is also directed at more fully understanding the Perot phenomenon. That Perot's popularity is a political phenomenon is hardly open to question. Following an eccentric if not quixotic on and off and on again campaign, and in spite of the formidable hurdles which the American system places before third-party candidates, Perot won nearly one in five votes cast in 1992. In this respect, Perot did better than all but one third party candidate since the Civil War split the nation. Perot's pockets are deep enough to finance a continued high public profile. Perot's likely continued presence quickens interest on our part in understanding the maintenance and decay of his coalition as well. Even without the trappings and formal powers of the Presidency, Perot, like Clinton, faces the identical political problem of somehow hanging on to his supporters while recruiting still others as they become disenchanted with the alternatives. To what extent does Perot's continued support rest upon an ideological base? Or upon disenchantment with business as usual, a continuing protest against politics itself? Or upon the failure of government to deal with the economy or the budget deficit? Or should the Perot movement be understood in more personal terms, dependent upon continuing public displays of a winning style and personality? Or, finally, does it turn on contempt for the alternatives? A number of items which attempt to tap the sources and strength of Perot support have been included in the study. They include: Ross Perot and United We Stand feeling thermometers (V7131, V7149, V7150) Liberal-conservative placement for Perot (V7220- 7221) Traits and affects batteries (V7231-7235, V7271- 7274) Attitudes toward political parties ((V7295-7296, V7305, V7366-7370) Attitudes toward media, special interests, government in Washington (V7306-V7308) Membership in, contact by United We Stand America (V7312-7314) To examine the maintenance and decay of electoral coalitions, we have empaneled the cross-section respondents to the 1992 NES Post-Election Survey, interviewing them again in the fall of 1993, and proposing to interview them one final time in the weeks following the 1994 midterm election. The panel design is a powerful one for several reasons. First, an absolute requirement for a study of electoral coalitions is the successful identification of Clinton, Perot, and Bush voters (and non-voters as well). For Clinton, the immediate political challenge has several aspects: to maintain the support of those who voted for him in 1992; to build support among those who voted for his opponents, especially those who went Perot's way in 1992; and to awaken interest and eventually support among those millions who, in 1992, voted for no one at all. Attempting to assess vote a year or more away from the election, as we would have to do absent a panel design, invites error of the most pernicious sort. For example, citizens who in fact voted for Clinton in 1992 but who have since recoiled in horror at what he has done, might now report that they had voted for Bush. To get this project off the ground, we need to know what citizens did on election day 1992, and to know that, we treat the 1992 NES Survey as a first wave of a panel. Second, coalition maintenance and decay may be a classic case of little detectable movement at the aggregate level obscuring lots of off-setting movement at the individual level, as citizens move in and out of various partisan camps. Determining the fluidity of the Clinton and Perot coalitions can be uncovered with panel evidence. Finally, panel data will also permit the testing of alternative theories of political learning. Whether such theories come from formal, statistical formulations, as in Bayesian models, or from various psychological theories, a claim held in common is that what people absorb from their political experiences depends on their prior beliefs and sentiments. Learning is conditional on what citizens already know. This means that we must have baseline readings on citizens before Clinton's coming to power. The 1992 NES survey of course delivers handsomely on precisely this point. These data tell us what citizens thought in 1992 about the necessity of new taxes, the seriousness of the federal budget deficit, the need for health care reform, the conditions under which women should be permitted to have abortions, whether gays should be allowed to serve in the armed forces, the responsiveness of government institutions, the performance of the major parties, and much, much more. And this means that, having returned to these same citizens in 1993 and 1994, we will be in excellent position to understand in a fine-grained way how electoral coalitions are held together and how they fall apart. Securing the District Due to a combination of re-districting, scandal, and retirement, the 1992 House elections resulted in a dramatic turnover in personnel. More than one-quarter of the House was replaced: 110 new Representatives won in November, the most in nearly half a century This turnover provides an the opportunity of examining the ways in which new members of the House secure their districts against challenge in the next election. For the first time, we can examine the relationship that develops between representatives and their constituents in its formative stages during the first term in office. The advantages of incumbency have been a central theme of research on House elections and on the institution itself. Defections from party-line voting in House elections have increasingly favored the incumbent. These days, incumbent Representatives almost always win, often by overwhelming margins. Despite all the talk about anti-incumbent feelings in 1992, fully 93 percent of House incumbents seeking re-election were returned to office. Taking into account primary election defeats, this figure remains an impressive 88 percent. On the other hand, this re-election rate was the lowest since the Watergate election of 1974 and fell just 2 points short of being the lowest in forty years. Moreover, it does not take into account the unusually large number of representatives who choose not to run again in 1992, some of whom certainly would have been defeated. It is also true that winning incumbents were much more likely to find themselves in close contests in 1992 than in previous years. Still, in the face of re-districting, scandal, and widespread popular disdain for the institution of Congress, incumbents seeking re-election were rarely turned away. Success at under these highly unfavorable conditions testifies to the continuing electoral benefits of incumbency. We know that incumbent advantage accrues quickly: it is well-established, perhaps established in full, by completion of the first term in office. Indeed, a common measure of incumbency advantage is the "sophomore surge:" the gain typically registered in the representative's first re-election try. What happens during these first two years? How do newly elected members of the House consolidate their victories? Is the incumbency advantage secured as a result of the actions that members of Congress engage in during their first term of office, or is it secured as a result of their first re-election campaign? As it is typically investigated, the problem is impossible to unravel. The data we rely on are always investigated in the context of an election campaign. Moreover, it is precisely those incumbents who are deepest in trouble at election time who work their district the most. The study we propose here provides a clean test of the inherent (as opposed to campaign-related) advantages of incumbency. Many new members are precarious, and most no doubt believe that they are. Under these circumstances, do in fact new members of the House concentrate their attention and activities on their home district during their first term and, most important, do their constituents take notice? As a general matter, we know next to nothing about the impressions created by Representatives -- whether they are new to Congress or not--between elections. What in fact happens to the visibility of newly-elected representatives over the critical period of their first term? Do constituents tend to forget about their representatives between elections, and then learn about them again as the next campaign takes off? Or do constituents learn more and more about their representatives as the first term proceeds, a response to what Richard Fenno has called "the permanent campaign?" The 1992-1993-1994 panel data provide sharp tests of the alternative theoretical interpretations of the incumbency advantage. Of the 1005 respondents who make up the 1992 NES post-election cross-section, over a quarter (n=275) resided in congressional districts that sent a new member to Congress in 1992. Thus, the high turnover that occurred in the House in 1992 provides sufficient numbers of respondents to support detailed analysis of the processes by which newly-elected representatives (compared to returning incumbents) shore up their support during their first term in office. The panel design provides efficient measurement of the evolution of new Representatives' reputations among their constituents. With panel evidence in hand, patterns of learning and forgetting and alterations in trust and support, conditional on the views held by constituents before their Representatives went off to Washington, can be traced. The survey included extensive content on evaluations of incumbent members of Congress. Much of the content repeats the now-familiar congressional batteries. Also embedded in the study is an experiment designed to give us more information about whether the use of the ballot card has contributed to over-reporting. Half of the respondents were supplied with the names as well as parties of the candidates for congress when asked for whom they voted. This emulates the ballot card. The other half of the respondents were simply asked whether they voted for the Democrat or the Republican candidate. Recall of candidates running in "this district this past November" (V7121-7129) Thermometer rating of incumbent; recall what job he/she holds? (V7136-7137) Likes/dislikes of incumbent (V7162-7173) Contact with U.S. Representative incumbent (V7174-7183) Vote for Representative (V7184-7185) Approve of way Representative handling job (V7191-7194) Does R's representative support Clinton's legislative proposals (V7195-V7199) Did he/she vote for Clinton's deficit reduction package (V7200-7202) Does Representative do a good job of keeping in touch (V7203) Liberal-conservative placement of Representative (V7222-7223) Developing New Instrumentation The design of the 1993 Pilot Study replicates one NES successfully implemented in 1990-91-92 to assess the political impact of the Persian Gulf War. In this design, the odd- year Pilot Study serves double duty as a platform both from which to conduct the second wave of the panel and to carry out the research and development work for the subsequent year's election study. One section of development work (variables 7371- 7422) follows a proposal made by Laura Stoker, to study the interest basis of political attitudes. Questions are asked about perceived interests of several groups (wealthy, poor, middle class, blacks, whites), as well as self and national interest, in three domains: National health insurance (V7374-7384) Affirmative action (V7405-7422) School choice (V7385-7404) Half of the respondents received the questions about affirmative action in lieu of the school choice questions while the other half got the school choice questions instead of those relating to affirmative action. Douglas Strand proposed a number of questions relating to attitudes toward homosexuals and about policies affecting homosexuals. The attitudes toward homosexuals are measured by asking Rs whether: Parents should encourage boys to be masculine and girls to be feminine (V7289-7294) Homosexuality is a matter of choice (V7336-7339) Homosexuals try to seduce non-homosexuals (V7340- 7343) The idea of homosexuality disgusting or uncomfortable (V7348-7351) He/she worries about getting AIDS or other disease from homosexuals (V7348-7351) Homosexuality is unnatural (V7352-7355) Homosexuals have too much/too little influence (V7356-7360) Homosexuality is against the will of God (V7361-7365) Attitudes towards policy relating to homosexuals are measured by these items: Favor or oppose laws protecting homosexuals from job discrimination (V7324-7327) Whether homosexuals should serve in military (V7328-7331) Should homosexual couples be allowed to adopt children (V7332-7335) A number of experiments in the survey response also are included in the Pilot Study. These include: Budget package vs. deficit reduction package (V7200) Experiment in wording of the vote choice for Representative question-reading candidate name as well as party, versus reading only party labels (V7185, V7283) Reversing order of self versus political object placement on liberal conservative 7-pt scale (V7205-7219) Certainty probe on liberal-conservative scale; self and other objects (V7208, V7211, V7216, V7219, V7221, V7223) Experiments on nature of follow-up: strength versus amount (lot, little) (V7263, V7266, V9\7291, V7294, V7300, V7308) Experiments on length of follow-ups: short versus verbose ((V7102-7104, V7349-7351) order in which groups were presented in the interest basis of politics section was reversed for half the sample (V7374-7422) STUDY CHARACTERISTICS AND ADMINISTRATION The 1993 Pilot Study was a telephone reinterview of (cross-section) respondents to the NES 1992 Pre- and Post-Election Study. Interviewing was carried out by the Telephone Facility of the Survey Research Center, the Institute for Social Research. Field period was Sept. 23 --Nov. 24, 1993 Average interview length was 42 minutes 750 interviews were taken, including 4 partials Response rate was 74.6 percent; cooperation rate was 88.4 percent (See below) The study was CATI -- there is no paper version of the Questionnaire Response Rate Calculations This is a Panel Study, and response rate calculations are somewhat different than those for an initial contact study, primarily because there is no "non-sample" category. Every one of the 1005 persons we originally interviewed in the 1992 Post -election study is, by definition, eligible for a reinterview. (1992 respondents who were interviewed in the Pre-election study only were not part of the 1993 study sample.) We reinterviewed 750 of these 1005 respondents to the 1992 study, for a strictly construed reinterview rate of 74.6 percent. 98 respondents from the 1990 Study refused to be reinterviewed. An additional 157 respondents could not cooperate because they were ill or for some other reason physically unable to complete a telephone interview; because they were not locatable; or because they did not have a telephone and did not respond to our requests to call the Telephone Facility. A cooperation rate, which excludes the 157 noninterview cases, is calculated at 88.4 percent. The Telephone Facility and NES staff collaborated on a several step plan to boost response rate and to reduce panel attrition. There were several mailings to the respondents, including a thank-you letter, a respondent report, and an advance contact letter enclosing a small clock as an incentive. The field period was long enough to provide time to track respondents. Persuasion letters were sent, to those who were initially reluctant to participate. An 800-number was set up for respondents to call for further information about the study. In the late stages of interviewing, monetary incentives were offered to 42 reluctant respondents. Finally, the study benefitted from having a highly committed and skilled cadre of interviewers. Interviewer training, pretesting and debriefings The first draft of the questionnaire was pretested by picking at random telephone numbers from local (not Ann Arbor) telephone exchanges. 30 inter- viewers were taken in this way by a mixture of experienced and new interviewers. Study staff "debriefed" the interviewers on their own and respondents' reactions to each question in the pretest instrument. These pretest interviews were also tape recorded, and new questions were "behavior coded" for more quantitative indications of problems with these questions. A separate debriefing was held for the behavior coders. Information from both of these debriefings (which were contradictory on certain points) was incorporated into the production instrument. Standard practice for an SRC study calls for a study guide, listing study objectives and procedures, as well as any special information that interviewers need to know about specific questions. (A copy of this document, as well as study guides for all previous studies, is available from NES Project Staff.) Prestudy conferences with all interviewers and NES staff and PIs gave an opportunity to train on specific questions, and answer concerns of interviewers. Midway through the interviewing, NES staff and PI met with interviewers to hear directly from them how the study was proceeding and how, in their opinion, new sections of the questionnaire were working. A full report of this debriefing is included in Appendix A. Forms Assignment When the Board began planning for this study, we were budgeted for about 40 minutes of interview time, and a number of experiments were proposed. In order to meet these objectives, respondents were randomly assigned to one of four forms. (Variable 7003 records the form assignment.) Randomization Responses to survey questions can be affected by questions that have been asked previously in the survey. There are many survey questions, like the feeling thermometers, where lists of objects are presented for evaluation by respondents. It is extremely difficult, if not impossible, to identify a single order for the items which eliminates response effects. An alternative is to randomize the order in which items on a list are presented to respondents. The CATI system used by the SRC Telephone Facility, AUTOQUEST, has a randomizing function and this was implemented for the feeling thermometer (variables V7130-7136, 7138-7141). No information as to the order in which the thermometer items were asked for a given respondent was preserved. Congressional District Identification for Movers One of the goals of the multiple advance mailings to the 1992 respondents was to get change of address information from local post offices. When we got information that a respondent had moved, and to where, study staff attempted to determine, from what was known of the respondent's new location, in which congressional district the respondent now lived. The name of the member of Congress for that district was then substituted throughout the questionnaire for the name of the member of Congress who was elected in the district in which the respondent lived at the time of the 1992 interview. In a few cases, the information that the respondent had moved was not elicited until the interview was actually underway. When this happened, the interview continued, using the original member of Congress. Organization and Documentation of the Dataset Data for all of the variables and all of the cases in the first wave of the panel, i.e., the 1992 Pre- and Post-election Study, are included in this dataset. Please note that this means that although there are 750 respondents in the 1993 Pilot Study, there are actually 1005 records in the Pilot dataset; one for each (cross-section) respondent to the 1992 Post-election Study. Respondents in the 1992 study who were not re-interviewed in the 1993 wave are assigned missing data codes on the 1993 variables. Documentation for the 1993 Study is separate from the documentation (i.e., codebook) for the 1992 Election Study. Since the variable numbers for the 1992 wave of the study re the same in the Study Staff and the Consortium Releases of the 1992 Election Study, users may use whichever version of that documentation they now have. Users who do not have any 1992 documentation available to them should specify that fact when ordering. The documentation for the 1993 wave is hard-copy, but users may also receive the documentation as WordPerfect 5.2 files or as an ASCII text file. The dataset is an ASCII, raw data file accompanied by SAS/SPSS control cards. There is no OSIRIS dataset. Documentation and dataset are available through the Inter-university Consortium for Political and Social Research. ICPSR User Services may be contacted by phone (313.763- 5010) or by Internet E-Mail (icpsr_netmail@um.cc.umich.edu) for further information. >> Study Design, Content, and Administration 1994 POST STUDY DESIGN The 1994 Election Study was designed to be simultaneously the third wave in a three wave panel, which began in 1992, and also a stand-alone cross-section data collection in the traditional NES time-series. Thus, there are two components to the 1994 Post-election Study: one is a fresh cross-section component, comprising 1136 respondents who were interviewed for the first time in the weeks following the November 8, 1994 general election, and the other is a set of 759 respondents who were initially interviewed in the 1992 Pre-election Study. All of these respondents were interviewed in the 1992 Post-Election Study, and 635 of the panel respondents also gave us an interview in the 1993 Pilot Study. The full set of 1795 respondents may be treated, with appropriate weighting, as a fully representative national cross-section. The three-wave study was designed to exploit the special features of the 1992-1994 elections; a minority president struggling to forge a majority coalition in the face of a strong third-party challenge, and the replacement in 1992 of fully one-quarter of the House of Representatives. The design themes of the 1992-1994 Panel became particularly salient because of the electoral earthquake of the 1994 election, when the Republicans gained control of both houses of Congress first time since 1952. The datafile has been enhanced, for panel respondents, with data from the 1992 and 1993 studies. Data from these earlier studies may be thought of as 'lagged' measures, for use in analysis of 1994 panel respondents. For a full description of the 1992 and 1993 study designs and content, the user is referred to the Appendices to this documentation, which contain the complete original study descriptions as they appear in the documentation for these studies. Of the 1005 respondents who make up the 1992 NES post-election cross-section, (from which the 1992-93-94 Panel respondents were drawn) over a quarter resided in congressional districts that sent a new member to congress in 1992. Thus, the high turnover that occurred in the House in 1992 provides sufficient numbers of respondents to support detailed analysis of the processes by which newly- elected representatives shore up, or fail to shore up their support during their first term in office. The congressional battery that has been in place in NES studies since 1978 was the chief vehicle used in 1992, 1993 and 1994 to evaluate respondents' attitudes towards Congress and their congressional representatives. (For 1993, these questions were modified as necessary to refer to "last November"s election and to the incumbent rather than to the congressional candidates). These questions include: * what respondents like and dislike about congressional candidates * whether and how they have been contacted by the candidates for summary evaluations ( feeling' thermometers) of the candidates, whether they can recall congressional candidates (1993: running in this district this past November') * whether they have had contact with the incumbent candidate * where they place congressional candidates on several issue dimensions * for their evaluations of congressional performance * what the most important issue discussed in the congressional campaign in their district The core battery of congressional evaluations was supplemented by questions on term limits, (1992 and 1994) on the representative's vote on President Clinton's crime bill,(1994), or on the Persian Gulf war resolution (1992), on Clinton's deficit reduction package (1993), whether their Representative was implicated in the House banking scandal (1992) and on whether the respondent felt that his representative cared more about prestige and influence for him/herself rather than solving the problems of the congressional district(1994). Another major theme of the 1992-1993-1994 Panel is the assessment of how well the "Clinton coalition" is faring. The 1992 Study, since it occurred in a Presidential year, had a full set of items bearing on the evaluation of candidate Clinton, some of which were repeated in 1993 and 1994. These repeated items include: * Summary evaluations (feeling thermometer) of Clinton * Traits and affects for Clinton * Placement of Clinton on several issue dimensions (92 and 94 only) * Placement of Clinton on liberal-conservative dimension * approval ratings of several aspects of Clinton's performance in office (93 and 94 only) * For whom R voted (92); recall of Presidential vote (94) * Evaluation of the economy Each of the studies includes specific measures relating to evaluation of Clinton, including likes/dislikes in 1992, opinion about NAFTA and the federal budget deficit in 1993, who the respondent would vote for if the election were held today (1993). Emphasis on the panel aspects of the design should not obscure the fact that the 1994 data can be used to support cross-sectional analyses of the 1994 electorate. Note that almost all of the items listed below were also asked in the 1992 Election Study. * Campaign interest * Media exposure * Measures of partisanship (party likes/dislikes and party identification), which party would better handle certain public problems * Summary evaluations (feeling thermometers) on major political figures and social groups * Voting behavior * Views on issues: most important problem and several issue dimensions, including defense spending, assistance to blacks, spending and services trade-off, health insurance, women's role, and recent proposals to reform welfare. * Preferences on federal budget allocations * Electoral participation * Retrospective and prospective national and personal economic evaluations * Liberal-conservative self-placement * Political information held by respondent * Values, including moral traditionalism, egalitarianism, and attitudes toward race, as well as individual items on school prayer and abortion * Religious affiliation and behavior * Occupation, work force status, home ownership and residential mobility, nationality, education, income, and number of children being raised. The 1992 Election Study, in addition to the topics already mentioned, included questions on social altruism and social connectedness of the respondent; assessments of U.S. involvement in the Persian Gulf War and U.S. foreign policy goals; opinions of the respondent about racial and ethnic stereotypes, on school integration and affirmative action; attitudes towards immigrants (particularly Hispanics and Asians); opinions on immigration policy and bilingual education; and opinions on the rights of homosexuals; on sexual harassment and women's rights. In addition to the congressional and Clinton evaluations already mentioned, the 1993 Pilot Study included a number of items intended to tap the sources and strength of support for Ross Perot. As a pilot study, the 1993 Study included developmental work in a number of areas. One such area is the interest group basis of political attitudes. Questions were asked about the perceived interests of several groups (wealthy, poor, middle class, blacks, whites),as well as self and national interest in three domains: national health insurance, affirmative action, and school choice. The 1993 Study also includes a number of questions relating to attitudes toward homosexuals, and about policies affecting homosexuals. Finally, a number of experiments in the survey response were implemented in the study, including: * an experiment in wording of the vote choice for Representative question * reversing order of self versus political object placement on liberal-conservative 7-pt scale * certainty probe on liberal-conservative scale; self and other objects NOTES ON SURVEY ADMINISTRATION FOR 1992, 1993 AND 1994 STUDIES Field Periods Like the 1992 Pre-and Post-Election Study, the 1994 study design involved face-to face, paper and pencil interviews of respondents randomly selected from the SRC's national area probability sample. The 1994 field period was November 9, 1994 through January 9, 1995, with 40% of the 1795 interviews taken in the first week, and 68% of the interviews within three weeks of the November 8 General Election. This is a significant improvement over the performance of the 1992 Post Election Study, in which only 42.3% of the Post-Election interviews were taken at the end of three weeks. In the 1992 Pre-Election Study, 2485 citizens were interviewed in person in the 9 weeks prior to the November 3, 1992 election of whom 1126 were cross section respondents. To permit analysis of the impact of the unfolding election campaign, a random half of the sample was released to the field on September 1 and the other half on October 1. In the weeks following the election, 2255 pre-election respondents were reinterviewed; 1005 of them were cross-section. Sample Replicates To more closely tailor the field effort to the actual sample performance, NES samples are randomly divided into "replicates" of varying sizes. The usual practice is hold some replicates in reserve. In 1992, additional replicates for both panel and cross section were released midway through the Pre-Election field period; in 1994, all panel sample was released at the beginning of the field period. It did not prove necessary to release additional cross-section replicates. Response Rates for the 1994 Election Study 1994 Post Election N Resp. Rate X-Section 1036 72.1% Panel 759 77.0% Overall 1795 74.1% Notes on the 1993 Pilot Study The 1993 Pilot Study was a telephone reinterview of cross-section respondents to the NES 1992 Pre- and Post- election Study. Interviewing was carried out by the Telephone Facility of the Survey Research Center, the Institute for Social Research. The Field period was Sept. 23 - Nov. 24, 1993, roughly halfway between the 1992 and 1994 Election Studies. 750 interviews were taken, with a response rate of 74.6%. The study was CATI. The average interview length was 42 minutes. Because there were a number of experiments, each respondent was randomly assigned to one of four forms. Randomization of the Feeling Thermometers in the 1993 Pilot Study There are many survey questions, like the feeling thermometers, where lists of objects are presented for evaluation by respondents. It is extremely difficult, if not impossible to identify a single order for the items which eliminates response effects. An alternative is to randomize the order in which items on a list are presented to respondents. The AUTOQUEST CATI system has a randomizing function, and this was implemented for the feeling thermometers in the 1993 Pilot Study. No information as to the order in which the thermometer items were asked for a given respondent was preserved. CONGRESSIONAL DISTRICT IDENTIFICATION AND CONGRESSIONAL CANDIDATES Congressional Ballot Cards, Candidate Lists, and Candidate Numbers In all NES Post-Election surveys since 1978, respondents have been asked several questions about their particular Congresspersons and Senators. These questions in the survey require the insertion by the interviewer, during pre-editing, of the names of candidates. See, for example, question B11, which includes feeling thermometers for the various candidates. Each candidate and Senator is assigned a unique number that reflects his or her incumbency status and party. (See MASTER CODES Candidate Number). The Candidate Lists used by the interviewers, which show which candidates are associated with which congressional district and with which numbers they are tagged, are Notes 4 and 5 in the Master Codes section of this documentation. Congressional District Determination From 1978 through 1990, the congressional district in which an NES sample segment was located was determined by the SRC's sampling section. This was usually done by comparing very detailed maps of the sample segment and of congressional districts. Congressional district determination for the 1992 and 1994 studies presented complications due to the massive redistricting following the 1990 U.S. Census, and due to its panel nature -- movers had to be tracked and their new district determined. Handling of Congressional Incumbency Where Redistricting has Occurred (1992) Throughout the documentation for the 1992 study, whenever the word "incumbent" is used, its referent is a representative who was a member of the 102nd Congress; i.e., the Congress in session prior to the November 1992 General Election. Due to redistricting, any given incumbent's district for the 103rd Congress may consist of a fairly different geographical area from the area covered by the district prior to the boundary changes. Therefore, prior to 1992, the "incumbent"may or may not have been the representative for the particular piece of geography (the sample segment or census tract) in which the respondent lives. For each sample segment, we have included in the dataset its 1992 congressional district number, v3019, and its congressional district number in 1990, v3020. By comparing the two, it can be determined whether the "incumbent" in question was actually the respondent's incumbent prior to the 1992 general election. Congressional District Assignments For Movers Respondents to the 1992 Post-election Study were the recipients of several mailings, which we used to track address changes, and minimize panel attrition due to "lost" respondents. When the United States Post Office returned information indicating that respondents had changed their addresses, the study staff attempted to determine, usually by calling local election offices, in which congressional district the respondent now lived. The substantive decision, for 1993 and for 1994 panel waves, was to ask the R to evaluate the congressional candidates in the district in which h/she was now living, and about whom h/she was presumably receiving information. In some instances, information about where a respondent was now living was not available until the field period, when interviewers were able to track the respondents by talking to former neighbors, etc. In 1994, the interviewers were instructed to contact local electoral offices directly to determine if R's change of address involved also a change of congressional district. A candidate list for R's new district was then prepared, and used to pre-edit the respondent's questionnaire. A similar procedure was used in the 1993 Pilot Study. A Reliability Check of Congressional District Assignments Since one of the chief themes of the 1992-93-94 Panel Study is the evaluation over time of respondents' attitudes toward their congressional representatives, and because of the complications of following movers and of redistricting, NES staff made an intensive effort to assess the both the accuracy and the stability of congressional district assignments. Their findings will appear as Technical Report 52, "Accuracy and Stability of Congressional District Assignments in the 1992-93-94 National Election Studies." That report will be available by early June, 1995. For the 1994 Election Study, we decided to send the entire set of sample segment selections to an outside source for computerized matching of congressional district boundaries and the Census geocodes for the SRC sample segments. In this process, we completely checked the 1992 Congressional District assignments. Approximately 71, or 2.8% of the 1992 respondents (N = 2485) were assigned to the wrong congressional district, because of errors in the original determination of the district (misreading maps, incorrect information from local election offices, etc.) These misassignments were corrected for the 1994 field work, but not for the 1993 Pilot Study, where 4.5% of the 750 respondents were misassigned. In both the 1992 and 1994 Studies, all other causes of being asked about the wrong congressional candidates (e.g., wrongly pre-edited questionnaires, inappropriate determination of congressional districts for movers) totaled less than one percent of the respondents. More important than these errors is the simple question of the stability of the congressional objects themselves. The candidates we ask the respondent to evaluate can change, because a) the respondent moves; b) his/her incumbent does not stand for re-election, or c) there is redistricting. respondent lives. 3% of the panel respondents were affected by 1993 and 1994 redistricting, so that they were not asked to evaluate the same candidates in 1992 and 1994. Incumbents did not run in the general election for 15% of the panel cases, so the congressional candidates they were asked to evaluate in 1994 were different than 1993 or 1992. About 8% of the panel respondents moved between their initial 1992 interview and the 1994 Election Study. Finally, it should be noted that about 3% of the 1994 respondents are registered in districts different than the one in which they were interviewed. Consequently, their vote choice was between a different set of candidates than those about whom they were asked. This dataset contains a number of variables, v22-v32, v80, v7004, and V7007, which record the various contingencies discussed above. Users interested in more detail about these matters should request Technical Report 52 from NES Project Staff. >> Study Design, Content, and Administration 1995 PILOT Study Design The 1995 Pilot Study was conducted between August 3 and September 10th, 1995. The study is a one-wave reinterview of a randomly selected subset of respondents with telephones from the fresh cross section portion of the 1994 Post-Election Study. 1994 "panel" respondents who had been interviewed in 1992 were not eligible for reinterview in the 1995 Pilot Study. The randomly selected sample consisted of 704 respondents from 1994; 486 of these respondents agreed to be interviewed in 1995. The response rate is thus .69 (486/704). The number of refusals was 95. The remainder of the non-interviews are persons with whom contact was lost, or who were unavailable during the study period, for such reasons as illness or absence from home. The study mode was Computer Assisted Telephone Interviewing, or CATI. The average interview length was 44.8 minutes. Study Content The content of the study reflects the NES commitment to improve measures of candidate evaluation, the impact of the campaign, values and predispositions, the comparative study of elections, and other responses to a stimulus letter calling for ideas for content sent to the user community on November 4, 1994. Specific topic areas in the study include: * an experiment between different measures of affective reactions to political figures * a module of items that are being concurrently tested in many other nations as part of a comparative study of politics * a set of 12 items asking respondents to make tradeoffs between programs, taxes and the budget deficit * a set of items designed to measure attitudes toward the environment and environmental policy * a new measure of "humanitarianism" * an extensive set of items re attention to the media, intended to capture exposure to the political campaigns. In order to include all of the content, and also in order to test between competing instrumentation, there were two forms of the questionnaire. Rosters of items, such as the thermometer, were randomized in administration, to minimize order effects. Data and Documentation Because the 486 Pilot Study respondents had also been interviewed in the 1994 Post Election Study, their data from that study has been merged onto the datafile. There are 486 cases in the dataset (in other words, 1994 respondents who were not reinterviewed in 1995 are not included in the dataset). The dataset is an ASCII, or "raw" dataset, accompanied by SAS and SPSS control cards. Missing data definition cards are also included. Documentation for the 1995 Pilot Study is available as an ASCII text file. 1994 Post-Election Study documentation is available on the NES CD-ROM. It will shortly be accessible at http://www.icpsr.umich.edu or through the NES Home Page: http://www.umich.edu/~nes. It is not included as part of the 1995 Pilot Study release. >> Study Design, Content, and Administration 1996 PRE-POST STUDY DESIGN The 1996 National Election Study entailed both a pre-election interview and a post-election re-interview. About three-fourths of the 1996 cases consist of empaneled respondents who were first interviewed in the 1994 or 1992 National Election Study. A freshly drawn cross-section sample makes up the balance of the 1996 cases. (Details of the sample design are given in "Sample Design of the 1996 Pre/Post Election Studies", in Appendix C. Altogether, 1714 citizens were interviewed in the 9 weeks prior to the November 5, 1996 election. To permit analysis of the impact of the unfolding election campaign, the pre-election sample was divided into four subsample replicates, which were released approximately two weeks apart. 