The same old story?

Since my last post, one week ago, the tone of commentary on Ron DeSantis has become more negative--there have been comparisons to also-rans like Scott Walker and Ted Cruz.  A piece in the New York Times by Sam Adler-Bell suggests that the race is falling into the same pattern as 2016.  DeSantis "polls considerably higher than Mr. Trump with wealthy, college-educated, city- and suburb-dwelling Republicans. Mr. Trump, meanwhile, retains his grip on blue-collar, less educated and rural conservatives. For the G.O.P., the primary fight has begun to tell an all-too-familiar story: It’s the elites vs. the rabble."  

A recent Quinnipiac University poll  of likely Republican primary voters finds 40% for Trump, 36% for DeSantis, with Nikki Haley and Mike Pence in a tie for third place, with 6% each.  Education and income do make a difference: Trump leads DeSantis 41-23% among white people without a college degree and trails (38-43% among whites with a college degree).  There's a similarly large split by income.  (They don't report breakdowns by urban/suburban/rural residence, but that's pretty dubious as a measure of "elite" status.)  What about other factors?  Trump leads by 48-28% among people aged under 45, and trails 40-35% among people aged 45 and up.  He leads among women (40%-32%) and has a smaller lead (40-39%) among men.  Trump is slightly behind among whites (36-38%) and has a substantial lead (56-27%) among what Quinnipiac calls "people of color" (probably mostly Latino).   

The survey also asks about second choice:  DeSantis is first with 31%, followed by Trump with 20%, Pence with 17%, and Nikki Haley with 9%. Relative to Haley, Pence does better among people with lower incomes, less educated people, younger people, women, and "people of color."  Some of the differences are small, and many of them could reasonably be attributed to chance.  However, there's a pattern that applies to both the Trump/DeSantis and Pence/Haley choices:  the more familiar name does better among groups that generally pay less attention to politics.  This provides an alternative  explanation of the education and income differences in Trump/DeSantis support--people with less education and lower incomes generally pay less attention to politics, and people who pay less attention to politics tend to go with the candidate that they've heard of.  These explanations aren't mutually exclusive--Trump probably retains some "populist" appeal.  But some of his strength, especially his strength among "working class" voters, at this point is just because of familiarity.  

In my view, a major reason that Trump won in 2016 is that other candidates focused on attacking each other, and didn't really go after Trump until it was too late.  Ironically, this could happen again in 2024 for different reasons--his opponents could interpret his early strength as positive appeal, be reluctant to antagonize his voters, and all leave the job of attacking Trump to someone else.

How could they?

A few weeks ago, Thomas Edsall had a column that began "Donald Trump’s 2016 election victory continues to confound election experts. How could American voters put such a fractious figure into the White House?"  Although most of the discussion in the column was about the general election, I don't think that his victory in November was particularly puzzling (admittedly, I don't claim to be an "election expert").  Trump got 45.9% of the vote, which is a pretty weak performance given the strength of partisanship today (only a little better than the 45.6 Michael Dukakis managed in 1988).  But Hillary Clinton only got 48% (partly because of the strength of partisanship and partly because of her own problems as a candidate), and the Electoral College did the rest.  

The more difficult question is why Trump won the Republican nomination.  One idea that has become popular is that he benefitted from the large number of candidates.  The reasoning is that they split the mainstream vote, so Trump won the early primaries without getting majorities and developed enough momentum that he couldn't be stopped.  This analysis is based on the assumption that there was a basic division between Trump and non-Trump voters, so that if a non-Trump candidate dropped out, his votes would go to other non-Trump candidates.  A CBS News/NY Times survey from December 15 asked people who said they were likely to vote in a Republican primary about their first and second choices, and also about which candidate "you would be most dissatisfied with as the Republican nominee."  A summary of the results for the top candidates:

                      First      Second     Dissatisfied
Trump             35%       12%          23%
Cruz                16%       19%            4%
Carson            13%        13%            4%
Rubio              10%        13%           4%
Bush                  3%        10%         18%

The race was closer if you look at combined first and second place votes, but Trump still had a solid lead (47%-35%).  And while Trump "led" in the dissatisfied choices, he only got 23%--apart from Trump and Bush, they were spread widely.  So there at this point, there wasn't a strong division into Trump and anti-Trump voters.  

