Saturday, August 03, 2019

Why Kevin Williamson is wrong about poverty and bad behavior

I recently wrote a post at Bloomberg Opinion arguing that "bad behavior" - drug use, violence, single parenthood, and idleness - is not the main cause of poverty in advanced nations. As evidence, I cited the country of Japan, which has extremely low rates of drug use, violence, single parenthood, and idleness, and yet which has a poverty rate almost as high as that of the U.S., and significantly higher than those of wealthy European countries. Since Japan has so little bad behavior and still has a pretty high poverty rate as advanced nations go, it must be the case that bad behavior, in aggregate, is not the major cause of poverty.

Kevin Williamson of the National Review took issue with my post. In a strongly worded rebuttal, he calls my piece "a stale slab of conventional wisdom", "sloppy thinking", "tedious writing", a "mishmash of tendentious platitudes and misunderstood truisms", and "sloppy analysis, if it counts as analysis at all." Yet despite this vitriol, Williamson fails to substantively rebut any of the points I made. In some cases, his arguments contain logical errors; in others, he simply misunderstands my argument.

Let's go through Williamson's post, and show why it fails to rebut my arguments.

1. Absolute poverty or relative poverty?

Williamson's first attempted rebuttal relies on the idea that Japanese poor people are not really poor:
 It is not obvious that Japan “has lots of poverty.”...Smith here relies on a useless measure of “relative” poverty, the share of the population earning less than half of the median income. You can see the limitations of that approach: A uniformly poor society in which 99 percent of the people live on 50 cents a day and 1 percent live on 49 cents a day would have a poverty rate of 0.00; a rich society with incomes that are rising across-the-board but are rising much more quickly for the top two-thirds would have a rising poverty rate, and some people who are not classified as being in poverty this year might be in poverty next year even though their incomes are higher, etc. It would be far better to consider poverty in absolute terms, but our progressive friends are strangely resistant to that.
Let's leave aside the question of whether poverty is best conceived of in absolute or relative terms. There are good arguments on both sides (and perhaps room for even more definitions of poverty than those two!). But it's definitely true that the average Japanese poor person enjoys a significantly higher standard of living than the average poor person in, say, Ethiopia.

But this is also true of American poverty! The kind of deprivation that conservatives like Williamson often blame on bad behavior in the United States is also relative poverty, not absolute. "Poor" Americans, including the residents of Garbutt, NY that Williamson famously blamed for their own economic difficulties, are also much richer than the average poor person in Ethiopia. So when evaluating conservative beliefs about behavior and poverty, we should look at relative poverty, because that's exactly the kind of poverty conservatives are generally talking about.

Now, if Japan were richer than the U.S., it might not be an appropriate country to compare ourselves to. If Japanese people making 50% of Japanese median income were materially better off than Americans making 50% of American median income, then Williamson's argument would have some bite. But in fact, the exact opposite is true. In purchasing power parity terms, Japan's median household income is only about 63% of Americans' median household income.

In other words, in absolute terms, someone at the Japanese relative poverty line is doing even worse than someone at the relative American poverty line. So it's not clear why Williamson thinks that insisting on an absolute standard of poverty, rather than a relative one, will advance his case.

2. What does looking at national averages tell us?

Williamson's next argument is that instead of looking at national averages of behavior (violence, drug use, etc.), we should look specifically at the behavior of the Japanese poor:
Secondly, it is not entirely clear that the Japanese are as free from the pathologies that attend poverty in many other places as Smith suggests. It is true that Japan as a whole has low rates of chronic unemployment, drug use, single motherhood, etc., but the relevant question here would be how Japanese who are poor compare on these metrics with Japanese at large. To assume that the situation with the poor can be approximately deduced from national averages is pretty sloppy analysis, if it counts as analysis at all.
Williamson is wrong about the relevant question. The relevant question is P(relative poverty | absolute bad behavior). In other words, the relevant question is: "How much does bad behavior change my chances of falling into the lower echelons of my developed country?". The percentage of Japanese people who are badly behaved is much smaller than the percentage of Americans who are badly behaved. Yet about the same fraction of people there fall into the lower echelons of that society (which, as noted above, are actually lower in absolute terms than the lower echelons of American society, at least at the 25th percentile). Thus, even if there are individual Japanese people who become poor due to bad behavior, it can't statistically be the biggest factor, as long as poverty has roughly the same causes in both countries.