1316 of the pre-election interviews were conducted with panel respondents; 398 with cross-section respondents. In the weeks following the election, 1534 pre-election respondents were reinterviewed: 1197 panel, 337 cross-section. This post-election survey included a mode experiment in which respondents were randomly assigned to be interviewed either by telephone or face-to-face. Further details of the administration of the surveys are given in "Study Administration", below. The two components of the study -- the panel and the new cross-section -- were designed to be used together to create a combined nationally representative sample of the American electorate. The 1996 NES data set includes a weight which incorporates sampling, nonresponse and post- stratification factors, (V3), for analysis of the 1996 NES combined sample (Panel component cases plus Cross-section supplement cases). A Time Series Weight (V5) which corrects for Panel attrition (but does not incorporate sampling, nonresponse and post-stratification adjustments) should be used in analyses comparing either the panel respondents or the combined panel and new cross-section respondents to previous (unweighted) National Election Studies data collections. See "Sample Design of the 1996 Pre- and Post-Election Study", and the documentation for V3, and V5 for further information. The frequencies that appear in this codebook are unweighted. A set of files, data, weights, and data documentation, designed to enable panel analyses of the 1992-94-96 data become available sometime late in 1997; announcements concerning the release of data for panel analysis are found at the NES website, www.umich.edu/~nes. The present release has been prepared for cross-section and time series analyses. STUDY CONTENT AND SUBSTANTIVE THEMES The content for the 1996 Election Study reflects its double duty, both as the traditional presidential election year time-series data collection and as a panel study. Substantive themes represented in the 1996 questionnaires include: * interest in the political campaigns; concern about the outcome; and attentiveness to the media's coverage of the campaign * information about politics * evaluation of the presidential candidates and placement of presidential candidates on various issue dimensions * partisanship and evaluations of the political parties * knowledge of and evaluation of House candidates * political participation: turnout in the November general election; other forms of electoral campaign activity * vote choice for President, the U.S. House, and the U.S. Senate, including second choice for President * personal and national economic well-being * positions on social welfare issues including: government health insurance; federal budget priorities, and the role of the government in the provision of jobs and good standard of living * positions on social issues including: abortion; women's roles; prayer in the schools; the rights of homosexuals and the death penalty * racial and ethnic stereotypes; opinions on affirmative action; attitudes towards immigrants * opinions about the nation's most important problem * values and predispositions: moral traditionalism; political efficacy; egalitarianism; humanitarianism individualism; trust in government * social altruism and social connectedness * feeling thermometers on a wide range of political figures and political groups; affinity with various social groups * detailed demographic information and measures of religious affiliation and religiosity. Several new themes are included in the 1996 study: THE CONGRESSIONAL CAMPAIGN: To better understand the dynamics of congressional campaigns, the pre-election wave contains a core battery of campaign-related congressional items (including candidate recall, thermometer ratings, ideological placements, and vote intention). ISSUE IMPORTANCE AND UNCERTAINTY: Several issue questions include "uncertainty" and "importance" follow-ups for both respondent self-placements ("How certain are you of your position on this scale?" "How important is this issue to you?") and candidate placements (e.g. "How certain are you of Bob Dole's position on this scale?" "How important is this issue to Bob Dole?"). COMPARATIVE STUDY OF ELECTORAL SYSTEMS: An eight-minute module of questions developed by a consortium of electoral scholars from 52 polities is included in the post-election interview. Designed to facilitate comparative analysis of political attitudes and voting behavior, the same questions are being asked in similar form in national election studies around the world, and the resulting survey data will eventually be merged with contextual data on electoral laws and political institutions to produce a rich cross-national data set. This module is included as questions T1-T16 in the post-election survey. ISSUE COVERAGE: New issue items in the areas of crime, the environment, gun control, and income inequality are included. A six-item battery carried forward from the 1995 Pilot Study taps respondents' reactions to proposed trade-offs among domestic spending, deficit reduction, and tax cuts. THE ENVIRONMENT: New items from the 1995 Pilot Study tap perceptions of environmental conditions (air quality and the safety of drinking water in the nation and in the respondent's own community), environmental priorities (ranging from global warming to cleaning up lakes and parks), self-placements and placements of candidates and parties on environmental issues (trading off environmental protection against jobs and living standards, and supporting or opposing government environmental regulations on businesses), and the relative effectiveness of national, state, and local governments in dealing with environmental problems. SOCIAL CAPITAL: Several measures of social connectedness are repeated from the 1992 survey. Items tapping trust in people and trust in government are repeated in the pre- and post-election waves to facilitate analysis of the effect of the campaign and election on broader social attitudes. A battery of items on membership and activity in a wide variety of social, political, religious, and civic organizations is included in the post-election questionnaire. This battery includes several questions on as many as four groups in each of twenty-two categories of organizations. Because of the large number of variables produced from these questions, two means of accessing these data are provided; one set of variables which summarize the groups data is available without any unusual effort by the user. A full complement of variables of interest to the specialist in groups membership and participation is also readily available by following instructions provided in Appendix A. MEDIA EXPOSURE: New media exposure, reception, and attention items developed in the 1995 Pilot Study include talk radio items, more specific exposure items for network and local television news, and reception items asking respondents to match news anchors with the networks they work for. A battery of exposure items for entertainment television programs provides an indirect measure of exposure to campaign advertisements. There is also a new open-ended item on recollection of a memorable campaign ad, some expansion and reorganization of items tapping attention to the campaign in various media. Congressional Ballot Cards, Candidate Lists, and Candidate Numbers In the usual NES Post-Election survey, and for 1996, in the Pre-Election survey as well, respondents are asked several questions about their particular Congressperson and Senators. In previous years, interviewers pre-edited questionnaires to fill in the names appropriate for the state and congressional district in which the respondent was living (or was living during the pre-election interview). The use of Computer-Assisted Interviewing software means that information about respondents' congressional district and about candidates and incumbents names (including retiring incumbents) and parties is maintained and periodically updated in a computerized database; this information is loaded into the laptop computers used by interviewers and accessed to provide the correct CD and candidate information for displaying and entering responses to the relevant questions. Each candidate and Senator is assigned a unique number that reflects his or her incumbency status and party. (See Candidate Lists) Particular questions in the survey, which include feeling thermometers for the various candidates, automatically appear on screen with the correct name filled in. The Candidate Lists stored in the database, which show which candidates are associated with which congressional district and with which numbers they are tagged, can also be found here, as can a sample ballot card. Candidates' names were identified by referring to the results of primary elections published in Congressional Quarterly. In the Pre-Election survey there is the additional complication that a number of states held their Congressional primaries after the Pre-Election field work had started. In these cases, the names of those candidates with the greatest chance of winning their party's nomination were loaded into the database. Forecasts of likely winners assumed that incumbents were likely to win their primaries and that unopposed non-incumbents would win. Other races were forecast by Board member Charles Franklin, using a probit model of all 1996 contested primaries involving non-incumbents and utilizing FEC data from August 1, 1996. As soon as the outcome of the primary was known, the correct candidate information was entered into the database and the new version was loaded onto the appropriate interviewers' laptop computers. In nearly all races the forecasted winner was correct. Further details can be found at the documentation for Pre-Election questions B2a and B2b. Features of a CAI questionnaire Using the capabilities of computer-assisted interviewing (CAI) in the 1996 NES enabled the introduction of several features that would not be feasible using a paper-and-pencil questionnaire. The most significant of these for users of this data are: randomization within batteries or sequences of questions; application of half-sampling to some questions; and random order of presentation of blocks of questions. Randomization within batteries refers to presenting, in a randomly determined order, a series of questions about the same objects (or people). An example would be the questions about the respondent's likes and dislikes of the three main Presidential candidates where the names of Clinton, Dole and Perot were inserted randomly as the first, second or third person to be asked about in this series. Randomization of names/objects in this way avoids ordering effects that might be obtained if, for example, the candidates were always asked about in the same order in every series of questions where a parallel question is asked about each of the three. Questions where randomization of order within a series was in force are clearly identified in the codebook. Randomization variables, which allow the user to identify the order of presentation, are provided for all instances of randomized presentation. A few questions, primarily open-ended questions, were half-sampled, so that a randomly selected half of respondents were asked the question. Finally, an order experiment, where a sequence of closed-ended questions was asked early in the interview for a random half of respondents and late in the interview for the other half, was included as part of the mode comparison experiment described below. For both of these features, the relevant codebook entries contain explanatory notes. All random selections were programmed into the computer application of the questionnaire and occurred automatically and independently of other circumstances of the interview. CAI eliminates the preparation of a paper and pencil version which would previously have been published in the codebook. STUDY ADMINISTRATION Interviewing for the pre-election survey began on September 3, 1996 and concluded on November 4, 1996. The average length of interview in the pre-election survey was 74 minutes. The overall response rate was 71%. (See "Response Rates" below for a complete discussion.) The post-election interviewing occurred between November 6 and December 31, 1996 inclusive, with an average interview length of 70 minutes. The overall reinterview rate was 90%, with further details available in the Response Rate section below. Sample "Releases" in the Pre-election survey Both parts of the sample (panel and cross-section) were randomly subdivided into four quarter sample releases, each of which is a proper, random subsample of the NES sample. Two additional 'reserve' replicates of cross-section cases were held in abeyance until it was determined that the additional sample lines would be needed to attain study goals. Replicates 1 through 4 were considered the "base sample," certain to be released. The release dates for sample replicates were: Replicate Date of release 1 September 3, 1996 2 September 12, 1996 3 September 26, 1996 4 October 10, 1996 5 (Reserve) September 26, 1996 (with replicate 3) 6 (Reserve) October 10, 1996 (with replicate 4) For a full description of the sample design and implementation, see "1996 SAMPLE DESIGN". Pre-election study: assignment to telephone mode One of the administrative problems in fielding a panel study is that respondents have had an intervening period of time in which to relocate, perhaps at some remove from areas where field study staff are available to interview them. We estimated that between 40 and 80 respondents might have moved to areas in which SRC did not have interviewers in the field. Our priority was to interview as many panel respondents as possible, but we did not want to incur the additional costs associated with interviewer travel. Accordingly, panel respondents who had moved 'out-of-range' for a face-to-face interview were converted to phone mode. The criterion set for deciding if a case was 'out-of-range' was 90 minutes driving time one-way from the interviewer's home to the respondent's address under local usual driving conditions. The total number of panel respondents that we interviewed who were "out of range" for this reason was 47. Post-Election Mode Experiment: Design and Implementation In contrast with the usual NES practice of conducting all post-election interviews in person, half of the respondents in the post-election wave of the 1996 survey were interviewed by telephone, with post-election respondents randomly assigned (except in extreme circumstances) to phone or face-to-face administration. The telephone mode used the same computerized questionnaire developed for the face-to-face post-election interviews and was conducted by the same interviewers. The mode experiment provides a direct comparison of the effects of mode of interview on important indicators of data quality and comparability. Cases were assigned to either telephone or face-to-face mode at the sample segment level. Every effort to retain randomly assigned cases in their assigned mode was made. Respondents who had been interviewed by telephone in the pre-election study were disqualified from random assignment to mode; all those reinterviews were done by telephone, a total of 47 cases. Respondents who did not have telephones and respondents who were not able to participate in the mode experiment because of a physical limitation that prevented them from being interviewed by one mode or another were also excluded, which totaled 130 additional cases (24 of these were completed by telephone). No changes in mode of interview because of respondent preference or for ease of administration were permitted. All prospective respondents received two incentives in the mail: a check for $10 and a small gift. Included in the mailing to telephone mode respondents was a sealed respondent booklet with the candidate ballot folded inside. The contact letter instructed respondents to set these materials aside until told to open them by the interviewer. Interviewers followed procedures to ascertain that respondents were using the booklet and ballot card appropriately and to note deviations from the instructions. Evaluation of problems in study implementation Two implementation problems arose in the post-election field administration. This resulted in two unintended systematic deviations from standard administration. 145 cases in the phone mode were mailed a respondent booklet that included the wrong ballot card. As soon as this problem was discovered, new respondent booklets with correct ballot cards were mailed by overnight mail to these respondents. Approximately 50 interviews were conducted where the respondent had the incorrect ballot card; in these cases interviewers read the correct ballot card information to the respondent. A full report to be issued will analyze these data to identify any systematic differences related to this implementation error. It was discovered early in the data collection period that 39 interviews were completed using the training version of the survey instrument, due to a technical problem in transmitting files to the field. The training version contained no randomized presentation of questions and lacked several last minute changes to the interview. Call-backs to 37 of these 39 respondents allowed us to collect data on the several missed questions. A report analyzing these cases for differential impact of the use of the training questionnaire is in preparation. RESPONSE RATES The response rate in the pre-election study was 71%. Among panel respondents the response (reinterview) rate was 76%; among cross-section respondents it was 60%. The overall reinterview/response rate in the post-election interviewing was 90%. Among panel respondents in the post-election survey, the response rate was 91% and among cross-section respondents it was 85%. The response rate in face-to-face mode (including all cases in this mode, experimentally assigned and excluded) was 89% and for telephone mode it was 91%. INTERVIEW COMPLETION RATE Completion rates for the pre-election sample releases, for pre-election time periods, and for post-election time periods are presented here. Table 1 presents the percentage completions per quarter sample replicate (replicates 3 and 4 include the reserve cases added to those replicates); table 2 shows the percentage of completions per two week time period in the pre-election survey. Table 3 lays out the number of interviews taken for each week elapsing after the Nov. 5 General Election. In 1996, 29% of the interviews were completed in the first week after the election and 86% in the first three weeks; progress was evenly divided between face-to-face and telephone modes. Table 1: % Completions by release (pre-election survey) RELEASE Total Panel Cross-section 1 28% 28% 18% 2 27 27 24 3+5 23 23 23 4+6 23 22 25 Table 2: Percent Completions by two week period (pre-election survey) DATES Total Panel Cross-section 9/3-9/16 19% 20% 18% 9/17-10/1 24 24 22 10/2-10/16 23 23 23 10/17-10/30 24 24 26 10/31-11/4 10 10 12 Table 3: Number of and Cumulative Percent of Interviews Taken in the Post-Election Study by Week of Interview DATES NUMBER OF CUMULATIVE CUMULATIVE INTERVIEWS NUMBER OF PERCENT OF INTERVIEWS INTERVIEWS Nov. 6-Nov.12 449 449 29% Nov.13-Nov.19 551 1000 65 Nov.20-Nov.26 314 1314 86 Nov.27-Dec. 3 91 1405 92 Dec.4- Dec. 10 84 1489 97 Dec.11-Dec.17 32 1521 99 Dec.18-Dec.24 10 1531 99 Dec.25-Dec.31 3 1534 100% FILE STRUCTURE The AMERICAN NATIONAL ELECTION STUDY, 1996 PRE- AND POST-ELECTION SURVEY are available in logical record length (LRECL) format. The data are sorted in ascending order by respondent number, and contain 1,657 variables for 1714 respondents. The machine-readable codebook, which provides complete formatting and other information for all variables accompanies the data. In addition, a set of SAS and SPSS control statements has been prepared for this collection. The control statements contain formatting information as well as variable labels, value labels and missing data specifications for all variables in the collection. The data can also be accessed directly through software packages that do not use SAS or SPSS control statements by specifying the record locations of the desired variables. The record locations for all variables are provided in the codebook. NOTES ON CONFIDENTIALITY Starting with the 1986 Election Study, NES has released occupation code variables in somewhat less detail than in years past. This dataset includes a two-digit code with 71 categories corresponding to Census Bureau occupational groupings. Those who need the full occupation code for their research should contact the NES project staff for information about the conditions under which access may be provided. Similarly, the National Election Studies have not included information for census tracts or minor civil divisions since 1978. Permission to use the more detailed geographic information for scholarly research may be obtained from the Board of Overseers. More information about this is available from NES project staff. Coding of the new religious denomination variable is in some cases based on an alphabetic "other, please specify" variable. This variable is restricted for reasons of confidentiality, but access may be provided to legitimate scholars under established NES procedures. OPEN-ENDED MATERIALS Traditionally, the National Election Studies have contained several minutes of open-ended responses (for example, the candidate likes and dislikes). These questions are put into Master Codes by the SRC coding section. Other scholars have developed alternative or supplemental coding schemes for the questions (for example, the levels of conceptualization, released as ICPSR 8151). The Board of Overseers wishes to encourage these efforts but in ways which respect the NES and SRC obligation to protect the privacy and anonymity of respondents. Circumstances under which individuals may have access to transcribed versions of these questions have been worked out and those interested should contact the NES project staff for further details. 1996 SPECIAL NOTE - CHANGES IN CODING BETWEEN PRE AND POST Several questions which were asked in the Pre-election interview and then asked again in the Post-Election interview had some differences between the versions used. Variables where pre and post codes (and some code labels) don't match on repeated questions: Pre Post 960369 1273 (Code 4 label; Respon. Booklet identical both waves) 960370 1274 (Code 4 label; Respon. Booklet identical both waves) 960371 1275 (Code 4 label; Respon. Booklet identical both waves) 960375 1277 (Code 4 label; Respon. Booklet identical both waves) 960376 1278 (Code 4 label; Respon. Booklet identical both waves) 960377 1279 (Code 4 label; Respon. Booklet identical both waves) 960378 1280 (Code 4 label; Respon. Booklet identical both waves) 960566 1251 (Code 7 in Pre; code 4 in Post) 960569 1259 (Codes 1, 2 in Pre; codes 1, 5 in Post) For variables 960369-371, 960375-378 / 961273-1275, 961277-1280 (7 point liberal-conservative scale questions) the differences appeared onscreen to the interviewer but the labeled Respondent Booklet was accurate (complete code 4 label) and identical for both interviews. NEW WEIGHTS FOR THE 1996 STUDY (RELEASED MARCH 1998) The steps taken to address the 1996 NES overestimation of voting in the 1996 presidential election resulted in the development of post-stratified weights which account for individual selection probability, regional differences in household nonresponse, and misrepresentation of any age by education subgroups. These revised, CPS-standardized weights were computed for the 1994 NES Post and 1996 NES Pre and Post Election data sets. Users of previous weights released with the 1994 and 1996 data will find that these weights extend and combine the features of previously released weights. Similar weights were computed for the 1992 cross-section cases; these weights will be included in an upcoming combined 1992-1994-1996 data file. V960005A and V960005B, the two new weight variables for 1996, are released for the first time for use with the 1996 NES data. A review of the findings that led to the construction of these new weights and full details of their development and effect are described in new Appendix B, "Post-Stratified Cross-sectional Analysis Weights for the 1992, 1994 and 1996 NES Data." The recommendation to explore developing these weights was made by Warren Miller and Merrill Shanks and authorized by the NES Board of Overseers at its September 1997 meeting. The SRC Sampling Section, under the direction of Steve Heeringa, completed the work and the technical report in consultation with the NES Director of Studies. There are two weights, one to be applied to the pre-election sample (V960005A) and the other which is for use with the post-election sample (V960005B). The post-election sample weight takes into account attrition that occurred between the pre- and post-election surveys. In analyses using variables from both the pre- and post-election data, the post-election weight should be applied. Use of either weight is appropriate only for the full sample, cross-section and panel cases combined. >> Study Design, Content, and Administration 1997 PILOT Study Design The 1997 Pilot Study was conducted between September 5 and October 1, 1997. The study is a reinterview of a subset of respondents with telephones from the 1996 Post-Election Study. All fresh-cross section cases for 1996 that completed a post-election interview and for which telephone numbers were available were included in the 1997 pilot. The balance of cases consisted of cases from the two previous waves, the 1994 'panel' cases and the 1992 'panel' cases for which telephone numbers were available and a post-election interview was conducted in 1996. Each of these panel components was represented proportionally in the initial sample for 1997. The initial sample consisted of 724 respondents from 1996; 551 of these respondents completed an interview in 1997. The response rate is thus .76 (551/724). The number of refusals was 22. The remainder of the non-interviews are persons with whom contact was never made, or who were unavailable during the study period, for such reasons as illness or absence from home. The study mode was Computer Assisted Telephone Interviewing ("CATI".) The average interview length was 45.3 minutes. Study Content The content of the study reflects the NES commitment to improve measures of group mobilization, interest articulation and representation, group-based political reasoning, race and racial attitudes and policy, issue attitudes, human and social capital, social choice, theories of the survey response, and other responses to a stimulus letter calling for ideas for content sent to the user community on November 11, 1996. Specific topic areas in the study include: MOBILIZATION AND NON ELECTORAL PARTICIPATION: A battery designed to improve NES instrumentation on non-electoral political participation and mobilization; specifically, respondents' efforts to contact public officials at different levels of government during the non electoral season and their reasons for contact. GROUP-BASED POLITICS: Elaborated testing of long-standing NES instrumentation on group closeness designed to evaluate both "traditional" NES instrumentation and investigate possible additions and improvements. Group difference and group conflict as a basis of current mass politics: Perceptions of paired "opposing" social groups on issue, ideology, party placements and vote choice. The groups include black and white people, Christian fundamentalists and gays and lesbians, and men and women. There is an embedded experiment testing the effects of focusing on group difference and conflict on social trust and political trust and interest. Group threat as a basis of group-based politics: A split ballot of items involving an experimental manipulation of the level of threat in different domains and prejudices about Blacks and Christian Fundamentalists. RESPONSE LATENCY: Activated timings of response latencies on several questions to extend recent NES work on certainty. EVALUATIONS OF THE PRESIDENT, CONGRESS AND THE SUPREME COURT: Exploration of a new battery of items to improve current NES instrumentation and extend parallel measurement across governmental institutions. RELIGION AND POLITICS: Further Pilot work on the role of religion in citizens' political thinking; attitudes toward the role of religion and religious institutions in American society and politics. The use of CATI enabled a number of experimental treatments within the survey instrument. Random assignment to question wording, early-late placement and presentation order were applied to numerous question sequences. Rosters of items, such as the thermometer and placements of groups and individuals on scales, were randomized in administration, to minimize order effects. Indicator variables that document the use of split-ballot and randomization features are found in the codebook. Data and Documentation Because the 551 Pilot Study respondents had also been interviewed in the 1996 Pre- and Post Election Studies, their data from those studies has been merged onto the datafile. There are 551 cases in the dataset (in other words, it contains 1996 data only for those respondents who were reinterviewed in 1997). The dataset is an ASCII, or "raw" dataset, accompanied by SAS and SPSS control cards. Missing data definition cards are also included. Documentation for the 1997 Pilot Study is available as an ASCII text file on the NES website (http://www.umich.edu/~nes) and from the ICPSR (http://www.icpsr.umich.edu). 1996 Election Studies documentation is also available (separately) on both websites; it is not included as part of the 1997 Pilot Study release. >> 1996 - ACCESSING GROUP-SPECIFIC DATA Please also see related paragraphs in the Introductory section of the codebook for general information about the 'Groups' section in the Post. For data users interested in a greater level of detail for the Post 'Groups' section (R3-R7w), the raw data for the 1996 Study includes additional data which are not represented in the codebook and are not included in the SAS and SPSS data definition files provided with the Study data. To access these additional variables, the column numbers may be cut and pasted from the listings below and then inserted into the SAS or SPSS data definition file that the user is submitting. SAS and SPSS missing data assignments also may be cut and pasted into the user's file. This additional information provides the specific responses to questions about individual groups in each category--Group1, Group2, Group3 or Group4. GROUP SPECIFIC DATA For the question on group membership, the category summary variable identifies the specific groups of which R is a member and additional group-specific vars are not necessary. For the questions on dues/contributions, meetings/activities, and political discussion, however, more than 2 responses were possible and the summary variables could not identify the particular response for an individual group. The responses categories are: "In the past 12 months have you paid dues or given any money to this group? Which is that? (Dues, contributions, or both?)" 1. Dues 3. Contributions 5. Both 7. Other (specify) 8. DK 9. NA 0. Inap, R is not involved with any group in this category; no further group mentioned in this category (Groups 2-4 only); no post IW "In the last 12 months have you taken part in any activities sponsored by this group or attended a meeting of this group?" 1. Attended a meeting 3. Taken part in activities 5. Both 8. DK 9. NA 0. Inap, R is not involved with any groups in this category; no further group mentioned in this category (Groups 2-4 only); no post IW "How often does this group discuss politics-- often, sometimes, rarely, or never?" 1. Often 2. Sometimes 3. Rarely 4. Never 8. DK 9. NA 0. Inap, R is not involved with any groups in this category; no further group mentioned in this category (Groups 2-4 only); no post IW COMPLETE SET OF COLUMN LOCATIONS (Within each group category, Group 1 is identified in the "A" variable, Group 2 in the "B" variable, Group 3 in the "C" variable", and Group 4 in the "D" variable). SEE MISSING DATA HEADINGS BELOW TO IDENTIFY QUESTIONS WITHIN GROUP CATEGORIES. V961344 5756 V961345 5757-5760 V961346 5761-5764 V961346A 5765 V961346B 5766 V961346C 5767 V961346D 5768 V961347 5769-5772 V961347A 5773 V961347B 5774 V961347C 5775 V961347D 5776 V961348 5777-5780 V961348A 5781 V961348B 5782 V961348C 5783 V961348D 5784 V961349 5785 V961350 5786-5789 V961351 5790-5793 V961351A 5794 V961351B 5795 V961351C 5796 V961351D 5797 V961352 5798-5801 V961352A 5802 V961352B 5803 V961352C 5804 V961352D 5805 V961353 5806-5809 V961353A 5810 V961353B 5811 V961353C 5812 V961353D 5813 V961354 5814 V961355 5815-5818 V961356 5819-5822 V961356A 5823 V961356B 5824 V961356C 5825 V961356D 5826 V961357 5827-5830 V961357A 5831 V961357B 5832 V961357C 5833 V961357D 5834 V961358 5835-5838 V961358A 5839 V961358B 5840 V961358C 5841 V961358D 5842 V961359 5843 V961360 5844-5847 V961361 5848-5851 V961361A 5852 V961361B 5853 V961361C 5854 V961361D 5855 V961362 5856-5859 V961362A 5860 V961362B 5861 V961362C 5862 V961362D 5863 V961363 5864-5867 V961363A 5868 V961363B 5869 V961363C 5870 V961363D 5871 V961364 5872 V961365 5873-5876 V961366 5877-5880 V961366A 5881 V961366B 5882 V961366C 5883 V961366D 5884 V961367 5885-5888 V961367A 5889 V961367B 5890 V961367C 5891 V961367D 5892 V961368 5893-5896 V961368A 5897 V961368B 5898 V961368C 5899 V961368D 5900 V961369 5901 V961370 5902-5905 V961371 5906-5909 V961371A 5910 V961371B 5911 V961371C 5912 V961371D 5913 V961372 5914-5917 V961372A 5918 V961372B 5919 V961372C 5920 V961372D 5921 V961373 5922-5925 V961373A 5926 V961373B 5927 V961373C 5928 V961373D 5929 V961374 5930 V961375 5931-5934 V961376 5935-5938 V961376A 5939 V961376B 5940 V961376C 5941 V961376D 5942 V961377 5943-5946 V961377A 5947 V961377B 5948 V961377C 5949 V961377D 5950 V961378 5951-5954 V961378A 5955 V961378B 5956 V961378C 5957 V961378D 5958 V961379 5959 V961380 5960-5963 V961381 5964-5967 V961381A 5968 V961381B 5969 V961381C 5970 V961381D 5971 V961382 5972-5975 V961382A 5976 V961382B 5977 V961382C 5978 V961382D 5979 V961383 5980-5983 V961383A 5984 V961383B 5985 V961383C 5986 V961383D 5987 V961384 5988 V961385 5989-5992 V961386 5993-5996 V961386A 5997 V961386B 5998 V961386C 5999 V961386D 6000 V961387 6001-6004 V961387A 6005 V961387B 6006 V961387C 6007 V961387D 6008 V961388 6009-6012 V961388A 6013 V961388B 6014 V961388C 6015 V961388D 6016 V961389 6017 V961390 6018-6021 V961391 6022-6025 V961391A 6026 V961391B 6027 V961391C 6028 V961391D 6029 V961392 6030-6033 V961392A 6034 V961392B 6035 V961392C 6036 V961392D 6037 V961393 6038-6041 V961393A 6042 V961393B 6043 V961393C 6044 V961393D 6045 V961394 6046 V961395 6047-6050 V961396 6051-6054 V961396A 6055 V961396B 6056 V961396C 6057 V961396D 6058 V961397 6059-6062 V961397A 6063 V961397B 6064 V961397C 6065 V961397D 6066 V961398 6067-6070 V961398A 6071 V961398B 6072 V961398C 6073 V961398D 6074 V961399 6075 V961400 6076-6079 V961401 6080-6083 V961401A 6084 V961401B 6085 V961401C 6086 V961401D 6087 V961402 6088-6091 V961402A 6092 V961402B 6093 V961402C 6094 V961402D 6095 V961403 6096-6099 V961403A 6100 V961403B 6101 V961403C 6102 V961403D 6103 V961404 6104 V961405 6105-6108 V961406 6109-6112 V961406A 6113 V961406B 6114 V961406C 6115 V961406D 6116 V961407 6117-6120 V961407A 6121 V961407B 6122 V961407C 6123 V961407D 6124 V961408 6125-6128 V961408A 6129 V961408B 6130 V961408C 6131 V961408D 6132 V961409 6133 V961410 6134-6137 V961411 6138-6141 V961411A 6142 V961411B 6143 V961411C 6144 V961411D 6145 V961412 6146-6149 V961412A 6150 V961412B 6151 V961412C 6152 V961412D 6153 V961413 6154-6157 V961413A 6158 V961413B 6159 V961413C 6160 V961413D 6161 V961414 6162 V961415 6163-6166 V961416 6167-6170 V961416A 6171 V961416B 6172 V961416C 6173 V961416D 6174 V961417 6175-6178 V961417A 6179 V961417B 6180 V961417C 6181 V961417D 6182 V961418 6183-6186 V961418A 6187 V961418B 6188 V961418C 6189 V961418D 6190 V961419 6191 V961420 6192-6195 V961421 6196-6199 V961421A 6200 V961421B 6201 V961421C 6202 V961421D 6203 V961422 6204-6207 V961422A 6208 V961422B 6209 V961422C 6210 V961422D 6211 V961423 6212-6215 V961423A 6216 V961423B 6217 V961423C 6218 V961423D 6219 V961424 6220 V961425 6221-6224 V961426 6225-6228 V961426A 6229 V961426B 6230 V961426C 6231 V961426D 6232 V961427 6233-6236 V961427A 6237 V961427B 6238 V961427C 6239 V961427D 6240 V961428 6241-6244 V961428A 6245 V961428B 6246 V961428C 6247 V961428D 6248 V961429 6249 V961430 6250-6253 V961431 6254-6257 V961431A 6258 V961431B 6259 V961431C 6260 V961431D 6261 V961432 6262-6265 V961432A 6266 V961432B 6267 V961432C 6268 V961432D 6269 V961433 6270-6273 V961433A 6274 V961433B 6275 V961433C 6276 V961433D 6277 V961434 6278 V961435 6279-6282 V961436 6283-6286 V961436A 6287 V961436B 6288 V961436C 6289 V961436D 6290 V961437 6291-6294 V961437A 6295 V961437B 6296 V961437C 6297 V961437D 6298 V961438 6299-6302 V961438A 6303 V961438B 6304 V961438C 6305 V961438D 6306 V961439 6307 V961440 6308-6311 V961441 6312-6315 V961441A 6316 V961441B 6317 V961441C 6318 V961441D 6319 V961442 6320-6323 V961442A 6324 V961442B 6325 V961442C 6326 V961442D 6327 V961443 6328-6331 V961443A 6332 V961443B 6333 V961443C 6334 V961443D 6335 V961444 6336 V961445 6337-6340 V961446 6341-6344 V961446A 6345 V961446B 6346 V961446C 6347 V961446D 6348 V961447 6349-6352 V961447A 6353 V961447B 6354 V961447C 6355 V961447D 6356 V961448 6357-6360 V961448A 6361 V961448B 6362 V961448C 6363 V961448D 6364 V961449 6365 V961450 6366-6369 V961451 6370-6373 V961451A 6374 V961451B 6375 V961451C 6376 V961451D 6377 V961452 6378-6381 V961452A 6382 V961452B 6383 V961452C 6384 V961452D 6385 V961453 6386-6389 V961453A 6390 V961453B 6391 V961453C 6392 V961453D 6393 ............................................... GROUP-SPECIFIC MISSING DATA ************************************************************ ************************************************************ FOR PAYMENT OF DUES/CONTRIBUTIONS: the group-specific data may be found in the columns below. For each group category, the first variable corresponds to the response for Group1 mention, the second variable corresponds to the response for Group2 mention, etc. LABOR UNIONS DUES/CONTRIBUTIONS if v961346a =0 then v961346a =.; if v961346b =0 then v961346b =.; if v961346c =0 then v961346c =.; if v961346d =0 then v961346d =.; v961346a (0) v961346b (0) v961346c (0) v961346d (0) BUSINESS OR WORK-RELATED DUES/CONTRIBUTIONS if v961351a =0 then v961351a =.; if v961351b =0 then v961351b =.; if v961351c =0 then v961351c =.; if v961351d =0 then v961351d =.; v961351a (0) v961351b (0) v961351c (0) v961351d (0) VETERANS DUES/CONTRIBUTIONS if v961356a =0 then v961356a =.; if v961356b =0 then v961356b =.; if v961356c =0 then v961356c =.; if v961356d =0 then v961356d =.; v961356a (0) v961356b (0) v961356c (0) v961356d (0) CHURCH/SYNAGOGUE DUES/CONTRIBUTIONS if v961361a =0 then v961361a =.; if v961361b =0 then v961361b =.; if v961361c =0 then v961361c =.; if v961361d =0 then v961361d =.; v961361a (0) v961361b (0) v961361c (0) v961361d (0) OTHER RELIGIOUS DUES/CONTRIBUTIONS if v961366a =0 then v961366a =.; if v961366b =0 then v961366b =.; if v961366c =0 then v961366c =.; if v961366d =0 then v961366d =.; v961366a (0) v961366b (0) v961366c (0) v961366d (0) ELDERLY/SENIOR DUES/CONTRIBUTIONS if v961371a =0 then v961371a =.; if v961371b =0 then v961371b =.; if v961371c =0 then v961371c =.; if v961371d =0 then v961371d =.; v961371a (0) v961371b (0) v961371c (0) v961371d (0) ETHNIC/NATIONALITY DUES/CONTRIBUTIONS if v961376a =0 then v961376a =.; if v961376b =0 then v961376b =.; if v961376c =0 then v961376c =.; if v961376d =0 then v961376d =.; v961376a (0) v961376b (0) v961376c (0) v961376d (0) WOMEN'S RIGHTS OR WELFARE DUES/CONTRIBUTIONS if v961381a =0 then v961381a =.; if v961381b =0 then v961381b =.; if v961381c =0 then v961381c =.; if v961381d =0 then v961381d =.; v961381a (0) v961381b (0) v961381c (0) v961381d (0) POLITICAL ISSUE DUES/CONTRIBUTIONS if v961386a =0 then v961386a =.; if v961386b =0 then v961386b =.; if v961386c =0 then v961386c =.; if v961386d =0 then v961386d =.; v961386a (0) v961386b (0) v961386c (0) v961386d (0) NONPARTISAN OR CIVIC DUES/CONTRIBUTIONS if v961391a =0 then v961391a =.; if v961391b =0 then v961391b =.; if v961391c =0 then v961391c =.; if v961391d =0 then v961391d =.; v961391a (0) v961391b (0) v961391c (0) v961391d (0) LIBERAL OR CONSERVATIVE DUES/CONTRIBUTIONS if v961396a =0 then v961396a =.; if v961396b =0 then v961396b =.; if v961396c =0 then v961396c =.; if v961396d =0 then v961396d =.; v961396a (0) v961396b (0) v961396c (0) v961396d (0) POLITICAL PARTY OR CAND SUPPORT DUES/CONTRIBUTIONS if v961401a =0 then v961401a =.; if v961401b =0 then v961401b =.; if v961401c =0 then v961401c =.; if v961401d =0 then v961401d =.; v961401a (0) v961401b (0) v961401c (0) v961401d (0) CHILDRENS' ACTIVITY DUES/CONTRIBUTIONS if v961406a =0 then v961406a =.; if v961406b =0 then v961406b =.; if v961406c =0 then v961406c =.; if v961406d =0 then v961406d =.; v961406a (0) v961406b (0) v961406c (0) v961406d (0) LITERARY, ART OR DISCUSSION DUES/CONTRIBUTIONS if v961411a =0 then v961411a =.; if v961411b =0 then v961411b =.; if v961411c =0 then v961411c =.; if v961411d =0 then v961411d =.; v961411a (0) v961411b (0) v961411c (0) v961411d (0) HOBBY OR LEISURE DUES/CONTRIBUTIONS if v961416a =0 then v961416a =.; if v961416b =0 then v961416b =.; if v961416c =0 then v961416c =.; if v961416d =0 then v961416d =.; v961416a (0) v961416b (0) v961416c (0) v961416d (0) NEIGHBORHOOD OR COMMUNITY DUES/CONTRIBUTIONS if v961421a =0 then v961421a =.; if v961421b =0 then v961421b =.; if v961421c =0 then v961421c =.; if v961421d =0 then v961421d =.; v961421a (0) v961421b (0) v961421c (0) v961421d (0) SERVICE/FRATERNAL DUES/CONTRIBUTIONS if v961426a =0 then v961426a =.; if v961426b =0 then v961426b =.; if v961426c =0 then v961426c =.; if v961426d =0 then v961426d =.; v961426a (0) v961426b (0) v961426c (0) v961426d (0) SERVICE TO NEEDY DUES/CONTRIBUTIONS if v961431a =0 then v961431a =.; if v961431b =0 then v961431b =.; if v961431c =0 then v961431c =.; if v961431d =0 then v961431d =.; v961431a (0) v961431b (0) v961431c (0) v961431d (0) EDUCATIONAL DUES/CONTRIBUTIONS if v961436a =0 then v961436a =.; if v961436b =0 then v961436b =.; if v961436c =0 then v961436c =.; if v961436d =0 then v961436d =.; v961436a (0) v961436b (0) v961436c (0) v961436d (0) CULTURAL SERVICE DUES/CONTRIBUTIONS if v961441a =0 then v961441a =.; if v961441b =0 then v961441b =.; if v961441c =0 then v961441c =.; if v961441d =0 then v961441d =.; v961441a (0) v961441b (0) v961441c (0) v961441d (0) SELF-HELP DUES/CONTRIBUTIONS if v961446a =0 then v961446a =.; if v961446b =0 then v961446b =.; if v961446c =0 then v961446c =.; if v961446d =0 then v961446d =.; v961446a (0) v961446b (0) v961446c (0) v961446d (0) OTHER DUES/CONTRIBUTIONS if v961451a =0 then v961451a =.; if v961451b =0 then v961451b =.; if v961451c =0 then v961451c =.; if v961451d =0 then v961451d =.; v961451a (0) v961451b (0) v961451c (0) v961451d (0) ************************************************************ ************************************************************ FOR MEETINGS/ACTIVITIES: the group-specific data may be found in the columns below. For each group category, the first variable corresponds to the response for Group1 mention, the second variable corresponds to the response for Group2 mention, etc. LABOR UNIONS MEETINGS/ACTIVITIES if v961347a =0 then v961347a =.; if v961347b =0 then v961347b =.; if v961347c =0 then v961347c =.; if v961347d =0 then v961347d =.; v961347a (0) v961347b (0) v961347c (0) v961347d (0) BUSINESS OR WORK-RELATED MEETINGS/ACTIVITIES if v961352a =0 then v961352a =.; if v961352b =0 then v961352b =.; if v961352c =0 then v961352c =.; if v961352d =0 then v961352d =.; v961352a (0) v961352b (0) v961352c (0) v961352d (0) VETERANS MEETINGS/ACTIVITIES if v961357a =0 then v961357a =.; if v961357b =0 then v961357b =.; if v961357c =0 then v961357c =.; if v961357d =0 then v961357d =.; v961357a (0) v961357b (0) v961357c (0) v961357d (0) CHURCH/SYNAGOGUE MEETINGS/ACTIVITIES if v961362a =0 then v961362a =.; if v961362b =0 then v961362b =.; if v961362c =0 then v961362c =.; if v961362d =0 then v961362d =.; v961362a (0) v961362b (0) v961362c (0) v961362d (0) OTHER RELIGIOUS MEETINGS/ACTIVITIES if v961367a =0 then v961367a =.; if v961367b =0 then v961367b =.; if v961367c =0 then v961367c =.; if v961367d =0 then v961367d =.; v961367a (0) v961367b (0) v961367c (0) v961367d (0) ELDERLY/SENIOR MEETINGS/ACTIVITIES if v961372a =0 then v961372a =.; if v961372b =0 then v961372b =.; if v961372c =0 then v961372c =.; if v961372d =0 then v961372d =.; v961372a (0) v961372b (0) v961372c (0) v961372d (0) ETHNIC/NATIONALITY MEETINGS/ACTIVITIES if v961377a =0 then v961377a =.; if v961377b =0 then v961377b =.; if v961377c =0 then v961377c =.; if v961377d =0 then v961377d =.; v961377a (0) v961377b (0) v961377c (0) v961377d (0) WOMEN'S RIGHTS OR WELFARE MEETINGS/ACTIVITIES if v961382a =0 then v961382a =.; if v961382b =0 then v961382b =.; if v961382c =0 then v961382c =.; if v961382d =0 then v961382d =.; v961382a (0) v961382b (0) v961382c (0) v961382d (0) POLITICAL ISSUE MEETINGS/ACTIVITIES if v961387a =0 then v961387a =.; if v961387b =0 then v961387b =.; if v961387c =0 then v961387c =.; if v961387d =0 then v961387d =.; v961387a (0) v961387b (0) v961387c (0) v961387d (0) NONPARTISAN OR CIVIC MEETINGS/ACTIVITIES if v961392a =0 then v961392a =.; if v961392b =0 then v961392b =.; if v961392c =0 then v961392c =.; if v961392d =0 then v961392d =.; v961392a (0) v961392b (0) v961392c (0) v961392d (0) LIBERAL OR CONSERVATIVE MEETINGS/ACTIVITIES if v961397a =0 then v961397a =.; if v961397b =0 then v961397b =.; if v961397c =0 then v961397c =.; if v961397d =0 then v961397d =.; v961397a (0) v961397b (0) v961397c (0) v961397d (0) POLITICAL PARTY OR CAND SUPPORT MEETINGS/ACTIVITIES if v961402a =0 then v961402a =.; if v961402b =0 then v961402b =.; if v961402c =0 then v961402c =.; if v961402d =0 then v961402d =.; v961402a (0) v961402b (0) v961402c (0) v961402d (0) CHILDRENS' ACTIVITY MEETINGS/ACTIVITIES if v961407a =0 then v961407a =.; if v961407b =0 then v961407b =.; if v961407c =0 then v961407c =.; if v961407d =0 then v961407d =.; v961407a (0) v961407b (0) v961407c (0) v961407d (0) LITERARY, ART OR DISCUSSION MEETINGS/ACTIVITIES if v961412a =0 then v961412a =.; if v961412b =0 then v961412b =.; if v961412c =0 then v961412c =.; if v961412d =0 then v961412d =.; v961412a (0) v961412b (0) v961412c (0) v961412d (0) HOBBY OR LEISURE MEETINGS/ACTIVITIES if v961417a =0 then v961417a =.; if v961417b =0 then v961417b =.; if v961417c =0 then v961417c =.; if v961417d =0 then v961417d =.; v961417a (0) v961417b (0) v961417c (0) v961417d (0) NEIGHBORHOOD OR COMMUNITY MEETINGS/ACTIVITIES if v961422a =0 then v961422a =.; if v961422b =0 then v961422b =.; if v961422c =0 then v961422c =.; if v961422d =0 then v961422d =.; v961422a (0) v961422b (0) v961422c (0) v961422d (0) SERVICE/FRATERNAL MEETINGS/ACTIVITIES if v961427a =0 then v961427a =.; if v961427b =0 then v961427b =.; if v961427c =0 then v961427c =.; if v961427d =0 then v961427d =.; v961427a (0) v961427b (0) v961427c (0) v961427d (0) SERVICE TO NEEDY MEETINGS/ACTIVITIES if v961432a =0 then v961432a =.; if v961432b =0 then v961432b =.; if v961432c =0 then v961432c =.; if v961432d =0 then v961432d =.; v961432a (0) v961432b (0) v961432c (0) v961432d (0) EDUCATIONAL MEETINGS/ACTIVITIES if v961437a =0 then v961437a =.; if v961437b =0 then v961437b =.; if v961437c =0 then v961437c =.; if v961437d =0 then v961437d =.; v961437a (0) v961437b (0) v961437c (0) v961437d (0) CULTURAL SERVICE MEETINGS/ACTIVITIES if v961442a =0 then v961442a =.; if v961442b =0 then v961442b =.; if v961442c =0 then v961442c =.; if v961442d =0 then v961442d =.; v961442a (0) v961442b (0) v961442c (0) v961442d (0) SELF-HELP MEETINGS/ACTIVITIES if v961447a =0 then v961447a =.; if v961447b =0 then v961447b =.; if v961447c =0 then v961447c =.; if v961447d =0 then v961447d =.; v961447a (0) v961447b (0) v961447c (0) v961447d (0) OTHER MEETINGS/ACTIVITIES if v961452a =0 then v961452a =.; if v961452b =0 then v961452b =.; if v961452c =0 then v961452c =.; if v961452d =0 then v961452d =.; v961452a (0) v961452b (0) v961452c (0) v961452d (0) ************************************************************ ************************************************************ FOR POLITICAL DISCUSSION: the group-specific data may be found in the columns below. For each group category, the first variable corresponds to the response for Group1 mention, the second variable corresponds to the response for Group2 mention, etc. LABOR UNIONS POLITICAL DISCUSSION if v961348a =0 then v961348a =.; if v961348b =0 then v961348b =.; if v961348c =0 then v961348c =.; if v961348d =0 then v961348d =.; v961348a (0) v961348b (0) v961348c (0) v961348d (0) BUSINESS OR WORK-RELATED POLITICAL DISCUSSION if v961353a =0 then v961353a =.; if v961353b =0 then v961353b =.; if v961353c =0 then v961353c =.; if v961353d =0 then v961353d =.; v961353a (0) v961353b (0) v961353c (0) v961353d (0) VETERANS POLITICAL DISCUSSION if v961358a =0 then v961358a =.; if v961358b =0 then v961358b =.; if v961358c =0 then v961358c =.; if v961358d =0 then v961358d =.; v961358a (0) v961358b (0) v961358c (0) v961358d (0) CHURCH/SYNAGOGUE POLITICAL DISCUSSION if v961363a =0 then v961363a =.; if v961363b =0 then v961363b =.; if v961363c =0 then v961363c =.; if v961363d =0 then v961363d =.; v961363a (0) v961363b (0) v961363c (0) v961363d (0) OTHER RELIGIOUS POLITICAL DISCUSSION if v961368a =0 then v961368a =.; if v961368b =0 then v961368b =.; if v961368c =0 then v961368c =.; if v961368d =0 then v961368d =.; v961368a (0) v961368b (0) v961368c (0) v961368d (0) ELDERLY/SENIOR POLITICAL DISCUSSION if v961373a =0 then v961373a =.; if v961373b =0 then v961373b =.; if v961373c =0 then v961373c =.; if v961373d =0 then v961373d =.; v961373a (0) v961373b (0) v961373c (0) v961373d (0) ETHNIC/NATIONALITY POLITICAL DISCUSSION if v961378a =0 then v961378a =.; if v961378b =0 then v961378b =.; if v961378c =0 then v961378c =.; if v961378d =0 then v961378d =.; v961378a (0) v961378b (0) v961378c (0) v961378d (0) WOMEN'S RIGHTS OR WELFARE POLITICAL DISCUSSION if v961383a =0 then v961383a =.; if v961383b =0 then v961383b =.; if v961383c =0 then v961383c =.; if v961383d =0 then v961383d =.; v961383a (0) v961383b (0) v961383c (0) v961383d (0) POLITICAL ISSUE POLITICAL DISCUSSION if v961388a =0 then v961388a =.; if v961388b =0 then v961388b =.; if v961388c =0 then v961388c =.; if v961388d =0 then v961388d =.; v961388a (0) v961388b (0) v961388c (0) v961388d (0) NONPARTISAN OR CIV96IC if v961393a =0 then v961393a =.; if v961393b =0 then v961393b =.; if v961393c =0 then v961393c =.; if v961393d =0 then v961393d =.; v961393a (0) v961393b (0) v961393c (0) v961393d (0) LIBERAL OR CONSERVATIVE POLITICAL DISCUSSION if v961398a =0 then v961398a =.; if v961398b =0 then v961398b =.; if v961398c =0 then v961398c =.; if v961398d =0 then v961398d =.; v961398a (0) v961398b (0) v961398c (0) v961398d (0) POLITICAL PARTY OR CAND SUPPORT POLITICAL DISCUSSION if v961403a =0 then v961403a =.; if v961403b =0 then v961403b =.; if v961403c =0 then v961403c =.; if v961403d =0 then v961403d =.; v961403a (0) v961403b (0) v961403c (0) v961403d (0) CHILDRENS' ACTIVITY POLITICAL DISCUSSION if v961408a =0 then v961408a =.; if v961408b =0 then v961408b =.; if v961408c =0 then v961408c =.; if v961408d =0 then v961408d =.; v961408a (0) v961408b (0) v961408c (0) v961408d (0) LITERARY, ART OR DISCUSSION POLITICAL DISCUSSION if v961413a =0 then v961413a =.; if v961413b =0 then v961413b =.; if v961413c =0 then v961413c =.; if v961413d =0 then v961413d =.; v961413a (0) v961413b (0) v961413c (0) v961413d (0) HOBBY OR LEISURE POLITICAL DISCUSSION if v961418a =0 then v961418a =.; if v961418b =0 then v961418b =.; if v961418c =0 then v961418c =.; if v961418d =0 then v961418d =.; v961418a (0) v961418b (0) v961418c (0) v961418d (0) NEIGHBORHOOD OR COMMUNITY POLITICAL DISCUSSION if v961423a =0 then v961423a =.; if v961423b =0 then v961423b =.; if v961423c =0 then v961423c =.; if v961423d =0 then v961423d =.; v961423a (0) v961423b (0) v961423c (0) v961423d (0) SERVICE/FRATERNAL POLITICAL DISCUSSION if v961428a =0 then v961428a =.; if v961428b =0 then v961428b =.; if v961428c =0 then v961428c =.; if v961428d =0 then v961428d =.; v961428a (0) v961428b (0) v961428c (0) v961428d (0) SERVICE TO NEEDY POLITICAL DISCUSSION if v961433a =0 then v961433a =.; if v961433b =0 then v961433b =.; if v961433c =0 then v961433c =.; if v961433d =0 then v961433d =.; v961433a (0) v961433b (0) v961433c (0) v961433d (0) EDUCATIONAL POLITICAL DISCUSSION if v961438a =0 then v961438a =.; if v961438b =0 then v961438b =.; if v961438c =0 then v961438c =.; if v961438d =0 then v961438d =.; v961438a (0) v961438b (0) v961438c (0) v961438d (0) CULTURAL SERVICE POLITICAL DISCUSSION if v961433a =0 then v961433a =.; if v961433b =0 then v961433b =.; if v961433c =0 then v961433c =.; if v961433d =0 then v961433d =.; v961443a (0) v961443b (0) v961443c (0) v961443d (0) SELF-HELP POLITICAL DISCUSSION if v961448a =0 then v961448a =.; if v961448b =0 then v961448b =.; if v961448c =0 then v961448c =.; if v961448d =0 then v961448d =.; v961448a (0) v961448b (0) v961448c (0) v961448d (0) OTHER POLITICAL DISCUSSION if v961453a =0 then v961453a =.; if v961453b =0 then v961453b =.; if v961453c =0 then v961453c =.; if v961453d =0 then v961453d =.; v961453a (0) v961453b (0) v961453c (0) v961453d (0) >> POST-STRATIFIED CROSS-SECTIONAL ANALYSIS WEIGHTS FOR THE 1992, 1994 AND 1996 NES DATA Prepared by the Sampling Section Division of Surveys and Technologies Survey Research Center Institute for Social Research University of Michigan 1. Overview: Why is NES issuing new weight variables? A new set of weights has been constructed for use with the series of National Election Studies beginning with the 1992 Pre-Election Study. This series includes the 1992 Pre and Post, the 1994 Post, and the 1996 Pre and Post Election Studies. The main difference between these and the previously released weights is found in the post-stratification criteria. The new weights post-stratify the National Election Study data to match the Current Population Study (CPS) estimate of the distribution of age group by education level. The previous set of weights adjusted the NES sample to the CPS distribution for Census Region, sex, and age group. These new weights correct for an under-representation of younger and less educated respondents in each year's sample of respondents mainly due to attrition of these categories of respondents in the panel component. The previous set of analysis weights developed for the 1996 NES public use data sets led to overestimation of reported voter turnout in the 1996 presidential election. A comparison between the 1992 and the 1996 presidential vote turnout estimates from the NES samples does not to reflect the trend of declining participation that has been evident from external sources, such as the Current Population Survey turnout estimates. Several sources of bias caused of this problem, leading to under-representation of 18-22 year olds in the 1996 NES sample, respondents with no high school diploma, or both. The significance of this under-representation becomes clear when the rates of voting participation by age and education subgroups are examined. The results are summarized in Tables 1a and 1b, below. Table 1a clearly demonstrates the well-known strong relationship between education and voting: people with less education are less likely to vote. Table 1b shows that reported voter turnout is higher among older people. Since the age and education groups with the lowest voting rates are underrepresented, estimates of 1996 presidential election participation are skewed in the direction of higher rates of turnout. Table 1a: Reported turnout in the 1996 presidential election by education level of respondent (source: 1996 NES). Education % reporting having voted No HS diploma 57.1 High school diploma 69.1 Some college 80.7 College Graduate 89.9 Total 76.6 Table 1b: Reported turnout in the 1996 presidential election by age group of the respondent (source: 1996 NES). Age % reporting having voted 18-21 54.6 22-29 59.2 30-39 73.3 40-49 80.7 50-59 81.0 60-69 81.8 70+ 84.5 Total 76.6 The following three sections describe the three major factors which contribute to the under-representation of specific age or education groups. These include "initial contact non-response bias," "coverage bias resulting from longitudinal sample design" and "education related attrition bias." Subsequent sections describe in detail the procedures used in the construction of the new weights. 2. Initial Contact Nonresponse Bias The first important source of age and education related bias is nonresponse bias at the initial interview. Initial contact nonresponse bias occurs when people with a certain characteristic in common have a significantly different response rate from the overall response rate. For example, if women are found to have a much higher response rate than the combined response rate for men and women, then there is an initial contact nonresponse bias based on gender. If there were no nonresponse bias based on age or education we would expect the NES cross-section samples to have age by education distributions similar to that of the Current Population Survey (CPS) population estimates. There would be minor differences attributable to sampling error, but we would not expect to find large or systematic differences. Table 2, which compares the weighted distributions of education for the 1992, 1994 and 1996 NES cross-section samples to CPS population estimates for the same years suggests that systematic differences are present. The weight used in Table 2 is the calculated base weight. This weight is the product of a person-level selection weight and a household-level nonresponse adjustment factor. Since the selection probability of an eligible adult is inversely proportional to the number of eligible adults in the household it is important to use the selection weight based on the number of eligible adults in the household when comparing NES person-level statistics to CPS person-level distributions. The base weight also adjusts for the difference in response rates by region and by urbanicity. The construction of these weight factors is described in Sections 5 through 8. This part of the NES weight is essentially the same for the old and new weights. In Table 2, CPS estimates for 1992, 1994 and 1996 are included in the shaded columns. Comparisons of the weighted cross-section data from 1992, 1994 and 1996 to the corresponding CPS estimates reveal clear systematic differences which cannot be wholly attributed to sampling error. In all three cross-section groups there is a strong relationship between the level of education achieved by the respondent and the nonresponse rate. Specifically, people with less education -- especially people without a high school diploma - tend to be underrepresented in the weighted cross-section samples. Table 2: Summary of weighted cross-section distributions by education 1992 CPS 1992 pre 1994 CPS 1994 post 1996 CPS 1996 pre propor- (weighted) propor- (weighted) propor- (weighted) tion tion tion No HS Diploma 0.208 0.144 0.195 0.161 0.189 0.103 HS Diploma 0.355 0.321 0.339 0.356 0.332 0.338 Some College 0.243 0.270 0.264 0.258 0.264 0.323 College Graduate 0.195 0.265 0.203 0.226 0.215 0.236 3. Coverage Bias Resulting from Longitudinal Sample Design The longitudinal design of the National Election Study results in a coverage bias in the 1992 and 1994 cross-section component of the 1996 sample. Respondents age 18-19 had no chance of being observed in the panel. Respondents age 20 or 21 years old had a chance of inclusion in only the 1994 cross-section component of the 1996 panel. This structural bias in cross-sectional estimates based on the combined 1996 NES sample is an additional contributor to under-representation of the younger population. The age 18-21 bias in the sample also affects education since the youngest group (e.g., 18-22) has a natural constraint on the level of education that a respondent could have achieved by the time he or she was interviewed. 4. Education Related Attrition Bias Differential reinterview rates (pre to post as well as across election year waves) based on education also contribute to over- estimation of voting in the 1996 presidential election. The relationship between education and cumulative attrition is shown in Tables 3a-3c. Table 3a tracks the 1992 cross-section cases across subsequent interviews. The age groups listed in the left-most column refer to the respondent's age at the initial interview. Thus, a 29 year old respondent in 1992 would not move into the next higher age group in 1994. Columns labeled "%" indicate the percent of the original sample that was reinterviewed. For example, in Table 3a, under 1996 (pre), there is a column labeled "n" and a column labeled "%". The value in the top row in the "%" column is 71.4%. This means that 71.4 percent of the seven 18-21 year olds with no HS diploma were included in the panel component of the 1996 pre election interview. Sample Tables 3b and 3c show the attrition for the 1994 and 1996 cross-section components. The summaries of cumulative attrition by education group portray a strong relationship between education and reinterview rate. Respondents with more education are more likely to participate in subsequent interviews. This difference in attrition rate is found between pre and post interviews of the same year (Table 3a - 1992 Post, Table 3c - 1996 Post) as well as across interview years (Table 3b - 1996 Pre). Initially biased samples are subjected to further nonresponse bias at every subsequent interview, causing significant under-representation of less educated, eligible voters. Since eligible adults with low education are less likely to vote and are under-represented in the sample, predictions of voting participation will be biased upward. Table 3a: Cumulative attrition for the 1992 NES Cross-section sample 1992 1994 1996 (pre) (post) (post) (pre) (post) AGE HIGHEST (in EDUCATION n n % n % n % n % 1992) 18-21 No HS Diploma 7 7 100.0 7 100.0 5 71.4 3 42.9 HS Diploma 30 27 90.0 18 60.0 11 36.7 6 20.0 Some College 24 23 95.8 18 75.0 15 62.5 14 58.3 College Graduate 1 1 100.0 1 100.0 0 0.0 0 0 TOTAL 62 58 93.5 44 71.0 31 50.0 23 37.1 22-29 No HS Diploma 15 15 100.0 8 53.3 6 40.0 6 40.0 HS Diploma 53 47 88.7 29 54.7 17 32.1 15 28.3 Some College 63 56 88.9 44 69.8 38 60.3 34 54.0 College Graduate 42 38 90.5 29 69.0 26 61.9 23 54.8 TOTAL 173 156 90.2 110 63.6 87 50.3 78 45.1 30-39 No HS Diploma 23 22 95.7 16 69.6 11 47.8 11 47.8 HS Diploma 89 78 87.6 56 62.9 44 49.4 41 46.1 Some College 93 86 92.5 72 77.4 54 58.1 49 52.7 College Graduate 107 103 96.3 78 72.9 62 57.9 58 54.2 TOTAL 312 289 92.6 222 71.2 171 54.8 159 51.0 40-49 No HS Diploma 13 13 100.0 9 69.2 6 46.2 5 38.5 HS Diploma 52 48 92.3 35 67.3 28 53.8 24 46.2 Some College 48 40 83.3 27 56.3 21 43.8 20 41.7 College Graduate 70 62 88.6 50 71.4 41 58.6 38 54.3 TOTAL 183 163 89.1 121 66.1 96 52.5 87 47.5 50-59 No HS Diploma 27 24 88.9 17 63.0 15 55.6 14 51.9 HS Diploma 43 40 93.0 33 76.7 26 60.5 22 51.2 Some College 28 25 89.3 18 64.3 14 50.0 14 50.0 College Graduate 45 39 86.7 33 73.3 30 66.7 29 64.2 TOTAL 143 128 89.5 101 70.6 85 59.4 79 55.2 60-69 No HS Diploma 37 30 81.1 23 62.2 17 45.9 16 43.2 HS Diploma 50 39 78.0 30 60.0 24 48.0 24 48.0 Some College 19 14 73.7 10 52.6 9 47.4 9 47.4 College Graduate 16 16 100.0 13 81.3 12 75.0 11 68.8 TOTAL 122 99 81.1 76 62.3 62 50.8 60 49.2 70+ No HS Diploma 54 42 77.8 28 51.9 22 40.7 21 38.9 HS Diploma 31 30 96.8 22 71.0 15 48.4 14 45.2 Some College 27 24 88.9 20 74.1 16 59.3 14 51.9 College Graduate 19 16 84.2 15 78.9 12 63.2 10 52.6 TOTAL 131 112 85.5 85 64.9 65 49.6 59 45.0 1126 1005 759 597 545 Summary by Education level: 1992 pre 1992 post 1994 post 1996 pre 1996 post n n % n % n % n % No HS Diploma 176 153 86.9 108 61.4 82 46.6 76 43.2 HS Diploma 348 309 88.8 223 64.1 165 47.4 146 42.0 Some College 302 268 88.7 209 69.2 167 55.3 154 51.0 College graduate 300 275 91.7 219 73.0 183 61.0 169 56.3 Total 1126 1005 89.3 759 67.4 597 53.0 545 48.4 Table 3b: Cumulative attrition for the 1994 NES Cross-section sample 1994 1996 (post) (pre) (post) AGE HIGHEST n n % n % (at EDUCATION 1994) 18-21 No HS Diploma 13 8 61.5 4 30.8 HS Diploma 24 13 54.2 9 37.5 Some College 18 10 55.6 7 38.9 College Graduate 0 0 0 TOTAL 55 31 56.4 20 36.4 22-29 No HS Diploma 14 6 42.9 4 28.6 HS Diploma 45 31 68.9 26 57.8 Some College 58 37 63.8 33 56.9 College Graduate 35 24 68.6 22 62.9 TOTAL 152 98 64.5 85 55.9 30-39 No HS Diploma 21 16 76.2 13 61.9 HS Diploma 93 59 63.4 53 57.0 Some College 73 45 61.6 40 54.8 College Graduate 59 44 74.6 40 67.8 TOTAL 246 164 66.7 146 59.3 40-49 No HS Diploma 14 10 71.4 8 57.1 HS Diploma 53 39 73.6 37 69.8 Some College 52 40 76.9 37 71.2 College Graduate 67 54 80.6 51 76.4 TOTAL 186 143 76.9 133 71.5 50-59 No HS Diploma 16 11 68.8 10 62.5 HS Diploma 43 33 76.7 27 62.8 Some College 24 19 79.2 19 79.2 College Graduate 29 21 72.4 21 72.4 TOTAL 112 84 75.0 77 68.8 60-69 No HS Diploma 42 30 71.4 28 66.7 HS Diploma 62 42 67.7 40 64.5 Some College 21 16 76.2 15 71.4 College Graduate 19 17 89.5 17 89.5 TOTAL 144 105 72.9 100 69.4 70+ No HS Diploma 51 32 62.7 31 60.8 HS Diploma 42 30 71.4 29 69.0 Some College 22 12 54.5 11 50.0 College Graduate 26 20 76.9 20 76.9 TOTAL 141 94 66.7 91 64.5 1036 719 652 Summary by Education level: 1994 post 1996 pre 1996 post n n % n % No HS Diploma 171 113 66.1 98 57.3 HS Diploma 362 247 68.2 221 61.0 Some College 268 179 66.8 162 60.4 College Graduate 235 180 76.6 171 72.8 Total 1036 719 69.4 652 62.9 Table 3c: Cumulative attrition for the 1996 NES Cross-section sample 1996 (pre) (post) AGE HIGHEST (at EDUCATION n n % 1996) 18-21 No HS Diploma 3 2 66.7 HS Diploma 9 7 77.8 Some College 23 21 91.3 College Graduate 0 0 TOTAL 35 30 85.7 22-29 No HS Diploma 4 2 50.0 HS Diploma 19 13 72.2 Some College 13 10 76.9 College Graduate 17 16 94.1 TOTAL 52 41 78.8 30-39 No HS Diploma 4 4 100.0 HS Diploma 36 29 80.6 Some College 31 29 93.5 College Graduate 28 23 82.1 TOTAL 99 85 85.9 40-49 No HS Diploma 5 4 80.0 HS Diploma 23 18 78.3 Some College 25 20 80.0 College Graduate 22 19 86.4 TOTAL 75 61 81.3 50-59 No HS Diploma 7 6 85.7 HS Diploma 17 15 88.2 Some College 17 15 88.2 College Graduate 15 15 100.0 TOTAL 56 51 91.1 60-69 No HS Diploma 9 9 100.0 HS Diploma 12 11 91.7 Some College 9 7 77.8 College Graduate 7 6 85.7 TOTAL 37 33 89.2 70+ No HS Diploma 13 10 76.9 HS Diploma 22 18 81.8 Some College 6 5 83.3 College Graduate 3 3 100.0 TOTAL 44 36 81.8 398 337 Summary by Education level: 1996 pre 1996 post n n % No HS Diploma 45 37 82.2 HS Diploma 137 111 81.0 Some College 124 107 86.3 College Graduate 92 82 89.1 Total 398 337 84.7 5. Construction of the new weights The revised NES final analysis weight is based on the product of a calculated base weight and a post-stratification factor. The base weight is constructed to adjust for selection probability and geographic differences in response rates at the time of the initial interview with each sample component. This weight is the product of a selection probability weight and the household nonresponse factor. The base weights for 1992, 1994, and 1996 cross-section cases are initially determined using the corresponding year's household nonresponse factor. Panel cases use this same base weight, carried over from the original interview. Since differences in selection probabilities for the NES sample household are due only to random selection of a single adult from households of various sizes, the selection probability weight is the number of eligible people in the household (up to three). The post-stratification factor is the ratio of the census proportion for each age by education subgroup, to the corresponding weighted ( base weight ) sample proportion. Multiplication of the base weight by this post-stratification factor adjusts the weighted sample distribution to conform to the CPS population estimates. The following sections describe the base weight and post-stratification factors in further detail. Final Weight = base weight x post-stratification factor where: Base weight = selection weight x household nonresponse factor and: Selection weight = the number of eligible adults in household (up to three) 6. Construction of a Base Weight The base weight is the product of two factors: the selection weight and the household nonresponse adjustment factor. Although the National Election Study uses an area probability sample design to achieve an equal probability sample of U.S. households, the NES design does not produce an equal probability sample of persons. Since only one person is chosen from each selected household, any particular individual's probability of selection is inversely proportional to the number of eligible adults in the household. The selection weight which is equal to the number of eligible persons in the household (inverse of the selection probability) adjusts for the under-representation of persons in larger households. The household nonresponse factor is used to adjust for the differential nonresponse rates found in different regions and PSU types (Self-representing MSA, Nonself-representing MSA, and non-MSA. Self-representing MSAs are the largest Metropolitan Statistical Areas in the nation and are therefore self-representing in the 1990 SRC National Sample; Nonself-representing MSAs are medium and smaller sized MSAs, and the non-MSAs are counties which are not designated as MSAs and are less urban. 7. Selection Probability Weight: The National Election Study uses an area probability sample design to achieve an equal probability sample of U.S. households. If a household has only one eligible adult, that person is included in the sample. If a selected household has more than one eligible adult, one is selected at random. Since the number of eligible adults varies across households, the probability of selection for individuals is unequal and a weight which is the reciprocal of the probability of selection should be used. In the interest of limiting the variation of the weights, respondents selected from households with more than three eligible adults were assigned a weight of three; otherwise the selection weight is equal to the number of eligible adults. 8. Household Nonresponse Adjustment Factor: Nonresponse bias is a potential source of nonsampling error in the NES data. It has been found that response rates vary significantly by geographic region and PSU type (MSA/non-MSA status). In an effort to counteract this potential source of bias, adjustment factors have been constructed at the household level to account for the geographic and urban/rural differences in response rates. Table 4 shows the initial contact response rates in the 1992, 1994 and 1996 NES by PSU type and region. The nonresponse adjustment factor was determined by dividing the cross-section cases among twelve cells of four regions (Northeast / Midwest / South /West) by three PSU types (SR MSA, NSR MSA, NSR Non- MSA). The cases in each cell share a nonresponse adjustment factor calculated as the inverse of the response rate of the cell. These response rates are for the initial cross-section components only. They do not include the panel cases. Table 4: Initial contact response rates by PSU type and region 1992 Response 1994 Response 1996 Response PSU Type Region rate rate rate SR MSA Northeast 0.683 0.570 0.423 Midwest 0.759 0.651 0.533 South 0.724 0.620 0.539 West 0.471 0.517 0.507 NSR MSA Northeast 0.741 0.577 0.526 Midwest 0.699 0.717 0.678 South 0.727 0.813 0.646 West 0.723 0.782 0.625 NSR Non-MSA Northeast 0.820 0.725 0.600 Midwest 0.917 0.878 0.721 South 0.830 0.736 0.687 West 0.762 0.946 0.810 9. Comparison of Weighted NES and CPS Age Group by Education Level Distributions Table 5a below shows the current interview age by education distributions of 1992 cross-section cases in initial and subsequent interviews. The table includes weighted (base weight) percentages and unweighted percentages with estimates of the population percentages according to the Current Population Study included for comparison. We can see for example, that in the 1992 NES pre election sample there were 15 respondents age 22-29 with no high school diploma. These represent approximately 1.3 percent of the 1126 total respondents in this sample. When the base weight is used, the weighted percent for this group increases to about 1.6 percent. The 1992 CPS population estimates are listed in a column on the left. It is estimated that in 1992 about 2.4 percent of all eligible adults were 22-29 year-olds with no high school diploma. The shaded rows indicate totals by age group and a summary by education is provided at the bottom of the page. Table 5b gives the same information for the 1994 cross-section cases and Table 5c shows the 1996 cross-section distributions. Table 5a: Distribution of the 1992 NES Cross-section sample by current age and education AGE HIGHEST 1992 Unwtd Wghted Unwtd Wghted (Cur- EDUCATION CPS n & % n % % rent ) (Sel,NR) (Sel,NR) 18-21 No College 4.3 37 3.3 4.6 34 3.4 4.7 College 3.1 25 2.2 2.3 24 2.4 2.6 TOTAL 7.3 62 5.5 7.0 58 5.8 7.3 22-29 No HS Diploma 2.4 15 1.3 1.6 15 1.5 1.8 HS Diploma 6.1 53 4.7 4.5 47 4.7 4.6 Some College 4.8 63 5.6 5.6 56 5.6 5.6 College Graduate 3.5 42 3.7 3.7 38 3.8 3.8 TOTAL 16.7 173 15.4 15.4 156 15.5 15.8 30-39 No HS Diploma 3.0 23 2.0 1.6 22 2.2 1.7 HS Diploma 8.7 89 7.9 8.0 78 7.8 7.8 Some College 6.1 93 8.3 8.0 86 8.6 8.3 College Graduate 5.7 107 9.5 9.2 103 10.2 10.0 TOTAL 23.4 312 27.7 26.8 289 28.8 27.8 40-49 No HS Diploma 2.4 13 1.2 1.2 13 1.3 1.3 HS Diploma 6.1 52 4.6 5.1 48 4.8 5.2 Some College 4.7 48 4.3 4.7 40 4.0 4.2 College Graduate 5.0 70 6.2 6.3 62 6.2 6.2 TOTAL 18.1 183 16.3 17.2 163 16.2 16.9 50-59 No HS Diploma 2.8 27 2.4 2.5 24 2.4 2.4 HS Diploma 4.7 43 3.8 4.6 40 4.0 4.8 Some College 2.4 28 2.5 2.4 25 2.5 2.5 College Graduate 2.5 45 4.0 4.2 39 3.9 4.1 TOTAL 12.3 143 12.7 13.7 128 12.7 13.7 60-69 No HS Diploma 3.5 37 3.3 3.0 30 3.0 2.7 HS Diploma 4.2 50 4.4 4.0 39 3.9 3.5 Some College 1.8 19 1.7 1.8 14 1.4 1.4 College Graduate 1.7 16 1.4 1.5 16 1.6 1.7 TOTAL 11.1 122 10.8 10.2 99 9.9 9.3 70+ No HS Diploma 4.8 54 4.8 3.8 42 4.2 3.1 HS Diploma 3.6 31 2.8 2.2 30 3.0 2.4 Some College 1.5 27 2.4 2.3 24 2.4 2.2 College Graduate 1.2 19 1.7 1.5 16 1.6 1.5 TOTAL 11.1 131 11.6 9.8 112 11.1 9.2 1126 1005 by Education Summary level: 1992 pre 1992 post 92 CPS n Unwtd % Wtd % n Unwtd % Wtd % No HS Diploma 20.8 176 15.6 14.4 153 15.2 13.9 HS Diploma 35.5 348 30.9 32.1 309 30.7 32.1 Some College 24.3 302 26.8 27.0 268 26.7 26.7 College Graduate 19.5 300 26.6 26.5 275 27.4 27.4 Total 1126 1005 Table 5a: (cont.): Distribution of the 1992 NES Cross-section sample by current age and education 1994 post 1996 pre 1996 post AGE HIGHEST 1992 Unwtd Wghtd Unwtd Wghtd Unwtd Wghtd (Cur- EDUCATION CPS n % % n % % n % % rent) (Sel,NR) (Sel,NR) (Sel,NR) 18-21 No College 4.3 13 1.7 2.5 0 0.0 0.0 0 0.0 0.0 College 3.1 4 0.5 0.7 1 0.2 0.3 1 0.2 0.3 TOTAL 7.3 17 2.2 3.2 1 6.2 0.3 1 0.2 0.3 22-29 No HS Diploma 2.4 9 1.2 1.1 4 0.7 0.8 3 0.6 0.7 HS Dip- loma 6.1 27 3.6 4.2 20 3.4 4.1 15 2.8 3.2 Some College 4.8 46 6.1 6.1 21 3.5 3.8 18 3.3 3.6 College Graduate 3.5 16 2.1 2.1 22 3.7 4.0 20 3.7 4.0 TOTAL 16.7 98 12.9 13.5 67 11.2 12.7 56 10.3 11.5 30-39 No HS Diploma 3.0 16 2.1 1.7 10 1.7 1.6 10 1.8 1.7 HS Dip- loma 8.7 54 7.1 7.2 40 6.7 6.3 37 6.8 6.5 Some College 6.1 77 10.1 9.7 54 9.0 8.7 47 8.6 8.2 College Graduate 5.7 74 9.8 9.6 54 9.0 9.3 50 9.2 9.4 TOTAL 23.4 221 29.1 28.2 158 26.5 25.9 144 26.4 25.8 40-49 No HS Diploma 2.4 11 1.4 1.3 6 1.0 0.7 5 0.9 0.6 HS Dip- loma 6.1 39 5.1 5.7 40 6.7 7.3 35 6.4 7.1 Some College 4.7 26 3.4 3.5 20 3.4 3.8 20 3.7 4.2 College Graduate 5.0 63 8.3 8.1 59 9.9 9.4 53 9.7 9.3 TOTAL 18.1 139 18.3 18.6 125 20.9 21.2 113 20.7 21.2 50-59 No HS Diploma 2.8 13 1.7 1.8 10 1.7 1.9 10 1.8 2.1 HS Dip- loma 4.7 35 4.6 5.1 29 4.9 5.3 24 4.4 4.6 Some College 2.4 23 3.0 3.2 22 3.7 4.0 22 4.0 4.3 College Graduate 2.5 32 4.2 4.7 28 4.7 4.8 27 5.0 5.1 TOTAL 12.3 103 13.6 14.8 89 14.9 15.9 83 15.2 16.1 60-69 No HS Diploma 3.5 21 2.8 2.8 13 2.2 2.1 12 2.2 2.2 HS Dip- loma 4.2 28 3.7 3.6 22 3.7 3.6 22 4.0 3.9 Some College 1.8 10 1.3 1.2 10 1.7 1.6 10 1.8 1.8 College Graduate 1.7 15 2.0 1.8 18 3.0 2.9 17 3.1 3.1 TOTAL 11.1 74 9.7 9.3 63 10.6 10.2 61 11.2 10.9 70+ No HS Diploma 4.8 35 4.6 3.5 32 5.4 4.2 30 5.5 4.3 HS Dip- loma 3.6 30 4.0 3.4 25 4.2 3.6 23 4.2 3.7 Some College 1.5 23 3.0 2.9 21 3.5 3.2 19 3.5 3.2 College Graduate 1.2 19 2.5 2.6 16 2.7 2.8 15 2.8 2.8 TOTAL 11.1 107 14.1 12.4 94 15.7 13.8 87 16.0 14.1 759 597 545 Summary by Education level: 1994 post 1996 pre 1996 post 92 CPS n unwtd % wtd % n unwtd % wtd % n unwtd % wtd % No HS Diploma 20.8 108 14.2 12.7 75 12.6 11.2 70 12.8 11.6 HS Diploma 35.5 223 29.4 31.1 176 29.5 30.1 156 28.6 29.1 Some College 24.3 209 27.5 27.2 149 25.0 25.4 137 25.1 25.6 College Graduate 19.5 219 28.8 29.0 197 33.0 33.2 182 33.4 33.7 Total 759 597 545 Table 5b: Distribution of the 1994 NES Cross-section sample by current age and education 1994 post 1996 pre 1996 post AGE HIGHEST 1994 wghtd wghtd wghtd EDUCATION CPS n unwtd % % n unwtd % % n unwtd % % (Sel,NR) (Sel,NR) (Sel,NR) 18-21 No College 4.2 37 3.6 4.2 12 1.7 1.8 8 1.2 1.3 College 3.1 18 1.7 2.4 6 0.8 1.1 5 0.8 1.0 TOTAL 7.3 55 5.3 6.6 18 2.5 3.0 13 2.0 2.3 22-29 No HS Diploma 2.3 14 1.4 1.3 6 0.8 1.0 3 0.5 0.5 HS Diploma 5.5 45 4.3 4.5 23 3.2 3.8 17 2.6 3.0 Some College 5.3 58 5.6 5.7 31 4.3 4.0 27 4.1 3.9 College Graduate 3.4 35 3.4 3.3 22 3.1 3.0 20 3.1 3.1 TOTAL 16.5 152 14.7 14.7 82 11.4 11.7 67 10.3 10.5 30-39 No HS Diploma 2.9 21 2.0 2.1 12 1.7 1.7 9 1.4 1.4 HS Diploma 8.1 93 9.0 9.0 57 7.9 7.5 51 7.8 7.1 Some College 6.6 73 7.1 6.8 53 7.4 7.3 47 7.2 7.2 College Graduate 5.7 59 5.7 5.7 41 5.7 5.9 38 5.8 6.3 TOTAL 23.3 246 23.7 23.7 163 22.7 22.4 145 22.2 22.0 40-49 No HS Diploma 2.3 14 1.4 1.6 11 1.5 1.9 9 1.4 1.7 HS Diploma 6.1 53 5.1 6.0 43 6.0 6.5 41 6.3 6.8 Some College 5.2 52 5.0 5.0 43 6.0 6.3 39 6.0 6.4 College Graduate 5.4 67 6.5 6.6 57 7.9 8.1 53 8.1 8.4 TOTAL 19.0 186 18.0 19.2 154 21.4 22.8 142 21.8 23.3 50-59 No HS Diploma 2.4 16 1.5 1.6 12 1.7 1.6 12 1.8 1.8 HS Diploma 4.6 43 4.2 4.4 36 5.0 5.4 29 4.4 4.9 Some College 2.8 24 2.3 2.2 16 2.2 2.1 16 2.4 2.3 College Graduate 2.8 29 2.8 3.1 25 3.5 3.8 25 3.8 4.2 TOTAL 12.5 112 10.8 11.1 89 12.4 13.0 82 12.6 13.3 60-69 No HS Diploma 3.0 42 4.1 3.7 25 3.5 3.3 23 3.5 3.4 HS Diploma 3.8 62 6.0 5.5 39 5.4 5.2 35 5.4 5.0 Some College 1.9 21 2.0 1.9 21 2.9 3.1 21 3.2 3.4 College Graduate 1.7 19 1.8 2.0 14 2.0 1.9 14 2.2 2.1 TOTAL 10.3 144 13.9 13.2 99 13.8 13.4 93 14.3 13.9 70+ No HS Diploma 4.6 51 4.9 4.1 37 5.1 4.4 36 5.5 4.9 HS Diploma 3.7 42 4.1 3.6 33 4.6 4.1 32 4.9 4.4 Some College 1.7 22 2.1 1.8 22 3.1 2.4 21 3.2 2.6 College Graduate 1.3 26 2.5 2.0 22 3.1 2.8 21 3.2 2.9 TOTAL 11.2 141 13.6 11.5 114 15.9 13.7 110 16.9 14.7 1036 719 652 Summary by Education level: 1994 post 1996 pre 1996 post 94 CPS n Unwtd % Wtd % n Unwtd % Wtd % n Unwtd % Wtd % No HS Diploma 19.5 171 16.5 16.1 110 15.3 15.2 96 14.7 14.4 HS Diploma 33.9 362 34.9 35.6 236 32.8 33.1 209 32.1 31.8 Some College 26.4 268 25.9 25.8 192 26.7 26.3 176 27.0 26.8 College Graduate 20.3 235 22.7 22.6 181 25.2 25.4 171 26.2 27.0 Total 1036 719 652 Table 5c: Distribution of the 1996 NES Cross-section sample by current age and education 1996 pre 1996 post AGE HIGHEST 1996 Unwtd Wghtd Unwtd Wghtd (Cur- EDUCATION CPS n % % n % % rent) (Sel,NR) (Sel,NR) 18-21 No College 4.4 12 3.0 4.1 9 2.7 3.6 College 2.9 23 5.8 7.5 21 6.2 8.2 TOTAL 7.3 35 8.8 11.6 30 8.9 11.8 22-29 No HS Diploma 2.0 4 1.0 0.8 2 0.6 0.5 HS Dip- loma 4.9 18 4.5 3.9 13 3.9 3.3 Some College 5.0 13 3.3 2.9 10 3.0 2.9 College Graduate 3.7 17 4.3 4.0 16 4.8 4.4 TOTAL 15.6 52 13.1 11.5 41 12.2 11.0 30-39 No HS Diploma 2.9 4 1.0 0.8 4 1.2 0.9 HS Dip- loma 7.6 36 9.0 9.0 29 8.6 8.7 Some College 6.3 31 7.8 7.6 29 8.6 8.4 College Graduate 5.9 28 7.0 6.6 23 6.8 6.3 TOTAL 22.8 99 24.9 24.1 85 25.2 24.4 40-49 No HS Diploma 2.4 5 1.3 1.0 4 1.2 0.9 HS Dip- loma 6.6 23 5.8 6.2 18 5.3 5.6 Some College 5.5 25 6.3 6.8 20 5.9 6.3 College Graduate 5.7 22 5.5 5.5 19 5.6 5.7 TOTAL 20.1 75 18.8 19.6 61 18.1 18.5 50-59 No HS Diploma 2.3 7 1.8 1.7 6 1.8 1.7 HS Dip- loma 4.6 17 4.3 4.9 15 4.4 4.9 Some College 2.9 17 4.3 3.6 15 4.4 3.8 College Graduate 3.0 15 3.8 4.8 15 4.4 5.7 TOTAL 12.8 56 14.1 15.2 51 15.1 16.1 60-69 No HS Diploma 2.8 9 2.3 1.9 9 2.7 2.3 HS Dip- loma 3.7 12 3.0 2.3 11 3.3 2.6 Some College 1.9 9 2.3 2.5 7 2.1 2.2 College Graduate 1.8 7 1.8 2.2 6 1.8 2.3 TOTAL 10.1 37 9.3 8.9 33 9.8 9.3 70+ No HS Diploma 4.3 13 3.3 2.8 10 3.0 2.5 HS Dip- loma 3.7 22 5.5 4.6 18 5.3 4.5 Some College 1.9 6 1.5 1.3 5 1.5 1.4 College Graduate 1.5 3 0.8 0.5 3 0.9 0.6 TOTAL 11.3 44 11.1 9.2 36 10.7 8.9 398 337 Summary by Education level: 1996 pre 1996 post 96 CPS n Unwtd% Wtd% n Unwtd% Wtd% No HS Diploma 18.9 45 11.3 10.3 37 11.0 9.8 HS Diploma 33.2 137 34.4 33.8 111 32.9 32.1 Some College 26.4 124 31.2 32.3 107 31.8 33.1 College Graduate 21.5 92 23.1 23.6 82 24.3 25.0 Total 398 337 9. Post-stratification Factor for the Revised Weights: The post-stratification factor for the revised NES cross-sectional weights was developed to address problems caused by under-representation of age or education groups. To do this, the corresponding CPS estimates were used as the benchmark standard. The post-stratification factor was calculated by dividing the CPS percent by the weighted (base weight) NES percent for each of the age by education subgroups. Note that the youngest age group consists of only two education groups (no college / at least some college) because of the small number of 18 to 21 year-olds in the samples (especially in 1994 and 1996)and because level of education is not as meaningful for the youngest age group since they may still be in school. Tables 6a, 6b and 6c show the data used to construct the post- stratification factors for the combined panel and cross-section NES samples for each year. As an example of the calculation, in the 1994 NES sample (Table 6b) there were fifty 18-21 year olds with no college education. These people represent approximately 2.8 percent (unweighted) of the 1994 sample. When the base weight is applied, the weighted percent is about 3.5. On the left side of each table the CPS statistics for the corresponding year are listed. These are used as estimates of the population percentages by age and education. The post- stratification factor is calculated for each subgroup by dividing the CPS estimate by the weighted percent. In the 1994 example this is 4.2 divided by approximately 3.5. Although the percentages in the tables are shown to the nearest tenth of a percent, the calculation of the post-stratification factors used percents to the nearest hundredth of a percent. Table 6a: Distributions and post-stratification factors for the combined 1992 samples 1992 pre 1992 post AGE HIGHEST 1992 Unwtd Wghtd Post-strat Unwtd Wghtd Post-strat (Cur- EDUCATION CPS n % % factor n % % factor rent) (Sel,NR) (92 cps) (Sel,NR) (92 cps) 18-21 No College 4.3 37 3.3 4.6 0.918 34 3.4 4.7 0.900 College 3.1 25 2.2 2.3 1.313 24 2.4 2.6 1.200 TOTAL 7.3 62 5.5 7.0 58 5.8 7.3 22-29 No HS Diploma 2.4 15 1.3 1.6 1.506 15 1.5 1.8 1.343 HS Dip- loma 6.1 53 4.7 4.5 1.354 47 4.7 4.6 1.319 Some College 4.8 63 5.6 5.6 0.857 56 5.6 5.6 0.864 College Graduate 3.5 42 3.7 3.7 0.935 38 3.8 3.8 0.908 TOTAL 16.7 173 15.4 15.4 156 15.5 15.8 30-39 No HS Diploma 3.0 23 2.0 1.6 1.833 22 2.2 1.7 1.747 HS Dip- loma 8.7 89 7.9 8.0 1.083 78 7.8 7.8 1.109 Some College 6.1 93 8.3 8.0 0.763 86 8.6 8.3 0.733 College Graduate 5.7 107 9.5 9.2 0.615 103 10.2 10.0 0.567 TOTAL 23.4 312 27.7 26.8 289 28.8 27.8 40-49 No HS Diploma 2.4 13 1.2 1.2 2.009 13 1.3 1.3 1.794 HS Dip- loma 6.1 52 4.6 5.1 1.204 48 4.8 5.2 1.180 Some College 4.7 48 4.3 4.7 1.013 40 4.0 4.2 1.113 College Graduate 5.0 70 6.2 6.3 0.791 62 6.2 6.2 0.797 TOTAL 18.1 183 16.3 17.2 163 16.2 16.9 50-59 No HS Diploma 2.8 27 2.4 2.5 1.118 24 2.4 2.4 1.155 HS Dip- loma 4.7 43 3.8 4.6 1.020 40 4.0 4.8 0.973 Some College 2.4 28 2.5 2.4 0.959 25 2.5 2.5 0.955 College Graduate 2.5 45 4.0 4.2 0.594 39 3.9 4.1 0.609 TOTAL 12.3 143 12.7 13.7 128 12.7 13.7 60-69 No HS Diploma 3.5 37 3.3 3.0 1.182 30 3.0 2.7 1.282 HS Dip- loma 4.2 50 4.4 4.0 1.055 39 3.9 3.5 1.199 Some College 1.8 19 1.7 1.8 1.000 14 1.4 1.4 1.250 College Graduate 1.7 16 1.4 1.5 1.114 16 1.6 1.7 0.994 TOTAL 11.1 122 10.8 10.2 99 9.9 9.3 70+ No HS Diploma 4.8 54 4.8 3.8 1.268 42 4.2 3.1 1.540 HS Dip- loma 3.6 31 2.8 2.2 1.633 30 3.0 2.4 1.490 Some College 1.5 27 2.4 2.3 0.642 24 2.4 2.2 0.671 College Graduate 1.2 19 1.7 1.5 0.791 16 1.6 1.5 0.818 TOTAL 11.1 131 11.6 9.8 112 11.1 9.2 1126 1005 Summary by Education Level: 1992 pre 1992 post 92 CPS n Unwtd% Wtd% n Unwtd% Wtd% No HS Diploma 20.8 176 15.6 14.4 153 15.2 13.9 HS Diploma 35.5 348 30.9 32.1 309 30.8 32.1 Some College 24.3 302 26.8 27.0 268 26.7 26.7 College Graduate 19.5 300 26.6 26.5 275 27.4 27.4 Total 1126 1005 Table 6b: Distributions and post-stratification factors for the combined 1994 samples 1994 post AGE HIGHEST 1994 Unwtd Wghtd Post-strat (Cur- EDUCATION CPS n % % factor rent) (Sel,NR) (94 cps) 18-21 No College 4.2 50 2.8 3.5 1.206 College 3.1 22 1.2 1.7 1.838 TOTAL 7.3 72 4.0 5.2 22-29 No HS Diploma 2.3 23 1.3 1.2 1.924 HS Diploma 5.5 72 4.0 4.4 1.252 Some College 5.3 104 5.8 5.9 0.898 College Graduate 3.4 51 2.8 2.8 1.230 TOTAL 16.5 250 13.9 14.2 30-39 No HS Diploma 2.9 37 2.1 2.0 1.503 HS Diploma 8.1 147 8.2 8.2 0.979 Some College 6.6 150 8.4 8.1 0.822 College Graduate 5.7 133 7.4 7.4 0.776 TOTAL 23.3 467 26.0 25.6 40-49 No HS Diploma 2.3 25 1.4 1.5 1.575 HS Diploma 6.1 92 5.1 5.9 1.041 Some College 5.2 78 4.4 4.4 1.189 College Graduate 5.4 130 7.2 7.2 0.750 TOTAL 19.0 325 18.1 18.9 50-59 No HS Diploma 2.4 29 1.6 1.7 1.407 HS Diploma 4.6 78 4.4 4.7 0.983 Some College 2.8 47 2.6 2.6 1.069 College Graduate 2.8 61 3.4 3.7 0.736 TOTAL 12.5 215 12.0 12.7 60-69 No HS Diploma 3.0 63 3.5 3.3 0.895 HS Diploma 3.8 90 5.0 4.7 0.805 Some College 1.9 31 1.7 1.6 1.175 College Graduate 1.7 34 1.9 1.9 0.869 TOTAL 10.3 218 12.1 11.6 70+ No HS Diploma 4.6 86 4.8 3.8 1.188 HS Diploma 3.7 72 4.0 3.5 1.046 Some College 1.7 45 2.5 2.2 0.744 College Graduate 1.3 45 2.5 2.3 0.559 TOTAL 11.2 248 13.8 11.9 1795 Summary by Education level: 1994 post 94 CPS n Unwtd% Wtd% No HS Diploma 19.5 279 15.5 14.7 HS Diploma 33.9 585 32.6 33.7 Some College 26.4 477 26.6 26.4 College Graduate 20.3 454 25.3 25.3 Total 1795 Table 6c: Distributions and post-stratification factors for the combined 1996 samples 1996 pre 1996 post AGE HIGHEST 1996 Unwtd Wghtd Post-strat Unwtd Wghtd Post-strat (Cur- EDUCATION CPS n % % factor n % % factor rent) 18-21 No College 4.4 24 1.4 1.8 2.383 17 1.1 1.5 3.007 College 2.9 30 1.8 2.6 1.140 27 1.8 2.6 1.118 TOTAL 7.3 54 3.2 4.4 44 2.9 4.1 22-29 No HS Diploma 2.0 14 0.8 0.9 2.349 8 0.5 0.6 3.673 HS Diploma 4.9 61 3.6 3.9 1.245 45 2.9 3.1 1.554 Some College 5.0 65 3.8 3.6 1.388 55 3.6 3.5 1.424 College Graduate 3.7 61 3.6 3.6 1.025 56 3.6 3.8 0.981 TOTAL 15.6 201 11.7 12.0 164 10.7 11.0 30-39 No HS Diploma 2.9 27 1.6 1.5 2.000 24 1.6 1.5 2.028 HS Diploma 7.6 133 7.8 7.5 1.013 117 7.6 7.3 1.041 Some College 6.3 138 8.1 7.9 0.805 123 8.0 7.9 0.804 College Graduate 5.9 123 7.2 7.2 0.811 111 7.2 7.4 0.799 TOTAL 22.8 421 24.6 24.1 375 24.4 24.0 40-49 No HS Diploma 2.4 22 1.3 1.3 1.865 18 1.2 1.1 2.080 HS Diploma 6.6 106 6.2 6.7 0.979 94 6.1 6.6 0.992 Some College 5.5 88 5.1 5.6 0.979 79 5.1 5.6 0.982 College Graduate 5.7 138 8.0 7.8 0.726 125 8.2 8.0 0.706 TOTAL 20.1 354 20.7 21.4 316 20.6 21.4 50-59 No HS Diploma 2.3 29 1.7 1.8 1.331 28 1.8 1.9 1.233 HS Diploma 4.6 82 4.8 5.2 0.880 68 4.4 4.8 0.958 Some College 2.9 55 3.2 3.1 0.914 53 3.5 3.4 0.847 College Graduate 3.0 68 4.0 4.4 0.672 67 4.4 4.9 0.606 TOTAL 12.8 234 13.7 14.5 216 14.1 15.0 60-69 No HS Diploma 2.8 47 2.7 2.5 1.096 44 2.9 2.7 1.030 HS Diploma 3.7 73 4.3 3.9 0.956 68 4.4 4.0 0.923 Some College 1.9 40 2.3 2.4 0.778 38 2.5 2.5 0.744 College Graduate 1.8 39 2.3 2.3 0.771 37 2.4 2.5 0.715 TOTAL 10.1 199 11.6 11.1 187 12.2 11.7 70+ No HS Diploma 4.3 81 4.7 3.9 1.098 75 4.9 4.0 1.063 HS Diploma 3.7 80 4.7 4.1 0.912 73 4.8 4.2 0.890 Some College 1.9 49 2.9 2.4 0.789 45 2.9 2.5 0.757 College Graduate 1.5 41 2.4 2.2 0.694 39 2.5 2.3 0.664 TOTAL 11.3 251 14.6 12.5 232 15.1 12.9 1714 1534 Summary by Education level: 1996 pre 1996 post 96 CPS n Unwtd% Wtd% n Unwtd% Wtd% No HS Diploma 18.9 230 13.4 12.5 203 13.2 12.2 HS Diploma 33.2 549 32.0 32.3 476 31.0 31.0 Some College 26.4 465 27.1 27.6 420 27.4 28.0 College Graduate 21.5 470 27.4 27.5 435 28.4 28.8 Total 1714 1534 10. "Trimming of weights The new weights for each sample -- 1992 pre and post, 1994 post and 1996 pre and post - were calculated as the product of the corresponding base weight and the post-stratification factor. The resulting products were then "trimmed" at the 1st and 99th percentiles in order to control the potential for high variation caused by these weights. The results of trimming at the 1st and 99th percentiles are shown in Table 7. The column labels "Before" and "After" indicate whether the statistics refer to the weight before or after trimming. Table 7: Comparison of final weight statistics before and after trimming 1992 pre 1992 post 1994 post Before After Before After Before After 1126 1126 1005 1005 1795 1795 mean 2.4136 2.4038 2.4092 2.4015 2.4201 2.4129 std dev 1.1252 1.0841 1.1075 1.0773 1.1817 1.1494 max 9.6008 5.5521 8.5612 5.2942 8.8935 6.5143 99th 5.5521 5.5521 5.2942 5.2942 6.6514 6.5143 1st 0.7796 0.7796 0.7471 0.7471 0.7999 0.7999 min 0.6480 0.7796 0.6644 0.7471 0.6370 0.7999 1996 pre 1996 post Before After Before After n 1714 1714 1 534 1534 mean 2.5241 2.5018 2.5112 2.4727 std dev 1.3853 1.2720 1.5714 1.3387 max 13.277 7.5774 16.753 8.4760 99th 7.5774 7.5774 8.4760 8.4760 1st 0.8930 0.8930 0.8496 0.8496 min 0.7104 0.8930 0.6406 0.8496 11. Results: The steps taken to address the 1996 NES overestimation of voting in the 1996 presidential election resulted in the development of post- stratified weights which account for individual selection probability, geographic related household nonresponse, and misrepresentation of any age by education subgroups. These revised, CPS-standardized weights were computed for the 1992 NES Pre and Post, 1994 NES Post and 1996 NES Pre and Post Election data sets. Users of previous weights released with the 1992, 1994 and 1996 data will find that these weights extend and combine the features of previously released weights. Table 8 compares the weighted ( final weights ) distributions by age and education to the CPS estimates. It is evident that the use of the final weights results in a distribution which is more similar to CPS population estimates. Table 8: Comparison of weighted (final weights) NES distribution to CPS population estimates for age by education subgroups. AGE HIGHEST '92 '92pre '92post '94 '94post '96 '96pre '96post (Cur- EDUCATION CPS NES NES CPS NES CPS NES NES rent) 18-21 No College 4.3 4.27 4.27 4.2 4.22 4.4 3.63 3.38 College 3.1 3.06 3.08 3.1 2.85 2.9 2.97 2.99 TOTAL 7.3 7.33 7.33 7.3 7.07 7.3 6.61 6.36 22-29 No HS Diploma 2.4 2.15 2.19 2.3 2.25 2.0 1.90 1.55 HS Diploma 6.1 6.10 6.09 5.5 5.47 4.9 4.93 4.95 Some College 4.8 4.86 4.85 5.3 5.30 5.0 5.09 5.11 College Graduate 3.5 3.48 3.48 3.4 3.43 3.7 3.72 3.73 TOTAL 16.7 16.60 16.61 16.5 16.45 15.6 15.63 15.35 30-39 No HS Diploma 3.0 2.99 2.99 2.9 2.94 2.9 2.96 2.99 HS Diploma 8.7 8.69 8.68 8.1 8.09 7.6 7.68 7.73 Some College 6.1 6.13 6.13 6.6 6.63 6.3 6.38 6.42 College Graduate 5.7 5.68 5.69 5.7 5.72 5.9 5.92 5.96 TOTAL 23.4 23.49 23.48 23.3 23.38 22.8 22.94 23.11 40-49 No HS Diploma 2.4 2.19 2.23 2.3 2.27 2.4 2.37 2.39 HS Diploma 6.1 6.11 6.11 6.1 6.13 6.6 6.61 6.65 Some College 4.7 4.75 4.74 5.2 5.18 5.5 5.56 5.59 College Graduate 5.0 4.97 4.97 5.4 5.45 5.7 5.73 5.76 TOTAL 18.1 18.02 18.05 19.0 19.03 20.1 20.27 