The survey also asked potential Republican voters to rate Trump, Cruz, Carson, Rubio, and Bush in several areas:  "ability to make the right decisions about the economy," "ability to handle the threat of terrorism," whether "he says what he believes," and "whether he has presented specific ideas about how he would accomplish his goals if he were elected president."  The percent giving favorable ratings:

                            Trump   Cruz   Carson   Rubio   Bush

Economy              82%    70%       65%    60%      52%
Terrorism              71%    72%       59%    59%     53%
Believes                76%    60%       72%    51%     41%
Specifics               59%    58%       47%    49%     42%

Trump was first on three of those, and a close second on the fourth.  Comparing him to his closest competitor (Ted Cruz), they were pretty much tied on terrorism and presenting specific ideas, and Trump had substantial leads on the economy and saying what he believes.  

So overall, I don't see support that for the idea that Trump benefitted from a splintered opposition--the main reason that he won was that Republican voters liked him (although Republican elites were less enthusiastic).  Carson's strong showing is also noteworthy--people seem to have forgotten about it, but he did well in the polls even though he didn't put much effort into campaigning.  His lack of effort and organization hurt him once the primaries started, but he clearly had an appeal to Republicans.  So that makes the contest for the nomination even more confounding:  two candidates who seemed weak on paper did well against what was supposed to be an unusually strong field.

A lot has changed since 2016, and there now seems to be a real division between Trump and non-Trump support.  But the 2016 campaign may still have lessons for today (hint:  I think DeSantis if overrated; my money is on Tim Scott).  

[Data from the Roper Center for Public Opinion Research]

Red district, blue district

 It's become common to speak of "red" and "blue" states, and to note that the "blue" (Democratic) states are richer than the "red" ones.  But the states are of widely different sizes, and some are very large, so they are probably not the best units to look at.  Representative Marcy Kaptur's office has produced a chart that shows median household income and party control by Congressional district.  There is a clear tendency for Democratic districts to have higher incomes--e. g., Ruy Teixeira says "The first page is heavily dominated by blue but the second, poorer page is a sea of red."  However, on looking at the chart, it seemed that it wasn't a simple matter of "the richer the district, the more likely to elect a Democrat."  Specifically, it seemed that the probability of electing a Democrat first declined and then increased as you went from low to high income.  But it's easy to imagine patterns when there's really nothing but random variation, so I looked more systematically.  After ranking the districts from low to high income, I computed a 50-seat moving average--seats 1-50, then seats 2-51, 3-52, ....392-441.*  The proportion of Democrats against the income of the middle district in the group:

The pattern is even more complicated than I thought--the chance of electing a Democrat first falls, then rises, then remains steady once the median income reaches about $80,000 (about a quarter of districts are above that level).  There are hints that it increases again at the very top (9 of the 10 districts with the highest median income are represented by Democrats), but it's not possible to be sure.  The Democratic share is at a minimum when the median income is about $65,000.  

I don't have any explanation--I'm just offering it as something to think about.  A few other observations:

1.  The discussion of this chart has treated it as an example of the Democrats' problems with working-class voters, but the individual and district level relationships are different issues.   Congressional districts are large (about 800,000 people), so they all contain a range of classes.  Consequently, it would be possible for the Democrats to do better in richer districts and better among working-class voters--in fact, this is what a classical Marxist analysis would predict.  Of course, this is not the case in the United States today, but it means that the district-level relationship needs its own explanation.

2.  Back in the 1950s, Seymour Martin Lipset observed that conservative parties sometimes did better in the poor regions of a nation (e. g., southern Italy).  His explanation was that those regions hadn't moved fully into a capitalist economy, so that their voters followed the lead of traditional elites, but this isn't applicable to the US today.  Curiously, the issue hasn't received much attention in subsequent research--I don't recall having seen anything, and my searches in Google Scholar drew a blank.  

3.  A common method for detecting non-linearity is to include a squared term in addition to the original.  But when you do that with these data, the t-ratio for the squared term was less than one.  In fact, you had to go up to the fifth power before one was statistically significant, and people rarely go that high.  So in this case, if you used a polynomial regression, you'd probably miss the non-linearity.   

  

*The chart includes non-voting delegates as well as representatives.  

Transvaluation of some values

 The values poll continues to get attention--since my last post, I've seen three New York Times columns discussing it (Ross Douthat, Peter Coy, David Brooks).  In the Dispatch, Nick Catoggio calls it >the hottest topic in political commentary."  My last post discussed specific issues with that survey--this one will address the general question of changes in values.