Just to show how this works, let's do a simple example. Take one measure of bad behavior, e.g. violent crime. Suppose, hypothetically, that the violent crime rates per 100,000 population were:
Poor Japanese people: 2
Non-poor Japanese people: 1
Poor Americans: 20
Non-poor Americans: 10

In this hypothetical, there is a behavior gap between poor and non-poor Japanese people. In fact, the ratio of bad behavior between poor and non-poor is the same between the two countries! But as long as the causal relationship between bad behavior and poverty is the same for both countries (and I'll talk more about this assumption in point #9 below), American poverty could not be reduced much simply by telling the poor Americans to stop being violent; in fact, they'd have to lower their violent crime rate by 95% just to catch up with their poor Japanese counterparts!

Statistically, it could also be the case that Japanese poor people are about as badly behaved as poor Americans. In other words, there could be one cohort of really badly behaved Japanese people hogging the lower end of the income distribution, while everyone higher in the distribution acts like a saint. In this case, the differences in the averages of Japan and America all come from the middle and upper classes of the two countries. (This example is totally unrealistic and false, but let's run with it, just for fun). In this example, non-poor Americans are much much worse behaved than non-poor Japanese people...yet still, this bad behavior doesn't cause them to fall into poverty. In other words, even this imaginary and extremely contrived situation would support my argument instead of Williamson's!

These are just two examples. But in general, you won't be able to find a single function relating bad behavior to poverty that fits both the Japanese data and the American data. The math just doesn't work.

So as long as we're willing to assume that the causes of poverty are similar from country to country, then just from looking at the national averages, we can conclude that bad behavior is statistically not the main cause of poverty in rich societies.

3. What about alcohol?

In my article I focused on four dimensions of bad behavior - illegal drug use, out-of-wedlock births, violence, and idleness. Williamson suggests there is another type of bad behavior I failed to consider: alcohol.
Third, it emphatically is not the case that Japan is a society that is largely free from substance abuse. In Japan, as in the United States, the most socially significant and destructive mode of substance abuse is legal: alcohol abuse. Japan has a big problem with alcohol, and alcohol abuse is related to joblessness and poverty, although the question of causality (Are they unemployed because they drink, or do they drink because they are unemployed?) gets complicated, and some studies suggest that in Japan some kinds of destructive drinking increase with income.
It's true that alcohol is typically the drug of choice in Japan. But even here, America is worse-behaved. Several different data sources all agree that the number of alcoholic drinks consumed per person in Japan is around 7, while the number in America is around 9. France, Germany and Australia, which have lower poverty rates than the U.S., are around 12.

It's also instructive to examine the Reuters article that Williamson links to regarding alcohol in Japan. The article notes that alcoholism is a problem in the country. But its three examples of alcoholics are 1) a civil servant who remained employed after six months in the hospital, 2) the country's finance minister, who died in office, and 3) a prince. These examples illustrate how differences in institutions affect the relationship between behavior and economic outcomes. Despite alcoholism, the civil servant was still employed, the finance minister was still finance minister, and the prince was still a prince. More on this later.

But in any case, suppose Williamson has a point, and alcohol, not illegal drug use, out-of-wedlock births, violence, or idleness is the big behavior-related cause of poverty. That would imply that conservatives' emphasis on the former four types of behavior is probably misplaced, and they should focus more on encouraging temperance and taxing alcohol more highly.

