Friday, January 27, 2017

Lies, opinions, probabilities and the 2016 election

Well, I wasn't going to write about this. I am trying to get back to gleanings from my reading.

But being heavily involved in decision-making and epistemology, I was heavily tempted. Then, listening to Terry Gross, she made a statement that pulled me in.

The original report was Trump Repeats Lie About Popular Vote in Meeting With Lawmakers in the New York Times.
President Trump used his first official meeting with congressional leaders on Monday to falsely claim that millions of unauthorized immigrants had robbed him of a popular vote majority.
To lie, from Webster:
Definition of lie
lied; lyingplay \ˈlī-iŋ\
intransitive verb
: to make an untrue statement with intent to deceive
: to create a false or misleading impression
Now to know whether Trump was lying (making an untrue statement) we would have to know 1) How many people voted and 2) of those who voted, how many voted fraudulently. In order to get the latter number, we have to agree on what constitutes fraudulent voting.

Any one of these numbers is extraordinarily hard to get but I'll go into the details in a moment. The key issue is that at this moment in time, we do not have either a fixed number as to how many voted nor do we know how many voted fraudulently. Without that information, we don't know that Trump is lying because we don't know, can't know whether what he is claiming is untrue.

For reasons that I think will become clear shortly, I think it highly unlikely that millions of fraudulent votes were cast. But it is possible. We are talking probabilities not facts. We can agree that Trump made an improbable statement but the NYT is sloppily claiming that he lied when they do not, and cannot, know what the truth is.

Now was this simply a headline writing error? Was it their bias? Was it their poor grasp of mathematic principles? I have no idea, but to claim someone else is lying when no one yet knows the truth is simply irresponsible journalism.

The only reason I am posting on this is that this sloppy thinking seems to be spreading. I heard Terry Gross interviewing a Guardian journalist, Luke Harding, and she repeated the same claim as the NYT.
He's saying that there were millions of votes cast for Hillary Clinton that were cast by illegal immigrants who shouldn't have voted. The facts say that those allegations are not true. The facts don't support it.
Does Terry Gross know how many people voted fraudulently? No, of course not. No one does. She presents her conviction as truth when there is no fact, only opinion.

So let's start at the beginning. As soon as I went to validate the number of voters in 2016, I immediately encountered the malleability of "facts." Based on the best number I could find, it appears that 137 million people voted in the presidential election on November 8, 2016, representing a 59.0% turnout. But that's a pretty obscure site, though credible. That number was not in any of the first dozen mainstream media accounts I accessed.

If you do a search on "how many people voted in 2016 presidential election" you get all sorts of noise. The leading headline is Over 90 Million Eligible Voters Didn't Vote in 2016. Then you have Voter turnout at 20-year low in 2016 from CNN. Makes you think not many people voted, doesn't it.

In fact, according to Fivethirtyeight, No, Voter Turnout Wasn’t Way Down From 2012. This was written on November 15th and the numbers are even higher now.
Approximately 58.1 percent of eligible voters cast ballots in last week’s presidential election, according to the latest estimates from Michael McDonald, associate professor at the University of Florida, who gathers data at the U.S. Elections Project. That’s down only slightly from 2012, when turnout was 58.6 percent, and well above 2000’s rate of 54.2 percent. Turnout may end up being higher than in any presidential election year between 1972 and 2000. (It’s already higher than in any midterm election since 1896, according to McDonald’s numbers, including the paltry 35.9 percent of voters who turned out two years ago.)
The Washington Post gets in on the game with the seemingly contradictory headline More votes were cast in 2016 than in 2012 — but that doesn’t mean turnout was great. As opposed to CNN and Washington Post, the real numbers indicate that this election saw the highest number of voters ever and the third highest voter participation rate in 44 years, since 1972.

The MSM headlines are clearly trying to tell a different story than what the numbers indicate.

OK, we know how many people voted. Now we get into the interesting analysis. How many fraudulent votes were there? We don't know. No one checks or tracks that number.

We do know some other things which give credence to the possibility that there was fraud, though almost certainly not of the magnitude that Trump is claiming.

What are the sources of uncertainty?
1) The US, by-and-large, and unlike European countries, conducts its elections broadly on the honor system. Yes, some states have some sorts of voter identification requirements but even those pale to the rigor of European poll identification.

2) It is well established that most electoral rolls (names of people deemed eligible to vote) are significantly flawed. Many places fail to routinely purge, sometimes for decades, voters who have died, voters who have left the district, voters who have been convicted of felonies, etc.

