Response: Continuity in NES Studies

Abstract

If the subjects of our inquiries-the parties, issues, and candidates of each election period-were not so given to change, and if the theoretical and methodological foundations of our inquiries were not almost equally in flux, a straightforward replication of measurements through time might be relatively easy to achieve in sustaining the ongoing study of American national elections. Such is not the case. Those who design each new NES data collection are bombarded by a multitude of claims based in theoretical developments, improvements in instrumentation, changes in the electoral setting, and the emergence of new candidates promoting fresh issues. In the face of all these changes, perfect continuity in the details of the NES collections is neither realistic nor desirable. Nevertheless, we share with Professor Abramson a deep concern for over-time analysis and appreciate the careful job he has done in documenting a decline in the sheer number of questions from early Michigan studies that are repeated in identical form in current NES data collections. In some respects, we take his report as a compliment: we are pleased that the study documentation we try so hard to make both comprehensive and accessible has enabled him to pinpoint discontinuities in question wording and format. Such discontinuities will not be news to diligent readers of the regularly updated Continuity Guide to the American National Election Studies, produced by the Center for Political Studies and distributed by the ICPSR, nor to careful users of the recently published Data Sourcebook for the American National Election Studies, 1952-1986 (Harvard University Press, 1989). Both the Continuity Guide and the new Data Sourcebook corroborate Abramson's conclusion regarding the frustration likely to be experienced by researchers interested in tracing the evolution of public opinion on matters of government policy. Still, we seem to differ with Abramson over how to establish the comparability of measures. Continuity in survey measures cannot be assured through mechanical repetition of questions, word-for-word as