Using the Error in Pre-election Polls to Test for the Presence of Pork

Polls are used by politicians, voters, and businesspeople alike to form expectations over political outcomes. In an electoral system with ‘pork’ at the national level, local economic prospects will be a direct function of political support. Because every poll comes with its margin of error, we can exploit the error in sub-national polls to test for whether forward-looking investment in the economy adjusts to shocks in local political behavior. The covariance between the strength of this adjustment and local-level characteristics allows us to identify the locations in which strong ‘pork’ contracts exist. To illustrate the technique, we provide a theoretical foundation and an empirical exercise using data on Ugandan micro-entrepreneurs during the 2001 presidential election. We find evidence of transfers which are explicitly conditional on electoral support, and that businesses are most sensitive to surprise voting patterns in ‘core’ districts.