When after all it was you, not me

 Since the murder of Charlie Kirk, people have been reposting one of his tweets from earlier this year, in which he said "assassination culture is spreading on the left.  Forty-eight percent of liberals say it would be at least somewhat justified to murder Elon Musk.  Fifty-five percent said the same about Donald Trump."  The source of these numbers is a report from the Network Contagion Research Institute published in April.  He didn't mention another figure from the report that I find even more remarkable--20% of conservatives said that it would be at least somewhat justified to murder Donald Trump.  Who were the people answering these questions?  A footnote says "using Prime Panels, we collected 2651 respondents.  Based on attention checks and those who took the survey in an unreasonable amount of time, we cleaned the dataset down to a final dataset of 1264 responses."  It's standard practice for online surveys to discard some cases (e. g., those that pick the first answer for every question), but more than half?  I didn't know anything about Prime Panels, but it's described here.  The key point is that it's a collection of "opt-in" panels--that is, people who sign up to do surveys in return for compensation.  That is, there's no effort to get a representative sample in the first place, and the extremely high rate of discarded answers suggests that respondents are not taking the survey very seriously (I wonder if some of the responses were generated by AI).  So we can continue the cleaning process by dropping the remaining 1264 cases and moving on.  

JD Vance also cited some survey data when he hosted the Charlie Kirk show.  He said that 24% of people who call themselves very liberal say it is acceptable to be happy at the death of  a political opponent, against only 3% of people who call themselves very conservative.  Also, 26%  of young liberals and only 7% of young conservatives say that political violence is sometimes justified.  These data  are from a legitimate survey organization, YouGov.  However, they are from surveys taken on September 10 and 11th, so they represent some combination of feelings about general principles and the particular case of violence that had just happened.  As far as I can tell, the question about whether it is acceptable to feel happy has never been asked before; there are a few general questions about whether political violence is ever justified, but they are all at least 20 years old.  There's also a problem of interpretation.  It's easy to think of examples of "violence in order to achieve political goals" that most people would regard as justified and even admirable--e. g., the American Revolution.  So I don't think that we can learn much from questions of this kind--questions about reactions to particular assassinations or assassination attempts would be more informative, but fortunately we haven't had many cases to ask about.    Moreover, there have been only few questions about those cases and they aren't comparable.

So far, my conclusions have all been negative, but there is one interesting additional piece of information in the YouGov report.  After the attack on Paul Pelosi in October 2022, they asked " How big of a problem do you think political violence is in the U.S. today?"  They repeated the question after the two attempts on Donald Trump, the attempt on Josh Shapiro, and the assassinations of Melissa Hortman and Charlie Kirk.  The figure shows the percent saying "a very big problem" broken down by party:  

Among Democrats. the level doesn't change much; among Republicans, it's substantially higher when the victim is a Republican.  Of course, with just five cases you can't draw strong conclusions, but I've found a similar pattern before--Republican views about the future of the next generation are more affected by the party of the president than Democratic views.  I think this has happened because Republican leaders have taken a catastrophist approach--saying that the Democrats, the universities, the media, etc.  are dominated by the radical Left, which is evil, vicious, and even demonic (all words used by Trump or Vance).  

The data I needed

 A new paper by Craig Volden, Jonathan Wai, and Alan E. Wiseman looks at the educational background of members of Congress over a 50-year period (1973-2021).  Their major conclusion is given in the title:  "On the Decline of Elite-Educated Republicans in Congress."  Over the years, Congressional Republicans have become less likely to have degrees from elite universities, while Democrats have become a little more likely.  As a result, a "diploma divide" has appeared, with Democrats more likely to have elite degrees.  The figure shows the percent of Democratic and Republican members of the House of Representatives who have elite degrees (see this Washington Post story for more figures).

  

Volden, Wai, and Wiseman also report that members with elite degrees have more liberal voting records than those without them.  The difference is larger for Republicans, suggesting that the declining representation of elite universities is related to the rightward move of Congressional Republicans over the period.

Taking both parties together, representation of elite degrees in the House has declined (it's been roughly constant in the Senate).  

I had done similar research on a smaller scale, comparing two Congresses (1953-5 and 2017-9).  I had expected to find a substantial increase in the percent with elite degrees, but that wasn't the case.  By my definition of elite, it only went from 17% to 18%.  I noticed that the share of elite degrees  declined among Republicans and increased among Democrats.  

Considering both studies together, it seems that the share from elite universities is about the same in recent years as it was in the early 1950s, but lower that in the 1970s:  that is, it peaked in the 1970s.  The Republican share of those with elite degrees was higher in the 1950s than the 1970s--that is, the changes since the 1970s continue a longer trend.  (Of course, our definitions of elite universities weren't identical, but they were similar enough so I'm pretty confident in these claims).  

