Abstract
We examine the academe–industry wage gap. Once self-selection and different personal characteristics of academic and industrial scientists have been taken into account the wage gap narrows from 28 to 13 %. The counterfactual wage faced by an academic scientist increases with time spent on development and decreases with time spent on research. This finding challenges the idea of a solely negative relationship between science and wages. We further find that preferences for science augment the relationship between research orientation and wages. Overall, the results have implications for policy makers that aim to increase development oriented research activities at universities, individual scientists thinking about whether to pursue a career in industry or academe, and managers trying to hire academic scientists.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
See for instance Sauermann and Roach (2014) or Stern (2004) for industrial wage dynamics. Many studies consider academic wage dynamics, including factors such as gender (McNabb and Wass 1997; Barbezat 1987; Bayer and Astin 1968), seniority (Moore et al. 1998), and international wage differences (Altbach et al. 2012; Stevens 2004; Ong and Mitchell 2000). An earlier stream of research estimates academics’ earnings functions, finding that scientists’ earnings are concave, peaking late in their career (Stephan 1996; Creedy 1988; Diamond 1986; Laitner and Stafford 1985; Lillard and Weiss 1979; Weiss and Lillard 1978).
Some academics do manage to increase their income with consulting, speaking fees, or prizes.
For brevity, we only show those parts of Stern’s model that help to clarify our thoughts and argumentation. For details and all the implicit assumptions made to keep the model simple we refer to Stern (2004) and the cited literature there.
As such, the profit function cannot explain why some firms choose to engage only in research. For simplicity reasons we refrained from incorporating this in the model, but it could, for instance, be realized by accounting for different functional forms of the costs of research and development activities. As noted above, the assumption of higher productivity of development is not essential for the model, but does help us to derive implications. In practice both activities are complementary and face diminishing returns, and which activity has the highest marginal productivity depends on the relative levels of investment in both.
However, academic scientists who successfully focus on research could have higher chances for promotion, and hence increase earnings in the long run.
It is worthwhile to mention that these findings do not indicate that there are no productivity effects at all, but rather that the preference effect on average dominates potential productivity effects with regard to basic research.
Scientists’ taste for business, on the other hand, represents motivation through extrinsic factors, such as wages, extralegal benefits, and job security. Noteworthy, both preferences are not mutually exclusive, meaning that scientists can have a high taste for science as well as a high taste for business (Sauermann and Roach 2012; Agarwal and Ohyama (2013); Roach and Sauermann 2010). We include the researchers’ taste for business into our empirical analysis to account for potential confounding effects.
Several other methods are available to deal with selection, including difference-in-difference designs, regression discontinuity designs, instrumental variable estimation, and control functions. The difference-in-difference method requires panel data on scientists that switch from academe to industry (or the other way around) due to an exogenous shock. As the data we employ is a cross-section and no exogenous shock is observed we cannot apply this estimator. A regression discontinuity design is also not applicable because there is no (arbitrary) rule that would determine the treatment (being an academic scientist). IV estimators as well as selection models rely on the idea to instrument selection into industry or academe through an instrumental variable, which affects the treatment but not the outcome, i.e. the wage gap. It is notoriously difficult to find a valid instrument in this particular case and we could not find a suitable instrument for this application either.
Detailed information on the project can be found in Unesco (2012), Eurostat (2012), or OECD (2013). Summary statistics based on the full survey have been published in Auriol (2007, 2010). Statistics on the Belgian data collection can be found in Moortgat and Van Mellaert (2011). For the methodological background and core questionnaire see Auriol et al. (2012).
The results change only slightly when we take hourly wage as the main variable of interest instead of the annual wage. The corresponding results are presented in “Appendix 3”.
In order to ensure a clear differentiation between academic and industrial scientists and control for systematic differences caused through multiple employments, we discarded scientists and researchers with multiple jobs.
Agarwal and Ohyama (2013) define scientists’ activities as primarily basic or primarily applied in nature, while Sauermann and Stephan (2013) make a distinction between scientists primarily engaged in basic research, those primarily engaged in applied research, and those primarily engaged in development. Our measure of research activities is most likely mainly driven by basic research activities but also picks up applied research components.
Our results are robust to alternatively taking the sums of challenge, contribution, and independence to represent taste for science.
For instance, one of the main Flemish institutions distributing Ph.D. scholarships (FWO 2013) lists “research ability and potential (including course results)” and “research skills and methodology” as the first two selection criteria for its Ph.D. fellowships.
We consider the researchers’ detailed research domain (listed in Table 1) for this calculation.
“Appendix 3” provides results based on the hourly wage instead of the annual wage as an alternative to control for differences in scientists’ efforts.
Marginal effect calculated as exp(−0.2196) − 1; analogue for the other marginal effects described in this section.
Columns five and six of Table 6 below show that this assumption does not hold for all covariates.
Note that all results are even more pronounced if one considers the median wages instead of the means.
