College electors, part 3

 I had two previous posts on voting preferences among college students in the 1940s-70s.  This post will look at faculty at two elite universities, Yale and Princeton.   Polls of faculty were less common than polls of students, but were taken in some elections.   In 1948, a poll of Princeton faculty found 47% for Thomas Dewey, 31% for Harry Truman, 16% for Norman Thomas (Socialist), 3% for Henry Wallace (Progressive), and less than 1% (1 out of 220) for Strom Thurmond.  The biggest difference from the national vote was with Thomas, who got only 0.3% in the election.*  I think that some of his support in the poll was sentimental (he was a Princeton graduate) and would have gone to Truman in a real election.  But either way, Princeton faculty were well to the left of the students, who went 70% for Dewey and 10% for Thurmond.  Dewey did better in the sciences and engineering.  

In 1952, 57% of Yale faculty favored Adlai Stevenson and 41% went for Dwight Eisenhower.  Engineering went for Eisenhower by a 3:1 margin, but all of the other schools went for Stevenson, with the Law and Divinity schools especially one-sided (they didn't break it down by department within Arts and Sciences).  The survey also asked about preference in three Senate races, two in Connecticut (one was to fill a vacancy) and one in Wisconsin, where Joe McCarthy was running for re-election.  The results for the Connecticut races were similar to the presidential results, but McCarthy got only 11%, while his Democratic opponent got 82%.

In 1964, 93% of Princeton faculty favored Johnson and only 4% favored Goldwater.  Four of the ten Goldwater supporters in the poll were from Engineering.  The survey also asked about party identification:  55% said they were Democrats, 30% independents, and 15% Republican.  

In 1968, a poll of Princeton faculty found 71% for Hubert Humphrey, 6% for Richard Nixon, 8% for Dick Gregory (Freedom & Peace Party, 0.1% nationwide), and 7% probably wouldn't vote.  The poll also gave space for comments, and "many expressed disenchantment, even disgust, with the choices offered for president."  One physics instructor wrote "I do not support Humphrey-Muskie; I oppose the others.  Wallace, Agnew, and LeMay appeal to the kinds of hatred that put Hitler in power, and Nixon welcomes the fruits of those kinds of hatred."   Engineering gave Nixon his strongest support, although even there he was well behind.  There was no clear difference between the science, humanities, and social science departments--Nixon was below 5% in all.  

Despite the scattered nature of the information, some points are clear.  First, there were differences by field, and they weren't just Humanities and Social Sciences vs. STEM.  This suggests that there is some intrinsic connection between intellectual interests, or certain kinds of intellectual interests, and political views.    Second, there was a big gap between faculty at elite universities and the general public by the late 1960s.  This means that the recent decline in public confidence in higher education isn't a straightforward matter of conservatives reacting to their underrepresentation on university faculties:  if it were, it would have happened long ago. Of course, the general distribution of votes is a not a complete measure of political climate, and you could argue that the decline is a reaction to some other political change on campus, like the rise of DEI.  But if people didn't react to a clear and simple change--the collapse of Republican support among university faculty--it seems unlikely that they would react to a more subtle one.    I think the decline in confidence has been driven by Republican elites:  they became more critical of higher education in recent years, and that has influenced public opinion.    Third, there were large shifts from one election to the next until the 1960s.  From personal observation I'm confident that faculty support for Republican candidates in presidential elections has been consistently low since the 1980s.  But  in 1980 John Anderson, a Republican running as an independent, got a lot of support among students at elite universities.   Anderson represented a strand of Republicanism that was once important but has now disappeared, so that raises a question of how much of the decline of Republican support among university faculty was the result of changes in the Republican party.  That is, when they had an opportunity to pick a liberal/moderate Republican, did they take it?  Unfortunately, I didn't find any faculty polls for the 1980 election.

[Data from the archives of the Daily Princetonian and Yale Daily News]

Interruption

My last two posts were on polls of college student voting from the 1940s-70s.  I intended my next post to be on polls of faculty voting, but I found more information than I expected, so it will take some time to put it together, so this post will fill the gap.

Earlier this year, David Shor posted the following figure on Twitter and wrote "Explicit antisemitic attitudes are now much more common among young voters" 

This was surprising to me, so I looked for similar questions to see if they showed the same pattern.  In 2022, a Pew survey asked if your opinion of Jews was very favorable, somewhat favorable, neither favorable nor unfavorable, somewhat unfavorable, or very unfavorable.  It asked parallel questions about six other groups:  Evangelical Christians, Catholics, Mormons, Muslims, mainline Protestants, and atheists.  The data set didn't include exact age, but classified people into four groups:  18-29; 30-49;50-64; and 65 and up.  

