||[Aug. 6th, 2012|11:27 pm]
I recently finished The Spirit Level, subtitled "Why More Equal Societies Almost Almost Do Better", although "Five Million Different Scatter Plot Graphs Plus Associated Commentary" would also have worked. It was a pretty thorough manifesto for the best kind of leftism: the type that foregoes ideology and a priori arguments in exchange for a truckload of statistics showing that their proposed social remedies really work.
(the following post is based on a few days of reading and a few hours' thought. I don't endorse anything below with much confidence, except when stated otherwise)
The book's argument was predicated on a bunch of graphs like these one:
The x-axis is a standard measure of income inequality; the y-axis is the level of some social problem. Imprisonment, psychiatric disease, teenage pregnancy, drug abuse, obesity, poor school performance, whatever. The graphs all look a bit like a more or less steep line, which means there's a strong relationship between the two measures: ie countries (or US states) with more inequality have more social problems. Although they don't do a great job of reporting p-values, if you dig a little further they all exist and they're all significant, some of them very much so.
Their theory is pretty simple: inequality causes social problems. We usually think of poverty as causing social problems, and they agree that in Ethiopia or somewhere that's true, but they say that in First World countries material poverty is no longer a particularly good predictor of social problems; their graphs of poverty vs. social problems just look like a bunch of dots scattered all over. Instead, the main predictor of social problems is inequality.
Why would inequality cause social problems? They refer to a famous study about London civil servants, in which civil servants at the bottom of the ladder, who have to deal with constant monitoring and Pointy-Haired-Bosses and always being ordered about have heart attack levels several times those of their higher-status colleagues who feel relatively "empowered"; the determining factor here was self-reported job stress rather than income. This remains true in monkeys; primatologists who artificially manipulate dominance within a group of monkeys find that newly dominant monkeys improve rapidly on various health indicators (but see here). And because giving monkeys cocaine is totally ethical, they find low-status monkeys become easily addicted and high-status monkeys are less susceptible to addiction.
So, they conclude, feelings of stress and misery associated with being low status cause people to become less physically healthy, less mentally healthy, and more likely to abuse drugs - among other things. And because there are decent evo psych reasons why low status should promote risky behavior, it also leads to violence and short-term thinking. And too much competitiveness and division between people lowers levels of social trust, which is known to be pretty much the most important factor in preserving a healthy society. Hence the observed association between inequality and all social problems - which, if you combined all the social problems into one big Index Of Social Problems, looks very impressive, like so:
(they then go on to have several chapters that seem to be only tangentially related cheerleading for leftism in general, including one talking about how much we need to fight global warming. In my opinion these detract from the book, so I'm going to gloss over them).
And their final and most fascinating finding is that inequality doesn't just hurt the poor. Middle class and rich people in unequal countries suffer more social problems than their middle-class and rich counterparts in more equal countries. In some cases, the rich in unequal countries have worse health, mental health, etc than even the poor in an equal country. So the strong message of Spirit Level is: it's not the economy, stupid - at least not the part measured in how much money you have. It's the equality level, plain and simple.
So of course then I go online to read the commentary and reviews and watch people demolish all of it.
Some guy named Christopher Snowdon has written a book called The Spirit Level Delusion in which he attacks the heck out of it, but he was too snide and irritating for me to read him for more than a few minutes (he's the kind of person who says things like 'Oh, leftist intellectuals must love this' as if that were damning criticism) so instead I read a slightly less grating report by (surprise!) a conservative think tank titled Beware False Prophets.
I won't bore you with the discussion of exactly which countries get included in the analysis, except to say that the author, Peter Saunders, considers it suspicious that for example Portugal (an unequal country with horrible social problems) was included but many other larger and wealthier countries were not. But other than that Saunders makes two good points.
First, the Scandinavian countries are always the best at everything. The USA is always the worst at everything, usually followed closely by the other English-speaking countries. It so happens that the Scandinavian countries are unusually equal, and the Anglophone countries (especially the US) are unusually unequal.
This sounds like it supports Spirit Level's theories, but it's not so simple. There are four Scandinavian countries and five Anglophone countries in the set of 23 countries analyzed (and most analyses didn't include all 23). Most of the strongest associations in Spirit Level were driven by the four Scandinavian countries driving up the average for the equal countries, and the US and other Anglophone countries driving down the average for the unequal countries.
If Spirit Level does ten analyses and all come out positive for a correlation, it looks like they have 230 or so data points - 23 countries times ten analyses. But Saunders is saying that all of these 230 data points are driven by only two facts: Scandinavian countries are always the best, and the Anglophone countries, especially America, are always worst. And then there are ten other countries scattered haphazardly somewhere around the middle. It's silly to try to make a theory on only two facts; there could be a bunch of reasons the Scandinavian countries are really awesome, and a bunch of reasons America sucks.
Saunders gives one such possibility: Scandinavian countries are unusually ethnically homogenous with a strong sense of ethnic identity, and Anglophone countries are unusually ethnically diverse with a weak sense of ethnic identity. More diverse countries are known to have lower levels of public trust and lower cooperation, which in turn causes various social problems. This alternate hypothesis at a stroke explains the three strongest data points in Spirit Level: the constant success of Scandinavia, the constant success of Japan (which also does very well on most measures), and the constant dismal failure of the USA and other Anglophones (although as far as I can tell it doesn't explain Portugal, the other country that fails at everything).