1.  What has happened:  In the 21st century, there's clearly been a turn away from religion in the United States, although you can argue about whether it's a new thing or the continuation of a longer trend.  For example, the Gallup Poll has a question on whether "religion can answer most or all of today's problems," which was asked in 1957, 1974, and pretty regularly starting in the 1980s.  The results through 2017:

It was asked again in 2018 and agreement fell below 50% for the first time.   There's also been a decline, or at least a moderation, of patriotism.  I wrote about questions comparing the United States with others in this post.  One of those questions was asked in the recent WSJ poll:  21% said it "stands above all countries in the world," 50% that is is "one of the greatest countries in the world, along with some others, and 27% said "there are other countries better than the United States."  When that was first asked in 2011, the numbers were 38%-53%-8%; in 2018 they were 24%-55%-21%.  Finally, there's also been a decline in the belief that people need to have children in order to have a full life.  I don't think I've posted on that, but it's pretty clear.  

2.  What hasn't happened:  a dramatic change in values since 2019.  This was one thing that brought attention to the survey, and made it especially appealing to the right.  From their point of view, most of the big developments since 2019 have been bad--the response to Covid, the rise in "wokeness," most of the policies of the Biden administration.  So evidence of a general collapse in values fit with their sense that things had gone badly wrong in the last few years.  But rapid changes in opinions are rare, and some of the apparent changes in the WSJ polls just aren't credible (e. g., a fall from 80% to 58% rating "tolerance for others" as very important).  So they are probably the result of the change in survey procedures.  Although social desirability bias may be a factor, I don't think it's a big factor with these opinions, especially when it comes to the difference between "very important" and "somewhat important."  But in addition to the issue about response order I mentioned last time, I think the chance that people will be careless or not take the survey seriously is greater in online surveys.  The documentation to the survey mentions that they removed about 5% of the surveys because of high rates of non-response, completing the survey too fast, or "straight-lining all grid questions."  But it seems likely that attention and effort are matters of degree, so that if 5% were clearly not paying attention, some larger number were not paying much attention.   It's hard to say exactly what affect this would have--it just means that we can't be sure about how much of the apparent change is real.  

3.  What's uncertain:  how much people value obligations to their family versus their own interests, or money versus other goals.  I think that there's a long term trend towards emphasizing non-material goals (like wanting a job that's fulfilling or does good for the world), but I'm not really confident of that.  The focus on money also goes up and down depending on short-term economic conditions, making it harder to distinguish any trend.  People clearly feel less obligation to get married and have children, but beyond that--for example, whether parents are more or less willing to sacrifice for the benefit of their children--it's hard to say.  I will look for other questions and write about them in a future post.  

What's wrong?

 A series of questions from a new survey sponsored by the Wall Street Journal is getting a lot of attention.  The figure that's been circulating shows the percent who says that the following values are very important "to you personally" in surveys taken in 1998, 2019, and 2023:

So it appears that patriotism, religion, having children, and community involvement are becoming less important to people, while money is becoming more important.  The WSJ story on the survey also mentions that the percent who say that tolerance and hard work is very important to them fell between 2019 and 2023.  So the results seem to show a dramatic turn towards selfishness in the past few years.

A few people have raised questions about this interpretation.  Patrick Ruffini notes that the 1998 and 2019 surveys were taken by telephone, while the 2023 survey was online.  He suggests that phone surveys are affected by "social desirability bias"--you want the interviewer to think well of you, so you give answers that will make you look better.  Social desirability bias might have inflated the numbers for values like hard work, patriotism, and community involvement, and held down the numbers for money.

Another new feature of the 2023 poll was that half of the sample were asked the response options ("very important," "somewhat important," "not that important," and "not at all important") in reverse order (see p. 7 here).  This may also make a difference--making a distinction "very important" and "somewhat important" may be harder than making a distinction between important and not important.  In this case, whether people say "very" or "somewhat" may depend on which answer is offered first.  In the phone surveys, "very" was always offered before "somewhat"--in half of the internet surveys, "somewhat" came first.  So my hypothesis is that the distribution of answers in 2023 differed depending on the order of the response options.  If so, comparison with the earlier years should be based only on the part of the sample that offers them in the same order as the phone surveys.  