The means by age group, with 5=very favorable, 4=somewhat favorable.....1=very unfavorable.

Although Jews got lower ratings from younger people, they were still the highest-rated group, as they were among all age groups.  Moreover, Catholics, Evangelical Christians, "mainline Protestants (such as United Methodists, Episcopalians, etc.," and Mormons also got lower ratings among younger people, and the slopes were as large or larger than the slopes for the rating of Jews.  

In the oldest group, 51.2% said they had a favorable view of Jews and 5.2 said unfavorable; in the youngest, it was 39.7% and 7.5%.  The Blue Rose question didn't include a "neither favorable nor unfavorable option," and if you allocate that in proportion that gives about 8% unfavorable in the oldest and 16% unfavorable in the youngest.  That's a smaller difference than the Blue Rose data, but still substantial.  However, the comparison with the other religions suggests that the favorable/unfavorable question is not a particularly useful measure of antisemitism.  Other surveys have tried to measure antisemitism by asking people if they agree with various stereotypes about Jews, and this seems like a more promising approach in principle, but those questions are less common.

How should the differences among age groups in favorability ratings be interpreted?  I would say it's a mix of two things:  younger age groups have less favorable views of religion in general, and more favorable about the two groups that groups that are outside the bounds of traditional American religious beliefs.**  If you just extrapolate, it seems that atheists and Muslims will soon be the highest-rated groups among younger people.  However, I don't think that will happen--instead, the ratings of atheists and Muslims will level off, and there will be a tendency for ratings of different groups to cluster around 3.0:  i.e., more people treating religion or lack of religion as a private choice, not for them to judge either way.

*I learned about it from a recent Substack post by Matthew Yglesias.

**"Traditional" in terms of the memory of people who are alive today.

[Data from the Roper Center for Public Opinion Research]

College electors, part 2

 My last post said that there were large moves towards the Democrats at several elite universities between 1960 and 1964.  According to the polls, in 1960 the Republican vote share was higher at Yale, Princeton, and Stanford than it was in the general public; in 1964, it was lower.  What about other universities?  The Daily Illini led an effort to poll students at Big Ten universities in 1960 and 1964.  The results  (I just show Republican share, since they reported the two-party vote--they don't mention any significant write-ins).  

                       1960      1964
Illinois            52%        36%
Indiana           63%           
Iowa               58%          42%
Michigan        54%         
Michigan St.                   30%
Minnesota                       45%
Northwestern  65%         48%
Ohio St.           64%         39%
Wisconsin        56%

I found results for the University of North Carolina to represent the South.  In 1960, a poll found 52% for Kennedy and 42% for Nixon, with the rest undecided.  In 1964, it was 57% Johnson, 37% Goldwater, 5% write-ins, and 1% undecided.   Those were close to the overall results for the state in those elections (52% for Kennedy in 1960 and 56% for Johnson in 1964).  The swings at Ohio State, Iowa, and Illinois were similar to the swings in their states and considerably smaller than the swings at the elite colleges.  

 One of the stories on the 1964 Stanford poll breaks it down by class:  support for Goldwater was 37% among freshmen, 31% among sophomores, and 22% among juniors and seniors.  It also had information on parents' political preferences:  51% said both parents were Republican, 20% that both were Democrats, the the rest that the parents were independents or had different political preferences.  Taken together, these points suggest that political preferences changed while at college.  

I also ran across a few surveys of faculty, which I'll talk about in my next post.  

College electors

In a review of a biography of William F. Buckley, Louis Menand wrote: "In 1948, eighty-eight per cent [of Yale students] supported Thomas E. Dewey for President; four per cent backed Harry Truman."  Although the New Yorker has a reputation for thorough fact-checking, that seemed unlikely to me--you rarely find a margin that large in any group.   I looked in the archives of the Yale Daily News and found they reported a survey of Yale students which found 68% for Dewey, 21% for Truman, 7.5% for Norman Thomas (Socialist), 2.5% for Henry Wallace, and 1% for Strom Thurmond.*  The story said that graduate students were evenly divided--34% for Dewey and 34% for Truman--but there's no mix of 34-34 and 88-4 that produces 68-21 for the total.  By itself, this is just a piece of trivia, but the changing relationship between college education and vote is an important issue, so I while I was at it I looked for data on subsequent elections.  Procedures were not uniform, so there's some extra margin of error, but here they are:   

                R         D
1948      68%      21%   7.4% Thomas  2.5% Wallace 1% Thurmond
1952      67%      33%
1956      71%      29%
1960      64%      33%
1964      30%      70%
1968      27%      45%    11% neither   9% undecided
1972      12%      76%

I've written about similar surveys at Harvard before.  Harvard students were more favorable to the Democrats, but showed the same general pattern:  mostly Republican in the 1930s and 1940s and heavily Democratic since the 1960s.  

