1994 Time Series Study
About the Dataset
- Type of study: Time Series Study
- Sample universe: U.S. eligible voters (cross-section)
- Sample composition: fresh cross-section cases; cross-section panel component
- Number of waves: 1 (post-election)
- Modes used: face-to-face
- Instrument format: CAPI
- Number of dataset respondents: 1,795
- Weights: V940004, V940005, V940006
- Additional files: 1994 Behavior Recording Experiment File
- Associated studies or derivative files: ANES 1992-1997 Merged File
Study Content Highlights
In addition to content on electoral participation, voting behavior, and public opinion, the 1994 ANES contains questions in other areas such as political knowledge, evaluation of parties, and values and predispositions. Special-interest and topical content includes new questions on welfare reform and the Clinton crime bill, as well as new questions in the battery on respondent health care and health care concerns. In addition, with particular benefit to the panel aspect, some emphasis is placed on candidate, incumbent, and presidential evaluations, including performance questions and issue placements.
Study Design Highlights
Besides being a stand-alone time series data collection in which the full sample comprises a representative cross-section, the ‘panel’ component of the cross-section sample the 1994 Time Series study also re-interviews respondents from the ANES 1992 Time Series Study, many of whom were also interviewed in the ANES 1993 Pilot Study. The empanelment was specifically designed to exploit the special features of elections and politics during the 1992-1994 time period, in particular through the careful inclusions of evaluations at the congressional and presidential levels
Interviewing began immediately after the November 8 elections and continued until January 9, 1995. The interviews, which averaged 72 minutes, were conducted face-to-face except in 36 cases where administration was possible only by telephone, primarily for panel component respondents who had moved “out of range”. A ballot-card experiment resulted in two randomly-assigned versions or “forms” of the questionnaire, which corresponded to alternative designs (though identical content) of the ballot card. Questionnaires were otherwise undifferentiated. Special features of the study design included use of “certainty” follow-ups on liberal-conservative scale placements, and the coding of specific probes used by interviewers for many individual questions.
The 1994 sample consisted of two components. The 1994 panel component included all fresh cross-section respondents from the 1992 Time Series study (respondents first interviewed in 1992) who had provided both pre-election and post-election interviews in 1992, for a total of 1,005 panel eligible cases. The 1994 fresh cross-section component provided an additional 1,436 eligible cases. The 1994 eligible total of 2,441 cases yielded 1,795 completions (759 panel, 1,036 fresh cross-section).
V940004 and V940005 are the weights originally provided with the 1994 Time Series final data release. V940004 was created for optional use with the full sample data, incorporating a household nonresponse adjustment factor, the within-household selection weight, and a post-stratification factor by age, sex, and region. Panel weight V940005 was constructed for longitudinal analysis when using only the panel component of the 1994 sample. In addition, for users comparing either the panel cases or the total sample to previous unweighted studies, the “time series” weight V940006 was added to the release to adjust for panel aging and attrition only.
In 1997, V940006a was constructed as an additional full sample weight to correct for an underrepresentation of younger and less educated respondents, and the resulting contribution to voting overestimation, which were discovered as a consequence of the long-term empanelment across the Time Series studies 1992-1994-1996 (similar weighting was also constructed for the 1996 Time Series study). V940006a adjusts for household nonresponse, weights for within-household selection, and is post-stratified by age and education. NOTE: it is not intended that analysis be undertaken using the 1994 fresh cross component alone.
A Behavior Reporting Experiment accompanied the 1994 Time Series Study during an identical field period, and was administered to test methods for reducing over-report of voter turnout and church attendance. In 10-minute interviews conducted with 700 respondents selected from voter registration lists, 4 questionnaire versions were administered containing different combinations of standard Time Series or experimental question wordings for turnout and church attendance. The actual turnout of each respondent was verified using official city registration and voting records. Data for the ANES 1994 Behavior Reporting Experiment File are available from the Data Center.
Associated Studies or Derivative Files:
An ANES derivative file is a file which merges together related ANES studies (having common questionnaires or cases), for the convenience of users.
The ANES 1992-1997 Merged File is an ANES derivative file that comprises merged data from 7 ANES Time Series and Pilot studies. Over years 1992-1997, ANES conducted a process of empanelment over the ANES 1992 Time Series Study, the ANES 1993 Pilot Study, the ANES 1994 Time Series Study, the ANES 1995 Pilot Study, the ANES 1996 Time Series Study, and the ANES 1997 Pilot Study. The 1994 Time Series study re-interviewed cases from the ‘fresh’ component of the 1992 Time Series study sample as part of its overall cross-section, which also included ‘fresh’ 1994 cases. The 1996 Time Series study similarly combined in its cross-section re-interviews of respondents from the preceding two Time Series studies together with ‘fresh’ cases. The 1993, 1995, and 1997 pilot studies each re-interviewed a set of cases from the preceding Time Series Study, although the 1995 pilot study only included cases from the ‘fresh’ component of the 1994 Time Series sample. The ANES 1992-1997 Merged File is downloadable from the Data Center.