Wednesday, November 25, 2020


 A couple of points to follow up on recent posts:

1.  My last post mentioned a question on whether you "think [name's] presidency will bring different groups of Americans together, or do you think that it will divide them?"  It had been asked early in the presidencies of George W. Bush and Barack Obama, but I discovered that it was also asked under Donald Trump:

                      together       divide

12/2016              37%            58%

1/2017                34%            61%

That was lower than Bush (46% bring together vs. 40% divide in December 2000) and much lower than Obama.  In January 2020, there was a similar question:  "So far, do you think Donald Trump's presidency has helped to bring different groups of Americans together, has helped to divide Americans, or hasn't it made much difference?"  17% said bring together, 54% divide, and 26% that it hadn't made much difference.  Despite the difference in response categories, it's pretty clear that he'd fallen below even the low expectations he started with, which is interesting given that assessments of Trump on many issues remained almost constant throughout his presidency.  

2.  I've seem several more claims that Republican support among minority voters in 2020 was higher than in any year since 1960.  I've forgotten the source for most of them, but Andrew Sullivan said "more non-whites voted for a Republican candidate than in any election since 1960!", which is almost exactly what he said a couple of weeks ago, down to the exclamation point.  But it still isn't true!  It seems to have originated with a tweet from Adrian Gray, who describes himself as a "Center-Right Strategist."  He's deleted the original tweet, and replaced it with what he calls "updated" figures going back to 1968.  According to those, the Republican vote among non-whites was no longer the highest since 1960, but was still high--the second highest in the period, trailing only 2004.  I had a post which said that Republican support among ethnic minorities was just average in 2020.  Why do we come to different conclusions?  First, the composition of "non-white" voters has changed--before the 1970s, it was almost equivalent to black voters--most surveys didn't even have a Latino category until the 1970s, and many didn't distinguish Asians until sometime after that.  The Republican vote is higher among those ethnic groups than among blacks, so as they grew Republican support among "non-whites" improved, even though it stayed the same or declined among each individual group.  Second, I used exit polls, which only go back to 1972.  Gray's figures seem to use the American National Election Studies to some point and then switch to exit polls.  I'm not sure of this, but I have the impression that exit polls show more Republican support among ethnic minorities than the ANES does--if that's right, then the change in data sources would also make Republican performance in 2020 look better.

But regardless of what the exact numbers are, it's interesting that this claim has come to be widely accepted.  Many people want to believe that the Democrats are the party of "elites" and the Republicans are the party of the people.  (Or maybe I should say that conservatives want to believe it and liberals are afraid it might be true).  And multiculturalism has advanced far enough so that "the people" includes ethnic minorities.  So they jump at any evidence of strong performance among non-white voters.

[Data from the Roper Center for Public Opinion Research]

Sunday, November 15, 2020

All the right people, part 2

 Since writing my last post, I've seen two more pieces saying that the purported* Trump realignment (more support among minorities and the working class, less among the white middle and upper classes) represents the way forward for the Republican party.  The first is by Jonathon Van Maren in the American Conservative, and the second is by Kevin Williamson in the Washington Post.  An interesting point is that both of them treat the central claim as self-evident, even though Trump just accomplished the rare feat of losing as an incumbent and the unprecedented feat of never having had an approval rating of over 50%.  Why would someone want to preserve a coalition that produced unpopularity and defeat?  You could say that it offers promise for the future because the non-white population is bound to increase.  On the other hand, so is the population with a college degree, so focusing on appeal to less educated people works against your long-term prospects.  As I've observed, there just seems to be a general sense that getting votes from disadvantaged groups is more worthy that getting the same number of votes from advantaged groups.  That view has always been prevalent on the left, but now the right seems to have adopted it too.  

In my last post, I suggested that Trump's appeal to less educated people was partly a matter of style rather than polices.  Williamson makes a similar point:  "Trump’s followers may thrill to the promise of trade wars and border walls, but what excites them even more is his gleeful transgression. They do not embrace him in spite of the schoolyard insults, Twitter tantrums and conspiracy nonsense, but because of these things..."   He says that they are reacting against elite progressive culture:  Trump's "constituency consists . . . ordinary people who woke up one morning to find themselves re-christened 'Latinx' or were sent to a corporate reeducation seminar because they didn’t get the memo on nonbinary pronouns."  But the problem with this argument is that elite progressive culture is an elite concern--most people have probably never even heard of the term "Latinx" or nonbinary pronouns. 