20.39 50-59 No HS Diploma 2.8 2.76 2.75 2.4 2.36 2.3 2.36 2.37 HS Diploma 4.7 4.68 4.68 4.6 4.61 4.6 4.64 4.67 Some College 2.4 2.36 2.36 2.8 2.78 2.9 2.89 2.92 College Graduate 2.5 2.51 2.51 2.8 2.77 3.0 3.01 3.03 TOTAL 12.3 12.31 12.30 12.5 12.51 12.8 12.90 12.99 60-69 No HS Diploma 3.5 3.52 3.50 3.0 2.99 2.8 2.78 2.79 HS Diploma 4.2 4.24 4.24 3.8 3.81 3.7 3.72 3.75 Some College 1.8 1.76 1.75 1.9 1.89 1.9 1.91 1.92 College Graduate 1.7 1.67 1.67 1.7 1.66 1.8 1.80 1.81 TOTAL 11.1 11.19 11.17 10.3 10.35 10.1 10.21 10.27 70+ No HS Diploma 4.8 4.84 4.83 4.6 4.57 4.3 4.28 4.32 HS Diploma 3.6 3.52 3.53 3.7 3.68 3.7 3.75 3.78 Some College 1.5 1.48 1.48 1.7 1.67 1.9 1.88 1.90 College Graduate 1.2 1.22 1.22 1.3 1.30 1.5 1.52 1.53 TOTAL 11.1 11.06 11.06 11.2 11.22 11.3 11.44 11.53 Summary by Education level: '92pre '92post '94post '96pre '96post '92CPS NES NES '94CPS NES '96CPS NES NES No HS Diploma 20.8 19.19 19.32 19.5 18.83 18.9 18.25 17.63 HS Diploma 35.5 36.88 36.77 33.9 34.53 33.2 33.37 33.69 Some College 24.3 24.26 24.24 26.4 26.31 26.4 26.69 26.85 College Graduate 19.5 19.68 19.68 20.3 20.33 21.5 21.70 21.84 The final check on the revised weight is to use this trimmed final weight to estimate presidential election voting rates in 1992 and 1996. Table 9 shows that in both 1992 and 1996 the use of the final weight results in significantly lower estimates of voting. Table 9: Calculated Voting Rates in the 1992 and 1996 Presidential elections 1992 1996 unwghtd base weight final weight unwghtd base weight final weight 0.77 0.78 0.75 0.77 0.77 0.72 >> SAMPLE DESIGN 1992 ELECTION STUDY STUDY POPULATION The study population for the 1992 National Pre/Post Election Study (NES) is defined to include all United States citizens of voting age on or before the 1992 Election Day. Eligible citizens must have resided in housing units, other than on military reservations, in the forty-eight coterminous states. This definition excludes persons living in Alaska or Hawaii and requires eligible persons to have been both a United States citizen and eighteen years of age on or before the 3rd of November 1992. MULTI-STAGE AREA PROBABILITY SAMPLE DESIGN The 1992 NES is based on a multi-stage area probability sample selected from the Survey Research Center's (SRC) National Sample design. Identification of the 1992 NES sample respondents was conducted using a four stage sampling process--a primary stage sampling of U.S. Metropolitan Statistical Areas (MSAs) and counties, followed by a second stage sampling of area segments, a third stage sampling of housing units within sampled area segments and concluding with the random selection of a single respondent from selected housing units. A detailed documentation of the SRC National Sample is provided in the SRC publication titled, 1980 SRC National Sample: Design and Development. Primary Stage Selection The selection of primary stage sampling units (PSUs), which depending on the sample stratum are either MSAs, single counties or groupings of small counties, is based on the county-level 1980 Census Reports of Population and Housing. Primary stage units were assigned to 84 explicit strata based on MSA/Non-MSA status, PSU size, and geographic location. Sixteen of the 84 strata contain only a single self-representing PSU, each of which is included with certainty in the primary stage of sample selection. The remaining 68 nonself-representing strata contain more than one PSU. From each of these nonself-representing strata, one PSU was sampled with probability proportionate to its size (PPS) measured in 1980 occupied housing units. The full SRC National Sample of 84 primary stage selections was designed to be optimal for surveys roughly two to three times the size of the 1992 NES. To permit the flexibility needed for optimal design of smaller survey samples, the primary stage of the SRC National Sample can be readily partitioned into smaller subsamples of PSUs. Each of the partitions represents a stratified subselection from the full 84 PSU design. Since the 1992 NES desired comparison of data over time from 1990 NES respondents, as well as an expanded representative sample of eligible 1992 respondents, a combined panel/cross-section sample was designed for the 1992 Pre/Post-Election Study. The Panel portion of the 1992 sample was selected from the original 1990 NES sample which, at the Primary stage had been selected from the "one-half" partition of the 1980 SRC National Sample. The"A" one-half sample of the 1980 National Sample design includes 11 of the 16 self-representing MSA PSUs and a stratified subsampling of 34 (of the 68) nonself-representing PSUs of the SRC National Sample. The Panel portion of the 1992 NES is designed to allow longitudinal analysis of individual change since the panel cases follow the original proportionate distribution to the 1990 "A" one-half sample areas. The 1992 NES Cross-Section encompasses both the panel cases and a new selection of cases from the two-thirds partition of the 1980 National Sample (that is the "A" plus the "B1" PSUs). The two-thirds 1980 National Sample design includes all 16 self-representing PSUs and 11 additional nonself-representing PSUs for a total of 45 (of 68) nonself-representing PSUs. The additional cases were added to the 1992 NES to supplement the Panel selections such that when the Panel and new Cross-section selections are combined for analysis a representative cross-section of the study population has been maintained. Table 9 identifies the PSUs for the 1992 National Election Study by MSA status and Region. The PSUs in the Panel portion of the sample design are shown in standard print on this table while those PSUs added for the two-thirds Cross-section are shown in italics. Table 9: PSUs in the 1992 NES Pre- and Post-Election Survey By: MSA Status and Region. REGION Self-representing MSAs Northeast New York, NY-NJ Philadelphia, PA-NJ Boston, MA* Nassau-Suffolk, NY Pittsburgh, PA* North Chicago, IL Central Detroit, MI St. Louis, MO* Minneapolis, MN-WI South Washington, DC-MD-VA Dallas-Ft Worth, TX Houston, TX* Baltimore, MD* Atlanta, GA West Los Angeles, CA San Francisco, CA REGION Nonself-representing MSAs Northeast Buffalo, NY Newark, NJ Haven, CT Atlantic City, NJ Manchester, NH North Milwaukee, WI Central Dayton, OH Kansas City, MO-KS Des Moines, IA Grand Rapids, MI Fort Wayne, IN Steubenville, OH Saginaw, MI South Birmingham, AL Columbus, GA-AL Miami, FL Jacksonville, FL Lakeland, FL McAllen, TX Waco, TX Wheeling, WV Knoxville, TN Richmond, VA West Seattle, WA Denver, CO Anaheim, CA Riverside, CA Fresno, CA Eugene, OR Phoenix, AZ REGION Non-MSAs Northeast Schuyler, NY Gardner, MA North Sanilac, MI Central Decatur, IN Phillips, KS/Saline, NE Mower, MN South Bulloch, GA Sabine, LA Hale, TX Monroe, AR/Ashley, AR Bedford, TN Montgomery, VA Robeson, NC West ElDorado-Alpine, CA Carbon, WY NOTE: The PSU's marked with an asterisk are Self-Representing for sample designs which use the two-thirds or larger portion of the sample (i.e., in this case, the combined cross-section and panel design). For the half-sample design (i.e., in this case, the panel portion alone) only 6 of the 16 Self-Representing areas remain Self-Representing. The other ten Self-Representing PSU's are paired and only five are used in the half-sample design, each representing both itself and the PSU it is paired with. Second Stage Selection of Area Segments The second stage of the 1980 National Sample was selected directly from computerized files that were prepared from the 1980[8] Census summary tape file series (STF1-B). The designated second-stage sampling units (SSUs), termed "area segments", are comprised of census blocks in the metropolitan primary areas and enumeration districts (EDs) in the rural areas of both non-MSA and MSA primary areas. Each SSU block, block combination or enumeration district was assigned a measure of size equal to the total 1980 occupied housing unit count for the area (minimum = 50). Second stage sampling of area segments was performed with probabilities proportionate to the assigned measures of size. A three-step process of ordering the SSUs within the primary areas produced an implicit stratification of the area segments in the second stage sampling frame, stratified at the county level by geographic location and population. Area segments were stratified within county at the Minor Civil Division (MCD) level by size and income, and at the block and ED level by location within the MCD or county. (For details, refer to the SRC publication, 1980 NATIONAL SAMPLE: DESIGN AND DEVELOPMENT.) Systematic PPS sampling was used to select the area segments from the second stage sampling frame for each county. In the self-representing (SR) PSUs the number of sample area segments varied in proportion to the size of the primary stage unit, from a high of 12 Cross-section and 12 Panel area segments in the SR New York MSA, 6 Cross-section segments and 5 Panel segments in the San Francisco MSA, to a low of 4 Cross-section and no Panel area segments in the smaller SR PSUs such as Minneapolis and Atlanta MSAs. Most Nonself-representing (NSR) half-sample (A) PSUs were represented by 2 Cross-section and 6 Panel area segments; most of the eleven other (B1) NSR PSUs had 6 Cross-section area segments (and, of course, no Panel segments). A total of 487 area segments were selected, 206 Cross-section and 281 Panel segments, 151 in the sixteen self-representing PSUs and 336 in the nonself-representing PSUs as shown in Table 10. Table 10: Number of Cross-Section and Panel Area Segments in the 1992 NES Sample Showing PSU Name, National-Sample Stratum and Partition, and MSA Status 1980 1980 National Sample # of 1992 NES # of 1992 NES N. Samp PSU Name Cross-section Panel Sample PSU# Sample Segs. Segments Six Largest Self-representing PSUs 1 A New York, NY-NJ 12 12 2 A Los Angeles, CA 12 9 3 A Chicago, IL 8 8 4 A Philadelphia, PA-NJ 6 6 5 A Detroit, MI 6 6 6 A San Francisco, CA 6 5 Ten Remaining Self-representing PSUs 7 B1 Washington, DC-MD-VA 6 0 8 B1 Dallas-Ft Worth, TX 6 0 9 A Houston, TX 0 7 10 A Boston, MA 0 6 11 B1 Nassau-Suffolk, NY 4 0 12 A St Louis, MO-IL 0 6 13 A Pittsburgh, PA 0 6 14 A Baltimore, MD 0 6 15 B1 Minneapolis, MN-WI 4 0 16 B1 Atlanta, GA 4 0 Nonself-representing MSAs: Northeast 17 A Buffalo, NY 2 6 18 B1 Newark, NJ 6 0 21 A New Haven, CT 2 6 23 A Atlantic City, NJ 2 6 24 A Manchester, NH 2 6 Nonself-representing MSAs: North Central 26 A Milwaukee, WI 2 6 27 A Dayton, OH 2 6 28 B1 Kansas City, MO-KS 6 0 29 A Des Moines, IA 2 6 31 A Grand Rapids, MI 2 6 32 A Fort Wayne, IN 2 6 33 A Steubenville, OH-WV 2 6 34 B1 Saginaw, MI 6 0 Nonself-representing MSAs: South 36 A Birmingham, AL 2 6 39 A Columbus, GA-AL 2 6 40 A Miami, FL 2 6 42 B1 Jacksonville, FL 6 0 43 A Lakeland, FL 2 6 44 A McAllen, TX 2 6 45 B1 Waco, TX 6 0 47 A Wheeling, WV-OH 2 6 49 A Knoxville, TN 2 6 50 A Richmond, VA 2 6 Nonself-representing MSAs: West 53 A Seattle, WA 2 6 55 A Denver, CO 2 6 56 A Anaheim, CA 2 6 57 B1 Riverside-San Bernardino, CA 6 0 58 A Fresno, CA 2 6 59 A Eugene, OR 2 6 60 B1 Phoenix, AZ 6 0 Nonself-representing Non-MSAs: Northeast 63 A Schuyler, NY 2 6 64 B1 Gardner, MA 6 0 Nonself-representing Non-MSAs: North Central 65 A Sanilac, MI 2 6 66 B1 Decatur, IN 6 0 68 A Phillips, KS/ ** 6 Saline, NE 2 ** 70 A Mower, MN 2 6 Nonself-representing Non-MSAs: South 73 A Bulloch, GA 2 6 74 B1 Sabine, LA 5 0 76 A Hale, TX 2 6 77 A Monroe, AR/ ** 6 Ashley, AR 2 ** 78 A Bedford, TN 2 6 80 B1 Montgomery, VA 5 0 81 A Robeson, NC 2 6 Nonself-representing Non-MSAs: West 82 A ElDorado-Alpine, CA 2 6 84 A Carbon, WY 2 6 Total 206 281 ** In two Non-SMSA National Sample strata (68 and 77) the 1980 materials from which the Panel area segments had been selected was exhausted (i.e., there were insufficient remaining SSUs from which to select new Cross-section area segments), so a new Primary selection had to be made from those two strata. Therefore, the Panel area segments for stratum 68 are from PSU Phillips County, KS, and the Cross-section area segments are from Saline County, NE; the Panel area segments for stratum 77 are from PSU Monroe County, AR, and the Cross-section area segments are from Ashley County, AR. Although 281 segments were used in the 1990 NES, only 272 Panel segments appear in the 1992 NES Panel. The difference is due to some segments used in 1990 not having any interviews completed in 1990 and, therefore, not becoming part of the 1992 Panel. Third Stage Selection of Housing Units For each area segment selected in the second sampling stage, a listing was made of all housing units located within the physical boundaries of the segment. For segments with a very large number of expected housing units, all housing units in a subselected part of the segment were listed. The final equal probability sample of housing units for the 1992 NES was systematically selected from the housing unit listings for the sampled area segments. The overall probability of selection for 1992 NES Cross-Section households was f=.00003988 or .3988 in 10,000. The equal probability sample of households was achieved for the combined Cross-Section/Panel design by using the standard multi-stage sampling technique of setting the sampling rate for selecting housing units within area segments to be inversely proportional to the PPS probabilities (see above) used to select the PSU and area segment. Five 1992 Panel replicates were designated for the entire "frame" of households in which a complete interview was obtained in the 1990 NES study (2000 - 11 partial interviews = 1989 1990 interview HUs). The original 1990 sample lines had been selected from the National Sample ("A" or "half-sample" PSUs) to be inversely proportional to the PPS probabilities used to select the area segments as described in the previous paragraph. The new Cross-Section component of the 1992 NES sample design was disproportionately allocated to the "B1" PSUs to supplement the Panel cases such that when cross-sectional analysis was undertaken, combining new cross-section cases with panel cases would yield an equal probability sample of households. The distribution of the combined sample would be that required by the two-thirds design. Fourth Stage Respondent Selection Within each sampled new cross-section housing unit, the SRC interviewer prepared a complete listing of all eligible household members. Using an objective procedure described by Kish (1949)[9] a single respondent was then selected at random to be interviewed. Regardless of circumstances, no substitutions were permitted for the designated respondent. This technique had also been used in 1990 to select the original Panel respondents. In 1992 the same Panel respondent (R) was sought for interview as had been interviewed in 1990. SAMPLE DESIGN SPECIFICATIONS The targeted completed interview sample size for the 1992 NES Pre/Post-Election Survey was n = 2,057 total cases. In the original sample size computation, the following assumptions were made for the cross-section component of the sample: response rate for the pre-election interview = .72 and of these 95% were assumed to be available and cooperative for the post-election interview, combined occupancy/eligibility rate = .83. These assumptions were derived from survey experience in the 1986 NES Post Election Survey[10]. The assumptions made for the panel component were: .913 recontact rate and .75 response rate for the pre-election interview. The same .95 response rate for the post-election interview was assumed for both the panel and the cross-section component. To most closely tailor the field effort to the sample field experience during this study, both parts of the selected sample had five replicates designated. Replicates 1 and 2 were considered the "base sample", certain to be released. 55% of this base was designated as Replicate 1 to be released September 1, 1992 and 45% designated as Replicate 2 to be released October 1, 1992. The other three replicates were designated "Reserve" replicates, one or more to be released for field work October 1, 1992 at the discretion of NES study staff. Replicate 3 (Reserve replicate 1) was never, in fact, released. Replicates 4 and 5 (Reserve replicates 2 and 3) were released with Base sample replicate 2 on October 1, 1992. Each replicate is a proper subsample of the NES sample. A subsampling of one-third of selected addresses was made in certain cases when selected lines were determined to be within locked buildings, in gated subdivisions or in areas which posed a danger to interviewing staff. This allowed concentration of greater field effort in these circumstances to obtain at least some interviews. In cases where this was done, appropriate weighting of the results will be used to compensate. (This is not reflected in the following tables however). Table 11 provides a full description of the original sample design specifications applied to the Base Sample and also indicates the number of HU listings assigned to each replicate. As stated above, Replicates 1 and 2 constitute the Base Sample; Replicates 3, 4 and 5 are reserve replicates. Replicate 3 was, in fact, never released for field work. Table 11: Original Sample Design Specifications and Assumptions 1992 National Pre/Post-Election Survey Cross-Section Component (Supplemental) Original Specifications and Assumptions Completed Post/ interview 1,000 Contact/Response Rate .95 Completed Pre/ interview 1,052 Response Rate .72 Eligible sample households 1,462 Occupancy/Eligibility Rate[11] .83 Panel Recontact Rate Sample HU listings Replicates 1 and 2 1,760 Replicate 1 (incl above)[12] 961 Replicate 2 (incl above)[13] 799 Replicate 3 (Reserve)[14] 200 Replicate 4 (Reserve) 75 Replicate 5 (Reserve) 51 Total Sample lines 2,086 Panel Component Total Original Specifications and Assumptions Completed Post/ interview 1,057 2,057 Contact/Response Rate .95 Completed Pre/ interview 1,112 2,164 Response Rate[15] .75 Eligible sample households 1,483 2,945 Occupancy/Eligibility Rate[11] Panel Recontact Rate .913 Sample HU listings Replicates 1 and 2 1,625 3,385 Replicate 1 (incl above)[12] 900 Replicate 2 (incl above)[13] 725 Replicate 3 (Reserve)[14] 208 Replicate 4 (Reserve) 104 Replicate 5 (Reserve) 52 Total Sample lines 1,989[16] SAMPLE DESIGN OUTCOMES Table 12 compares the original sample design specifications and assumptions for the new Cross-Section Component of the 1992 NES as applied to the Base Sample (as in Table 11) and as applied to the actually released sample (Replicates 1, 2, 4 and 5) to the actual outcome for that component. Table 13 makes a similar comparison for the Panel Component of the 1992 NES Sample and Table 14 presents a summary of the figures for the combined Cross-Section/Panel Sample. The response rates which appear in these tables are calculated using both complete and partial (short-form) interviews. An alternative response rate which excludes short-form interviews is described in "Response Rates", above. Table 12: Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Cross-Section Component of the 1992 National Pre/Post-Election Survey Original Original S & A Specifications Applied to & Assumptions Actual Release (Reps. 1 & 2) (Reps. 1,2,4 & 5) Completed Post/Interviews 1,000 1,103 Contact/Response Rate .95 .95 Released for Recontact 1,052 1,161 Completed Pre/ Interviews 1,052 1,161 Response Rate .72 .72 Eligible Sample Households 1,462 1,613 Occupancy/Eligibility Rate[17] .83 .83 Subsampling for dangerous/ locked areas -- -- Sample HU listings 1,760 1,943 Sample growth from update[18] -- 1.03 Selected Sample lines 1,760 1,886 Actual Outcome Completed Post/Interviews 1,005 Contact/Response Rate .89 Released for Recontact 1,126 Completed Pre/ Interviews 1,126 Response Rate .74 Eligible Sample Households 1,522 Occupancy/Eligibility Rate .80 1,900 Subsampling for dangerous/ locked areas .99[19] Sample HU listings 1,923 Sample growth from update 1.02 Selected Sample lines 1,886 Table 13: Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Panel Component of the 1992 National Pre/Post-Election Survey Original Original S & A Specifications Applied to & Assumptions Actual Release (Reps 1 & 2) (Reps 1,2,4 & 5) Completed Post/ Interviews 1,057 1,158 Contact/Response Rate .95 .95 Released for Recontact 1,112 1,219 Completed Pre/ Interviews 1,112 1,219 Response Rate .75[20] .75 Eligible Sample Households 1,483 1,626 Panel Recontact Rate .913 .913 Sample HU listings Released 1,625 1,781 Total Panel cases 1,989 1,989 Actual Outcome Completed Post/ Interviews 1,250 Contact/Response Rate .92 Released for Recontact 1,361 Completed Pre/ Interviews 1,361 Response Rate .78 Eligible Sample Households 1,752 Panel Recontact Rate .979 Sample HU listings Released 1,789 Total Panel cases 1,989 Table 14: Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Combined Cross-Section/Panel Sample. 1992 National Pre/Post-Election Survey Original Original S & A Specifications Applied to & Assumptions Actual Release (Reps. 1 & 2) (Reps. 1,2,4 & 5) Completed Post/ Interviews 2,057 2,261 Released for Recontact 2,164 2,380 Completed Pre/ Interviews 2,164 2,380 Eligible Sample Households 2,945 3,239 Total Sample HU listings 3,385[21] 3,724 Growth from update of Cross-Section component 1.015 Selected Sample lines 3,667 Actual Outcome Completed Post/ Interviews 2,255 Released for Recontact 2,487 Completed Pre/ Interviews 2,487 Eligible Sample Households 3,274 Total Sample HU listings 3,712 In comparing the second column of Table 12 with the third column, it can be seen that, for the 1992 Cross-Section component, the sample growth from the update procedure was slightly less than expected; this was perhaps due to the fact that many of the new cross-section segments had been listed within the year previous to field dates for the 1992 NES study. The original sample design specifications also overestimated the actual occupancy/eligibility rates resulting in 91 fewer eligible HUs than estimated. However, since the actual response rate was higher than estimated, completed pre-election interviews fell only 35 short of the number estimated. The assumptions for response rate and occupancy/eligibility rate were based on the 1986 NES field experience for a probability sample based on the entire two-thirds design of the National Sample. The actual response rate for the 1992 cross-section component (.74), as well as the occupancy/eligibility rate very likely reflects the disproportionate allocation of the new cross-section segments in the B1 areas of the National Sample which may well have different occupancy/eligibility and response rates than any overall past NES rates on which the original assumptions were based. The number of Post-election interviews obtained, 1,005, was closer to the target of 1000 interviews projected for the Base Sample alone than the 1,103 projected for the actual 1,886 sample lines released. For the Panel Component (see Table 13), both the Panel recontact rate and the response rate exceeded assumptions resulting in 142 more pre-election interviews than expected. A lower than assumed response rate for the post-election interview reduced the excess to 92 more post-election interviews than projected for the release of the Panel base sample plus replicates 4 and 5 (reserve replicates 2 and 3). The figures for the combined cross-section sample shown in Table 14 show completed pre-election interviews of 107 over expected. Due to lower than assumed response rate for the post-election interview, combined with lower cross-section and higher panel overall response and occupancy/eligibility rates, the final total number of post election interviews was 6 fewer than the projected outcome for the sample lines released. WEIGHTED ANALYSIS OF 1992 NES DATA The area probability sample design for the 1992 NES results in an equal probability sample of U.S. households. However, within sample households a single adult respondent is chosen at random to be interviewed. Since the number of eligible adults may vary from one household to another, the random selection of a single adult introduces inequality into respondents' selection probabilities. In analysis, a respondent selection weight should be used to compensate for these unequal selection probabilities. The value of the respondent selection weight is exactly equal to the number of eligible adults in the household from which the random respondent was selected. The use of the respondent selection weight is strongly encouraged, despite past evaluations which have shown these weights to have little significant impact on the values of NES estimates of descriptive statistics. The Sampling Section has provided two final person level analysis weights which will incorporate sampling, nonresponse and post-stratification factors. One weight variable (#3009) is for use with Panel cases only; the other weight variable (#3008) is for the 1992 NES Cross-section (which includes both panel and new cross-section cases.) Analysts interested in developing their own nonresponse or post-stratification adjustment factors must request access to the necessary sample control data from the NES Board. CONSTRUCTION OF ANALYSIS WEIGHTS Nonresponse adjustment factors were constructed at the household level separately for Panel and new Cross-Section component cases. Nonresponse adjustment cells were formed by crossing PSU type (Self-representing, Nonself- representing MSA or non-MSA) by the nine Census divisions (New England, Middle Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain, and Pacific). A nonresponse factor equal to the inverse of the response rate in each cell was applied to the interview cases. In order to have a minimum of approximately 25 cases in each nonresponse adjustment cell, some cells were collapsed across Census divisions in the same Census region. An intermediate weight was constructed by multiplying the probability of selection of the household by the nonresponse adjustment factor by the number of eligible persons in the household[22]. This intermediate weight was used to produce a weighted sex by age category by Census Region table. The age categories used were: 18-44, 45-64, and 65+. Post-stratification factors were constructed to match the sample proportions in the 24 sex by age by Region cells to the July 1991 Census population totals (United States Department of Commerce News Public Information Office Press Release - CB92-93). The two final analysis weights were each centered to a mean of 1.0 so that the sum of the weights equals the number of respondents (1,359 for the 1990-92 Panel and 2,485 for the 1992 Cross-section). COMPARING THE 1992 NES TO PREVIOUS NATIONAL ELECTION STUDIES Earlier National Election Studies data collections did not include weights to adjust for nonresponse and the unequal probability of selection at the household level. Thus, weighting the 1992 NES data by V3009 (for analysis of the Panel cases) or by V3008 (for combined analysis of the panel and new cross-section cases) produces estimates that are not strictly comparable to those obtained from previous National Election Studies that were not weighted to incorporate sampling, nonresponses and post-stratification factors. Analysis comparing data from the 1992 NES data to previous NES data collections should employ V7000. Because approximately half of the respondents to the 1992 NES were part of a panel first interviewed in 1990, to be comparable with previous NES cross-section data collections, the combined 1992 panel and new cross-section data must be weighted to correct for panel attrition and the aging of the panel respondents. Panel attrition is not uniform across demographic groups. Some respondents (the mobile and those with the least amount of formal education) are more susceptible to panel attrition. By definition, panel respondents are two years older than the cross-section respondents. And by definition, there are almost no 18 or 19 year-olds among the panel respondents interviewed in 1992 (because an 18 year-old in 1992 would have been 16 years-old in 1990 and ineligible for the 1990 study). Weighting of the panel respondents is necessary to ensure comparability with past NES data collections. V7000 corrects the combined panel and cross-section cases for the panel attrition and aging that occurred among the panel respondents. This weight should be used when comparing estimates made on the 1992 NES data to estimates made on previous (unweighted) NES data collections. V7000 does not appear in the April 1993 CPS Early Release Version of the 1992 National Election Study. CONSTRUCTION OF V7000 To construct this weight, panel respondents were classified by age (17-24, 25-39, 40- 64, 65-74, 75 and over), education (less than high school, high school diploma, and more than high school education), and mobility (whether or not the respondent had moved between 1990 and 1992). Cross-classification of these three variables produced a 30-celled table (5 x 3 x 2) for each of the following: (1) 1990 panel respondents who comprised the panel portion of the sample "universe" for the 1992 study (N=1769); and (2) panel respondents interviewed in 1992 (N=1359). The weight was constructed by dividing the value of each cell in the 1990 table (1) by the value of the corresponding cell in the 1992 table (2). (For example, 10.9 percent of the 1,769 1990 panel respondents were age 40-64/had more than high school education/ had not moved. In 1992, respondents in the cell defined by these same categories comprised 11.8 percent of the 1359 panel respondents interviewed. The case weight for this group of respondents is 10.9/11.8 = .9237.) In order to have a minimum of approximately 25 cases in each cell, some cells were collapsed. This procedure centers the weight variable V7000 so that it has a mean of 1.0 and the sum of the weights (2488) is approximately equal to the actual number of combined panel and cross-section respondents (2,485). Respondents who are part of the new cross-section have the value "1.0000" on V7000. SAMPLING ERRORS OF 1992 NES ESTIMATES SAMPLING ERROR CALCULATION PROGRAMS The probability sample design for the 1992 National Election Study permits the calculation of estimates of sampling error for survey statistics. For calculating sampling errors of statistics from complex sample surveys, the OSIRIS statistical analysis and data management software system offers the PSALMS and REPERR programs. PSALMS is a general purpose sampling error program which incorporates the Taylor Series approximation approach to the estimation of variances of ratios (including means, scale variables, indices, proportions) and their differences. REPERR is an OSIRIS program which incorporates algorithms for replicated approaches to variance estimation. Both Balanced Repeated Replication (BRR) and Jackknife Repeated Replication (JRR) are available as program options. The current version of REPERR is best suited for estimating sampling errors and design effects for regression and correlation statistics. Sampling Error Codes and Calculation Model Estimation of variances for complex sample survey estimates requires a computation model. Individual data records must be assigned sampling error codes which reflect the complex structure of the sample and are compatible with the computation algorithms of the various programs. The sampling error codes for the 1992 NES are included as variables #3068 and #3069 in the ICPSR Public Use data set. The assigned sampling error codes are designed to facilitate sampling error computation according to a paired selection model for both Taylor Series approximation and Replication method programs. For the Panel Component segments, two sampling error (SE) codes have been included for analysis of 1992 data. For longitudinal analysis of Panel data alone, the original 1990 SE code should be used since this reflects the half-sample design of the 1990 NES sample. For any cross-sectional analysis, where Panel data is combined with new cross-section data, the 1992 SE code must be used. Table 15 provides a description of how individual sampling error code values for Panel only data are to be paired for sampling error computations. Thirty (30) pairs or strata of sampling error computation units (SECUs) are defined. Each SECU in a stratum pair includes cases assigned to a single sampling error code value. The exceptions are the second SECU in stratum 27 which is comprised of cases assigned sampling code values 36 AND 55 and the second SECU in stratum 29 which is comprised of cases with SECUs 61 AND 63. Table 15: 1992 Pre/Post-Election Survey: Panel-Only Analysis Paired Selection Model for Sampling Error Computations (1990 Sampling Error Codes - Variable #3069) Pair (SECU) (SECU) (Stratum) 1 of 2 2 of 2 Codes Codes 1 103 104 2 105 106 3 99 100 4 101 102 5 95 96 6 97 98 7 93 94 8 91 92 9 89 90 10 83 84 11 81 82 12 77 78 13 75 76 14 73 74 15 2 6 16 7 8 17 14 16 18 17 18 19 19 21 20 24 28 21 11 29 22 30 33 23 37 43 24 40 48 25 42 45 26 50 51 27 52 36 + 55 28 57 64 29 60 61 + 63 30 67 68 Table 16 shows the Strata and SECU codes to be used for the paired selection model for sampling error computations for any 1992 cross-sectional analyses using the combined cross-section/panel data. The 42 strata reflect the expanded 2/3rds National Sample design used in 1992. Table 16: 1992 Pre/Post-Election Survey: Cross-Section Analysis[23] Paired Selection Model for Sampling Error Computations (1992 Sampling Error Coded - Variable #3068) Pair (SECU) (SECU) (SE Stratum) 1 of 2 2 of 2 1 1 2 2 1 2 3 1 2 4 1 2 5 1 2 6 1 2 7 1 2 8 1 2 9 1 2 10 1 2 11 1 2 12 1 2 13 1 2 14 1 2 15 1 2 16 1 2 17 1 2 18 1 2 19 1 2 20 1 2 21 1 2 22 1 2 23 1 2 24 1 2 25 1 2 26 1 2 27 1 2 28 1 2 29 1 2 30 1 2 31 1 2 32 1 2 33 1 2 34 1 2 35 1 2 36 1 2 37 1 2 38 1 2 39 1 2 40 1 2 41 1 2 42 1 2 It can be seen from this table that the three-digit 1992 SE code is comprised of: first the two-digit SE Stratum code followed by the one-digit SECU code. Generalized Sampling Error Results for the 1992 NES To assist NES analysts, the OSIRIS PSALMS program was used to compute sampling errors for a wide-ranging example set of means and proportions estimated from the 1988 NES Pre-election Survey data set[24]. For each estimate, sampling errors were computed for the total sample and for fifteen demographic and political affiliation subclasses of the 1988 NES Pre-Election Survey sample. The results of these sampling error computations were then summarized and translated into the general usage sampling error table provided in Table 17. Incorporating the pattern of "design effects" observed in the extensive set of example computations, Table 17 provides approximate standard errors for percentage estimates based on the 1988 NES. To use the table, examine the column heading to find the percentage value which best approximates the value of the estimated percentage that is of interest[25]. Next, locate the approximate sample size base (denominator for the proportion) in the left-hand row margin of the table. To find the approximate standard error of a percentage estimate, simply cross-reference the appropriate column (percentage) and row (sample size base). Note: the tabulated values represent approximately one standard error for the percentage estimate. To construct an approximate confidence interval, the analyst should apply the appropriate critical point from the "z" distribution (e.g. z=1.96 for a two-sided 95% confidence interval half-width). Furthermore, the approximate standard errors in the table apply only to single point estimates of percentages not to the difference between two percentage estimates. The generalized variance results presented in Table 17 are a useful tool for initial, cursory examination of the NES survey results. For more in depth analysis and reporting of critical estimates, analysts are encouraged to compute exact estimates of standard errors using the appropriate choice of a sampling error program and computation model. Table 17: Generalized Variance Table. 1992 NES Pre-Election Survey. APPROXIMATE STANDARD ERRORS FOR PERCENTAGES For percentage estimates near. Sample n 50% 40% or 30% or 20% or 10% or 60% 70% 80% 90% The approximate standard error of the percentage is: 100 5.385 5.277 4.933 4.308 3.231 200 3.912 3.824 3.581 3.128 2.343 300 3.278 3.210 3.006 2.260 1.962 400 2.905 2.846 2.661 2.324 1.743 500 2.663 2.603 2.437 2.128 1.593 750 2.294 2.244 2.094 1.657 1.250 1000 2.078 2.039 1.907 1.657 1.250 1500 1.846 1.803 1.688 1.474 1.102 2000 1.722 1.691 1.568 1.368 1.030 2500 1.637 1.604 1.506 1.310 0.982 >> SAMPLE DESIGN 1994 NATIONAL ELECTION STUDY STUDY POPULATION The study population for the 1994 National Post-Election Study (NES) is defined to include all United States citizens of voting age on or before the 1994 Election Day. Eligible citizens must have resided in housing units, other than on military reservations, in the forty-eight coterminous states. This definition excludes persons living in Alaska or Hawaii and requires eligible persons to have been both a United States citizen and eighteen years of age on or before the 8th of November 1994. MULTI-STAGE AREA PROBABILITY SAMPLE DESIGN The 1994 NES is based on a multi-stage area probability sample selected from the Survey Research Center's (SRC) National Sample design. Identification of the 1994 NES sample respondents was conducted using a four stage sampling process--a primary stage sampling of U.S. Metropolitan Statistical Areas (MSAs) and counties, followed by a second stage sampling of area segments, a third stage sampling of housing units within sampled area segments and concluding with the random selection of a single respondent from selected housing units. A detailed documentation of the SRC National Sample is provided in the SRC publication titled 1980 SRC National Sample: Design and Development. Primary Stage Selection The selection of primary stage sampling units (PSUs), which depending on the sample stratum are either MSAs, single counties or groupings of small counties, is based on the county-level 1980 Census Reports of Population and Housing. Primary stage units were assigned to 84 explicit strata based on MSA/Non-MSA status, PSU size, and geographic location. Sixteen of the 84 strata contain only a single self-representing PSU, each of which is included with certainty in the primary stage of sample selection. The remaining 68 nonself-representing strata contain more than one PSU. From each of these nonself-representing strata, one PSU was sampled with probability proportionate to its size (PPS) measured in 1980 occupied housing units. The full SRC National Sample of 84 primary stage selections was designed to be optimal for surveys roughly two to three times the size of the 1994 NES. To permit the flexibility needed for optimal design of smaller survey samples, the primary stage of the SRC National Sample can be readily partitioned into smaller subsamples of PSUs such as one-half sample or two-thirds sample partition. Each of the partitions represents a stratified subselection from the full 84 PSU design. The one-half partition of the 1980 National Sample (i.e., the "A" primary sampling units or PSUs) includes 11 of the 16 self-representing MSA PSUs and a stratified subsampling of 34 of the 68 nonself-representing PSUs of the SRC National Sample. The two-thirds partition includes all of the "A" PSUs plus "B1" PSUs, i.e., 5 additional self-representing PSUs and 11 additional nonself-representing PSUs. Since the 1994 NES desired comparison of data over time from 1992 NES respondents, as well as a representative sample of eligible 1994 respondents, the 1994 NES sample design includes both a panel and a cross-section component. The panel component of the 1994 design consists of all [1] respondents from the cross-section component of the 1992 NES sample. The 1994 cross-section component is a new selection of respondents from an area probability sample of households taken from the two-thirds partition of the SRC National Sample. Due to sample design decisions in 1992, when the NES sample moved from using the one-half sample partition to the two-thirds sample partition of the SRC National Sample, the cross-section portion of the 1992 NES sample included a disproportionate number of selections from segments in "B1" PSUs (see Table 2). This same disproportionate distribution is, of course, reflected in the 1994 Panel component of the 1994 NES sample. While this does lead to some statistical inefficiency in the form of increased variance of survey estimates relative to that of an even distribution across the two-thirds partition primary areas, since the "B1" PSU areas do represent a proper subsample of the 1980 National Sample design, separate longitudinal analysis of the Panel component of the 1994 NES may be undertaken as well as analysis of combined Panel and Cross-section data [2]. Table 1 identifies the PSUs for the 1994 National Election Study by MSA status and Region. The "B1" PSUs in the Panel portion of the sample design which received the disproportionate allocation in 1992 to supplement the half-sample are shown in italic print on this table; all PSUs on this table are proportionately represented in the 1994 two-thirds Cross- Section Sample. Table 1: PSUs in the 1994 NES Post-Election Survey By MSA Status and Region (B1 PSUs are marked *) REGION Self-representing MSAs Northeast New York, NY-NJ Philadelphia, PA-NJ Boston, MA Nassau-Suffolk, NY* Pittsburgh, PA North Chicago, IL Central Detroit, MI St. Louis, MO Minneapolis, MN-WI* South Washington, DC-MD-VA* Dallas-Ft Worth, TX* Houston, TX Baltimore, MD Atlanta, GA* West Los Angeles, CA San Francisco, CA REGION Nonself-representing MSAs Northeast Buffalo, NY Newark, NJ* New Haven, CT Atlantic City, NJ Manchester, NH North Milwaukee, WI Central Dayton, OH Kansas City, MO-KS* Des Moines, IA Grand Rapids, MI Fort Wayne, IN Steubenville, OH Saginaw, MI* South Birmingham, AL Columbus, GA-AL Miami, FL Jacksonville, FL* Lakeland, FL McAllen, TX Waco, TX* Wheeling, WV Knoxville, TN Richmond, VA West Seattle, WA Denver, CO Anaheim, CA Riverside, CA* Fresno, CA Eugene, OR Phoenix, AZ* REGION Non-MSAs Northeast Schuyler, NY Gardner, MA* North Sanilac, MI Central Decatur, IN* Saline, NE Mower, MN South Bulloch, GA Sabine, LA* Hale, TX Ashley, AR Bedford, TN Montgomery, VA* Robeson, NC West ElDorado-Alpine, CA Carbon, WY Second Stage Selection of Area Segments The second stage of the 1994 NES National Sample was selected directly from computerized files that were prepared from the 1990 [3] Census file (PL94-171 file on CD Rom) which contains the block-level 1990 Census total housing unit (HU) data. The designated second-stage sampling units (SSUs), termed "area segments", are comprised of census blocks in the metropolitan (MSA) primary areas and either census blocks or enumeration districts (EDs) in the rural areas of non-MSA primary areas. Each SSU block, block combination or enumeration district for non-MSA PSUs was assigned a measure of size equal to the total 1980 occupied housing unit count for the area; MSA SSU block(s) were assigned a minimum measure of 72 1990 total HUs per SSU. Second stage sampling of area segments was performed with probabilities proportionate to the assigned measures of size (PPS). A three-step process of ordering the SSUs within the primary areas produced an implicit stratification of the area segments in the second stage sampling frame, stratified at the county level by geographic location and population. Area segments were stratified within county at the Minor Civil Division (MCD) level by size and income, and at the block and ED level by location within the MCD or county. (For details, refer to the SRC publication, 1980 National Sample: Design and Development.) For the 1994 NES Panel/Cross-section sample the number of area segments used in each PSU varies. In the self-representing (SR) PSUs the number of sample area segments varied in proportion to the size of the primary stage unit, from a high of 12 Cross-section and 7 Panel area segments in the self-representing New York MSA, to a low of 4 Cross-section and no Panel area segments in the smaller self-representing PSUs such as Pittsburgh and Boston MSAs. Most Nonself-representing (NSR) PSUs were represented by 6 Cross-section and 2 Panel area segments except for "B1" PSUs for which there are either 5 or 6 Panel segments. A total of 554 area segments were selected, 191 Panel and 363 Cross-Section segments, 157 in the sixteen self-representing PSUs and 397 in the nonself-representing PSUs as shown in Table 2. In most cases, both Cross-Section and Panel selections were been made from the same area segments within each PSU, so in actual fact a total of 376 distinct National Sample area segments have been used for the 1994 NES Post-Election Study. Table 2: Number [4] of Cross-Section and Panel Area Segments in the 1994 NES Sample Showing PSU Name, National-Sample Stratum and Partition, and MSA Status N. Samp National Sample # of 1994 NES # of 1994 NES PSU #/ PSU Name Cross-section Panel Sample Partition Sample Segs. Segments Six Largest Self-representing PSUs 501 A New York, NY-NJ 12 (7) 12 502 A Los Angeles, CA 12 (5) 12 503 A Chicago, IL 8 8 504 A Philadelphia, PA-NJ 6 6 505 A Detroit, MI 6 6 506 A San Francisco, CA 6 (5) 6 Ten Remaining Self-representing PSUs 507 B1 Washington, DC-MD-VA 6 6 508 B1 Dallas-Ft Worth, TX 6 6 509 A Houston, TX 6 0 510 A Boston, MA 4 0 511 B1 Nassau-Suffolk, NY 4 4 512 A St Louis, MO-IL 4 0 513 A Pittsburgh, PA 4 0 514 A Baltimore, MD 4 0 515 B1 Minneapolis, MN-WI 4 4 516 B1 Atlanta, GA 4 4 Nonself-representing MSAs: Northeast 517 A Buffalo, NY 6 2 518 B1 Newark, NJ 6 6 521 A New Haven, CT (5) 6 2 523 A Atlantic City, NJ (5) 6 2 524 A Manchester, NH 6 2 Nonself-representing MSAs: North Central 526 A Milwaukee, WI 6 2 527 A Dayton, OH 6 2 528 B1 Kansas City, MO-KS 6 6 529 A Des Moines, IA 6 2 531 A Grand Rapids, MI 6 2 532 A Fort Wayne, IN 6 2 533 A Steubenville, OH-WV 6 2 534 B1 Saginaw, MI 6 6 Nonself-representing MSAs: South 536 A Birmingham, AL 6 2 539 A Columbus, GA-AL 6 2 540 A Miami, FL 6 (1) 2 542 B1 Jacksonville, FL 6 6 543 A Lakeland, FL 6 2 544 A McAllen, TX 6 2 545 B1 Waco, TX (5) 6 6 547 A Wheeling, WV-OH 6 2 549 A Knoxville, TN 6 2 550 A Richmond, VA 6 2 Nonself-representing MSAs: West 553 A Seattle, WA 6 2 555 A Denver, CO 6 2 556 A Anaheim, CA 6 2 557 B1 Riverside-San Bernardino, CA 6 6 558 A Fresno, CA 6 2 559 A Eugene, OR 6 2 560 B1 Phoenix, AZ 6 6 Nonself-representing Non-MSAs: Northeast 463 A Schuyler, NY 6 2 464 B1 Gardner, MA 6 6 Nonself-representing Non-MSAs: North Central 465 A Sanilac, MI 6 2 466 B1 Decatur, IN 6 6 468 A Saline, NE 6 2 470 A Mower, MN 6 2 Nonself-representing Non-MSAs: South 473 A Bulloch, GA 6 2 474 B1 Sabine, LA 6 5 476 A Hale, TX 6 2 477 A Ashley, AR 6 2 478 A Bedford, TN 6 2 480 B1 Montgomery, VA 6 5 481 A Robeson, NC 6 2 Nonself-representing Non-MSAs: West 482 A ElDorado-Alpine, CA 6 (1) 2 484 A Carbon, WY 6 2 Total Number of Segments (363) 366 (191) 206 Third Stage Selection of Housing Units For each area segment selected in the second sampling stage, a listing was made of all housing units located within the physical boundaries of the segment. For segments with a very large number of expected housing units, all housing units in a subselected part of the segment were listed. The final equal probability sample of housing units for the 1994 NES was systematically selected from the housing unit listings for the sampled area segments. The new Cross-Section component of the 1994 NES sample design was selected from the SRC National Sample to yield an equal probability sample of households. The distribution of the 1994 cross-section sample is that required by the two-thirds design of the SRC National Sample. The overall probability of selection for 1994 NES Cross-Section households was f=.00001885 or .1885 in 10,000. The equal probability sample of households was achieved for the Cross-Section design by using the standard multi-stage sampling technique of setting the sampling rate for selecting housing units within area segments to be inversely proportional to the PPS probabilities used to select the PSU and area segment [5]. The 1994 Panel consists of all respondents for whom a complete interview was obtained in the 1992 NES Cross-section sample. 