Two problems have been noted with the variables w11ze3a and w11ze3b, one concerning the variable labels and one concerning the missing data codes.
The variable label descriptions for w11ze3a and w11ze3b are reversed in the 20100903 version of the dataset.
The correct description for w11ze3a is "How many more opinions does R have".
The correct description for w11ze3b is "How many fewer opinions does R have".
In SPSS the labels can be fixed by running the following code:
var lab w11ze3a "w11ze3a. How many more opinions does R have".
var lab w11ze3b "w11ze3b. How many fewer opinions does R have".
The nonresponse codes for w11ze3a and w11ze3b fail to differentiate item nonresponse from cases that legitimately skipped the question because it was inapplicable. ze3a and ze3b show data from followup questions to ze2 ("Compared to the average person do you have fewer opinions about whether things are good or bad, about the same number of opinions, or more opinions?"). ze3a should have the -1 "inapplicable" code for respondents whose answers to ze2 were anything but "more." ze3b should have the -1 code for respondents whose answers to ze2 were anything but "fewer."
The following SPSS code will fix the non-response codes.
missing values w11ze2 w11ze3a w11ze3b ().
if w11ze2=-7 or w11ze2=1 or w11ze2=2 w11ze3a=-1.
if w11ze2=-7 or w11ze2=2 or w11ze2=3 w11ze3b=-1.
missing values w11ze2 w11ze3a w11ze3b (-9 thru -1).