4.  Why isn't Japan's system preventing more poverty?

Williamson asks, if Japan has higher employment rates and national health insurance, why is their poverty only slightly less than that in the U.S.?
Smith is correct that Japan has high work-force participation, and that it has a universal(ish) national health-insurance scheme. To which he adds: “Too many people fall through the cracks in the capitalist system because of unemployment, sickness, injury or other forms of bad luck.” This is an odd thing to write immediately after noting that Japan has 1. low unemployment and 2. a national health-care system that helps people through sickness and injury. 
Perhaps those things are not sufficient?
This is absolutely true; these things are not sufficient. But nowhere did I say they are.

National health insurance helps reduce individual bankruptcy risk and (if done right, as in Japan) control costs. Maintaining low unemployment also reduces bankruptcy risk, and the risk of losing one's skills and connections. But Japan does show that neither of these will be sufficient to attain low European levels of poverty. For that, we will need more - other forms of income transfers, and/or institutions to ensure that more of society's income flows to lower-wage workers.

5. Is poverty due to "capitalism"?

Williamson argues that to attribute poverty to "capitalism" ignores the differences between countries' systems:
“Capitalism” is a very broad term. The United States is a capitalist country, and a rich capitalist country at that. So is Japan. So is Singapore. So is Sweden. So is Switzerland. These countries have radically different health-care systems, tax codes, family lives, cultural norms, etc. Unsurprisingly, these produce different outcomes on a great many social fronts — but all of them are comprehended by “capitalism.”...To argue that the problem is “the capitalist system” is to retreat into generality and to refuse to consider the facts of the case, each on its own merits.
It is true that different advanced countries have different systems, and that labeling them "capitalism" tends to obscure more than it clarifies. In fact, I recently wrote a whole article about that topic, which Williamson should read!

But since the time of Vilfredo Pareto it has been well-known that every country has a substantial amount of market poverty - that is, poverty before taxes and transfers. Here is a graph from the Economic Policy Institute of market poverty rates vs. post-transfer poverty rates:

In other words, no matter how you set up your system, you're going to get a lot of people who will experience relative poverty without government transfers. Hence, the social safety net matters a lot.

Capitalism is not a bad thing. Capitalism, in some form, is an amazing engine of wealth creation. Capitalism of some sort, as far as we know, is absolutely necessary to maintaining high standards of living and eliminating absolute poverty.

But capitalism is not omnipotent. Drowning government in a bathtub and leaving individuals to sink or swim on their own in a free market economy will result in some people failing and being poor, no matter how well they behave. Thus, any capitalist system can be improved with a social safety net.

That was my point. I think most people got it!

6. Extreme poverty and mental illness

Williamson notes that much extreme poverty in the U.S., including chronic homelessness, is caused by mental illness:
In New York, Los Angeles, and other big cities, it is common for people to sleep on the streets even as beds in shelters go unoccupied. There are many reasons for that, but the main one almost certainly is mental illness (and substance abuse as a subset of that). That is the nearly universal opinion of the professionals who work with the urban homeless. 
There are better and worse ways to deal with mental illness in a wealthy, complex society, and we in the United States have settled on one of the worst: After the “deinstitutionalization” of the 1960s and 1970s, in which left-wing liberationist thinking combined with right-wing penny-pinching to gut the public mental hospitals, we punted the problem to the police and to the jailers, who are ill-equipped to handle it. The United States is not alone in this. Many (perhaps most) Western European countries have more effective social-welfare systems than we do, but even in Sweden, with its fairly comprehensive welfare state, mental illness is the leading cause of “work force exclusion,” as they call it.
As far as I know, this is all absolutely true. But notice that mental illness is not generally a behavioral issue - alcohol, drugs and violence can exacerbate your chances of becoming a schizophrenic, but there are tons of people with mental illness even in well-behaved countries like Japan. As Williamson notes, the most effective way of dealing with mental illness is government-funded treatment. And as Williamson notes, "right-wing penny-pinching" was indeed one reason the American government stopped providing the classic form of such treatment - compulsory institutionalization (which still exists in Japan). The question of how best to treat mental illness is an open one, but it's basically certain that the answer will be carried out by government intervention rather than by some improvement in personal responsibility - again, speaking to my thesis.

