Thursday, November 16, 2017

Revise and resubmit

After reading my last post, my esteemed friend Robert Biggert urged me to take a look at rates of immigration, and pointed to this column by Thomas Edsall, which argued that there was a turn towards Trump in places where there were few immigrants but the number was increasing rapidly.  Then this morning I saw another by Edsall which argued that there was a turn towards Trump in places where there were few "minorities" but the number was increasing rapidly.  The idea is that whites feel most threatened when they are first exposed to a significant number of "others."  After reading the first one, I got figures on the number of foreign-born people by state in 2000, 2010, and 2015.  I found the 2000-10 changes didn't make any difference to 2016 vote, but the 2010-15 changes did.  So I decided to limit my attention to changes in the black and Hispanic population since 2010.  The results from regressions with Republican gain over 2012 as the dependent variable (standard errors in parentheses):

Constant                 .026            -.062
                              (.013)           (.030)

Utah                         -.145         -.135
                                 (.021)        (.019)

Home                      -.040           -.046
                                (.017)          (.015)

Black (2010)           -.109            -.088
                                 (.028)          (.026)

Hispanic (2010)       -.131          -.096
                                 (.042)          (.038)

Foreign-born (2010)  -.084         -.041
                                   (.076)        (.068)

Disability                   .437            .649
                                  (.248)          (.226)

Grow10-15                                    .023

Adj. R-square            74.5%         75.4%

The last variable is the sum of the growth in black, Hispanic, and foreign-born population.  For each one, "growth" is the ratio of 2015 share to 2010 share.  The reason I used the sum is that when growth in each population was included separately, the group was statistically significant, but none of the individual estimates were.  Almost by definition, the largest values for growth occurred in places that had few "minorities" in 2010:  the states that ranked highest on Grow10-15 were North Dakota, South Dakota, Maine, Vermont, and Minnesota.  But the estimates for percent black and Hispanic in 2010 are still negative and statistically significant, so the conclusions from my previous two posts stand up.  (The estimate for foreign-born is negative although not statistically significant).  Going back to the "defense of whiteness" idea, these analysis suggest in the contemporary United States it's a temporary thing--it applies when "others" are first appearing but weakens or reverses as they remain.

Saturday, November 11, 2017

and why were they there?

As I said in my last post, Donald Trump did better than Mitt Romney had in states with few "minorities" (blacks or Hispanics) and worse in states with large numbers of "minorities."  I have read a lot of articles about the 2016 election, but don't recall anyone noticing this.  In a post from December 2016  I noted that some of Trump's biggest gains took place in states with few minorities, but I didn't follow up on it.  Here is a figure showing gain or loss over 2016 by percent black or Hispanic.  The correlation is about 0.5 for all states and 0.7 if you exclude Utah.  I don't think that it's likely to go away if you control for other factors, so how should it be explained? 

One popular analysis of the 2016 election was that it was about the defense of "whiteness," which is a mixture of prejudice and the rational (although selfish) defense of group interests.  People who advocate this don't pay much attention to regional differences--they generally focus on making a case that there's a historical pattern of white backlash after every move towards racial equality.   But it seems that this account implies regional differences too.  Racial interests will make more difference to white voters if the percentage of racial minorities is larger:  that is, minorities will be more of a threat where they are more numerous.  To the extent that Trump appealed to "whiteness," he should have gained more where the percentage of minorities was higher.  That is the opposite of what happened.*  So how can this pattern be explained?  The assumption of the usual account is that an appeal to "whiteness" will gain votes among whites.  But white opinions on race have changed a lot over time.  The decline in straightforward prejudice is well known, but whites have also seem to have become less likely to think that they are the ones who are discriminated against.  That suggests that an increasing number of whites will have a negative reaction to a "defense of whiteness" appeal--they will regard it as unfair.  Even whites in the middle--those who don't think there's much discrimination against either blacks or whites--may have a negative reaction because they think it will increase racial conflict. 

So my view is that to the extent that Trump appealed to "whiteness," that hurt him in much of the country.  The thing that helped him was an appeal to nationalism, as discussed in this post, among others.

*Another variant of this analysis is that Trump was no different from Romney, Bush, etc.:  they all appealed to "whiteness" and he was just cruder.  But this would imply no relationship. 

Sunday, November 5, 2017

Where were they?

By and large, the states in which Trump did well in last year's election were the states in which Mitt Romney did well in 2012.  The figure shows the Republican share of the Democratic plus Republican vote in the 2012 and 2016 election (the District of Columbia is not shown because it is off the scale).

