Before the vaccine was available, the only substantial variation among Democrats was that intention was lower in September 2020--the ratio in December 2020 was just a little larger than it had been in May (68%-17% in May and 75%-16% in December). For Republicans, there was a drop between May and August, and it stayed pretty much constant after that point. The ratio was somewhat lower in December than it had been in May (54%-29% in May and 49%-39% in December). Once the vaccine was available, the ratio increased for both--that is, more people got the vaccine than had said they would--but it increased more for Democrats. In the last survey (August 2021) 84% of Democrats said they had gotten the vaccine and 4% said they they didn't intend to; among Republicans it 56% said they had and 29% that they didn't intend to. So there was a gap from the beginning, but it grew during 2020 and grew even more after the vaccine became available. Of course, there's no way to be sure why it grew, but it's reasonable to think that it was because many leading Republican politicians played along with anti-vaccine sentiment.
Just the social facts, ma'am
Tuesday, April 1, 2025
Motes and beams
Before the vaccine was available, the only substantial variation among Democrats was that intention was lower in September 2020--the ratio in December 2020 was just a little larger than it had been in May (68%-17% in May and 75%-16% in December). For Republicans, there was a drop between May and August, and it stayed pretty much constant after that point. The ratio was somewhat lower in December than it had been in May (54%-29% in May and 49%-39% in December). Once the vaccine was available, the ratio increased for both--that is, more people got the vaccine than had said they would--but it increased more for Democrats. In the last survey (August 2021) 84% of Democrats said they had gotten the vaccine and 4% said they they didn't intend to; among Republicans it 56% said they had and 29% that they didn't intend to. So there was a gap from the beginning, but it grew during 2020 and grew even more after the vaccine became available. Of course, there's no way to be sure why it grew, but it's reasonable to think that it was because many leading Republican politicians played along with anti-vaccine sentiment.
Sunday, March 23, 2025
Which way is up?
A few days ago, the New York Times printed a discussion among David Brooks, Ross Douthat, David French and Bret Stephens. The moderator, Patrick Healy, started off by saying "more registered voters think America is on the right track than at any point since 2004, according to a new NBC News poll." It's not actually the highest since 2004 (I think NBC News was just counting their own polls), but it's definitely above average. In the latest NBC News poll, 44% said the country was going in the right direction and 54% said it was on the wrong track. I have recorded results from about 100 surveys that asked that question between 1971 and 2022 and the median was about 35% right direction and 59% wrong track. The figure below shows results from 2024-2025:
Blue dots are before the election, red dots are after the election but before Trump's inauguration, and green dots are after the inauguration. There was little or no trend before the election and not much change after the election, but since January 11 the "right direction" numbers have been high--what's more, they seem to be rising. Trump's approval ratings, in contrast, have been declining.
How to you reconcile these different patterns? The New York Times columnists didn't really try--they seemed to assume that "right direction" was equivalent to strong approval and went on to talk about why Trump has a lot of dedicated supporters. But if you look at the whole period since 1971, the correlation between presidential approval and "right direction" isn't very strong: for example, Obama had higher approval ratings than Trump, a lot of enthusiastic supporters, but averaged only 36%-58% in the right direction/wrong track question. I can think of two possibilities. One is that "right direction" is a combination of presidential approval and beliefs about social and economic conditions, and that Republican beliefs about social and economic conditions are more influenced by partisanship than Democratic beliefs are (as Paul Krugman has claimed). That is, the plus factor of Republicans turning around and thinking conditions are good is bigger than the minus factor of Democrats turning and thinking conditions are bad. The other is that there's a third factor, which is the extent to which a president is able to implement his agenda. That is, if the president is meeting a lot of opposition, someone who approves of the president might say that the country is on the wrong track. This account might help to explain why the "right track" is not just relatively high but rising--Trump has been able to get things done. That is, almost everyone who approves of Trump will be happy with the direction of the country. In contrast, under Obama or Trump I, there were more people who approved of the president but were frustrated because he wasn't getting his way.
