Lifting The Election Fog

 I didn't really want to write another post looking at the claims that the election was stolen, but the claims keep coming.  In particular, several people have pointed me towards this "Reasons why the 2020 presidential election is deeply puzzling" article.  Then someone even went to the length of looking up my non-blog email address (admittedly not actually hard to do) and emailing it to me.  So, I guess I'll take that as a challenge.  

The challenge in answering this is that it takes the approach, if I may use the crude phrase, of throwing a whole lot of shit at the wall to see what will stick.  Answering all of these fully would take some time, but I think I can answer almost all of them fully enough.  And while this form of argumentation is not proof, I will at least ask skeptical readers to consider that if the solid majority of the claims turn out to be erroneous, perhaps the author is not actually much in command of the facts in general.

Working from first to last, the author (one Patrick Basham) says:

First, consider some facts. President Trump received more votes than any previous incumbent seeking reelection. He got 11 million more votes than in 2016, the third largest rise in support ever for an incumbent. By way of comparison, President Obama was comfortably reelected in 2012 with 3.5 million fewer votes than he received in 2008.

This is a true fact, but it doesn't actually mean a whole lot.  There are two factors which determine the raw number of votes cast in American elections: the number of voters available and the percentage of eligable voters who actually register and vote.  (The vast majority of people who are registered do in fact vote, the issue is more that a lot of people don't actually register.  In 2016, 86.8% of registered voters voted, but that is only 55.7% of eligible voters.)   In 2020 an estimated total of 158 million votes were cast (CA and NY are, to my knowledge, still counting at tortoise pace) but that's still only about 66.5% of eligible voters.  Over the last 24 years turnout has been increasing from a low point in 1996.  Arguably, this is a product of our increasing political polarization.  The Democrats put huge efforts into registering and turning out voters in 2020, and doubtless they expected they would coast to victory against an unpopular President Trump as a result.  The reason why the election was in fact just as narrow as 2016 (though with a narrow Dem victory instead of a narrow GOP one) is that they failed to account for the fact that they are nearly as divisive as Trump himself.  Just knowing that the Democrats were working so hard to increase voter turnout got more Republican leaning voters to the polls as well.

Trump’s vote increased so much because, according to exit polls, he performed far better with many key demographic groups. Ninety-five percent of Republicans voted for him. He did extraordinarily well with rural male working-class whites.

Trump grew his support among black voters by 50 percent over 2016. Nationally, Joe Biden’s black support fell well below 90 percent, the level below which Democratic presidential candidates usually lose.

This conflates something that is true with something that isn't.  

It's true that Trump improved his performance with black voters, but even with that improvement Trump only got the support of 12% of black voters (19% of black men and 9% of black women).  It's also true that Trump won white working class voters by a large margin -- he beat Biden by 35% among white voters with no college degree.  But in a sign of trouble for Trump, that was a decline in his core constituency from 2016 when he beat Clinton by 37% among whites with no college degree.  Also a significant problem for Trump is that fact that while he won college educated white men by 14% in 2016 he only won that demographic by 3% in 2020, while Biden won among white college educated women by 9% which was actually an increase over Clinton's win among the demographic of 7%.  

So yes, Trump got lots of votes from working class whites, and he increased his support among blacks and Hispanics, but if we look at all the demographics we see a picture of Trump as a candidate who lost more support than he gained in terms of percentages of voters, even though partisanship drove record turnout numbers and thus a record number of ballots cast for both candidates.

Trump increased his share of the national Hispanic vote to 35 percent. With 60 percent or less of the national Hispanic vote, it is arithmetically impossible for a Democratic presidential candidate to win Florida, Arizona, Nevada, and New Mexico. 

 Interestingly, if you take the exit polls here and filter them by state to look at Florida, Arizona, and Nevada (New Mexico, unfortunately is not available) you'll see that Biden got 61% support from Latinos in Nevada (and won), 61% of Latinos in Arizona (and won) and 53% of Latinos in Florida (and lost.)  I don't know why the author thinks that the national Latino exit poll average has some magical mathematical certitude, but his formula actually holds at a state level.

