Friday, January 11, 2013

On Predicting Voter Turnout Based On Early Voting


In July, 2012, as the runoff in the Republican and Democratic primary election approached, Professor Mark Jones argued that voter turnout in the Republican primary runoff election could be predicted based on the early voting turnout in the most populous fifteen counties in Texas (his blog post is here). He provided early voting figures from the first Republican primary, which allowed twelve days for early voting, compressed the data into five days to be comparable to the five days of early voting allowed in the runoff primary election. He stated that if a comparison of early voting (early voting in the runoff/early voting in the first primary) was a low percentage, then turnout would likely be closer to the low estimate for voter turnout in the runoff primary. On the other hand, if the proportion were large, then turnout would likely be closer to the high estimate for voter turnout in the runoff primary. His figure is reproduced below:


Professor Jones admits that accurately predicting turnout and the percentage of the vote that would be cast during early voting is difficult, but he offers the following:

It is difficult to predict both what actual turnout will be in the marquee race of the GOP primary, the Senate runoff between David Dewhurst and Ted Cruz (in the first round in May it was 1.4 million), and what proportion of the voters will vote early or by mail (in the 15 most populous counties, 51 percent of voters cast a ballot in the Republican Senate primary early or by mail during the early voting period of May 14 to May 25). A reasonable estimate of overall turnout in the GOP Senate primary would fall between 750,000 and one million, while it is very likely that the proportion of these Texans casting a vote early or by mail will be somewhere between 55 percent and 60 percent.

How well did the prediction work? Here are the data:

County
Rep Primary
Rep Runoff
% RR/RP
Harris
78,441
70,481
89.85%
Dallas
31,312
33,895
108.25%
Tarrant
40,463
34,837
86.10%
Bexar
36,413
32,764
89.98%
Travis
17,535
16,056
91.57%
Collin
26,581
22,650
85.21%
El Paso
5,857
2,301
39.29%
Denton
18,774
14,338
76.37%
Fort Bend
20,884
14,451
69.20%
Hidalgo
4,293
2,022
47.10%
Montgomery
22,792
17,268
75.76%
Williamson
16,303
9,821
60.24%
Nueces
6,473
3,731
57.64%
Galveston
14,437
11,544
79.96%
Cameron
2,939
1,825
62.10%
Total
343,497
287,984
83.84%




EV + ED Vote
1,406,648
1,111,938
79.05%
Predicted Vote

1,179,333

Difference

67,395
6.06%

As you can observe from the table, the turnout in the Republican runoff primary was 1.1 million voters, and the early vote in the fifteen largest counties constituted 25.90 percent of the total votes cast, which is greater than the 24.42 percent provided by the early vote in the fifteen largest counties in the first primary. The total vote cast in the runoff primary in the fifteen largest counties was nearly 84 percent of the vote cast in the first primary. Using the early vote in the largest fifteen counties to predict the total vote in the runoff primary resulted in a predicted vote of 1.179 million votes, which are only 67,395 votes more than the actual vote and a reasonably small error.

Using the early vote to predict the final vote total in this case was reasonably accurate, but can the early vote be used generally to predict the final vote total? In the general elections between 2004 and 2012, the correlation between early vote in the 15 largest Texas counties and the final vote total is .95, which is quite high (see figure below). Why is the relationship even greater (.99) in this election? There are several reasons. First, the importance of the contest is probably most important. Whoever wins the Republican primary for U.S. senator is almost assured of winning the general election. Therefore, the incentive to participate in the primary runoff is great. Second, both candidates campaigned vigorously, both in the first primary and in the runoff and spent a considerable amount of money to make the case for their election to the voters. Third, the potential participants in the runoff primary are very restricted. The only voters who are allowed to participate are those voters who voted in the first primary and those individuals that did not vote in any party’s primary or participate in another party’s nomination process. 

 

1 comment: