Monday, November 23, 2015


1.  On Sunday, the New York Times had a piece by Alec MacGillis called "Who Turned My Blue State Red?" on the question of why poor states vote Republican.  I started reading it with low expectations, but perked up when I read "the people who most rely on the safety-net programs secured by Democrats are, by and large, not voting against their own interests by electing Republicans. Rather, they are not voting, period."

Those of us who are interested in politics tend to forget that a lot of people, and probably a majority of working class people, don't vote even in Presidential elections.  In a post last year, I suggested that working-class turnout might have declined in states like West Virginia because of the decline of manufacturing and mining, so the people who continued to vote would be the local middle class.  Although they might have relatively low incomes by national standards, they would compare themselves to the local people who didn't have steady jobs.  From that perspective, they could be drawn to the Republicans, and especially sympathetic to the idea that poor people got that way because they made bad choices:  they could think of examples from personal experience.

MacGillis  seemed to be saying pretty much the same thing.  He basically took the traditional journalist's approach of going out and talking to people--he offered some survey evidence, but it seemed sort of tangential.  But it makes me think that someone should look at changes in the relationship between class and turnout by state.

2.  In my last post, I talked about the relationship between height, weight, and earnings for men and women. The obvious way to do it would have been to look at the relationship between body mass index (weight over height squared) and earnings.  I thought about doing it, but didn't want to to impose a constraint on the form of the relationship.   But then I realized that I had wound up imposing a constraint on the relationship anyway, so why not try it?  BMI turns out to be almost uncorrelated with height, so you can just look at the bivariate relationship between BMI and earnings.  I rounded BMI to the nearest whole number, took mean earnings, and smoothed the figures to reduce the influence of sampling error, producing:

There are some interesting and unexpected things here:
1.  There is a penalty for having a high BMI, but it's not much different for men and women.
2.  Contrary to what I thought based on the figures in the previous graph, women do not benefit from being thin. Predicted income is about the same in the whole "normal" range (actually a little higher in the upper reaches, although I don't think the difference is large enough to be worth taking seriously).
3.  Men suffer a substantial loss from being "underweight."  That does match the results from my previous post.

Putting it together, it looks like men are the ones who lose the most if they don't have the "ideal" body.  I don't really believe this--the numbers of men with low values of BMI are small, so I suspect that most or all of the apparent difference is just chance variation.  But it's worth considering.

Saturday, November 21, 2015

Too rich or too thin?

The Behavioral Risk Factor Surveillance System, despite its ominous-sounding name, is just a survey focusing on health.  It asks people for their height and weight.  As a result of an example I used for a class, I started thinking about the relationship between weight and earnings.  The BRFSS doesn't have a measure of individual earnings, but it has a question on family income, and for people who are single (never married) and employed, it seems safe to assume that income and earnings are usually very close.  Since the BRFSS obtains very big samples, that still leaves over 20,000 cases.  I controlled for gender, height and weight.  For both height and weight, I included the original and its square and allowed both to differ by gender.  The following figures show the predicted values for a range of weights at selected heights.

 For women, being thinner goes with having a higher income.  The relationship seems to flatten out at the lowest weights, but basically if your goal is to be rich (or at least reasonably well paid), you can't be too thin.  For men, the relationship is different--the highest expected earnings are at the middle weights, and the penalty for being below the "ideal" weight is at least as large as the penalty for being above it.  For example, at 6 feet, the highest predicted earnings are at a bit over 200 pounds, and the predicted values for a man who weighs 150 and a man who weighs 295 are about the same.

Although I was primarily interested in weight, I noticed that the influence of height also differs by gender.  Women benefit from being taller through the whole range. but the predicted values for tall men (6'4") and fairly tall men (6') are almost identical.

I wouldn't put too much faith in the exact shape of the curves, since I didn't experiment much with alternative specifications, but the possibility that "underweight" men earn less is intriguing.  I put "underweight" in quotes because men who are in what is officially classified as the "normal or healthy" range earn less than men who are in the "overweight" range and sometimes even the lower reaches of "obese."

Thursday, November 12, 2015

Coulda been a contender

A few days ago, the New York Times had a story called "Boxing is a brutal, fading sport.  Could football be next?"  It pointed out that boxing used to be very popular, but said "there came a time when the fight game’s hold on the American spirit began to loosen, when it stood widely condemned as plain brutal."  As the title suggests, it went on to say that maybe the same thing was starting to happen to football.

I looked for questions about whether boxing should be banned, and found a number going back to the 1950s.  Several gave a simple choice:

                 Banned         Not              DK
6/1955         39%           45%             16%
4/1962         37%           47%             16%
4/1963         41%           43%             16%
6/1965         43%           45%             12%
3/1981         23%           70%               7%
7/1997         19%           74%               7%

One in 1984 gave people a choice of banning only amateur boxing, only professional boxing, or banning both.  39% favored both, 8% favored banning one of them, and 50% said both should be allowed.  In 1999, a Fox survey gave a choice between regulated better (42%), left alone (24%), or abolished altogether (24%).

