Top Two Impact Limited by Ballot Dropoff
As I am preparing to head down to my old stomping grounds at the Institute of Governmental Studies at UC Berkeley for their election post-mortem, one panel that caught my eye is “The Top-Two: So What?". Most analysis has been on whether the new system is living up to its authors’ promises of moderating the Legislature (maybe) and engaging independent voters (not yet).
Precious little has been invested in answering practical questions important to practitioners about the top-two beginning with: In contests that feature two candidates of the same party, how often do voters of the opposite party choose a candidate and how often do they skip the contest?
It is an important question. Intraparty general elections did not exist before top-two. They made up one in six of the contests in November. Faulty assumptions on crossover rates can shape outcomes. If pollsters include too many or too few members of the opposing party, it will throw off their projections. Campaigns can waste precious resources trying to appeal to voters who are not going to vote for either candidate.
To answer this question we first need to understand a bit about voter behavior. Why would a voter make a proactive decision NOT to vote in a contest when they already have a ballot and pen in their hands? Most voters fall into a pyramid of partisanship when it comes to crossover voting:
- Loyalists- Will only vote for candidates of their party.
- Oppositionists- Will vote for anyone other than members of the other major party.
- Crossovers- Will vote for member of other parties if there is no member of their party on the ballot.
- Independents- Will vote for candidates regardless of party. Note Independents here refers to a form of voting behavior, not registering as having no party preference or decline to state.
Consider Republicans faced with a Dem vs. Dem contest. Some will just move on when they do not see an “R” (Loyalists). Others could have stomached a candidate with no party preference but cannot bring themselves to vote for Obama’s party (Oppositionists). Still others will vote for the more moderate Democrat when they have to (Crossovers). Some voted for a mix of candidates from different parties even before top-two (Independents).
These categories are not absolute and the circumstances of individual campaigns matter. A Latino may be less partisan when evaluating Latino candidates. That said these broad categories are reflected in dropoff statistics, the difference between the number of voters who cast a ballot in a district and the number of votes cast for candidates in that contest. There were 153 Congressional and Legislative contests held in November falling into four categories:
The lowest dropoff contests are those where voters can choose between candidates of the two major parties and as partisan options decrease dropoff increases. Notably the highest dropoff in any traditional general was 8.5%. Only one-fifth of intraparty general had lower drop-offs than this high watermark for traditional generals. Almost exclusively these were competitive Dem vs. Dem elections in districts where Republican registration is so anemic that high interest levels among Democrats trumped any dropoff among Republicans.
While it remains unclear whether top-two will ever increase overall turnout, it is already very clear that intraparty generals can reduce participation in individual contests. How much? Traditional elections can help provide the answer. Dropoff rates for most voters are probably not affected by an intraparty general. Democrats facing two Democrats will pick one. Minor party and unaffiliated voters are used to sometimes having to pick candidates from one of the major parties. Thus the bulk of the observed increase in dropoff in these elections likely comes from members of the major party not represented. Thanks to data from Political Data we know how many people voted in each district and their partisan associations.
We do have to make some assumptions about how voters not affiliated with one of the major parties behave. If one assumes these voters are more likely than members of the major parties to dropoff in a traditional primary, than you have to assume these voters will also be more likely to dropoff in an intraparty primary. Thus two models are used for calculating dropoff in traditional primaries. The first assumes these unaffiliated voters are no more or less likely to dropoff than other voters and thus had the same dropoff rate as the average: 4.6%. The second assumes that no members of the major parties dropped off in traditional primaries and any voters skipping a contest were unaffiliated or minor party voters. This would translate to an average dropoff rate for these voters of 21%. The truth is almost certainly somewhere in the middle. A more complex model could refine these assumptions, but these extremes provide bookends for this analysis.
With this in place we can now calculate how much ballot dropoff came from the party with two candidates on the ballot, how much from the major party with none, and how much from others. For those who enjoy algebra, to calculate the percentage of voters who crossed over:
[(Total Voted - Votes Cast in Contest) – (Total Voted Represented Major Party * Traditional Primary Dropoff Rate) - (Total Voted Unaffiliated and Minor Party * Traditional Primary Dropoff Rate)] / Total Voted Unrepresented Major Party
This formula suggests on average 53-65% of members of the unrepresented major party crossed over to cast a ballot. There are important caveats. The model does not consider other factors that a cursory review suggests also correlate to dropoff rates. For example as income goes up dropoff rates appear to go down. And as discussed, dropoff rates for unaffiliated voters could be calculated more precisely. However the data still strongly suggests that, when faced with a choice between two members of the opposition party, at least a third of voters choose not to make a decision.
These numbers provide important guideposts to practitioners. If a poll shows less than half of members of the unrepresented party participating, it may be skewed. Conversely if a campaign plan assumes getting more than two-thirds of the voters from what would normally be the opposition, it it probably is overly optimistic.
It is interesting to note some of the factors that do and do not appear to influence these crossover rates. There was little difference in dropoff rates between competitive and uncompetitive general elections. Further research is warranted, but trying to convince voters of the benefits of crossing over may be a relatively inefficient task. Rather the focus should be on those already inclined to do so.
While the sample size is small, crossover rates were higher for Congressional contests than state Legislative contests. Perhaps this is reflective of the importance the average voter places on each. Data was inconclusive on whether Republicans showed a potentially greater willingness to crossover than Democrats. Socioeconomic factors may be more important and clouding the data.
These relatively simplistic calculations are merely a first step. By applying some of the same logic to precinct level results one could control for more factors. This would allow for more refined conclusions about crossover rates and may even allow for individualized rankings of voters’ likelihood to crossover in intraparty elections. Such calculations would provide an important tool to practitioners. A message to an undecided voter with a 100% likelihood of voting is twice as valuable as one to a voter with a 50% likelihood of voting. However, if the former is only half as likely to crossover as the latter, then the two voters are in fact of equal value to a campaign.
As debate continues on what the impacts of top-two are on candidates and governance, questions of how to run campaigns under top-two will only increase in importance.