7. Would more antipoverty programs reduce poverty in America?

Williamson writes:
[T]he big changes that progressives generally propose for the United States — a national health-care system like Japan’s, an enlarged welfare state more like Sweden’s — do not seem to have been entirely effective in the places where they have been tried. And there is good reason to believe that Swedish or Swiss practice cannot simply be imported into Eastern Kentucky or Baltimore and replicated locally. That does not mean that there is nothing to learn from Japanese or European practice — perfection is not our criterion — but it does complicate the conversation. We have, in fact, spent a tremendous amount of money on anti-poverty and economic-development programs, and much of that has not delivered anything like the promised return.
Well, it is the case that the U.S. spends less on social welfare than most European countries:

But it's also true that Canada and Australia have lower poverty rates than we do, despite lower rates of welfare spending (at least according to this measure). Williamson is almost certainly right that the U.S. has different challenges than these other countries. That may mean we need to spend more in order to achieve the same effect. But before we go claiming that government social spending won't work for the U.S., we should at least try spending as much as the UK or Germany (or Japan!). 

As for Williamson's statement that "We have, in fact, spent a tremendous amount of money on anti-poverty and economic-development programs, and much of that has not delivered anything like the promised return", that may or may not be the case, but much of it has delivered the promised return. Just looking at recent examples, we see that government transfers have been entirely responsible for the fall in the child poverty rate since the early 1990s:

Also, thanks to policies begun by George W. Bush and continued by Barack Obama, homelessness in the U.S. has fallen substantially:

These approaches are working. Tell me again why we shouldn't double down on things that work, especially if we can afford to do so?

8. Sentimentality and poverty reduction

Williamson, whose work I cited in my original post as an example of a conservative blaming poverty on bad behavior, writes:
In my own reporting on poverty in the United States, I have tried to present the facts as unsparingly as I can. Perhaps Noah Smith thinks that I do this in order to savor the exquisite delights of moral condemnation. But the intended purpose is to scour away the crust of sentimentality that poverty has acquired in order that we may deal with the actual facts of the case in a way that is productive and that does not end up deepening the very problems we hope to mitigate. (emphasis mine)
Here are some quotes from the Williamson article I cited:
Yes, young men of Garbutt — get off your asses and go find a job... 
[N]obody did this to them. They failed themselves...There wasn’t some awful disaster. There wasn’t a war or a famine or a plague or a foreign occupation. Even the economic changes of the past few decades do very little to explain the dysfunction and negligence — and the incomprehensible malice — of poor white America... 
The truth about these dysfunctional, downscale communities is that they deserve to die. Economically, they are negative assets. Morally, they are indefensible. Forget all your cheap theatrical Bruce Springsteen crap. Forget your sanctimony about struggling Rust Belt factory towns and your conspiracy theories about the wily Orientals stealing our jobs. Forget your goddamned gypsum, and, if he has a problem with that, forget Ed Burke, too. The white American underclass is in thrall to a vicious, selfish culture whose main products are misery and used heroin needles.
 And in the article Williamson himself cites, he wrote:
Thinking about the future here and its bleak prospects is not much fun at all, so instead of too much black-minded introspection you have the pills and the dope, the morning beers, the endless scratch-off lotto cards, healing meetings up on the hill, the federally funded ritual of trading cases of food-stamp Pepsi for packs of Kentucky’s Best cigarettes and good old hard currency, tall piles of gas-station nachos, the occasional blast of meth, Narcotics Anonymous meetings, petty crime, the draw, the recreational making and surgical unmaking of teenaged mothers, and death...
If this represents "scouring away the crust of sentimentality", what does poverty reporting look like with the crust of sentimentality still on??

