On Bryan Caplan’s recommendation, I just read The Persistence of Poverty, by Charles Karelis (2007). Karelis seeks to explain these patterns:
Five patterns that have been common among the poor in many times and places, then, and that played a role in keeping them poor or making them poorer, are: 1. not working much for pay; 2. not getting much education; 3. not saving for a rainy day; 4. abusing alcohol; and 5. taking risks with the law. … [And] having children early and out of wedlock … is doubtless a big factor in poverty in the United States today.
His explanation is that often we experience diminishing returns for “relievers” that reduce our pain. Which actually gives us increasing, not diminishing returns for getting “more” in that area. For example:
Consider housing. Suppose we take the perspective of a couple whose house has a bedroom for them and each of their six children, plus adequate room for entertaining and other functions besides. Clearly they are consuming or using housing in the more-than-sufficient range. Their house is a source of positive experience. As each child leaves for college (let us say), the amount of space available for the use of the couple goes up by roughly equal amounts, but probably their enjoyment of the house goes up by smaller and smaller amounts. …
Now imagine a couple whose dwelling is a one-bedroom house that is barely adequate for themselves. If a child arrives, then given the crowded conditions, the couple’s privacy is much reduced, their peace and quiet is disturbed, and they may have to start sleeping in shifts. Whatever the compensating joys of parenthood may be, these are impressive deteriorations in their physical comfort. By the time child number six arrives, the couple may hardly notice the further deterioration in their situation that occurs as a result. One more voice outside the bedroom door when you cannot sleep anyway on account of the five that are already audible probably will not make much difference.
That is, you don’t notice as much of a difference between very and mildly crowded, as you do between mildly and not at all crowded. Karelis further claims (though without sufficient support in my opinion) that this sort of things happens much more often for people who are poor relative to their culture:
Whether something is functioning as a reliever for a given consumer is relative. It is only partly a question of objective economic circumstances, because it depends too upon how the consumer sees those circumstances. … A more anthropological view responds that the difference lies in neither discipline nor opportunity but in the values of the various sub-cultures. … No one doubts that different cultural groups within the United States have different histories, and that these different histories create different economic norms and expectations. For instance, having come from much poorer countries, Asian immigrants to the United States often have relatively low norms and expectations. By contrast, African-Americans, who are closely acquainted with the lifestyles of middle-class whites, and who have long been exposed to “the American dream” and all it implies, often have relatively high norms, if not exactly expectations. … this … predicts that the felt relief of the marginal dollar will be greater for poor Asian immigrants than the felt relief of the marginal dollar for similarly poor African-Americans.
Karelis first published this idea in 1986, and he notes that similar ideas were published by van Praag in 1968 and Friedman & Savage in 1948. I find the idea coherent and mildly plausible, but just based on Karelis’ arguments, I wouldn’t put it much above other common poverty explanations, such as stupidity, sickness, impatience, impulsiveness, and lack of self-control.
However, I gave more weight to this account after I realized that it is implied by my usual favorite account of income status: utility linear in income rank. As income is usually distributed lognormally, the function relating rank to log income must be convex below and concave above median income. This implies diminishing returns in income above median income, and a range below median income with increasing returns to income. When you have increasing returns to income, you value each unit of income more when you have more of it, rather than the usual diminishing returns case, where you value each unit of income less, the more income you have.
Karelis suggests some policy implications:
Seeing that the income effect and the substitution effect of strategies to make work pay will be mutually reinforcing, making work pay [via increased wages] is a double-barreled anti-poverty approach. By contrast, no-strings assistance is a single-barreled approach, since it lacks a positive substitution effect. …
Thirty-five years ago the speeches and writings of American civil rights leaders often framed or interpreted the circumstances of their audiences by “comparing them up”—measuring them against the circumstances of the middle-classes and the upper-middle classes, or even against the images of the good life found within the American Dream. This was openly done for the sake of energizing audiences with discontent. The goal was reasonable enough, but according to our theory, the strategy was probably counterproductive. …
Increasing the differential between the income from crime and the income from honest work—by raising the odds of punishment, lengthening sentences, or making (honest) work pay better—is likelier to be effective than strategies built on the assumption that criminals are dysfunctional and hence unresponsive to sticks and carrots of this kind; and it is likelier to be effective than strategies built on the assumption of atypical preferences. …
Our utility function could [correctly] be used to justify putting the least poor people ahead of the very poorest people in distributing assistance.
Though Karelis didn’t mention it, the same logic says to allow and even promote more gambling among the poor; within the convex region gambles look like a net gain. For the same reason, it would be good to promote inequality among the poor.
However, all these policy recommendations are based on assuming that the preferences of other poor people don’t change when you help one poor person move up their utility function. But if the transition from convex to concave utility, and other aspects of the utility of money, result from the actual distribution of income in one’s reference culture, then helping one person changes the distribution against which others compare themselves.
For example, if utility is linear in rank, the help you give one person is exactly cancelled by the hurt you thereby inflict on the others who this one person has jumped over in rank. Yes, with some other functional form, the help might outweigh the hurt, but with other forms the hurt might outweigh the help. This effect of changing the reference distribution is not small, and shouldn’t be ignored as Karelis does.
Finally, Karelis focuses entirely on immediate choices, rather than on long-term strategies. Young poor people who care about the long run should focus on trying to dig their way out of poverty, and so much less display the six patterns of poverty that Karelis tries to explain. So to predict typical poor behaviors, we need to add a substantial degree of impatience or lack of self-control to Karelis’ account.
Added 4p: Karelis responded to my email, and asked me to post this comment:
Thank you for blogging so thoughtfully about my book. One comment. The hypothesis of the book can fairly easily be extended to the putative fact that impulsiveness, impatience, and lack of self-control are commoner among the poor by introducing the idea that overcoming these weaknesses is a kind of work. The idea would be that this “will-work” will have less appeal when the material gain produced yields a smaller rather than larger addition to utility, as (the hypothesis contends) is the case in the lower income ranges. As I recall, Andrew Sullivan, Ezra Klein, and Ta-nehisi Coates all noted this extrapolation of my theory around the time the book appeared. The amendment was made possible, really, by research into the nature of willpower subsequent to the publication of my book.
This story can work re efforts to make small gains while poor, but doesn’t work re making big gambles or long-term efforts to dig oneself out of poverty. Those should be seen as large gains, even for someone who can’t find sufficient motivation to work for small gains. So to explain a poor person uninterested in either of those, we need to add something else to our story.
Added 6p: Karelis further responds:
Yes, we need to add something else to our story, but maybe not too much. My hypothesis says that medium-sized additions to the consumption of someone at the low end of the income scale (additions that leave them shy of sufficiency) will raise and not lower their marginal utility for the good in question, and thereby increase their motivation to secure more of that good. You are right to see that these medium additions need not come from an external source. Self-help can be as effective as exogenous help in raising marginal utility in this way. For instance, if you have ten unwashed dishes in the sink, and someone washes eight of them, the psychological benefit you will get from washing the last two will be more than you would have gotten from washing the first two, and it doesn’t matter whether that “someone” is you or a friend.
So why don’t poor people perform this kind of self-help more often? They may have internalized the law of diminishing marginal utility, just like the policy folks who resist helping the poor for fear of undermining their motivation for self-help. More likely, none of us, rich or poor, is very good at self-gaming, i.e. figuring out and acting upon the likely impact of possible actions on what will seem rational to us when we have completed those actions.
One might plausibly argue that no one ever really makes long term plans. People who seem to be doing so, such as students going to college, are really just executing standard cultural plans, doing “what you are supposed to do”. Then the “extra” we’d need to add, to explain why the poor don’t seem to have long term plans to dig them out of poverty, is to say that the cultures of poor people often don’t have standard cultural plans that induce them to so dig.
What culture? Many poor people, and I ran into quite a few of them as a registered nurse, work their rear ends off just to survive. Not everyone has the intelligence or the option to obtain a higher education and end up with a high paying job. I saw people who never missed a day of work, but the state's minimum wage was a whopping $7.25 an hour, plus many worked for years in the textile industry and the plants shut down kicking them to the curb, shafting them on their retirement, and leaving them with only Social Security when they became of age. Many were "frequent flyers" to the hospital because they simply couldn't afford their medications. Our problem is we've allowed those at the top to increase their income exponentially, while the vast majority of the population's income hasn't even kept up with inflation for decades.
"However, I gave more weight to this account after I realized that it is implied by my usual favorite account of income status: utility linear in income rank. As income is usually distributed lognormally, the function relating rank to log income must be convex below and concave above median income. This implies diminishing returns in income above median income, and a range below median income with increasing returns to income. When you have increasing returns to income, you value each unit of income more when you have more of it, rather than the usual diminishing returns case, where you value each unit of income less, the more income you have."
This paragraph could use a graph to help the explanation. I found it very insightful, but had to spend 3 minutes going back and forth between here and wikipedia to clearly understand what you were saying.