Key Points
- The large premium that college degree holders earn relative to workers with only a high school diploma suggests that a better-educated workforce would increase U.S. output.
- Barriers to borrowing against future income, though, may make it difficult to acquire a college education, implying a potential role for using policy to increase access to college, especially if it is appropriately targeted.
- However, college education is costly, and the payoff is uncertain and realized only after a lengthy absence from the workforce. Optimal policy, therefore, aims to balance these costs against the potential benefits, requiring the explicit modeling of education attainment when making budget projections.
Education and Income Growth
Introduction
It is generally well accepted that college pays: In 2012, college graduates’ wages exceeded those of high school graduates by 77 percent, on average.1 This college wage premium has been increasing since 1980 when it was about 40 percent.2
An old debate on the value of college education and on the government’s role in subsidizing it has now reemerged.3 Concerns about “overeducation” were raised during the late 1970s and early 1980s as more people began to attend college. For example, one study argues that a mismatch between job skills and too much education causes workers to be less productive. However, notwithstanding such concerns, the average college wage premium grew during the 1990s.4
Total compensation — including wages and fringe benefits (such as health insurance and pensions) — reflect a worker’s productivity. So, the higher compensation for college educated workers could mean that educating more workers would increase the Gross Domestic Product (GDP). But the benefits of college accrue slowly in the future whereas going to college reduces wages today. Therefore, it is important to calculate the expected present value of future income from a college education.
Many potential students have a large present value of future earnings relative to the direct tuition costs of attending college. If they are unable to borrow against their future income to pay for tuition, public policies that enable those individuals to attend college would likely increase GDP. Nonetheless, care must be taken to correctly assess the actual value of promoting additional college education.
What is Education?
From an economic standpoint, education is the accumulation of “human capital.” Human capital includes the skills to work with physical capital (plant and equipment), and to organize and manage productive activity, including the use of new tools, technology, and organizational methods to increase output per hour worked. A worker with more skills and expertise produces more output than a similar worker with less of such human capital, all else equal.
A long standing debate, however, is whether more schooling actually imparts such abilities. Some academic studies claim that schooling is largely just “signaling.” Under this view, a college degree does not materially add to a worker’s productive capability but merely signals the existence of such intrinsic ability.5 For example, billionaire tech entrepreneur Peter Thiel has offered fellowships for certain students to delay college and instead work at technology companies.6
Indeed, studies show that 45 percent of college students made no significant improvements in their critical thinking, reasoning, and writing skills during the first two years of college. After four years, 36 percent showed no significant gains in those skills.7 However, this evidence is also inconsistent with the idea that college delivers zero improvements to human capital. Generic tests of critical thinking, reasoning, and writing skills might not capture and accurately measure all specialized skills learned in college.
Nonetheless, diminishing returns appear to be setting in as the U.S. population seeks more college education. One study finds that increases in college enrollment have led to a decline in the average quality of college graduates between 1960 and 2000, resulting in a decrease of 6 percentage points in the college premium.8 Another study finds that hours spent studying have decreased over the past half century from an average of 40 hours to just 27 hours per week.9
Historical Change in College Wage Premium
Average compensation, adjusted for inflation, has been flat for full-time workers without a college degree since 1990. Figure 1 shows that the inflation-adjusted average wage plus employer-paid health insurance of workers with at least a college degree grew through the early 2000s, but has plateaued since then. At first glance, periods of flat real labor compensation growth are puzzling in the presence of ongoing productivity growth from technological change. However, our recent brief on capital intensity shows that a decline in capital per worker during the last couple decades has offset the gains in other sources of productivity improvements.
Figure 1: Average Wage Plus Employer-Paid Health Insurance for Full-Time Workers Aged 40 to 55 in 1995 Dollars
Health insurance costs provided by data source are imputed.
Source: PWBM calculations based on the Current Population Survey.
Looking at wage and employer-paid health insurance data from the March Current Population Survey (CPS), Table 1 reports the average annual increase in the average total compensation of different birth-cohorts of workers as they age. However, we first divide this compensation by the average compensation of workers with only a high school education who were born in the same year and demographic category. Therefore, the values shown for different education (e) classes track the yearly growth rate in the compensation premium from going to college. Results from the 1974 birth cohort have only about 10 years of earnings, and therefore, those results might not be as representative. Nonetheless, having at least a college education has produced significantly higher yearly growth in the total compensation premium, and the growth rate itself has increased over time.
Table 1. Trend-Line Yearly Increase in the Average Compensation Premium (Average Compensation Divided by Average High School Compensation) for Each Birth Cohort from 1988 to 2012.