At Princeton:  

                 R          D
1948       72%        8%       8% Thomas   1.5% Wallace  10%  Thurmond
1952       73%      27%     
1956       73%      27%
1960       71%      29%
1964       27%      66%
1968       28%      40%      11%  Gregory   4% Wallace    4% Paulsen
1972      (35%      65%)

Princeton traditionally had a lot of students from the South, which explains Thurmond's strength in 1948.  In 1968, comedian Dick Gregory was on the ballot as candidate of the Peace and Freedom party, getting 0.1% of the national vote.  Pat Paulsen was another comedian who was not on the ballot but was running a joke campaign.  In 1972, there was a story about a poll, but it didn't give the totals, so I combined the numbers they gave to get an estimate.  Democratic support was consistently lower at Princeton than at Harvard or Yale, but it had a similar swing in the 1960s.  

Moving outside the Ivy League, here are figures for Stanford:

                R          D
1948       68%      10%       4% Thomas   8% Wallace  
1952       68%      28%     
1956       No data
1960       57%      36%      5% Pauling
1964       30%      70%
1968       24%      33%      18%  McCarthy 2.5% Wallace
1972       23%      67%

"Pauling" is the chemist Linus Pauling, who was not a candidate but was included on the student ballot.   He was an opponent of nuclear testing and supporter of nuclear disarmament, so he could be regarded as a leftist option.  Stanford shows the same pattern as the others.  

The long-term movement from Republicans to Democrats is no surprise, but the timing is interesting.  Rather than a gradual shift, there was a sudden swing between 1960 and 1964.  In 1968, there were a lot of protest or "none of the above" votes, but the 1972 distribution was similar to what it had been in 1964.  That is, in less than a decade, the campuses went from solidly Republican to solidly Democratic.  Presumably the change in 1964 was a reaction against Goldwater and/or his supporters, not a positive attraction to Johnson.  That raises a couple of questions.  First, how much difference did the candidates make?  If the Republicans had nominated different candidates (say William Scranton in 1964 and George Romney in 1968), would the eventual change have been smaller or would it just have been more spread out?  Second, why didn't Republicans make more effort to win back the universities, especially the "elite" ones?  Political analysis wasn't as data-driven back in the 1970s, but this change was big enough to be visible without high-quality data or elaborate analysis.  College graduates were a minority, but still a substantial group, and one that was clearly going to grow.  Moreover, college graduates, and particularly graduates of elite colleges, have extra political weight--they are more likely to vote, donate money, appear in the media, and run for office.  But I don't have the impression that Republicans ever made much effort to reverse or even contain the shift.  

*I've said this before, but I'll say it again:  It's remarkable that online editions of newspapers and magazines haven't developed reasonable conventions about including links to sources.  The New Yorker didn't have any link for the 88-4 numbers, but it did have a link for "Harry Truman"--another New Yorker story about his general impact. 

Transformed by Trump?

 The New York Times recently had an analysis of changes in county-level voting between 2012 and 2024.  As they summarized it, Trump's victory in 2024 was "the culmination of continuous gains by Republicans in much of the country each time he has run for president, a sea of red that amounts to a flashing warning sign for [the] Democratic Party."  Although there were a lot of interesting things in the story, I think there was a basic problem with their analysis.  They looked at what they called "triple-trending" counties:  that is, counties that had moved in the same direction politically in 2016, 2020, and 2024.  There were about 1,500 counties that moved towards Trump in all three elections, and only 57 that moved towards the Democrats in all three.  Of course, most of the Trump-trending counties had small populations, but some were large, and they had a combined population of over 40 million, compared to 8 million for the Democratic-trending counties.  