Still, there were some shifts, and they made a difference to the outcome.  What sort of states shifted towards the Republicans?  One possibility is that it was ones that were struggling economically.  I saw a lot of stories about Midwestern cities where factories had closed down, or were threatened with closure, and people turned to Trump.  Another possibility is that it was places that were suffering from high levels opioid addiction--people could think that "get tough" policies were needed or just that the Obama administration hadn't been successful in dealing with the issue, either of which would help Trump.  High rates of disability could produce sentiment against "unworthy" beneficiaries of government programs (I proposed something along those lines in this post).  Although people sometimes talk like all of these go together, they are not all that highly correlated at the state level--the correlation between deaths from drug overdose and percent of the working-age population receiving disability benefits is about 0.4, and the correlations of the unemployment rate with overdoses and disability are both about 0.2.  Some regressions of the Republican share of the two-party vote in 2016 on these three variables plus various controls (t-ratios in parentheses). 

      u       od      d    Controls
1.  .006   -.0005   .013      2012
   (-1.2)  (-0.6)   (3.4)

2. -.0098   .0006   .006      plus Utah
    (-2.5) (1.0)    (1.9)

3. -.012    .001    .006      plua home state
   (-3.2)   (1.6)   (2.1)

4.  -.001   .0004   .004      plus % black, % Hispanic
     (0.3)  (0.8)   (1.4)

5.                  .005      same as #4

Utah was an outlier, and there was an obvious reason for that--independent conservative candidate Evan McMullin, who was a native of Utah and a Mormon, got over 20% of the votes there.  "Home state" was defined as +1 for Massachusetts and -0.5 for Illinois and Hawaii.  Both 2016 candidates were residents of New York, so only 2012 needed to be considered.  When dealing with American politics, it's always advisable to consider race and ethnicity, so I put in controls for black and Hispanic.  

Considering all of the regressions, it's clear that Trump did not do relatively well where the unemployment rate was high.  The rate of overdose deaths and disability always had the same sign, but disability was consistently stronger, so in the last regression I just considered disability.  The estimate is statistically significant, although not overwhelmingly so.  The disability rates range from about 3% to about 9%, so going from low to high is estimated to increase Trump's share by .03 (relative to Romney in 2012).   That's enough to be of interest.  

However, the effects of percent black and Hispanic, which I just threw in as controls, were much more substantial.  In states where there were few blacks and Hispanics, Trump generally did better than Romney had; in states with lots of black and Hispanics, he generally did worse.  The estimates were both negative and of similar size--if you combine them into a single "minority" proportion, the estimate is about -.13, and going from the states with the lowest percent minority to the highest (Texas and New Mexico) would reduce Trump's share by about .07.  

 Black voters are overwhelmingly Democratic everywhere; Hispanic voters are pretty heavily Democratic and Trump actually did a little better among them than Romney did, according to exit polls.  Black turnout was probably down from 2012, but Hispanic turnout was probably up.  So it seems likely that most of the shift resulted from changes among white voters. 

I will offer some thoughts on the meaning of this pattern in my next post.

Sunday, October 29, 2017


I looked for surveys that were relevant to the issues discussed in my last post, and found one sponsored by CNN and conducted by the Opinion Research Corporation in 2010.  It included a series of questions on whether people favored or opposed:

a. Building a 700-mile long fence on the border with Mexico
b. Creating a program that would allow illegal immigrants already living in the United States for a number of years to stay here and apply to legally remain in this country permanently if they had a job and paid back taxes
c. Imposing fines of tens of thousands of dollars on employers who hire illegal immigrants
d. Sending employers who hire illegal immigrants to jail 
e. Putting more Border Patrol and federal law enforcement agents on the U.S. border with Mexico

The distribution of opinions:

                            Favor        Oppose     Mixed/DK
Fence                    45%           55%          1%
Stay                       83%          17%           0%
Fines                     62%           38%           1%
Jail                        41%           59%           0%
Border                   78%           21%           0%

The survey also contained a series of questions asking about whether certain statements apply to immigrants moving here in the past ten years:

a. Are basically good, honest people (82% say yes)
b. Add to the crime problem (58% yes)
c. Take jobs away from Americans  (52% yes)
d. Are hard-working (87% yes) 
e. Are a burden on American taxpayers (65% yes)