In another post, I suggested that there has been a change in the relationship between elite university background and political orientation.  " At one time, elite colleges played an important part in creating an 'establishment'.... that's no longer the case, at least on the conservative side."  That is, I thought that Republicans with an elite educational background used to be more moderate, but that this was no longer the case.  The Volden, Wai, and Wiseman data lets me test that idea.  I divided it into four periods and compared the DW-Nominate scores of elite graduates and others (controlling for race, sex, and Hispanic status).  Negative numbers mean members with elite degrees have more liberal voting records.

                            House                        Senate
                            R            D            R            D
1973-80         -.045      -.052          -.201    -.043
1981-94        -.083       -.056           -.180    -.035
1995-2008    -.049       -.023            -.085    -.048
2009-21          .007      -.002               .021    -.070

Since 2009, Republicans with elite degrees have been just about the same as those without.  With Democrats, the difference has disappeared in the House, but stayed the same or maybe even grown in the Senate. So my hypothesis is supported:  among Republicans, having an elite degree is no longer associated with more more moderate views.  The reasoning behind the hypothesis was that as the political climate at elite universities has moved left, conservatives increasingly feel like an embattled minority, and consciously reject prevailing views rather than being influenced by them.  That suggests that the change should be mostly a generational one.  An alternative explanation is that party-line voting has become more common in recent years, especially among Republicans, so party is coming to dominate all other differences.  

What do you know?

 I've had several posts about the connection between tastes in reading and political views.  The basic conclusion is that people with more "sophisticated" tastes tend to be more liberal.  I recently ran across another relevant survey (from 1999) which asked people if they could name the authors of the following books:  The Cat in the Hat, Huckleberry Finn, The Shining, The Old Man and the Sea, The Firm, A Tale of Two Cities, A Farewell to Arms, The Great Gatsby, Moby Dick, and Crossings (I list them in order of the percent who gave the correct answer).  The mean was about three correct answers (it was open-ended, not multiple choice).   

The few political questions in the survey were not of much general interest, so I'll just look at ideological self-rating (very liberal to very conservative).  Among people who got 0-2 right (about 30% of the sample), 45% said they were conservative, 37% moderate, and 17% liberal; among people who got 6-10 right (about 20%), it was 27%, 48%, 25%.  If you regress ideology on age, sex, race, Hispanic status, education, and "knowledge" (number of correct answers), the estimate for knowledge was positive and statistically significant (.054 with a standard error of .014).  The estimate for education (1=no HS diploma... 4=college graduate) was .01 and not significantly different from zero.  

Why would knowledge of the names of authors be associated with ideological self-rating?  One possibility is that some people understood the terms in a non-political sense, e. g., "liberal" as meaning something like broad-minded.  Incorrect (or unconventional) understandings are more common among less educated people, so this would suggest that the relation between knowledge and ideology would be stronger among the less educated, but if anything it was stronger among more educated people.  So I think that there really is a connection between knowledge and political views (people who knew more authors were also more likely to say they were Democrats, although the connection was weaker).  This doesn't mean that this knowledge affects political views; rather the general curiosity that makes people learn and remember the names could also make them more critical of tradition. 

PS:  This is the 15th anniversary of my first post.   I didn't really plan to keep the blog running this long, but coming up with new material has been easier than I expected.  

[Data from the Roper Center for Public Opinion Research]

 

Long term, short term, part 2

 My last post showed that conservative confidence in science, education, and medicine dropped over the last several years.  What about other institutions?  Between 2016 and 2024, conservative confidence declined for 11 out of the 13 institutions that the GSS asks about, and one of the increases (banks and financial institutions) was very small.  The only substantial increase was for the Supreme Court.  Among liberals, confidence increased for six institutions and declined for seven.  The figure shows average liberal and conservative confidence across all institutions*:

Until 2008, conservative and liberal confidence was about the same.  Since then, liberal confidence has consistently been higher, but the size of the gap increased substantially in the 2020s.  

One of the institutions is "the executive branch of the federal government."  That is closely related to approval of the current president, so also calculated the average for all institutions except the executive:

This was close until diverging since 2018.  

So liberalism is now associated with more confidence in institutions in general.  Is this related to the growth of educational divisions in politics?  Sometimes people say that education is associated with more positive views of institutions--they'll be more likely to think that the people in charge ("the elites") know what they're doing.  There are several possible reasons, but the most obvious one is that educated people are in charge of most institutions.  Of course, the average educated person doesn't have much influence, but they'll have more shared understanding and chance of personal connections with the people who do.  On the other side, there is the argument that education makes people more critical.  I computed correlations of confidence in different institutions with degree (no high school diploma.....graduate education).

