Sunday, August 30, 2015

Who supports Donald Trump?

Although many people are trying to explain Donald Trump's popularity, there's not much information about who  supports him.  The most extensive discussion I've seen is in a New York Times story on August 22, which said that it was based on "a review of public polling . . . and a new private survey that tracks voting records."  It reported "Mr. Trump leads among women . . . among evangelical Christians . . . among moderates and college-educated voters . . . among the most frequent, likely voters, even though his appeal is greatest among those with little history of voting."  However, in the passage I've quoted, there's only one comparison (more support among those with little history of voting) and maybe one implied one (it says his message is "thought to resonate most with conservatives and less-affluent voters"), and they didn't link to any of the polls.  And given that the vote is so splintered, "leading" could mean less than 20 percent support in a group.  

After a little searching, I found a Fox News poll (conducted Aug 11-13) that gave breakdowns for some groups.   Trump got more support among men than among women (28% to 21%)*.  I don't think anyone will be surprised at that, but the difference is smaller than I would have expected.   Women generally are less attracted by "toughness" or "outsider" appeals, and when you add Trump's various offensive comments, it seems surprising that the gap isn't bigger.  27% of evangelical Christians supported Trump; they didn't give numbers for non-evangelicals, but he got only 25% support overall.  So in this poll evangelicals were somewhat more likely to support Trump than non-evangelicals, but given the size of the sample I'm sure that the difference isn't statistically significant. For age and income, there's no evidence of a difference.  Education, however, makes a big difference:  Trump got 30% support from people without a college degree and only 18% from those with a college degree.  So clearly some candidate did better among people with college degrees:  who was it?   The obvious answer is the establishment candidate, Jeb Bush, but Bush was at 10% among people without a college degree and 8% among people with a degree.  The ones who actually did better among people with college degrees were Ben Carson (16% to 8%), Marco Rubio (7% to 2%), and Scott Walker (8% to 4%).  Apart from Trump, the only other candidate who did noticeably worse among the college-educated was Ted Cruz (8% to 12%).  

The differences by education are intriguing--Rubio and Walker can be regarded as establishment candidates, plus Rubio has tried to cultivate an image as a serious policy thinker, but Carson is just as much of a populist and outsider as Trump.   One possibility is that the desire for "diversity" is stronger among college-educated voters, but then the numbers for Cruz are a puzzle.  (There's also Bobby Jindal, but he gets virtually no support among any group, so we don't have to worry about him).  

The margin of error in group differences from a single poll is large, but the association between education and support for Carson bears watching.  Every story about Bernie Sanders seems to point out that his supporters tend to be well educated.  According to the same poll, Sanders gets 37% among likely Democratic primary voters with college degrees, and 26% among those without.   So in a relative sense, Carson's support is more concentrated among college graduates than Sanders's is.

*The figures refer to likely Republican primary voters.

Monday, August 24, 2015

Resentment, complacency, and question wording

A couple of years ago, I gave figures for a question on whether blacks "are receiving too many special advantages, receiving fair treatment, or are being discriminated against?"  The proportions choosing the three answers stayed about the same between 1990 and 2000, but between 2000 and 2008 there was a decline in "discriminated against" and a rise in "fair treatment."  My conclusion then was "'racial resentment'  seems to be a minority view.  'Racial complacency' is more common and may be on the rise."  Unfortunately, that question hasn't been asked since then, but as I looked for more data on the general topic, it occurred to me that the question about "blacks" up through 2000 and "African-Americans" in 2008, and that people might react differently to those terms.  "African-Americans" may make people think of "hyphenated" white ethnic groups like Irish-Americans,which could account for the lower numbers saying "discriminated against."

It seemed like it would be pretty easy to check this--surely some survey would have asked a question about discrimination and randomly switched between the terms.  But despite a good deal of searching, I couldn't find anything.  Then I turned to questions that were identical or almost identical except for the terms.  There have been a number of the basic form "Would you say there is a great deal of discrimination, some discrimination, only a little discrimination, or none at all against...blacks/African Americans?"  The results, with the responses scored as 4, 3, 2, 1, so higher numbers indicate more perceived discrimination.