The problem is the focus on the "trend."  In 2012, Obama got 51% of the vote and Romney got 47.2%, for a Democratic lead of 3.8%.  The Democratic lead was 2.1% in 2016, 4.5% in 2020, and -1.5% (48.2% to 49.7%) in 2024.  So in the nation as a whole, the changes from one election to the next were -1.7,+2.4, and -6.0, and the change from 2012 to 2024 was -5.3.  Suppose you had a county that was tied in 2012, then was -1.5 in 2016, -3.0 in 2020, and -4.5 in 2024 and another county which at zero in 2012, -20 in 2016, -17.5 in 2020, and -23.5 in 2024.   The first one would qualify as triple-trending and the second would not, but the second is clearly the one that was more affected by Trump.  

Those are just hypothetical examples--what about real counties?  The article didn't include a list of triple-trending counties, but I identified a few.  One (Pike County, Ohio) was on a list of counties that have moved most strongly Republican over 2012-24.  I haphazardly chose five (Hudson, NJ; Cuyahoga, OH; Palm Beach, FL; Clark, NV; and Grant, WI) to represent different parts of the country.  A few days later, another Times story mentioned the small town of Kennett, MO and said  it was a strongly pro-Trump area, so looked up its county (Dunklin) and found it qualified.  The figures show the Democratic margins in presidential elections since 1972:

Dunklin, Grant, and Pike Counties all moved strongly towards the Republicans in 2016.  They continued moving Republican in 2020 and 2024, but if you were to pick the election that stood out it would be 2016--the first one that Trump was on the ballot.  Trump clearly did change voting patterns in these counties.  

    

Hudson, Clark, Cuyahoga, and Palm Beach are different.  None of them had particularly large changes in 2016.  Palm Beach, Clark, and especially Hudson had large shifts towards the Republicans in 2024, but if you want to explain that, the natural place to look is at something unique to 2024.

Another way to look at it is to plot the margin in a county against the margin in the national vote.  For Pike:

The three Trump elections clearly stand out.  For Hudson:

The Trump elections don't stand out.  I won't repeat the figures, just say that the Trump elections also stand out for Dunklin and (less strongly) Grant, but not for the others.

I'm not just making the obvious point that "triple-trending" Republican counties aren't all the same--I'm saying that they don't represent a Trump effect, even in a rough sense.  If a county moved strongly towards the Republicans the first time Trump was on the ballot and stayed there the next two times he was on the ballot, then it's reasonable to think that he had an effect, even if it followed the national trend and moved back a little in 2020.  But there are lots of things that could produce a trend over those elections (especially in fast-growing counties, where newcomers may have different political inclinations).  Trump's presence on the ballot was a constant in 2016, 2020, and 2024, so it's not clear why it should produce a trend.  You could say that he was something of an unknown quantity in 2016, so he had more effect as people got to know him better.  But that wouldn't produce a difference between 2020 and 2024.  

My explanation of the difference in the number of Democratic and Republican "triple-trending" counties is that it's a combination of two things:  increasing association between population and Democratic share of the vote in 2016 and 2020 (it stayed about the same from 2020 to 2024) and a larger overall vote shift between 2020 and 2024 than between 2016 and 2024.  The first means that a lot of small counties moved Republican in 2016 and 2020; the second means that fewer counties (of all sizes) bucked the national trend in 2024 than in 2020.  If you look closely at the geographical distribution of recent shifts, there are probably some implications for the electoral college, although it's not clear which party they would favor.  But they don't have any significant implications for the popular vote.

Mood Indigo, part 2

In December, I had a post criticizing the idea that school closures during the Covid years were imposed by "elites." I observed that the public was divided on the issue, and that most people were satisfied with the way their local schools were handling it.  A few days ago, Andrew Gelman discussed that post on his blog, so this one is some additional thoughts suggested by the comments.  

Here's some more information on public opinion.  On several occasions, a Fox News poll asked what the local public schools should do in the upcoming term:  "open fully in-person as usual"; "open in-person with social distancing and masks," "combine in-person and remote learning," or "be fully remote."

                     usual       distancing    combined    remote

July 2020       15%         21%             31%             25%
May 2021*     51%        27%             19%                1%
Aug  2021*    36%         33%             21%               7%
Jan    2022      28%         27%             30%             14%

The two marked with an asterisk just offered the first three options--"fully remote" was a volunteered response.  Data on actual practices in the 100 largest school districts can be found here.  I haven't looked at them closely, but it appears that in January 2022, only about 5 of the largest 100 school districts were fully remote, and about 15 required masks.  

