Left turn?, part 3

I thought I just had a few odds and ends to clear up regarding my last post, but then I ran across a post from January.  That was about a claim by Thomas Edsall that "as the gulf between rich and poor widens, voters become increasingly mean spirited and hostile to the welfare state, progressive taxation and regulations designed to protect consumers, workers and the environment."  I selected some items from the GSS that dealt with those issues.  I was concerned with trends rather than year-to-year changes, but on looking at it again I saw that opinions on all of them moved in a liberal direction between 2014 and 2016.  That is, the liberal move involved a range of opinions, not just opinions about race.  That led to the question of whether there was really anything unusual about opinions on race, as I had been assuming.   

I had a total of twelve items--four on race discussed in my post from last week, and eight others from January.  They were measured on different scales, so you have to standardize them in order to compare the size of the changes.  I did that by using the standard deviation of differences among annual means.   The three largest differences all involved questions about race--the fourth question on race (affirmative action) was in the middle of the pack.  So it seems that opinions about race did change more than other opinions.  Still, there was also a general liberal movement. 

I made a simple index by adding together all the means.  Here are the values form 1994-2018, with lower meaning more liberal:

That looks like a liberal trend, with 2010, 2012, and 2014 as exceptions and the big change from 2014-6 as a return to the trend.  That is a little misleading, since research has shown that there is a tendency for opinions on many issues to move right when a Democrat is in office and left when a Republican is in office.  Still, I think it points at part of the explanation--there was a reaction against Obama that was running out of steam by 2016. 

Five of the items, all involving spending on various domestic issues, go back to the 1970s.  The index for them:

The red line includes an adjustment for the estimated effect of party control.  There's no clear trend since the late 1980s, just ups and downs.  But overall, it seems like opinion on these issues is as liberal as it's ever been. 

Left turn?, part 2

In my last post, I observed that four different measures of racial attitudes in the General Social Survey all had big moves in a liberal direction between 2014 and 2016, and all moved same direction between 2016 and 2018.  What about other surveys?  It turns out I've written on some relevant ones before:

1.  In September, I discussed this question:  "How serious a problem do you think racial discrimination against blacks is in this country--a very serious problem, a somewhat serious problem, not too serious, or not at all serious?"  The last three times it was asked:

                       Very           Somewhat       Not too    Not at all
July 2010         18%             44%                 28%         10%
June 2015         37%            37%                 16%           9%
Sept 2016         42%            37%                 13%           6%

My title for that post was "Not sure I believe this":  It seemed like an awfully large change, and I wondered if it reflected the context of the questions in the surveys--that is, the questions in 2015 and 2016 might have appeared after questions that got people thinking about discrimination.  The 2015 questionnaire is available, but I haven't checked, or if I did I've forgotten.

2.  On December 28, I wrote about this "IN YOUR OPINION, HOW WELL DO YOU THINK NEGROES ("blacks" starting in 1978) ARE TREATED IN THIS COMMUNITY--THE SAME AS WHITES ARE, NOT VERY WELL, OR BADLY?"  Here are results from the last two times it was asked:
                           Same       Not well     Badly

March 2001         64%          24%          3%

Feb 2018              53%         38%          8%

A substantial move towards seeing worse treatment.  I noted that the 2018 sample was different from the earlier ones, though.  

3.  The next day, I wrote about this question:  "In general, do you think the criminal justice system in the United States is biased in favor of blacks, or is it biased against blacks, or does it generally give blacks fair treatment?"  Results from the most recent surveys:

                 

                      Favor  Against   Fair

Aug  2013        5%    35%       43%

July 2015        5%    49%       40%

Sept 2017        8%    47%       36%

A big change between 2013 and 2015--note that the percent of "don't knows" was quite a bit lower in 2015 than it had been.

