Turnout in the runoff: The Problem with Reading the Tea Leaves
We’re closing in on the final days of early voting, and Election Day is barely more than a week away. It is the time when a lot of us will attempt to read the tea leaves and make prognostications about who will cast a ballot. The problem is that we may know a lot less about this stuff than we think we do.
How Does Turnout in Runoffs Typically Compare to Turnout in Primaries?
There is no systematic trend in turnout from primaries to runoffs in Louisiana. On one hand, you might think that with the focus of a two-candidate runoff the stakes are clearer and voters are more likely to show up. On the other hand, you might think that when asking voters to voluntarily give up time to cast a ballot twice there will be inevitable roll off. In fact, neither guess turns out to be a general rule.
Louisiana has had only seven statewide runoff elections at the top of the ticket (either governor or U.S. Senate) since adopting its election system in the 1970s. Consider the percentage change in the total number of votes cast from the primary to the runoff in these elections [Note: This is not the percentage point change in the turnout rate, it is the percent change in total votes cast]:
- 1996 U.S. Senate: +38%
- 1986 U.S. Senate: +15%
- 1991 Governor: +12%
- 1995 Governor: +5%
- 2003 Governor: +3%
- 2002 U.S. Senate: -1%
- 2014 U.S. Senate: -13%
More often than not turnout rose from the primary to the runoff, but the variation across election years is large. In Mary Landrieu’s first election for U.S. Senate in 1996, for example, 38% more votes were cast in the runoff than in the primary. No surprise there; that runoff was held on the day of the presidential election. Then, last year, in Landrieu’s final senate election, 13% fewer votes were cast in the runoff than in the primary.
The wide spread of these primary-to-runoff trends suggests that the particular circumstances of each race plays an important role, but with such a small sample size it is impossible to state any claims with certainty.
The data are even sparser when looking at turnout among specific groups. The Louisiana Secretary of State’s website has turnout data by race going back to 1998, which covers just three runoff elections: 2002, 2003, and 2014. Here is the percent change in total votes cast from primary to runoff for blacks:
- 2002 U.S. Senate: +2%
- 2003 Governor: +2%
- 2014 U.S. Senate: -11%
And here is the percent change in total votes cast from primary to runoff for whites:
- 2002 U.S. Senate: -4%
- 2003 Governor: +1%
- 2014 U.S. Senate: -17%
In 2002, black voters cast 7,500 more ballots in the runoff than in the primary, a 2% percent increase. This represented a slight rise in their turnout rate from 40.0% of black registered voters to 40.8% of black registered voters. In contrast, white voters cast 37,000 fewer ballots, a 4% decline. Their turnout rate fell from 48.4% to 46.4%. But, last year, the number of total votes fell among both blacks and whites. The overall turnout rate among registered voters fell from 51.5% to 43.6%. Among black voters it fell from 47.3% to 42.2% and among white voters from 54.8% to 45.5%.
Based on these few elections there does not appear to be any pattern where black turnout systematically suffers in a runoff relative to white turnout. But, again, we’re talking about three elections. There is no way to discern systematic patterns from such a small sample.
What Will We Learn From Early Voting?
Not much. There is plenty that I or anyone else could claim about what will happen come Election Day based on what we see in early votes, but the truth is we have no idea about how well early voting in Louisiana predicts overall turnout. Early voting in Louisiana as we know it today has not been around all that long - neither in terms of policies governing its availability nor in terms of voters' willingness to use it. In the late 1990s and early 2000s only about 3% to 6% of total votes either in the presidential elections or in gubernatorial primaries were not cast on Election Day. In presidential elections in Louisiana, the share of early votes jumped to 15% in 2008 and 18% in 2012. In gubernatorial primaries, the share doubled to 11% in 2007, to 16% in 2011, and to 21% in 2015 (more than five times the share of early/absentee votes cast twenty years earlier).
The changes in early voting leave us with just a handful of recent elections for comparison. We do not have sufficient data to identify if/how early voting predicts Election Day turnout.
Like I said, I can venture a guess. During the primary, early vote totals were compared to last year’s early vote totals and found to be wanting. From that, folks guessed that overall turnout in 2015 would be lower than it was in 2014. The guess turned out to be right. That happens sometimes. But it was still a guess based on how we think things might be associated, not a prediction based on what we know to be associated. As I argued then we just do not have enough elections yet under the current early voting system to establish what the typical pattern is.
It probably makes more sense to compare the trend in early voting within a single election (from primary to runoff) than to compare the trend in early voting from one election year to another. But, with this more constrained comparison we have even less data. There has been just one runoff in Louisiana since early voting became the phenomenon it is today. In 2014, the number of early votes dropped by 5% from the primary to the runoff. Total turnout dropped too, by 15%.
Based on last year, it seems a reasonable guess to me that if we see a change in the amount of early voting from the primary to the runoff this year, then we’ll see a change in the same direction for Election Day turnout too. But, this is a guess based on one data point. It is not an empirical regularity that has been vetted in any rigorous way. And we certainly have no way of predicting with any accuracy what the magnitude of the shift in overall turnout would be based on the magnitude of the early voting shift.
It is also risky to predict the makeup of the Election Day electorate based on the makeup of the early voting electorate. Consider the black share of the 2014 electorate: Blacks cast 32.5% of early votes but just 28.8% of all votes in the 2014 primary. Their share of all votes was 3.7 percentage points lower than their share of early votes - indicating early votes over-represented the actual black share of the electorate. But, in the runoff things went the other way. The black share of total turnout was 2.4 percentage points higher than the black share of the early vote - indicating early votes under-represented the black share of the electorate.
Looking across all statewide elections since 2008 no systematic pattern emerges for how the share of the black vote in early voting relates to the share of the black vote overall. In some elections blacks make up a larger share of early voters than they make up of all voters (2008, 2012, and 2014 primary), in other elections the black share of early voters is lower than the black share of all voters (2010, 2011, and 2014 runoff).
There is more consistency for Republicans. They are over-represented in early voting in every statewide election since 2008. However the size of the gap fluctuates widely from half a percentage point to nine percentage points.
As Long As We Are All Just Guessing Here…
Like others, I expect higher turnout in the runoff than in the primary. This is just a hunch based on the fact that black turnout was relatively low in the primary when the Democrat was all but assured a spot in the runoff without having to launch a heavy GOTV effort. Now the Democrats have a stronger incentive to mobilize voters. If they do (and I assume they will), then we'll see turnout rise where it was lowest in the primary.
But this is just a hunch. It's based on what seems logical, not something I (or anyone else) can draw out of the data. So let's offer up our prognostications, but we should not pretend the tea leaves are anything more than, well, tea leaves.