Investing in Human Capital: Why Does Postsecondary Educational Attainment Lag behind for the Poor?
Theory and Evidence
The theory of human capital investment, as developed by Gary Becker (1996), Jacob Mincer (1974) and others, posits that (all else equal) a rise in labor market returns to any particular skill, or an educational credential that signals the attainment of that skill, should generate higher investments in that skill or credential. So if demand for those with higher education rises in the labor market, and the earnings premium for having a college diploma (relative to high school) increases, more students will enroll in college and obtain that degree. This increase in the supply of college graduates should, in turn, reach a point that it offsets the higher demand and causes the earnings premium to fall to its earlier level.
Of course, this scenario assumes no other complications in the adjustment process, including market failures of any kind, and no other limits on the potential supply of skilled labor. If, for example, there are lags in the time needed for such skill development, then the adjustment process might take many years to complete, and in the presence of imperfect information and foresight among students, the supply of skilled workers over time could potentially overshoot the new equilibrium, causing wages of skilled workers to oscillate, as they have in some markets for highly educated workers (Freeman 1971).
On the other hand, the ability of students to make these additional investments at all might be limited—if, for example, the marginal students in these markets have lower scholastic ability, their information about market returns is incomplete, or they face higher costs of investing in the skills. Indeed, among low-income students, it is quite possible that all of these complications could limit their investment decisions over time. 
If the theoretical responses of investments in skills to market increases in pay premia for those skills are therefore somewhat ambiguous, what does the empirical evidence show? The important and well-known book by Claudia Goldin and Lawrence Katz, The Race between Education and Technology (2008), offers us perhaps the clearest long-term evidence on this issue. They show that, due to technological developments in a variety of industries, the labor market return to high school diplomas rose sharply in the early part of the twentieth century, and in response, the supply of high school graduate labor rose over the first several decades of the century, just as predicted by the human capital model.
Indeed, the process continued until the higher wage premium associated with high school graduation had disappeared by mid-century. Goldin and Katz note that the rise in high school enrollments and graduation reflected not only private investment decisions but also a major public policy response to increase the teaching capacities of public high schools and encourage (or require) more such enrollments. 
In the last few decades of the twentieth century, a similar process occurred in which technological change (plus globalization and other institutional forces) likely increased the demand for college graduates and caused their relative wages to rise as well.  But, unlike the earlier episode, there was relatively little rise in the supply of highly skilled workers until the end of the century. Though Autor (2014) notes that higher enrollments in college finally increased the supply of highly educated labor after the year 2000, and especially after the onset of the Great Recession in 2007, this increase was sufficient only to stabilize the premium associated with college rather than reduce it. 
Furthermore, Bailey and Dynarski (2011) have shown that the response of college enrollments and attainments to the higher college wage premiums of the 1980s and 1990s varied strongly by family income, with higher responses among highincome students than lower-income ones. Accordingly, the gap in B.A. attainments that already existed by family income grew larger over time. Other evidence (e.g., Holzer and Dunlop 2013) also showed rising enrollments in A.A. programs among poorer students and minorities after 2000, while Whites/nonpoor students showed greater increases in B.A. enrollments and attainments, thus contributing to widening earnings gaps as well.
-  This discussion assumes that the market return to a completed degree is at least as high for the disadvantaged as for other students, which appears to be the case (Backes, Holzer, and Velez 2014).
-  Mandatory high school enrollment up to a certain age (usually 16) in most states was a mechanism by which higher high school enrollment was required.
-  College enrollments and supply actually rose substantially in the late 1960s and early 1970s in response to the Vietnam War because college students were deferred from being drafted; this caused the college wage premium to decline substantially in the 1970s (Freeman 1976). But enrollments declined after the war ended, and the positive shift in labor demand for college graduates appears to have begun around 1980. The associated rise in the college premium was not sufficient to dramatically raise the supply of such graduates for the next few decades. Labor economists have long debated the extent to which the rising college premiums of this period mostly reflect labor demand and supply factors (Goldin and Katz 2008); (Autor et al. 2008) or other institutional forces like weaker unions and lower statutory minimum wages (Card and Dinardo 2007).
-  By most accounts, real wages did not rise for college or high school graduates after 2000, only rising for those with graduate degrees beyond the B.A. (e.g., Mishel et al. 2012–2013).