The inspiration for my original post was a column by Nate Silver called "The expert class is failing, and so is Biden’s presidency."  He began by saying that "the expert class" was responsible for "the response to September 11 — the Afghanistan and Iraq Wars . . . the financial crisis and the bank bailouts. . . . Then the pandemic: what was supposed to be a triumph of management for a technocratic elite instead wound up as a worst-of-all-worlds scenario . . .  massive inflation, which was supposed to be a thing of the past."  But decisions on how to respond September 11, the financial crisis, and Covid were made by elected officials, not by an "expert class."  The decisions may have been influenced by experts, but the defining feature of expertise is that it's specialized, so that experts on the Middle East and experts on macroeconomics don't form a class with a common outlook.  Moreover, extent of expert influence differed from one case to another:  on the financial crisis, the Bush administration followed expert advice, but on the response to September 11, it seemed to decide what to do and then assemble expert opinion in support of the decision.  On Covid, I don't think expert opinion had that much effect either way:  from an early point, the basic Republican position was that the threat wasn't all that serious, and partly in response, the basic Democratic position was that it was very serious.  Here I can appeal to some data:  a CNN survey conducted May 7-10, 2020 asked "Which comes closer to your view about where the U.S. stands in the coronavirus outbreak:  the worst is behind us or the worst is yet to come?"   Republicans favored "the worst is behind us" by 68%-28%; Democrats favored "the worst is yet to come" by 73%-24.  As a result, on policy it came down to Republicans being in favor of doing less and Democrats in favor of doing more.   So overall, I think Silver's analysis made the mistake of ignoring the biggest factor--partisanship--in order to focus on something that was at most a secondary factor.

[Data from the Roper Center for Public Opinion Research]

Separate ways, part 2

 A few weeks ago, I had a post about liberal and conservative confidence in medicine, science, and education.  There was a gradual divergence starting in the 1990s, which accelerated in the last few year.  This post will look at confidence in other institutions.  The figures will show the gap between confidence among self-rated liberals and conservatives:  positive values mean conservatives have more confidence than liberals, negative mean that liberals have more confidence than conservatives.  First, "the executive branch of the federal government."

This operates like a measure of presidential approval:  liberals have more confidence when a Democrat is president, conservatives have more when a Republican is president, and the swings have increased pretty steadily.

Next, the institutions I talked about in my previous post:

I show smoothed estimates in order to give a general sense of the trends (although for some reason the lines don't show up well).  

Next, major companies, banks and financial institutions, and organized labor.

The gaps generally declined from the 1970s to 1990s and have grown since them.  For business, the gap is now about as large as it was in the 1970s; for finance and labor, the gaps are somewhat bigger.  

For the press and TV, the gaps grew gradually until recently, when they started to grow more rapidly, with a big jump in the Trump/Biden years.  

For organized religion, the gap has grown steadily since the 1970s.  For the military, it declined from the 1970s to 1980s and has grown since then.  It's now a little larger than it was in the 1970s. (I put these two together simply to reduce the number of separate figures, not because I think they have anything in common).

 

Finally, two other parts of the federal government:  Congress and the Supreme Court.  

There is one big outlier:  the Supreme Court in 2022, when liberal confidence dropped sharply, presumably because of the abortion decision (I think the shift from 2021 to 2022 is the largest year-to-year change in the data).  Before then, there was a gradual shift from liberals generally being more confident to conservatives being more confident, which makes sense given the ideological drift of the court.  For Congress, there's not much trend, but the year to year changes seem reasonable given what was happening.  

What's the overall picture?  Historically, many people didn't understand the terms "liberal" and "conservative," or interpreted them in idiosyncratic ways.  Understanding has grown over the years, so the match between the labels and people's views on specific issues has grown:  for example, people who like organized labor and distrust business are more likely to call themselves liberals.  This development would mean that the gaps would tend to gradually increase, so I'll focus on the departures from the general trends.  First, there have been some sharp increases in the last decade or so--this is important, but not surprising.  Second, between the 1970s and about 1990, the gaps in confidence for business, finance, labor, and the military all declined--that is, there are some cases of "depolarization."  This happened despite the apparent shift of the Republican party to the right under Reagan.  Why?  In a previous post, I suggested that the Democrats became less critical of the military.  Although I'm less sure, you can also make a case that the Democrats became less critical of business and less closely aligned with organized labor.  The way the story is told now, this didn't happen until Bill Clinton came along, but Clinton was just the most successful example of a type that had been around for a while.