2
hand, individual-level analysis cannot easily delineate the specific impact of government policy
on the individual’s decision to bestow his or her vote as a reward.
To conclude this tale of disciplinary woe, the post-Kramer literature has yet to resolve
the pocketbook-sociotropic question despite important methodological advances, such as
analysis exploiting pooled cross-sectional times series data (e.g., Rivers 1986; Markus 1988)
and analysis investigating the dependence of pocketbook voting on other cognitive orientations
(Kinder and Mebane 1983; Feldman 1984). The problems blocking further progress go
beyond the purely statistical question of whether individual-level data can be used to
distinguish systematic government impact from idiosyncratic noise. Even if the relative
contribution of sociotropic factors to political support were properly estimated, we would still
be in the dark about the motivations behind the orientations we uncovered.
For suppose, as is standard, that sociotropic voting is measured as a statistical
association between voting and an expressed concern with national economic conditions. As
Kinder and Kiewiet understood, this concern might not be altruistic but instead might reflect an
enlightened prediction about the respondent’s future. Thus complicating any verdict about the
meaning of sociotropic voting is the fact that aggregate economic conditions are aggregates of
individual conditions and have implications for each individual’s prospects. Even the most
self-interested rational voter would be sociotropic in part.
As a result, sociotropic influences on voting cannot be interpreted independently of
pocketbook influences. Interpreting pocketbook influences, in turn, requires understanding
how policy affects the individual’s economic condition and translates into political preferences.
In other words, when pocketbook and sociotropic influences on voting are conceived simply as
Convention