[Data from the Roper Center for Public Opinion Research]
Saturday, March 15, 2025
Where the ducks are?
Since the 2024 election, a lot of people have said that the way for the Democrats to win is to move towards the center. A few days ago, Thomas Edsall had a column in which said it would be hard to do that, and they might not gain many votes even if they did. I think he made a strong case on the first point, but not on the second. Edsall started by observing that in the 1990s, 55% of Democrats had not attended college and 21% were college graduates, and now the numbers have reversed: 25% no college and 45% college graduates. He then said "These widely recognized changes in levels of educational attainment have coincided with an ideological shift." In 1994, 25% of Democrats said they were conservative, 48% said they were moderate, and 25% said they were liberal; in 2024, it was 9%, 34%, and 55%.
It's true that the educational shift "coincided" with an ideological shift, but the way he puts it suggests that the educational shift caused the ideological shift. Here is the correlation between party (strong Democrat.... Strong Republican), and ideology (extremely liberal....extremely conservative) in the GSS:
It's consistently higher among college graduates, so the change in educational composition has made some difference. But it's increased in all educational groups: the correlation among people with no college today is about as large as the correlation among college graduates was in the 1970s. But this just involves what people call themselves: what if we look at opinions? The GSS also has a question that starts "Some people think that the government in Washington is trying to do too many things that should be left to individuals and private businesses. Others disagree and think that the government should do even more to solve our country's problems" and asks people to put themselves on a five-point scale. This is a pretty good measure of conservative vs. liberal principles.
It/s also increased for all educational groups, but the increase is generally slower than the party/ideology correlation.
Finally, here's the correlation between self-rated ideology and views on the scope of government:
Again, an upward trend for all educational groups, but even smaller (especially for people without college degrees).
So it seems that some of the "ideological shift" is just a matter of people learning the correct ideological label for their party. The rising association between ideology and party also has implications for Republicans and independents as well as Democrats. According to the GSS, in the 1970s 20% of Republicans described themselves as liberals, 35% moderates, and 46% conservatives; in the Trump era (2018-22) it was 5%, 27%, and 68%. But as Democrats became more liberal and Republicans became more conservative, independents became more moderate: they went from 28/45/27 in the 1970s to 16/64/20 in the Trump era. Another way to look at it is that moderate independents went from 6% of the population in the 1970s to 10% in the Obama era (2008-16) to 14% in the Trump era. This suggests that the gains to moderation might actually be larger than before--there are more people who are completely up for grabs.
Friday, March 7, 2025
Dictatorship of The Donald?
A few weeks ago, I had a post on the relationship between people's main source of news and their opinions on three questions--whether Trump's victory in 2024 was legitimate, whether Biden's victory in 2020 was legitimate, and whether there was fraud in the 2024 election. I reduced those three variables to two--% thinking Trump's victory was legitimate minus % thinking Biden's was; and % thinking Trump's victory was legitimate plus % thinking Biden's was + % thinking that there was not widespread fraud in 2024. The first can be thought of as Democratic vs. Republican orientation and the second as general confidence in or cynicism about the political process. There was no consistent difference between consumers of new and traditional media in Democratic vs. Republican orientation, but consumers of all the new media sources were high in cynicism.
The survey also contained the following question: "A dictator is a leader who has total power over a country, with no checks and balances. Do you think (Donald) Trump will try to rule as a dictator or not?" Overall, 40% said he would, 41% that he wouldn't, and 19% weren't sure. Breaking that down by reported source of news:
The x-axis shows the average confidence/cynicism score for audiences of the different sources; new media outlets are in blue and the traditional ones are in red. People who got their news from new media sources didn't think it was very likely that he would try to become a dictator. Of course, views are also related to Democratic/Republican orientation. I adjusted for this got the following figure:
The audiences that were more cynical (Trump not legitimate plus Biden not plus fraud in 2024) were less likely to think that Trump would try to rule as a dictator. Another way to say this is that if you regress percent thinking that Trump would try to rule as a dictator on Democratic/Republican and confident/cynical orientation, both had strong and about equal effects, but confidence rather than cynicism goes with belief that Trump would try to rule as a dictator, which is the opposite of what I would have predicted. Why do we have this relationship? I would say that cynicism is associated with a sense that this is all "just politics": people said that about Biden, Obama, Bush.... so there's no reason to be especially worried now.
