4
able to protect or increase their future consumption. The implied functional relationships
between parties and utility have testable implications for individual voting behavior.
One important task then is to spell out the way in which income translates into
consumption. The model developed here provides a framework for measuring the impact of
partisan government on individual consumption. To do this, as just noted, I abandon a crucial
assumption motivating much of the economics of voting literature, namely, that in a voting
regression the coefficient on the relevant income variable is a fixed parameter. The standard
econometric estimate ignores the restriction that income itself is a stochastic process partly
reflecting government policy. In this sense the effort to find a parameter invariant to that
stochastic process is simply misdirected. The rules rational individuals use to maximize their
consumption are not fixed but are contingent on the processes determining their income (see
Lucas 1976).
For the actual analysis, I supplement data from the Panel Study of Income Dynamics
(PSID) with appropriate political variables capturing partisan variations in the national
government. This measurement exercise generates a regressor reflecting partisan impact on
utility. This regressor allows estimation of the vote function by matching the appropriate value
of the relevant individual and household characteristics from the PSID with characteristics
sampled by the American National Election Studies (NES). On this basis I compare the
political impact of government induced changes in individual economic conditions with the
political impact of aggregate economic conditions.
I do not allege that the resulting estimates are substitutes for the usual survey-response
measures since, as already indicated for sociotropic voting, it is not clear what the measures
Convention