So what were people reacting against?  I think it is political conflict and gridlock.  Obama appealed to some of the same sentiments as Trump--he suggested that he would get things done by transcending traditional ideological divisions and bring people together, and at one time a lot of people, including people who didn't vote for him, believed that he could.  Between 2000 and 2009, CBS News had a number of survey questions that went "do you think George W. Bush's/Barack Obama's presidency will bring different groups of Americans together, or do you think that it will divide them?"  With Bush, opinions were consistently split, about 45% saying bring together and 45% saying divide.  Immediately after Obama's election, 69% said that they thought he would bring groups together, and only 16% said that they thought he would divide them.  Even in April 2009, it was 63% to 25%.  However, as Republicans maintained a united front of opposition, conflict increased rather than declined, and people turned to someone who said he would get things done by being tough and not caring if he offended people. 

*I say "purported" because it's not clear that there's been any change among racial minorities.  See my post from Nov 9.

[Data from the Roper Center for Public Opinion Research]

Saturday, November 14, 2020

All the right people

 It used to be pretty common for parties to re-nominate a candidate who had lost a previous presidential election, most recently Richard Nixon in 1968.  But in recent decades the tendency has been to denounce any losing candidate as inept, dismiss him as a future contender, and sometimes to excise his/her legacy from the party.  Trump is an exception to this (as to many other things)--he's being mentioned as a possible candidate in 2024, and even many opponents of Trump say that "Trumpism" will remain a force.  Why?  It's not because of his overall performance--he got only 46% of the vote in 2016 against an unpopular opponent, and lost by a significant margin in 2020 despite the advantages of incumbency.  Rather. it seems to be because of the perceived composition of his votes--he had "a coalition that was more blue-collar and nonwhite" than those of previous Republican candidates.  As I have said before, there seems to be widespread sense that getting support from racial minorities and working-class people is more worthy than getting support from whites or the middle and upper classes. 

It's not clear that he actually did better than previous Republican candidates among non-white voters, or among blue-collar voters, but he did do well among whites without a college degree.  Republicans have been gaining among this group for a long time, but support jumped in 2016 and stayed high in 2020.  One popular view about why this happened was that he appealed to their economic interests with his positions on immigration and trade; another is that he diverted them from their economic interests by talking about threats to their status.  A third possibility, which has received less attention, is that his general style or image was appealing to less educated people.  I've discussed this a few times, but want to come back to it by considering a historical parallel.

In December 1953, the Gallup Poll had a survey which included several questions about Joe McCarthy--whether you approved of him, of his methods, whether you thought he should be the Republican nominee in 1956, whether you'd vote for him in a three-way race with Eisenhower, Stevenson, and McCarthy, and whether you thought he hurt or helped our relations with allies.  I divided education into not a high school graduate (53% of the sample), high school graduate (37%), and college graduate (10%).  On all of the questions, less educated people showed more support for McCarthy.  For example, 51% of people without high school diplomas, 44% of those with high school, and 30% of those with college degrees said they approved of his methods.  Of course, Republicans had more favorable attitudes, but at that time Republican support was higher among more educated people, so the relationship with education is stronger when you take account of partisanship.  Here is percent approving of his methods by education and 1952 vote:

                             Eisenhower       Stevenson        Non-voter

No HS                         74%                  45%                    71%

HS                               71%                  28%                   56%

College                        59%                  26%                   47%

 The appealing feature of this comparison is that there's no reason to think that views of McCarthy were connected to economic interests,  or that the status of working class people was threatened at the time.  So the educational difference is likely to be a direct result of education, not a reflection of something else.  The general point about the relationship between education and support for McCarthy was known at the time, but was generally seen in terms of "authoritarianism," which is a more complicated and controversial idea.  The characteristic view of less educated people might better be called something like "impatient" rather than "authoritarian":  someone could say that some of McCarthy's charges were excessive, but at least he was doing something about a problem, and not being held back by concerns about what people would think.   That's similar to what Trump's defenders said.   

Returning to the point, I doubt that a more conventional Republican--someone who did more or less what Trump did but didn't tweet about it--could maintain Trump's appeal to less educated voters.  The idea that he didn't care about offending people was part of his appeal.