Science               .184
Supreme Court   .109
Business             .088
Medicine            .047
Executive           .031
Press                   .002

Banks                -.017
Congress           -.026
Religion            -.027
Education         -.042
Labor                -.095
Military             -.098

But these are averages over the whole time from the 1970s to 2020s.  In several cases, the correlation shifts over time.  The biggest changes:

The correlation was near zero in the early years, but is now negative.  That is, educated people have lost confidence faster than less educated people.

It started out negative and is now near zero--educated people have gained relative to less educated people.  

More and less educated people were about equally confident in the 1970s; now more educated people are more confident. This is different that the pattern for science, where more educated people have been more confident all along.

More educated people were less confident in the military in the early years,  but the difference has pretty much disappeared.  

With organized labor, you could say it reflects changes in composition--blue collar unions have declined and white collar unions (mostly in the public sector) have held steady or even grown.   But as many people have observed, the educational divide in military service has grown.  That is, educated people have become less likely to have military experience, but more likely to have confidence.  I think the changes are evidence of a point I've made in a number of posts:    that there's been a growth of social egalitarianism among more educated people.    Of course, it's possible that more educated people have less knowledge or understanding of less educated people than they used to.  But, contrary to what is often said, they have become less likely to express negative views of less educated people, or the institutions associated with less educated people.

*Except "banking and financial institutions"--that question didn't start until later.

Long term, short term

 The General Social Survey has a series of questions about confidence in "the people running" various institutions that has been asked since the early 1970s.  But the GSS was not the first to ask these questions--they were originally from the Harris Poll.  The very first time was in the 1965 survey of college students that I wrote about recently, but they were also asked of a national sample in 1967.  The figure shows average confidence in institutions in 1967 and 2024 (the most recent round of the GSS):

The line indicates equal confidence in both years.  With one exception, they are below the line, meaning average confidence was lower in 2024 than in 1967.  The exception is organized labor, where average confidence was almost exactly the same.  I think that is because organized labor has become less prominent:  there are fewer strikes or the threat of strikes, or large wage increases that might lead to inflation or tax increases.  In 1967, 23% said they had a great deal of confidence and 33% said that they had very little confidence:  in 2024, "great deal" had fallen to 18%, but "very little" had also fallen, to 24%.  So you could say that there's been a move towards indifference.

The other institutions all have substantial declines, but several stand out:

Congress has had the largest decline (-0.82; changes for the executive branch and the Supreme Court are about -0.6).  Science, military, and television have had relatively small declines (-.25 to -.33).  I've discussed confidence in the military before.  For TV, I think it's because people entertainment programming as well as news (questions specifically about TV news show a large decline).  The relatively small decline for "the scientific community" is interesting, since a lot of people have talked about the growth of anti-science attitudes in recent years.  The GSS also asks about medicine, and the decline is considerably larger than the decline for science (-.52 to -.24).  

I looked at confidence in science, medicine, and education in an earlier post, and noted that there was a growing split between liberals and conservatives in the last few years, presumably because of the reaction to Covid.  That was before the 2024 GSS came out, so here are the updated figures:

Very little change from 2022 to 2024.

A drop from 2022 to 2024 among both liberals and conservatives, but larger among conservatives.  Also, liberal confidence increased in 2018 and 2021, so the decline leaves them about where they were in 2016.  With conservatives, confidence is at its lowest level ever.  

Finally, education:

A slight increase from 2022 to 2024 among both liberals and conservatives.  

The liberal/conservative gap for confidence in education and science grew between 2016/8 and 2022, but stayed about the same between 2022 and 2024.  In contrast, the gap for confidence in medicine kept growing.  Why?  My thought is that more people are interested in medicine and feel like they have some basis for offering opinions--they can talk about what happened when they got a vaccine, or someone they know got a vaccine.  As a result, there's more momentum in public opinion when doubts start to grow.  Another possibility is that it was driven by the prominent role of RFK, Jr. in the presidential race.

