In North Carolina, voters must decide whether to replace democratic senator Kay Hagan with her republican opponent, Thom Tillis. Hagan, who was elected in 2008 with 53% of the vote, once again faces a difficult challenge. Per an October 11th SurveyUSA poll, Hagan leads by just three points, and 538 reports that only one poll has ever had her ahead of Tillis by more than six points. Some consider this one of the most competitive races in 2014, and it’s in one of the nation’s most purple states. In such a close race, every vote counts.
We can better understand both the current Senate race and the new election legislation if we look at early and by-mail absentee voting in North Carolina. Not only does the state have an active “one-stop” (early) and by-mail voting electorate, the new legislation affects (among many other changes) early voting (by cutting a week off the early voting period and codifying early voting hours) and the ability for traditionally democratic voters (young and minority voters) to turnout (by, to name just two examples, eliminating same day voter registration and paid voter registration drives).
These facts about the NC election landscape mean that if we compare early and absentee voting numbers from the past to the 2014 numbers as they become available, we can anticipate in real-time whether Hagan should be worried about her chances of success, and whether the consequences of the new legislation are as restrictive as some suspect. Early voting in NC begins on October 23rd, and while absentee by-mail began on September 5th, there are still too few returned ballots to draw conclusions from the returns.
Over the next few days, I’ll point out a few trends that we should keep our eyes on. Unless otherwise noted, my data comes from the North Carolina State Board of Elections’ website.
(1) Total early and by-mail turnout
Here’s my proposal: If the total number of pre-Election Day voters shifts off from expected trends, then we have reason to suspect that the new legislative changes affected turnout. Simple enough, right?
But how will we know whether the 2014 results diverge from “expected trends”? One possibility is that turnout this year follows the pattern set by past midterm elections. This would make a lot of sense, since research suggests that voting trends are more predictable across similar election types. Midterm turnout has always been lower than turnout for presidential elections, and nationwide CPS data on early voting that begins in the mid-1990s (figure 1) suggests a similar trend: even though early and absentee voting has steadily increased across the country, the increase has a jigsaw look, consistent every four years, not every two.
Figure 1: National CPS data on early and absentee voters as percentage of all voters (data only available up until 2006, figure from Gronke, McDonald, 2008).
So, if we assume the pre-Election Day turnout numbers in NC correspond with past midterm numbers, what should we expect? Figure 2 uses data from elections in 2002, 2006, and 2010 to model the predicted early and by-mail turnout this year (click on the figure for a clearer view).
Figure 2: Predicted pre-Election Day 2014 turnout in NC, using only midterm results (data from NCSBE).
The model predicts that around 1.26 million voters will turnout before Election Day in 2014. But even this conservative estimate (I use a linear model, but a polynomial model would predict almost two million early and by-mail voters) may be too high. That’s because pre-Election Day turnout in 2012 suggests that early voting may have hit some sort of ceiling. Figure 3 displays early and by-mail turnout in NC during the presidential elections since 2000.
Figure 3. Pre-Election Day turnout during presidential election in NC since 2000 (data from NCSBE).
Note that between 2000 and 2008, turnout during presidential elections followed a similar pattern as turnout during midterms–an upward J-curve. This makes sense, since early and by-mail voting has gained incredible momentum across the country (but especially in the South) since 2008, when the democrats took advantage of convenience voting methods to help elect Obama. Even though turnout was lower in 2010 than 2008 (as we’d expect), early and by-mail voting numbers increased in a similar way across the midterm elections as they did across the presidential elections.
But in 2012, pre-Election Day voting (at least in NC) tapered off. What explains this? It’s unlikely that people forgot about early and by-mail voting. It’s also unlikely (though not out of the question) that election legislation suppressed the pre-Election Day vote, since (as far as I know) there were no major changes to NC election law between 2010 and 2012.
One possibility, then, is that NC has basically maxed out on it’s early and by-mail voting pool. The idea is that out of all the voters who will vote during a presidential election, nearly every one of them who would use early or by-mail voting has decided to do so. Given the amount of publicity surrounding pre-Election Day voting in the past four years, this possibility shouldn’t be ruled out.
Now, if we assume that presidential and midterm election voting populations are basically similar, it’s also possible that the midterm pre-Election Day voting electorate has maxed out. If that’s the case, the above model’s prediction would be much too high.
Figure 4, therefore, presents the midterm elections prediction (from figure 2) along with a prediction that early and by-mail voting this year will peter out in the same way it fell off in 2012. To determine this second number, I assume that the rate of change from 2010-2014 will be same as the rate of change from 2008-2012. Just a tad over one million voters would return ballots before Election Day if this second model is correct.
Figure 4. Predicted pre-Election Day turnout in NC, using midterm results and ceiling hypothesis (data from NCSBE).
Given the above work, what conclusions can we draw? First off, it provides two different conservative estimates for early and by-mail turnout in NC. If turnout is below what the midterm hypothesis predicts, then we have reason to believe that either the ceiling hypothesis is correct, or that the new election legislation has affected turnout. If turnout is below what the ceiling hypothesis predicts, then that’s even stronger evidence for the new legislation’s impact on pre-Election Day voting.
Does this tell us anything about Hagan’s chances? We’d have to know the relationship between early and by-mail voting in NC and votes for, well, democrats. We’ll address that more directly in tomorrow’s post.