9. What an actually good conservative critique of my article would look like

It's true that an 800-word comparison of two countries can't make a watertight case that bad behavior is not the biggest cause of poverty. It would definitely be possible for a conservative to read my article and then poke holes in my arguments. I don't think Kevin Williamson has done this. But if he did, I think it would look something like this:

"The U.S. and Japan don't have comparable economic systems, so comparing their behavior and their poverty rates is inappropriate. America is a meritocracy where you can climb high if you work hard and act right. Therefore if you're poor in America, it must mean that you chose to sabotage yourself. But Japan's combination of lifetime employment, an inefficient corporate system that doesn't reward effort and achievement, sexism, and permanent social stigma for failure means that poverty there is usually caused by bad luck."

In other words, different countries, even different developed countries, might have fundamentally different reasons for allowing people to fall into poverty.

This is a real possibility. If this is true, it would mean that Japan has lots of room to reduce poverty by making their system more like America's, so that their well-behaved, hard-working poor people can escape poverty by the sweat of their brows. And it would mean that America is already doing about as well as it can do in terms of institutional set-up, and that the best it can do is to urge its poor people to try harder and be more moral.

Of course, I could also marshal plenty of data against that case - for instance, the aforementioned success of U.S. government transfers in reducing child poverty and homelessness in recent decades. Or the big decreases in American violence and drug use since 1990 that haven't been accompanied by decreases in market poverty. But at least I would need more than just the example of Japan!

(In fact, this is the counterargument that I was prepared for, since people often respond to Japan-based arguments by saying "Japan is just different". But Williamson didn't use it.)

So even though my article does not completely settle the question of the root causes of poverty (and indeed, no article will), Kevin Williamson's attempted rebuttal does not hit the mark. For a good follow-up to my article, see this post by Scott Sumner. And thanks to Sumner for the photo at the top of this post.


  1. I have two thoughts. First, compared to you, Williamson is relatively innumerate. He sees a relative poverty measure, so he criticizes it not considering whether the same relative poverty line corresponds to more severe absolute poverty in Japan. He says you can't infer anything about the rates of bad behavior among the poor by looking at national averages without doing the arithmetic and seeing that, even if all Japanese misbehavior is committed by the poor, they still behave better (by his standards) than the average American.

    As is typical of US conservatives, he considers current policy in the USA to be way lefty because it's moved left since 1932. If he compared the huge US antipoverty efforts to the huger efforts in other developed countries he wouldn't make the silly argument he makes. Also I would exclude old age pensions (not related to dealing with the sort of poverty he denounces) and healthcare (not counted as income and not appearing in poverty rates). If you looked at means tested assistance other than Medicaid (or NHS or whatever) the USA would stand out more.

    Yes on Medicaid not counted. As is typical of those who say we fought a war on poverty and poverty won, he ignores all of the actual battles. If one measures poverty with pre tax cash income, one will conclude that all the money going to the EITC, SNAP and Medicaid was wasted. This just because none of it is considered when calculating the meaningless traditional poverty rate. In fact, as you note, anti poverty spending is correlated with a reduction in poverty.

    Also causation and correlation. Poverty is correlated with welfare program participation. That's true by law. But that doesn't mean that welfare causes the culture of poverty. In fact, the evidence suggests welfare spending promotes independence decades later. I am thinking of food stamps (introduced at different times in different counties) and an index of economic independence (high school graduation, not getting pregnant before 20, not getting welfare, having a job, and not being convicted of a felony). The evidence is food stamps for one's family about when one is born promotes self reliance 20 years later (not super strong and as usual for girls not boys who are just hopeless (as we know having been boys)).

    But there is similar evidence about the EITC, Medicaid and widows pensions (the first huge national welfare program in the USA).