Midwestern states Michigan, Pennsylvania, and Wisconsin always swing in the same direction as Ohio and Iowa, their regional peers. Ohio likewise swings with Florida. Current tallies show that, outside of a few cities, the Rust Belt swung in Trump’s direction.

 Really...?

In 2004 IA and OH swung Republican while MI, WI, and PA went Democrat. In 2000 IA voted Democrat as well, leaving OH as the GOP hold out.  Can I gripe for a moment about how sloppy it is to suggest that something is "fishy" about MI, WI, and PA voting Democrat while OH and IA vote Republican because that never happens when you've failed to note that in two out of the previous five presidential elections those states did in fact split?  Does Mr. Basham even care if the stuff he's writing is accurate?  Is he trying?  There is nothing remotely interpretive about this.  It is a straight up fact which takes less time to look up than it's taking me to type this complaint.  Yet he got it wrong and thousands of people have shared his article.

Yet, Biden leads in Michigan, Pennsylvania, and Wisconsin because of an apparent avalanche of black votes in Detroit, Philadelphia, and Milwaukee. Biden’s ‘winning’ margin was derived almost entirely from such voters in these cities, as coincidentally his black vote spiked only in exactly the locations necessary to secure victory. He did not receive comparable levels of support among comparable demographic groups in comparable states, which is highly unusual for the presidential victor.

This is not true.  As I've written previously, Biden was not put over the top by an avalanche of votes from those four cities.  The percentage of votes he got in those cities was virtually unchanged from 2016.  Indeed, in Philadelphia County in particular Trump actually increased his share of the vote by 2.5% versus 2016.  And while Biden did do very well in those counties which are heavily urban counties with a large number of black voters, he did not do unusually well in those counties compared to other urban counties that historically vote strongly Democrat.  Biden got 82% of the vote Philadelphia County, 73% of the vote in Hudsen County NJ, 85% in New York County, NY, 83% on Bronx County, NY, 91% in Prince George's County, MD, etc.  These are typical urban county numbers and we see them in totally uncontested states like NY and NJ as well as the hotly contested PA.

Moving to the midwestern comparisons, Biden's 69% support in Detroit is not very different from his 67% in Cleveland, OH or 66% on Columbus, OH.  And note that Wayne County (Detroit) is 40% black while Cuyahoga County (Cleveland) is only 30% black and Franklin County (Columbus) is 21%.  

So, is there something fishy here?  No.  These cities voted much like they did in 2016 in terms of percent breakdown and they voted much like similar counties in states that were not battleground states.

We are told that Biden won more votes nationally than any presidential candidate in history. But he won a record low of 17 percent of counties; he only won 524 counties, as opposed to the 873 counties Obama won in 2008. Yet, Biden somehow outdid Obama in total votes.

Sigh...  "We are told", eh?  Well, it's actually not hard to explain.  Should Mr. Basham wish to take his intellectual curiosity over to the US Census website, he can consult the tables of US population by county and there discover that the 500 largest counties contain 76% of the total US population.  As he has already noted, Biden won solid victories in many urban counties.  He also significantly improved his performance in mid-density suburban counties.  

Let's go to a non-swing state to look at an example.  In Cook County, IL (Chicago) Biden for 76% of the vote versus Hillary's 78%.  However, when we look at the surrounding counties, Biden outperformed Hillary. (using this handy county vote data visualization)

Will County (Clinton 53%, Biden 54%)
DuPage County (Clinton 58%, Biden 59%)
Lake County (Clinton 61%, Biden 62)
McHenry County (Clinton 46%, Biden 49%)
Kendall County (Clinton 49%, Biden 53%)

If we roll back to 2004, every one of those suburb and exurb counties went for Bush.

Will County (Bush 52%)
DuPage County (Bush 54%)
Lake County (Bush 51%)
McHenry County (Bush 60%)
Kendall County (Bush 61%)

These are all counties with hundreds of thousands of people in them, and they've all shifted blue since the days of George W Bush.  The median US county has less than 30,000 people, and a huge number of those very low population, rural counties went for Trump.