1005 1992 cross-section interview HUs make up the 1994 Panel. Fourth Stage Respondent Selection Within each sampled new cross-section housing unit, the SRC interviewer prepared a complete listing of all eligible household members. Using an objective procedure described by Kish (1949) [6] a single respondent was then selected at random to be interviewed. Regardless of circumstances, no substitutions were permitted for the designated respondent. This technique had also been used in 1992 to select the original Panel respondents. In 1994 the same Panel respondent (R) was sought for interview as had been interviewed in 1992. SAMPLE DESIGN SPECIFICATIONS The targeted completed interview sample size for the 1994 NES Post-Election Survey was n = 1,750 total cases. In the original sample size computation, the following assumptions were made for the cross-section component of the sample: response rate for post-election interview = .74, combined occupancy/eligibility rate = .83 and change from updating the sample HU listings = 1.02. The updating was to include only "Type II" updating, i.e., only changes found at selected lines at the time of interviewing; no pre-study update was felt to be necessary due to the fact that most of the selected segments had been used and updated recently on other SRC studies (Health and Retirement Survey and the Asset and Health Dynamics Survey). The assumption as to occupancy/eligibility rate was derived from survey experience in the 1986 NES Post Election Survey [7] and that regarding response rate was based on the 1992 cross-section component outcome for the pre-election interview [8]. The assumptions made for the panel component were: .915 recontact rate based on the .923 recontact rate in the 1993 NES Pilot Study for 1992 cross-section respondents (i.e., same respondents as the current 1994 Panel), .691 response rate for the post-election interview based on NES experience from 1990-1992 in recontacting respondents three times over a two year period, and at .975 change from the update assuming some loss of HUs among panel respondents and inability to track the respondent to a new address. Table 3 provides a full description of the original sample design specifications. Table 4 shows those specifications and assumptions applied to the actual selected Cross-section component of the 1994 NES Sample and also indicates the number of HU listings assigned to each replicate. Table 3: Original Sample Design Specifications and Assumptions 1994 National Post-Election Survey Cross-Section Component Panel Component Total Completed Post interviews 1,130 620 1,750 Response Rate .74 .691 Eligible sample households 1,527 897 2,945 Occupancy/Eligibility Rate[9] .83 NA Panel Recontact Rate NA .915 Sample Units 1,840 980 3,385 Change from Update 1.02 .975 Total Sample lines 1,804 1,005 2,809 [9] Based on field experience in 1986 NES study. Table 4: Original Sample Design Specifications and Assumptions Applied to the Selected Cross-Section Sample Lines for the 1994 National Post-Election Survey Base Reserve Sample Sample Replicates Total Rep 1 Rep 2 Rep 3 Rep 4 Completed Interviews 1,097 31 31 31 1,190 Response Rate .74 .74 .74 .74 .74 Designated Respondents 1,482 42 42 42 1,608 Occupancy/ Elig Rate[10] .83 .83 .83 .83 .83 Sample Units 1,783 51 51 51 1,939 Change from Update 1.02 1.02 1.02 1.02 1.02 Total Sample lines 1,751 50 50 50 1,901 SAMPLE DESIGN OUTCOMES Table 5 compares the original sample design specifications and assumptions for the new Cross-Section Component of the 1994 NES (as in Table 3) applied to the released cross-section sample (Replicate 1) to the outcome for the final Cross-Section sample. Table 6 makes a similar comparison for the Panel Component of the 1994 NES Sample and Table 7 presents a summary of the figures for the combined Cross-Section/Panel Sample. Table 5: Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Cross-Section Component of the 1994 National Post-Election Survey Original Actual Outcome Specifications & Assumptions Applied to: Actual Release (Replicate 1) Completed Interviews 1,097 1,036 Response Rate .74 .721 Designated Respondents 1,482 1,436 Occupancy/Eligibility Rate .83 .824 1,740 Subsampling for dangerous/ locked areas -- .99[11] Sample HU listings 1,786 1,757 Sample growth from update[12] 1.02 1.00 Selected Sample lines 1,751 1,751 Based on the daily monitoring of field results, on November 21, 1994 NES study staff decided that it would be a better use of study resources to raise the cross-section response rate rather than to release additional cross-section sample which might have had the effect of further reducing the response rate. Therefore no reserve replicates of the cross-section sample were released. Table 6 /s shows the panel component sample outcome for the 1994 NES Post-Election Survey. Of course, in this component all sample lines were released; no reserve replicates were designated to be withheld. Due to extremely conservative original assumptions, the actual number of interviews obtained exceeded even the most optimistic projection by nearly 60 interviews. This has more than made up for the fewer than anticipated cross-section interviews which can be seen in Table 7, where entire 1994 NES sample design projections are compared with the combined sample outcome. Table 6: Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Panel Component of the 1994 National Post-Election Survey Original Actual Outcome Specifications & Assumptions Applied to Release Completed Interviews 620[13] 759 Response Rate .691[14] .770 HHs with Eligible Resp 897 986 Panel Recontact Rate .917 .981 Sample Units 980 1,005 Change from update .975 Total Panel Cases 1,005 1,005 Table 7: Figures for Original Sample Design Specifications and Assumptions and Actual Sample Design Outcomes for the Combined Cross-Section/Panel Sample. 1994 National Post-Election Survey Original Actual Outcome Specifications & Assumptions Completed Interviews 1,750 1,795 Overall Response Rate .722 .741 Eligible Sample HH 2,424 2,422 Occ/Elig/Recontact Rate .860 .877 Total Sample HU listings 2,820 2,762 Overall Change from update 1.004 1.002 Selected Sample lines 2,809 2,756 WEIGHTED ANALYSIS OF 1994 NES DATA The area probability sample design for the 1994 NES results in an equal probability sample of U.S. households. However, within sample households a single adult respondent is chosen at random to be interviewed. Since the number of eligible adults may vary from one household to another, the random selection of a single adult introduces inequality into respondents' selection probabilities. In analysis, a respondent selection weight should be used to compensate for these unequal selection probabilities. The value of the respondent selection weight is exactly equal to the number of eligible adults in the household from which the random respondent was selected. The use of the respondent selection weight is strongly encouraged, despite past evaluations which have shown these weights to have little significant impact on the values of NES estimates of descriptive statistics. The Sampling Section has provided two final person-level analysis weights which incorporate sampling, nonresponse and post-stratification factors. One weight variable (#5) is for use with Panel cases only; the other weight variable (#4) is for the 1994 NES Cross-section (which includes both panel and new cross-section cases.) In addition, a Time Series Weight (variable #6) which corrects for panel attrition was constructed. This weight should be used in analyses which compare the 1994 NES to earlier unweighted National Election Study data collections. Analysts interested in developing their own nonresponse or post-stratification adjustment factors must request access to the necessary sample control data from the NES Board. CONSTRUCTION OF ANALYSIS WEIGHTS Nonresponse adjustment factors were constructed at the household level separately for Panel and new Cross-Section component cases. Nonresponse adjustment cells were formed by crossing PSU type (Self-representing, Nonself-representing MSA or non-MSA) by the nine Census divisions (New England, Middle Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain, and Pacific). A nonresponse factor equal to the inverse of the response rate in each cell was applied to the interview cases. In order to have a minimum of approximately 25 cases in each nonresponse adjustment cell, some cells were collapsed across Census Divisions in the same Census Region. Tables 8 and 9 show the nonresponse adjustment factors for the Panel and for new cross section respectively. An intermediate weight was constructed by multiplying the reciprocal of the probability of selection of the household by the nonresponse adjustment factor and by the number of eligible persons in the household [15]. This intermediate weight was used to produce a weighted sex by age category by Census Region table. The age categories used were: 18-44, 45-64, and 65+. Post- stratification factors were constructed to match the sample proportions in the 24 sex by age by Region cells to the July 1993 Census population projections (Current Population Reports, P25- 1111, Table 4). Table 10 shows the post-stratification factors for the 1994 NES Panel. Table 11 shows the post-stratification factors for the complete cross-section (both panel and new cross section cases.) The two final analysis weights were each centered to a mean of 1.0 so that the sum of the weights equals the number of respondents. CONSTRUCTION OF TIME SERIES WEIGHT The 1994 NES Panel consists of 759 respondents originally selected for the 1992 Pre- Election Study. Of 1,126 1992 Pre-Election respondents, 1,005 were also respondents on the 1992 Post-Election Study. All 1,005 1992 Post-Election respondents were eligible for the 1994 NES Panel. In order to adjust for panel attrition, a Time Series Weight was constructed which adjusts the proportions for 30 demographic cells: Education (3) by Age Group (5) by Years of Residence (2) to the 1992 proportions. New 1994 cross-section cases have a Time Series weight of 1.0. In forming the panel attrition weight cells, the following definitions were used: Age Group: 17-24, 25-39, 40-64, 65-74, 75 or more. Education: Less than high school graduate, high school graduate, more than high school education. Years of Residence: Less than 3 years at current residence, 3 or more years at current residence. Table 8 Computation of Nonresponse Adjustment Weights -- Panel Nonresponse Adjustment PSU Type Census Region Response Rate Weight SR-MSA Middle Atlantic 74.6 1.340 East North Central 84.0 1.190 West North Central 92.9 1.077 South Atlantic 71.8 1.392 West South Central 75.0 1.333 Pacific 66.7 1.500 NSR-MSA New England & Middle Atlantic 70.8 1.413 East North Central 78.8 1.269 West North Central 71.4 1.400 South Atlantic 75.0 1.333 East South Central & West South Central 77.6 1.289 Mountain 92.8 1.078 Pacific 72.2 1.386 NSR-non MSA New England & Middle Atlantic 58.7 1.704 East North Central & West North Central 81.0 1.234 South Atlantic 82.7 1.210 East South Central & West South Central 81.8 1.222 Mountain & Pacific 66.7 1.50 Table 9 Computation of Nonresponse Adjustment Weights -- New Cross Section Nonresponse Adjustment PSU Type Census Region Response Rate Weight SR-MSA New England & Middle Atlantic 56.0 1.787 East North Central & West North Central 65.1 1.536 South Atlantic 72.0 1.389 West South Central 52.0 1.923 Pacific 48.4 2.067 NSR-MSA New England 44.0 2.273 Middle Atlantic 65.6 1.524 East North Central 68.6 1.458 West North Central 71.1 1.406 South Atlantic 82.7 1.209 East South Central 80.4 1.243 West South Central 82.5 1.212 Mountain 85.3 1.172 Pacific 71.3 1.402 NSR-non MSA New England & Middle Atlantic 72.5 1.379 East North Central & West North Central 87.8 1.139 South Atlantic 72.4 1.382 East South Central & West South Central 74.7 1.339 Mountain & Pacific 94.6 1.057 Table 10 1994 NES Panel Post-Stratification Weight Census Age Census Est. 94 Nat'l Post- Sex Region Group July 1, 1993 Election Strat. Study Weight Male Northeast 18-44 10,652,000 8,676,130 1.2277 45-64 4,867,000 5,246,960 0.9276 65+ 2,815,000 2,880,610 0.9772 Midwest 18-44 12,679,000 13,912,400 0.9113 45-64 5,626,000 6,229,820 0.9031 65+ 3,211,000 5,109,480 0.6284 South 18-44 18,797,000 16,207,280 1.1598 45-64 8,177,000 9,324,160 0.8770 65+ 4,574,000 3,440,280 1.3295 West 18-44 12,611,000 8,973,210 1.4054 45-64 4,908,000 2,573,920 1.9068 65+ 2,580,000 2,295,480 1.1239 Female Northeast 18-44 10,844,000 8,032,420 1.3500 45-64 5,338,000 3,233,370 1.6509 65+ 4,329,000 3,012,940 1.4368 Midwest 18-44 12,783,000 11,746,140 1.0883 45-64 5,990,000 6,753,230 0.8870 65+ 4,789,000 4,847,570 0.9879 South 18-44 18,950,000 17,179,490 1.1031 45-64 8,882,000 9,486,140 0.9363 65+ 6,753,000 5,970,310 1.1311 West 18-44 11,979,000 10,117,500 1.1840 45-64 5,077,000 3,416,980 1.4858 65+ 3,543,000 2,752,280 1.2873 Totals 190,754,000 171,418,100 Table 11 1994 NES Cross-section Post-Stratification Weight Census Age Census Est. 94 Nat'l Post- Sex Region Group July 1, 1993 Election Strat. Study Weight Male Northeast 18-44 10,652,000 7,780,520 1.3691 45-64 4,867,000 3,562,080 1.3663 65+ 2,815,000 2,807,870 1.0025 Midwest 18-44 12,679,000 13,282,300 0.9546 45-64 5,626,000 6,435,320 0.8742 65+ 3,211,000 3,968,760 0.8091 South 18-44 18,797,000 16,523,490 1.1376 45-64 8,177,000 8,230,300 0.9935 65+ 4,574,000 4,023,460 1.1368 West 18-44 12,611,000 9,120,530 1.3827 45-64 4,908,000 3,867,010 1.2692 65+ 2,580,000 2,414,850 1.0684 Female Northeast 18-44 10,844,000 8,160,800 1.3288 45-64 5,338,000 3,776,480 1.4135 65+ 4,329,000 4,027,800 1.0748 Midwest 18-44 12,783,000 11,222,760 1.1390 45-64 5,990,000 6,169,130 0.9710 65+ 4,789,000 4,186,580 1.1439 South 18-44 18,950,000 17,375,850 1.0906 45-64 8,882,000 7,917,440 1.1218 65+ 6,753,000 5,942,100 1.1365 West 18-44 11,979,000 10,060,750 1.1907 45-64 5,077,000 4,359,910 1.1645 65+ 3,543,000 3,088,300 1.1472 Totals 190,754,000 168,304,380 In order to obtain a minimum of approximately 15 cases per cell, some of the cells were collapsed across age groups. Table 12 shows the panel attrition factors for the 25 Years in Residence by Education Level by Age Group cells. Table 12 Panel Attrition (Time Series) Weight Factors Years of Panel Attrition Residence Education Level Age Group Weight Factor < 3 < HS Graduate 25-39 1.750 40+ 1.818 < HS Grad, HS Grad 17-24 1.428 HS Graduate 25-39 1.933 40+ 1.562 HS Graduate 17-24 1.375 25-39 1.376 40+ 1.326 3+ < HS Grad 17-39 1.308 40-64 1.423 65-74 1.583 75+ 2.125 HS Graduate 17-24 1.571 25-39 1.533 40-64 1.443 65-74 1.417 75+ 1.500 > HS Graduate 17-24 1.417 25-39 1.354 40-64 1.564 65-74 1.269 75+ 1.769 PROCEDURES FOR SAMPLING ERROR ESTIMATION The 1994 NES is based on a stratified multi-stage area probability sample of United States households. Although smaller in scale, the NES sample design is very similar in it basic structure to the multi-stage designs used for major federal survey programs such as the Health Interview Survey (HIS) or the Current Population Survey (CPS). The survey literature refers to the NES, HIS and CPS samples as complex designs, a loosely-used term meant to denote the fact that the sample incorporates special design features such as stratification, clustering and differential selection probabilities (i.e., weighting) that analysts must consider in computing sampling errors for sample estimates of descriptive statistics and model parameters. This section of the 1994 NES sample design description focuses on sampling error estimation and construction of confidence intervals for survey estimates of descriptive statistics such as means, proportions, ratios, and coefficients for linear and logistic linear regression models. Standard analysis software systems such SAS, SPSS, OSIRIS assume simple random sampling (SRS) or equivalently independence of observations in computing standard errors for sample estimates. In general, the SRS assumption results in underestimation of variances of survey estimates of descriptive statistics and model parameters. Confidence intervals based on computed variances that assume independence of observations will be biased (generally too narrow) and design-based inferences will be affected accordingly. Sampling Error Computation Methods and Programs Over the past 50 years, advances in survey sampling theory have guided the development of a number of methods for correctly estimating variances from complex sample data sets. A number of sampling error programs which implement these complex sample variance estimation methods are available to NES data analysts. The two most common approaches to the estimation of sampling error for complex sample data are through the use of a Taylor Series Linearization of the estimator (and corresponding approximation to its variance) or through the use of resampling variance estimation procedures such as Balanced Repeated Replication (BRR) or Jackknife Repeated Replication(JRR). New Bootstrap methods for variance estimation can also be included among the resampling approaches. See Rao and Wu (1988). 1. Linearization Approach If data are collected using a complex sample design with unequal size clusters, most statistics of interest will not be simple linear functions of the observed data. The objective of the linearization approach is to apply Taylor's method to derive an approximate form of the estimator that is linear in statistics for which variances and covariances can be directly estimated. Kish, 1965; Woodruff, 1971). Linearized variance approximations are derived for estimators of ratio means (Kish and Hess, 1959); finite population regression coefficients and correlation coefficients (Kish and Frankel, 1974); and many other non-linear statistics. Software packages such as SUDAAN and PC CARP (see below) use the Taylor Series linearization method to estimate standard errors for the coefficients of logistic regression models. In these programs, an iteratively reweighted least squares algorithm is used to compute maximum likelihood estimates of model parameters. At each step of the model fitting algorithm, a Taylor Series linearization approach is used to compute the variance/covariance matrix for the current iteration's parameter estimates (Binder, 1983). Available sampling error computation software that utilizes the Taylor Series linearization method includes: SUDAAN and PC SUDAAN, SUPERCARP AND PC CARP, CLUSTERS, OSIRIS PSALMS, OSIRIS PSRATIO, and OSIRIS PSTABLES. PC SUDAAN and PC CARP include procedures for estimation of sampling error both for descriptive statistics such as means, proportion, totals and for parameters of commonly used multivariate models (least squares regression, logistic regression). 2. Resampling Approaches In the mid-1940's, P.C. Mahalanobis (1946) outlined a simple replicated procedure for selecting probability samples that permits simple, unbiased estimation of variances. The practical difficulty with the simple replicated approach to design and variance estimation is that many replicates are needed to achieve stability of the variance estimator. Unfortunately, a design with many independent replicates must utilize a coarser stratification than alternative designs--to achieve stable variance estimates, sample precision must be sacrificed. Balanced Repeated Replication (BRR), Jackknife Repeated Replication (JRR) and the Bootstrap are alternative replication techniques that may be used for estimating sampling errors for statistics based on complex sample data. The BRR method is applicable to stratified designs in which two half-sample units (i.e., PSUs) are selected from each design stratum. The conventional "two PSU-per-stratum" design in the best theoretical example of such a design although in practice, collapsing of strata (Kalton, 1977) and random combination of units within strata are employed to restructure a sample design for BRR variance estimation. The half-sample codes prepared for the 1994 NES data set require the collapsing of nonself-representing strata and the randomized combination of selection units within self-representing (SR) strata. When full balancing of the half-sample assignments is employed (Wolter, 1985), BRR is the most computationally efficient of the replicated variance estimation techniques. The number of general purpose BRR sampling error estimation programs in the public domain is limited. The OSIRIS REPERR program includes the option for BRR estimation of sampling errors for least squares regression coefficients and correlation statistics. Research organizations such as Westat, Inc. and the National Center for Health Statistics have developed general purpose programs for BRR estimation of standard errors. Another option is to use SAS or SPSS Macro facilities to implement the relatively simple BRR algorithm. The necessary computation formulas and Hadamard matrices to define the half-sample replicates are available in Wolter (1985). With improvements in computational flexibility and speed, jackknife (JRR) and bootstrap methods for sampling error estimation and inference have become more common (Rao and Wu, 1988 ). Few general purpose programs for jackknife estimation of variances are available to analysts. OSIRIS REPERR has a JRR module for estimation of standard errors for regression and correlation statistics. Other stand alone programs may also be available in the general survey research community. Like BRR, the algorithm for JRR is relatively easy to program using SAS, SPSS or S-Plus macro facilities. BRR and JRR are variance estimation techniques, each designed to minimize the number of "resamplings" needed to compute the variance estimate. In theory, the bootstrap is not simply a tool for variance estimation but an approach to actual inference for statistics. In practice, the bootstrap is implemented by resampling (with replacement) from the observed sample units. To ensure that the full complexity of the design is reflected , the selection of each bootstrap reflects the full complexity of the stratification, clustering and weighting that is present in the original sample design. A large number of bootstrap samples are selected and the statistic of interest is computed for each. The empirical distribution of the estimate that results from the large set of bootstrap samples can then be used to a variance estimate and a support interval for inference about the population statistic of interest. In most practical survey analysis problems, the JRR and Bootstrap methods should yield similar results. Most survey analysts should choose JRR due to its computational efficiency. NES data analysts interested in the bootstrap technique are referred to LePage and Billard (1992) for additional reading and a bibliography for the general literature on this topic. One aspect of BRR, JRR and bootstrap variance estimation that is often pushed aside in practice is the treatment of analysis weights. In theory, when a resampling occurs (i.e., a BRR half sample is formed), the analysis weights should be recomputed based only on the selection probabilities, nonresponse characteristics and post-stratification outcomes for the units included in the resample. This is the correct way of performing resampling variance estimation; however, in practice acceptable estimates can be obtained through use of the weights as they are provided on the public use data set. Sampling Error Computation Models Regardless of whether linearization or a resampling approach is used, estimation of variances for complex sample survey estimates requires the specification of a sampling error computation model. NES data analysts who are interested in performing sampling error computations should be aware that the estimation programs identified in the preceding section assume a specific sampling error computation model and will require special sampling error codes. Individual records in the analysis data set must be assigned sampling error codes which identify to the programs the complex structure of the sample (stratification, clustering) and are compatible with the computation algorithms of the various programs. To facilitate the computation of sampling error for statistics based on 1994 NES data, design-specific sampling error codes will be routinely included in all public-use versions of the data set. Although minor recoding may be required to conform to the input requirements of the individual programs, the sampling error codes that are provided should enable analysts to conduct either Taylor Series or Replicated estimation of sampling errors for survey statistics. Table 13 defines the sampling error coding system for 1994 NES sample cases. Two sampling error code variables are defined for each case based on the sample design primary stage unit (PSU) and area segment in which the sample household is located. Sampling Error Stratum Code (Variable #63). The Sampling Error Computation Stratum Code is the variable which defines the sampling error computation strata for all sampling error analysis of the NES data. With the exception of the New York, Los Angeles and Chicago MSAs, each self-representing (SR) design stratum is represented by one sampling error computation stratum. Due to their population size, two sampling error computation strata are defined for each of the three largest MSAs. Pairs of similar nonself-representing (NSR) primary stage design strata are "collapsed" (Kalton, 1977) to create NSR sampling error computation strata. The SRC National Sample design uses Controlled selection and a "one-per-stratum" PSU allocation are used to select the primary stage of the 1994 NES national sample. The purpose in using Controlled Selection and the "one-per-stratum" sample allocation is to reduce the between-PSU component of sampling variation relative to a"two-per-stratum" primary stage design. Despite the expected improvement in sample precision, a drawback of the "one-per- stratum" design is that two or more sample selection strata must be collapsed or combined to form a sampling error computation stratum. Variances are then estimated under the assumption that a multiple PSU per stratum design was actually used for primary stage selection. The expected consequence of collapsing design strata into sampling error computation strata is the overestimation of the true sampling error; that is, the sampling error computation model defined by the codes contained in Table 13 will yield estimates of sampling errors which in expectation will be slightly greater than the true sampling error of the statistic of interest. SECU - Stratum-specific Sampling Error Computation Unit code (Variable #64) is a half sample code for analysis of sampling error using the BRR method or approximate "two-per-stratum" Taylor Series method (Kish and Hess,1959). Within the SR sampling error strata, the SECU half sample units are created by dividing sample cases into random halves, SECU=1 and SECU=2. The assignment of cases to half-samples is designed to preserve the stratification and second stage clustering properties of the sample within an SR stratum. Sample cases are assigned to SECU half samples based on the area segment in which they were selected. For this assignment, sample cases were placed in original stratification order (area segment number order) and beginning with a random start entire area segment clusters were systematically assigned to either SECU=1 or SECU=2. In the general case of nonself-representing (NSR) strata, the half sample units are defined according to the PSU to which the respondent was assigned at sample selection. That is, the half samples for each NSR sampling error computation stratum bear a one-to-one correspondence to the sample design NSR PSUs. The particular sample coding provided on the NES public use data set is consistent with the "ultimate cluster" approach to complex sample variance estimation (Kish, 1965; Kalton, 1977). Individual stratum, PSU and segment code variables may be needed by NES analysts interested in components of variance analysis or estimation of hierarchical models in which PSU-level and neighborhood-level effects are explicitly estimated. Table 13 shows the sampling error stratum and SECU codes to be used for the paired selection model for sampling error computations for any 1994 NES analyses; the same codes can be used when using the combined cross-section/panel data or when using either panel or cross- section data separately. The 42 strata reflect the two-thirds National Sample design used in 1994. It can be seen from this table that the three-digit 1994 SE code is comprised of: first the two-digit SE Stratum code followed by the one-digit SECU code. Table 13. 1994 National Election Study Sampling Error Codes Sampling Error SECU Code Segment Segment Stratum Code (Half Sample) PSU Numbers Numbers Number Cross- Panel Section 01 1 501 103 119 135 103 103 103 2 501 107 123 139 123 02 1 501 111 127 143 111 127 143 2 501 115 131 148 131 148 03 1 502 110 123 136 136 2 502 101 114 126 114 04 1 502 104 117 129 117 129 2 502 107 120 133 120 05 1 503 112 129 112 129 2 503 117 134 117 134 06 1 503 103 120 103 120 2 503 107 125 107 125 07 1 504 102 110 117 102 110 117 2 504 106 113 121 106 113 121 08 1 505 105 112 119 105 112 119 2 505 101 108 115 101 108 115 09 1 506 104 110 116 104 110 116 2 506 101 107 113 107 113 10 1 507 105 111 115 105 111 115 2 507 103 107 113 103 107 113 11 1 508 101 107 110 101 107 110 2 508 103 109 114 103 109 114 12 1 509 104 109 114 2 509 101 107 111 13 1 510 105 111 2 510 101 107 14 1 511 105 111 105 111 2 511 102 108 102 108 15 1 512 102 108 2 512 105 111 16 1 513 101 107 2 513 104 110 17 1 514 104 110 2 514 101 107 18 1 515 105 111 105 111 2 515 102 108 102 108 19 1 516 102 108 102 108 2 516 105 111 105 111 20 1 517 101 103 105 105 111 107 109 111 2 518 101 103 105 101 103 105 107 109 111 107 109 111 21 1 521 103 105 107 103 109 109 111 2 523 103 105 107 105 111 109 111 22 1 524 102 104 106 102 108 108 110 112 2 534 102 104 106 102 104 106 108 110 112 108 110 112 23 1 526 101 103 105 105 111 107 109 111 2 527 101 103 105 103 109 107 109 111 24 1 528 102 104 106 102 104 106 108 110 112 108 110 112 2 529 102 104 106 106 112 108 110 112 25 1 531 102 104 106 106 112 108 110 112 2 532 102 104 106 104 110 108 110 112 26 1 533 102 104 106 106 112 108 110 112 2 547 101 103 105 101 107 107 109 111 27 1 536 101 103 105 105 111 107 109 111 2 539 101 103 105 105 111 107 109 111 28 1 540 101 103 105 109 107 109 111 2 542 102 104 106 102 104 106 108 110 112 108 110 112 29 1 543 102 104 106 104 106 108 110 112 2 545 103 105 107 101 103 105 109 111 30 1 544 101 103 105 103 109 107 109 111 2 476 001 004 006 001 012 007 010 012 31 1 549 101 103 105 103 109 107 109 111 2 550 103 105 105 103 109 107 109 111 32 1 553 102 104 106 106 112 108 110 112 2 555 101 103 105 105 111 107 109 111 33 1 556 101 105 107 101 107 109 111 2 557 102 104 106 102 104 106 108 110 112 108 110 112 34 1 558 102 104 106 102 108 108 110 112 2 559 101 103 105 105 111 107 109 111 35 1 560 104 108 112 104 108 112 2 560 102 106 110 102 106 110 36 1 463 001 003 005 002 008 007 009 011 2 464 002 004 005 001 004 005 009 010 012 009 011 012 37 1 465 001 003 005 005 011 007 009 011 2 466 002 004 005 001 004 008 008 010 012 010 011 012 38 1 468 001 002 006 006 012 007 008 011 2 470 002 003 005 002 012 007 011 012 39 1 473 001 005 008 006 012 009 011 012 008 011 2 474 002 004 007 001 004 007 008 011 012 008 011 40 1 477 001 003 005 006 012 007 010 012 2 478 002 005 006 005 010 008 010 012 41 1 480 002 006 007 002 005 007 008 010 012 010 011 2 481 001 004 005 001 008 007 009 011 42 1 482 002 004 005 007 007 009 012 2 484 001 004 006 004 011 009 011 012 Generalized Sampling Error Results for the 1994 NES To assist NES analysts, the OSIRIS PSALMS program was used to compute sampling errors for a wide-ranging example set of means and proportions estimated from the 1988 NES Pre-election Survey data set [16]. For each estimate, sampling errors were computed for the total sample and for twenty demographic and political affiliation subclasses of the 1988 NES Pre-Election Survey sample. The results of these sampling error computations were then summarized and translated into the general usage sampling error table provided in Table 14. Incorporating the pattern of "design effects" observed in the extensive set of example computations, Table 14 provides approximate standard errors for percentage estimates based on the 1988 NES. To use the table, examine the column heading to find the percentage value which best approximates the value of the estimated percentage that is of interest [17]. Next, locate the approximate sample size base (denominator for the proportion) in the left-hand row margin of the table. To find the approximate standard error of a percentage estimate, simply cross-reference the appropriate column (percentage) and row (sample size base). Note: the tabulated values represent approximately one standard error for the percentage estimate. To construct an approximate confidence interval, the analyst should apply the appropriate critical point from the "z" distribution (e.g., z=1.96 for a two-sided 95% confidence interval half-width). Furthermore, the approximate standard errors in the table apply only to single point estimates of percentages not to the difference between two percentage estimates. The generalized variance results presented in Table 14 are a useful tool for initial, cursory examination of the NES survey results. For more in depth analysis and reporting of critical estimates, analysts are encouraged to compute exact estimates of standard errors using the appropriate choice of a sampling error program and computation model. Table 14: Generalized Variance Table. 1994 NES Post-Election Survey. APPROXIMATE STANDARD ERRORS FOR PERCENTAGES For percentage estimates near: Sample n 50% 40% 30% 20% 10% or 60% or 70% or 80% or 90% The approximate standard error of the percentage is: 100 5.406 5.297 4.955 4.325 3.244 200 3.853 3.775 3.531 3.082 2.312 300 3.170 3.106 2.905 2.536 1.902 400 2.766 2.710 2.535 2.213 1.660 500 2.492 2.442 2.284 1.994 1.495 750 2.072 2.030 1.899 1.658 1.243 1000 1.826 1.789 1.674 1.461 1.096 1250 1.661 1.628 1.523 1.329 0.997 1500 1.542 1.511 1.413 1.233 0.925 1800 1.434 1.405 1.315 1.147 0.861 References Binder, D.A. (1983), "On the variances of asymptotically normal estimators from complex surveys," International Statistical Review, Vol. 51, pp. 279-292. Kalton, G. (1977), "Practical methods for estimating survey sampling errors," Bulletin of the International Statistical Institute, Vol 47, 3, pp. 495-514. Kish, L. (1965), Survey Sampling. New York: John Wiley & Sons, Inc. Kish, L., & Frankel, M.R. (1974), "Inference from complex samples," Journal of the Royal Statistical Society, B, Vol. 36, pp. 1-37. Kish, L., & Hess, I. (1959), "On variances of ratios and their differences in multi-stage samples," Journal of the American Statistical Association, 54, pp. 416-446. LePage, R., & Billard, L. (1992), Exploring the Limits of Bootstrap. New York: John Wiley Sons, Inc. Mahalanobis, P.C. (1946), "Recent experiments in statistical sampling at the Indian Statistical Institute," Journal of the Royal Statistical Society, Vol 109, pp. 325-378. Rao, J.N.K & Wu, C.F.J. (1988.), "Resampling inference with complex sample data," Journal of the American Statistical Association, 83, pp. 231-239. Wolter, K.M. (1985 ). Introduction to Variance Estimation. New York: Springer -Verlag. Woodruff, R.S. (1971), "A simple method for approximating the variance of a complicated estimate," Journal of the American Statistical Association, Vol. 66, pp. 411-414. NOTES [1] The Panel consists of all 1005 Respondents from the 1992 NES study Cross-Section sample. Of these, 925 were recontacted in the 1993 NES Pilot Study (a follow-up of the 1992 NES survey), of which 750 were re-interviewed, 98 refused to be re-interviewed and 77 could not be re-interviewed at that time due to some 'permanent' condition. 