The ‘political budget cycle’ is the technical name for the way incumbent politicians manipulate fiscal policy to try to get re-elected; there is abundant evidence for its existence (Drazen 2008a). In a refinement of this, an incumbent president might wish to use his power to boost spending in swing states. As Drazen (2001) writes: "Aggregate economic conditions before an election, specifically per capita output or income growth (and to a lesser extent inflation), have a significant effect on voting patterns in the US and other countries", which suggests that presidents can perhaps boost the economy and therefore boost votes.
The fundamental question is whether voters reward such spending. In other words, could incumbent presidents "purchase" important votes by diverting resources towards swing states at the expense of other states?
US voters as fiscal conservatives
The answer is not as obvious as it may seem, since voters have been shown to be fiscally conservative. Peltzman (1992) studies how US voters respond to shifts in government spending. Using a happiness index to control for macroeconomic conditions, he examines how various cyclically-adjusted federal-spending variables in the periods before the election affect the two-party vote share of the incumbent president (or the president’s party’s candidates).
Fiscal conservatism would be captured by a negative coefficient on the federal-spending variable. Of course, median-voter theory generally suggests no benefit or punishment for spending (as it reflects the median voter's preference), but many commentators assume that states would reward a president for bringing federal spending to their state.
Peltzman (1992) finds that:
- Voters penalise the president's party for federal spending during his term.
- It is not the case that voters "like spending but dislike taxes".
- A dollar is a dollar, in the sense that different types of spending yield the same effect.
Brender (2003) and Drazen and Eslava (2005) similarly find that voters are fiscally conservative in Israel and Colombia, respectively.
Conservative in aggregate but not at the state level?
There are two main problems with this approach, however.
- First, Peltzman includes a lagged dependent variable alongside state-fixed effects meaning that his estimates are biased.1
- Second, the regressor may not be exogenous. For example, if incumbent presidents feel that they are likely to lose, they might spend more money in the hope of winning more votes. It might be that such a reverse causality accounts for the negative finding in the Peltzman regressions.2
In work in progress (Giavazzi and McMahon 2012), we re-examine the effect of federal spending on two-party vote share outcomes using a measure of government spending that varies across states. In particular, we use Department of Defence Prime Military Contracts; these are the same data used to study the macroeconomic effects of fiscal policy in Nakamura and Steinsson (2011).3
Our state-level spending variables measure all military purchases with a value greater than $10,000 (from 1966 to 1983) and greater than $25,000 (1983 to 2006). The variables cover all sorts of spending with private contractors; for example, about 10% (on average nationally) is guns, missiles, and bullets, while about 50% is services such as architects, engineers, or someone to clean the gutters. We use data from all 50 states as well as DC, and we normalise the change in state-level spending by state GDP.
The key to obtaining consistent estimates of the effect of government spending on voting outcomes is getting suitable instruments for government spending which purge it of its correlation with the omitted variables. The main endogeneity worry is that the variation in fiscal spending may not be random across states even if aggregate military spending is (as discussed in numerous studies on fiscal policy). One particular concern is that a president and his staff may have allocated private military contracts precisely to those states which are doing relatively poorly economically (and which, as a result, he is expected to lose votes in). To address this we use variation in aggregate spending multiplied by the average allocation to each state; this means that our identification comes off the fact that aggregate variation in defense spending has a bigger effect in some states than in others. We can also instrument for aggregate variation in federal defense spending using Ramey’s (2011) defense news variable. Regardless of the approach we use, we obtain a very similar pattern of shocks across states and the results are robust to using any of the measures of state fiscal shock.
Given these data, we can control for state-, time-, and president-fixed effects to capture the average two-party vote share in a given state and to allow for election-specific movements across states in each election period. In our analysis we measure increases in state s (exogenous) federal defense spending on private military contracts and look at those increases that are large relative to the normal spending in that state and also large relative to increases in other states in that election period. We then check whether these increases lead to increases in the two-party vote share for the incumbent relative to the normal vote share in state s, controlling for increases in vote share common across all states in that election.
The main finding is that we find no evidence that voters punish the president when their state benefits from spending under his control. A one level, these results indicate that voters are not that fiscally conservative; they don’t punish the president for spending more on them (though they may still punish the president for spending more in general).4 Equally, however, we find no evidence that voters reward presidents from such disproportionate gains in spending, which is an important point for presidents and their staff, and the media, to note. As such, at least via more federal spending on private contracts in a given state, it does not seem like Obama could buy votes in swing states such as Ohio using his power as the incumbent.
Brender, A (2003) “The effect of ﬁscal performance on local government election results in Israel: 1989–1998”. Journal of Public Economics 87, 2187–205.
Drazen, A (2001) “The Political Business Cycle After 25 Years”, in NBER Macroeconomics Annual 2000, Volume 15, edited by B. S. Bernanke and K. Rogoff, MIT Press.
Drazen, A (2008a) “Political Budget Cycles,” in S.N. Durlauf and L.E. Blume, eds., in The New Palgrave Dictionary of Economics (2nd edition), London and New York: Macmillan Palgrave, 2008.
Drazen, A and M Eslava (2005) “Electoral manipulation via voter-friendly spending: theory and evidence”. NBER working paper 11085. Cambridge, MA: NBER.
Giavazzi, F and M McMahon (2012) ”How Fiscally Conservative are US voters?”. in progress, 2012.
Levitt, S D and J Snyder Jr (1997) “The Impact of Federal Spending on House Election Outcomes”, Journal of Political Economy, University of Chicago Press, vol. 105(1), pages 30-53, February.
Nakamura, E and J Steinsson (2011) “Fiscal Stimulus in a Monetary Union: Evidence from US Regions”, NBER Working Papers 17391, National Bureau of Economic Research, Inc.
Peltzman, S (1992) “Voters as Fiscal Conservatives”, The Quarterly Journal of Economics, MIT Press, vol. 107(2), pages 327-61, May.
1 Nickell (1981) demonstrated that the fixed-effects estimator for autoregressive panel data models is inconsistent. He shows that if the coefficient on the lagged dependent variable is positive (as Peltzman finds), then the bias on the coefficient for an exogenous variable is biased in a direction determined by the sign of the regression relation between the exogenous variable and the lagged dependent variable.
2 In fact, we can show using quarterly data that there is a negative and significant correlation between presidential approval and different measures of federal spending shocks. While this could reflect fiscal conservatism, it could equally reflect government responses to faltering popularity.
3 Levitt and Snyder (1997) also instrument for federal spending to investigate the effect on votes for members of Congress rather than the president.
4 Because of time-fixed effects, any punishment of the president common across all states would be picked up in the time fixed effects.