Do professors’ wages depend on students’ earnings?

ABSTRACT

Survey data indicate that the average salary of full professors in the U.S. and the average starting salary of new college graduates in the U.S. have both grown over each of the past eight years. The positive association between these survey data is intuitively appealing in that if greater demand for education comes as a consequence of higher annual earnings accruing to individuals with more education, then an increase in the starting salaries of new college graduates should lead to an increase in the wages of college and university faculty. This paper is the first to study whether the demand for professors is derived from the demand for higher education. To do so, we model the interactions between the academic labour market and the market for college and university graduates, and we show that, in equilibrium, academic wages are independent of students’ earnings. This main characteristic of the equilibrium conditions of our formal model is supported by empirical analysis.

KEYWORDS:

• Professors’ wages
• new college graduates’ earnings
• labor economics
• academic labor market
• private sector labor market

JEL CLASSIFICATION:

• J01
• J24
• J30

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Relatedly, Ehrenberg, McGraw, and Mrdjenovic (2006) report that differences in faculty quality present in different fields at a university are important predictors of the field differences in average faculty salaries that exist at the full professor level at the university.

² Gibson, Anderson, and Tressler (2014) use the same data to examine the impact of journal rankings on academic economists’ salaries.

³ Actually, earnings differences across fields rival college earnings premiums, which implies that the choice of field of study is potentially as important as the decision to enrol in college. Kirkeboen, Leuven, and Mogstad (2016) find that different fields have widely different payoffs, field of study drives the heterogeneity in the payoffs to postsecondary education, and the estimated payoffs are consistent with individuals choosing fields in which they have comparative advantage. The presence of comparative advantage is consistent with the Roy model (Heckman and Sedlacek 1985).

⁴ Relatedly, Broecke (2012) reports that a one standard deviation in selectivity across universities in the United Kingdom leads to a seven percent increase in earnings 3.5 years after graduation. Interestingly, Sjoquist and Winters (2016) find that state merit aid programs induce some students to forgo attending out‐of‐state schools and shifts them towards lower quality in‐state schools, so that the net effect is a reduction in academic quality.

⁵ Interestingly, a contemporaneous study by Diette and Raghav (2016) using student-level data from a selective private liberal arts college finds a negative relationship between university grades and graduates’ salaries.

⁶ Ehrenberg and Zhang (2005) report that usage of such faculty has been growing over the decades leading up to their study.

⁷ Interestingly, Figlio et al. (2013) find that university students perform better after the delivery of live-only instruction than after online-only delivery of instruction, and that the difference in performance between these two modes is especially pronounced for Hispanic students, male students and low-performing students. This issue is likely magnified in light of Ehrenberg and Zhang (2005) and Ran and Xu (2019), given that online delivery of instruction better facilitates the use of non-tenure track faculty.

⁸ Fortin and Ragued (2017) report a gender gap in the effect of temporary interruption of post-secondary education on post-graduate starting salaries. More specifically, they find a positive effect on wages of temporary interruption for men who held a full-time job during their out-of-school spell(s).

⁹ These explicit functional forms are often used given that they are in line with optimal allocation models with convex adjustment costs and yield crisp and quite simple explicit solutions that allow us to compare and reach conclusions in an easier way.

¹⁰ Note that a student’s earnings are independent of his or her ability, A. Although H depends on A, A is constant in H, implying that student ability is the same for all $i$.

¹¹ Note that from (13) ${ϵ}_{it\mathit{+}\mathit{1}}=\beta H_{it\mathit{+}\mathit{1}}-p{S}_{it\mathit{+}\mathit{1}}$, and from (14) $H_{it+1}=H_{it}+F_{it}\left(A,S_t,H_t,\phi \left(R_t\right),P\left(\frac{T_t}{n}\right)\right)-\text{bet}$, yielding ${\in }_{it+1}=\beta \left[{\mathrm{H}}_{it}+F_{it}\left(A,S_t,H_t,\phi \left(R_t\right),P\left(\frac{T_t}{n}\right)\right)-\text{bet}\right]-p{S}_{it+1}$. Differentiation with respect to yields (16).

¹² We thank an anonymous reviewer for this important remark.

¹³ Benefits to elite schools are better understood for high schools than colleges (e.g. Pop-Eleches and Urquiola 2013; Dobbie and Fryer, 2013; Estrada and Gignoux 2017).

¹⁴ This dilution is evidenced by the fact that the variance in R&D for the private colleges and universities in our sample is 4.8 times larger than that for the public colleges and universities in our sample. Relatedly, the variance of ProfSal for private colleges and universities in our sample is 3.9 times larger than that for public colleges and universities.

¹⁵ This is the specification used in the model employing the full sample, and seems to us to be the more reasonable specification for Tuition in this case given that state governments vary in their support of higher education institutions.

¹⁶ In the full sample case, logAge is only marginally insignificant.

¹⁷ For the latter of these two variables, HBCU is a dummy variable equal to 1 for historically black colleges and universities, and 0 otherwise.