80 of the 1005 1992 Cross-section respondents could not be found for re-interview in 1993. [2] Analysis of pooled data from respondents from both components of the 1994 NES sample requires a strong assumption about the nature of the attrition of the 1992 cross-section sample. It must be assumed that panel attrition is not correlated with variables under consideration in the analysis. [3] Non-MSA segments were selected from the 1980 Census summary tape file series STF1B file, with minimum SSU size of 50 occupied H.U.s. [4] The number of segments shown for the 1994 Panel is the expected count; it is based on the number of 1992 Cross-Section segments. It is possible that some of these 1992 segments yielded no 1992 interviews and so do not actually show up in the 1994 Panel. The total number of segments shown for the 1994 Cross-section sample also includes three segments from which no listed HU was selected for the 1994 cross-section, due to few or no HU listings for that segment. Where different, the actual number of segments having selections in 1994 is shown in parentheses to the left. [5] Kish, L. (1965). Survey Sampling, John Wiley & Sons, New York, NY. [6] Kish, L. (1949). "A procedure for objective respondent selection within the household," Journal of the American Statistical Association, Vol 44, pp. 380-387. [7] The 1986 NES was the most recent NES sample using the two-thirds National Sample without alteration (e.g., increasing number of segments in the B1 areas as in 1992). Occupancy/eligibility rate was .835. [8] The response rate in 1986 had been unusually low, and it was felt that the more recent experience in the two-thirds partition PSUs would be the best estimate--less affected than occupancy/eligibility rate by the increased number of segments in B1 areas. [9] Based on field experience in 1986 NES study. To most closely tailor the field effort to the sample field experience during this study, the cross-section sample had four replicates designated (see Table 4). Each replicate is a proper subsample of the NES sample. Replicate 1, considered the "base sample", was to be released for interviewing to begin November 9, 1994, the day following Election Day 1994. The other three replicates of the cross-section sample (Replicates 2-4) were designated "Reserve" replicates, none, one or more to be released for field work no later than November 21, 1994 at the discretion of NES study staff based on daily monitoring of field results from Release 1. Reserve replicates 2-4 of the cross-section component of the NES sample were never, in fact, released for field work. [10] A subsampling of one-third of selected addresses was made in certain cases when selected lines were determined to be within locked buildings, in gated subdivisions or in areas which posed a danger to interviewing staff. This allowed concentration of greater field effort in these circumstances to obtain at least some interviews. In cases where this was done, appropriate weighting of the results will be used to compensate. (See Table 5.) [11] One percent of the released sample was lost due to subsampling in three locked and dangerous segment areas; 17 of the 20 selected lines excluded from these six segments were in replicate 1. These lines were assigned a result code of '75' and considered 'Non-Sample' lines. [12] Since only the Type II updating process was applied to the cross-section component of the 1994 NES Sample, the update inflation factor was set at 1.02 -- slightly lower than the usual factor of 1.03 typical of combined Type I (pre-study) and Type II updating inflation applied to the National Sample. [13] Actually the projection ranged from 620-700 completed interviews. See comments in following footnote. [14] An overall Panel response rate of 69.1% was assumed, based on previous recontact experience (response rate of 1990 Pilot Study respondents to the 1992 NES Pre-Election Study follow-up): 750 cases were interviewed twice previously at 76.6% response rate = 575 cases, and 255 other cases combined 17.6% response rate = 45 cases. Removing the change from update and recontact rate (1005 - 25 - 83 = 897), overall response rate: 620/897 = .691. This was admittedly a very conservative estimate and actual projection of expected number of interviews was a range of 620-700. [15] In constructing the analysis weight, a maximum of three eligible adults was allowed [16] The design effects from the 1988 NES are expected to be similar to those for the 1994 NES. Sampling errors for the 1994 NES have not been run. [17] The standard error of a percentage is a systematic function with its maximum centered at=50%; i.e., the standard error pf p=40% and p=60% estimates are equal. >> SAMPLE DESIGN 1996 ELECTION STUDY STUDY POPULATION The study population for the 1996 National Pre/Post-Election Study (NES) is defined to include all United States citizens of voting age on or before the 1996 Election Day. Eligible citizens must have resided in housing units in the forty-eight coterminous states. This definition excludes persons living in Alaska or Hawaii and requires eligible persons to have been both a United States citizen and eighteen years of age on or before the 5th of November 1996. MULTI-STAGE AREA PROBABILITY SAMPLE DESIGN The 1996 NES is based on a multi-stage area probability sample selected from the Survey Research Center's (SRC) National Sample design. Identification of the 1996 NES sample respondents was conducted using a four stage sampling process--a primary stage sampling of U.S. Metropolitan Statistical Areas (MSAs) or New England County Metropolitan Areas (NECMAs)[1] and counties, followed by a second stage sampling of area segments, a third stage sampling of housing units within sampled area segments and concluding with the random selection of a single respondent from selected housing units. A detailed documentation of the 1980 SRC National Sample, from which the 1996 NES Panel was originally drawn is provided in the SRC publication titled 1980 SRC National Sample: Design and Development. A detailed documentation of the 1990 SRC National Sample, from which the 1996 NES Cross-section supplement was drawn, is provided in the SRC publication titled 1990 SRC National Sample: Design and Development. The 1996 NES sample design called for a 1996 NES Panel component consisting of all respondents to the 1994 NES study, originally drawn from the 1980 SRC National Sample, and a 1996 NES Cross-section component drawn from the 1990 SRC National Sample. Although both of these SRC National Samples are multi-stage area probability samples as described above, there are differences in specific details at the various stages of the two SRC National Samples which will be described below. Figure 1 shows in schematic detail the original sources of the components of the 1996 NES Sample. On this figure the "n" indicated in the 1992 and 1994 boxes is actually the number of Respondents from that year and component that became the Panel component two years later. Of course the "n" shown for the 1996 NES Panel and Cross-section components does not refer to 1996 Respondents but, for the 1996 Panel, to the total number of sample eligible households (i.e. the total of the Respondents from both components of 1994) and, for the Cross-section supplement, to the total selected number of listed housing units used in the 1996 NES. Figure 1: Source of 1996 NES Sample Cases 1980 SRC 1990 SRC National Sample National Sample 1992 NES Cross-section (n=1,005) 1994 NES 1994 NES Panel Cross-section (n=759) (n=1,036) 1996 NES 1996 NES Panel Cross-section (n=1,795) (n=803)[2] Both 1980 & 1990 National Samples 1996 NES Combined Sample (n=2,598) Selection Stages for the 1996 NES Panel Component: 1980 SRC National Sample[3] Primary Stage Selection: 1996 NES Panel Component The selection of primary stage sampling units (PSUs), which depending on the sample stratum are either MSAs, single counties or groupings of small counties, is based on the county-level 1980 Census Reports of Population and Housing. Primary stage units were assigned to 84 explicit strata based on MSA/non-MSA status, PSU size, and geographic location. Sixteen of the 84 strata contain only a single self-representing PSU, each of which is included with certainty in the primary stage of sample selection. The remaining 68 nonself-representing strata contain more than one PSU. From each of these nonself-representing strata, one PSU was sampled with probability proportionate to its size (PPS) measured in 1980 occupied housing units. The full SRC National Sample of 84 primary stage selections was designed to be optimal for surveys roughly two to three times the size of the 1994 NES. To permit the flexibility needed for optimal design of smaller survey samples, the primary stage of the SRC National Sample can be readily partitioned into smaller subsamples of PSUs such as a one-half sample or two-thirds sample partition. Each of the partitions represents a stratified subselection from the full 84 PSU design. The one-half partition of the 1980 National Sample (i.e., the "A" primary sampling units or PSUs) includes 11 of the 16 self-representing MSA PSUs and a stratified subsampling of 34 of the 68 nonself-representing PSUs of the SRC National Sample. The two-thirds partition includes all of the "A" PSUs plus "B1" PSUs, i.e., 5 additional self-representing PSUs and 11 additional nonself-representing PSUs. Since the 1994 NES desired comparison of data over time from 1992 NES respondents, as well as a representative sample of eligible 1994 respondents, the 1994 NES sample design included both a Panel and a Cross-section component. The Panel component of the 1994 design consisted of all[4] respondents from the NES Cross-section component of the 1992 NES sample. The 1994 NES Cross-section component was a new selection of respondents from an area probability sample of households taken from the two-thirds partition of the SRC National Sample. The Panel component of the 1996 NES sample consists of all 1994 respondents from both of these 1994 NES components. See Figure 1. Due to sample design decisions in 1992, when the NES sample moved from using the one-half sample partition to the two-thirds sample partition of the SRC National Sample, the Cross-section portion of the 1992 NES sample included a disproportionate number of selections from segments in "B1" PSUs (see Table 1). This same disproportionate distribution was, of course, reflected in the Panel component of the 1994 NES sample and, thus carried to the 1996 NES Panel. While this led to some statistical inefficiency in the form of increased variance of survey estimates relative to that of an even distribution across the two-thirds partition primary areas, since the "BI" PSU areas do represent a proper subsample of the 1980 National Sample design, separate longitudinal analysis of the 1996 NES Panel (i.e., analysis of combined 1994 Panel and 1994 Cross-section data)[5] can be undertaken. Table 1 identifies the PSUs for the Panel component of the 1996 National Election Study by MSA status and Region. The "B1" PSUs in the Panel portion of the sample design which received the disproportionate allocation in 1992 to supplement the half-sample are also indicated on this table as well as the number of area segments carried over to the 1996 NES Panel component (see next section); all PSUs on this table are proportionately represented in the 1994 NES two-thirds Cross-section Sample. Second Stage Selection of Area Segments: 1996 NES Panel Component The second stage of the 1994 NES National Sample was selected directly from computerized files that were prepared from the 1990[6] Census file (PL94-171 file on CD Rom) which contains the block-level 1990 Census total housing unit (HU) data. The designated second-stage sampling units (SSUs), termed "area segments", are comprised of census blocks in the metropolitan (MSA) primary areas and either census blocks or enumeration districts (EDs) in the rural areas of non-MSA primary areas. Each SSU block, block combination or enumeration district for non-MSA PSUs was assigned a measure of size equal to the total 1980 occupied housing unit count for the area. MSA SSU block(s) were assigned a minimum measure of 72 1990 total HUs per SSU; non-MSA SSU blocks were assigned a minimum measure of 50 1980 occupied HUs per SSU. Second stage sampling of area segments was performed with probabilities proportionate to the assigned measures of size (PPS). A three-step process of ordering the SSUs within the primary areas produced an implicit stratification of the area segments in the second stage sampling frame, stratified at the county level by geographic location and population. Area segments were stratified within county at the Minor Civil Division (MCD) level by size and income, and at the block and ED level by location within the MCD or county. (For details, refer to the SRC publication, 1980 National Sample: Design and Development.) For the 1994 NES combined Panel/Cross-section sample the number of area segments used in each PSU varied. In the self-representing (SR) PSUs the number of sample area segments varied in proportion to the size of the primary stage unit, from a high of 12 Cross-section and 7 Panel area segments in the self-representing New York MSA, to a low of 4 Cross-section and no Panel area segments in the smaller self-representing PSUs such as Pittsburgh and Boston MSAs. Most Nonself-representing (NSR) PSUs were represented by 6 Cross-section and 2 Panel area segments except for "B1" PSUs for which there are either 5 or 6 Panel segments. A total of 554 area segments were selected for the 1994 NES, 191 Panel and 363 Cross-section segments, 157 in the sixteen self-representing PSUs and 397 in the nonself-representing PSUs as shown in the last column of Table 1. In most cases, both 1994 NES Cross-section and 1994 NES Panel selections were made from the same area segments within each PSU, so in actual fact a total of 376 distinct 1980 National Sample area segments were used for the 1994 NES Post-election Study. Of these, 364 segments had respondents in 1994 and were carried over to the Panel component of the 1996 NES Study. Table 1: PSU Name and Number[7] of Panel Area Segments in the 1996 NES Sample Showing 1980 SRC National-Sample Stratum, Partition and MSA Status National Sample National Sample # of 1996 NES PSU Number and PSU Name Panel Segments Partition Six Largest Self-representing PSUs 501 A New York, NY-NJ 11 502 A Los Angeles, CA 10 503 A Chicago, IL 8 504 A Philadelphia, PA-NJ 6 505 A Detroit, MI 6 506 A San Francisco, CA 6 Ten Remaining Self-representing PSUs 507 B1 Washington, DC-MD-VA 6 508 B1 Dallas-Ft Worth, TX 6 509 A Houston, TX 5 510 A Boston, MA 3 511 B1 Nassau-Suffolk, NY 4 512 A St Louis, MO-IL 3 513 A Pittsburgh, PA 4 514 A Baltimore, MD 4 515 B1 Minneapolis, MN-WI 4 516 B1 Atlanta, GA 4 Nonself-representing MSAs: Northeast 517 A Buffalo, NY 5 518 B1 Newark, NJ 6 521 A New Haven, CT 5 523 A Atlantic City, NJ 5 524 A Manchester, NH 6 Nonself-representing MSAs: Midwest (North Central in 1980 Census) 526 A Milwaukee, WI 6 527 A Dayton, OH 5 528 B1 Kansas City, MO-KS 6 529 A Des Moines, IA 6 531 A Grand Rapids, MI 6 532 A Fort Wayne, IN 6 533 A Steubenville, OH-WV 6 534 B1 Saginaw, MI 6 Nonself-representing MSAs: South 536 A Birmingham, AL 6 539 A Columbus, GA-AL 6 540 A Miami, FL 6 542 B1 Jacksonville, FL 6 543 A Lakeland, FL 6 544 A McAllen, TX 6 545 B1 Waco, TX 6 547 A Wheeling, WV-OH 6 549 A Knoxville, TN 6 550 A Richmond, VA 6 Nonself-representing MSAs: West 553 A Seattle, WA 6 555 A Denver, CO 6 556 A Anaheim, CA 5 557 B1 Riverside-San Bernardino, CA 6 558 A Fresno, CA 6 559 A Eugene, OR 6 560 B1 Phoenix, AZ 6 Nonself-representing Non-MSAs: Northeast 463 A Schuyler County, NY 8 464 B1 Gardner County, MA 8 Nonself-representing Non-MSAs: Midwest (North Central in 1980 Census) 465 A Sanilac County, MI 5 466 B1 Decatur County, IN 8 468 A Saline County, NE 7 470 A Mower County, MN 6 Nonself-representing Non-MSAs: South 473 A Bulloch County, GA 7 474 B1 Sabine County, LA 6 476 A Hale County, TX 5 477 A Ashley County, AR 7 478 A Bedford County, TN 6 480 B1 Montgomery County, VA 8 481 A Robeson County, NC 7 Nonself-representing Non-MSAs: West 482 A El Dorado-Alpine Counties, CA 6 484 A Carbon County, WY 5 Total Number of Segments 364 Third Stage Selection of Housing Units: 1996 NES Panel Component For each area segment selected in the second sampling stage, a listing was made of all housing units located within the physical boundaries of the segment. For segments with a very large number of expected housing units, all housing units in a subselected part of the segment were listed. The final equal probability sample of housing units for the 1994 NES was systematically selected from the housing unit listings for the sampled area segments. The Cross-section component of the 1994 NES sample design was selected from the 1980 SRC National Sample to yield an equal probability sample of households. The distribution of the 1994 NES Cross-section sample is that required by the two-thirds design of the 1980 SRC National Sample. The overall probability of selection for 1994 NES Cross-section households was f=0.00001885 or 0.1885 in 10,000. The equal probability sample of households was achieved for the 1994 NES Cross-section design by using the standard multi-stage sampling technique of setting the sampling rate for selecting housing units within area segments to be inversely proportional to the PPS probabilities used to select the PSU and area segment.[8] The 1994 NES Panel consisted of all 1005 respondents for whom a complete interview was obtained in the 1992 NES Cross-section sample. Respondents in 1994 from both the 1994 Cross-section and the 1994 Panel comprise the 1996 NES Panel. Fourth Stage Respondent Selection: 1996 NES Panel Component Within each sampled 1994 NES Cross-section housing unit, the SRC interviewer prepared a complete listing of all eligible household members. Using an objective procedure described by Kish (1949)[9] a single respondent was then selected at random to be interviewed. Regardless of circumstances, no substitutions were permitted for the designated respondent. This technique had also been used in 1992 to select the original Panel respondents. In 1994 the same Panel respondent (R) was sought for interview as had been interviewed in 1992. The 1996 Panel consists of all 1994 NES respondents for whom a complete interview was obtained in the 1994 NES Combined Cross-section and Panel sample. 1795 interviewed respondents make up the 1996 NES Panel component. Selection Stages for the 1996 NES Cross-section Supplement: 1990 SRC National Sample Primary Stage Selection: 1996 NES Cross-section Supplement The selection of primary stage sampling units (PSUs) for the 1990 SRC National Sample, which depending on the sample stratum are either MSAs, New England County Metropolitan Areas (NECMAs), single counties, independent cities, county equivalents or groupings of small counties, is based on the county-level 1990 Census Reports of Population and Housing.[10] Primary stage units were assigned to 108 explicit strata based on MSA/NECMA or non-MSA/NECMA status, PSU size, Census Region and geographic location within region. Twenty-eight of the 108 strata contain only a single self-representing PSU, each of which is included with certainty in the primary stage of sample selection. The remaining 80 nonself-representing strata contain more than one PSU. From each of these nonself-representing strata, one PSU was sampled with probability proportionate to its size (PPS) measured in 1990 occupied housing units. The full 1990 SRC National Sample of 108 primary stage selections was designed to be optimal for surveys roughly three to five times the size of the 1996 NES. To permit the flexibility needed for optimal design of smaller survey samples, the primary stage of the SRC National Sample can be readily partitioned into smaller subsamples of PSUs such as a one-half sample or a three-quarter sample partition. Each of the partitions represents a stratified subselection from the full 108 (representing the coterminous United States as does the NES study) PSU design. The one-half sample partition of the 1990 National Sample was designed to be roughly comparable in number of PSUs to the two-thirds partition of the 1980 National Sample. The one-half partition of the 1990 National Sample (i.e., the "A" primary sampling units or PSUs) includes 18 of the 28 self-representing MSA PSUs and a stratified subsampling of 40 of the 80 nonself-representing PSUs of the SRC National Sample. The remaining PSUs are divided in half and designated as either B1 or B2. The three-quarter partition includes all of the "A" PSUs plus "B1" PSUs, i.e., five additional self-representing PSUs and twenty additional nonself-representing PSUs. Since the 1996 NES desired comparison of data over time from 1994 NES respondents, as well as a supplement of eligible 1996 respondents, the 1996 NES sample design includes both a Panel and a Cross-section component. The Panel component of the 1996 NES design consists of all respondents from the both the Panel and the Cross-section components of the 1994 NES sample.[11] The 1996 NES Cross-section supplement component is a new selection of respondents from an area probability sample of households taken from the one-half partition of the new 1990 SRC National Sample. Since emphasis in the 1996 NES Study was to be on the Panel component and a rather small number of 1996 NES Cross-section respondents was sought, a subselection was made from the non-self representing PSUs in the 1990 half-sample partition; seven nonself-representing MSA PSUs and seven non-MSA PSUs were randomly eliminated. Table 2 identifies the 44 PSUs in the 1996 NES Cross-section supplement by MSA status and Region and also indicates the number of area segments used for the 1996 NES Cross-section supplement (see next section on second stage selection). Second Stage Selection of Area Segments: 1996 NES Cross-section Supplement The second stage of the 1990 SRC National Sample, used for the 1996 NES Cross-section supplement, was selected directly from computerized files that were extracted for the selected PSUs from the 1990 U.S. Census summary file series STF1-B. These files (on CD Rom) contain the 1990 Census total population and housing unit (HU) data at the census block level. The designated second-stage sampling units (SSUs), termed "area segments", are comprised of census blocks in both the metropolitan (MSA) primary areas and in the rural areas of non-MSA primary areas. Each SSU block or block combination was assigned a measure of size equal to the total 1990 occupied housing unit count for the area; SSU block(s) were assigned a minimum measure of 72 1990 total HUs per MSA SSU and a minimum measure of 48 total HUs per non-MSA SSU. Second stage sampling of area segments was performed with probabilities proportionate to the assigned measures of size (PPS). Prior to the second-stage selection, the SSUs were ordered or implicitly stratified within each selected PSU. Block Groups were stratified by household income and, within these income groups, by geography (county, tract, and block). Counties within MSA PSUs having more than one county were ordered by size and distance from the central city of the MSA. (For details, refer to the SRC publication, 1990 National Sample: Design and Development.) For the 1996 NES Cross-section supplement the number of area segments used in each PSU varies. In the self-representing (SR) PSUs the number of area segments varies in proportion to the size of the primary stage unit, from a high of 13 area segments in the self-representing New York MSA and 12 area segments in Los Angeles MSA, to a low of 4 area segments in the smaller self-representing PSUs such as Cleveland, Miami-Hialeah or Nassau-Suffolk MSAs. All nonself-representing (NSR) PSUs were represented by 4 area segments each. A total of 210 NES Cross-section area segments were selected, 106 in the 18 self-representing PSUs and 104 in the nonself-representing PSUs as shown in Table 2. Table 2: PSU Name and Number of Area Segments in the 1996 NES Cross-section Supplement Showing 1990 SRC National-Sample Stratum, Partition, and MSA Status National Sample National Sample # of 1996 NES PSU Number and PSU Name Panel Segments Partition Eight Largest Self-representing PSUs 120 A New York, NY MSA 13 190 A Los Angeles-Long Beach, CA MSA 12 130 A Chicago, IL MSA 9 121 A Philadelphia, PA-NJ MSA 7 131 A Detroit, MI MSA 6 150 A Washington DC-MD-VA MSA 6 110 A Boston, MA NECMA 6 171 A Dallas and Ft Worth, TX CMSA 6 Ten Remaining Self-representing PSUs 170 A Houston, TX MSA 5 191 A Seattle-Tacoma, WA CMSA 4 141 A St Louis, MO-IL MSA 4 152 A Baltimore, MD MSA 4 122 A Nassau-Suffolk, NY MSA 4 194 A Anaheim-Santa Ana, CA MSA 4 132 A Cleveland, OH MSA 4 154 A Miami-Hialeah, FL MSA 4 181 A Denver, CO MSA 4 196 A San Francisco, CA MSA 4 Nonself-representing MSAs: Northeast 211 A New Haven-Waterbury-Meriden, CT NECMA 4 213 A Manchester-Nashua NH NECMA 4 220 A Buffalo, NY MSA 4 226 A Atlantic City, NJ MSA 4 Nonself-representing MSAs: Midwest 230 A Milwaukee, WI MSA 4 236 A Madison, WI MSA 4 239 A Steubenville-Wheeling, OH[12] 4 240 A Des Moines, IA MSA 4 Nonself-representing MSAs: South 250 A Richmond-Petersburg, VA MSA 4 255 A Columbus, GA-AL MSA 4 257 A Jacksonville, FL MSA 4 258 A Lakeland, FL MSA 4 260 A Knoxville TN MSA 4 262 A Birmingham, AL MSA 4 273 B1[13] Waco, TX MSA 4 274 A McAllen-Edinburg-Mission, TX MSA 4 Nonself-representing MSAs: West 280 A Salt Lake City-Ogden etc, UT MSA 4 292 A Fresno, CA MSA 4 293 A Eugene-Springfield, OR MSA 4 Nonself-representing Non-MSAs: Northeast 320 A Elk County, PA 4 Nonself-representing Non-MSAs: Midwest 332 A Switzerland County, IN 4 342 A Taney County, MO 4 Nonself-representing Non-MSAs: South 351 A Harrisonburg IC, VA 4 354 A Wheatfield County, GA 4 370 B1 Jim Wells County, TX 4 Nonself-representing Non-MSAs: West 381 A Sandoval County, NM 4 Total Number of Segments 210 Third Stage Selection of Housing Units: 1996 NES Cross-section Supplement For each area segment selected in the second sampling stage, a listing was made of all housing units located within the physical boundaries of the segment. For segments with a very large number of expected housing units, all housing units in a subselected part of the segment were listed. The final equal probability sample of housing units for the 1996 NES Cross-section supplement was systematically selected from the housing unit listings for the sampled area segments. The Cross-section supplement of the 1996 NES sample design was selected from the 1990 SRC National Sample to yield an equal probability sample of 803 listed housing units. The 1996 NES Cross-section supplement drawn was ten percent larger than the expected required sample size of 730 lines to allow for additional "reserve" sample replicates to be released if necessary to meet interview goals. The overall probability of selection for 1996 NES Cross-section households was f=0.000007500 or 0.07500 in 10,000. The equal probability sample of households was achieved for the 1996 NES Cross-section supplement by using the standard multi-stage sampling technique of setting the sampling rate for selecting housing units within area segments to be inversely proportional to the PPS probabilities used to select the PSU and area segment.[14] Fourth Stage Respondent Selection: 1996 NES Cross-section Supplement Within each sampled 1996 NES Cross-section housing unit, the SRC interviewer prepared a complete listing of all eligible household members. Using an objective procedure described by Kish (1949)[15] a single respondent was then selected at random to be interviewed. Regardless of circumstances, no substitutions were permitted for the designated respondent. This technique had also been used in 1992 and 1994 to select the original Panel respondents. In 1996 the same Panel respondent (R) was sought for interview as had been interviewed in 1992 and 1994. 1996 NES SAMPLE DESIGN SPECIFICATIONS The 1996 Pre/Post-election Study sought a total of 1750 interviews in the Pre-election phase, all of which were to be contacted for reinterview in the Post-election phase. THE PRE-ELECTION PHASE: The 1996 NES sample design included both Panel and Cross-section components for the Pre-election phase, but emphasis in the 1996 NES design was on obtaining a maximum number of Panel interviews. To this end, the 1996 NES Panel component included the full set of 1795 1994 NES respondents, 1036 from the 1994 NES Cross-section component and 759 from the 1994 NES Panel component. Given sample design assumptions for the 1996 NES Panel of an eligibility rate of 0.98 and response rate of 0.75, this component was expected to yield 1320 interviews in 1996. The 1996 NES Cross-section supplement was intended to yield 430 interviews. It was estimated that this would require a NES Cross-section sample draw of 730 housing units. This assumed an occupancy/growth rate of 0.86, an eligibility rate of 0.95 and a response rate of 0.72. The overall 1996 NES Pre-election sample Design is set out in Table 3, below. Table 3: Sample Design Specifications and Assumptions 1996 Pre/Post-election Survey Cross-section Panel Component Total Component Completed Interviews 430 1320 1750 Response Rate 0.72 0.75 Eligible Sample Households 597 1760 2357 Eligibility Rate 0.95 NA Panel Recontact Rate NA 0.98 Occupied Households 628 1795 2423 Occupancy/growth Rate 0.86 1.0 Total Sample Lines 730 1795 2525 Sample Design, and Assignment of Replicates The Cross-section supplement of the 1996 NES sample was drawn from the recently listed "A" or half-sample partition of the 1990 SRC National Sample. Because of the small size of this NES sample component, both the number of PSUs (selected primary areas) and the Secondary Selection Units (area segments) in the National half-sample were reduced by subselection for the 1996 NES sample design.[16] The 18 self-representing areas in the 1990 SRC National half-sample were all retained for the Cross-section supplement (8 of these remained self-representing in the half-sample and 10 represent not only their own MSA but their "pair" among the twenty additional self-representing primary areas of the full 1990 SRC National Sample design). Nineteen of the 26 non-selfrepresenting MSAs and 7 of the 14 non-MSAs were retained for the 1996 NES Cross-section supplement (or 26 of 40 NSR PSUs). The number of second stage units (SSUs or area segments) was also reduced for the 1996 NES Cross-section supplement. In self-representing PSUs, the number of segments was reduced by one-half with a minimum of four segments in any PSU. In the nonself-representing PSUs, the number of segments was reduced to two-thirds, from six to four segments per PSU. This resulted in a total of 210 segments or SSUs from which the 1996 NES Cross-section supplement was selected. There could be no reduction of the total number of segments or of persons in the 1996 NES Panel component since all 1994 NES respondents were to be recontacted for interview in 1996. The number of area segments represented by the 1795 respondents to the 1994 study eligible for the 1996 NES Panel was 364. Both the 1996 NES Cross-section supplement and the 1996 NES Panel were divided by segment into two replicate samples. Replicates 1 and 2 of the 1996 NES Cross-section supplement each included 105 segments. The original replicate assignment of Panel segments also resulted in an even division of those segments by replicate. 1996 NES Cross-section Supplement Selection and Assignment of Releases The 1996 NES Cross-section supplement drawn was ten percent larger than the expected required sample size of 730 listed housing units to allow for additional "reserve" sample replicates. Final number of housing units in the Cross-section supplement was 803 spread over the 210 area segments as outlined below. Selected lines in each of the two replicates were divided into two equal parts to accommodate 4 quarterly releases. The quarterly releases were designed to assess effect on voter opinion formation of news events which occurred at various times over the course of the study. The first replicate sample was divided into release 1 and 2; the second replicate sample into release 3 and 4. An additional two reserve releases (5 and 6) equal to 73 lines, or 10% of the total 1996 NES Cross-section supplement, were also drawn from Replicate 2 to be released with releases 3 and 4, if necessary to meet study interview goals. Both reserve releases 5 and 6 were, in fact, released. Although Replicates 1 and 2 are each made up of different area segments (except as modified by the request to include Panel Rs needing tracking in Releases 1 and 2), all 1996 NES Cross-section and Panel Primary Areas are included in each Replicate if they contained more than a single segment. In contrast to the assignment of replicates by area segment, releases were originally specified in the 1996 NES sample design to be assigned across the HU-level file, rather than by area segment so any segment having more than one selection will have the selections distributed across Releases 1 and 2 (or 3, 4, 5 and 6 for Replicate 2 segments). In order to increase the efficiency of the field interviewing effort, original releases 3 and 4 were later revised such that their assignment was based on area segment, rather than across all Replicate 2 segments. 1996 NES PRE-ELECTION SAMPLE OUTCOME: Table 4: 1996 NES Pre-Election Sample Design Specifications and Assumptions Compared to Sample Outcome. 1996 Pre/Post-election Survey [17] Cross-section Panel Component Total Component Design Outcome Design Outcome Design Outcome Completed Interviews 430 398 1320 1316 1750 1714 Response Rate 0.72 0.60 0.75 0.76 Eligible Sample Households 597 666 1760 1741 2357 2407 Eligibility Rate 0.95 0.96 NA NA Panel Recontact Rate NA NA 0.98 0.98 Occupied Households 628 692 1795 1781 2423 2473 Occupancy/growth Rate 0.86 0.85 1.00 1.00 Total Sample Lines 730 817 1795 1788 2525 2605 A comparison of the total design figures compared to the Pre-election outcome figures in Table 4 indicates the following: for the 1996 NES Panel component, where there was no option for reserve releases, and where primary field effort was placed, eligibility and response rates equal to those anticipated resulted in a number of completed interviews very close to that projected by the sample design. On the other hand, for the 1996 NES Cross-section supplement, even with the release of reserve replicates, a lower than expected response rate resulted in a seven percent shortfall in number of completed interviews. Since the Cross-section supplement made up less than one-quarter of the total sample design, the overall shortfall in number of completed interviews was only two percent. THE POST-ELECTION PHASE: The study design for the 1996 Post-election component of the NES Study called for recontact of all respondents to the 1996 NES Pre-election survey (both those originally in the Panel component and those in the Cross-section supplement.) The Post-election phase of the 1996 NES included a mode experiment which called for the random assignment, by area segment, of the majority of these respondents, to be recontacted after the election for an interview either by phone or in person. Those to be excluded from this mode experiment were those respondents either 1) who were interviewed by phone during the Pre-election study or 2) who were known to not have a phone. The assignment to either the phone or the in-person mode was made on the basis of segment, such that approximately half of the Post-election recontacts made by phone and the other half in person. Since the Post-election phase of the study involved no new respondents--all respondents were considered Panel respondents for this phase. A combined recontact and response rate of 85% was assumed for the Post-election phase of the 1996 NES to yield a total of 1460 interviews. Of the total of 1714 interviews completed for the 1996 Pre-election study, the sample released for Post-election recontact was distributed as shown in Table 5. Post-election interview outcome is also shown on this table. The combined recontact and response rate exceeded expectations resulting in a total number of Post-election interviews over the 1460 goal. Table 5. Post-election Mode Distribution and Interview Outcome for 1996 NES.[18] Mode # Released NI NIP Refusal Interviews Recontact/ Response Rate Face-to-Face: 875 35 23 42 774 0.89 Include in Experiment 742 22 17 34 668 0.90 Exclude from Experiment 133 13 6 8 106 0.80 Telephone: 839 25 17 37 760 0.90 Include in Experiment 759 21 16 33 689 0.91 Exclude from Experiment 80 4 1 4 71 0.89 Total 1714 60 40 79 1534 0.90 WEIGHTED ANALYSIS OF 1996 NES DATA The 1996 NES data set includes two final person-level analysis weights which incorporate sampling, nonresponse and post-stratification factors. One weight (variable #4) is for longitudinal micro-level analysis using the 1996 NES Panel. The other weight (variable #3) is for analysis of the 1996 NES combined sample (Panel component cases plus Cross-section supplement cases). In addition, a Time Series Weight (variable #5) which corrects for Panel attrition was constructed. This weight should be used in analyses which compare the 1996 NES to earlier unweighted National Election Study data collections. Analysts interested in developing their own nonresponse or post-stratification adjustment factors must request access to the necessary sample control data from the NES Board. CONSTRUCTION OF ANALYSIS WEIGHTS Sample Selection Weight The area probability sample design for the 1996 NES results in an equal probability sample of U.S. households. However, within sample households a single adult respondent is chosen at random to be interviewed. Since the number of eligible adults may vary from one household to another, the random selection of a single adult introduces inequality into respondents' selection probabilities. In analysis, a respondent selection weight should be used to compensate for these unequal selection probabilities. The value of the respondent selection weight is exactly equal to the number of eligible adults in the household from which the random respondent was selected. The use of the respondent selection weight is strongly encouraged, despite past evaluations which have shown these weights to have little significant impact on the values of NES estimates of descriptive statistics. Household Nonresponse Adjustment Factor Nonresponse adjustment factors were constructed at the household level separately for Panel and Cross-section component cases. Nonresponse adjustment cells for the relatively small 1996 NES Cross-section supplement were formed by crossing PSU type (Self-representing, Nonself-representing MSA or non-MSA) by the four Census regions (Northeast, Midwest, South, and West). A nonresponse factor equal to the inverse of the response rate in each cell was applied to the interview cases. For the larger number of Panel cases, 1996 nonresponse adjustment cells were initially formed by crossing PSU type by the nine Census divisions (New England, Middle Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain and Pacific). However, in order to have a minimum of approximately 25 cases in each nonresponse adjustment cell, some cells were collapsed across Census Divisions in the same Census Region. Tables 6 and 7 show the 1996 nonresponse adjustment factors for the Cross-section supplement and for the Panel respectively. The 1996 NES Panel nonresponse prior to 1996 was reflected in the 1994 full sample weight which was used to construct 1996 NES Panel final sample weights. Table 6 Computation of Nonresponse Adjustment Weights -- 1996 NES Cross Section Supplement Nonresponse PSU Type Census Region Response Adjustment Rate (%) Weight SR-MSA Northeast 42.31 2.364 Midwest 53.33 1.875 South 53.85 1.857 West 50.70 1.972 NSR-MSA Northeast 52.63 1.900 Midwest 67.80 1.475 South 64.55 1.549 West 62.50 1.600 NSR-non MSA Northeast 60.00 1.667 Midwest 72.09 1.387 South 68.67 1.456 West 80.95 1.235 Table 7 Computation of Nonresponse Adjustment Weights -- 1996 NES Panel Component Nonresponse PSU Type Census Division Response Adjustment Rate (%) Weight SR-MSA New England & Middle Atlantic 72.90 1.372 East North 72.50 1.379 Central West North 86.05 1.162 Central South Atlantic 77.91 1.284 West South 63.64 1.571 Central Pacific 65.85 1.519 NSR-MSA New England & Middle Atlantic 71.96 1.390 East North 76.03 1.315 Central West North 70.77 1.413 Central South Atlantic 76.71 1.304 East South 64.71 1.545 Central West South 70.59 1.417 Central Mountain 76.98 1.299 Pacific 76.67 1.304 NSR-non MSA New England & 81.82 1.222 Middle Atlantic East North 84.62 1.182 Central West North 72.73 1.375 Central South Atlantic 84.96 1.177 East South 76.53 1.307 Central & West South Central Mountain & 70.73 1.414 Pacific 1996 Combined NES Post-stratification Factor As a first step in post-stratifying the sample to 1990 Census proportions, an intermediate weight for the 1996 NES combined sample (Cross-section plus Panel cases) was constructed as follows. First an intermediate weight for Cross-section supplement cases was constructed by multiplying the 1996 Cross-section nonresponse adjustment (Table 6) by the number of eligible persons in the sample household[19] by an inflation factor which is the 1995 estimated U.S. households divided by the number of eligible households (97,061,000/661). This initial weight was used to produce a weighted sex by age group by Census Region table for the 1996 NES Cross-section supplement. The age categories used were: 18-44 years, 45-64 years, and 65+ years. Post-stratification factors were constructed to match the sample proportions in the 24 sex by age by Region cells to the July 1995 Census population projections (Current Population Reports, P25-1111, Table 4) by dividing the Census total by the weighted sample estimate for each post- stratification cell. Because of the small number of Cross-section supplement cases, it is not intended that Cross-section only analysis be undertaken. An intermediate weight factor for the 1996 NES Panel cases was similarly constructed by multiplying the 1996 nonresponse adjustment (Table 7) by the 1994 full sample weight times the reciprocal of the constant used to center the 1994 weights (1993 estimated U.S. population 18 or more years of age / number of 1994 respondents).[20] For the 1996 NES Panel respondents, the number of eligible persons in the household and nonresponse prior to 1996 was reflected in the 1994 full sample weight. The last element in this computation was necessary to restore the Panel intermediate weight to its full representation of the population. This intermediate weight was used for Panel cases to produce a weighted sex by age group by Census Region table as described above. Again, post-stratification weights were constructed to match the sample proportions in the 24 sex by age group by Census Region cells to the July 1995 Census population projections. 1996 NES Panel Post-Stratification Factor For 1996 NES combined Panel and Cross-section analysis, the proportion of respondents contributed to the total sample was adjusted for by multiplying the Panel case intermediate weight by the proportion of Panel cases (1316/1714) and multiplying the Cross-section case intermediate weight by the proportion of Cross-section cases (398/1714). Thus a combined Cross-section and Panel post-stratification weight was produced, by dividing the 1995 Census estimated totals in the 24 sex by age group by Census Region cells by the corresponding weighted estimates for the combined sample. The figures for this combined post-stratification factor are shown in Table 8. It is these figures, centered as explained below, which are used for the final 1996 combined sample weight (V3). The final analysis weight (V4 ) for longitudinal analysis of the 1996 NES Panel is the product of the 1994 full sample weight, the 1996 Panel household nonresponse adjustment factor, and the Panel post-stratification factor. FINAL ANALYSIS WEIGHTS The final analysis weights are the product of the household level non-response adjustment factor, the number of eligible persons, the sample selection (inflation) weight and the post-stratification factor. The final analysis weight for the Panel-only analysis (V4) is centered so that the sum of the weights is equal to the total number of Panel respondents, 1316. The final analysis weights for the combined 1996 NES sample (V3) sums to 1714, the total number of respondents. These weights were constructed using the 1996 NES Pre-election data set. The nonresponse and attrition between the Pre and Post-election studies are not incorporated. Table 8: 1996 NES Combined (Cross-section and Panel) Sample Post-Stratification Factor Census Age Census Est. 1996 NES Post- Sex Region Group July 1, 1995 Weighted[21] Stratification Factor Male Northeast 18-44 10,440,000 9,885,067 1.056 45-64 5,019,000 5,329,059 0.942 65+ 2,892,000 3,152,420 0.917 Midwest 18-44 12,645,000 10,248,770 1.234 45-64 5,870,000 7,553,155 0.777 65+ 3,310,000 3,215,352 1.029 South 18-44 18,919,000 15,799,320 1.197 45-64 8,691,000 8,455,024 1.028 65+ 4,789,000 5,216,866 0.918 West 18-44 12,778,000 9,478,170 1.348 45-64 5,298,000 5,349,446 0.990 65+ 2,708,000 2,347,394 1.154 Female Northeast 18-44 10,630,000 8,990,888 1.182 45-64 5,503,000 5,895,540 0.933 65+ 4,378,000 3,556,867 1.231 Midwest 18-44 12,749,000 11,606,790 1.098 45-64 6,234,000 6,622,310 0.941 65+ 4,871,000 4,952,220 0.984 South 18-44 19,077,000 20,443,010 0.933 45-64 9,397,000 9,362,888 1.004 65+ 7,016,000 6,738,762 1.041 West 18-44 12,169,000 11,691,630 1.041 45-64 5,454,000 5,937,677 0.919 65+ 3,686,000 3,664,183 1.006 Totals 194,523,000 185,492,800 CONSTRUCTION OF TIME SERIES WEIGHT The 1996 NES Panel consists of 759 respondents originally selected for the 1992 NES Pre-election Study (1994 NES Panel) and 1036 respondents originally selected for the 1994 NES Study (1994 NES Cross-section). All of the 1005 1992 Post-election respondents were eligible for the 1994 NES Panel and 759 of these responded in 1994 and remained eligible for the 1996 NES Panel. Of these 759 respondents from the 1992 NES (1994 Panel), 597 were interviewed for the 1996 NES. Of the 1036 respondents from the 1994 Cross-section, 719 were interviewed in 1996 for an overall 1996 NES Panel response rate of 1316/1795 or 0.733.[22] Table 9: Time Series Weight Factors Years of Education Level Age Group Time Series Residence Weight Factor < 3 < HS Graduate 18-24 1.168 25-39 1.087 40-64 1.284 65 + 1.073 HS Graduate 17-24 1.169 25-39 1.060 40-64 0.897 65 + 1.748 > HS Graduate 17-24 0.958 25-39 0.978 40-64 0.950 65 + 0.791 3+ < HS Grad 17-39 1.205 40-64 0.917 65-74 1.018 75+ 1.605 HS Graduate 17-24 1.171 25-39 1.172 40-64 0.990 65-74 1.010 75+ 0.960 > HS Graduate 17-24 1.236 25-39 0.931 40-64 0.908 65-74 0.761 75+ 1.057 PROCEDURES FOR SAMPLING ERROR ESTIMATION The 1996 NES sample design is based on a stratified multi-stage area probability sample of United States households. Although smaller in scale, the NES sample design is very similar in it basic structure to the multi-stage designs used for major federal survey programs such as the Health Interview Survey (HIS) or the Current Population Survey (CPS). The survey literature refers to the NES, HIS and CPS samples as complex designs, a loosely-used term meant to denote the fact that the sample incorporates special design features such as stratification, clustering and differential selection probabilities (i.e., weighting) that analysts must consider in computing sampling errors for sample estimates of descriptive statistics and model parameters. This section of the 1996 NES sample design description focuses on sampling error estimation and construction of confidence intervals for survey estimates of descriptive statistics such as means, proportions, ratios, and coefficients for linear and logistic linear regression models. Standard analysis software systems such SAS, SPSS, OSIRIS assume simple random sampling (SRS) or equivalently independence of observations in computing standard errors for sample estimates. In general, the SRS assumption results in underestimation of variances of survey estimates of descriptive statistics and model parameters. Confidence intervals based on computed variances that assume independence of observations will be biased (generally too narrow) and design-based inferences will be affected accordingly. Sampling Error Computation Methods and Programs Over the past 50 years, advances in survey sampling theory have guided the development of a number of methods for correctly estimating variances from complex sample data sets. A number of sampling error programs which implement these complex sample variance estimation methods are available to NES data analysts. The two most common approaches to the estimation of sampling error for complex sample data are through the use of a Taylor Series Linearization of the estimator (and corresponding approximation to its variance) or through the use of resampling variance estimation procedures such as Balanced Repeated Replication (BRR) or Jackknife Repeated Replication(JRR). New Bootstrap methods for variance estimation can also be included among the resampling approaches. See Rao and Wu (1988). 