3) It is relatively easy for non-citizens to obtain sufficient identification to register to vote even though, as non-citizens, they are ineligible to do so.

4) With the rise of early voting, it is easier than ever to vote on behalf of someone else.
Is there any evidence that this happens? Sure, plenty. It is a party line of the Democrats and their allies in the mainstream media that there is no evidence of voter fraud. (Well, that was the party line until they briefly decided that Russian voter hacking might have caused the 2016 outcome, though they now seem to have backed away from that claim.) They also actively seek to prevent investigations into voter fraud which is probably the clearest indication that there is fraud to be found if you look for it.

The potential magnitude of the problem is clear from Inaccurate, Costly, and Inefficient: Evidence That America’s Voter Registration System Needs an Upgrade from the Pew Center on the States. Findings:
Approximately 24 million—one of every eight—voter registrations in the United States are no longer valid or are significantly inaccurate.

More than 1.8 million deceased individuals are listed as voters.

Approximately 2.75 million people have registrations in more than one state.
So 29 million names on the electoral rolls are not accurate for voting purposes; remember, there were 137 million voters. 29/137 = 21% of electoral roll voter names are faulty or prohibited from voting.

In this most recent 2016 election, we have this from The Detroit News, Records: Too many votes in 37% of Detroit’s precincts. In 37% of their voting precincts, there were more votes cast than there were voters on the rolls. If Detroit were experiencing rapid population growth, this might be possible. Given that Detroit has been rapidly shrinking in population, this outcome is almost impossible without fraud.

It is not uncommon for there to be claims of excess voters over those on the voter rolls: see Philadelphia, Florida, Illinois as a fairly representative sample. Sometimes there are innocent explanations, particularly if the voting area includes population centers with high churn such as universities or military bases. But often there are not good explanations.

Other instances:

Early voting fraud: Florida, Iowa, Virginia

Votes from Beyond the Grave

Nevada GOP Chair Cries Election Fraud in Early Voting

Do non-citizens vote in U.S. elections? by Jesse T. Richman, Gulshan A. Chattha, and David C. Earnest concluded that "up to 2.8 million votes could have been dropped by illegal aliens into ballot boxes during elections held in 2008 and 2010."

From the Washington Post in 2014, Could non-citizens decide the November election? by Jesse Richman and David Earnest.
Could control of the Senate in 2014 be decided by illegal votes cast by non-citizens? Some argue that incidents of voting by non-citizens are so rare as to be inconsequential, with efforts to block fraud a screen for an agenda to prevent poor and minority voters from exercising the franchise, while others define such incidents as a threat to democracy itself. Both sides depend more heavily on anecdotes than data.

In a forthcoming article in the journal Electoral Studies, we bring real data from big social science survey datasets to bear on the question of whether, to what extent, and for whom non-citizens vote in U.S. elections. Most non-citizens do not register, let alone vote. But enough do that their participation can change the outcome of close races.

Our data comes from the Cooperative Congressional Election Study (CCES). Its large number of observations (32,800 in 2008 and 55,400 in 2010) provide sufficient samples of the non-immigrant sub-population, with 339 non-citizen respondents in 2008 and 489 in 2010. For the 2008 CCES, we also attempted to match respondents to voter files so that we could verify whether they actually voted.

How many non-citizens participate in U.S. elections? More than 14 percent of non-citizens in both the 2008 and 2010 samples indicated that they were registered to vote. Furthermore, some of these non-citizens voted. Our best guess, based upon extrapolations from the portion of the sample with a verified vote, is that 6.4 percent of non-citizens voted in 2008 and 2.2 percent of non-citizens voted in 2010.
From FoxNews in 2016, Why Trump's probe of voter fraud is long overdue by John Fund and Hans A. von Spakovsky has a very long list of recent voter fraud cases.

At the federal level, there have been consistent efforts over the past eight years to both fight voter identification laws and to prevent investigations into voter fraud. The most egregious example perhaps being the targeting of the Texas based True the Vote organization by the IRS.

