The anemic response of skill investment to skill premium growth

Joseph Altonji, Prashant Bharadwaj, Fabian Lange

06 May 2008



Since 1980, the demand for skilled labour has risen faster than the supply of skills, fuelling a steady increase in the earnings premia found for measures of skills such as schooling or cognitive test scores.1 The rapid rise in the skill premium represents a substantial increase in the economic incentive to acquire skills. For example, Heckman, Lochner, and Todd (2008) show that between 1980 and 2000 the internal rate of return for completing high school rather than dropping out after tenth grade has increased from approximately 40% to 55%. Standard economic theory suggests that such an increase in the skill premium should induce young adults to invest more in their skills and supply more skills to the labour market.

How rapidly and how much young adults respond to this increase in the returns to skills and how this response varies across the population have important implications for the development of the US economy. Young adults’ investment in their skills determines directly how much the US economy will be able to benefit from ongoing technological progress. And, whether earnings inequality within and between groups will decline during the next few decades depends, to a large extent, on how those who traditionally have acquired few labour market skills respond to the increase in the skill premium.

The supply of skill

In Altonji, Bharadwaj, and Lange (2008), we examine how the skills of young American adults have changed since 1980. We compare various skill measures of participants in the NLSY-1979 panel survey and participants in the NLSY-1997 panel survey at age 22.2 We consider the standard skill measures of completed schooling and cognitive test scores, but we also look at factors that influence skill acquisition, such as parental education and growing up in a two-parent family. We also make use of measures of the ease with which young adults transition from schooling into the labour market. Because both surveys follow individuals over time, we can use the wages between 1998 and 2004 of those who were young in 1979 to aggregate the different skills indicators into a single wage based index of skills. Figure 1 shows how the supply of skills of young adults has changed between 1980 and 2004 across the distribution of skills.

The figure shows that overall the 1997 youth cohort is more skilled than the 1979 cohort. For instance, the amount of skill of an individual at the median of the skill distribution has increased by about 6.5 percent.

Is the above increase largely a behavioural response by youth to the widening skill premium? The answer is no. We find that much of the increase in skills observed between the 1979 and 1997 cohorts can be attributed to the fact that members of the more recent cohort have significantly more educated parents than young people in 1979. About two-thirds of the increase in skills documented in figure 1 is due to the increase in parental education. Holding parental education, race and gender, and family structure constant, the supply response to the increase in skill premia between cohorts was small: about 1% on average and about 1.5% at the median. Thus, if parental education, race, gender, and family structure are largely exogenous with respect to the skill premium, then our results indicate that the supply response to the increase in the skill premium during the last few decades has been surprisingly small.

This small increase in the supply of skills conditional on exogenous factors contrasts with the large increase in the skill premium alluded to above. Autor and Katz (1999) report that annual wage growth of skilled labour exceeded wage growth of unskilled labour by about 1.51% in the 1980s and about 0.4% annually between 1990 and 1996. This implies that between 1980 and 1996 the ratio of skilled to unskilled wages increased by about 20%. It seems that very large increases in skill premia are necessary to induce young workers to increase their investments in skills substantially.

Our finding that the supply of skills has increased surprisingly little across the two cohorts is consistent with empirical findings reported by other authors. For instance, Heckman and Lafontaine (2008) find that at the same time that the internal rate of return for high school graduation increased dramatically between 1980 and 2004, the high school graduation rate declined by about four to five percentage points. Similarly, the college graduation rate has risen by only a couple of percentage points over this time period, and in our data we find that cognitive test scores have likewise only increased by small amounts. Overall, the empirical picture that emerges is that the increase in the skill premia was not accompanied by a commensurate supply response. This implies that, all else equal, the large degree of earnings inequality observed today is likely to persist far into the 21st century.

Inequality in skills is widening

If today’s earnings differentials across skill types persist during the next decades, then inequality in earnings in the population will only decline if the dispersion of skills itself declines. That is, even if the skill premium paid in the labour market remains constant, the dispersion in earnings could fall if the skills of those toward the bottom of the skill distribution rose by more than the skills of those toward the top of the skill distribution.

