In yesterday’s post, I described the data and methods that Chris Warshaw and I use in a new paper on the effects of partisan gerrymandering on parties’ associational activities. Today I’ll discuss our findings and their implications. In a nutshell, parties that are victimized by gerrymandering are impaired in their performance of several key associational functions. These handicaps, moreover, are substantively large, statistically significant no matter how gerrymandering is measured, and roughly equal in size at the congressional and state house levels. Justice Kagan’s speculation in Whitford about the associational harms of gerrymandering thus appears to be entirely correct.
Start with our first outcome variable: the difference between the parties’ respective shares of contested seats. When our aggregate measure of gerrymandering increases by one standard deviation, the disadvantaged party contests three percentage points fewer state house seats relative to the advantaged party. The point estimates are very similar at the congressional level and if gerrymandering is assessed using a particular (rather than an aggregate) metric.
Next consider our other proxy for candidate recruitment: the difference between the overall quality of each party’s candidates. A one-standard-deviation rise in our aggregate gerrymandering metric is associated with roughly a nine-percentage-point decline in the relative quality of a party’s candidates. A party that is the victim of gerrymandering, in other words, runs many fewer incumbents and quality challengers than its opponent.
The story is much the same with fundraising. When our aggregate gerrymandering metric goes up by one standard deviation, the share of total campaign contributions received by a party’s candidates goes down by about five percentage points. Donors, that is, give substantially less money to candidates with lower odds of winning seats and capturing a legislative majority.
Lastly, our results for ordinary voters (as opposed to candidates and donors) are smaller in scale but still statistically significant. A one-standard-deviation increase in our aggregate gerrymandering metric is linked to around a half-percentage-point drop in a party’s statewide vote share. Thus through some combination of lower turnout and actual conversion, voters become somewhat less likely to support a party that is disadvantaged by a district map.
In our view, our findings strongly confirm Justice Kagan’s hypothesis in Whitford. She predicted that backers of a “‘disfavored party’” would “face difficulties fundraising . . . generating support from independents, and recruiting candidates to run for office”—and that is precisely what our analyses showed. Accordingly, our findings should be helpful to plaintiffs currently pursuing associational challenges to district plans. To date, these litigants have relied primarily on qualitative testimony from voters, candidates, and party officials about how their associational activities have been impeded. This evidence may now be complemented by our data-driven conclusion that, across many states and years, partisan gerrymandering systematically undermines party health.
We also think our findings are relevant to political scientists. Until now, they have mostly focused on other gerrymandering issues, like measuring the concept and distinguishing it from the impact of political geography. These are certainly important topics, but unfortunately they have overshadowed the equally vital question of how gerrymandering affects the rest of our political system. Our paper makes progress toward answering this question by exploring the relationships between gerrymandering and various associational functions performed by parties. But this literature is still in its infancy—despite Justice Kagan’s plea for exactly these sorts of studies—and much more work remains to be done.