Going back to the issue of new versus traditional media, I think this is the main problem with the new media: not that it has a bias towards the right, but that it leads to a sense of confusion and doubt. Or putting it another way, whatever their faults, the traditional media provide a structure for interpreting the news and have some kind of agreement about more and less reliable sources.
Note: this is all about the average views of groups defined by reported source of news, since I don't have the individual-level data. Of course, this isn't necessarily the same as the individual-level relationship, but I think it's still meaningful in this case.
[Data from the Roper Center for Public Opinion Research]
Wednesday, February 26, 2025
Getting tired of winning
There have been a lot of news stories about Donald Trump since he took office, but I don't recall seeing many about his approval ratings. I looked up surveys in the Roper Center's iPoll database and calculated net approval (approve-disapprove):
I was primarily interested in questions about "the way he is handling his job as president," but before his inauguration there were a few about how he was handling the transition and about his first term, so I show them too. Those were all positive, and his ratings for the second term started out positive, but there seems to have been a steady decline, and all of the surveys in February found more disapproval than approval.
A more comprehensive collection is maintained by 538: it includes about 60 polls with the approval question, while the Roper Center has only 10. Again, there is a downward trend.
I did some regressions of the net approval rating on time in days*:
estimate se
iPoll -.35 (.095)
538 -.32 (.071)
538w -.30 (.071)
538r -.24 (.049)
The 538 collection rates the quality of different survey organizations, so I did a regression with the cases weighted by quality (538w). I also did one adding dummy variables for each survey organization, which amounts to basing the estimates differences among surveys taken by the same organization (538r). All agreed in showing a downward trend. The 538r estimate means that his net approval rating has declined by 1 about every four days, and even the lower end of the 95% confidence interval means by 1 every 6 or 7 days. This is a pretty strong trend, so why hasn't it received more attention? One reason is we're a long way from the next election--polls get more attention as elections get closer. Another is competition from all of the other things that have been happening. But I think that there's another factor--the degree to which journalists have adopted the "vibe shift" story. Trump has gained support from some influential people, and in the election did relatively well with some groups of voters that are regarded as important or interesting, but he has never been very popular with the public.
*Most of the surveys were taken over several days; I use the final day for time.
[Data from the Roper Center for Public Opinion Research and 538]
Saturday, February 22, 2025
Vibe shift?, part 4
I wasn't planning to write another post on this topic, but then I ran across some questions on immigration policy, which was a major issue in the election. In February 2017, I wrote about three questions on immigration. One was about "illegal immigrants who are currently working in the United States": whether you "favor deporting as many as possible or do you favor setting up a system for them to become legal residents?" The graph shows percent choosing "set up a system" minus percent choosing "deporting as many as possible":
Opinion was pretty evenly divided in 2007-10, but in those years the introduction just said "illegal immigrants" rather than "illegal immigrants who are working." More recently, large majorities have favored "setting up a system for them to become legal residents." Unfortunately, the last time the question was asked was in August 2020, so there's no way to know whether opinions changed during Biden's time in office.
Another question on deportation is "do you favor or oppose...deporting [all] illegal immigrants* back to their home countries?" The graph shows percent opposed minus percent favor, so as with the first graph, higher numbers represent the "softer" view.
The red dots are questions that include "all." This seems to make a difference--opinions are evenly divided or mostly opposed to "deporting all illegal immigrants" but mostly in favor of "deporting illegal immigrants."