Public opinion didn't turn against boxing in a straightforward sense--in fact, support for abolishing boxing was generally lower in the 1980s and 1990s.  Or looking at it another way, boxing's hold on the American spirit wasn't secure even in its glory days:  about 40% wanted it abolished.

There are no questions about whether football should be abolished, but several in the past few years have asked if the respondent would allow a son to play football:  about 70-75% say they would.  People do seem to be aware of the dangers of football:  in 2011, a survey asked "in which one of the following professional sports do you think athletes are most likely to sustain a permanent injury that will affect them after they retire?...NFL (National Football League) football, martial arts, boxing, hockey."  More chose football (49%) than boxing (36%).

The lesson is that changes in the fortunes of the sports didn't directly reflect changes in public opinion.  I think the basic reason that football flourished while boxing declined is that football developed an organization that could supply a consistent "product," and boxing never did.

[Source:  iPOLL,  Roper Center for Public Opinion Research.  The Roper Center has moved from UConn, where it had been since 1977, to Cornell University.  I'm sorry to see it go (I was interim director from 2004-6), but I'm glad that it found what promises to be a good home.]

Friday, November 6, 2015

False consciousness in the donor class?

Paul Krugman recently had a column asking why Republicans claim that they will deliver faster economic growth, even though growth has actually been considerably faster under Democratic presidents since at least the 1940s.  His answer is that it's partly "epistemic closure: modern conservatives generally live in a bubble into which inconvenient facts can’t penetrate" and partly that they "need to promise economic miracles as a way to sell policies that overwhelmingly favor the donor class."  It seems to me that he's trying too hard to make a mystery out of this.  I think Republicans promise faster growth because they believe they can deliver it, and they believe they can deliver it because they have a "common sense" argument that they can:  there are tradeoffs between different goals, and if you insist on things like equality or environmental protection, you'll pay a price in economic growth.  And in my experience, most people (probably all people most of the time) base their opinions on what seems like common sense to them rather than on careful analysis of the available data.

However, his column led me to wonder if the "donor class" actually does better under the Republicans, or if a rising tide lifts all boats, so that all income groups do better under the Democrats.  I obtained data from the World Top Incomes Database on the average income (adjusted for price levels) of the bottom 90%, the 90-5 percent, the 95-99 per cent, and the top one percent.  Then I did a regression with percent change over the previous year as the dependent variable and party of the president as the independent variable.  The classification of years when party control changes is difficult:  the president is inaugurated in January, so the new party is in control for the great majority of the year, but the American economy is big and there's a lot of inertia.  So I counted those years as half one party and half the other.

The average annual change in real income under the two parties:

                           Dem                Rep
0-90%                  1.8%              0.0%
91-95%                2.0%              1.1%
95-99%                2.0%              1.2%
99-100%              2.0%              2.0%

The numbers in the World Top Incomes Database are pre-tax incomes, so it's likely that the top 1% really has done better under Republican presidents in terms of after-tax income.

Saturday, October 31, 2015

See the happy extremist

Since my series of posts on changes the relationship between political views and happiness trailed off into "something happened, but it's hard to say what it was," I wanted to do something that came to a definite conclusion.  Arthur Brooks said "people at the extremes are happier than political moderates."  He speculated that this was because "extremists have the whole world figured out, and sorted into good guys and bad guys. They have the security of knowing what’s wrong, and whom to fight."  As you can see in this post, no such pattern is present in the GSS:  average happiness increases as you go from left to right, except recently it's fallen off among people who say they are "extremely conservative."  What if you look at happiness by party identification?
Now there is some pattern for strong identifiers to be happier (higher scores on the vertical axis).  And since Democrats are more likely to be single, lower income, and less religious, controlling for those characteristics, as Brooks did, would make the pattern stronger.  But what about the people who identify with another party?  They're not a large group, even in the cumulative GSS, and are usually treated as missing values, but they are less happy than everyone except independents.  And it seems likely that they are more "extreme" than any of the other groups--the two biggest enduring minor parties over the period have been the Libertarians and Greens.  14.7% of the "others" put themselves in one of the extreme ideological categories, higher than any other party group (11.3% of strong Republicans say they are extremely conservative and 1.0% say they are extremely liberal, for a total of 12.3%).  And of course, there's a plausible explanation for why they would be less happy than Democrats or Republicans:  they're dissatisfied with the way things are.

So it doesn't seem to be the case that extremists are happier than moderates.  Rather, as I said a few posts ago, the pattern seems to be that people who are interested in politics are happier than people who aren't.  Further evidence can be seen in the relationship with political views, people who answered "don't know" to the question on political ideology are the least happy group.