Questions a, c, d, and e from the first group all involve enforcement of immigration laws.  If you add them together to get an index of support for more vigorous enforcement and regress them on the views of immigrants, all have a statistically significant relationship except the first.  That is, there is no evidence that whether or not people think most immigrants are "basically good, honest people has any connection to support for stronger enforcement (the t-ratio is about 0.4) of policies against illegal immigration.  If you regress the "path to citizenship" question (b) from the first series on the same views of immigrants, "basically good, honest people" has a significant relationship (t=3.2), and appears to be the second or third strongest predictor, depending on how you measure.  That is, views of the character of immigrants matter for opinions about whether they should be allowed to stay, but not (or not nearly as much) for opinions about stronger enforcement of the immigration laws.  This pattern illustrates a point I've made before, which is that there are (at least) two immigration issues, and that people who favor stronger enforcement of immigration laws are not necessarily "anti-immigrant" in a general sense.

I didn't find any survey questions that directly involved rights to immigrate or exclude immigrats, but an opinion piece appeared in the New York Times today (it's been online for a few days) about unaccompanied children fleeing violence in Central America and trying to enter the United States.  There are currently about 350 reader comments, and I looked at the ten most liked ones.   None of the them questioned the individual accounts or said that they represented only a small fraction of illegal immigrants. Nine offered some variant of "the United States can't (or doesn't have an obligation) to take in everyone"--the other one said that Mexico was the closest neighbor, so they had the primary responsibility to help.  Although the most liked comments have no value as a measure of the general distribution of opinions, I think they say something about how people explain the opinions they have. 

[Data from the Roper Center for Public Opinion Research]

Tuesday, October 24, 2017

The policy that dare not speak its name

A summary of public opinion on immigration:
1.  Strong majorities think that people who were brought here as children should be allowed to become citizens.  Majorities say that people who came as adults who have been working and don't have a criminal record should at least be allowed to stay, and possibly to become citizens.
2.  Opinion is now pretty evenly divided on whether the level of legal immigration should be increased, reduced, or kept the same.  
3.  Large majorities say that immigration laws should be more strongly enforced about people who are now trying to come in.  

That is, give a break to people who are currently here, but try to stop further illegal immigration.  The policy of the Obama administration was pretty much in line with prevailing public opinion--he supported a proposal for a "path to citizenship," established DACA, left the laws on legal immigration alone, and deported a lot of people.  However, he didn't say much about the deportations.  The attention came from critics, mostly on the left, but including Donald Trump in one of the debates:  "President Obama has moved millions of people out. Nobody knows about it. Nobody talks about it. But under Obama, millions of people have been moved out of this country. They've been deported."  Hillary Clinton said even less about them--as I recall, she just ignored Trump's statement.  Why not talk about a policy that would be popular and refute Trump's claims about how we had "open borders"?  There were some immediate reasons, which are discussed in this article.  But I think there was also a deeper reason.  

In popular moral thinking, nations are important.  We have obligations to other members of our nation that we don't have to people in other nations.  That raises the question of who is a member of our nation.  Regardless of their views on what the right level of immigration should be, the great majority of people would agree that "we" (the current citizens) have a right to decide on their number and the conditions of joining our nation.  If you asked people to give reasons for these beliefs, I think most would say that they are just common sense.

What you could call "sophisticated" thought is not satisfied with appeals to common sense--it demands justification in terms of principles.  Sophisticated thought is not limited to intellectuals in a narrow sense--it also includes most journalists and politicians, and a significant number of educated people more generally.  The major principle that is accepted today is human rights:  people are endowed with certain inalienable rights, among which are life, liberty, and the pursuit of happiness.  Nations and governments are just a means to secure these individual rights.  Living where you want is a basic part of liberty and the pursuit of happiness, so on a straightforward interpretation of this principle, it's hard to justify any restrictions on immigration.  At the same time, most sophisticated people also share the intuitive sense that the nation is more than just an instrument for securing individual rights.  By and large, they deal with this conflict by avoiding it:  Democrats denounce anti-immigrant policies but don't say much about what they think immigration policy should be, and Republicans call for the "rule of law" but don't try to give a justification for those laws.  

I can't offer any direct evidence for any of this, but I think it is a way to make sense of a number of things about public opinion and politics today.   

Tuesday, October 17, 2017

Raise the bar?

A paper published in Nature Human Behavior proposes changing "the default P-value threshold for statistical significance for claims of new discoveries from 0.05 to 0.005"--in terms of t-ratios, from about 2 to 2.8.  The paper seems to have been written with experimental social psychology in mind, but its 72 listed authors include economists, political scientists, and sociologists.  They are a distinguished group--the sociologists are from the University of Pennsylvania, Univ of North Carolina, Michigan, Harvard, Princeton, and Stanford. 