The black dots represent questions that asked about "blacks," red "African Americans," and green "blacks or African Americans."  Questions that included both terms got the highest perceived discrimination; questions that asked about "blacks" may have gotten higher perceived discrimination than those that asked about "African Americans," but it's not clear.  Overall, it seems like asking about "blacks" rather than "African Americans" may make some difference, but not much.

It seems like there's a pattern over time:  a rise in perceived discbetween the 1970s and about 2000, and a decline since about 2000.  But to complicate things, there were also some differences in the response categories, as shown in this figure:

The numbers are the same, but now the different colors and letters a,b,c,d indicate different responses that were offered:  a is "a great deal, a fair amount of discrimination, some discrimination but not much, or practically none," b is " a great deal, some, only a little, or none at all," c is "a lot, some, only a little, or none at all," and d is "a lot, some, a little, or none."  There's not enough overlap between the times they were asked to come to any definite conclusions, but form c was asked often enough to be pretty sure that there was a decline in perceived discrimination between 2000 and the present.  The 1978 and 1985 description of the lowest category as "practically none" definitely seems weaker than "none" or "none at all," so some of the rise in perceived discrimination between the 1970s  may be spurious (people who would accept "practically none" but not "none at all" as the best description).  However, given the big change between 1978 and 1985 it seems safe to say that there was a change from the 1970s.

Considering these data, and my previous posts on the general topic (see here for a list), I think there are two distinct trends.  First, "racial resentment"--that is, the belief that blacks are actually advantaged--is declining and continues to decline.  Second, that the belief that blacks are discriminated against rose from the 1970s until around 2000, but has declined since then.  My original suggestion that "racial complacency" is rising seems pretty accurate.

Wednesday, August 19, 2015


As promised in my last post, here are the rationales for the probabilities I gave for the Republican nomination.  On Jeb Bush, I'm just going along with prevailing opinion.  He has a big lead in fundraising and endorsements, and most people who are supposed to know seem think that this is important.  He's been slipping in the polls in the last week, and if I were to give numbers now he would be lower.  Beyond that point, my thinking is that the main issue is the domestic economy.  On that, almost all candidates are trying to show how conservative they are--ie, how opposed to taxes and spending.  There are some differences in the exact promises:  income tax rates of 17%, 14%, 12%, 10%, tax returns that fit on one page, tax returns that fit on a postcard..... but it's hard to keep them straight and I doubt that many voters take them seriously.

The result is that there are only a few candidates who stand out.  One is Donald Trump, who has obvious liabilities.  Another is John Kasich, who suggested that government spending can sometimes accomplish good things, and that when it does he'd support it.  He's the only one (except maybe Trump) who is trying to appeal to voters who have conservative inclinations but aren't ideologically committed.  There are a lot of voters like that, so I could put Kasich higher, but he starts with a disadvantage in money and name recognition.  Also, many activists and donors regard this position as heresy, so I think if he gets close there will be a strong "anyone but Kasich" movement.  Finally, there's Marco Rubio, who is Hispanic.  "Diversity"  has become a mainstream value, and I think that given the choice between candidates who are reliably conservative and about equally qualified,  a lot of Republican voters would pick one who's black or Hispanic.  Rubio doesn't have the negatives of Ben Carson (no political experience) and Ted Cruz (not likable enough), so he's the one who can benefit from that.

That leaves the 12.5% for a late entrant.  Candidates who came in late have a very poor track record in recent primaries (Rick Perry, Fred Thompson, Wesley Clark), but since a lot of Republicans think that 2016 is a good opportunity and no one in a large field has caught fire (except  Trump), it seems like there's an unusually good opportunity.

Thursday, August 13, 2015

The Man of the Hour

Paul Krugman questions what he sees as the conventional conservative and liberal explanations of Donald Trumps support and offers one of his own.  I agree that the two he questions--that people mistakenly believe Trump is a real conservative and that he appeals to the white working class--are implausible.  Krugman proposes that Trumpism is a continuation of the Tea Party--it appeals to "angry, fairly affluent white racists."  One problem with this explanation is that the Tea Party movement wasn't particularly about race (see this post for some evidence).  Another is that Trump hasn't said much about race.  But what about immigration--isn't that a "coded" appeal to racism?  In the general public, opposition to immigration and anti-black racism aren't closely connected--immigration is an issue in its own right.  Beyond immigration, Trump is an economic nationalist:  he sees a competition among nations in which America is losing (in fact, his most inflammatory comment on immigration was directed at the Mexican government).  This is a position that has little support among political elites, but is popular among the public.  Trump's other major theme is his contempt for politics and politicians, which puts him in line with public sentiment.  Some of the other candidates are trying to make a similar appeal, but one of the things that average voters dislike most about politics is conflict.  When Ted Cruz threatens to shut down the government, he thinks he's taking a stand against politics as usual; most voters think he's engaging in politics as usual.

