Thursday, March 21, 2013

Intergenerational Inequality

Here's some food for thought, in chart form. All of the data is from the General Social Survey.**

If your father didn't graduate high school, then you are 8 times more likely not to graduate high school than if your father did. If your father graduated college, then you are 3 times more likely to graduate college than if your father didn't.

You would think this would be a result of income inequality. Fathers who don't graduate high school, with children who don't either, tend to be much poorer than fathers who have, with children who have. That's true.

But you'd be wrong to think that this effect is largely explained by income. It's not. Intergenerational inequality in educational attainment persists after adjustment for income inequality.

Here's a graph of educational attainment by real income. Sorry about the x-axis labels; I can't figure out how to fix that. The scale runs from zero to approximately $180,000, with income rising steeply in the fourth quartile. Let's just stop for a minute to observe the graph. Notice that college graduation rates are approximately 10 percent until the midpoint -- and the high school dropout rate is close to 50 percent -- that's the $20,000 mark, roughly speaking. The peak of high school degrees occurs at the 75 percentile -- that's the $50,000 mark. Note how concentrated college degrees are at the top end of the income distribution.

OK, here's the terrifying part about intergenerational inequality. I'm going to divide the sample which you see graphed above into three groups: (1) people whose fathers did not graduate high school; (2) people whose fathers did graduate high school but not college; and (3) people whose fathers graduate high school and college.

What I want you to observe is the striking difference in educational attainment regardless of income level. And what you should realize is that inequality is, on balance, way more of an intergenerational phenomenon than one determined by current-generation income.

Group 1: People whose fathers did not graduate high school

Group 2: People whose fathers did graduate high school but not college

Group 3: People whose fathers graduate high school and college

The data suggest inequality is a problem of families, not of individuals. Even among the poor, basically every person whose father graduated college graduates high school themselves, and 40 percent or so graduate college. Whether your father graduated high school is much, much more predictive of whether you have than how much you make.

Without tempting the Lucas critique,* I would suggest that these results can be exploited by policy.  The value of a high school or college degree is not just the value of that person holding the degree. It's the geometric sequence of that person and all of the succeeding family members who earn degrees because of him.

Here's why that math is important. The high school dropout rate among people whose fathers were dropouts is 22.2 percent. The dropout rate with high-school-grad fathers is 2.9 percent. Let's assume that the social value of a high school degree is $30,000 per graduate; that's roughly the difference in average income between non-grads and grads. Public policy that supposes they are helping one person assesses the value of that degree at $30,000, obviously. Public policy that supposes they are helping an infinite succession of people assesses the value of that degree at $819,000.

(For math people, that is the difference in the sums of the geometric sequences of 30,000 * .971^n and 30,000 * 0.88^n, plus 30,000 for the original degree. The intuition here is that "one degree" is not really one degree, but 27 degrees because of the effect on the likelihood of future degree holders.)

We'd do a lot more on education policy margin, clearly, if we thought that the value of a degree was $819,000 rather than $30,000.

I know, I know -- intergenerational inequality is a very uncomfortable subject for Americans. But we need to talk about it more. Only when we recognize that the Dream is largely hollow can we begin to do something about it. Progressive taxation is no substitute for real policies to address intergenerational inequality. We need to have a conversation about how we achieve increases in the high school graduation rate and the college graduation rate, as well as all of the variables which impact those rates, such as household breakup, crime, lead, etc.

* The reason I'm not running into Lucas critique territory is because I think these results are micro-founded. My view is that education actually changes the expected utility functions of the agents to favor education.

** For the people who want to do more research, the relevant database queries are DEGREE, PADEG, and CONINC.


  1. If 30,000 is the difference between average income, the geometric multiplier would be on a far, far higher base than just that – considering that one earns his income over a lifetime.

    I would argue 11,000 is a better, and more appropriately conservative estimate of "value-added". Your figure includes the value from increased potential of going to college (hence 20k vs about 50k). A lot of trends lately show that college doesn't add nearly that much value, furthermore that potential isn't randomly distributed among high school graduates.

    In terms of net value added to society the difference is between about 19k and 30k, by that measure. If you account for overall lifetime gains, the figure increases to over 10 million with a more conservative estimate, and 26 million with yours.

    If we're considering infinite horizons, I don't see any reason not to do this. The policy goal I parse from this is the relative unimportance of college and the increased need to focus on high school. I'm still somewhat unconvinced, though, that this is a causal relationship.

    I have more thoughts, about your methodology as well as conclusion here:

  2. Evan, I am really not doubting your conclusions, but have you thought through the conditions that populate the intersection of father with high school and bachelors and below national median family income. Since bachelors generally codes for way above median income, this must be the extreme low income tail of that sub-group (so a tiny percentage of the overall sample). So why do these families have such low incomes relative to their education. I am betting there are a plethora of odd circumstances such as temporary unemployment, incarceration and unreported income that might might also suggest that there are other resources available to many of these families.

  3. Increasing graduation rates are easy - just pass everyone! I kid, as I believe you have quality controlled increase in graduation rates in mind implicitly. Nevertheless I think that should be explicit as it highlights how hard it is to actually increase the rate.

  4. Definitely interesting. There's a bit of potential for confounding by age and by increasing education: older people are less likely to have fathers with degrees, and tend to have higher income.

    The pattern still seems to be similar if I restrict to ages 25-40. Will post at when I have more: in particular, graphs that smooth over income.

  5. Economists lower wages predictions for more education have come true Abstract

    1) P. Krugman: rising wage premium for higher education is over This prediction has come true! Appendix 1
    2) L.Thurow: little/no payoff from a little bit more education This explains why so many college graduates earn so little.
    3) C. Murray: encouraged more investing in our best and brightest President disagrees stressing CC system.
    4) A. Greenspan: solving skills shortage with education reform will take years Bill Gates want more H-1B visas.
    5) P. Drucker: stress what learners do well, schools don't do that Peter is definitely against testing. See Appendix 2


  6. This is consistent with adoption studies, which find very little relationship between adoptive parents' characteristics such as income and their adopted children's characteristics as adults.

  7. I believe what you're seeing is basically that IQ is much more strongly heritable than traits which lead to wealth.

    We can argue that IQ is perhaps not a true measure of intelligence. But it does track very well with academic performance overall. Studies have shown that while biological children have IQs very close to their parents, adopted children do not. They also get next-to-no boost to their own likelihood to attend college based upon their parents college attainment.

    While in general higher IQs tend to lead to higher earnings, there are other elements which matter as well. One of the most important is the personality trait of conscientiousness, which basically determines where you fall on the type A/lazy spectrum. Unlike IQ, conscientiousness seems to be mostly not heritable, although it's unclear what causes it (parents don't seem to influence personality in any clear way beyond genetics). IQ and conscientiousness alone mostly determine job performance, but other effects (many unrelated to performance entirely) may determine your ability to actually translate good performance into recognition and increased wages.

    Put simply, if your dad was smart, you'll probably be around as smart. But that's no guaranteeing you a ticket to the middle class.