When Andrew sent his inquiry, I looked and found a number of questions of the form "How much discrimination do you think there is against African Americans in our society today--a lot, some, only a little, or none at all?" (there are some slight differences in question wording, but the answer categories are always the same).  The recent examples:

                   Lot      Some      Little   None

Aug 2013    19%      49%       22%     7%

June 2014    17%      49%       27%     5%

Nov 2014     21%     47%       24%      8%

May 2015     41%      41%       14%      3%

Aug 2015     36%      40%        16%      6%

*Nov 2016   41%     35%        17%      6%

*Feb 2016    39%      43%       14%      5%

*Mar 2019   46%      34%        14%     5%

The ones marked with an asterisk asked about "blacks," which could make a difference.

You could reasonably raise doubts about any of these results, but the overall pattern is clear:  there was a substantial change in opinions between 2014 and 2015.  In fact, you could be more precise and say the middle of 2015.  

The shooting of Michael Brown occurred in August 2014, and lead to protests which got a lot of media coverage.  There was no video in that case, and people had different interpretations.  On Nov 26, 2014, a surveillance video of the shooting of Tamir Rice was released, which also got a lot of media coverage and was widely shared on social media.  Since then, there have been many other videos of police shootings.  More recently there have been videos of black people doing ordinary things and being confronted or having the police called on them. I think that these have made a difference--most news stories, even ones that get a lot of media coverage, don't change many opinions because they are complicated and can be interpreted in many different ways.  But a video of an everyday situation leaves less room for interpretation (although there is always some) and there was not just one, but a series of them.  

In conclusion, I think that there is strong evidence that opinions have changed:  people see more discrimination against blacks than they did just a few years ago.  That raises a few other questions, which I will discuss in my next post.  

[Data from the Roper Center for Public Opinion Research] 

Left turn?, part 1

A few days ago, Andrew Gelman sent me a question from Elin Litman on a puzzling finding from the 2018 GSS.  He included my reaction in a post on his blog--this post is an elaboration of my remarks there.   The items which prompted the original question were about whether the nation was spending too much, too little, or about the right amount on "Improving the conditions of Blacks" and "assistance to blacks."  The means, with lower numbers corresponding to the "too little" end:

So opinions on both apparently made a big move in a "liberal" direction between 2014 and 2018.  The change from 2016 to 2018 might be partly the result of the change in administration:  responses to questions like this tend to move left with a Republican president and right with a Democrat.  But of course that doesn't explain the 2014-6 changes, which are very large.  Could it just be some kind of mistake?

I looked at the means for two other variables related to race:  "Some people think that (blacks/negroes/African-Americans) have been discriminated against for so long that the government has a special obligation to help improve their living standards; they are at point 1. Others believe that the government should not be giving special treatment to (blacks/negroes/African-Americans); they are at point 5. Where would you place yourself on this scale?" and "Some people say that because of past discrimination, blacks should be given preference in hiring and promotion. Others say that such preference in hiring and promotion of blacks is wrong because it discriminates against whites. What about your opinion - are you for or against preferential hiring and promotion of blacks?"  The means--again, low numbers mean more liberal views:

Again, there's a large move in a liberal direction on both opinions between 2014 and 2016, and a smaller move in a liberal direction between 2016 and 2018.  On all questions, opinions in 2018 are more liberal than in any previous year. 

Why would this have happened?  One possibility is that it had to do with Donald Trump--the interviews for the 2016 GSS started in April 2016, after it had become pretty clear that he would be the Republican nominee, and went on until November.  Maybe Democrats reacted against Trump by strongly affirming things that he seemed to threaten.  So I broke the means down by party identification.  In all cases, the changes were in the same direction among supporters of all parties, although larger among Democrats.  That is, Independents and Republicans also became more liberal on these issues between 2014 and 2018.  An example:

Until 2014, Democrats and Republicans had been moving in different directions, but between 2014 and 2018 they both moved towards the "special obligation" end.  That doesn't fit the idea that the change was a Democratic reaction against Trump. 

So what caused the change?  I'll address that in my next post.

  PS:  Steve Morgan made some similar points on Twitter.

