Bryan Caplan  

Goldin-Katz and the Education Plateau

Tear Down These Walls... Retired U.S. Admiral: Iranians...

Education has not plateaued.  Does this mean that Goldin and Katz's The Race Between Education and Technology is wrong?  No.  Goldin-Katz never claim that the stock of education has stagnated.*  When Goldin-Katz talk about a "standstill" in educational attainment, they mean a standstill of flow, not a standstill of stock:

For cohorts born from the 1870s to about 1950, every decade was accompanied by an increase of about 0.8 years of education.  During that 80-year period the vast majority of parents had children whose educational attainment greatly exceeded theirs.  Educational change between the generations then came to an abrupt standstill.

But if stock is the right measure of labor supply, how can Goldin-Katz explain the sharp increase in the college premium?  Their story is that the college premium rose because educational attainment is rising at a slower rate than it used to.  Katz sent me a figure to illustrate:


Goldin-Katz aren't explaining a big change (the large rise in the college premium) with another big change (stagnant educational attainment).  They're explaining a big change (the large rise in the college premium) with a small change (a decline in the rate of growth of educational attainment).  If the stock of education had actually stagnated, Goldin-Katz's model predicts that the college premium would have risen far more than it did.  That's internally consistent, but I think it strains credulity. 

Last point: Using the Goldin-Katz measure, even straight-line progress implies relative stagnation.  Imagine a society where the fraction of the population with college degrees starts at 1%, and rises by .5% per year.  Here's educational attainment over the course of 50 years:


Here's what log relative supply will look like over the course of 50 years:

To win Goldin-Katz's "race between education and technology," education attainment must permanently exceed straight-line increase.  Perhaps I'm biased, but that strikes me as completely unrealistic goal.  If that's our choice, we might as well just let technology win.

* Goldin and Katz kindly confirmed this over email.

Comments and Sharing

COMMENTS (9 to date)
RPLong writes:

Is it possible these graphs are underestimating the true amount of schooling occurring these days? Out here in the workforce, many of us have to seek continuing education to build our skills. I'm talking about professional certificates (such as a CFA, PMP, CPM, etc.), courses in programming various computer languages, foreign language certification, and so on.

I have to keep taking such courses to keep myself marketable in the workforce. Is such schooling included in the analysis?

ajb writes:

This is a really excellent post of Bryan's. Coupled with his last on plateaus, they really point to a weakness in the Goldin-Katz story.

@RPLong. If the graphs underestimate extra schooling then Bryan's point holds even more strongly.

Jonathan writes:

Check out Autor & Acemoglu's 2011 Handbook of Labor Economics article

The first figure in this blog post is Figure 3, and the fact that the Katz-Murphy (1992) estimates overpredict the college premium since 1992 is shown in Figure 19.

MG writes:

What do Goldin-Katz say about changes in the demand for education?

Costard writes:

2 points:

1. 1890-1970 (when these cohorts entered the workforce) most likely saw greater changes in professional educational requirements - ie prerequisites for the bar, for a medical license, etc. - than did 1980-present. Perhaps the growth in education-time was a sign of progress, as the authors seem to think; but it could also have been a result of the increasingly important (legal) role played by professional societies, unions and, for lack of a better word, guilds. Increasing educational requirements is probably the easiest way to restrict entry into a given field.

2. The college premium is quite possibly a *product* of the increase in education-time. The labor pool becomes less homogenous as education increases. Specialized fields of study become not only encouraged, but oftentimes required for a specific job. The result is that candidates are either qualified or they are not. In good times the demand for them is high and they are paid well; whereas in bad times they are simply laid off. In other words, changes in demand for these highly-specialized workers (especially in highly-cyclical industries like O&G) manifests itself in employment rather than wages.

Steven Kopits writes:

Education is not a goal in and of itself. It is investment, and not either production or consumption. So all things being equal, we'd like to either avoid the need for formal education (eg, apprenticeship), reduce the time required to educate someone, or increase the efficacy of that education.

Increasing education for its own sake is not an appropriate policy goal.

Mark Little writes:

@Costard, you seems to be thinking along lines similar to my thinking.

As to your point 1, I agree but would put the case less tentatively. The summary tag I use is the white collar union card. But while the "guilds" (or occupational licensure) play a part, the major factor in the US since the 1970s is a legal regime that provides a monopoly to academic credentialing, by effectively forbidding the use of alternative signals.

As to your point 2, note that this relates to the recent debate on this blog between Arnold Kling and Bryan Caplan regarding PSST recalculation versus wage stickiness as an explanation for unemployment in the recent economic unpleasantness.

More generally, I think it is far too little recognized that in the long run economic progress and technological advancement take place through a greater articulation of the division of labor. It is fashionable to speak vaguely of "high tech" and of "human capital", but at bottom it is really just Smith's pin factory writ very large. Growth comes from exploiting new opportunities for specialization, not from increasing educational attainment.

Les Cargill writes:

Isn't education a technology?

Mark Little writes:


I assume your post re Goldin-Katz is intended in support of the signaling model alternative. But how does the signaling model explain the sharp increase in the college premium?

I don't recall you discussing this in previous posts, and your Brief Letter on Signaling summarizing your position doesn't seem to mention it.

Unless the signaling model can explain both the huge level of waste in higher education investment and, especially, the changes in the college premium, then I don't see how that model can do much to strengthen The Case Against Education.

Would you please post on this question?


Comments for this entry have been closed
Return to top