Victorious presidential candidates, especially challengers, usually have down-ballot coattails; Biden did not. The Republicans held the Senate and enjoyed a ‘red wave’ in the House, where they gained a large number of seats while winning all 27 toss-up contests. Trump’s party did not lose a single state legislature and actually made gains at the state level.

Trump actually ran behind most Republicans in 2016 as well.  Despite his reputation for being a "winner", he's not as good a vote getter as a lot of other Republican in competitive states.  For instance in 2016 Trump won Georgia with 50.4% of the vote, but Republican Senator Johnny Isakson won his senate race with 54.8% of the vote. In Ohio, Trump won with 51.3% of the vote in 2016, but Rob Portman won his senate seat with 58% of the vote.  In Wisconsin Trump won with 47.2% of the vote, but Ron Johnson won the senate race with 50.2% of the vote.  

Further, the GOP was recovering from a particularly bad 2018 performance, when highly motivated Democrats turned out in a "blue wave" election which won the House back from the GOP, reducing the GOP seats from 241 to 197.  In 2020 we won back some, but not all, of those lost seats, still remaining in the minority.

And honestly, if the theory is that the Democrats "stole" the election, isn't it awfully strange that they only thought to steal the White House and let themselves get beaten down to a razor thin majority in the House while failing to take the Senate as so many polls had predicted?

Despite poor recent performances, media and academic polls have an impressive 80 percent record predicting the winner during the modern era. But, when the polls err, non-polling metrics do not; the latter have a 100 percent record. Every non-polling metric forecast Trump’s reelection. For Trump to lose this election, the mainstream polls needed to be correct, which they were not. Furthermore, for Trump to lose, not only did one or more of these metrics have to be wrong for the first time ever, but every single one had to be wrong, and at the very same time; not an impossible outcome, but extremely unlikely nonetheless.

 This is a rather silly, tea-leaf-reading kind of objection, not totally unlike the complain that certain counties always vote for the winner (until they don't) or that you can predict the presidential election via how some baseball team does.  This is referring to models that try to predict the outcome of a presidential election based on economic metrics (inflation, unemployment, etc.) rather than by looking at polls.  These models are usually based on the assumption that if the economy is good, the president will be elected.

Of course, the issue in 2020 is that this is such an extraordinary year.  We do not have a good model for how elections behave when the economy was strong but is now struggling under the pressure of a pandemic which the president is accused by the other party of having badly mishandled. 

Having provided all that background, this claim is really odd.  The author argues that for Trump to lose the election, that would mean that the polls would have to be correct, but that in years when the polls were not right in predicting the presidential election, the economic based models have always been right.  So let's break this claim into parts:

1) Were the polls wrong?  Well, Trump did do better than pre-election polling suggested.  Polls suggested he would lose Florida, that Ohio was a dead heat, and that the Democrats might have a chance of capturing Texas.  Trump won Florida and Ohio and won Texas by a solid margin (though much less than GOP candidates have in prior years.)  But on a national basis the final RCP polling average predicted the results would be 51% for Biden and 44% for Trump with the rest undecided or third party.  The final results were 51% Biden 47% Trump.  That's not massively off, though it is certainly true that it's worth noting that state and congressional polls were off in key ways.

2) Did econometric models predict Trump would win?  I'm sure some did, but when I googled for these, the very first result I came up with was on predicting the Democrats would win.  

3) So if the author's incorrect assertion that either the polls or the econometric models MUST be right, then if the polls were wrong that Trump would lose, we'd expect the econometric models to be right that Trump would lose.  But of course, if Trump did in fact lose, there's not even a problem here by the author's own incorrect set of claims about how one of these must be right. 

Then Basham goes on to list, "The following peculiarities also lack compelling explanations."  If you're already getting tired of this, and you're beginning to trust my analysis over his, I'll tell you that there are nine of these and that none of them are actually very compelling objections, but I'm going to shoot for completeness here without utterly exhausting my readers.