From Nixon vs. McGovern to Trump vs. Clinton

The 2018 General Social Survey is out.  Since the GSS is given every two years in the early part of the year, 2018 is the first one since the election of Donald Trump.  The GSS contains questions on whether you voted in the most recent presidential election and, if you did, who you voted for.  It also has a number of questions which have been included since the 1970s, making it possible to look at changes in the the association between opinions and voting choices over a long period of time.  I considered four opinions:
1.  Some people think that the government in Washington ought
to reduce the income differences between the rich and the poor,
perhaps by raising the taxes of wealthy families or by giving
income assistance to the poor. Others think that the government
should not concern itself with reducing this income difference
between the rich and the poor. [Place yourself on 7 point scale]
2.  In general, some people think that it is
the responsibility of the government in Washington to see to it
that people have help in paying for doctors and hospital bills;
they are at point 1. Others think that these matters are not the
responsibility of the federal government and that people should
take care of these things themselves; they are at point 5.  [Place yourself]
3.  Some people think that (blacks/negroes/African-Americans) have been discriminated
against for so long that the government has a special obligation
to help improve their living standards; they are at point 1.
Others believe that the government should not be giving special
treatment to (blacks/negroes/African-Americans); they are at
point 5. [Place yourself]
4.  Please tell me whether or not you think it should be
possible for a pregnant woman to obtain a legal abortion if:
The woman wants it for any reason.

I computed the correlation of opinions on each with vote for the Democratic or Republican candidate (third party voters were excluded) in elections from 1972 to 2016. 

The correlations for equalization, helping blacks, and helping with medical bills track each other very closely--that wasn't why I picked them, it just turned out that way.  In the 1970s, opinions on abortion had almost no correlation with vote, but then it started to catch up and it's now comparable to the others.  There's an upward trend for all of them--that is, supporters of the Democratic and Republican candidates are more distinct.  People have known about the increasing association between opinion and party for long time, but some new things emerge from a closer look.  First, there is a difference between the trends for abortion and the other issues.  It's close to a linear trend for abortion, but not for the others.

I took an average of the correlations for the other three items and plotted it against time.  There's clearly a more rapid increase in the 21st century.  It's hard to date it exactly, but there was a large increase between the 2000 and 2004 elections, and another one between the 2004 and 2008 elections.  In 2016, the correlations for general equality and helping with medical costs declined slightly, and the correlation with helping blacks rose.  Those changes fit with common accounts of Trump's appeal, but they were not very large.

Second, there has not been a shift from emphasis on "economic" issues (the first two) to "social" issues (abortion and arguably race), which people often say has happened.  "Economic" issues are less important in a relative sense because some new social issues have emerged (just abortion here, but I would guess that some others have followed a similar course), but more important in an absolute sense--that is, vote can be predicted more accurately from economic opinions than it could back in the 1970s.   

Third, there has not been much variation between individual elections, suggesting that specific characteristics of the candidates or campaigns have not have much impact. 

PS:  The correlations for 1996 and especially 1992 are probably exaggerated because of votes for Ross Perot.  I expect that Perot voters were mostly in the middle ideologically, making the separation between Democratic and Republican voters bigger.  I don't think that John Anderson had much impact in 1980--the opinions of his voters were probably close to those of Carter's. 

More on the electoral college

I had some posts on the electoral college a few weeks ago.  After Andrew Gelman discussed them on his blog, I thought more about the issue.  In my post, I dismissed the arguments in favor of the Electoral College, and most of them are pretty bad, but there is one that deserves to be taken seriously.  It was given by Akhil Reed Amar in the New York Times a few days ago:   "Any reforms might backfire, with unforeseen and adverse consequences. The Electoral College is the devil we know."  Twenty years ago, this was a pretty strong argument:  the last time that there had been a conflict between the electoral college and the popular vote was over 100 years ago (1888).  So the possibility of a discrepancy seemed like just something to speculate about, not a serious possibility. But it's happened in two of the six elections since then, making the "devil we know" considerably less appealing.  You could say "so what?  A candidate winning the Electoral College while finishing behind in the popular vote is no different from a football team winning while finishing behind in yards gained."  [I'm pretty sure I read something like this shortly after the election, but I forget who wrote it].  The answer to that can be found in a survey question asked a few days before the 2000 election:

"As you may know, presidents are chosen not by direct popular vote, but by the electoral college system in which each state receives electoral votes based on its population. Over the past one hundred years, the winning presidential candidate has won both the popular and electoral vote. Let us suppose in this election that one candidate were to win the popular vote and one candidate were to win the electoral vote. Who would you favor serving as the next president of the United States--the one who wins the popular vote, or the one who wins the electoral vote?"