Or you could look at the the combined percentages for "very" and "somewhat" important:

                                        1998    2019   2023
Hard work                         98%    98%   94%
Self-fulfillment                  94%   94%   91%
Tolerance                           NA     96%   90%
Money                                 88%   92%   90%

Community involvement   94%   96%    80%
Patriotism                           93%   88%    73%
Children                              84%    77%   65%
Belief in God                      NA     72%    65%
Religion                              89%    75%   60%

For the first four, there's little change--they are almost universally regarded as important at all three times.  The results for "self-fulfillment," which isn't mentioned in the WSJ article, are particularly interesting--the percent rating it as very important fell from 64% in 2019 to 53% in 2023.  That's hard to square with either the growing selfishness or the social desirability interpretations, but is consistent with my hypothesis.  These figures indicate some changes in the last few years, but not the general collapse of values that is being claimed.

Unfortunately, the Wall Street Journal doesn't make the original survey data available.  If they ever do, I will check and see if the response order makes a difference.  

Wokeness

 There have been numerous efforts to define "woke" in the past week or two.  A common problem with the ones I've seen is that they have been too elaborate, treating it as if it were a political ideology rather than general current of opinion.   I would define it as an inclination to think that discrimination is the primary cause of group inequalities (except when a generally favored group is on the bad side, as with men being overrepresented in prison).   

With this definition, we have a question that can be used to measure the growth of "wokeness." The General Social Survey asks:  "On the average (negroes/blacks/African-Americans) have worse jobs, income, and housing than white people. Do you think these differences are mainly due to discrimination?" 

Agreement was generally declining among both whites and blacks until about 2014.   but there have been large increases in 2016, 2018, and 2021.  White agreement is at its highest level ever, and black agreement is almost equal to its highest level.   Of course, opinions are affected by experience and other evidence, but those haven't changed dramatically in the last ten years---that leaves what I called the "inclination."   Ibram X. Kendi's remarks are interesting in this context:  "The racist answer is 'no'—it presumes that racist discrimination no longer exists and that racial inequities are the result of something being wrong with Black people. The anti-racist answer is 'yes'—it presumes that nothing is wrong or right, inferior or superior, about any racial group, so the explanation for racial disparities must be discrimination."  That is, he doesn't appeal to evidence, but to principle:  "presumes" and "must be."

The previous figure showed that the turn towards wokeness occurred among both blacks and whites.  It also occurred among both more and less educated whites, although it was stronger among college graduates:

This is normal--usually changes of opinion are similar among different groups, although there may be subtle differences.  But the pattern by ideology is different:

The GSS asks people to rate their ideology on a seven-point scale:  I divided it into three groups, very and somewhat liberal; slightly liberal, moderate and slightly conservative; and somewhat and very conservative.  Liberals and moderates have both moved substantially towards agreement, but there has been little change among conservatives.  There's a good deal of sampling error in year-to-year changes in subgroups, but agreement rose among conservatives between 2014 and 2016, but then declined in 2018 and 2021.  

So the division over "wokeness" is not just something that has attracted the attention of political elites and people on Twitter:  it is clearly visible in the public.  

'Twas a famous victory

 In September 2015, a Pew survey asked "Thinking about the way things are going in politics today...on the issues that matter to you would you say your side has been winning more often than it's been losing, or losing more often than it's been winning?"  The question has been repeated a number of times since then, most recently in September 2022.  The figure shows the percent who said that their side was generally winning broken down by party--the colors distinguish between supporters of the President's party (Democrats in 2015 and 2016, Republicans 2017-2020, and Democrats in 2021-2).  

Unsurprisingly, supporters of the President's party are more likely to think that they are winning than supporters of the opposition party are.  The unexpected part is that the views of supporters of the president's party are more variable--in the opposition party, "winning" ranged from 14-22%, which is probably not much more than would be expected from sampling variation; in the president's party, it ranges from 31% to 69%.  The highest value occurred in May 2019.  Trump supporters generally say that his greatest accomplishments were the tax bill, the confirmation of Supreme Court Justices, and the Abraham accords, and maybe the program to develop Covid vaccines, none of which occurred around that time.  The most plausible explanation for the positive feelings among Republicans is the release of the Mueller report, which was generally regarded as a political victory for Trump because it wasn't as damaging as many had expected.  On this interpretation, Republicans had the greatest sense of success after avoiding a loss rather than after a positive accomplishment, which shows the strength of negative partisanship.    

[Data from the Roper Center for Public Opinion Research]