1. Linearization Approach If data are collected using a complex sample design with unequal size clusters, most statistics of interest will not be simple linear functions of the observed data. The objective of the linearization approach is to apply Taylor's method to derive an approximate form of the estimator that is linear in statistics for which variances and covariances can be directly estimated. (Kish, 1965; Woodruff, 1971). Linearized variance approximations are derived for estimators of ratio means (Kish and Hess, 1959); finite population regression coefficients and correlation coefficients (Kish and Frankel, 1974); and many other non-linear statistics. Software packages such as SUDAAN and PC CARP (see below) use the Taylor Series linearization method to estimate standard errors for the coefficients of logistic regression models. In these programs, an iteratively reweighted least squares algorithm is used to compute maximum likelihood estimates of model parameters. At each step of the model fitting algorithm, a Taylor Series linearization approach is used to compute the variance/covariance matrix for the current iteration's parameter estimates (Binder, 1983). Available sampling error computation software that utilizes the Taylor Series linearization method includes: STATA, SUDAAN and PC SUDAAN, SUPERCARP AND PC CARP, and CLUSTERS. PC SUDAAN, PC CARP and STATA include procedures for estimation of sampling error both for descriptive statistics such as means, proportions, totals and for parameters of commonly used multivariate models (least squares regression, logistic regression). 2. Resampling Approaches In the mid-1940's, P.C. Mahalanobis (1946) outlined a simple replicated procedure for selecting probability samples that permits simple, unbiased estimation of variances. The practical difficulty with the simple replicated approach to design and variance estimation is that many replicates are needed to achieve stability of the variance estimator. Unfortunately, a design with many independent replicates must utilize a coarser stratification than alternative designs--to achieve stable variance estimates, sample precision must be sacrificed. Balanced Repeated Replication (BRR), Jackknife Repeated Replication (JRR) and the Bootstrap are alternative replication techniques that may be used for estimating sampling errors for statistics based on complex sample data. The BRR method is applicable to stratified designs in which two half-sample units (i.e., PSUs) are selected from each design stratum. The conventional "two PSU-per-stratum" design in the best theoretical example of such a design although in practice, collapsing of strata (Kalton, 1977) and random combination of units within strata are employed to restructure a sample design for BRR variance estimation. The half-sample codes prepared for the 1994 NES data set require the collapsing of nonself-representing strata and the randomized combination of selection units within self-representing (SR) strata. When full balancing of the half-sample assignments is employed (Wolter, 1985), BRR is the most computationally efficient of the replicated variance estimation techniques. The number of general purpose BRR sampling error estimation programs in the public domain is limited. The OSIRIS REPERR program includes the option for BRR estimation of sampling errors for least squares regression coefficients and correlation statistics. Westat, Inc. has developed the Westvar PC for BRR estimation of standard errors. Another option is to use SAS or SPSS Macro facilities to implement the relatively simple BRR algorithm. The necessary computation formulas and Hadamard matrices to define the half-sample replicates are available in Wolter (1985). With improvements in computational flexibility and speed, jackknife (JRR) and bootstrap methods for sampling error estimation and inference have become more common (Rao and Wu, 1988 ). Few general purpose programs for jackknife estimation of variances are available to analysts. OSIRIS REPERR has a JRR module for estimation of standard errors for regression and correlation statistics. Other stand alone programs may also be available in the general survey research community. Like BRR, the algorithm for JRR is relatively easy to program using SAS, SPSS or S-Plus macro facilities. BRR and JRR are variance estimation techniques, each designed to minimize the number of "resamplings" needed to compute the variance estimate. In theory, the bootstrap is not simply a tool for variance estimation but an approach to actual inference for statistics. In practice, the bootstrap is implemented by resampling (with replacement) from the observed sample units. To ensure that the full complexity of the design is reflected , the selection of each bootstrap reflects the full complexity of the stratification, clustering and weighting that is present in the original sample design. A large number of bootstrap samples are selected and the statistic of interest is computed for each. The empirical distribution of the estimate that results from the large set of bootstrap samples can then be used to a variance estimate and a support interval for inference about the population statistic of interest. In most practical survey analysis problems, the JRR and Bootstrap methods should yield similar results. Most survey analysts should choose JRR due to its computational efficiency. NES data analysts interested in the bootstrap technique are referred to LePage and Billard (1992) for additional reading and a bibliography for the general literature on this topic. One aspect of BRR, JRR and bootstrap variance estimation that is often pushed aside in practice is the treatment of analysis weights. In theory, when a resampling occurs (i.e., a BRR half sample is formed), the analysis weights should be recomputed based only on the selection probabilities, nonresponse characteristics and post-stratification outcomes for the units included in the resample. This is the correct way of performing resampling variance estimation; however, in practice acceptable estimates can be obtained through use of the weights as they are provided on the public use data set. Sampling Error Computation Models Regardless of whether linearization or a resampling approach is used, estimation of variances for complex sample survey estimates requires the specification of a sampling error computation model. NES data analysts who are interested in performing sampling error computations should be aware that the estimation programs identified in the preceding section assume a specific sampling error computation model and will require special sampling error codes. Individual records in the analysis data set must be assigned sampling error codes which identify to the programs the complex structure of the sample (stratification, clustering) and are compatible with the computation algorithms of the various programs. To facilitate the computation of sampling error for statistics based on 1996 NES data, design-specific sampling error codes will be routinely included in all public-use versions of the data set. Although minor recoding may be required to conform to the input requirements of the individual programs, the sampling error codes that are provided should enable analysts to conduct either Taylor Series or Replicated estimation of sampling errors for survey statistics. Table 10 defines the sampling error coding system for 1996 NES sample cases. Two sampling error code variables are defined for each case based on the sample design primary stage unit (PSU) and area segment in which the sample household is located. Sampling Error Stratum Code (Variable #2125). The Sampling Error Computation Stratum Code is the variable which defines the sampling error computation strata for all sampling error analysis of the NES data. With the exception of the New York, Los Angeles and Chicago MSAs, each self-representing (SR) design stratum is represented by one sampling error computation stratum. Due to their population size, two sampling error computation strata are defined for each of the three largest MSAs. Pairs of similar nonself-representing (NSR) primary stage design strata are "collapsed" (Kalton, 1977) to create NSR sampling error computation strata. For both the 1980 and 1990 SRC National Sample design controlled selection and a "one-per-stratum" PSU allocation are used to select the primary stage of the 1996 NES national sample. The purpose in using controlled selection and the "one-per-stratum" sample allocation is to reduce the between-PSU component of sampling variation relative to a"two-per-stratum" primary stage design. Despite the expected improvement in sample precision, a drawback of the "one-per-stratum" design is that two or more sample selection strata must be collapsed or combined to form a sampling error computation stratum. Variances are then estimated under the assumption that a multiple PSU per stratum design was actually used for primary stage selection. The expected consequence of collapsing design strata into sampling error computation strata is the overestimation of the true sampling error; that is, the sampling error computation model defined by the codes contained in Table 14 will yield estimates of sampling errors which in expectation will be slightly greater than the true sampling error of the statistic of interest. SECU - Stratum-specific Sampling Error Computation Unit code (Variable #2126) is a half sample code for analysis of sampling error using the BRR method or approximate "two-per-stratum" Taylor Series method (Kish and Hess, 1959). Within the SR sampling error strata, the SECU half sample units are created by dividing sample cases into random halves, SECU=1 and SECU=2. The assignment of cases to half-samples is designed to preserve the stratification and second stage clustering properties of the sample within an SR stratum. Sample cases are assigned to SECU half samples based on the area segment in which they were selected. For this assignment, sample cases were placed in original stratification order (area segment number order) and beginning with a random start entire area segment clusters were systematically assigned to either SECU=1 or SECU=2. In the general case of nonself-representing (NSR) strata, the half sample units are defined according to the PSU to which the respondent was assigned at sample selection. That is, the half samples for each NSR sampling error computation stratum bear a one-to-one correspondence to the sample design NSR PSUs. The particular sample coding provided on the NES public use data set is consistent with the "ultimate cluster" approach to complex sample variance estimation (Kish, 1965; Kalton, 1977). Individual stratum, PSU and segment code variables may be needed by NES analysts interested in components of variance analysis or estimation of hierarchical models in which PSU-level and neighborhood-level effects are explicitly estimated. Table 10 shows the sampling error stratum and SECU codes to be used for the paired selection model for sampling error computations for any 1996 NES analyses; the same codes can be used when using the 1996 NES combined Cross-section/Panel data or when using 1996 NES Panel data separately. The first 42 strata reflect the two-thirds 1980 National Sample design used in 1994 and apply to the 1996 NES Panel. Strata 51 through 89 reflect the half sample 1990 National Sample design used for the 1996 NES Cross-section supplement. It can be seen from this table that the three-digit 1996 SE code is comprised of: first, the two-digit SE Stratum code followed by the one-digit SECU code. Table 10: 1996 National Election Study Sampling Error Codes SE SEC SE PSU Segment #s Total Stratum U Code Panel Respondents (1992,1994) (In 1996) 01 1 011 501 103 119 135 8 2 012 501 107 123 139 3 02 1 021 501 111 127 143 13 2 022 501 115 131 148 8 03 1 031 502 110 123 136 4 2 032 502 101 114 4 04 1 041 502 117 129 4 2 042 502 107 120 133 5 05 1 051 503 112 129 7 2 052 503 117 134 12 06 1 061 503 103 120 8 2 062 503 107 125 7 07 1 071 504 102 110 117 13 2 072 504 106 113 121 9 08 1 081 505 105 112 119 10 2 082 505 101 108 115 14 09 1 091 506 104 110 116 8 2 092 506 101 107 113 2 10 1 101 507 105 111 115 17 2 102 507 103 107 113 24 11 1 111 508 101 107 110 13 2 112 508 103 109 114 6 12 1 121 509 104 114 4 2 122 509 101 107 111 5 13 1 131 510 101 111 2 2 132 510 107 1 SE SEC SE PSU Segment #s Total Stratum U Code Panel Respondents (1992,1994) (In 1996) 14 1 141 511 105 111 6 2 142 511 102 108 8 15 1 151 512 102 3 2 152 512 105 111 4 16 1 161 513 101 107 2 2 162 513 104 110 5 17 1 171 514 104 110 4 2 172 514 101 107 2 18 1 181 515 105 111 15 2 182 515 102 108 15 19 1 191 516 102 108 10 2 192 516 105 111 10 20 1 201 517 103 105 13 107 109 111 2 202 518 101 103 105 28 107 109 111 21 1 211 521 103 105 107 12 109 111 2 212 523 103 105 107 13 109 111 22 1 221 524 102 104 106 11 108 110 112 2 222 534 102 104 106 18 108 110 112 23 1 231 526 101 103 105 19 107 109 111 2 232 527 101 103 105 13 109 111 24 1 241 528 102 104 106 30 108 110 112 2 242 529 102 104 106 16 108 110 112 25 1 251 531 102 104 106 29 108 110 112 2 252 532 102 104 106 18 108 110 112 26 1 261 533 102 104 106 14 108 110 112 2 262 547 101 103 105 12 107 109 111 27 1 271 536 101 103 105 14 107 109 111 2 272 539 101 103 105 17 107 109 111 28 1 281 540 101 103 105 11 107 109 111 2 282 542 102 104 106 31 108 110 112 29 1 291 543 102 104 106 29 108 110 112 2 292 545 103 105 107 42 109 111 30 1 301 544 101 103 105 18 107 109 111 2 302 476 001 004 006 9 007 012 31 1 311 549 101 103 105 18 107 109 111 2 312 550 101 103 105 24 107 109 111 32 1 321 553 102 104 106 15 108 110 112 2 322 555 101 103 105 30 107 109 111 33 1 331 556 101 105 107 18 109 111 2 332 557 102 104 106 33 108 110 112 34 1 341 558 102 104 106 24 108 110 112 2 342 559 101 103 105 25 107 109 111 35 1 351 560 104 108 112 44 2 352 560 102 106 110 23 36 1 361 463 001 002 003 005 14 007 008 009 011 2 362 464 001 002 004 005 31 008 009 010 012 37 1 371 465 001 005 22 007 009 011 2 372 466 001 002 004 005 44 008 010 011 012 38 1 381 468 001 002 006 23 007 008 011 012 2 382 470 002 003 005 25 007 011 012 39 1 391 473 001 005 006 008 31 009 011 012 2 392 474 001 002 004 007 20 008 011 40 1 401 477 001 003 005 006 26 007 010 012 2 402 478 002 005 006 20 008 010 012 41 1 411 480 002 005 006 007 44 008 010 011 012 2 412 481 001 004 005 007 21 008 009 011 42 1 421 482 002 004 005 18 007 009 012 2 422 484 001 004 009 11 011 012 1996 NES Cross-section Segments (from 1990 National Sample Frame): SE SEC SE PSU Segment #s Total Rs (1996) Stratum U Code (1996 Cross Section) 51 1 511 120 003, 019, 035, 051 4 067, 083, 099 2 512 120 011, 027, 043, 4 059, 075, 091 53 1 531 190 003, 019, 035, 4 196[23] 051, 067, 083 002, 014 2 532 011, 027, 043, 3 190 059, 075, 091 196[24] 010, 022 SE SEC SE PSU Segment #s Total Rs (1996) Stratum U Code (1996 Cross Section) 55 1 551 130 008, 024, 040, 4 056, 072 2 552 130 016, 032, 048, 3 064 57 1 571 121 006, 022, 038, 054 4 2 572 121 014, 030, 046 3 58 1 581 131 004, 020, 036 2 2 582 131 012, 028, 044 4 60 1 601 150 003, 019, 035 1 2 602 150 011, 027, 043 2 61 1 611 171 006, 022, 038 1 2 612 171 014, 030, 046 3 62 1 621 170 003, 019, 035 9 2 622 170 011, 027 5 63 1 631 110 008, 024, 040 2 2 632 110 016, 032, 048 3 64 1 641 122 004, 020 1 2 642 122 012, 028 1 65 1 651 141 008, 024 4 2 652 141 016, 032 4 66 1 661 132 001, 013 2 2 662 132 009, 021 1 67 1 671 152 008, 024 1 2 672 152 016, 032 4 68 1 681 154 003, 015 1 2 682 154 007, 019 1 69 1 691 194 004, 020 4 2 692 194 012, 028 3 70 1 701 191 005, 013, 021, 029 14 2 702 181 005, 009, 017, 021 8 71 1 711 220 005, 009, 017, 021 13 2 712 226 002, 006, 014, 018 9 72 1 721 211 003, 011, 015, 023 1 2 722 213 004, 008, 016, 020 7 73 1 731 230 002, 010, 014, 022 12 2 732 236 002, 010, 014, 022 12 76 1 761 239 001, 005, 013, 017 7 2 762 240 006, 010, 018, 022 9 77 1 771 262 002, 010, 014, 022 19 2 772 255 008, 012, 020, 024 10 78 1 781 257 004, 012, 016, 024 5 2 782 258 002, 006, 014, 018 12 79 1 791 273 003, 011, 015, 023 4 2 792 274 002, 006, 014, 018 5 81 1 811 260 003, 011, 015, 023 9 2 812 250 007, 011, 019, 023 7 84 1 841 292 001, 009, 013, 021 10 2 842 293 007, 011, 019, 023 10 85 1 851 280 002, 014 6 2 852 280 006, 018 4 86 1 861 320 006, 018 5 2 862 320 010, 022 7 87 1 871 332 004, 008, 016, 020 22 2 872 342 008, 012, 020, 024 9 88 1 881 351 001, 009, 013, 021 32 2 882 354 008, 012, 020, 024 13 89 1 891 370 005, 009, 017, 021 12 2 892 381 001, 005, 013, 017 17 Total: 1714 Generalized Sampling Error Results for the 1996 NES To assist NES analysts, the PC SUDAAN program was used to compute sampling errors for a wide-ranging example set of proportions estimated from the 1996 NES Pre-election Survey data set. For each estimate, sampling errors were computed for the total sample and for twenty demographic and political affiliation subclasses of the 1996 NES Pre-election Survey sample. The results of these sampling error computations were then summarized and translated into the general usage sampling error table provided in Table 11. The mean value of deft, the square root of the design effect, was found to be 1.346. The design effect was primarily due to weighting effects (Kish, 1965) and did not vary significantly by subclass size. Therefore the generalized variance table is produced by multiplying the simple random sampling standard error for each proportion and sample size by the average deft for the set of sampling error computations. Incorporating the pattern of "design effects" observed in the extensive set of example computations, Table 11 provides approximate standard errors for percentage estimates based on the 1996 NES. To use the table, examine the column heading to find the percentage value which best approximates the value of the estimated percentage that is of interest.[25] Next, locate the approximate sample size base (denominator for the proportion) in the left-hand row margin of the table. To find the approximate standard error of a percentage estimate, simply cross-reference the appropriate column (percentage) and row (sample size base). Note: the tabulated values represent approximately one standard error for the percentage estimate. To construct an approximate confidence interval, the analyst should apply the appropriate critical point from the "z" distribution (e.g., z=1.96 for a two-sided 95% confidence interval half-width). Furthermore, the approximate standard errors in the table apply only to single point estimates of percentages not to the difference between two percentage estimates. The generalized variance results presented in Table 11 are a useful tool for initial, cursory examination of the NES survey results. For more in depth analysis and reporting of critical estimates, analysts are encouraged to compute exact estimates of standard errors using the appropriate choice of a sampling error program and computation model. Table 11: Generalized Variance Table. 1996 NES Pre/Post-election Survey. APPROXIMATE STANDARD ERRORS FOR PERCENTAGES For percentage estimates near: Sample n 50% 40% 30% 20% 10% or 60% or 70% or 80% or 90% The approximate standard error of the percentage is: 100 6.730 6.594 6.168 5.384 4.038 200 4.759 4.663 4.362 3.807 2.855 300 3.886 3.807 3.561 3.108 2.331 400 3.365 3.297 3.084 2.692 2.019 500 3.010 2.949 2.758 2.408 1.806 750 2.475 2.408 2.252 1.966 1.474 1000 2.128 2.085 1.951 1.703 1.277 1250 1.904 1.865 1.745 1.523 1.142 1500 1.738 1.703 1.593 1.390 1.043 1714 1.626 1.593 1.490 1.300 0.975 References Binder, D.A. (1983), "On the variances of asymptotically normal estimators from complex surveys," International Statistical Review, Vol. 51, pp. 279-292. Kalton, G. (1977), "Practical methods for estimating survey sampling errors," Bulletin of the International Statistical Institute, Vol 47, 3, pp. 495-514. Kish, L. (1949). A procedure for objective respondent selection within the household, Journal of the American Statistical Association, Vol. 44, pp. 380-387. Kish, L. (1965), Survey Sampling. New York: John Wiley & Sons, Inc. Kish, L., & Frankel, M.R. (1974), "Inference from complex samples," Journal of the Royal Statistical Society, B, Vol. 36, pp. 1-37. Kish, L., & Hess, I. (1959), "On variances of ratios and their differences in multi-stage samples," Journal of the American Statistical Association, 54, pp. 416-446. LePage, R., & Billard, L. (1992), Exploring the Limits of Bootstrap. New York: John Wiley & Sons, Inc. Mahalanobis, P.C. (1946), "Recent experiments in statistical sampling at the Indian Statistical Institute," Journal of the Royal Statistical Society, Vol 109, pp. 325-378. Rao, J.N.K & Wu, C.F.J. (1988.), "Resampling inference with complex sample data," Journal of the American Statistical Association, 83, pp. 231-239. Rosenstone, Steven J., Kinder, Donald R., Miller, Warren E., & the National Election Studies Sample Design: Technical Memoranda, 1994 Election Study pp. 882-905 in Rosenstone, Steven J.,Kinder, Donald R., Miller, Warren E., & the National Election Studies, AMERICAN NATIONAL ELECTION STUDY, 1994: POST-ELECTION SURVEY (ENHANCED WITH 1992 AND 1993 DATA) (Computer file). Conducted by University of Michigan Center for Political Studies. 2nd ICPSR ed. Ann Arbor MI: University of Michigan, Center for Political Studies, and Inter-university Consortium for Political and Social Research (producer), 1995. Ann Arbor MI: Inter-university Consortium for Political and Social Research (distributor), 1995. Wolter, K.M. (1985 ). Introduction to Variance Estimation. New York: Springer-Verlag. Woodruff, R.S. (1971), "A simple method for approximating the variance of a complicated estimate," Journal of the American Statistical Association, Vol. 66, pp. 411-414. Footnotes 1 NECMAs are used in the 1996 NES Cross-section component only, which is drawn from the 1990 SRC National Sample. 2 The 730 listed housing units projected to be necessary to produce the 430 interviews from the 1996 NES Cross-section supplement were increased by 10% (73) for reserve releases. The 803 listed housing units selected for this component of the 1996 NES Sample actually yielded 666 eligible households within which an interview was attempted. 3 Further description of the 1994 sample design can be found in "Sample Design: Technical Memoranda, 1994 Election Study" pp. 882-905 in Steven J. Rosenstone, Donald R. Kinder, Warren E. Miller and the National Election Studies. AMERICAN NATIONAL ELECTION STUDY, 1994: POST-ELECTION SURVEY. 4 The 1994 NES Panel consisted of all 1005 Respondents from the 1992 NES Cross-section sample. Of these, 925 were recontacted in the 1993 NES Pilot Study (a follow-up of the 1992 NES survey), of which 750 were re-interviewed, 98 refused to be re-interviewed and 77 could not be re-interviewed at that time due to some 'permanent' condition. 80 of the 1005 1992 NES Cross-section respondents could not be found for re-interview in 1993. 5 Analysis of pooled data from respondents from both components of the 1994 NES sample requires a strong assumption about the nature of the attrition of the 1992 NES Cross-section sample. It must be assumed that Panel attrition is not correlated with variables under consideration in the analysis. 6 Non-MSA segments were selected from the 1980 Census summary tape file series STF1B file, with minimum SSU size of 50 occupied HUs. 7 The number of segments shown for the 1996 NES Panel is the expected count; it is based on the number of 1994 NES Cross-section and Panel segments having selected lines. It is possible that some of these segments yielded no 1994 interviews and so do not actually show up in the 1996 Panel. 8 Kish, L. (1965). Survey Sampling, John Wiley & Sons, New York, NY. 9 Kish, L. (1949). "A procedure for objective respondent selection within the household," Journal of the American Statistical Association, Vol 44, pp. 380-387. 10 Office of Management and Budget (OMB) June 1990 definitions of MSAs, NECMAs, county, parish, independent city. These, of course, differ in some respects from the primary stage unit (PSU) definitions used in the 1980 SRC National Sample so will not be strictly comparable to the 1996 NES Panel PSUs--particularly in New England where MSAs were used as PSUs in the 1980 National Sample and NECMAs were used as PSUs in the 1990 National Sample. 11 For more detailed description of original Panel component selection, see appropriate sections earlier in this document. 12 In the 1990 SRC National Sample, U.S. Census Region boundaries were maintained for purposes of stratification at the Primary Stage of selection. Since some MSA definitions cross Region boundaries, such MSAs were split and the MSA counties recombined in ways that maintained the Region boundary. This PSU actually contains the Ohio counties from both the Steubenville- Wierton, OH-WV MSA (Jefferson County, OH) and the Wheeling, WV-OH MSA (Belmont County, OH) and although it is made up of MSA counties--it is not a cohesive MSA by OMB 1990 definition. 13 For efficiency of field work the substitution of two "B1" PSUs was allowed for the "A" areas in the normal 1990 half-sample -- Waco, TX MSA for Oklahoma City, OK MSA and Jim Wells County, TX for Lavaca County, TX. 14 Kish, L. (1965). Survey Sampling, John Wiley & Sons, New York, NY. 15 Kish, L. (1949). "A procedure for objective respondent selection within the household," Journal of the American Statistical Association, Vol 44, pp. 380-387. 16 See appropriate sections earlier in this report for details of the Cross-section supplement of the 1996 NES sample. 17 Outcome figures are from the 1996 National Pre-election Study Field Progress Report, February 28, 1997. 18 Figures in this table are from the 1996 National Post-Election Study Field Progress Report, April 18, 1997. 19 In constructing the analysis weight, a maximum of three eligible adults was allowed. 20 See 1994 NES sample weight documentation. 21 Weighted by `Intermediate factor' for Cross-section and Panel cases weighted proportionately as described above for 1996 NES combined Cross-section and panel analysis. 22 This 1996 Panel response rate appears lower than the 0.76 reported on Table 4 which was computed based on recontacted households having the eligible R from the 1994 study and actual 1996 NES sample release and interview figures from the 1996 NES final field report. 23 The four San Francisco (separated from Oakland, CA in the 1990 OMB definition), CA MSA area segments were considered as part of the Los Angeles-Long Beach, CA MSA for purposes of SE Code assignment to avoid having empty SE CODE cells since there were very few 1996 NES Cross-section respondents in this MSA. 24 See footnote #23. 25 The standard error of a percentage is a symmetric function with its maximum centered at p=50%; i.e., the standard error of p=40% and p=60% estimates are equal. >>1997 NES Pilot Technical Note - Randomization Problem April 24, 1998 The Surveycraft CATI system's 'Random Number Generation' features and their Effects on Analysis of the 1997 NES Pilot "Group threat" Experiment. Steve Heeringa, Division of Survey Technologies, Survey Research Center Executive Summary: A problem has been identified in the random assignment of treatments in an experimental question module of the 1997 NES Pilot survey instrument. The randomization problem has been linked to unexpected correlation in sequences of random number calls made within the Surveycraft computer-assisted interviewing system. The problem does produce an unbalanced distribution of sample cases to the cells of the factorial experimental design but does not lead to a bias in the interpretation of the experimental results. Details are provided below. A report that analyzes these items is the 1997 pilot study report by J. Bowers. A portion of the 1997 NES Pilot questionnaire (section 'J') includes a "group threat" factorial experimental design to study question order and 'threat level' treatment effects in a series of items that explore respondent views and prejudices toward African-Americans and Christian Fundamentalists. The full design involves 2 question sequence orderings - African-Americans first or Christian Fundamentalists first; 2 levels of intended "threat" - high and low; and 3 'threat domains': political, social and economic. The Survey Craft computer assisted interview (CAI) application used an internal random number generator to determine each subject's assignment to target group order and threat level for the questions about each target group. A different Surveycraft function was used to randomize the order of the three threat domains, once the group and threat level were determined. The intent of the CAI programming was to randomly assign the group order, threat level by group and threat domain for each respondent. Complete randomization of choice for each of these three experimental components is expected to yield equal numbers of cases at each combination of treatment for the 2 x 2 x 3 factorial design. In practice, due to sampling variability inherent in the randomization process, the actual counts in each experimental cell will be distributed about the expected sample size for each experimental cell. Within the Surveycraft CAI questionnaire for the 1997 NES Pilot, the random assignment of group order and threat level was determined by a call to an internal system random number generator. Examination of the final sample size distribution across the cells of this experimental question module suggests significant departures from the equal sample size per cell assumption. Specifically, there appears to be a problem in the randomization assignment for group order and threat level. Table 1 compares the expected and actual distributions of 1997 NES Pilot sample to experimental cells: Table 1 1997 NES Pilot Section J Question Experiment. Expected and Actual Distribution of Respondents to Treatment Categories. Target Group Order Threat Expected Actual Level Respondents Respondents First Series African Americans High 138 181 Low 138 116 Christian Fundamentalists High 138 53 Low 138 202 Second series African Americans High 138 100 Low 138 197 Christian Fundamentalists High 138 114 Low 138 141 Through analysis of actual random numbers generated in the course of the 1997 NES Pilot computer-assisted interviews and communication with the authors of Surveycraft, the randomization problem has been traced to Surveycraft's handling of random number seeds in sequential calls of the random number function. Our review finds that the initial random number draws to determine the target group for the first question sequence were performed correctly. Observed variation in numbers of cases assigned at random to the African-American (n=297) and Christian Fundamentalists (n=255) target group question order are due to sampling error in the random draws of binomial (0,1) indicator variables. Since the random draws to determine threat level in the first and second question sequences are correlated with this initial random draw they also are pure random numbers (albeit not independent of the initial draw). The randomization of the experiment is therefore not affected by the problem-the joint probability that a respondent receives a particular configuration of experimental treatments is independent of respondent characteristics or the sample design. Unfortunately, the correlated sequence of random numbers does affect the balance of the distribution of subjects to the experimental design cells. This will have an unspecified, but negative effect on the power to detect effects of target group ordering and threat level that are the object of the factorial experimental design. The third factor in the experimental design, random ordering of each question representing a threat domain, was performed by a separate Surveycraft internal function. To the best of our ability to test the mechanism, this dimension of the experiment appears free of the randomization problem identified for the group order and threat level experimental conditions. ISR/SRC has corrected the problem which created this situation, working with Surveycraft authors to identify programming changes and conventions that now permit independent random number sequence generation directly within the system. Random numbers to determine assignments to experimental treatment in question sequences were drawn in advance, tested for independence and preloaded for use by the interviewing application. These simulations demonstrated that sequences of independent random assignments to treatments are now functioning within the SRC Surveycraft CATI system. >> LIST OF TECHNICAL REPORTS AND OTHER OCCASIONAL PAPERS THROUGH 1996 1. Sanchez, Maria. (July 1982) "7-Point Scales." 2. Shanks, J. Merrill, Maria Sanchez, and Betsy Morton. (March 1983). "Alternative Approaches to Survey Data Collection for the National Election Studies." 3. Lake, Celinda. (September 1983) "Similarity and Representativeness of 1983 Pilot Samples." 4. Lake, Celinda. (November 1983) "Comparison of 3-point, 5-point, and 7-point Scales from the CATI Experiment 1982 Election Study." 5. NES Staff. (December 1983) "1980 Precinct Data Returns Project." 6. Lake, Celinda. (February 1984) "Coding of Independent/Independents and Apoliticals in the Party Identification Summary Code and Apoliticals in the Rolling Cross-Section." 7. Morchio, Giovanna and Maria Sanchez. (February 1984) "Creation of a Filter Variable to be Used When Analyzing Questions about Congressional Candidates in the 1982 Integrated Personal/ISR CATI/Berkeley CATI Dataset: A Report to the Board of Overseers, National Election Studies." 8. Morchio, Giovanna and Maria Sanchez. (March 1984) "Comparison of the Michigan Method of District Assignment on the Telephone with the Personal Interview Simulated Data: A Report to the Board of Overseers, National Election Studies." 9. Traugott, Santa. (June 1984) "Two Versions of the Abortion Question." 10. Sanchez, Maria.(July 1984) "Branching versus 7-point scale measurements." 11. NES Staff. (August 1984) "Weekly Field Report for the National Election Studies Continuous Monitoring, Jan. 11 - Aug. 3, 1984: A Report to the Board of Overseers, National Election Studies." 12. NES Staff. (August 1984) "Questions and Versions in NES Continuous Monitoring, 1984: A Report to the Board of Overseers, National Election Studies." 13. NES Staff. (n.d) "Years of Schooling." 14. NES Staff. (n.d) "Newspaper Code." 15. Traugott, Santa. (n.d.) "The Political Interest Variable on the 1984 Election Study." Unpublished Staff Memo to NES Planning Committee. 16. Sanchez, Maria and Giovanna Morchio. (n.d.) "Probing Don't Know Answers -- Do We Always Want to Do This?" 17. NES Staff. (February 1985) "Progress of the Rolling Cross Section." 18. Bowers, Jake. (February 1995) NES Pilot Study Efforts to Measure Values and Predispositions. Full text of paper in WordPerfect 6.0 is available via the NES FTP server. 19. Traugott, Santa. (February 1985) "Some Analysis of Hard-to-Reach Rolling Thunder Respondents." 20. Traugott, Santa. (April 1985) "Sample Weighting in NES Continuous Monitoring, 1984: A Report to the Board of Overseers, National Election Studies." 21. Traugott, Santa. (April 1985). "Sample Weighting in NES Pre-Post Election Survey,1984: A Report to the Board of Overseers, National Election Studies." 22. Brehm, John. (June 1985) "Report on Coding of Economic Conditions Series in the 1984 Pre-Post Election Study" 23. Brehm, John. (July 1985). "Question Ordering Effects on Reported Vote Choice." 24. Traugott, Santa. (July 1985) "Assessment of Media Measures in RXS." 25. Traugott, Santa. (July 1985) "Assessment of Media Measures in Pre-Post" 26. Brehm, John. (August 1985). "Analysis of Result Code Disposition for Continuous Monitoring by Time in Field: Report to the Board of Overseers, National Election Studies." 27. Morchio, Giovanna, Maria Sanchez and Santa Traugott. (November 1985). "Mode Differences: DK Responses in the 1984 Post-Election Survey: A Report to the Board of Overseers, National Election Studies." 28. Morchio, Giovanna and Santa Traugott. (February 1986) "Congressional District Assignment in an RDD Sample: Results of 1982 CATI Experiment." 29. Brehm, John and Santa Traugott. (March 1986) "Similarity and Representativeness of the 1985 Pilot Half-samples." 30. Gronke, Paul. (September 1986) "NES Question C2: R's Party Registration." 31. Brehm, John. (March 1987) "How Representative is the 1986 Post-Election Survey?" 32. Morchio, Giovanna. (May 1987) "Trends in NES Response Rates." 33. Brehm, John. (December 1987) "Who's Missing? an Analysis of NonResponse in the 1986 Election Study: A Report to the Board of Overseers, National Election Studies." 34. Traugott, Santa. (August 1989) "Validating Self-Reported Vote: 1964-1988." 35. -- open -- 36. Traugott, Santa and Giovanna Morchio. (March 1990) "Assessment of Bias Due to Attrition and Sample Selection in the NES 1989 Pilot Study." 37. -- open -- 38. Gronke, Paul. (May 1990) "Assessing the Sample Quality of the 1988 Senate Election Study: A response to Wright." 39. Presser, Stanley, Michael W. Traugott and Santa Traugott. (November 1990). "Vote 'Over' Reporting in Surveys: The Records or the Respondents?" 40. Bloom, Joel. (March 1991) "Sources of Pro-incumbent Bias in NES Survey Estimates for U.S. House Races since 1978: A Second Look." 41. Mayer, Russell. (November 1991) "Identifying Bias in Voting Models." 42. Traugott, Michael W., Santa Traugott and Stanley Presser. (May 1992) "Revalidation of Self-Reported Vote." 43. Rosenstone, Steven J., Margaret Petrella and Donald R. Kinder. (April 1993) "The Consequences of Substituting Telephone for Face-to-Face Interviewing in the 1992 National Election Study." 44. Luevano, Patricia. (March 1994) "Response Rates in the National Election Studies, 1948-1992." 45. Traugott, Santa and Steven J. Rosenstone. (Nov. 1994) "Panel Attrition Among the 1990-1992 Panel Respondents." 46. Traugott, Santa and Steven J. Rosenstone. (Nov. 1994) "Demographic Characteristics of Respondents to the 1980, 1984 and 1988 NES Pre-Election Studies by Week of Interview." 47. Traugott, Santa. (Nov. 1994) "Candidate Traits Used in NES Studies, 1979-1994." 48. Traugott, Santa. (Nov. 1994) "Affects Towards Candidates Used in NES Studies, 1979-1994." 49. Traugott, Santa. (Nov. 1994) "Candidate Placements Used in NES Studies, 1968-1994." 50. Sheng, Shing-Yuan. (Jan. 1995) "NES Measurements of Values and Pre-Dispositions, 1984-1992." 51. Traugott, Santa. (Feb. 1995) "NES Question Batteries: Measuring Values and Dispositions, 1983-1994." 52. Tolleson-Rinehart, Sue, et.al. (May 1994) "The Reliability, Validity, and Scalability of Indicators of Gender Role Beliefs and Feminism the 1992 National Election Study: A Report to the ANES Board of Overseers." >> LIST OF PILOT STUDY REPORTS 1991 Pilot Study Reports Beebe, Tim. The Effects of Pre-Notification and Incentive on Panel Attrition. Undated. Brady, Henry E. Report on Feeling Thermometer for "Moderates." January 13, 1992. Citrin, Jack, Donald P. Green, Beth Reingold and David O. Sears. A Report on Measures of American Identity and New "Ethnic" Issues in the 1991 NES Pilot Study. Undated. Conover, Pamela J., and Virginia Sapiro. Gender Consciousness and Gender Politics in the 1991 Pilot Study: A Report to the ANES Board of Overseers. January, 1992. Delli Carpini, Michael X., and Scott Keeter. An Analysis of Information Items on the 1990 and 1991 NES Surveys: A Report to the Board of Overseers for the National Election Studies. January 14, 1992. Highton, Benjamin, and Raymond E. Wolfinger. Estimating the Size of Minority Groups. January 13, 1992. Huddy, Leonie. Analysis of Old-Age Policy Items in the 1991 Pilot Study. Undated. _____. Addendum. February 2, 1992. Knack, Stephen. Social Connectedness and Voter Participation: Evidence from the 1991 NES Pilot Study. January 1992. _______. Social Altruism and Voter Turnout: Evidence from the 1991 NES Pilot Study. January, 1992. _______. Performance and Recommendations Summary for 1991 NES Pilot Variables #2828-2847. January 24, 1992. _______. Deterring Voter Registration Through Juror Source Practices: Evidence from the 1991 NES Pilot Study. January, 1992. Oliver, Eric, and Raymond E. Wolfinger. Jury Duty as a Deterrent to Voter Registration. January 22, 1992. Zaller, John. Report on 1991 Pilot Items on Environment. February 2, 1992. 1993 Pilot Study Reports Dennis, Jack. The Perot Constituency: A Report to the Board of Overseers of the National Election Studies. March 10, 1994. Franklin, Charles H. Report on the 1993 NES Pilot Study. March 16, 1994 Jacobson, Gary and Doug Rivers. Overreport of Vote for House Incumbent in NES Surveys. March 11, 1994. Strand, Douglas. Homosexuality, Gay Rights, and the Clinton Coalition: Report to the National Election Studies on Results from the 1993 NES Pilot Study. March 16, 1994. Stoker, Laura. New Items on the 1993 Pilot Study. March 9, 1994. Stoker, Laura. A Reconsideration of Self-Interest in American Public Opinion. Presented at the annual meeting of the Western Political Science Association. Albuquerque, New Mexico. (March 10-12, 1994) Zaller, John. Securing the District. March 11, 1994. 1995 Pilot Study Reports Alvarez, R. Michael. Survey Measures of Uncertainty: a Report to the National Election Studies Board on the Use of Certainty Questions to Measure Uncertainty about Candidate Traits and Issue Positions. Bartels, Larry M. Budget Items on 1995 Pilot Study. ________. Entertainment Television Items on 1995 Pilot Study. ________. Humanitarianism Items on 1995 Pilot Study. ________. Issue Scales Versus Effort Items on the 1995 Pilot Study ________. Talk Radio Items on 1995 Pilot Study. ________. Television News Items on 1995 Pilot Study. Berinsky, Adam and Steven Rosenstone. Evaluation of Environmental Policy Items on the 1995 NES Pilot Study. Buhr, Tami, Ann Crigler and Marion Just. Media Questions on the 1996 election study and related content analysis of media cover of the presidential campaign. Hansen, John Mark. Revealed Preference Budget Items on the 1995 National Election Pilot Study: a Report. Marcus, George E. And Michael Mackuen. Measuring Mood in the 1995 NES Pilot Study. Rabinowitz, George and Stuart Elaine Macdonald. New Issues on the 95 Pilot Study. Rahn, Wendy W. And John Transue. The Political Significance of Fear of Crime. Richardson, Amy. Questions on Public Attitudes Toward the Environment. Steenbergen, Marco R. Compassion and American Public Opinion: An Analysis of the NES Humanitarianism Scale. Zaller, John. Analysis of News Exposure Items from the 1995 Pilot 1997 Pilot Study Reports Barker, David. "Measures of Talk Radio Exposure and Attention." Burden, Barry C. and Janet M. Box-Steffensmeier. "Vote Likelihood and Institutional Trait Questions in the 1997 NES Pilot Study." Carman, Christopher and Christopher Wlezien. "Ideological Evaluations of Government Institutions and Policy." Cirksena, Kathy. "Report to the Board of Overseers on Respondent Preferences for Cash Incentive in the 1997 Pilot (from Panel Debriefing" Questions) Rahn, Wendy and Christina Wessel. "Perceptions of the Partisan Homogeneity of Social Groups: A Report to the NES Board of Overseers." Sapiro, Virginia. "Pro-Life People or Opponents of Abortion? Pro-Choice People or Supporters of Abortion? A Report on the NES 1997 Pilot Study." Wald, Kenneth D., et al. "Evaluation of the New Religious Items on the NES 1997 Pilot Study: A Report to the NES Board." Wlezien, Christopher. "Liberal-Conservative Evaluations of Groups." Wong, Cara. "Group Closeness: 1997 National Election Study Pilot Report." >> MASTER CODE CAMPAIGN ISSUES 001 "Domestic issues" 006 Child care; DAY CARE; child support 045 ABORTION; any reference 010 UNEMPLOYMENT, jobs, retraining -- general or national 011 Unemployment, lack of jobs in specific area/region/state/industry 012 More help for the unemployed 020 EDUCATION -- any mention, including quality of schools, cost of college, students not learning anything 030 AGED/ELDERLY -- any mention, including Social Security, Medicare, eldercare. 040 HEALTH PROBLEMS -- quality of medical care, cost of medical care, availability of medical care, catastrophic health insurance (except AIDS, code 048) 048 AIDS 050 HOUSING -- providing housing for the poor, the homeless, young people can't buy homes, any mention. 055 INFRASTRUCTURE -- Build/maintain roads, bridges, railroads, mass transit systems; transportation - NFS "POVERTY" has the general thrust of helping the underprivileged; the 'welfare' code 090 may have connotation of undeserving people on welfare. Thus, 'do more for people on welfare' is a 060 rather than 90. WELFARE --NFS is a 090. 