Sample restricted to full-time workers aged 30 to 60 (35 to 60 for e>College). Compensation consists of wages and employer-paid health insurance (imputed by data source).
| Male (year of birth) | ||||||
|---|---|---|---|---|---|---|
| 1949 | 1954 | 1959 | 1964 | 1969 | 1974 | |
| e<College | -0.63% | -0.02% | 0.06% | -0.48% | 0.48% | 1.51% |
| e=College | 1.59% | 1.49% | 2.20% | 1.34% | 2.35% | 0.33% |
| e>College | 3.90% | 4.48% | 2.23% | 2.43% | 5.05% | -2.64% | Female (year of birth) |
| 1949 | 1954 | 1959 | 1964 | 1969 | 1974 | |
| e<College | -0.23% | 0.50% | 0.25% | 0.27% | 1.28% | 1.36% |
| e=College | 1.67% | 0.69% | 1.15% | 0.52% | 0.56% | 0.32% |
| e>College | 2.02% | 3.99% | 1.05% | 1.78% | 5.65% | 5.42% |
Source: PWBM calculations based on the Current Population Survey.
The U.S. workforce has been attaining higher levels of education. Figure 2 shows that the portion of the U.S. workforce with a high school or less than high school education has been decreasing over the last two decades. Those with some college education without a degree has remained steady. The college graduate population has increased 6 percentage points (36 percent higher). The fraction of the population with more than a college degree has had the fastest growth, increasing from 9.7 percent of the workforce to 14.1 percent (46 percent higher).
Figure 2: Percent of U.S. Workforce Aged 30 to 50 by Education Category Over the Period 1991-2012
Source: PWBM calculations based on the Current Population Survey.
The Present Value of College
The value of college depends on a number of factors, including ability for job-market work, the cost of college, and the chosen field of study. For example, in 2009, median lifetime earnings for a college graduate working in the STEM (science, technology, engineering, mathematics) sector exceeded $3 million versus $1.2 million for a peer in the health-support sector.10
Nonetheless, averaging across all fields of study, an 18 year-old today can expect to earn considerably more in total compensation between ages 18 to 65 by graduating from college. Using the annual March Current Population Survey (CPS), we can calculate total compensation, which is equal to total wages plus employer-paid health insurance benefits. A male attending college can expect to earn an additional $2.32 million, and a female can expect $1.96 million. These values project future earnings from historical trends and account for lost wages while in college.11
However, this simple comparison exaggerates the value of college. The gains in compensation to college graduates often accrue slowly in the future, whereas the lost wages while in college happen sooner. A valid comparison, therefore, requires computing the present value of future income earned by an average college graduate. We then compare this amount against the present value of future income earned on average by someone who graduates only from high school. We call this difference the present value of the college premium.
Table 2 shows the present value of the college premium under different interest rate assumptions. The nominal interest rate alternatives (as shown in the top row of the table) are used to discount the value of future compensation to place them on par with 2012 dollars. In other words, if the present value premium were received as a lump-sum today and invested at the stated interest rate, it would produce the same amount of lifetime resources as the future salary earned by a college graduate.
Table 2: Present Value of the College Premium in Thousands of 2012 Dollars for Different Interest Rates
| 2% | 4% | 6% | 8% | 10% | |
|---|---|---|---|---|---|
| Male | $1,261 | $714 | $418 | $249 | $149 |
| Female | $1,050 | $591 | $347 | $210 | $129 |
Source: PWBM calculations based on the Current Population Survey.
Notice that the interest rate plays a major role in determining the present value of the college premium. Using a low discount rate, such as 2 percent, implies that the additional compensation from going to college is almost guaranteed. Under that assumption, a male is expected to earn a net premium of $1,261,000 while a female is expected to earn $1,050,000. In contrast, a high interest rate, such as 10 percent, implies that future wages are almost as risky as the stock market. The value of college drops sharply to $149,000 for males and $129,000 for females. Even these calculations are somewhat conservative because these do not include the larger future taxes that will be paid by college graduates, as the U.S. tax system is progressive. Allowing for taxes reduces the college premium by another 15 percent or so.
Determining the correct interest rate to use is difficult because future labor income is not traded on financial markets as an asset, and therefore its market rate of return cannot be observed. However, the historical data shows that the additional compensation to college is not risk free. The standard deviation of changes in compensation over successive age cohorts in 2012 is about 12 percent for high school educated males, 10 percent for college educated males, 9 percent for high school educated females, and 13 percent for college educated females. The standard deviation of changes by age cohort in the expected college premium is 25 percent for males and 24 percent for females.
However, this volatility is measured for the average premium across all workers in a cohort and not for the actual incomes of individual workers. Since workers cannot easily diversify their individual labor income across multiple occupations, the associated risk at the individual level is likely to be much higher.