1. Late on election night, with Trump comfortably ahead, many swing states stopped counting ballots. In most cases, observers were removed from the counting facilities. Counting generally continued without the observers

The key issue here is that in several key states (which include PA, MI, and WI where GOP controlled legislatures had wrangled over how to process the expected avalanche of mail-in ballots during the pandemic and had ended up ruling that unlike in OH and FL, mail in ballots could not be pre-processed to make counting easier and faster on election day) there were big backlogs of mail in ballots that were still needing to be processed election night.  Because Trump had been telling his supporters that mail in voting could not be trusted and they should vote on election day, mail in ballots trended heavily Democratic while in-person ballots trended heavily Republican.  This meant that as this backlog of mail in ballots was processed, the vote would swing in Biden's direction.  

However, a whole mythology has built up in loosely sourced pieces like this one that "the vote counting was stopped" and then numbers started to appear that were strongly in Biden's favor.  I don't know why it's shocking that some counties might slow their counting efforts as the night got into the small hours of the morning.  Having myself had to report to the precinct I was working at as a Republican precinct election official in Delaware County, OH at 5:30am, I would definitely have been pretty tired as thing got past midnight if I'd been having to supervise counting as well.  

This piece at The Dispatch rounds up a number sources to rebut the claim that there was a broad stop in counting.  They say:

There was no “stopping the voting” on Election Day. Several states paused vote counting late at night, hardly an unusual move considering the human need for sleep, but there have been no credible reports of voters being turned away at the polls before they closed. And how did everybody decide to stop all at the same time? They didn’t. Several examples from battleground states: Butler County, Pennsylvania, stopped their count at 11 p.m. because of a scanner breakdown. Fulton County, Georgia, stopped scanning absentee ballots at 10:30 p.m. Ballot counting did not stop at all in Wisconsin—where state law mandates no interruption can occur—or in Philadelphia, where officials announced they would stop issuing results at 9:30 p.m. but continued counting ballots. North Carolina stopped counting votes on Election Night for a very simple reason: They ran out of ballots to count.

Basham's next claim:

 2. Statistically abnormal vote counts were the new normal when counting resumed. They were unusually large in size (hundreds of thousands) and had an unusually high (90 percent and above) Biden-to-Trump ratio

The WSJ had a piece which addressed the "big vote drops" claim pretty well just the other day:

“Ballot dumps”: It’s being painted as suspicious that big batches of votes were reported in the early hours of Nov. 4. To take Wisconsin: Mr. Trump complained in a tweet that Joe Biden got “a dump of 143,379 votes at 3:42AM.” But the explanation is prosaic: Contemporaneous reporting says this is when Milwaukee’s central counting location finished with roughly 170,000 mail ballots. They included votes for both candidates but broke heavily for Mr. Biden.

The timing is unfortunate, but Wisconsin law doesn’t let counties process absentee ballots until Election Day, unlike states that reported early, including Florida. Still, the margin in Milwaukee County doesn’t look crazy: Mr. Biden won 69% to 29%, compared with Hillary Clinton’s victory of 65% to 29%. As a share of Wisconsin’s vote total, Milwaukee County fell to 13.9%, from 14.8%. A recount finished last week increased Milwaukee’s tally by only 382 votes.

The same goes for Michigan, which reported a similar batch of ballots in the wee hours of Nov. 4. State law says mail votes can’t be processed until one day before the election. The overnight jump for Mr. Biden appears to have come from Wayne County, which includes Detroit. But again the margins aren’t wild: Mr. Biden won there 68% to 30%, compared with Mrs. Clinton’s 67% to 29%. As a share of Michigan overall, Wayne County fell to 15.8%, from 16.2%.

Next up: 

3. Late arriving ballots were counted. In Pennsylvania, 23,000 absentee ballots have impossible postal return dates and another 86,000 have such extraordinary return dates they raise serious questions

These "impossible dates" reflect claims that the date the state recorded it received back a mail-in ballot was the same day or the next day after it was mailed.  

Given that PA had drop boxes where a voter could drop off a mail-in ballot, I wouldn't say it's necessarily impossible for a voter living near a mailing center to receive a mail-in ballot the same day or the day after it was mailed and then go drop the completed ballot off at a drop box on the same day he received it.  Indeed, some eager to make sure he didn't forget might do exactly that.  

