My prediction of the total vote in Texas was high by 1,180,324 votes with 98.5 percent of the precincts reporting. That is an error of 13.29 percent (totally unacceptable). So why was I so wrong?
First, my estimation of the percentage of the total Early Vote based on the Early Vote in the 15 SOS counties was wrong. I predicted that the Early Vote in those 15 counties would make up 69.29 percent of the total Early Vote in Texas' 254 counties. The actual percentage was 68.67, which was not too far off. However, I predicted that the total Early Vote would make up only 64.53 percent of the total vote cast in Texas, whereas the Early Vote made up 73.77 percent of the total vote, far in excess of the percentage in previous presidential elections in Texas. As a result, my estimate of the total vote in Texas was off by more than one million votes, as I noted above. My biggest error was underestimating by nearly ten percent the percentage of the total vote constituted by the Early Vote.
What can we learn from this exercise? Large early vote totals do not necessarily lead to large total votes. An increase in early votes may simply mean that people are anxious to cast their votes and early voting offers an opportunity to cast that vote before Election Day. Most research on the effect of Early Voting on voter turnout shows only a slight increase (a few percentage points). Thus, the total vote in 2016 exceeded the total vote in 2008 by only 800,359, and the percentage of registered voters was lower than in 2008. As a percentage of the voting-age population, the turnout was lower than in 2008 and 2012 (note that I used the VAP figures from Michael McDonald's US Elections Project Web site in the table below rather than the Secretary of State's Web site for this table. McDonald's figures seem more accurate.)