The Goldin-Katz Swindle
By Arnold Kling
On p. 96 of their new book, Claudia Goldin and Lawrence F. Katz provide a table on the composition of the U.S. labor force in terms of educational background, based on historical census data. The percentage of high school graduates and above is as follows:
1960 50.5 %
1980 79.3 %
2000 91.3 %
As you can see, the percentage of high school graduates rose much more rapidly from 1960-1980 (28.8 percentage points) than from 1980-2000 (12.0 percentage points). By this measure, the rate of improvement in human capital in the United States stalled after 1980. Because of this, as Goldin and Katz point out, in the “race” between the growth of skill-biased technology and the growth of education in the work force, the latter has slowed down. They argue, convincingly, that this slowdown has caused skill premiums to rise, resulting in more income inequality.
What went wrong?
Well, for one thing, it is arithmetically impossible for the graduation rate to rise 28.8 percentage points from a level of 79.3 percent. On p. 325, Goldin and Katz write, “An upper bound exists for a graduate rate; it cannot exceed one.”
So why did they write their book? Mostly, it is a swindle.Instead of focusing attention, as I just did, on the high-school graduation rate, they emphasize an estimate of average total number of years of schooling, which they call “educational attainment.” On p. 19, they write,
After increasing nonstop for the first three quarters of the twentieth century, educational attainment among the native-born population slowed considerably during the last quarter century. The educational attainment of a child born in 1975 was just 0.50 years more than that of his or her parents born in 1951, but the education attainment of a child born in 1945 was 2.18 years more that that of his or her parents born in 1921.
The swindle is that in order for the average years of schooling to increase at the same rate, we would have had to increase the college attendance and graduation rates in recent years as rapidly as we increased high school graduation rates from 1900 through 1960.
Goldin and Katz recognize this (they do understand arithmetic), but it leaves the question of whether it is possible to dramatically increase the college attendance and graduation rates. They say that it is possible, and so they would make it a huge policy priority to try to get more young people into and through college. But at this point, they stop using their own data and start relying on meager other sources, or none at all.
On p. 336, they write,
One possibility is that young people might not actually benefit from going to college. The rate of return we have estimated may not be applicable to some young people who do not currently attend or complete college…But that possibility appears not to be the case.
…carefully executed studies using plausibly exogenous variation in educational attainment find high rates of return to further schooling.
The only backup for this assertion is the following footnote:
Card (1999, 2001) provides critical and comprehensive reviews of recent studies using such “natural experiments” to estimate the returns to schooling. More recently, Oreopolous (2007) finds that post-1970 changes in state compulsory schooling laws to raise the minimum school-leaving age above 16 years modestly but statistically significantly raise the educational attainment of affected cohorts and generate substantial labor market returns to such increased schooling.
In other words, Goldin and Katz are suggesting that the issue of marginal vs. average return to college has been settled by a study of compulsory schooling laws for 16-year-olds.
On p. 348, they write,
The K-12 system is less than perfect for many students, but it is important to recognize that schools are essentially failing particular students. Those left behind by the system are mainly minority children in inner-city schools who become youths who are not college ready.
The implication is that everyone aged 18 other than inner-city blacks is college ready. They cite no sources to support this implication, and I beg to differ. The students I taught at George Mason were not college ready, and they were not inner-city blacks, either.
I give Goldin and Katz an A+ for their contribution to positive economics, meaning the analysis of data. The compilation of statistics on education by cohort is extremely valuable. Their assessment that the increase in the educational wage premium reflects a slowdown in supply growth in addition to continued demand growth is both original and well-supported by their research.
However, as you can tell, I would give their normative economics (policy recommendations) a much lower grade. The level of empirical support is as flimsy in that chapter as it is firm in the positive chapters.