In contrast, the annual historical standard deviation of broad stock indices is about 20 percent annualized. To compensate for this risk, stocks, on average produce an average annual return between 8 and 10 percent. Hence, a discount rate between 6 to 8 percent on future wages is not unreasonable.12 In that case, the present value of the college premium is between $249,000 - $418,000 for males and $210,000 - $347,000 for females.
However, college costs more than just the indirect lost wages while in college. Tuition and other college-related expenses directly reduce the value of a college education even more. The direct cost of college depends on the choice of school and type of financial aid. A New York Federal Reserve Bank study in 2014 put the average net tuition for a bachelor’s degree at $6,500 per year in 2013 and showed that government tax incentives reduces the average cost-per-student closer to zero.13 To be sure, some colleges cost substantially more. But graduates of these colleges typically make a larger present value premium than the average amounts reported above.
Other Factors
There might be other potential value from attending college that is not completely captured in the financial data reported earlier. For example, the marriage rate is only 52 percent for those with less than a high school education and 59 percent for those only with a high school education. But they increase to 60 percent for those with some college, to 70 percent for those with a college degree, and to 72 percent for those with more than a college degree.14 Other “non-pecuniary” benefits might also include a longer life-span, lower divorce rates, lower teenage birth rates, and other factors.
However, the college premium might also be overstated for policy purposes. The simple reason is selection bias: People with higher ability are more likely to attend college than those with lower ability. As a result, trying to push lower-skilled individuals into college would likely produce a smaller compensation premium for them than shown earlier. Moreover, even the “non-pecuniary” benefits noted might not be earned by them if a college degree does not produce a greater income.
Conclusion
On average, a college education has a positive and persistent benefit to the earnings of workers. For this reason, the Penn-Wharton Budget Model explicitly incorporates changes in educational attainment over time by various demographic groups. That tracking allows us to accurately estimate changes in labor effectiveness associated with different government policies. Standard government forecasting, in contrast, does not take this factor into account, potentially leading to bias.
However, considerable care must be taken when evaluating policies that encourage more education. To be sure, simple numerical comparisons suggest a large financial advantage of attending college. But those inter-education-group average return differences likely overstate the actual benefit of encouraging additional people to attend college. In contrast, federal government policy actively supports marginal educational institutions even as the workforce outcomes of their students deteriorate.15 Merit-based awards and grants targeted to STEM majors are more likely to realize positive returns for both prospective students and the wider U.S. economy.
-
Data from the 2012 Current Population Survey (CPS) for wage earnings of all full-time workers between ages 30-50. Accounting for employer health insurance brings the premium to 74%. ↩
-
Claudia Goldin and Lawrence F. Katz, The Race between Education and Technology (Cambridge: Harvard University Press, 2008). ↩
-
Megan McArdle, “What We’re Buying With $1 Trillion in Student Loans,” BloombergView, November 2, 2015, available at: http://www.bloombergview.com/articles/2015-11-02/what-we-re-buying-with-1-trillion-in-student-loans> ↩
-
Mun C. Tsang and Henry M. Levin, “The Economics of Overeducation,” Economics of Education Review 4, no. 2 (1985): 93–104, available at https://www.researchgate.net/publication/4837615_The_Economics_of_Overeducation ↩
-
Michael Spence, “Job Market Signaling,” The Quarterly Journal of Economics 87, no. 3 (1973): 355–374, available at: http://www.jstor.org/stable/1882010. ↩
-
Ben Wieder, “Thiel Fellowship Pays 24 Talented Students $100,000 Not to Attend College,” Chronicle, May 25, 2011, available at: http://chronicle.com/article/Thiel-Fellowship-Pays-24/127622/ ↩
-
Richard Arum and Josipa Roksa, Academically Adrift: Limited Learning on College Campuses (Chicago: University of Chicago Press, 2011). ↩
-
