ANES

TIME SERIES CUMULATIVE DATA FILE

How to Cite
DATA
Return to Data Center
Errata
CODEBOOK
Introduction pdf text
Variables pdf text
Appendices pdf text
 

ABOUT THE DATASET:
In the ANES Time Series Cumulative Data File, the project staff have merged into a single file all cross-section cases and variables for select questions from the ANES Time Series studies conducted since 1948. Questions that have been asked in three or more Time Series studies are eligible for inclusion, with variables recoded as necessary for comparability across years.

Important: The  ANES Time Series Cumulative Data File data cannot be used with codebooks other than its own.

STUDY OVERVIEW:
Questions that have been asked in three or more ANES Time Series studies may be included in the ANES Time Series Cumulative Data File.  Data from other ANES studies, such as the 1984 Continuous Monitoring Study, the 1988 Super Tuesday Study, or the 1988-90-92 Senate Election Study, are never included.

Because each variable in the Time Series Cumulative Data File incorporates data for the same question from each of the ANES surveys, the file is particularly useful in service to three kinds of analysis: 1) analysis that focuses on over time change in citizens, in their individual characteristics, in the opinions they hold, and in their political behavior; 2) analysis that looks at subgroups of citizens that are represented by few cases in a single, cross-section sample, but by many more cases when several samples are combined; and 3) analysis that is concerned with replicating results over several elections. For these types of analyses, the chief advantage of relying on the Time Series Cumulative Data File, as opposed to combining, on one's own data from several National Election Studies, is that in constructing the Time Series Cumulative Data File, the ANES Project Staff have already gone through the trouble of recoding variables so that the same question has the same variable number and the same coding scheme for each of the Election Studies. A great deal of effort has gone into checking and verifying these recodes.

Those who use the Time Series Cumulative Data File should keep two things in mind: 1) the wording of questions occasionally changes over time to reflect changes in the political context in which the question is being asked. The ANES Project Staff have done their best to document in the codebook any over time differences in question wording that have occurred; 2) even when a question is worded identically in successive surveys, analysts may still wish to examine the placement of the question in each questionnaire to ensure that changes in its placement are not contaminating one's results.

SAMPLES FOR ELECTION STUDIES IN THE TIME SERIES CUMULATIVE FILE:
Over the years, the most common ANES study design has been a cross-section, equal probability, sample. These designs are typically "self-weighting" -- i.e., the respondents do not need to be weighted to compensate for unequal probabilities of selection in order to restore the "representativeness" of the sample. On several occasions, however, ANES has departed from this standard design. In some years, ANES "oversampled" certain groups (African-Americans in 1964, for example). In other years, the Election Study combined a panel reinterview with a cross- section design (as in 1974, for example). It is important to understand that the Time Series Cumulative File is a file of pooled cross- section studies: any respondent for a particular study who is strictly "panel" or "supplement" has been deleted from the Time Series Cumulative File. For example, the sequence of studies 1972, 1974, and 1976, constitutes a panel, with cross-section respondents in 1972 and 1974 being reinterviewed in succeeding years. If a 1972 respondent moved out of the SRC sampling area, but was nevertheless reinterviewed, that respondent became a "panel only" respondent, and the representativeness of the 1974 cross-section was maintained by selecting a new respondent from the residents at the sample address from which the "panel only" respondent had moved. Such 1974 "panel only" respondents are not included among the 1974 respondents that appear in the Time Series Cumulative Data File.

Because not all of the cross-section samples included in the Time Series Cumulative Data File are equal probability and thus self- weighting, all pooled cross-section descriptive analyses should be run using Variable 9, the weight variable. For most years, the value of that variable for all respondents is simply "1.0"

The table below lists the cross-section sample sizes, weighted and unweighted, for each Election Study included in the Time Series Cumulative Data File.

Table 1

THE SAMPLE SIZES FOR ALL YEARS ARE AS FOLLOW:

Cross-section *
  WeightedUnweighted
1948: -- N=662
1952: -- N=1899
1954: -- N=1139
1956: -- N=1762
1958: N=1822 N=1450
1960: N=1954 N=1181
1962: -- N=1297
1964: -- N=1571
1966: -- N=1291
1968: -- N=1557
type 0*1970: -- N=1507
type 1*1970: N=835 N=758
type 2*1970: N=817 N=749
1972: -- N=2705
1974: N=2523 N=1575
1976: N=2869.5N=2248
1978: -- N=2304
1980: -- N=1614
1982: -- N=1418
1984: -- N=2257
1986: -- N=2176
1988: -- N=2040
1990: -- N=1980***
1992: N=2488 N=2485
1994: **** N=1795
1996: **** N=1714
1998: 1281 N=1281
2000: 1807 N=1807
2002: 1511 N=1511
2004: 1212 N=1212

*     note: the 1970 figures exclude 73 non-eligible Rs in the original dataset's cross-section N. The Time Series Cumulative File excludes all non-eligible respondents from its cross-section. For descriptions of type 0, type 1, and type 2 variables in 1970, see weight vars V9-11.

**     note: the weighted cross-section Ns are represented in the Guide to Public Opinion and Electoral Behavior, which was produced using data from the Time Series Cumulative File. To reproduce the data appearing in the Sourcebook, it is necessary to use appropriate weights (see V9-11).

***     note: 20 cases have been deleted from the 1990 Study data due to belated discovery of interview fabrication and ineligible Rs.

****    note: In 1994, 1996 and 2000 there are multiple weights which can be used with the data; see Study documentation.