5
most midterm elections) thereafter. (NES, individual years, 1948-2000) Yet if the scope
and quality of the NES trumps all other contenders, it still brings with it huge problems
for any initial effort to isolate an issue structure:
•
Under-sampling of policy realms makes it hard to create measures in some
issue domains, no matter how great their potential consequence to postwar
political history.
•
Over-sampling of policy realms means that other domains can easily generate
their own measures, however minimal their potential consequence to the historical
record.
•
When over-sampling meets under-sampling, over-sampled domains may even
‘ingest’ the substance of the under-sampled, despite the fact that they have no
evident substantive connection.
•
Changes in question wording make it difficult to separate structural shifts
from mere item development.
•
The same question wording can still align its substance differently, depending
on what else is in the survey.
Any hope of surmounting these difficulties must begin with a theoretical
framework sufficient to discipline the analysis. In particular, this framework requires a
set of hypotheses drawn from postwar political history about the main domains of policy
conflict during this extended era. We believe that four major policy domains would
make nearly any list of priorities in postwar politics: social welfare, international
relations, civil rights, and cultural values. (Barone 1990, Chafe 1991, Patterson 1996)
We adopt these four as our theoretical frame.
These domains then require a set of definitions sufficiently clear as to assign
opinion items to them. In practice, working definitions for these policy realms need to
meet two criteria, when the evolution of an issue structure is the principal concern. From
one side, they must be able to specify items that do and do not fall within the overall
Convention