Unfortunately, the empirical evidence summarized in Figure 1 shows that the difference in the skills of the 1980 and 2004 youth cohorts is larger at the top of the skill distribution than at the bottom. Again, we find that much of the increase in dispersion of skills is due to the changing distribution of parental education across cohorts rather than a supply response to skill prices. Overall, it appears that in the years ahead the changing distribution of skills in the population will exacerbate rather than counteract the trend towards increasing earnings disparities. For instance, all else equal, our findings indicate that increased dispersion in skills will raise the inequality that today’s young adults experience later in life by an additional five percent.

Why is the supply of skills so inelastic?

At this point we can only speculate as to why the response in skills to the increase in skill premia is so small. It is possible that for many young adults the economic incentives for acquiring skills are less important than the non-pecuniary costs of skill investments. For these young adults, a change in the economic incentives to invest into skills might therefore induce only a small supply response. Another plausible explanation is that young adults are liquidity constrained or myopic in their investment decisions for other reasons. If this is the case, they might forego valuable investment opportunities in order to protect their consumption while young. This explanation would, for instance, be consistent with a number of studies (e.g. Kane (1994), Dynarski (2003)) that find that schooling decisions are quite sensitive to direct costs of schooling and tuition subsidies. The importance of liquidity constraints is controversial, however. Cameron and Taber (2004) is one of a number of studies that finds that liquidity constraints are not important for explaining schooling decisions.

Research summarized in Cunha and Heckman (2007) suggests that part of the explanation might be that parental investment during early childhood shapes the potential to acquire additional skills later in life. Parents might not have responded to the increase in labour market returns, perhaps because they were not fully aware of the large increase in the returns to skills or because their children’s labour market success might not be the primary motivating factor in determining the time and resources they devote to their children.

At this point, the question of why the supply response to the increase in the labour market returns to skill has been so small is an open one. In our opinion, it ranks among the most important empirical issues facing labour economists today.


Altonji, Joseph G., Prashant Bharadwaj, and Fabian Lange (2008), “Changes in the Characteristics of American Youth: Implications for Adult Outcomes”, NBER Working Paper 13883.
Autor, David H., Lawrence F. Katz, and Melissa S. Kearney (2005), “"Trends in U.S. Wage Inequality: Re-Assessing the Revisionists”, NBER Technical Working Paper 11627.
Cameron, Stephen and Christopher R. Taber, “Borrowing Constraints and the Returns to Schooling,” Journal of Political Economy, Vol 112, February 2004.
Cunha, Flavio and James J. Heckman, “The Technology of Skill Formation”, IZA DP No. 2550.
Dynarski, Susan (2003) “Does Aid Matter? Measuring the Effect of Student Aid on College Attendance and Completion.” American Economic Review Vol. 93, 1 pp. 279-288.
Heckman, James J. and Paul A. LaFontaine (2007), “The American high school graduation rate: trends and levels.” NBER Working Paper 13670.
Heckman, James J., Lance J. Lochner, and Petra E. Todd (2008), “Earnings functions and rates of return.” Journal of Human Capital, vol. 1, 1 pp. 1-31.
Kane, Thomas (1994) “College Entry by Blacks since 1970: The Role of College Costs, Family Background, and the Returns to Education”, Journal of Political Economy Vol. 102, 5 pp . 878-911.
Katz, Lawrence and Kevin Murphy (1992), “Changes in Relative Wages, 1963-1987: Supply and Demand Factors”, Quarterly Journal of Economics 107 (February): 35-78.




1 See Katz and Murphy (1992), Autor and Katz (1999), and Autor, Katz and Kearney (2005).
2 NLSY-1979 tracks individuals who were 14 to 22 years old in 1979. NLSY-97 tracks individuals who were 12 to 16 in 1997.





Topics:  Labour markets

Tags:  Inequality, skill premium

Thomas DeWitt Cuyler Professor of Economics, Yale University

Assistant Professor in the Department of Economics, University of California San Diego

Assistant Professor of Economics, Yale University