I’ve been remiss in following up on my promise to discuss political science research on campaign finance. Fortunately, Nick Seabrook, a Ph.D. candidate in political science at the University of Buffalo, sends a back-of-the-envelope analysis that speaks to the potential consequences of the Citizens United decision:
bq. In following the coverage of Citizens United in the mainstream media and blogosphere, two major criticisms of the decision have become apparent: (1) that the massive influx of corporate money into the electoral process will effectively allow corporate sponsorship to dictate election outcomes (let’s call this the McCandidate hypothesis); and (2) that this surge in independent corporate expenditures will disproportionately benefit Republican office seekers (the Fat Cat hypothesis).
bq. While testing these propositions on federal elections will require waiting at least until the outcome of the 2010 midterms, state legislative elections offer an ideal quasi-experiment for analyzing the implications of the decision (24 states ban independent corporate expenditures in state races, 26 do not).
bq. For another project, I put together complete election data for 2002 upper and lower house state legislative districts in 9 states, 4 that ban corporate expenditures (AZ, KY, NY, SD) and 5 that allow them (CA, FL, MD, NV, OR). We can test these two hypotheses by running some simple regressions. Here, state legislative election results are modeled as a function of incumbency, district partisanship, and candidate expenditures.
bq. We can indirectly test the McCandidate hypothesis by splitting the sample between states that ban corporate electioneering and those that do not. If independent corporate expenditures have a significant effect, an observable implication of this would be that these other factors (incumbency, party, and money) should have less of an impact. However, there is no evidence that this is the case – the models explain just under 90% of the variation in the two-party vote in each case, and the coefficients and significance levels are pretty much the same (a Chow test confirms this). Elections appear to work similarly whether a state allows independent corporate expenditures or not, and if corporate money is affecting election results in these states, it is doing so at the margins.
bq. What about the Fat Cat hypothesis? Recombining the data into a single model and introducing a dummy variable for states that allow corporate expenditures, the results are surprising. All else being equal, allowing independent corporate money actually works to the detriment of Republican candidates to the tune of between 2% and 3% of the two-party vote (b=-2.3; s.e.=.6). If we exclude New York from the sample (NY places restrictions on corporate spending but does not ban it outright), the coefficient is statistically insignificant. Similarly, if we run a logit model where the dependent variable is which candidate wins rather than their vote shares (same independent variables), the corporate dummy again fails to reach statistical significance. From these data, there is simply no evidence that Republican candidates are benefiting from corporate campaign involvement at the state level.
bq. Of course, some caveats apply: we can’t really generalize too much from a sample of just 9 states in a single election year, and it’s unclear whether the dynamics of corporate electioneering may be different at the federal level than they are in state elections. Nevertheless, these results suggest that we should be cautious about simply assuming that Citizens United will inevitably produce wholesale changes in the functioning of our electoral system