In North Carolina, voters must decide whether to replace democratic senator Kay Hagan with her republican opponent, Thom Tillis. Hagan, who was elected in 2008 with 53% of the vote, once again faces a difficult challenge. Per an October 11th SurveyUSA poll, Hagan leads by just three points, and 538 reports that only one poll has ever had her ahead of Tillis by more than six points. Some consider this one of the most competitive races in 2014, and it’s in one of the nation’s most purple states. In such a close race, every vote counts.
On top of that, the state’s legislature has made significant (and restrictive) changes to North Carolina’s election law, and we’ll want to know how these changes (which it now appears will remain in effect for the current election) affect turnout as we anticipate a possible Supreme Court hearing and decision.
We can better understand both the current Senate race and the new election legislation if we look at early and by-mail absentee voting in North Carolina. Not only does the state have an active “one-stop” (early) and by-mail voting electorate, the new legislation affects (among many other changes) early voting (by cutting a week off the early voting period and codifying early voting hours) and the ability for traditionally democratic voters (young and minority voters) to turnout (by, to name just two examples, eliminating same day voter registration and paid voter registration drives).
These facts about the NC election landscape mean that if we compare early and absentee voting numbers from the past to the 2014 numbers as they become available, we can anticipate in real-time whether Hagan should be worried about her chances of success, and whether the consequences of the new legislation are as restrictive as some suspect. Early voting in NC begins on October 23rd, and while absentee by-mail began on September 5th, there are still too few returned ballots to draw conclusions from the returns.
Over the next few days, I’ll point out a few trends that we should keep our eyes on. Unless otherwise noted, my data comes from the North Carolina State Board of Elections’ website.
(1) Total early and by-mail turnout
Here’s my proposal: If the total number of pre-Election Day voters shifts off from expected trends, then we have reason to suspect that the new legislative changes affected turnout. Simple enough, right?
But how will we know whether the 2014 results diverge from “expected trends”? One possibility is that turnout this year follows the pattern set by past midterm elections. This would make a lot of sense, since research suggests that voting trends are more predictable across similar election types. Midterm turnout has always been lower than turnout for presidential elections, and nationwide CPS data on early voting that begins in the mid-1990s (figure 1) suggests a similar trend: even though early and absentee voting has steadily increased across the country, the increase has a jigsaw look, consistent every four years, not every two.
Figure 1: National CPS data on early and absentee voters as percentage of all voters (data only available up until 2006, figure from Gronke, McDonald, 2008).
So, if we assume the pre-Election Day turnout numbers in NC correspond with past midterm numbers, what should we expect? Figure 2 uses data from elections in 2002, 2006, and 2010 to model the predicted early and by-mail turnout this year (click on the figure for a clearer view).
Figure 2: Predicted pre-Election Day 2014 turnout in NC, using only midterm results (data from NCSBE).
The model predicts that around 1.26 million voters will turnout before Election Day in 2014. But even this conservative estimate (I use a linear model, but a polynomial model would predict almost two million early and by-mail voters) may be too high. That’s because pre-Election Day turnout in 2012 suggests that early voting may have hit some sort of ceiling. Figure 3 displays early and by-mail turnout in NC during the presidential elections since 2000.
Figure 3. Pre-Election Day turnout during presidential election in NC since 2000 (data from NCSBE).
Note that between 2000 and 2008, turnout during presidential elections followed a similar pattern as turnout during midterms–an upward J-curve. This makes sense, since early and by-mail voting has gained incredible momentum across the country (but especially in the South) since 2008, when the democrats took advantage of convenience voting methods to help elect Obama. Even though turnout was lower in 2010 than 2008 (as we’d expect), early and by-mail voting numbers increased in a similar way across the midterm elections as they did across the presidential elections.
But in 2012, pre-Election Day voting (at least in NC) tapered off. What explains this? It’s unlikely that people forgot about early and by-mail voting. It’s also unlikely (though not out of the question) that election legislation suppressed the pre-Election Day vote, since (as far as I know) there were no major changes to NC election law between 2010 and 2012.
One possibility, then, is that NC has basically maxed out on it’s early and by-mail voting pool. The idea is that out of all the voters who will vote during a presidential election, nearly every one of them who would use early or by-mail voting has decided to do so. Given the amount of publicity surrounding pre-Election Day voting in the past four years, this possibility shouldn’t be ruled out.
Now, if we assume that presidential and midterm election voting populations are basically similar, it’s also possible that the midterm pre-Election Day voting electorate has maxed out. If that’s the case, the above model’s prediction would be much too high.
Figure 4, therefore, presents the midterm elections prediction (from figure 2) along with a prediction that early and by-mail voting this year will peter out in the same way it fell off in 2012. To determine this second number, I assume that the rate of change from 2010-2014 will be same as the rate of change from 2008-2012. Just a tad over one million voters would return ballots before Election Day if this second model is correct.
Figure 4. Predicted pre-Election Day turnout in NC, using midterm results and ceiling hypothesis (data from NCSBE).
Given the above work, what conclusions can we draw? First off, it provides two different conservative estimates for early and by-mail turnout in NC. If turnout is below what the midterm hypothesis predicts, then we have reason to believe that either the ceiling hypothesis is correct, or that the new election legislation has affected turnout. If turnout is below what the ceiling hypothesis predicts, then that’s even stronger evidence for the new legislation’s impact on pre-Election Day voting.
Does this tell us anything about Hagan’s chances? We’d have to know the relationship between early and by-mail voting in NC and votes for, well, democrats. We’ll address that more directly in tomorrow’s post.
Thanks, Jimmy Brewer, for help on the post.