This combination of appeals is a lot like Ross Perot's in 1992, so my guess is that Trump's supporters are similar to Perot's, which I wrote about in this post.  I would also guess that he has a strong appeal to self-employed people, who would be more likely to think that someone with business experience could straighten things out in Washington.  I don't remember much about Perot's campaign, but my general impression is that he wasn't a particularly compelling candidate, but his support held up pretty well over the course of the campaign.

While I'm at it, here are my probabilities for various candidates winning the Republican nomination:

1.  Jeb Bush                              40%
2.  Marco Rubio                        20%
3.  John Kasich                         15%
4.  Other declared candidates   12.5%
5.  Not yet declared                  12.5%

I'll explain those in a later post.

Friday, August 7, 2015

Fox News was right

I wasn't going to post again this soon, but I couldn't resist the opportunity to use this title.  Justin Wolfers objects to Fox's statement that "Given the over 2,400 interviews contained within the five polls, from a purely statistical perspective it is at least 90% likely that the tenth place Kasich is ahead the eleventh place Perry." The shares in the polls were 3.2% for Kasich and 1.8% for Perry.  It seems to be easier to work with numbers than with small percentages, and given  2,400 total respondents, that's 77 and 43 people.  The observed difference is 34, and the standard deviation is about 11.

Wolfer's basic point is to calculate a probability that one candidate is ahead of another, you need to start with some assumptions about the way that things might be.  He illustrates this by what he acknowledges is a contrived example--"Either Republican voters like Mr. Kasich so much that he is beating Mr. Perry by 20 points, or alternatively they find them both pretty likable, but Perry’s new glasses are sufficient to give him a 0.1 percentage point lead. A small poll in which Mr. Kasich edges out Mr. Perry by a mere 1.4 percentage points is more consistent with the latter scenario than the former," so if we start by assuming the two scenarios are equally likely, we will end by concluding that Perry is probably (in fact, almost certainly) ahead.

However, Fox didn't start by saying that they would use polling results to decide between Perry and Kasich--they started by saying that they would use polling results to pick the ten participants.  So we're talking about comparing two unspecified candidates who are 10th and 11th in the polls., which may be what they mean by "from a purely statistical perspective."   Since the candidates are now just "one guy" and "another guy", it seems the prior distribution representing  possible values of the difference in support for them has to be symmetrical.

 Once you restrict your attention to symmetrical prior distributions, it's hard not to conclude that someone who leads in the poll by 34 has at least a 90% probability of being ahead.  For example, suppose the prior distribution is normal with mean zero and standard deviation of 100. This means that you think there could be a pretty big gap between them.  By definition, the #10 and #11 candidates have less than 10% of the vote each, so a difference of 100 out of 2400 (about 4%) is big.  Then the probability that the candidate who leads by 34 is really ahead is .999.

What about normal with mean zero and standard deviation of 25?  This would amount to saying that you expect the #10 and #11 candidates to be pretty close--it's unlikely that they're as much as 2% (48 out of 2400) apart.  The probability falls to .998.  What if the distribution is normal with mean zero and standard deviation of 10?  This means that you have a strong expectation that the #10 and #11 candidates are very close.  Now the probability falls to 0.98. What if the distribution is normal with mean zero and a standard deviation of 5?  Now it's down to 0.88.  But this distribution amounts to saying that you're almost sure that the #10 and #11 candidates are almost tied. It seems hard to justify that assumption when you just start by knowing that you're talking about the #10 and #11 candidates in a field of about 16.

You could fault Fox for not including a qualifier like "under any reasonable assumptions."  But the conclusion that there was at least a 90% chance that Kasich was really ahead was justified.  In fact, I'd commend them for putting it that way rather than giving an exact number.