I posted about one of the most compelling pieces of evidence that there is systematic fraud a couple of years ago in Political Statistics referencing an interesting analysis conducted by Dan McLaughlin, Do Democrats Always Win Close Statewide Elections?. It has long been an article of faith among Republicans that "You can't just win. You have to beat them by the margin of fraud." McLaughlin was investigating whether there was any statistical support for that view. There is.
To get a sense of the answer, I took a look at all the statewide Senate and governor’s races from 1998 through 2013 (thanks to Sean Trende of RealClearPolitics for a big assist with the data) as well as all the statewide results in the presidential elections during that period. Let’s begin with the very closest races, those decided by less than one percentage point. There have been 27 such races since 1998, and Democrats have won 20 out of 27
You would expect for all contested races, that either party would have a 50:50 chance of winning. However, what McLuaghlin turns up is that for the very tightest races (<1% difference in votes), Democrats win 74% of the recounts. This statistical advantage extends all the way up to not so close races. For all contested races up to a 6% difference in votes, Democrats win 60% of the recounts. Possibly that is just statistical noise as the sample size is only 147, i.e. there were only 147 contested elections with recounts. But the effect size is pretty compelling. There is plenty of evidence documenting that our voting infrastructure is flawed and susceptible to voter fraud. We have numerous examples where there is strong evidence that voter fraud has occurred. We have no aggregate picture of how common this is or how persistent. My personal estimate is that there are occasional, episodic, material, and localized instances of voter fraud. I am guessing that at an aggregate national level, the number of fraudulent votes is high in absolute terms but low in percentages, i.e. between 0.1 and 0.5%. I suspect that no national level (President) elections have been affected by these instances of fraud but that quite likely some Federal Senate and House elections have probably been swung because of voter fraud. I am guessing that voting fraud is much more consequential for State and Local races than it is for federal level races. My final forecast is that, if we do a detailed election integrity review, we will find that election fraud is most frequent within big cities that operate under a one party system. Since most big cities have been essentially one party voting zones for fifty years and more and since they are Democrat dominated, this will end up being primarily a Democratic Party issue (as echoed by the DNC's suppression of the Bernie Sanders campaign in the 2016 election) which is why they are so adamant against voter identification laws (despite their prevalence in Europe) and why they are so adamant that there is no fraud and that there should be no investigation. I do not believe this to be an indictment of the Democratic Party per se, i.e. they are not inherently fraudulent. This is simply a product of historical trends. Big cities have always been prey to corruption. One party systems are also inherently prone to corruption. If your party ends up being concentrated in urban areas (as the Democrats have become) and if you have no party competition for prolonged periods of time (as has been the case for most major cities), then you are inherently exposed to the possibility of voting fraud. The final twist in all this is whether Democrats have played into Trump's hands or whether this was simply happenstance. Republicans, as alluded to above, believe, and the statistics support them, that they lose in recounts with Democrats in close elections and they believe this is due to electoral fraud, primarily in big cities. With Democrats baseless claims of Russian Hacking and Trump's exaggerated claims of vote fraud, Trump's call for a bi-partisan investigation into voter fraud would seem to have backed the Democrats into a corner. Theoretically, Democrats would want to know that Russia did not hack the election and as they have declared that there is no voter fraud, they should have no concerns about an investigation. Was this a master strategic move on Trump's part to force Democrats to endorse a course of action that likely will be embarrassing and detrimental to them? I don't know and it seems improbable but that is certainly the effect. Finally, back to Terry Gross and the New York Times. When they claim there are no facts to support Trump's opinion that there might have been massive voter fraud, they are simply wrong. Trump has claimed that he might have won the popular vote had there not been voter fraud. In order to win the popular vote, which went to Clinton by some 2.9 million, there would have to have been (2.9/137 = ) 2.1% voter fraud. The above studies indicate that that is well within the realm of possibility (21% error rate on electoral rolls; 2.8 million fraudulent votes across the 2008 and 2010 elections, etc.) Gross and the NYT make a category error when they fail to distinguish between Trump's opinion that there might have been sufficient voter fraud for him to have won the popular vote and their opinion that there was little or no voter fraud. Opinions are not true or false, therefore they cannot be a lie. Failing to distinguish statements of opinions from statements of fact is the category error.

Gross and the NYT also make an error of fact by asserting (in Gross's case) or implying (in the case of the NYT) that there is no evidence to support Trump's opinion.

Trump's claim is improbable but it is conceivable. There is certainly evidence that there is voter fraud. There is even evidence that it might possibly be material. By declaring an opinion to be a lie and by denying that there is any evidence to support the opinion, Gross and the NYT markedly harm their credibility.

UPDATE: New Data: Illegal Voters May Have Decided New Hampshire in 2016 by J. Christian Adams provides evidence that though the absolute number of fraudulent votes cast might have been relatively small (roughly six thousand), they were strategically targeted enough in a small state to swing an election. In this instance, the number of fraudulent votes accounted for more than the margin of victory for the Senate democratic candidate over the Republican incumbent.

No comments:

Post a Comment