# Precinct analysis: Old reliables, newcomers, and everyone else

I have three more views of the 2015 electorate, now that I have a copy of the voter roster. With that, and with the past rosters that I have, I can try to paint a more detailed picture of who voted in this election, and perhaps make some comparisons to past elections. Today we’re going to look at voting history. How many voters this year were new, how many had voted in one or more recent elections, and how do those numbers compare to previous years?

```
Year    All 3   1and2   1and3   2and3   Just1   Just2   Just3    None
=====================================================================
2015   59,639  13,150  26,170   8,714  33,993   6,566  17,964 101,603
2013   46,582  22,044   4,721  13,148  12,239  20,690   6,046  48,662
2011   44,744   9,706  15,360   4,302  15,559   2,830   5,394  19,927
2009   55,117   5,818  22,122  25,227  10,907   7,684  20,218  38,755
```

Let me translate what those column headers mean. “All 3” is the number of people in that election who had voted in each of the three prior city elections. For the year 2015, that means the number of people who had voted in 2013, 2011, and 2009. For 2013, that means the number of people who had voted in 2011, 2009, and 2007. I trust you get the idea for 2011 and 2009; I have rosters going back to 2003, so that’s as far back as I can do this exercise. These are your old reliable voters – year in and year out, they show up and vote.

The next six columns specify one or more of these prior elections. A 1 refers to the election immediately before, a 2 refers to the election before that – i.e., two elections before – and a 3 is for three elections before. Again, for 2015, those elections are 2013 (“1”), 2011 (“2”), and 2009 (“3”). Thus, the column “1and2” means all the people who voted in 2013 and 2011, but not 2009. “1and3” means means all the people who voted in 2013 and 2009, but not 2011. “2and3” means all the people who voted in 2011 and 2009, but not 2013. Along similar lines, “Just1” means all the people who voted in 2013 but not 2011 or 2009, and so forth. Substitute other years as appropriate, and you’ve got it. Lastly, “None” means the people who had voted in none of the past three elections. These are your new voters.

I presume I don’t have to tell you that 2015 was indeed an outlier in this regard. We knew going in that years with high profile referenda have higher turnout than other years, and that’s what happened here. In addition, you have to remember that “high turnout” is a relative thing. Turnout for the Harris County portion of the city of Houston was 268,872, which is more than any odd-year election since 2003, but pales in comparison to the turnouts of recent even years in which city props have been on the ballot. In 2010, for example, 389,428 voters came out in the Harris County part of Houston – 40.9% turnout – with 343,481 casting a vote on the red light camera referendum. In 2012, for the four bond items and two charter amendments up for a vote, there were 565,741 voters, with as many as 435,836 ballots cast. Point being, there are a lot of even-year city voters. Some number of them decided to vote this year as well. I’m not in a position to quantify it further than that, but at a guess based on the other years, I’d say 30 to 50 thousand of those 101,603 were true newbies, while the rest had some prior voting history in Harris County. As we’ve discussed before, new people move in all the time, and some other people become newly eligible due to turning 18 or becoming citizens. If and when I get more details on that, I’ll be sure to share them.

Here’s another way of looking at the data: The proportion of each class of voter for these elections.

```
Year   All 3   2 of 3   1 of 3   0 of 3
=======================================
2015   22.3%    17.9%    21.9%    37.9%
2013   26.8%    22.9%    22.4%    27.9%
2011   38.0%    24.9%    20.2%    16.9%
2009   29.7%    28.6%    20.9%    20.9%
```

“2 of 3” and “1 of 3″ refers to voters who had voted in two of the previous three elections, and one of the previous three elections, respectively. Again, the share of new voters this year was clearly higher than in other years. It’s no surprise that the share of new voters was so low in 2011. It was a low turnout year – just over 117,000 voters in total – so you’d expect that a large majority of them would be the regulars. By the same token, the old reliable share this year was lower than usual, for the same reason. I’m fascinated by how stable the 1 of 3” share was across the four races. As we saw in the table at the top, the one prior election in question can be any of the three predecessors. It’s not just folks who’d been new the year before. That number is directly affected by the turnout levels of the election in question and the one before it.

So that’s our first look at this data. I don’t have any broad conclusions to draw here, I just find this stuff amazing. Who would have guessed that over 2,800 people who voted in the low-turnout 2011 election had also voted in the low-turnout 2007 election, but not the higher-turnout 2009 or 2005 elections? Well, now you know. I’ll have more tomorrow.