Using new long-run microdata, this article studies wealth and income trends of households with a college degree (college households) and without a college degree (noncollege households) in the United States since 1956. We document the emergence of a substantial college wealth premium since the 1980s, which is considerably larger than the college income premium. Over the past four decades, the wealth of college households has tripled. By contrast, the wealth of noncollege households has barely grown in real terms over the same period. Part of the rising wealth gap can be traced back to systematic portfolio differences between college and noncollege households that give rise to different exposures to asset price changes. Noncollege households have lower exposure to the equity market and have profited much less from the recent surge in the stock market. We also discuss the importance of financial literacy and business ownership for the increase in wealth inequality between college and noncollege households.
It is a well-documented fact that the college wage premium has increased substantially since the 1980s (see, e.g., Levy and Murnane,1992; Katz and Autor, 1999; and Goldin and Katz 2007). This trend can be traced back to differences in the growth of the demand for and the supply of college-educated workers that are driven by skill-biased technical change, socio-demographic factors, and institutional features (Card and Lemieux, 2001, and Fortin, 2006). Recent work has begun to analyze the relationship between college education and wealth inequality (Emmons, Kent, and Ricketts, 2018, and Pfeffer, 2018) and demonstrates an increasing association between college education and wealth.
In this article, we take a long-run perspective on college and noncollege income and wealth over almost the entire post-WWII period. We use a novel household-level dataset, the "SCF+", which combines the post-1983 Survey of Consumer Finances (SCF) with data from historical surveys going back to 1949. To ensure consistent coding of education groups, our analysis starts in 1956. Kuhn, Schularick, and Steins (forthcoming) have harmonized the data across the historical survey waves. The combined data provide long-run household-level information on income, assets, debt, and demographics. The SCF+ closes an important gap, as high-quality microdata were not previously available over longer time horizons. For instance, the Panel Study of Income Dynamics (PSID), which is one of the most important sources of household-level wealth data in the United States, has included questions on family wealth only since 1984 (see Pfeffer et al., 2016).
Our analysis confirms a strong increase in the college income premium since the early 1980s. The average income of college households has increased by about 50 percent in real terms since then. However, the increase of the college income premium is dwarfed by that of the college wealth premium. The wealth of college households has increased by a factor of three between 1983 and 2016, while that of noncollege households has barely grown at all. A substantial part of the widening of the wealth gap between college and noncollege households is driven by strong wealth gains within the top 10 percent of the wealth distribution. Moreover, the share of noncollege households making it to the top 10 percent of the wealth distribution has declined over time. We also document that households with two college-educated spouses have enjoyed particularly large gains in wealth. However, this trend is not driven by assortative matching, but by the overall growth in college education. Consistent with previous findings of Eika, Mogstad, and Zafar (forthcoming), assortative mating appears to have decreased among college graduates over time.
An important question raised by our findings is why the ratio of college to noncollege wealth has grown so much more than the income gap. The workhorse economic models of wealth accumulation imply tight comovement of income and wealth differences, as income is the sole determinant of wealth. We demonstrate that college and noncollege households exhibit systematic differences not only in the size of their asset holdings, but also in the composition of their portfolios. College households own a higher share of stocks and mutual funds. As a consequence, college and noncollege households are differentially exposed to asset price changes. College households reap disproportionately high capital gains during stock market booms. Importantly, such capital gains are unrelated to income. This is consistent with the decoupling of the evolutions of the college income and college wealth premiums since the 1980s (see also Kuhn, Schularick, and Steins, forthcoming). We also find some indication that business ownership matters for the increase in the college wealth premium, especially since the late 1990s. The fact that the college wealth premium is associated with equity holdings and business ownership may be related to higher levels of financial literacy and entrepreneurial skills among college households. We discuss the role of these factors and their potential to affect wealth via portfolio composition and differential returns.
While this article documents a sharply rising college wealth premium in recent decades, it is important to note that causality can run in both directions. College graduates may hold more wealth due to their higher educational attainment, but there is also evidence that it is easier to obtain college degrees when coming from a wealthy family. Wealthy families can afford more investment in their offsprings' educational careers. The cost of college has increased considerably since the 1980s, which constitutes an obstacle for children from poorer households (see, e.g., Haveman and Smeeding, 2006). Beyond providing a basis for inter vivos transfers, wealth may have an insurance function as a "safety net" (Pfeffer, 2018). In this sense, it can work as a "catalyzer," facilitating human capital investment in early life. This will typically lead to higher wealth, which may be augmented by gifts and bequests. The process can accumulate across generations, creating a succession of college-educated households with ever more wealth (Pfeffer and Killewald, 2018). Our SCF+ data consist of repeated cross sections and therefore remain silent on intergenerational wealth links. At the same time, the wealth information in the PSID is less detailed than that in the SCF+, and in particular the coverage of wealth at the top is not comprehensive (Pfeffer et al., 2016). New data sources are needed to address these questions.
Read the full article.