Unlike the first question, this one has been asked a number of times in the last year or so, making it useful for assessing the idea of a "vibe shift." The figure shows opinions in 2024-25. During that period, it had options for "strongly" and "somewhat" favor or oppose, so I count strongly as 2 and somewhat as 1. There was also a variant that was new in 2024, which added "even if they have lived here for a number of years, have jobs and no criminal record?" The two versions were asked to randomly selected halves of the sample.
For the basic version, most people favored deportation, and the level of support didn't vary much. For the "even if" version, most people opposed deportation, and it seems like there was a shift towards greater opposition between the second and third times it was asked (August and October 2024). There was no change after the election (the last two were in December 2024 and February 2025).
Finally, there is this question: "Which comes closest to your view about illegal immigrants who are currently working in the U.S.? 1. They should be allowed to stay in their jobs and to eventually apply for U.S. citizenship, or 2. They should be allowed to stay in their jobs only as guest workers, but not to apply for U.S. citizenship, or 3. They should be required to leave their jobs and leave the U.S." The figure shows the average counting the first option as +1, the second as zero, and the third as -1.
Opinions are consistently on the "allowed to stay" side. There is no trend over the whole period, but opinions moved towards "allowed to stay" after the 2024 election. In the most recent survey, 51% said allowed to stay and eventually apply for citizenship, 19% said allowed to stay as guest workers, and 30% said required to leave.
Putting these together, if there has been a recent "vibe shift," it's away from support for deportation. Over the long term, there's no clear trend.
*or sometimes "immigrants who are living in the United States illegally"
[Data from the Roper Center for Public Opinion Research]
Saturday, February 15, 2025
News and views
There has been a lot of speculation about the effect of "new media" on politics, but we don't have much information. This isn't just because of the difficulty of establishing causation--most surveys don't ask people where they get their news, so we don't even know that much about the association between news source and political views. A Washington Post/Univ. of Maryland survey from December asked about main sources of news--it gave a list of possibilities and asked people to choose all that applied--and also asked if they thought Trump's election in 2024 was legitimate, if Biden's election in 2020 was legitimate, and if there was "solid evidence that there was widespread voter fraud in the 2024 election." The voter fraud question is particularly interesting because it doesn't have an obvious connection to partisanship: Trump talked about the likelihood of fraud before the election but treated the results as accurate after he won, while the Democrats expressed more confidence before but had reason to be unhappy with the results.
Overall, 79% said Trump's election was legitimate, 63% said Biden's election had been legitimate, and 16% thought that there was solid evidence of fraud in 2024. I took the percentages who held these views for each news source and constructed two variables--Trump's election was legitimate minus Biden's election was legitimate, and Trump's election was not legitimate plus Biden's election was not plus there was widespread fraud in 2024.* The first variable can be taken as left/right orientation and the second as cynicism about the political process. The figure shows the relationship between these two variables:
The blue dots are new media and the red ones are traditional media.** The audiences of the new media are higher in cynicism, but scattered across the right/left spectrum. The audiences of the traditional media cover a wide range in both dimensions, but there is a strong pattern--audiences that are farther to the right are also more cynical. To make it more concrete, among people who said they got news from "The New York Times or another national newspaper" 86% said Trump's victory was legitimate, 90% that Biden's was, and 4% that there was widespread fraud; among those who got news from Fox, 81% said Trump's victory was legitimate, 41% that Biden's was, and 23% that there was widespread fraud. That is, the Fox audience was more likely to choose the cynical answer on every question, even Trump's victory. There was no such pattern with the new media.So this evidence suggests that the main effect of the new media is not to move people to the left or right, but to reduce confidence in the political process. Of course, some of this is selection, but probably not all of it--people choose sources that are in line with their views, but then those sources reinforce those views.
*I standardized the underlying variables first so they would have equal influence.
**Note that I don't say "legacy media." Use of that term is a pretty reliable sign that someone doesn't know what he/she is talking about.