Brooks didn't give a link to his source, or any information about it, but if I were a betting man, I'd bet he did one of two things:  (1) mixed up the patterns for party identification and ideology or (2) used a survey that counted "don't knows" on a liberal/conservative rating  as moderates, which is sometimes done.

Friday, October 30, 2015

What is to be explained?

In my last post, I said that the relationship between political views and happiness changed over time and that I'd have to think before offering an explanation.  My general idea was that political polarization had been growing, so that there was an increasing tendency for people to be less happy when the "wrong" party was in power.  The only problem was that there were signs of a growth in polarization in the 1990s, like the 1994 congressional election and the impeachment of Bill Clinton, so why wasn't anything visible then?  In the hope of getting a hint, I estimated the relationship between self-rated ideology and happiness by year, and got this:

Higher values mean a tendency for conservatives to be relatively less satisfied.  The zero value is arbitrary--it just represents the pattern in 2014.  The 95% confidence intervals for the annual estimates are about plus or minus .02 (about plus or minus .03 in the first half of the period), so you could not reject the hypothesis that all the values within the Bush or Obama presidencies were the same.  However, if you just look at the figure, it looks more like just random variation with a few exceptions--2004, 2008 and 2010.

So maybe the problem isn't to explain the patterns under recent presidents, but the patterns in a few particular years?  The GSS is conducted in March, and March 2010 was when the Affordable Care Act was passed.  So maybe that explains why conservatives (and especially "extremely conservative" people) were relatively unhappy then, but (maybe) not in 2012 and 2014.  With G. W. Bush, the 2002 pattern wasn't unusual, but 2004-8 were.  Those years had something in common--the United States was engaged in an extended war.  It seems plausible that liberals and moderates would be more worried about war and its implications.  The GSS started in 1972, but didn't include the question on political views until 1974, so being at war was a new event in the history of the available data (the first Gulf War took place in January/Feb 1991, but American troops were already leaving by March).

I'm not sure that what needs to be explained is a few unusual years rather than presidencies--the point is just that it's another way to look at it, and the data doesn't let us decide between the possibilities.  This is a case where we start with what seems like a lot of data--50,000 cases--which turns out to be a lot less than we need for some purposes.

Friday, October 23, 2015

Lost Happiness

This post is inspired by Andrew Gelman's comment on a previous post about partisanship and satisfaction with various things.  Andrew noted that the relationship between partisanship and general happiness had changed, a point first noticed by Jay Livingston, who was commenting on a piece by Arthur Brooks  entitled "Conservatives are Happier, and Extremists are Happiest of All."  Brooks said "the happiest Americans are those who say they are either 'extremely conservative' . . .  or 'extremely liberal' (35 percent). Everyone else is less happy, with the nadir at dead-center 'moderate' . . . ."  He didn't give a source, but Livingston noted that in the 2010 sample of the General Social Survey, people who called themselves "extremely conservative" were the least happy group.

I looked at the complete GSS data starting in 1972 to see how the relationship changed.  If you look at the combined sample, there's absolutely no sign that "extremists" are happier:  happiness increases pretty steadily as you go from left to right.  Brooks said he had controlled for a number of other variables, so maybe that accounts for it.  I doubt it, but I'm not going to pursue that issue.

Next I allowed the relationship to change from year to year, expecting that there would be some gradual trend--a tendency for conservatives to become relatively less happy.  There was no sign of that, but I thought I saw some variation that corresponded to party of the president--conservatives were relatively unhappy when Democrats were in office.  But on including a set of dummy variables for presidents, there was little variation until GW Bush and then Obama.  So I ended up with a three-period classification:  up to 2000, Bush (2002-8--the GSS is done only in even numbered years), and Obama (2010-4).  The relationship in the three periods is shown in this figure:

Higher numbers on the vertical axis mean happier.  In the 20th century, conservatives were happier than liberals, with those who called themselves extremely conservative being happiest of all.  Under Bush, conservatives and extreme conservatives stayed about the same, but everyone else became less happy.  Under Obama, conservatives and extreme conservatives became less happy, but everyone else stayed about the same.  The net result was that happiness went down across the board between the 20th century and Obama.  Liberals and moderates lost under Bush and didn't gain under Obama, and conservatives didn't gain under Bush and lost under Obama.  There seems to have been an especially large drop-off among extreme conservatives, who became a lot less happy under Obama.

I was surprised at these results--I thought that even among people with strong political views happiness would be influenced overwhelmingly by personal circumstances, so that any relationship with the party of the president would be so small as to be undetectable.    And if it is influenced by the party of the president, why just since GW Bush?  People had strong feelings about at least some of the 20th-century presidents, notably Nixon, Reagan, and Clinton.  So I'm going to have to think about this before proposing an explanation.