Now that I’ve sketched The Persistence of Poverty, how does it stand up?

The Right

1. Karelis blames persistent poverty on persistent poverty-inducing behavior: not working, not finishing school, not saving, abusing alcohol, committing non-lucrative crime.  While his evidence is a bit thin, almost everything else I’ve read on poverty confirms that such behavior (plus impulsive sex) is indeed one of poverty’s chief causes.

2. Karelis says that, contrary to standard economic theory, marginal utility is often increasing (and marginal disutility is often decreasing). He produces several convincing examples, mostly involving the alleviation of pain (bee stings, foot blisters), but also consumer products (like scratches on a car).  He could easily have produced numerous other examples of increasing marginal utility.  Any “acquired taste” works.  Almost all hobbies – everything from baseball to opera, chess to stamp collecting – become more enjoyable as you learn to appreciate their subtleties.  The same goes for friendship and love; they’re often awkward at first, then become increasingly enriching… at least for a while.

The Wrong

1. Karelis has a clear story about when utility is decreasing and when utility is increasing: utility is increasing for “relievers” but decreasing for “pleasers.”  For all its clarity, however, Karelis’ story is doubly wrong.

He’s wrong on relievers; it is child’s play to list important examples with decreasing marginal utility.

We normally take our vision for granted, so fixing blind eyes is a clear-cut reliever.  Now suppose you choose between: (a) one good eye for sure, or (b) a 50/50 gamble between normal vision and blindness.  Karelis’ story implies that most of us would choose (b)!

If that seems extreme, sleep is another clear case of a reliever.  Now suppose that tonight you choose between (a) sleeping seven hours a night for the next eight days, or (b) sleeping eight hours a night for seven of the next eight days, and sleeping zero hours on the eighth day.  Karelis’ story again implies that most of us would choose (b)!

He’s also wrong on pleasers.  As mentioned, most hobbies have increasing marginal utility.  Many luxury experiences – clear examples of pleasers – are ruined by one salient flaw.  Karelis mentions scratches on cars, but we can say the same about getting a sunburn on a tropical cruise, lacking alcohol at a party, or watching a movie on a screen with a single stain on it.*

2. So what?  If lots of relievers have decreasing marginal utility, and lots of pleasers have increasing marginal utility, we no longer have much reason to think that the poor are more likely to encounter increasing marginal utility than the rich.  As a result, Karelis has little reason to predict different behavior for the poor and the rich.  And different behavior for poor and rich is exactly what his theory is designed to explain!

3. Karelis loves talking about the impoverished ancient Lydians, who were so poor that they only ate every other day:

[L]ike many poor people today, the Lydians allocated their meager resources unevenly between time-slices of themselves. If we assume that “severe famine” means a condition where luxurious consumption cannot be achieved even by saving, their pattern must have been “skimp a little/skimp a lot”: eat no more than enough one day and nothing the next. Further indication of their preference for uneven consumption is the very fact that what they used to distract themselves from hunger were games of dice and similar games. Assuming they gambled on these games, and that they staked their meager holdings, the result for the individual would have been further oscillation in consumption, over and above the practice of eating on alternate days.


But there are worse things than having some days of deep poverty. For instance, there is having twice that many days on which the poverty is half as deep. That again is the point of Lydian prudence.

On reflection, however, the Lydians are a bizarre outlier at best.  Where on Earth do poor people regularly choose to starve on alternate days?  How many poor people who can afford a cheap permanent address choose instead to oscillate between homelessness and a comfy hotel room?  Actual behavior of the poor closely fits the standard economic model of decreasing marginal utility: The poor eat cheap meals on a daily basis, and reside in cheap apartments on a monthly basis.

4. Increasing marginal utility fails to explain why the poor don’t save.  Karelis’ model could just as easily predict that the poor will deprive themselves in the present in order to binge in the future.  (He does mention the poor saving for festivals, but still treats non-saving as the norm).

5. Increasing marginal utility can explain why the poor steal, rob, and deal drugs more; the average earnings may be low, but the variance is high.  However, increasing marginal utility does little to explain why the poor commit more crimes of pure violence, drunk driving, vandalism, and other offenses that almost never lead to financial gain.