[Data from the Roper Center for Public Opinion Research]

Monday, November 9, 2020

Except for all of the exceptions

 In a recent post, I mentioned a claim that the Republicans had their best performance among minority voters since 1960.  I said it seemed to be "either misleading or simply wrong."  I would probably have forgotten about it, but this morning I saw a column from Fareed Zakaria in the Washington Post saying "Trump appears to have won a larger share of the minority vote than any Republican since 1960." (It was dated Nov 5, so I was a few days late).  It included a link to the source (and he deserves credit for that), which was an article in the National Review, which quoted a tweet by someone who I think was described as a Republican strategist, but the original tweet was deleted.  When I looked back this afternoon, the link was no longer there, but there was an "Editor's note":  "An earlier version of this article and headline incorrectly stated that Trump won the largest share of non-white voters of any Republican presidential candidate since 1960. In fact, George W. Bush won a larger share of Hispanic and black voters, and Nixon, Ford, Reagan, and Dole matched or eclipsed Trump’s support among blacks." That led me to wonder about the actual changes of the "minority" vote.  Here is a figure, using election polls since 1972:

For Asian-Americans, there is a downward trend in the Republican vote.  (There has been a lot of immigration from Asia, so perhaps it's because the composition of that population has changed).   For blacks and Latino/Latina, no trend.  The Republican vote among blacks was over 15% in 1972 and 1976, and has bounced around between 9% and 12% since then, except when Obama was running.  The Hispanic vote has tracked the non-Hispanic white vote pretty well (correlation of .59), although it has varied more (perhaps because of a smaller sample).  Overall, the general picture is stability in racial/ethnic differences, except for the shift of Asian-Americans.  

As far as the 2020 election, you might regard it as interesting that Trump didn't lose ground among blacks and Latinos, but his performance was not especially good--it was about average for a Republican.  For me, the most interesting thing is why this "fact" got out there so quickly--not just the general point that Trump did better among minority voters in 2020 than in 2016, but the precise but ridiculous claim that it was the best performance since 1960.  

Note (Nov 15):  The figure with the original scale--

Saturday, November 7, 2020

Clearing the decks

 I'll have a number of posts on this election--this one will consider a number of things that I found interesting but don't need a full post.

1.  Many people are talking about the terrible failure of the polls, often reviving the idea of "shy Trump voters."  It does sound like there were some substantial mistakes in some state polls, but the national polls weren't far off.  The final Economist forecast based on an average of polls gave Biden 54.4% of the two-party vote.  If 2% voted for third parties, that would mean 53.3% Biden and 44.7 Trump.  Not all of the votes have been counted yet, but I estimate that the final will be 50.9% Biden and 47.1% Trump.  So the polls were off by about 2.5%.  That's not unusual--there are often errors of about that size.  They don't have any clear pattern--sometimes they favor the Democrat, sometimes the Republican, sometimes the incumbent, sometimes the challenger.  There's a Wikipedia page on the history of presidential election polling that gives figures from the Gallup poll.  For example, in 1996, the average of the last 10 polls (October and November) was 51.4% for Clinton and 36.3% for Dole.  The election returns were 49.2% for Clinton and 40.7% for Dole.  In 2000, almost all polls showed Bush ahead, usually by about 4-5 points.  Of course, Gore was slightly ahead in the actual results.    

So although the polls were off, they weren't off by an unusually large amount.  But in a close election, any difference becomes more noticeable.  People talked about the 1996 error at the time, but have forgotten it because Clinton won easily anyway.

2.  That leads to my second point--that recent presidential elections have all been close.  I knew this alread, but when I looked at the  data I was surprised at how large the change had been.  If you define a landslide as a win by a margin of 10% or more, 11 of the 25 elections from 1900 to 1996 were landslides.  There have been no landslides in the 21st century, in fact none since 1984, so we have set a record for the longest gap between landslide elections (the previous record was 1872-1904).  A plot  of the absolute value of the margin over time in elections from 1864-2016:

  My original purpose in compiling these data was to look at the relationship between popular vote and electoral vote.   Here it is, with the Republican share of the two party vote (centered at zero, so 60% Republican is 0.1, 45% Republican is -.05, etc) on the horizontal axis and the log of the ratio of Republican to Democratic electoral votes on the vertical.