The core argument of the case is about the chance of "false positives."  The great majority of the hypotheses proposed in the social sciences are of the form "x is associated with y" (controlling for other factors relevant to y).  If the observed data would be unlikely under the "null hypothesis" that "x is not associated with y" (controlling for other factors), you count it as support for the hypothesis that "x is associated."

Suppose that for every ten proposed hypotheses that are true, there are 100 that are false.  Using a .05 level means that we can expect a statistically significant association for five of the false ones.  Suppose a statistically significant association is found for 80% of the true hypotheses, which is the target that people usually aim for in designing experiments; then 5 out of 13, or almost 40% of the statistically significant associations will represent false hypotheses.  Their idea is that researchers should change the standard of statistical significance to 0.5% and continue to aim for 80% power (which would mean bigger experiments).  That would mean there would continue to be 8 statistically significant associations that represent real ones but only 0.5 (6% of the total) that are spurious.

The ratio of true to false proposed hypotheses is crucial here.  If it's 1:1, then with 80% power and a 5% significance level, we have only 6% spurious associations.  The authors offer some evidence that the ratio is about 1:10 for psychology experiments, and say that a "similar number has been suggested in cancer clinical trials, and the number is likely to be much lower in biomedical research."   They also address the possible objection that the "threshold for statistical significance should be different for different research committees."  They say that they agree, and that genetics and high-energy physics have gone for a higher standard--a t-ratio of about 5, but don't even address the possibility that a lower standard might be appropriate.  That is, they seem to take a 10:1 ratio of false to true hypotheses as the minimum, and recommend the .005 standard as a baseline suitable to all fields.  They return to this point in the concluding remarks, where they say that since the .05 level was established "a much larger pool of scientists are now asking a much larger number of questions, possibly with much lower prior odds of success."  This isn't convincing to me.  In the papers I read (published or for review), most of the hypotheses about relations between variables seem pretty plausible.  Even if I don't find the reasoning that leads to the prediction convincing, and often I don't, it's not hard to think of an alternative argument (or several arguments) that leads to the same prediction.  The idea that more scientists asking more questions means lower prior odds of success isn't  compelling either.  In some fields, theory has developed, and that should let you make reasonable predictions on more questions.  In others, there's at least more evidence,  meaning more examples to draw on in making predictions.  So I doubt there is a tendency for the prior odds in favor of proposed hypotheses to decline. 

If they were just making a suggestion about how to interpret the .05 significance level, I would not object, and in fact would generally agree (see my book Hypothesis Testing and Model Selection in the Social Sciences).  But realistically, a "default" of .005 would mean it would become difficult to publish work in which the key parameter estimates were not statistically significant at that level, just as it's now difficult to publish work in which the key parameter estimates aren't significant at the .05 level.*  That would be a loss, not a gain, especially with non-experimental data, where a bigger sample is usually not an option.

*They say results that didn't reach .005 "can still be important and merit publication in leading journals if they address important research questions with rigorous methods,"  but I'm confident that the great majority of reviewers and editors would say that about the .05 level today.  Importance and rigor are matters of judgment, so there's usually disagreement among reviewers; the "default" level of significance is objective, so it takes on outsize importance.

Wednesday, October 11, 2017

Then and now

Ta-Nehisi Coates has a piece called "Civil-Rights Protests Have Never Been Popular," in which he notes that in 1966, 63% of people polled had a negative opinion of Martin Luther King.  The question was asked five times by the Gallup Poll, asking people to rate him on a scale of +5 to -5.  A summary of the results, plus some historical events:

                        +        -        -5
May 1963      41%   37%  (20%)
                                                       March on Washington 8/1963
Aug  1964     44%   38%  (22%)
                                                       Selma march, 3/1965
May  1965     45%  46%   (27%)
                                                        Chicago open housing movement, mid- 1966
Aug   1966    33%  63%   (39%)
Aug   2011    95%    4%    (1%)

During King's life, there was always a significant number giving him the lowest possible rating, which I show in parentheses.

 There are some complications, which I will discuss in the future, but Coates is right in his general point--King was not particularly popular when he was alive, and among whites negative views probably always outnumbered positive views.  Strongly negative views were definitely more common than strongly positive views.

[Data from the Roper Center for Public Opinion Research]