6. Karelis is highly confident that unconditional transfers increase work effort.  Virtually no empirical researcher shares this view; they argue about the magnitude of the disincentives, not the sign.  In any case, you don’t need empirical research to see the absurdity of Karelis’ position.  If unconditional transfers increased work effort, then current recipients of these transfers – many of whom don’t work at all – would work even less if the government cut them off completely.  Are we really supposed to believe that able-bodied welfare recipients would voluntarily starve to death – and watch their children starve to death – rather than work at McDonald’s?  Indeed, Karelis’ model predicts that when famine strikes, people will struggle less aggressively for food than they do during normal times.  Not what we see in the real world, at all.

7. To his credit, Karelis open-mindedly invokes the reader’s introspection and the testimony of the poor as evidence.  Yet both introspection and testimony strongly contradict his idea that the poor are so miserable that more responsible behavior would barely help them.  Most obviously, every book I’ve ever read on single mothers says that being poor with a child is much harder than being poor without a child.  Poor childless single women only have to do one job to take care of themselves.  Poor single mothers have to juggle two jobs – their paid work and their parenting.  Poor childless single women who are down on their luck can just move back in with family.  Poor single mothers who are down on their luck have to find a relative willing to take them and their kids.  Poor childless single women have to get to their job on time; poor single mothers have to coordinate their work commute with their kids’ commute to day care or school.

The same goes, of course, for combining poverty with a drinking problem.  The non-drinking poor just have to get to work and pay the bills.  The drinking poor have to avoid getting fired despite drunkenness, hangovers, and so on.  Similarly, consider the volatile combination of poverty with crime.  The law-abiding poor just have to get to work and pay the bills.  The law-breaking poor have to evade the police and dodge bullets.

In all the ethnography I’ve read, I can’t recall a single example where a poor person said that any one of their problems was “no big deal, because they had so many other problems,” or “a few extra dollars a month would barely make a difference.”  One of the best treatments, Edin and Lein’s Making Ends Meet, shows that poor women energetically supplement their main income sources – welfare and legal work – with small sums from family, ex-boyfriends, current boyfriends, illegal work, and charity.  If Karelis were right, why would they be so eager for a few extra bucks?

The Overstated

1. Karelis criticizes almost all competing theories for lack of parsimony, which he regards as a dire intellectual fault: “Unusual tastes and preferences, whether located in the individual or in the culture to which the individual conforms, are precisely the sort of variable we should try to omit from our explanation of poverty, absent overwhelming evidence for them.” (emphasis added)  Yet on reflection, we can easily make competing theories elegantly parsimonious.  Instead of saying, “We have one theory for poor people, and another for everyone else,” why not just say, “Whatever causes persistent poverty is a continuous variable”?  Thus, if laziness causes persistent poverty, you don’t have to say that the poor are lazy and everyone else isn’t.  Just say that everyone is lazy to some degree, but the poor are lazier.  If impulsiveness causes persistent poverty, you don’t have to say that the poor are impulsive and everyone else isn’t.  Just says that everyone is impulsive to some degree, but the poor are more impulsive.

2. If I were Karelis, I would downplay the importance of parsimony.  After all, his theory is so complicated that it takes multiple blog posts just to explain.  In contrast, I was able to run through his six competing theories in a single post.

3. Intellectual salesmanship aside, persistent poverty really is a complicated problem.  Embracing any one explanation, however “parsimonious,” is awfully dogmatic.  Thus, suppose that Karelis’ story made perfect sense.  He would still need to acknowledge a major role for what he calls “restricted opportunities.”  The poor have lower IQs, worse health, and less inherited wealth.  Housing regulation forces them to either live in low-wage areas or spend most of their money on rent.  Indeed, if you take a global view, most of the poor can’t even legally work in the First World.  Similarly, even if Karelis’ story made perfect sense, he would still need to acknowledge a major role for preferences.  Some people dislike working more than others.  Some people like alcohol more than others.  Some people have more violent tempers than others.  All of these preferences – and many others – tend to make you poor.  Obviously.

* Yes, you can try to redefine luxuries as relievers, but only by making Karelis’ theory tautological.  “Things that relieve discomfort have increasing marginal utility” is supposed to be an empirical claim, not a definition.

Read Caplan’s next Karelis post here.