57% said the candidate who won the popular vote, 33% the candidate who wins the electoral vote, with 10% not sure.  Note that this question didn't ask about changing the system in the future--it asked about who should be president after the 2000 election.  It also pointed out that we currently did it by electoral vote, not by popular vote.  Nevertheless, a clear majority said that the president should be the one who won the popular vote. 

The implication is that every time there is a discrepancy, there is a risk of conflict--if leading supporters of the losing candidate don't accept the result, there is a lot of popular sentiment that they can appeal to.  In 2000 and 2016, leading Democrats accepted the outcome--there were only a few complaints, and no effort to mobilize against it. But can we count on that happening again if there's a discrepancy in 2020?  I wouldn't bet on it.  People could say that the Electoral College results are tainted by voter fraud or voter suppression in a few key states.  Of course, that wouldn't prevent a president from taking power, but it would make the losers feel justified in engaging in more extreme tactics, which would make the winners feel justified in engaging more extreme tactics themselves. 

[Data from the roper Center for Public Opinion Research]

Another quarter

In November 2018, I had a post about the difference between elite and popular ideologies.  Politicians can be classified as simply liberal, moderate, or conservative, but in the general public opinions on economic and social issues are almost uncorrelated.  As a result, there are substantial numbers of people who are liberal on economics and conservative on social issues or conservative on economics and liberal on social issues.  So why don't any ambitious politicians try to appeal to those combinations?   In February, Paul Krugman had a column on the same issue, which even had a similar title.  I have a couple of disagreements with his analysis:  (1) he claims that there are very few people with the economic conservative/socially liberal combination and (2) he says social conservatism can be reduced to racism.  But he also linked to a paper by Brian Feinstein and Eric Schickler that offered some interesting historical evidence.  Their idea was that even though Franklin Roosevelt stayed away from issues of race, the New Deal ideology (government should help the disadvantaged) naturally led to support for civil rights and the ideology of opposition to the New Deal (defending states' rights and individual property rights against an overbearing federal government) naturally led to modern racial conservatism.  They showed that state party programs gradually diverged on civil rights issues, and in the North the Democrats were considerably more liberal than the Republicans by the 1950s.  This is in contrast to a widespread view that party differences on civil rights were minimal until the 1960s.  Their analysis led me to wonder if there were party differences in the general public as well.  There weren't many questions on civil rights before the 1960s, but I found one from May 1952, which asked "Which of the following statements best describes what you'd like to see done about legislation to prohibit employers from turning people down solely because the their race or religion?"  The options were "there should be a national law to protect minorities from discrimination in hiring," "it should be left to each state to pass such a law if it wants one," and "there shouldn't be any law like that; the problem should be worked out in some other way."  This question deals with an important issue of principle and doesn't mention any party or political figure.

As expected, blacks were overwhelmingly in favor of a national law (73%, 6% for leaving it to the states, 10% opposed, and 10% don't know). There were no clear party differences among Southern whites:  16% favored a national law, 33% said there should be no law, and 39% said it should be up to the states.  But among Northern whites, there was a difference between supporters of the different parties:  39% of Democrats favored a national law and 28% said their should be no law; 26% of Republicans favored a national law and 40% said there should be no law.   

Since the question also mentioned religion, you might wonder if the party difference had to do with religion rather than race.  Jews and Catholics were more likely to favor a national law than Protestants were (64% of Jews, 36% of Catholics, and 24% of Protestants), but the party difference is still there if you restrict it to Protestants. 

In summary, Northern Democrats were significantly more liberal than Northern Republicans on this issue, even at this early stage.  The example seems to support Krugman's suggestion that "advocating economic inclusion seems to spill over into advocacy of racial and social inclusion, too"--that is, that the dominance of the overall liberal/conservative dimension is not an accident.

[Data from the Roper Center for Public Opinion Research]