    Another question is what about the morals and culture of non college educated whites changed. William Julius Wilson would say that when good jobs vanish, breadwinner norms soon follow. He said that about truly disadvanted African Americans, but his hypothesis implies that that which was true of blacks would become true of whites with automation, increased import competition and weakened unions. And that's what happened. There is a comprehensible even testable relationship between economics and morals which tends to support historical materialism. In contrast, the stories starting with morals don't fit the data. If the decline in religion is the problem, the Sweden would be a hell hole. But it isn't.

    I really shouldn't pile on. You can handle Williamson while typing with one hand tied behind your back. But piling on is fun.

  2. If one wants to answer the question: "How much does bad behavior change the chances of individuals falling into the lower echelons of my developed country?" then comparing a country with lots of bad behaviour to one with not much is not going to answer that question.

    Assume the answer is "bad behavior increase the chances of being in the lower echelons a great deal". Take a society with lots of poor, badly behaved people. They become aware of this and stop being badly behaved. Consequently many of them rise up the relative income scale. This country used to look like the US and ends up being like Japan.

    There would however have to be a fallacy involved to compare the society when it looked like the US to the same society at a later point in time when it looked like Japan and conclude 'this disproves the theory that bad behavior causes relative poverty' when it was this very fact that drove the change.

    1. then comparing a country with lots of bad behaviour to one with not much is not going to answer that question.

      Yeah, it really is.

      Take a society with lots of poor, badly behaved people. They become aware of this and stop being badly behaved. Consequently many of them rise up the relative income scale. This country used to look like the US and ends up being like Japan.

      You mean...a poorer country, with similar levels of relative poverty??

    2. This comment has been removed by the author.

  3. Anonymous10:25 AM

    For part 9, there’s also the possibility that reducing poverty with cash transfers won’t stop the bad behavior, and we should be finding ways to reduce that too, or life in the American underclass will still be pretty miserable.

    1. Sure, no one is saying that drugs, violence etc. are good for society.

    2. Anonymous4:03 PM

      Of course. But do you think cash transfers can help stop them? Your Japan analogy is a good rebuttal of the conservative argument that bad behavior causes poverty...and also of the liberal argument that poverty leads to bad behavior. So on social breakdown etc. Williamson is on stronger ground.

  4. OK, could I answer the question "How much does smoking change my chances of falling into the lower percentiles of lung health?" by comparing countries with different levels of smoking ?

    Lets say that in both countries 80% of smokers fall into the bottom half of lung health measures, but in one country 20% smoke and in the other only 5% smoke. The data would be consistent with (but would not prove) the statement "Smoking significantly increases my chances of falling into the lower percentiles of lung health". As far as I can see the relative smoking rates are totally irrelevant to this.

    Am I missing something obvious here ?

    1. This was meant to be a response to Noah's comment above (time stamped 3:36 PM)

  5. Blaming the poor for their poverty absolves us from having to understand it or do anything about it and makes us feel we earned our rewards. It is about feeling good, not about understanding. It is the stale slab of conventional understanding, sloppy thinking, and tedious writing...

  6. "The relevant question is P(relative poverty | absolute bad behavior)"
    No, it is not. The relevant question is P(relative poverty | do(absolute bad behavior)). If you try to answer a causal question, you have to articulate it using causal operators.
    In plain English, what you want to know is "what would be the poverty rate in US if it had no bad behavior (or the level of bad behavior of another country, say Japan)”. Both you and Williamson are not addressing properly that question.
    (Almost) everyone would agree that bad behavior, all other things being equal, can only increase poverty rate but the true question is by how much. Of course Williamson article is written with astonishing bad faith. He takes whatever fact goes his way, and sometimes even just make things up. However your points are suggestive that it is more complicated than that but don’t go much further.
    A purely statistical approach (which you did in implicitly) cannot answer that question, even controlling for confounding variables. You need a causal model.

  7. Noah 97.5, Kevin 2.5

    And anyway, we're looking at cumulative causation of various kinds, so most of what's discussed gives us clues but not much more. We're not hypothesising about planets here, but whole eco-systems in which the players have agency.