I was asked to write a review of Claudia Goldin and Lawrence F. Katz’s, The Race Between Education and Technology.Here is a long-winded, unedited version of it. Among the stuff that did not make it is that I don’t view it as important to have the world’s most educated public, nor do I view it as important to be a world leader in much of anything – the authors actually think those are important goals. A much shorter and neater version will be available in a few months.
For nearly 50 years the conventional wisdom among labor economists regarding the cause of increasing earnings inequality in the United States was that technical changes in the economy were “skill-biased.” In other words, the advent of computers, information technology, the advanced automation of manufacturing processes and the general deindustrialization that is reshaping developed economies into service economies has resulted in a labor market that favors workers with high levels of skill.
The movement from farm to factory and the subsequent electrification of the factories tended to benefit workers of all skill levels. Recent technical changes have both been occurring rapidly and have not been amenable to the employment of lesser-skilled workers – leading to a growing earnings disparity between those at the top of the earnings distribution and those at the bottom – or so the story goes.
In their impressive work of historical research and analysis, The Race Between Education and Technology, Goldin and Katz (GK) investigate two ideas that are conspicuously absent from the simple narrative above. First, they examine the precise nature of the technical change that has occurred in the United States since the turn of the 20th century in order to better understand how the demand for different types of labor has changed over time. Second, in what should be a lesson to all good economists, they also investigate the supply side of the market before drawing any inferences about how and why the wage structure seems to have changed so dramatically.
Combining a remarkable array of current and historical data with a scouring of the literature on the nature of work, the types of skills required for various jobs, the extent and timing of various technological innovations over time, GK determine that the demand for skill (skill-biased technical change) was not the driver of differences in outcomes for high-skilled workers and low-skilled workers across the 20th century. They argue that both the pace of technical change was steady throughout the 20th century and that the changes earlier in the century (such as the diffusion of electricity) were no more or less biased toward the skilled-labor force than the technical changes occurring today. While the rate of labor demand appears to have been steady, GK point to a relative increase in the supply of skilled and educated workers in the early part of the 20th century as the factor mitigating the tendency for technical change to exacerbate inequality, and a subsequent slowdown in the expansion of the supply of skilled workers since the 1970s as allowing earnings inequality to expand.
In “the race” between the demand and supply of high-skilled labor, the demand has been fast and steady, while the supply raced to an early (and substantial) lead, only to be exhausted and falter on the back stretch. Ultimately, we see that the college earnings premium is the same today as it was in 1915.
There is much to like in this work. The historical analysis of the rise of elementary, secondary and post-secondary education is a must-have for students in any economics of education or history of education course. Their analysis decomposing the skilled-unskilled wage premium into relative supply and relative demand factors is an indispensable tool for any student in labor economics. And their diligence in obtaining original data from multiple sources is an example of how high-quality scholarly work ought to be conducted.
While the supply-demand analysis and the historical presentations are top-notch, there is still much to be concerned with in the book. Some of these concerns are due to factors beyond the authors’ control. For example, GK equate educational attainment with the level of worker skill. Though it is clear that these are highly correlated, the latter is ultimately not observable despite it being a primary object of interest. It is not hard to imagine that the relationship between educational attainment and worker skills has changed throughout the course of the century, thus treating it as so may lead to an under- or over-statement of the authors’ main thesis.
Academic discussions about inequality can be contentious and I do not intend to initiate a food-fight here, but one would expect a book attempting to demonstrate the major cause of inequality to be a bit more careful in its discussion of inequality. Even if one were to concede that individual earnings inequality is the relevant metric of interest (as opposed to household consumption, or political power, or status, etc.) and that there is a particular level of inequality that makes it a public concern and economically significant (for example, Enrico Moretti recently pointed out that half of the returns to obtaining a college degree over a high school degree are “eaten-up” by the higher costs of living for college graduates), the authors should have made that clear, and at a minimum included three additional points.
First, the income distribution is not chosen by individuals, businesses or political leaders. It is the unintended result of the millions of economic transactions that happen each day, wedded with the institutional environment that these operate in. It is much easier to talk about it than it is to do something about it. Furthermore, the measured increase in inequality has not been the outcome of an evil capitalist plot to make the world unequal – but the authors’ interjections about the “inequality reinforcing” work of Ronald Reagan and Margaret Thatcher (p. 329) would seem to indicate otherwise. Second, some amount of inequality is not only desirable, but necessary to preserve incentives. No one knows the correct amount. For example, in physics, equality of temperatures means the heat death of the universe. Gradients are important in all realms of life, and this is no less true in the social sciences as it is in the hard sciences. Third, outside of showing that immigration changes since 1980 have had a minor impact on measured earnings inequality (about 10 percent), the authors virtually ignore the other major factors that are understood to have had an influence as well.
Other inequality enhancing suspects include mobility, household structure and winner-take-all markets. At about the same time as GK show the widening in earnings inequality, you can see a large degree of residential segregation by income. Richer households tend to live in the areas where better schools are located. Is this because richer families choose to live where better schools are? Or is it because richer families pay a premium to live near other richer families – and children from these families tend to have better cognitive abilities than children from other families? If the latter is part of the story, then focusing on school quality and educational attainment as “solutions” for the inequality problem is misplaced. Changes in household size and structure have had a large impact on measured inequality too. Many college educated women have entered the labor force since 1980 and have gone on to marry other professionals. At the same time, birth rates for single-mothers and divorce rates generally have increased (though steadying recently) creating many more single-parent households than there were earlier in the century. Finally, “winner-take-all” markets are increasingly common. These markets produce just a few winners, whose productivity is not much greater than the large number of losers – exacerbating inequality between individuals of otherwise similar skill. This is probably a better explanation for the expansion at the top end of the income distribution than a labor supply slowdown. Yet GK are silent on these topics.
