Monday, March 7, 2016

Occam's Razor versus spinning

I think both these articles are misleading but that their underlying point is still interesting. From Fewer Democrats Are Voting This Year In (Surprise!) States With Strict New Voter Laws by Janie Velencia and Alissa Scheller and from Dem turnout and voter ID: The dirty little secret by Thomas Lifson, respectively left-leaning and right-leaning magazines.

Scheller and Velencia are bringing to attention to an interesting phenomena that has received some, but not a lot, of media attention; the differences between the two parties in their turn-out so far in the election cycle.
GOP voter turnout in this year's presidential race is up 62 percent relative to 2008, the last time both parties had open contests. But Democratic voter turnout is down by 29 percent across all the primary and caucus states that have voted so far. In all but two states, fewer Democrats turned out to vote in 2016 than did in 2008.
My go-to explanation is that this reflects primarily different levels of competitiveness in the two races.

On the Democratic side you have two white, northeastern candidates long in the tooth. One is a principled and self-identified socialist and the other is a politician under FBI investigation on multiple issues and with a career clouded by controversy, ethically dubious decisions, and legally questionable actions. In addition, from the start of the contest, all the establishment powers were trying to force this into a coronation rather than a contest with super delegates and sycophantic press coverage and DNC shenanigans to provide coverage for the preferred candidate. The deck was stacked and there wasn't much choice. To the surprise of everyone, the principled socialist conducted a more than forlorn hope campaign. None-the-less, this was a campaign of few choices and little competition.

On the Republican side you have had a field of, what, seventeen candidates? Men, women, older, younger, establishment, outsiders, candidates with executive experience and executive neophytes, Tea Party conservatives, social conservatives, libertarians, religious conservatives, fiscal conservatives, candidates from the North, South, Midwest, and West, black, white, Hispanic and Asian candidates, etc. Lots of choice and lots of competition.

Competition tends to increase turnout. My Occam's Razor is that the Republicans have had much more choice, a much wider range of choice and fought much harder over their choices and that is why their turnout is up whereas the Democrat turnout is down.

Scheller and Velencia want to make this empirical phenomena (increased Republican turnout versus declining Democrat turnout) about voter identification. I think they have a poorly structured and unsupported argument.

My standing assumption about voter fraud is that it is more prevalent than Democrats want to acknowledge and less prevalent than Republican activists make it out to be. There are enough well documented (though rarely publicized) incidents to affirm that it is a real issue. Incidents such as counties with 104% turnout of registered voters and the like.

It does happen. But how prevalent is it? That is the critical question to answer and that is unanswered. And how consequential is it? Also unanswered. We have a lot of elections that come down to a swing of a few hundred votes or of a percentage point or two. It is quite conceivable that it is more prevalent than we realize and that it is consequential. Possible but not certain. It warrants investigation but Democrats have been reflexively resistant to even the lightest research to answer the question. A position which makes it appear that they have something to hide.

An impression that is increased by the fact that their claimed suppression of voting following implementation of voter ID laws has not showed up in actual election results. In fact, in the last research report I saw a couple of years ago, states with voter ID laws typically saw higher turnout than before the ID law was passed. Remembering that the sample size is still very small, I don't view that as conclusive by any means but it does call into question the Democrat concern that voter ID suppresses voting.

The Republicans have been more exercised about voting fraud because they appear to believe that it is more prevalent and more consequential. The Democrats have been more obstructionist to voter ID laws and also seem to have been the greater beneficiaries of elections where there are greater accusations of voter fraud. Sometime in the aughts, there was a string of Senate contests where the Republican came in with a victory of a couple of hundred votes which election was then contested by Democrats with one, two and three recounts, each recount increasing the number of Democrat ballots. Might have been a statistical fluke, but it looked suspicious.

Back to Scheller and Velencia's argument:
Eight out of the 16 states that have held primaries or caucuses so far have implemented new voter ID or other restrictive voting laws since 2010. Democratic turnout has dropped 37 percent overall in those eight states, but just 13 percent in the states that didn't enact new voter restrictions. To put it another way, Democratic voter turnout was 285 percent worse in states with new voter ID laws.
That's interesting and worth investigating. Scheller and Velencia don't investigate, they simply try and make a correlational argument from a small sample size, and then leave it at that. My eyeballing of their data suggests that there might be a different and stronger correlation than between states with and without voter ID.

In six of the nine states with Democrat turnout declines of greater than 20%, the states had a Democrat majority twenty years ago and have switched to Republican majorities. In other words, the current decline is part of a longer term trend of Democrat decline in those six states (South Carolina, Georgia, Alabama, Arkansas, Tennessee and Texas.) Iowa, Virginia, and Nevada are the possible exceptions (I simply don't know whether the Democrats are declining there or not).

So structural decline is likely a more probable explanation than voter ID laws. The other explanation that seems relevant arises from the observation that the Republican candidate Trump is a lifelong Democrat who seems to be drawing large numbers of Democrats into the Republican column. He has so far done quite well in open primaries (you don't have to be a registered Republican to vote in the Republican primary) and less well in closed primaries (only registered Republicans can vote in the Republican primary). It appears from various polls that 15-30% of Trumps voters are actually Democrats.

If you are a Democrat who votes for Trump in an open primary, you numerically decrease the Democrat turnout and increase the Republican turnout.

Scheller and Velencia make the argument that voter ID laws have been influential in reducing Democrat turnout but they cherry pick their data and do not present the readily available counterarguments that appear to be more explanatory. Turnout is down because 1) the Democrat choices are not appealing to swaths of voters, 2) the elections so far have been primarily in states where the Democrats are in systemic decline anyway, and 3) a segment of Democrats are defecting in open primaries to Donald Trump.

That's my read of the situation.

Lifson's argument is even weaker. His synopsis is:
The numbers tell a story, and you can draw the obvious conclusions. Because the mainstream media certainly won’t. Keep this statistic in mind the next time some progressive tries to claim voter fraud is not a serious problem.

Political Wire quotes the HuffPo:
Huffington Post: “Eight out of the 16 states that have held primaries or caucuses so far have implemented new voter ID or other restrictive voting laws since 2010. Democratic turnout has dropped 37 percent overall in those eight states, but just 13 percent in the states that didn’t enact new voter restrictions. To put it another way, Democratic voter turnout was 285 percent worse in states with new voter ID laws.”
Left unsaid: despite the “burden” of obtaining voter ID, GOP turnout was up.
In other words, for Lifson, the decline is simply a product of reduced opportunity for fraud. I understand the emotional appeal of that argument but it is inadequate as it does not explain the Democrat voting decline in states without voter ID laws.

One final note of interest from these articles and it has to do with journalist advocates and their, presumably, subconscious framing of issues.

The issue is voter ID laws. Scheller and Velencia are comparing states with and without voter ID laws. But notice how they frame it as an issue between states with and without voting restrictions. Very interesting. No state has limited voting. They have limited the opportunities of fraudulent voting. Given how they frame the issue, it would be easy to interpret Scheller and Velencia as being upset that opportunities for fraudulent voting have been reduced. Not what I think they intended to communicate but it seems plausible.

No comments:

Post a Comment