Judgment and opinion

 Exactly a year ago, I had a post about a survey question from 1993 on whether members of Congress should follow public opinion or their own judgment when voting on issues.  I wasn't planning on marking the anniversary, but by coincidence I recently ran across other questions on the same issue, from 1939 and 1940.  They aren't identical to the 1993 question, but seem similar enough to be compared.  The overall distributions:

            Own       Public     

1939     38            59          A
1940     32            64          A
1940     35            39          B
1993     23            70          C

The exact questions:
A.  Should members of Congress vote according to their own best judgment or according to the way the people in their districts feel?
B.  In cases when a Congressman's opinion is different from that of the majority of people in his district, do you think he should usually vote according to his own best judgment, or according to the way a majority of his district feels?
C.  When your representative in Congress votes on an issue, which should be more important:  the way that voters in your district feel about the issue, or the Representative's own principles and judgment about what is best for the country?

The percent choosing the "own judgment" option is substantially lower in the 1993 question than in all three of the 1939-40 questions. It seems to me that the addition of  "what is best for the country" in the 1993 question made the "own judgment" side sound more favorable, so if the differences in question wording mattered they probably understated the change.  In looking at the 1993 question, I had found that education didn't make much difference.  The 1939 and 1940 surveys didn't ask about education, but they had variables for occupation and interviewer's rating of social standing.  People of "higher" position were a bit more likely to say that representatives should follow their own judgement, but it was only a small difference.  I tried a few other demographic variables, which didn't make much difference.  So the major story is simply the difference in the overall distributions.  Of course, 1993 was 30 years ago, so we don't know what's happened since then.   It seems strange that no one has asked about the issue since then, so I'll make another attempt to find questions.

The 1939 survey also asked about a question I've written about before "Do people who are successful get ahead largely because of their luck or largely because of their ability?"  The same question was also asked in 1970 and then in 2016.  My previous post on this question reported the distribution (16% said luck in 1939, 8% in 1970, and 13% in 2016), but didn't look at group differences.  In 1939, there were large differences by economic standing:  

                                       Luck     Ability
Wealthy                           3%          97%
Average +                       7%           93%
Average                         11%          89%
Poor+                             17%          83%
Poor                                23%         77%
On relief                         30%         70%

Unfortunately, the individual data for the 2016 survey is not available in the Roper Center or ICPSR--I will try to track it down, although I think the odds are against me.

[Data from the Roper Center for Public Opinion Research]

It's all over now?, part 2

 I wasn't going to have another post on this topic, but then I read an article in the New York Times that drew parallels between Biden's position today and Obama's and George W. Bush's positions when they were running for re-election.  It suggested that discouraging early polls had led Obama and Bush to "retool" and "recast" their campaigns.  But the figures in my last post shows that both led in the polls at the corresponding point in the campaign, and that their performance in the election was very close to what would have been predicted from the polls a year before.  Of course, this doesn't mean that the campaign efforts didn't matter--holding onto a narrow lead is an accomplishment, but it's a different accomplishment from "turning around a struggling campaign."

The story contrasted GW Bush and Obama to "George H.W. Bush in 1992, [who] failed to heed polls showing voters distressed about the economy and ready for a change after 12 years of Republicans in the White House."  I hadn't included that race in my post because there were no surveys about Bush vs. Clinton in November 1991.  But there were surveys in October and December, and then more starting in January 1992.   In the October 1991 survey, Bush had a big lead:  58% said they would vote for Bush and 22% for Clinton.  Bush's lead in the surveys through early April 1992:

His lead diminished pretty steadily, with maybe an upturn in late March, but he was consistently ahead:  out of 25 surveys, 24 had Bush in the lead, and one had them tied.  So the early polls weren't showing warning signs.  

I had forgotten that Bush was far ahead for so much of the campaign, and not just against Clinton--he led by similar margins in matchups with other potential Democratic candidates.  I remembered that he had been very popular after the end of the Gulf War, but thought that faded pretty quickly and that the presidential race was competitive from the beginning.  It looks like the New York Times writers made the same mistake.  

The growth of partisan polarization means that a swing of this size couldn't happen today.  But the 1992 election may be relevant in  another way.  Going by basic economic statistics, things weren't great, but weren't that bad either, but popular perceptions of the economy were very negative.    As far as I know, there's no generally accepted explanation for the gap.  Either the Bush campaign didn't make enough effort to turn the perceptions around, or their efforts weren't successful.  Either way, the experience may have some lessons for today.

[Data from the Roper Center for Public Opinion Research]

It's all over now?

Donald Trump has generally been leading Joe Biden in recent polls of how you would vote if an election were held today.  How much does this tell us about their prospects for the actual election?  Questions about how you would vote in a hypothetical election go back to the early days of survey research, so we have a pretty long historical record to go on.  I collected questions from the November one year before the election that involved the eventual nominees.  I found them for most elections starting in 1944.  In 1952, 1968, 1976, 1988, and 1992, there were no surveys that asked about the actual matchup.  I also excluded 1972, when all surveys that asked about Nixon and McGovern also included George Wallace as a third party candidate, and 1964, when a survey taken just a few days after the Kennedy assassination showed Johnson with a 79%-15% lead over Goldwater.    That left thirteen elections.  The figure shows the Democratic lead in the election and in polls taken the previous November:

There is clearly a relationship:  if you regress the election lead on the poll lead, the estimate is about 0.5, and the estimate for the intercept is near zero.   So for the purposes of prediction, you should cut the current lead in half.  The standard error is about 6.   So while it's obviously better to be ahead than to be behind, a small lead at this point doesn't tell you much.   