Goldin and Katz seem to understand that their supply and demand framework only takes them so far in explaining changes in inequality. Therefore they mention several institutional factors that work to enhance inequality: a decline in the real power of the minimum wage, the decline of unions, the decreased progressivity of the tax code, and the decline of other institutions to protect workers. In contrast to the rigor with which they subject their supply-demand hypothesis, these institutional factors are presented without proof. For example, while marginal tax rates seem to be less progressive than in the 1970s, the actual payment of federal taxes is more progressive now than at any point in U.S. history (particularly since over 40 percent of eligible taxpayers pay zero federal taxes). Furthermore, it is puzzling to examine what institutional factors receive nary a footnote. For example, the decline in educational attainment and the increase in inequality in the U.S. comes on the heels of President Johnson’s “Great Society” programs. These declines also come during the Nixon era price controls (and WWII wage compression occurred under a regime of wage controls). And while the authors point out that the steep decline in unionization over the past 30 years has led to less protection for unskilled workers, they neglect to mention that the percentage of the workforce required to hold an occupational license from a government agency has increased virtually one-for-one with the decline in unionization. These special privileges serve to protect a class of workers over 2.5 times larger than the current unionized workforce and the minimum wage combined. It should at least be a matter of interest to ask whether these workers are of lesser, the same or higher skill levels than the previously covered union workforce – and what impacts this might have on inequality and equality of opportunity for the lesser-skilled.
The GK labor slowdown story is still very important, and should not be discounted because of the criticism above. However, I am not sure how far their story gets us in terms of understanding the role of education in explaining inequality increases. To illustrate, economists at the OECD put together a chart of earnings inequality changes since 1979 versus changes in educational attainment over the same period for the OECD countries – and when you remove the U.S. from the data, there is simply no relationship between the two.
In their final analysis, GK conclude that an alarming decline in educational attainment is the real culprit behind the U.S. earnings inequality increase since 1980. Even if you accept this conclusion fully, proceed with caution along their preferred policy paths. First, GK wish to increase early childhood interventions – particularly with access to quality pre-school. If differences in cognitive skills in adults derive from early childhood experience, then this proposal would have some traction, but if they arise from differences in cognitive skills from genetic endowments, then there is little such programs can do. In fact, there is little evidence that post-childhood education can reduce differences in cognitive skills that come from either genetics or environmental factors. Furthermore, it is one thing to ask for a quality pre-K education and quite another to deliver it. The families that would require the best pre-K educations are likely to be those suffering in the poorest quality K-12 school districts. I am not confident these same districts would be able to deliver a quality pre-K program or be immune from the destructive public choice problems that plague traditional school systems, but I would love to be proven wrong.
Second, GK recommend that we rekindle the virtues of American education (such as decentralization) and improving the operation of K12 systems so that students graduate from high school better prepared for college. This will be increasingly difficult in an era when for the first time in U.S. history federal tax collections exceeded the take from state and local governments, and when there are increasing calls for oversight at the federal level. And hopes for an improved K12 system will be dashed so long as the lack of educational entrepreneurship continues, the political class refuses to recognize the success of various school choice programs (such as the elimination of the successful DC voucher program), and an honest assessment of the shocking lack of educational productivity is acknowledged and dealt with.
The authors’ final recommendation is one that has been echoed for as long as I can remember in higher education – make financial aid sufficiently generous and transparent so that students who are college ready can complete a 4-year degree or gain skills at a community college. This recommendation raises two major concerns. First, it assumes that attending college, any college, enhances productivity and is not merely serving a signaling function. To the extent that institutions have arisen to limit the ability for employers to screen the skills of prospective workers, then the possibility that colleges serve merely as expensive signals increases. One such institution worthy of examination was the Supreme Court decision in the 1971 Griggs v. Duke Power case that severely limited the screening that employers could do – thereby decreasing the chances of high-skilled but credit constrained, information constrained, non-college students to secure higher paying jobs.
The second concern is that while there is economic evidence that high-ability, low-income students are under-represented in elite colleges and universities, it is simply no longer credible to argue with such urgency that students are not aware of financial aid opportunities, that colleges are generally out of reach for students of all income classes, or that credit constraints pose a significant problem in educational markets. There may be small pockets of students that this is true for, but on the whole consider that the average net-price of attending community colleges is only one-hundred dollars per year. Clearly community colleges are within reach of every American right now – and they are virtually in every sizeable community as well. And in terms of the generosity of financial aid, in the past decade alone, federal grant aid has increased 79 percent in real terms to $21 billion, federal loans by 70 percent to $67 billion, institutional grants increased by 78 percent to $29 billion, private loans by 536 percent to $19 billion and state grant aid increased by 78 percent (with 59% increase in need-based aid). In total, real student aid increased by 101 percent from 1998 to 2008, from $80 billion to $162 billion. In addition the average student debt for students getting aid from public-four year colleges is only $18,800. This is far less than the cost of most cars, and a pittance when one considers that the returns to receiving a college degree exceed 10 percent per year over the average wage of a high school graduate. Couple this evidence with the fact that the portion of family budgets, for families throughout the income distribution, dedicated to the necessities of food, shelter and clothing is lower now than at any time in U.S. history, and it is simply not credible to argue that college is out of reach for many people – particularly when well less than half the population does not graduate from college.
Ultimately, this superbly written and deeply researched book had the temerity to tackle an extremely important and difficult (if not unanswerable) question. Despite my quarrels with the authors’ policy recommendations and oversight of some institutional factors affecting inequality and educational attainment, the book is a must-read for anyone interested in understanding the relationship between human capital accumulation, economic growth, and inequality.
Michael J. Rizzo
Lecturer of Economics
University of Rochester