3
regressors in a statistical analysis, their own possible relation is merely a source of statistical
imprecision in estimates of their impact. But their relation is not a nuisance: it goes to the
heart of how individuals make political decisions.
Kramer, then, is perfectly correct to insist that the natural way to interpret the
connection between economic conditions and voting is in terms of causal connections between
government policy and individual circumstances. Yet for that very reason he is wrong to argue
that aggregate-level analysis can substitute for individual-level analysis. The causal ambition
of aggregate-level analysis is fully realized only at the individual level, by specifying the
consumption function with regard to the economy and the vote function with regard to politics.
Thus Kramer’s (1983) attempt to estimate a parameter translating government-induced changes
in individual income into votes is misdirected. A correct estimation procedure would
recognize the cross-equation restrictions between a model of the government’s impact on
individual income and a model of the way that impact translates into votes.
To put the point another way, Kramer’s arguments concerning causality are in tension
with his arguments concerning the relative autonomy and superiority of aggregate-level
analysis. Accordingly, this paper employs a different method for sorting out individual and
aggregate effects. It explores the political consequences of economic conditions by embedding
individuals in a model political economy determining their material and, in turn, political
interests. The premise underlying the model is simply an elaboration of the standard
pocketbook thesis. Income ultimately matters to people because it improves their consumption
or the consumption of those they care about. This indirect concern with income motivates
their political behavior. Specifically, voters are modeled as supporting the party seen as best
Convention