What this claim also misses is that the state may simply have been sloppy in recording the dates.  If so, that's bad, and it should be done better.  But the key thing is of course that these ballots needed to be processed on the merits: did they have a valid signature, was it a registered voter, etc.  Jumping straight to the assumption that these represent sloppy fraud which helpfully left clues for online sleuths rather than efficient voters or sloppy state record keeping of valid votes strikes me as unreasonable.

4. The failure to match signatures on mail-in ballots. The destruction of mail-in ballot envelopes, which must contain signatures

Basham doesn't even provide a specific objection here, but the claims I've seen about not matching signatures on mail in ballots have all related to claims that third party observers felt they didn't get enough of a chance to watch the signature matching work, not claims that the signatures didn't actually match on specific ballots.

Via a piece at The Dispatch, here's the Georgia signature verification process as described by the Atlanta Journal Constitution:

Before election officials counted absentee ballots in Georgia, they checked voter signatures to help make sure that ballots came from the voters who returned them. That verification process reviewed signatures on absentee ballot envelopes when they were received at county election offices. Then ballots are separated from envelopes to protect the secret ballot, leaving no way to link voters to the candidates they chose. The right to cast a ballot in secret is guaranteed by the state Constitution.

Next: 

5. Historically low absentee ballot rejection rates despite the massive expansion of mail voting. Such is Biden’s narrow margin that, as political analyst Robert Barnes observes, ‘If the states simply imposed the same absentee ballot rejection rate as recent cycles, then Trump wins the election’

 This results from confusing the percent of ballots rejected for signature issues with the percent of ballots rejected for arriving after the deadline.  I'll let The Dispatch do the legwork on this one:

According to a press release from the Georgia secretary of state’s office, the rejection rate for absentee ballots in the November 2020 general election over signature issues was .15 percent, on par with the rate for Georgia’s 2018 election.

Out of 1,322,529 absentee ballots, there were 2,011 absentee ballots rejected in November 2020 because of signature issues. In the 2020 primary election, 3,266 ballots were rejected among a total of 1,151,371 absentee ballots, a rate of .28 percent.

“The lower rejection rate in the general election compared to the primary is likely the result of both parties attempting to help voters cure their absentee ballots pursuant to the process set forth in Georgia statute,” the secretary of state’s office said in a statement.

In a November 17 press conference, Gabriel Sterling, a voting system implementation manager in Georgia, said that the rejection rate for signature issues in the 2020 election was also consistent with the rejection percentage in the 2016 election.

This 4 percent rejection rate that Trump cites is misleading because it includes ballots that were rejected for arriving after the 7 p.m. Election Day deadline, and not because of signature issues.

“The biggest chunk of ballots are rejected or counted as rejected showed up after the 7:00 PM deadline,” Sterling said. “That is where the majority of rejections come from. But for signature matches always run around 0.15 to 0.2%. That is the normal thing we’ve seen in Georgia for years and we also have a new situation now where Republicans and Democrats alike get these lists of people who have absentee ballots to cure.”

Ari Schaffer, the press officer for the secretary of state responded to Trump’s tweet to clarify that “He's confusing signature issues with ballots rejected for coming in late.”

Next:

6. Missing votes. In Delaware County, Pennsylvania, 50,000 votes held on 47 USB cards are missing

This refers to a case in Delaware County, PA where a Republican poll watcher claims he saw people working on the voting machines and loading USB thumb drives to do something to them.  It's not clear why Basham thinks that 50,000 votes are missing.  There were 328,329  total votes cast in Delaware County PA, of which Biden got 63%.  This represents an 18% increase from 2016, which is actually somewhat above the PA average vote increase of 12%.  Does Basham think that there are 50,000 more votes in this county, which would mean its vote count increased by 32%?  That would be kind of a crazy number.  But really, all we have here is a "some guy saw something he thought might be wrong" kind of claim.  It was testified about in front of the PA legislature.  If there's something here, let the courts sort it out.