On a similar note, Saunders notes that Spirit Level's US state data can all be explained better by percent African-American population in a state than it can by equality in the state (although in reality the two measures are very similar because the main source of inequality is racial inequality).
The authors of Spirit Level respond to Saunders' first point by saying that sure, if you artificially take out half of the countries they included in their analysis just because they give too good results, then you can make their results disappear (sometimes! not always!), but that's not very good science. I sympathize both with them and Saunders: given the small size of the sample (23 countries), any group of countries that's outside the norm for any reason other than inequality is going to screw up the results, but an analysis with more countries is difficult to impossible because it leaves the realm of rich First World countries to which the inequality-not-poverty theory properly applies.
I'm equally uncertain about the second point, the African-Americans in US states. On the one hand, the Spirit Level authors begin by dismissing it with a disappointing "That's racist and you are racist for considering it", but eventually mention a few studies (for example here) that weaken the association in a few specific cases but leave the possibility of a more general effect unclear. They also say that even if it were true it would prove their theory because the only possible difference between different races is economic inequality (I guess this is the problem with reading leftist books).
Saunders finishes up by saying that the beautiful graph above with the perfect relationship between inequality and the Index of Social Problems is entirely an artifact of what social problems they chose to include, and then creates a similar graph with some slightly different social problems - divorce rates, burglaries and suicides instead of teen pregnancies, imprisonment and generic mental health - with a slightly different cutoff point for "rich countries" that shows exactly the opposite effect - less equal countries doing better in every way. I'm a sucker for that kind of argument, although I think Saunders was a bit more manipulative than the original authors were.
Overall, I find Spirit Level's statistics slightly more convincing than its critics'. Although Saunders' re-analyses can make a lot of the effects disappear, even he admits there are quite a few that he can't get rid of, and I find him to be quite manipulative in a way that damages his credibility (for example, he removes certain outliers when doing so would discredit Spirit Level's results, but simultaneously preserves even worse outliers on the other side if removing them would re-credit the results). Although he has a good point about the Scandinavians and the African-Americans and so on, in most cases these only weaken Spirit Level's arguments without destroying them entirely.
But there's a big gap between agreeing that a correlation exists, and agreeing that it's caused by inequality. The main take-home lesson I got from this book is that every social problem and illness and disappointing aspect of society is correlated with every other social problem and illness and disappointing aspect of society, usually very highly. They're all best in Scandinavia and worst in the United States. Inequality is one particular social ill that gets included in this correlation, and Spirit Level can come up with decent explanations for why it would be the causative one, but who knows if it's a just-so story? I can think of many other possibilities:
THINGS ABOUT RACE AND CULTURE: As mentioned above. Looking at the Index of Social Problems graph, it goes very clearly from Japan to the Protestant European countries to the Catholic European countries to the Anglopshere and the US. I'm not sure exactly why more Protestantism (and let's face it; Japan is "Protestant" in everything but religion) leads to higher levels of social trust, greater cohesion, lower inequality, or why this totally breaks down in the US, but it's hard to escape. It also seems to be going from less diverse to more diverse, albeit not as obviously or monotonically. And that's not even mentioning the really racist stuff about population genetics, although that probably couldn't explain the inequality correlation anyway.
POLITICS: I like this one because it's the only thing that could possibly make conservatives angrier than the book's actual conclusion. But if you look at the data, the successful countries and US states are not only the ones that are most equal, but the ones that are most big-government liberal; the failed countries and states are the ones that are most small-government conservative. Massachussetts aka Taxachussetts tops the league of US states in pretty much everything; the Deep South is always at the bottom (I know the one graph I found online and put in this entry doesn't show this so well, but it's true in many of the others). Famously socialist Scandinavia is on top of the international league, and the famously free-market Anglophone countries are always at the bottom. It's easy to see how leftist policies could reduce inequality, and if leftism happens to work as well as leftists think it would also reduce social problems. Even if this turns out not to be the driving factor behind the data, it still seems to be enough of a real effect that it demands explanation (maybe in the "better" countries better education -> more leftism?)
NO, MAYBE IT REALLY IS MONEY: I understand that social problems don't correlate with base per capita income, but what if it's more complicated? If there are a bunch of rich people around, they raise the price of goods, and they vote in regulations that further raise the price of goods: that's why you can have an awesome meal at a nice restaurant for a dollar or two in a Third World country (as long as you don't mind e. coli). What if the inequality effects we're seeing here are just the effects of poor people living in rich areas with high cost of living, and so having less purchasing power to buy material goods than poor people in a poorer country? I haven't seen either Spirit Level or its critics address this possibility, which makes me think I'm missing something, but it sounds plausible.