[Data from the GSS and Odum Institute Data Archive]

Friendship recession, part 4

 I hadn't intended to have so many posts on this issue, but people keep talking about it.  In the New York Times today, Robert Putnam and Richard Reeves write "One in seven young men reports that he has no close friends, up from 3 percent in 1990."  The link goes to a report on the same 2021 survey that I wrote about before.*  In a previous post, I suggested that the apparent decline in number of close friends might reflect a difference in the survey format or the sample, and in this post I'll look at that possibility more closely.  The question about the number of close friends has been asked in two other recent surveys, one by Pew in July 2023 and another by Survey Center on American Life using the IPSOS Knowledge Panel in April 2024 (they also did the 2021 survey).  The percent reporting no close friends:

                   Men          Women

1990          4.4%           1.6% 
2021          16.4%        11.6%
2023            8.3%          7.4%

2024           17.3%       15.7%

The 2021, 2023, and 2024 surveys were conducted using online panels.  I suggested that people in online panels know that saying you had friends would increase the chance that you'd get additional questions, so that some of the "none" answers were just people who were in a hurry to get through.  The 2021 and 2024 surveys show about twice as many people reporting no friends as the 2023 survey, but that's consistent with my hypothesis--Pew and IPSOS may differ in their propensity to ask additional questions (or in how strictly they check for signs that a respondent was rushing through).  

My hypothesis implies that online panels will have more "nones," but doesn't suggest any changes in the relative frequency of other answers.  That is, if you say you have four friends rather than five, you'll still get follow-up questions.**  Here are the mean numbers of close friends for people who say they have at least one***.

                  Men          Women
1990          6.7               5.7
2021          5.0               4.5
2023          4.9               4.6
2024          5.4               5.3

So there does seem to be some decline, and it's larger for men, but the result not a growing gap, but a decline of the gap between men and women.  

These numbers don't fit the story about the social isolation of men today.  Of course, you could say there are offsetting differences--women "really" have more close friends, but also have higher standards for counting someone as a close friend.  That could be true, but you could make a similar argument about changes in the reported number of close friends--that is, people's standards might have risen over the years.  In my last post, I mentioned a question from 1950:  "When you have personal problems, do you like to discuss them with anyone to help clear them up, or not?" and said "Today, the general assumption is that it's a need, not something that you might or might not like to do."  It happens that the 2024 survey asked "If you were facing a personal problem, who is the first person you would turn to for help?" and offered some possibilities (spouse or partner, friend, parents...) , followed by "there is no one I could turn to."  The difference between those questions illustrates the nature of a change that I think has happened--growing emphasis on the ability to discuss "personal problems" as a defining feature of friendship. 

*The numbers in that report don't actually refer to young men, but to men in general.  
**You'll get fewer follow-ups if there are questions about the characteristics of individual friends, but those are unusual. 
***More than 10 is counted as 10.  

[Data from the Roper Center for Public Opinion Research, Survey Center for American Life, and Pew Research Center]

Friendship recession, part 3

 Some further notes on the evidence regarding a "friendship recession."

1.  My last post mentioned a survey from 2019 that asked about the number of "true friends."  That survey also had a question on loneliness:  "How much of the time do you feel lonely:  All of the time, most of the time, only sometimes, or hardly ever?" 3% said all the time, 5% most of the time, 23% only sometimes, and 67% hardly ever.  A 1990 Gallup survey asked "how often do you ever feel lonely,":  10% said frequently, 26% sometimes, 40% seldom, and 23% never.  A Los Angeles Times survey from 1989 asked "Everybody is lonely sometimes. Would you say that you are often lonely or seldom lonely?"  9% said often, 82% seldom, and 7% volunteered that they never felt lonely.    A NORC survey from July-August 2020 asked whether you had recently felt "very lonely or remote from other people."  32% said that they had. The same question was asked in a number of surveys from 1963 to 2001, with "yes" responses ranging from from 17% to 28%, so the 2020 figure was the highest ever.  But that was in the early stages of Covid, when there were significant restrictions on face-to-face interaction.  Considering all of those questions, I don't think that there's much evidence for a long-term rise in loneliness.  

2.  Claude Fischer had a post on changes in the reported number of friends.  In addition to the 2021 survey that I discussed in my previous post, he found a 2023 Pew survey in which similar numbers said that they had no close friends.  

3.  There were two surveys of teenagers (13-17) that asked about the number of close friends.  Unfortunately the categories used in the reports weren't identical, but they were close:

                      2018                  2024

0                       2%                   2%
1-5                  77%                 78%
6-9                  11%
6-10                                        15%
10+                    9%
more than 10                             5%

The results are very similar.  Moreover, the 2018 survey just asked about the number of close friends, while the 2024 survey included the qualifier "not counting family"--to the extent that matters, it would mean the comparison is biased towards finding a decline.  

4.    Back in 1950, the Gallup Poll asked ""When you have personal problems, do you like to discuss them with anyone to help clear them up, or not?" 64% said yes and 35% said no.  There was some sex difference between men and women, but it was not that large (69% yes among women, 62% among men).  Apart from the results, I think that it's interesting that they asked it that way.  Today, the general assumption is that it's a need, not something that you might or might not like to do.

[Data from the Roper Center for Public Opinion Research]