The largest residual is for 1980, when Jimmy Carter had a 10-point lead in a November 1979 poll, but lost badly in the election.  The major reason for this was probably that 1980 was a bad year in terms of both the domestic economy and foreign affairs.  Another factor is that John Anderson entered the race as an independent candidate, and probably took more votes from Carter than from Reagan.  The next biggest residual is in 1984, when Reagan led Mondale by 53%-36% in November 1983 (an average of three surveys), and won by an even bigger margin in 1984.  The economy was improving in 1984, and relations with the Soviet Union improved after Gorbachev came to power.  These two cases are obviously relevant to the current situation, although given increased partisanship the potential for change might be smaller.

  The third largest residual is 2000--George W.  Bush had a 14-point lead (54-40) in November 1999--and there were five surveys, which all were pretty consistent.   There were no dramatic developments in the economy or foreign affairs, so what happened to eliminate Bush's lead?  This is just speculation, but as I recall, Bush had very good press early on.  This was partly because reporters seemed to like him and admire his efficient campaign, but also because they seemed to think that "compassionate conservatism" was an idea whose time had come.  I don't mean that they supported it--most reporters were liberals--but they believed that Bush was in tune with voters.  So my thought is that Bush's early lead reflected favorable media coverage, and that as people got to know him better, they didn't like him as much.  This isn't directly relevant to 2024, since voters already know both Trump and Biden.  But Trump was barred from Twitter in January 2021, and Truth Social doesn't have nearly as large an audience, so to some extent voters will be rediscovering him as they start paying attention to the campaign.  My impression based on perusal of Truth Social is that Trump has become less effective as a communicator:  he goes on at length about he's being unfairly persecuted and how people love him (e. g., a series of posts about his rapturous reception at a UFC event).  On any other topic, even attacking the other Republican candidates, it seems like he's just going through the motions.  So it's possible  that he'll lose ground as voters get more exposure to the new Trump.

[Data from the Roper Center for Public Opinion Research]

Predistribution and redistribution

 The change in the connection between education and party--a shift of educated voters towards the Democrats and less educated voters towards the Republicans--has received a lot of attention.  One popular view is that working class voters have moved away from the Democrats because the party no longer pays attention to their economic interests.  Writing in the New York Times, Pamela Paul says "When it comes to economics, the authors say, Democrats have too often pursued the interests of their own elites and donors. Since the 1990s, the party has pursued policies that worsen the economic plight of Americans who are not well off."  However, although you can find examples that arguably support this analysis, if you look at spending on a range of social programs, the idea that the Democrats have stopped trying to help people with low and moderate incomes doesn't hold up:  see this post.  A new paper by Ilyana Kuziemko, Nicolas Longuet Marx & Suresh Naidu offers a more promising idea:  that education affects relative support for "predistribution"--policies designed to affect jobs and wages--versus "redistribution."  Less educated people tend to favor predistribution, while more educated people favor redistribution, so as educated people have come to have more influence in the Democratic party, policies have shifted towards redistribution.  Thus, although they are still trying to help the working class, they're doing it in a way that has less appeal to the working class.  

As an example of the effect of education on different kinds of opinions, here is the percent of college graduates and others who take the liberal position on some questions from a 2015 CBS News/NY Times survey:

          Not grad grad    Difference

Tax stock transactions 35 39        4
Tax million incomes     69 69      0
sick leave          86 85       -1
caregiver leave 83 79       -4
Union power         45 40      -5
trade restrictions         69 64         -5
Minimum wage $15         40 35         -5
Minimum wage $10     75 69          -6
limit CEO pay         56 45       -11
schedule notice 78 66       -12
Distribution fair 75 63        -12

Positive numbers in the "difference" column mean that college graduates are more liberal than less educated people; negative numbers mean they're more conservative.  Most of the figures are negative, but if you look more closely there's a pattern--more educated people are equally or more liberal when it comes to raising taxes on people with high incomes, but more conservative on things that involve direct regulation.  The biggest difference ("schedule notice") is for a question about whether hourly workers should be given two weeks notice of any change in hours worked or compensated with overtime pay.* None of the differences are especially large, but they are consistent, so they can contribute to a general image of the parties.  

I think that their analysis explains at least part of the shift in party support, and I've made a similar but less systematic account in this paper, although I think that the effect of education on economic opinions has also shifted in a liberal direction--definitely on redistribution, but probably on predistribution as well.  Finally, there's a question of whether the shift led to a change in overall support for the parties?  A New York Times article by Peter Coy on the Kuziemko et al. paper says it does--the title is "How Democrats Lost Voters With a ‘Compensate Losers’ Strategy."  But the paper doesn't actually discuss this issue, and in principle it could go in either direction--the gains among educated voters could be bigger, smaller, or equal to the losses among less educated voters.  I'll discuss this point more in a future post, but at this point I'll just observe that the assumption that this shift is bad for the Democrats is revealing in itself--there is now a general idea that it's better to appeal to the "working class" than to "elites."  So Democrats worry about the shift, while Republicans are proud of it.

[Data from the Roper Center for Public Opinion Research]

* I take support for trade restrictions as the liberal position.  In addition to the policy questions, I also show the results for a question on whether the overall distribution of income is fair.