Income Differences Between PhDs and Masters: Evidence from The Netherlands

Abstract

In this paper we investigate individual income differences between individuals with Master and PhD degrees over the first 20 years of a career. We compare monthly incomes of Dutch higher education graduates with the same years of experience, gender and field of study. It turns out that over the first 20 years of experience, the average annual return to a PhD education is not significantly different from zero. During the PhD track and the first years after PhD graduation PhDs earn less than Masters, but this is compensated by higher earnings in later years. Income differences differ strongly by sex.

Appendix: Underlying Mechanisms

We analyze three underlying mechanisms for the observed differences in the income profile between PhDs and non-PhDs: differences in employment, in sector of work, and in hours worked per week.

1.1 Employment

The employment variable is based on the main source of income as registered in the wage data. Employment is defined as either being employed or being self-employed. We run logit regressions using the same covariates as in Eq. (2) to estimate the odds ratio of employment for PhDs compared to Masters. The odds ratio is defined as

$$\begin{aligned} \frac{P(\textit{work}|\textit{PhD})/(1-P(\textit{work}|\textit{PhD}))}{P(\textit{work}|\textit{Master})/(1-P(\textit{work}|\textit{Master}))}. \end{aligned}$$

The odds ratio shows the relative difference in the odds of employment between PhDs and Masters. An odds ratio higher than one indicates that PhDs are more likely to be employed than Masters. Figure 5 shows the odds ratios of employment over experience classes. Results are reported for men and women together, as well as separately (where we compare the odds with Masters of the same sex).

Fig. 5

Predicted probability of (self-)employment: Odd Ratios of probability of employment (versus non-employment) for PhDs compared to Master graduates with the same experience and sex, by sex and experience class. 95% confidence intervals

1.2 Hours Worked

Differences in hours worked per week can also partially explain variation in income. Both the PhD survey and the Labor Force Survey (LFS) contain information on self-reported hours worked.¹³ For each respondent, we therefore have one observation on hours worked. This is for the year 2009. We run an OLS regression on hours worked per week on this cross-section sample to obtain estimates of the difference in hours worked between PhDs and non-PhDs, for men and women separately. Since the reported hours worked in the PhD survey, by design, only pertain to PhDs who already have their PhD degree, we exclude the first experience-class in this analysis.

Figure 6 shows the difference of hours worked for PhDs compared to Masters with the same sex and experience. Part of the positive female PhD return is driven by female PhDs working more hours per week than female Master graduates. They do so over the whole experience spectrum. The estimated difference is on average 3.8 h/week (statistically significant at a 5-percent significance level). This corresponds to a more than 10% difference as female Master graduates work on average 32.5 h per week. For men there is a small, but non-significant, negative association between being a doctorate holder and hours worked.

1.3 Sector of Work

Differences in income profiles might be the result of differences in the sector in which Masters and PhDs typically work. For respondents who are employees, we have additional information on type of job, including sector. We define four sectors: private sector, subsidized, government, and academia (science). Differences in sector of work are estimated between PhDs and non-PhDs over experience, using a multinomial logit model. We include the same covariates as in Eq. (2). The relative risk ratios are plotted against experience in Fig. 7. These are

$$\begin{aligned} \frac{P({\textit{sector}}_{i}|PhD)/P({\textit{sector}}_{i}|PhD))}{P({\textit{sector}}_{i}|Master)/P({\textit{sector}}_{i}|Master)} \end{aligned}$$

where the reference sector, ${ sector}_{1}$, is the private sector.

Fig. 6

Predicted differences in hours worked. Differences in weekly hours work of PhDs compared to Master graduates with the same experience, by experience class and sex. Based on cross-section of 2009 survey data. 95% confidence intervals

Fig. 7

Predicted probabilities of sector of work: relative risk ratio of the probability of working in the subsidized or government sector versusworking in the private sector for PhDs compared to Master graduates

Fig. 8

Predicted returns to a PhD education by sector of work: return as a percentage of income of Master graduates with the same level of experience, by experience class. Observations for PhDs are only included after obtaining of PhD degree

This difference in private sector shares contributes negatively to the returns to a PhD education. This is because incomes in the private sector are on average higher than those in other sectors, particularly as one grows in experience. Figure 8 provides insight in the relative income position of PhDs compared to Master graduates working in the same sector. The figure is based on similar regressions as in Eq. (2), ran separately for each sector for all employees in the sample. To make a fair comparison, we only include observation for PhDs after PhD graduation. The effect of the difference in sector of work between Masters and PhDs can be assessed by recalculating the returns under the assumption that PhDs have the same sector distribution as Masters. The AAR (for comparison only after PhD graduation) is 2.8%. Whit equal sector distribution for Masters and PhDs, the AAR is 6.7%.