Wednesday, August 5, 2015

Message: I Care

Liberals often charge that conservatives lack compassion--that they don't care about the poor, or don't think about what it's like to be poor.  The usual conservative reply is that the difference doesn't involve compassion, but views about what works--that measures that seem like they would help the poor may be ineffective or even counterproductive.  That's pretty much what Greg Mankiw says in his review of a new book by Arthur Brooks.  But Brooks goes farther, or at least he did in his 2006 book, Who Really Cares?:  he says that conservatives are actually a lot more compassionate.  For example, he says that data from the General Social Survey show that people who disagreed strongly with the statement, "the government has a responsibility to reduce income inequality" gave, on the average,  nine times as much to non-religious charities as those who strongly agreed (pp. 55-6).

As the example suggests, Brooks is concerned with support for what he calls "forced income redistribution  to achieve greater income equality," (p. 181) not things like same-sex marriage.  So from now on I'll refer to opinions on redistribution rather than "liberalism."  

In fact, the GSS never had the item he described.  However, it did ask for a response to "It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes."  The numbers on that question pretty much match those that he reports.  People who strongly agreed gave an average of $64, people who strongly disagreed gave an average of $542.

The GSS has several other questions about attitudes to income redistribution.  One goes:  "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.   Here is a card with a scale from 1 to 7. Think of a score of 1 as meaning that the government ought to reduce the income differences between rich and poor, and a score of 7 meaning that the government should not concern itself with reducing income differences. What score between 1 and 7 comes closest to the way you feel? " People who chose 1 gave an average of $87 and people who chose 7 gave an average of $208.  But people who chose 4 were way ahead of both, with an average of $506.

The problem with giving the averages is that the distribution of giving is extremely skewed--most people report giving nothing, while a few give very large amounts.  In a sample of about 1,000 people, the top ten people gave almost half of the total.  So apparently impressive differences in the means can reflect a handful of people.  So I'll give t-ratios from an appropriate model for skewed dependent variables (controlling for income).  With the first question (GOVEQINC) the t-ratio is 2.6.  With the second (EQWLTH), it is 0.6.  The usual standard for "statistical significance" is a t-ratio of about 2.  Some people would set it higher and some lower, but everyone would agree that 0.6 amounts to no evidence either way--the variable might make a difference in one direction, the other direction, or no difference at all.  So your conclusions about whether attitudes to "forced redistribution" are related to charitable contributions depend on which measure you use.

It seems to me that EQWLTH is better in principle because it simply asks what the government should do.  Some people might interpret EQINCOME as implicitly choosing between individuals and the government--in effect, asking if the government has the primary responsibility.  That would make the association almost automatic.

The GSS has still another question on the same general issue:  "On the whole, do you think it should or should not be the government's responsibility to . . . reduce income differences between the rich and poor?"  People who say "no" give more, and the t-ratio is 2.2.  That question (EQUALIZE) doesn't seem open to the same objection as GOVEQINC.  So now we're back to conclusions differing depending on which measure you use.

EQUALIZE is part of a series asking about the government's responsibility to do various things.  The questions and the associated t-ratios (positive numbers mean anti-redistribution answers go with more charitable giving):

Provide a job for everyone who wants one.                                             2.8
Keep prices under control.                                                             2.3
Reduce income differences between the rich and poor.                             2.2
Provide health care for the sick.                                                             1.8
Provide a decent standard of living for the unemployed                            1.1
Give financial assistance to college students from low-income families   0.8
Provide decent housing for those who can't afford it.                     0.5
Provide a decent standard of living for the old                                          0.1
Provide industry with the help it needs to grow.                                      -0.1

All of these, except the second and the last, involve egalitarian redistribution.  All, except again the second and last, are things you would expect liberals to be more favorable towards (those two are ambiguous--you could make arguments in either direction).  But only three (counting EQUALIZE) have a statistically significant association with reported giving, and one of those is the ambiguous "keep prices under control."

Putting it together, the evidence is weaker than Brooks suggests.  By luck or design, he happened to report just the one item that gave strongest support to his case.  Still, there seems to be something there--some measures of support for redistribution or a more expansive government are associated with lower charitable contributions, and none were associated with more.