Nevertheless, this data is probably making it into the second version of the Non-Libertarian FAQ that I'm working on. I mean, libertarians are always talking about how it doesn't matter what people's relative wealth is, what matters is making the pie bigger. But I think, whether or not inequality causes all our social ills, there is much to be said for the idea that the size of the pie is no longer important, and the distribution of wealth really does matter. I was, as usual, insufficiently cynical. It's not just that taxing the rich is a convenient way get money to use to help the poor. Taxing the rich and then putting all the money into a big hole might well still make everyone better off, including the rich themselves. Being able to give it to the poor is just a side bonus.
(I have about 50% credence in this proposition right now)
Equality Trust - pretty much a website version of the book, maintained by the authors. Lots of graphs.
Beware False Prophets - the least obnoxious and spiteful, and most interesting, rebuttal to Spirit Level I could find.
Spirit Level authors' response to various criticisms/rebuttals, including the one above.
2012-08-07 06:00 pm (UTC)
"The y-axis is a standard measure of income inequality; the x-axis is the level of some social problem."
I think you have 'x' and 'y' swapped.
"The authors of Spirit Level respond to Saunders' first point by saying that sure, if you artificially take out half of the countries they included in their analysis just because they give too good results, then you can make their results disappear (sometimes! not always!), but that's not very good science. I sympathize both with them and Saunders: given the small size of the sample (23 countries), any group of countries that's outside the norm for any reason other than inequality is going to screw up the results, but an analysis with more countries is difficult to impossible because it leaves the realm of rich First World countries to which the inequality-not-poverty theory properly applies."
AGH THIS IS WHY WE HAVE BAYESIANISM. Is P(What We See|Inequality Hypothesis) higher or lower than P(What We See|Ethnicity Hypothesis)?
For the one where there are few confounding factors- the percentage AA in American states- ethnicity solidly trounces inequality.
Basically, this is where you can conclude that Spirit Level's authors aren't sufficiently scholarly, in that they aren't constantly hunting for better explanations and willing to try any hypothesis. They've got a dog in the fight, and so everything they say needs to be taken with a grain of salt. That said, they could still be right- but it's much more likely that they'll be right about small issues (like high levels of stress being bad for health) than about large issues (like public policy prescriptions).
"I'm not sure exactly why more Protestantism (and let's face it; Japan is "Protestant" in everything but religion) leads to higher levels of social trust, greater cohesion, lower inequality, or why this totally breaks down in the US, but it's hard to escape."
Uh, Protestants are northern European, and Catholics are southern European, and latitude and IQ are related. It breaks down in the US because of ethnic heterogeneity.
"It's easy to see how leftist policies could reduce inequality, and if leftism happens to work as well as leftists think it would also reduce social problems. Even if this turns out not to be the driving factor behind the data, it still seems to be enough of a real effect that it demands explanation (maybe in the "better" countries better education -> more leftism?)"
Perhaps places that consume less social services can afford more of them. Scandinavia can give generous support to single mothers because there are so few of them- try to do with with sub-Saharan Africans, and you quickly bankrupt the state.
One way to think about this in general is that liberal policies *require* responsibility and conservative policies *create* responsibility. When marginal tax rates are 70%, and it's a better economic proposition to be a single mother than go to college, Swedes will generally pay the taxes and go to college. The budget will get balanced and the people who aren't responsible will get cared for. But try to implement that in Mississippi, and you'll find the responsible people leave and the irresponsible people arrive, and quickly the state will be broke and the welfare rolls overflowing.
But when you have a state like Texas, with few social services and low taxes, the individual incentives are all aligned towards being responsible, and so people behave more responsibly (than they would elsewhere; correcting for race is still important).
Uh, Protestants are northern European, and Catholics are southern European, and latitude and IQ are related. It breaks down in the US because of ethnic heterogeneity.
I'd be very hesitant to make claims about IQ and culture/ethnicity without very solid evidence, but there are well-documented cultural factors that seem likely to correlate with social problems: Catholic countries usually have more work-to-live cultures and present-time-orientation cultures, while Protestant countries tend to have more live-to-work cultures and future-time-orientation cultures.
However, cutting the other direction, Anglosphere cultures are within the normal range of Northern European Protestant countries in these respects.
Also possibly cutting the other direction is that there's a very strong statistical linkage in the US between 1) getting married and staying married, and 2) positive personal outcomes in terms of income, happiness, and other positive outcomes, and I've read some fairly persuasive arguments that the link is at least partially causal from marriage->success (economies-of-scale and division-of-labor benefits in housework and household expenses, increased financial stability from pooling two potential incomes, increased social stability from shared family/friend support networks, direct emotional support from your spouse, etc). I'm not familiar with data off the top of my head, but I'd expect Catholic countries to be more likely than Protestant countries to have cultures that encourage getting married and staying married.
If there's really a latitude-IQ correlation that isn't just a result of something boring like that people of African descent usually do worse on IQ tests, that would be about ten times more interesting than anything else in this post. I immediately thought "parasites!", but that doesn't explain countries like France and Italy that are advanced enough that parasites aren't a big problem.
The only other thing I can think of is that brain waste heat removal is such a limiting factor that 10,000 years in a cold climate where waste heat is less of a problem can allow Scandinavians to evolve more powerful brains, but that would have to be one heck of an evolutionary incentive.