That's close to a linear relationship, except for some outliers at the high end of Republican vote.  They occurred in the days of the "solid South," when the Southern states voted Democratic regardless of what the rest of the country did, which put a ceiling on the potential Republican majority in the electoral college  I excluded the four elections where the Republican share of the two party vote was over 60% (it would have made more sense to leave 1972 in, but that rule excluded it), and estimated a regression, then calculated the probability of winning the electoral college at various shares of the popular vote, assuming the errors have a normal distribution.

popular vote share        Chance of winning EC

50%                            50%

50.5%                        69%

51%                            84%

51.5%                        97.5%

52%                           99%

I didn't report this before because it was more of an amusement than a serious analysis, but given recent history, the question of how big a lead in the popular vote has to be before you can be pretty sure of winning the electoral college is important.    I assumed a constant error variance, which probably isn't right--it seems to increase as the election gets more lopsided.  Adjusting for that would increase the chance that the winner of the popular vote would lose the Electoral College.                         

Friday, November 6, 2020

What happened?

 In 2016, Donald Trump got about 46% of the vote.  In 2020, he got around 47% (47.7% according to the last figures I saw, but it will go down as California gets around to counting its ballots).  So hardly any change in overall support for Trump (although voters turnout was higher this time)--but has there been any change in who voted for him?  There's been a lot of attention to his increased support among non-white voters:  for example, Andrew Sullivan said "the GOP got the highest proportion of the minority vote since 1960!"  But I haven't seen any systematic comparison of changes in support among different groups, so I'll give one here, based on the Edison exit polls:

Republican Vote, 2016-2020


                                        2016               2020            Change
Men                             52%                     49%           -3     

Women                       41%                     43%           +2

GAP                              11                         6

White                            57%                    57%           0

Black                            8%                      12%            +4
Latino/a                      28%                   32%           +4
Asian-Am.                  27%                     31%            +4

GAP (B/W)                 49                        45    

GAP (L/W)                29                        25    

White Men                 62%                     58%           -4
White Women           52%                     55%           +3
Black Men                   13%                     18%           +5%

Black Women            4%                       8%             +4%

Latino                        32%                     36%           +4%

Latina                         25%                     28%           +3%


Age 18-29                   36%                     35%           -1%

        30-44                   52%                     55%           +3%

        45-65                   52%                     49%           -3%

        65+                       52%                     51%            -1%

GAP(Y/O)                   16                         16


Urban                          34%                     37%           +3%

Suburban                    49%                     48%           -1%

Rural                            61%                     54%           -7%

GAP (U/R)                 27                        17              



                                        2016               2020            Change
White Coll.                 48%                    49%           +1%
White non-C              66%                     64%           -2%
Non-W coll                22%                     27%           +5%
Non-W non-coll       20%                     26%           +6%
GAP (White)              18                        15

under $50,000         41%                     42%           +1%  
    $50K-49,999         49%                     43%           -6%
    $100,000+             47%                     54%           +7%
GAP (LO/HI)             6                          12


LGBT                           14%                     28%           +14%

Not LGBT                   47%                     48%           +1%
GAP                              33                        20

Veterans                     60%                    52%           -8%        
non-Vets                     44%                     46%           +2%

GAP                             16                         6


W. Evangelical           80%                    76%           -4%

All others                    34%                     37%           +3

GAP                              46                        39 

I highlight the ones for which the gap changed by more than five percentage points.

On race and ethnicity, Trump did improve among both blacks and Latinos, but not by much--the black/white and Latino/white gaps are still very large.  You could say that the increase among blacks--from 8% to 12%--is big in relative terms, but 12% isn't out of line with recent Republican performances--in 2004, George Bush got 11% of the black vote (and 44% of the Latino vote).  So Sullivan's statement about it being the highest Republican support among racial minorities since 1960 is either misleading or simply wrong. 

I can think of plausible explanations for two of the changes.  In 2016, there were almost no differences by income--in 2020, there was a small but clear tendency for the Trump vote to increase with income.  I would guess that this is because in 2016 there were some doubts about his economic policy, but as president he's followed a traditional Republican approach.  The decline in support among evangelicals might be a reaction against his personal conduct--in 2016, you could believe that he would "grow into the office," but by 2020 it's clear that he is what he is.   The other changes are puzzling--I can't think of a good reason why the urban/rural, LGBT/straight, or veteran/non-veteran gaps would decline.  In fact, I would have guessed that the urban/rural and veteran/non-veteran gaps would increase, since his policies and rhetoric favored the military and disfavored cities.  Of course, these figures affected by sampling error, and some of them are crudely measured, so they could be illusory.  Still, there's clearly a lot for social scientists to think about.