Pedro Carneiro and Sokbae Lee, “Trends in Quality-adjusted Skill Premia in the United States, 1960—2000,” The American Economic Review 101, no. 6 (October 2011): 2309–2349, available at: http://www.jstor.org/stable/23045644. ↩
-
Philip Babcock and Mindy Marks, “The Falling Time Cost of College: Evidence from Half a Century of Time Use Data,” Review of Economics and Statistics 93, no. 2 (2010): 468-78, available at: http://www.mitpressjournals.org/doi/pdf/10.1162/REST_a_00093 ↩
-
Philip Oreopoulos and Uros Petronijevic, “Making college worth it: A review of research on the returns to higher education,” National Bureau of Economic Research, Working Paper No. 19053, May 2013, available at: http://www.nber.org/papers/w19053.pdf. ↩
-
Part of the male-female total income difference may be due to the differential selection of fields of study and work—it is also well known that technical fields pay higher wages and are male-dominated. We use the historical growth rate of average earnings from 1994 to 2011 by gender and education. ↩
-
Technically, the standard deviation alone does not tell the entire story. What also matters is the correlation of these risks with the rest of the economy. Since wages are hard to diversify, however, the appropriate discount rate also recognizes the value of the so-called idiosyncratic component that we normally would not weight if it were associated with stock risks. ↩
-
Jaison R. Abel and Richard Deitz, "Do the Benefits of College Still Outweigh the Costs?" Federal Reserve Bank of New York 20, no. 3 (2014). Available at: https://www.newyorkfed.org/medialibrary/media/research/current_issues/ci20-3.pdf ↩
-
Education attainment for individuals between the ages of 30-65 years are based on data from the 2013 Current Population Survey. ↩
-
Andrea Fuller and Josh Mitchell, “U.S. Helps Shaky Colleges Cope With Bad Student Loans,” Wall Street Journal, December 22, 2015, available at: http://www.wsj.com/articles/u-s-helps-shaky-colleges-cope-with-bad-student-loans-1450752462 ↩
,Less than high school,High school diploma,Some college,College graduate,Postgraduate 1988,26249.26432,32953.84556,39761.82239,50483.38877,57894.34121 1989,24875.01945,32809.42629,39492.39842,51001.46552,57336.41371 1990,24352.39046,31528.44262,38179.40222,49781.00382,55467.65218 1991,23484.18563,31037.09202,37954.12366,48379.0263,58280.80931 1992,23218.92036,31458.44483,37546.15643,49041.81308,57838.07702 1993,22850.86845,31189.75526,37309.41181,48743.6913,60101.35734 1994,23866.2373,31621.86525,37437.23902,50405.55395,60049.14009 1995,23948.15731,31774.40004,38856.1793,53230.84692,73294.91911 1996,24229.84941,32579.74686,38956.39243,52442.1374,75560.39452 1997,25124.06646,32178.26575,39005.67736,53437.94688,75175.58861 1998,23359.4875,32342.52268,39175.12194,56955.74834,73852.29929 1999,22950.63291,32063.32837,40230.532,55873.07568,69828.087 2000,23395.64173,32697.49138,41350.36646,61159.01042,75782.25115 2001,23634.10112,33218.19604,41384.84502,61416.91595,80757.48515 2002,23719.90145,33260.23881,41417.54674,60956.15014,82087.54359 2003,24763.80582,34122.49356,41083.42657,60121.99476,80600.27656 2004,24135.18877,34176.17481,41855.03115,61668.19437,83949.27991 2005,24454.41043,34304.91195,41320.8201,61144.36476,85495.20992 2006,24324.36177,34324.16235,41455.88139,61996.92306,84508.58821 2007,23860.18793,33735.13948,40836.02053,60388.17076,79859.07347 2008,23989.32051,33036.1695,39859.8528,60141.24464,81594.11874 2009,24131.06733,33492.91246,41202.76415,61701.35358,83903.77549 2010,23924.97815,34254.86338,40481.73682,60110.06596,80354.70903 2011,24460.62977,34177.61411,40348.20452,61475.95395,81100.67874 2012,22881.70651,33064.98058,40308.80618,58716.77051,82076.11483
,Less than high school,High school diploma,Some college,College graduate,Postgraduate 1991,0.1023,0.3524,0.2706,0.1775,0.0971 1992,0.0954,0.3424,0.2805,0.1826,0.0991 1993,0.0945,0.3296,0.2883,0.1882,0.0994 1994,0.0965,0.3243,0.2865,0.1906,0.102 1995,0.0977,0.3243,0.2861,0.1917,0.1002 1996,0.0973,0.3298,0.2849,0.1934,0.0946 1997,0.0971,0.3259,0.282,0.2004,0.0946 1998,0.095,0.323,0.2778,0.2062,0.098 1999,0.0918,0.3199,0.2845,0.2042,0.0997 2000,0.0931,0.3133,0.2849,0.2102,0.0985 2001,0.0956,0.3068,0.2823,0.2123,0.103 2002,0.0967,0.3032,0.2796,0.2145,0.1059 2003,0.0945,0.3011,0.2803,0.215,0.1091 2004,0.096,0.3036,0.2787,0.2132,0.1086 2005,0.0952,0.298,0.2795,0.2182,0.1091 2006,0.097,0.2946,0.2704,0.224,0.114 2007,0.0906,0.2843,0.2789,0.2255,0.1207 2008,0.0903,0.2857,0.2766,0.2238,0.1236 2009,0.0882,0.2823,0.2763,0.2294,0.1238 2010,0.0831,0.2738,0.2785,0.2341,0.1305 2011,0.0867,0.2621,0.2799,0.236,0.1353 2012,0.0815,0.2541,0.2813,0.2417,0.1415