7. Non-resident voters. Matt Braynard’s Voter Integrity Project estimates that 20,312 people who no longer met residency requirements cast ballots in Georgia. Biden’s margin is 12,670 votes

This is based on analysis where someone compared names on the voter rolls to names on the post office's list of people who filed change of address forms to some destination out of state.  Of course, key things this approach misses would be members of the military who are from Georgia but are temporarily deployed to some other place, and students who are maintaining their Georgia residency while attending a university in another state.  (I myself was from 1997 to 2000 a California voter who had filed a change of address to have my bills forwarded to by college address in Ohio.)  Another far more prominent voter with a change of address is a Florida voter who is currently getting his mail at 1600 Pennsylvania Avenue.  

Georgia is investigating claims that people have been registering non-residents, but does not believe it was a significant factor in the election.  

8. Serious ‘chain of custody’ breakdowns. Invalid residential addresses. Record numbers of dead people voting. Ballots in pristine condition without creases, that is, they had not been mailed in envelopes as required by law

This is too vague to address, but refers to a host of claims which people have been eager to make in tweets and press conferences but which the Trump team has actually been quite hesitant to make in court.  We have an innocent until proven guilty system.  If people want to make the claim that 2020 was the result of dead people voting, etc., they need to make the case and show it was significant.

 9. Statistical anomalies. In Georgia, Biden overtook Trump with 89 percent of the votes counted. For the next 53 batches of votes counted, Biden led Trump by the same exact 50.05 to 49.95 percent margin in every single batch. It is particularly perplexing that all statistical anomalies and tabulation abnormalities were in Biden’s favor. Whether the cause was simple human error or nefarious activity, or a combination, clearly something peculiar happened.

First off, it's important to note that this kind of thing results from people trying to read the entrails of the NY Times election updates:

That can be mildly interesting, but it's important to understand that it's fundamentally not how we legally count election results.  Elections are carried out at the county level.  County election boards count their votes in batches and report unofficial totals as they go to the Attorney General and to the media.  But the time series results of a vote count are simply the result of which votes from which places are counted and reported when.

So what's his complaint here?  That in a state in which Biden held a slight majority over Trump, the last 53 updates to the data set all showed Biden with a slight lead over Trump.  In other words, he's complaining that the final updates to the vote total were all broadly typical of the votes of the state as a whole.  

Isn't that kind of an odd complaint?  

Georgia, at this point, is one of our surest results.  The entire state has undergone a hand count which should have resolved any doubt that somehow the vote counting machines could have mis-counted the state.  That hand count found the results of some human error, but it did not differ significantly from the initial count.  

This has been a long post.  I have addressed this piece of "just asking questions" fog generation point by point and tried to present links to plenty of sources so that the curious reader can research further on his own.  I have also, I feel, demonstrated how much less time and work it takes to do one of these vague "there are a bunch of things that bother me" lists than it takes to go through and answer point by point.  And yet the piece from The Spectator was widely shared, providing just enough detail to make readers who wanted to feel like perhaps Trump actually won with a fission of satisfaction that they were continuing to doubt based on lots of facts.

Elections are massive and complicated things, but their results are not unknowable.  This is not the closest election we've had in recent years, and there is not, at this point, reasonable cause to doubt the result: Trump lost.

While I may like Trump less than the people who are still questioning the election results (disclosure: I did not vote for Trump or Biden) I very much share their trepidation over how bad the Biden administration will likely be.  

But if there was ever good reason to doubt the election results, there is not now.  It's not unworthy to let the Trump team make their case in the courts, but they've been doing that for weeks now and failing again and again.  Not only have they failed to successfully make their case, but in many cases their claims in court have been far short of their claims in press conferences. So let the cases play out, but also look at the results.  Don't just read the questions people throw out, read the answers to those question, because there are answers.  Don't intentionally lose yourself in a fog of carefully nurtured doubt.  Not only are people running the risk of losing themselves in permanent crankery and self delusion, but now we have pro-Trump attorneys responsible telling people not to vote in the Georgia senate runoff elections in January -- potentially ceding control of the Senate to the Democrats and giving Biden far more power.  This is not rational, it is not helpful, it is not "winning" or "fighting".  It's a tantrum of self destructive behavior which needs to stop before it damages our country.