060 POVERTY; aid to poor, underprivileged people; help for the (truly) needy; general reference to anti- poverty programs; hunger/help for hungry people 090 SOCIAL WELFARE; "Welfare"; the welfare mess, too many undeserving on welfare 099 OTHER SPECIFIC MENTIONS OF DOMESTIC ISSUES ....................................... 100 Problems of the FARMERS; farm bankruptcies, poor prices for crops, effects of the drought 150 Protecting the ENVIRONMENT, POLLUTION, the ozone layer, the greenhouse effect. 151 Controlling/REGULATING GROWTH or land development; banning further growth/development in crowded or ecologically sensitive areas; preserving natural areas 154 TOXIC WASTE, RADIOACTIVE WASTE 160 Need to develop ALTERNATIVE ENERGY SOURCES 199 Other specific mentions of AGRICULTURE or ENVIRONMENT problems ....................................... 300 CIVIL RIGHTS/RACIAL PROBLEMS; affirmative action programs; relations between blacks and whites 310 WOMEN'S ISSUES -- ERA, equal pay for equal work, maternity leave (except day care, code 006) 320 DRUGS -- extent of drug use in U.S; "WAR ON DRUGS"; drugs--NFS; ALCOHOLISM, any mention 321 DRUGS -- stopping drugs from coming into this country 340 CRIME/VIOLENCE; streets aren't safe; respect for police; releasing criminals early; not enough jails; death penalty 367 GUN CONTROL - all mentions 370 EXTREMIST GROUPS/TERRORISTS 380 General mention of MORALITY/TRADITIONAL VALUES; sex, bad language, pornography, teenage pregnancy 381 Specific mention of FAMILY VALUES -- latchkey children, divorce; unwed mothers, working mothers 382 Homosexual/gay rights; gays in the military [code 048 for mentions of AIDS) 384 RELIGION (too mixed up in) and politics; prayer in schools 399 OTHER MENTION of race, public order, morality ....................................... 400 INFLATION, high prices, cost of living 405 WAGES TOO LOW; minimum wage 408 Recession/Depression in specific industries, states or regions -- slump in OIL/STEEL/AUTO INDUSTRY, etc. (except farm, code 101); hard times in this REGION or area 410 RECESSION; DEPRESSION, hard times -- no specific locale or industry 415 THE DEFICIT; BALANCING THE BUDGET; cutting government spending 416 TAXES -- any reference; tax reform 425 TOO MANY IMPORTS -- protectionism, competition, outsourcing, problems of auto industry relating to foreign competition; U.S. makes (too) few exports; (high) tariffs imposed by other nations; free trade; GATT 427 VALUE OF THE DOLLAR -- strengthening or weakening 428 STOCK MARKETS; investments; interest rates 440 CLASS ORIENTED ECONOMIC CONCERNS -- middle class getting squeezed; big business too powerful 453 Solvency/stability/regulation/control of the nation's FINANCIAL INSTITUTIONS. [1990] Savings and Loan scandals 460 IMMIGRATION 491 ECONOMICS, THE ECONOMY 493 BALANCE OF TRADE; balance of payments; foreign oil dependency (except supply of oil, see 524) 499 OTHER MENTION of economic, business or labor problems ....................................... 500 FOREIGN POLICY; FOREIGN AFFAIRS 514 LATIN AMERICA, Central America, AID TO CONTRAS (reference to IRAN-CONTRA coded 816) 516 AFRICA -- starving people, overpopulation 517 SOUTH AFRICA -- Apartheid 524 MIDDLE EAST -- Iran hostages, Persian Gulf, supply of mid-east oil (except oil dependency, see 493) 530 RUSSIA -- relations with, arms talks, detente; summit, etc. 540 FIRMNESS in foreign policy 550 U.S. military involvement abroad 560 FOREIGN AID; amount of money given to foreign countries; obligation to take care of our problems at home first 570 AVOID WAR, establish PEACE -- any reference 700 DEFENSE (SPENDING); the military; quality/cost of weapons 710 NUCLEAR ARMS RACE -- disarmament, SALT, INF, threat of nuclear war; arms control 712 STAR WARS 714 SPACE PROGRAM ....................................... 810 Honesty, sincerity of government officials; corruption 811 Honesty, sincerity of candidates in general; e.g., "just making promises," "saying whatever it takes to get elected" 812 Candidates are just talking (negatively) about each other, MUD SLINGING. 813 How well incumbent represents/candidate would REPRESENT THIS DISTRICT 814 Congressperson's personal life/morality 815 Candidate's ABILITY/EXPERIENCE 816 Candidate's (voting) RECORD 817 PRESIDENT CLINTON 818 BUSH and the IRAN-CONTRA affair 819 IRAN-CONTRA affair, mess, scandal, IRAN ARMS DEAL, without reference to Bush 850 Which party will control the House of Representatives; other partisan mentions 851 Need for change/new blood/fresh ideas in Congress; term limits for members of Congress 876 PHILOSOPHICAL DIFFERENCES between the candidates - liberal vs. conservative views; balance of authority between state and federal government; etc. 900 A local issue or concern -- the college, the dam, the auto-insurance initiative, the leak in our nuclear plant 991 1992: OTHER SPECIFIC MENTIONS OF CAMPAIGN ISSUES 995 1990: "There were no issues" (except 996); just party politics 997 1990: OTHER SPECIFIC MENTIONS OF CAMPAIGN ISSUES 996 1992: INAP 1990: "There was no campaign in my district" [Missing Data] 998 DK >> MASTER CODE CAMPAIGN POLITICAL ADVERTISEMENTS 1992 CODES (PART ONE) R Pays No Attention To Political Ads 001 R claims not to remember what the ads s/he saw were about - NFS says only "nothing", "very little/not much", "can't remember", "don't recall", etc. without further explanation or elaboration). 002 R deliberately and actively avoids watching political ads (I hit the mute button/change the channel; I go to the refrigerator, etc.). 003 R does watch the political ads but indicates s/he chooses to pay no attention to them (I don't pay much attention, they don't register on my mind, goes in one ear and out the other, I just laugh at them, I'm immune to them). R GIVE GENERAL ASSESSMENT OF POLITICAL ADS (NO CANDIDATE SPECIFIED) 010 AMOUNT/FREQUENCY OF ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 011 PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 012 PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 013 DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 014 HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 015 NEGATIVE CAMPAIGNING - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 016 POSITIVE CAMPAIGNING - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 017 HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 018 HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 028 OTHER POSITIVE GENERAL ASSESSMENT OF POLITICAL ADS (NO CANDIDATE SPECIFIED) 029 OTHER NEGATIVE GENERAL ASSESSMENT OF POLITICAL ADS (NO CANDIDATE SPECIFIED) R GIVES GENERAL ASSESSMENT OF BUSH POLITICAL ADS 030 AMOUNT/FREQUENCY OF BUSH ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 031 BUSH ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 032 BUSH ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 033 BUSH ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 034 BUSH ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 035 NEGATIVE CAMPAIGNING BY BUSH - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 036 POSITIVE CAMPAIGNING BY BUSH - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 037 BUSH ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 038 BUSH ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 039 R REFUSES TO LISTEN TO/WATCH BUSH ADS SPECIFICALLY 048 OTHER POSITIVE GENERAL ASSESSMENT OF BUSH POLITICAL ADS 049 OTHER NEGATIVE GENERAL ASSESSMENT OF BUSH POLITICAL ADS R GIVES GENERAL ASSESSMENT OF CLINTON POLITICAL ADS 050 AMOUNT/FREQUENCY OF CLINTON ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 051 CLINTON ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 052 CLINTON ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 053 CLINTON ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 054 CLINTON ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 055 NEGATIVE CAMPAIGNING BY CLINTON - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 056 POSITIVE CAMPAIGNING BY CLINTON - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 057 CLINTON ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 058 CLINTON ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 059 R REFUSES TO LISTEN TO/WATCH CLINTON ADS SPECIFICALLY 068 OTHER POSITIVE GENERAL ASSESSMENT OF CLINTON POLITICAL ADS 069 OTHER NEGATIVE GENERAL ASSESSMENT OF CLINTON POLITICAL ADS R GIVES GENERAL ASSESSMENT OF PEROT POLITICAL ADS 070 AMOUNT/FREQUENCY OF PEROT ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 071 PEROT ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 072 PEROT ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 073 PEROT ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 074 PEROT ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 075 NEGATIVE CAMPAIGNING BY PEROT - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 076 POSITIVE CAMPAIGNING BY PEROT - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 077 PEROT ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 078 PEROT ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 079 R refuses to listen to/watch Perot ads specifically 088 Other positive general assessment of Perot political ads 089 Other negative general assessment of Perot political ads R IDENTIFIES SPECIFIC BUSH POLITICAL ADS 130 Bush ad - no other details given. 131 Bush ad - no content given, but production details remembered (e.g., closeup of face, sitting on edge of desk, it was green). 132 Bush ad - "Two Faces of Clinton"/Time magazine cover highlighting two faces. 133 Bush ad - computer ad. 134 Bush ad - on Bush's record in general. 135 Bush ad - attacking Clinton's record in Arkansas. 136 Bush ad - on Clinton's draft record/anti-American activities. 137 Bush ad - about taxes; saying Bush won't raise taxes (again). 138 Bush ad - about Bush's economic plan/promises for the economy. 139 Bush ad - Florida relief; giving food to poor countries; Bush portrayed as a caring person. 140 Bush ad - family values; families coming together; Bush portrayed as a family man. 141 Bush ad - foreign policy accomplishments of the Bush administration; Bush shown as commander-in-chief. 142 Bush ad - needs four more years to finish the job. 143 Bush ad - clips from the Republican convention. 144 Bush ad - average people questioning Clinton's willingness and ability to keep his promised. 149 Bush ad - other R IDENTIFIES SPECIFIC CLINTON POLITICAL ADS 150 Clinton ad - no other details given. 151 Clinton ad - no content given, but production details remembered (e.g., closeup of face, waving to crowd, flag in background). 152 Clinton ad - attacking Bush's broken promise not to raise taxes; "read my lips -- no new taxes". 153 Clinton ad - attacking Bush's handling of the economy; "we can't afford four more years". 154 Clinton ad - about creating jobs/putting people back to work. 155 Clinton ad - about the need for change; about rebuilding America/putting American on the right course. 156 Clinton ad - defending Clinton's record in Arkansas/record on taxes as governor. 157 Clinton ad - reforming welfare. 158 Clinton ad - showing working people. 159 Clinton ad - defending Clinton's draft record. 160 Clinton ad - giving address to write to for Clinton's economic plan; experts endorsing Clinton's economic plan. 169 Clinton ad - other R IDENTIFIES SPECIFIC PEROT POLITICAL ADS 170 Perot ad - no other details given. 171 Perot ad - no content given, but production details remembered (e.g., sitting behind a desk, scroll with writing, 30 minutes long). 172 Perot ad - used a lot of charts and graphs. 173 Perot ad - describing in general terms problems with the economy/the deficit. 174 Perot ad - detailed how the deficit would affect future generations. 175 Perot ad - plans/promises to solve America's problems. 176 Perot ad - Purple Heart ad 189 Perot ad - other R IDENTIFIES A SPECIFIC EVENT THAT WAS NOT A PRESIDENTIAL POLITICAL AD 190 Other - R describes a new event that clearly was not part of a political ad (e.g., Quayle talking about Murphy Brown; Mary Matalin talking about Hillary Clinton). 191 Other - R describes a political ad, but one for a congressional, state or local candidate or one concerning a controversial issue (e.g., abortion, gay rights, etc.). MISCELLANEOUS 997 Other, miscellaneous 998 DK (except 001-003) 999 NA 1996 CODES (PART TWO) NOTE: The codes for political ads used in 1996 are different from the coding scheme used for political ads in 1992. As a result of experience with and recommendations about the wording of political ad questions in 1992, the Board of Overseers approved a different means of asking about recall of political advertisements in the 1996 NES. Two important differences set 1996 apart from 1992. One is that the question in 1996 asks the respondent to focus on recall of a single specific ad, the one you ad remember best'. In 1992 the question asked about "what do you remember about any of these ads"-- in the plural. Second, in 1992 the question concerned Presidential ads while in 1996 the questions did not restrict respondents to Presidential ads,. Thus the coding scheme for 1996, while developed from and similar to that of 1992, is not the same. Differing coding categories exist (specific ads mentioned in 1992 of course have no relevance in 1996) and the frequencies for similar or repeated categories are also different. The effort in 1996 was to code accurately the open-ended responses received in 1996 while producing codes that could be aggregated in ways that facilitate some kinds of comparisons between 1992 and 1996. R Pays No Attention To Political Ads 001 R claims not to remember what the ads s/he saw were about - NFS says only "nothing", "very little/not much", "can't remember", "don't recall", etc. without further explanation or elaboration). 002 R deliberately and actively avoids watching political ads (I hit the mute button/change the channel; I go to the refrigerator, etc.) 003 R does watch the political ads but indicates s/he chooses to pay no attention to them (I don't pay much attention, they don't register on my mind, goes in one ear and out the other, I just laugh at them, I'm immune to them). R GIVE GENERAL ASSESSMENT OF POLITICAL ADS (NO CANDIDATE SPECIFIED) 010 AMOUNT/FREQUENCY OF ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 011 PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE -too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 012 PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important)issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 013 DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 014 HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real)facts/all the facts; tries to clarify/face the issues; they make sense. 015 NEGATIVE CAMPAIGNING - (too negative); (too much)backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 016 POSITIVE CAMPAIGNING - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 017 HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 018 HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 028 OTHER POSITIVE GENERAL ASSESSMENT OF POLITICAL ADS (NO CANDIDATE SPECIFIED) 029 OTHER NEGATIVE GENERAL ASSESSMENT OF POLITICAL ADS(NO CANDIDATE SPECIFIED) R GIVES GENERAL ASSESSMENT/DESCRIBES GENERAL FEATURE(S) OF DOLE POLITICAL AD(S) 030 AMOUNT/FREQUENCY OF DOLE ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 031 DOLE ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 032 DOLE ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 033 DOLE ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 034 DOLE ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 035 NEGATIVE CAMPAIGNING BY DOLE - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 036 POSITIVE CAMPAIGNING BY DOLE - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 037 DOLE ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 038 DOLE ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 039 R REFUSES TO LISTEN TO/WATCH DOLE ADS SPECIFICALLY 040 DOLE AD NEGATIVE RE: CLINTON NFS ( badmouthing' downside of' Clinton) 048 OTHER POSITIVE GENERAL ASSESSMENT OF DOLE POLITICAL ADS 049 OTHER NEGATIVE GENERAL ASSESSMENT OF DOLE POLITICAL ADS R GIVES GENERAL ASSESSMENT/DESCRIBES GENERAL FEATURE(S) OF CLINTON POLITICAL AD(S) 050 AMOUNT/FREQUENCY OF CLINTON ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 051 CLINTON ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 052 CLINTON ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 053 CLINTON ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 054 CLINTON ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 055 NEGATIVE CAMPAIGNING BY CLINTON - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 056 POSITIVE CAMPAIGNING BY CLINTON - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 057 CLINTON ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 058 CLINTON ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 059 R REFUSES TO LISTEN TO/WATCH CLINTON ADS SPECIFICALLY 060 NEGATIVE RE: DOLE, NFS 068 OTHER POSITIVE GENERAL ASSESSMENT OF CLINTON POLITICAL ADS 069 OTHER NEGATIVE GENERAL ASSESSMENT OF CLINTON POLITICAL ADS R GIVES GENERAL ASSESSMENT/DESCRIBES GENERAL FEATURE(S) OF PEROT POLITICAL AD(S) 070 AMOUNT/FREQUENCY OF PEROT ADS - too many of them; they show too many in one evening/time period; see the same ones over and over. 071 PEROT ADS PROVIDE NO INFORMATION/SERVE NO VALUABLE PURPOSE - too vague/general; not specific (enough); not talking about real/important issues; contain only rhetoric/self-serving promotion/platitudes; point out problems but offer no solutions. 072 PEROT ADS PROVIDE INFORMATION/SERVE VALUABLE PURPOSE - talk about (important) issues/candidate's stands on issues; try to present solutions to issues; are enlightening; treat voters like grown-ups. 073 PEROT ADS DISHONEST/MISLEADING - (too) deceitful; tell lies/half-truths/only the facts that help them; try to confuse/hide/avoid the issues; say only what they think the voter wants to hear. 074 PEROT ADS HONEST/STRAIGHT-FORWARD - tells the truth; presents the (real) facts/all the facts; tries to clarify/face the issues; they make sense. 075 NEGATIVE CAMPAIGNING BY PEROT - (too negative); (too much) backbiting/mudslinging; only try to tear opponents down/make personal attacks on opponent. 076 POSITIVE CAMPAIGNING BY PEROT - doesn't make personal attacks on opponent; talk about the candidate/why the candidate should be elected. 077 PEROT ADS HAD NEGATIVE EFFECT ON R - made R angry/disgusted; destroyed R's interest in politics/the election; R finds them boring; R is tired of seeing them. 078 PEROT ADS HAD POSITIVE EFFECT ON R - helped R understand the candidate/issues; helped R decide who to vote for. 079 R refuses to listen to/watch Perot ads specifically 088 Other positive general assessment of Perot political ads 089 Other negative general assessment of Perot political ads R DESCRIBES SPECIFIC DOLE POLITICAL AD(S) 130 Dole ad - no other details given ("I know it was Dole's ad") 131 Dole ad - production details described (showed him in black and white, he was talking to some women) 132 Dole ad - 15% tax cut, would let people keep more of what they earn (i.e. would cut taxes) 133 Dole ad - war injuries, military service record 134 Dole ad - Russell KS values and community, personal history/life story (other than military record) 135 Dole ad - Dole's position on Medicare cuts 136 Dole ad - mention of Kemp 140 Dole ad - attacking Clinton for largest tax hike in history', criticizing Clinton for apologizing for raising taxes, general/other negative on Clinton's tax record 141 Dole ad - attacking Clinton re: Whitewater 142 Dole ad - attacking Clinton re: ethics of White House staff and cabinet 143 Dole ad - attacking Clinton re: immigration and border patrol 144 Dole ad - attacks Clinton as a liar-NFS; Clinton changes what he says from one time to the next; Clinton's inconsistencies; doesn't keep/breaks promises 145 Dole ad - Attacks Clinton re: drug policies, teen drug use going up, budget cuts for drug enforcement, Clinton on MTV re: pot use 146 Dole ad - Attacks Clinton re: family values 147 Dole ad - Attacks Clinton as a liberal, closet liberal; shows Clinton saying I'm not a liberal' 148 Dole ad - other negative re: Clinton 149 Dole ad - other specifics R DESCRIBES SPECIFIC CLINTON POLITICAL AD(S) 150 Clinton ad - no other details given 151 Clinton ad - production details described 152 Clinton ad - describing his stance on family values. 153 Clinton ad - describing the achievements of his first term in office 154 Clinton ad - describing his record on employment, jobs 155 Clinton ad - reforms welfare, makes jobs for unemployed/people on welfare 156 Clinton ad - saying Clinton makes up his own mind, is a leader 157 Clinton ad - Clinton's efforts on drugs; Dole criticisms wrong/unfair; appt. of drug czar; policies and funding to combat drugs 158 Clinton ad - Clinton's record on [illegal] immigration 159 Clinton ad - Clinton doing right on elderly health care, positive Record on Medicare 160 Clinton ad - supports education, supports student loan pgms, supports reading pgms 161 Clinton ad - support of issues affecting children (other than drug policy or education) 162 Clinton ad - record on gun control, puts more cops on streets, endorsed by police, tough on crime (excludes any drug-related--see 157) 163 Clinton ad - Other positive, not coded elsewhere 170 Clinton ad - compares Clinton's record favorably w/Dole's on multiple issues 171 Clinton ad - attacking Dole's stance on social security 172 Clinton ad - attacking Dole's position on school lunch, other children's issues, on education 173 Clinton ad - attacking Dole's Medicare voting record 174 Clinton ad - attacking Dole re: his comments on cigarettes, support of tobacco industry 175 Clinton ad - Attacking Dole's tax cut proposal 176 Clinton ad - negative attack on Dole/Gingrich 177 Clinton ad-neg re: Dole's voting record: wrong for the past, wrong for the future' 179 Clinton ad discussing Dole--NFS, other 169 Clinton ad - other specifics R DESCRIBES SPECIFIC PEROT POLITICAL AD(S) 180 Perot ad - no other details given. 181 Perot ad - production details described 182 Perot ad - used a lot of charts and graphs. 183 Perot ad - describing problems with the economy/the deficit/the budget, Perot will drop our taxes. 184 Perot ad - doesn't take special interest' money; not beholden to special interests 185 Perot ad - he'll abolish the IRS 186 Perot ad - announcing his candidacy ( I'm back'); announcing his VP candidate 187 Perot ad - re: not being in debates 189 Perot ad - other specifics R DESCRIBES A SPECIFIC EVENT THAT WAS NOT A POLITICAL AD 190 R describes a news event that clearly was not part of a political ad; mentions watching the convention or seeing a candidate on a news program or during debates. CANDIDATE NAMED IS NOT MAJOR PRESIDENTIAL CANDIDATE (INCLUDES STATE AND LOCAL RACES) 191 R describes a political ad, but one for a congressional, state or local candidate R DESCRIBES OTHER ADS: CANDIDATE NOT ASCERTAINED/AD SPONSOR NOT ELSEWHERE IDENTIFIED 192 R describes ad concerning a specific issue (e.g.Medicare, abortion, gay rights, etc.). R IDENTIFIES AD AS BEING BY THE DEMOCRATS' (NOT ASSOCIATED W/ SPECIFIC CANDIDATE) 301-General positive about Democrats/Democratic candidates, NFS 302-Negative towards the Republicans 397-Other R IDENTIFIES AD AS BEING BY THE REPUBLICANS' (NOT ASSOCIATED W/ SPECIFIC CANDIDATE) 401-General positive about republicans/Republican candidates, NFS 402-Negative towards the Democrats 497-Other DON'T RECALL CANDIDATE, NO SPECIFIC CANDIDATE BUT AD DESCRIPTION MENTIONS CLINTON, DOLE or BOTH Clinton: 502 positive about Clinton: other and NFS 503 Clinton and taxes 504 Clinton and pot 505 negative about Clinton: other, NFS 506 names Clinton Dole: 520 negative about Dole's past political stands, Dole's voting record 521 Dole and taxes; the budget/finances, will help the little people on taxes 523 Dole general, other, NFS 524 Dole, recalls production details 525 Dole in WWII, injuries 526 negative towards Dole other, nfs, general Both Clinton and Dole: 598 R mentions both Clinton and Dole, general, other, NFS 599 Dole and Clinton contradict each other MISCELLANEOUS 996 Miscellaneous production details recalled 997 Other, miscellaneous 998 DK 999 NA >> MASTER CODE CANDIDATE NUMBER SENATE: 10 Third party or independent Senate candidate ** 11 Democratic candidate in open Senate race 12 Republican candidate in open Senate race 13 Democratic Senate incumbent 14 Republican Senate incumbent 15 Democratic Senate challenger 16 Republican Senate challenger 17 Democratic Senator, no race in state 18 Republican Senator, no race in state 19 Democratic Senator, term not up in state with race 21 Democratic Senator--retiring (state with open race) 22 Republican Senator--retiring (state with open race) 27 Democratic Senator, no race in state 28 Republican Senator, no race in state 29 Republican Senator, term not up in state with race HOUSE: 30 Third party or independent House candidate ** 31 Democratic candidate in open House race 32 Republican candidate in open House race 33 Democratic House incumbent 34 Republican House incumbent 35 Democratic House challenger 36 Republican House challenger 41 Democratic Representative--retiring (district with open race) 42 Republican Representative--retiring (district with open race) GOVERNOR: [NOT USED 1992 and 1996] 50 Third party or independent Gubernatorial candidate ** 51 Democratic candidate in open Gubernatorial race 52 Republican candidate in open Gubernatorial race 53 Democratic Gubernatorial incumbent 54 Republican Gubernatorial incumbent 55 Democratic Gubernatorial challenger 56 Republican Gubernatorial challenger 57 Democratic governor, no race in state 58 Republican governor, no race in state 61 Democratic governor--retiring (state with open race) 62 Republican governor--retiring (state with open race) OTHER: 90 Both Democratic and Republican candidates (used in incumbency var only) 97 Name given not on Candidate List MISSING DATA: 98 DK; refused to name candidate 99 NA 00 INAP ++VOTED OUTSIDE DISTRICT OF IW: DISTRICT WITH NO RUNNING INCUMBENT: (VOTE VAR ONLY) 81 Democratic candidate 82 Republican candidate DISTRICT WITH RUNNING INCUMBENT: (VOTE VAR ONLY) 83 Democratic incumbent 84 Republican incumbent 85 Democratic challenger 86 Republican challenger ALL DISTRICTS: (VOTE VAR ONLY) 80 Third party or independent candidate ** 91 Democrat--no name given 92 Republican--no name given ** IF 3RD PARTY/INDEPENDENT CANDIDATE NAMED, THIS CODE IS USED ONLY IF NAME APPEARS ON CANDIDATE LIST (IF NAME NOT ON CANDIDATE LIST, CODE 97 IS USED). NOTE: CODE 97 INCLUDES INSTANCES WHERE R VOTED STRAIGHT MAJOR PARTY TICKET BUT NO CANDIDATE FOR R'S PARTY RAN FOR GIVEN OFFICE (OR: R INSISTS VOTED FOR A MAJOR PARTY'S CANDIDATE BUT NO CANDIDATE RAN FOR GIVEN OFFICE REPRESENTING NAMED MAJOR PARTY). ++ CODES 80-86,91,92 ARE NOT USED IN VARS OTHER THAN VOTE VARS. GENERAL NOTE: IN THOSE QUESTIONS WHERE R IS NOT READ NAMES OF CANDIDATES BUT R SUPPLIES A CANDIDATE NAME OF HIS/HER OWN CONSTRUCTION [I.E., IN RECALL, 'MOST IMPORTANT PROBLEM IN DISTRICT' HOUSE CANDIDATE], RESPONDENTS SOMETIMES IN ERROR GIVE NAMES OF CANDIDATES FOR OTHER OFFICES OR NAMES OF NONRUNNING OFFICEHOLDERS. IF SUCH A NAME IS DETERMINED TO BE APPROPRIATE FOR R'S STATE/CD AND THE NAME IS CODEABLE FROM THE CANDIDATE LIST USED, WHEREVER POSSIBLE THE 'INCORRECT' NAME IS STILL CODED. (However, see ** for 3rd/party and independent candidates). [NOTE: If R names candidates from districts other than district corresponding to R's sample location, those candidates' codes are not coded--97 is used.] >> MASTER CODE BALLOT CARDS AND CANDIDATE LISTS CANDIDATE LISTS AND BALLOT CARDS - 1992 STATE: Alabama CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 13. Richard C. Shelby Democratic incumbent 16. Richard Sellers Republican challenger 19. Howell T. Heflin Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Glen Browder Democratic incumbent 36. Don Sledge Republican challenger ============================================================ STATE: Alabama CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 13. Richard C. Shelby Democratic incumbent 16. Richard Sellers Republican challenger 19. Howell T. Heflin Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Tom Bevill Democratic incumbent 36. Mickey Strickland Republican challenger ============================================================ STATE: Alabama CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 13. Richard C. Shelby Democratic incumbent 16. Richard Sellers Republican challenger 19. Howell T. Heflin Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Ben Erdreich Democratic incumbent 36. Spencer Bachus Republican challenger ============================================================ STATE: Alabama CONGRESSIONAL DISTRICT: 07 (A) NAMES FOR U.S. SENATE: 13. Richard C. Shelby Democratic incumbent 16. Richard Sellers Republican challenger 19. Howell T. Heflin Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Earl F. Hilliard Democratic candidate 32. Kervin Jones Republican candidate 41. Claude Harris Jr. Democrat--retiring ============================================================ STATE: Arizona CONGRESSIONAL DISTRICT: 01 (A) NAMES FOR U.S. SENATE: 15. Claire Sargent Democratic challenger 14. John McCain Republican incumbent 19. Dennis DeConcini Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Sam Coppersmith Democratic challenger 34. John "Jay" Rhodes Republican incumbent ============================================================ STATE: Arizona CONGRESSIONAL DISTRICT: 02 (A) NAMES FOR U.S. SENATE: 15. Claire Sargent Democratic challenger 14. John McCain Republican incumbent 19. Dennis DeConcini Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Ed Pastor Democratic candidate 32. Don Shooter Republican candidate 41. Morris K. Udall Democrat--retiring ============================================================ STATE: Arizona CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 15. Claire Sargent Democratic challenger 14. John McCain Republican incumbent 19. Dennis DeConcini Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Roger Hartstone Democratic challenger 34. Bob Stump Republican incumbent ============================================================ STATE: Arizona CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 15. Claire Sargent Democratic challenger 14. John McCain Republican incumbent 19. Dennis DeConcini Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Walter Mybeck Democratic challenger 34. Jon Kyl Republican incumbent ============================================================ STATE: Arizona CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 15. Claire Sargent Democratic challenger 14. John McCain Republican incumbent 19. Dennis DeConcini Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Karan English Democratic candidate 32. Doug Wead Republican candidate ============================================================ STATE: Arkansas CONGRESSIONAL DISTRICT: 01 (A) NAMES FOR U.S. SENATE: 13. Dale Bumpers Democratic incumbent 16. Mike Huckabee Republican challenger 19. David Pryor Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Blanche Lambert Democratic candidate 32. Terry Hayes Republican candidate 41. Bill Alexander Democrat--retiring ============================================================ STATE: Arkansas CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 13. Dale Bumpers Democratic incumbent 16. Mike Huckabee Republican challenger 19. David Pryor Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Bill McCuen Democratic candidate 32. Jay Dickey Republican candidate 41. Beryl Anthony Democrat--retiring ============================================================ STATE: California CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Patricia Malberg Democratic challenger 34. John T. Doolittle Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Lynn Woolsey Democratic candidate 32. Bill Filante Republican candidate 41. Barbara Boxer Democrat--retiring ============================================================ STATE: California CONGRESSIONAL DISTRICT: 07 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. George Miller Democratic incumbent 36. Dave Scholl Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 08 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Nancy Pelosi Democratic incumbent 36. Marc Wolin Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 09 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Ronald V. Dellums Democratic incumbent 36. Billy Hunter Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 10 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Wendell H. Williams Democratic candidate 32. Bill Baker Republican candidate ============================================================ STATE: California CONGRESSIONAL DISTRICT: 12 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Tom Lantos Democratic incumbent 36. Jim Tomlin Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 13 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Pete Stark Democratic incumbent 36. Verne Teyler Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 19 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Richard H. Lehman Democratic incumbent 36. Tal L. Cloud Republican challenger =========================================================== STATE: California CONGRESSIONAL DISTRICT: 24 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Anthony C. Beilenson Democratic incumbent 36. Tom McClintock Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 26 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Howard L. Berman Democratic incumbent 36. Gary Forsch Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 27 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Doug Kahn Democratic challenger 34. Carlos J. Moorhead Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 28 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Al Wachtel Democratic challenger 34. David Dreier Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 29 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Henry A. Waxman Democratic incumbent 36. Mark A. Robbins Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 31 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Matthew G. Martinez Democratic incumbent 36. Reuben D. Franco Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 32 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Julian C. Dixon Democratic incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 33 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Lucille Roybal-Allard Democratic candidate 32. Robert Guzman Republican candidate ============================================================ STATE: California CONGRESSIONAL DISTRICT: 34 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Esteban E. Torres Democratic incumbent 36. J. "Jay" Hernandez Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 35 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Maxine Waters Democratic incumbent 36. Nate Truman Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 36 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Jane Harman Democratic candidate 32. Joan Milke Flores Republican candidate 41. Mel Levine Democrat--retiring ============================================================ STATE: California CONGRESSIONAL DISTRICT: 38 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Evan Anderson Braude Democratic candidate 32. Steve Horn Republican candidate 41. Glenn M. Anderson Democrat--retiring ============================================================ STATE: California CONGRESSIONAL DISTRICT: 39 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Molly McClanahan Democratic candidate 32. Ed Royce Republican candidate 42. William E. Dannemeyer Republican--retiring ============================================================ STATE: California CONGRESSIONAL DISTRICT: 40 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Donald M. Rusk Democratic challenger 34. Jerry L. Lewis Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 41 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Bob Baker Democratic candidate 32. Jay C. Kim Republican candidate ============================================================ STATE: California CONGRESSIONAL DISTRICT: 42 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. George E. Brown Jr. Democratic incumbent 36. Richard B. Rutan Republican challenger ============================================================ STATE: California CONGRESSIONAL DISTRICT: 43 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Mark A. Takano Democratic candidate 32. Ken Calvert Republican candidate ============================================================ STATE: California CONGRESSIONAL DISTRICT: 44 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Georgia Smith Democratic challenger 34. Al McCandless Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 45 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Patricia McCabe Democratic challenger 34. Dana Rohrabacher Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 46 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Robert John Banuelos Democratic challenger 34. Robert K. Dornan Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 47 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. John F. Anwiller Democratic challenger 34. C. Christopher Cox Republican incumbent ============================================================ STATE: California CONGRESSIONAL DISTRICT: 48 (A) NAMES FOR U.S. SENATE: 11. Barbara Boxer Democratic candidate 12. Bruce Herschensohn Republican candidate 11a. Dianne Feinstein Democratic candidate 14a. John Seymour Republican incumbent (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Michael Farber Democratic challenger 34. Ron Packard Republican incumbent ============================================================ STATE: Colorado CONGRESSIONAL DISTRICT: 01 (A) NAMES FOR U.S. SENATE: 11. Ben Nighthorse Campbell Democratic candidate 12. Terry Considine Republican candidate 29. Hank Brown Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Patricia Schroeder Democratic incumbent 36. Raymond Diaz Aragon Republican challenger ============================================================ STATE: Colorado CONGRESSIONAL DISTRICT: 02 (A) NAMES FOR U.S. SENATE: 11. Ben Nighthorse Campbell Democratic candidate 12. Terry Considine Republican candidate 29. Hank Brown Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. David E. Skaggs Democratic incumbent 36. Brian Day Republican challenger ============================================================ STATE: Colorado CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 11. Ben Nighthorse Campbell Democratic candidate 12. Terry Considine Republican candidate 29. Hank Brown Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Tom Kolbe Democratic challenger 34. Dan Schaefer Republican incumbent ============================================================ STATE: Connecticut CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 13. Christopher J. Dodd Democratic incumbent 16. Brooks Johnson Republican challenger 19. Joseph I. Lieberman Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Rosa DeLauro Democratic incumbent 36. Tom Scott Republican challenger ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Corrine Brown Democratic candidate 32. Don Weidner Republican candidate ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Mattox Hair Democratic candidate 32. Tillie Fowler Republican candidate 41. Charles E. Bennett Democrat--retiring ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Phil Denton Democratic challenger 34. Cliff Stearns Republican incumbent ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 12 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Tom Mims Democratic candidate 32. Charles T. Canady Republican candidate 42. Andy Ireland Republican--retiring ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 17 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Carrie Meek Democratic candidate 41. William Lehman Democrat--retiring ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 18 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Magda Montiel Davis Democratic challenger 34. Ileana Ros-Lehtinen Republican incumbent ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 20 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Peter Deutsch Democratic candidate 32. Beverly Kennedy Republican candidate 41. Dante B. Fascell Democrat--retiring ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 21 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 32. Lincoln Diaz-Balart Republican candidate 41. Larry Smith Democrat--retiring ============================================================ STATE: Florida CONGRESSIONAL DISTRICT: 22 (A) NAMES FOR U.S. SENATE: 13. Bob Graham Democratic incumbent 16. Bill Grant Republican challenger 29. Connie Mack III Republican--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Gwen Margolis Democratic challenger 34. E. Clay Shaw Jr. Republican incumbent ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 01 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Barbara Christmas Democratic candidate 32. Jack Kingston Republican candidate 41. Lindsay Thomas Democrat--retiring ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 02 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Sanford Bishop Democratic candidate 32. Jim Dudley Republican candidate 41. Charles Hatcher Democrat--retiring ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Richard Ray Democratic incumbent 36. Mac Collins Republican challenger ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Cathey Steinberg Democratic candidate 32. John Linder Republican candidate 41. Ben Jones Democrat--retiring ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 05 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. John Lewis Democratic incumbent 36. Paul R. Stabler Republican challenger ============================================================ STATE: Georgia CONGRESSIONAL DISTRICT: 07 (A) NAMES FOR U.S. SENATE: 13. Wyche Fowler Democratic incumbent 16. Paul Coverdell Republican challenger 19. Sam Nunn Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. George "Buddy" Darden Democratic incumbent 36. Al Beverly Republican challenger ============================================================ STATE: Iowa CONGRESSIONAL DISTRICT: 04 (A) NAMES FOR U.S. SENATE: 15. Jean Lloyd-Jones Democratic challenger 14. Charles E. Grassley Republican incumbent 19. Tom Harkin Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Neal Smith Democratic incumbent 36. Paul Lunde Republican challenger ============================================================ STATE: Illinois CONGRESSIONAL DISTRICT: 01 (A) NAMES FOR U.S. SENATE: 11. Carol Moseley Braun Democratic candidate 12. Richard Williamson Republican candidate 19. Paul Simon Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Bobby L. Rush Democratic candidate 32. Jay Walker Republican candidate 41. Charles A. Hayes Democrat--retiring ============================================================ STATE: Illinois CONGRESSIONAL DISTRICT: 02 (A) NAMES FOR U.S. SENATE: 11. Carol Moseley Braun Democratic candidate 12. Richard Williamson Republican candidate 19. Paul Simon Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 31. Mel Reynolds Democratic candidate 32. Ron Blackstone Republican candidate 41. Gus Savage Democrat--retiring ============================================================ STATE: Illinois CONGRESSIONAL DISTRICT: 03 (A) NAMES FOR U.S. SENATE: 11. Carol Moseley Braun Democratic candidate 12. Richard Williamson Republican candidate 19. Paul Simon Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. William O. Lipinski Democratic incumbent 36. Harry C. Lepinske Republican challenger ============================================================ STATE: Illinois CONGRESSIONAL DISTRICT: 05 (A) NAMES FOR U.S. SENATE: 11. Carol Moseley Braun Democratic candidate 12. Richard Williamson Republican candidate 19. Paul Simon Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 33. Dan Rostenkowski Democratic incumbent 36. Elias R. Zenkich Republican challenger ============================================================ STATE: Illinois CONGRESSIONAL DISTRICT: 06 (A) NAMES FOR U.S. SENATE: 11. Carol Moseley Braun Democratic candidate 12. Richard Williamson Republican candidate 19. Paul Simon Democratic--term not up (B) NAMES FOR U.S. HOUSE OF REPRESENTATIVES: 35. Barry W. Watkins Democratic challenger 34. Henry J. Hyde