What We Talk about When We Talk about Campaign Spending

While research on the effect of candidate spending on voting outcomes has been
extensive, researchers have been forced to rely on summaries of candidate spending
which includes pre-election spending, fundraising expenses, and numerous other ex-
penditures that are not tied directly to the general election success of a candidate.
Using transaction-level data on candidate disbursements, I replicate the specificiations from several of the most influential papers, using alternative measures of candidate spending. Ifind that eliminating pre-election spending, and limiting the measure of candidate spending to only types of spending that are used directly on the candidate's  own campaign, leads to signicantly higher estimates of the effect of candidate spending, and the disparate effect of challenger and incumbent spending largely disappears. Then, utilizing the 2012 American National Election Survey, which asked about voter intent before the election and actual voting behavior after, and candidate disbursement  data which allows me to track candidate spending throughout the campaign, I can determine the effect of candidate spending on individual voting behavior under very weak identifying assumptions. I find that
spending by candidates on advertising and campaign events increases the likelihood
that voters will change their preferences in favor of that candidate and this effect is
particularly strong among voters who did not participate in the 2012 Presidential primary election.