While we find some significant effects of covariates on wage after matching, this does not mean that the matching was unsuccessful. Rather, they should be interpreted as common trends across industry and academe.
It should be noted here that this evaluation of the wage gap assumes that the academic would work for the average firm, instead of matching to a firm which suits his needs. If this matching were to happen, the wage gap could become even smaller for scientists who spend much time on research, as they are more likely to sort into firms which offer high freedom and lower wages. Scientists who spend more time on development should be more likely to sort into companies which focus more strongly on development, leading to an even higher wage differential.
To avoid multicollinearity we include career_years but not age in the regression. The results are robust to employing age instead of career_years.
23 academics in the sample, or 5 %, spend at least 50 % of their time on development.
As noted above, these predictions assume that scientists would work for the average firm, instead of sorting into firms according to their preferences.
Since the turning point of the estimated relationship between career years and the relative wage gap lies far out of the observed range of career years (315 years), the influence is almost linear with only slightly decreasing marginal effects. Including only the linear term or taking a logarithmic specification does not alter the results.
Alternatively, one could include a corresponding interaction term in the regression. However, interaction term specifications imply the assumption that all other covariates exert the same effect within both sub-groups, which is in our case not true.
In that regard, Balsmeier and Pellens (2014) find that academic scientists that patent are more likely to leave academe, and also publish more than average.
For instance, our findings regarding motive-based selection into industry or academe are similar to those of Roach and Sauermann (2010), Sauermann and Roach (2012) and Agarwal and Ohyama (2013). Sauermann and Roach (2014) uncovers a similar relationship between motives and reservation wages for publishing.
The groups described here can overlap, e.g. they were all retained in the determination of which scientists to remove from the sample and then removed all the same time.
Similar methods have been employed by Sauermann and Roach (2012).
References
Agarwal, R., & Ohyama, A. (2013). Industry or academia, basic or applied? Career choices and earnings trajectories of scientists. Management Science, 59, 950–970.
Aghion, P., Dewatripont, M., & Stein, J. (2008). Academic freedom, private-sector focus, and the process of innovation. Rand Journal of Economics, 39(3), 617–635.
Altbach, P., Reisberg, L., Yudkevich, M., Androushcak, G., & Pacheco, I. (2012). Paying the professoriate: A global comparison of compensation and contracts. London: Routledge.
Angrist, J. (1998). Estimating the labor market impact of voluntary military service using social security data on military applicants. Econometrica, 66(2), 249–288.
Arora, A., & Gambardella, A. (1994). The changing technology and technological change: General and abstract knowledge and the division of innovative labour. Research Policy, 23(5), 523–532.
Auriol, L. (2007). Labour market characteristics and international mobility of doctorate holders: Results for seven countries. OECD Science, Technology and Industry Working Papers 2007/02. doi:10.1787/310254328811.
Auriol, L. (2010). Careers of doctorate holders: Employment and mobility patterns. OECD Science, Technology and Industry Working Papers 2010/04. doi:10.1787/5kmh8phxvvf5-en.
Auriol, L., Schaaper, M., & Felix, B. (2012). Mapping careers and mobility of doctorate holders: Draft guidelines, model questionnaire and indicators. OECD Science, Technology and Industry Working papers 2007/6. doi:10.1787/5k4dnq2h4n5c-en.
Balsmeier, B., & Pellens, M. (2014). Who makes, who breaks: Which scientists stay in academe? Economics Letters, 122(2), 229–232.
Barbezat, D. (1987). Salary differentials by sex in the academic labour market. Journal of Human Resources, 22, 422–428.
Bayer, A., & Astin, H. (1968). Sex differences in academic rank and salary among science doctorates in teaching. Journal of Human Resources, 2, 191–200.
Blundell, R., & Costa Dias, M. (2009). Alternative approaches to evaluation in empirical microeconometrics. Journal of Human Resources, 44, 565–640.
Card, D., & Sullivan, D. (1988). Measuring the effect of subsidized training-programs on movements in and out of employment. Econometrica, 56, 497–530.
Cohen, W. M., & Levinthal, D. A. (1989). Innovation and learning: The two faces of R&D. The Economic Journal, 99(397), 569–596.
Cohen, W., & Levinthal, D. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35, 128–152.
Creedy, J. (1988). Cohort and cross-sectional earnings profiles: Scientists in Britain and Australia. Journal of Economic Studies, 15(1), 44–52.
Dasgupta, P., & David, P. (1994). Towards a new economics of science. Research Policy, 23, 487–521.
Diamond, A. (1986). The life-cycle research productivity of mathematicians and scientists. Journal of Gerontology, 41, 520–525.
Eurostat. (2012). Careers of doctorate holders. Retrieved 05 March, 2012, from http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Careers_of_doctorate_holders.
Fini, R., & Lacetera, N. (2010). Different yokes for different folks: Individual preference, institutional logics, and the commercialization of academic research. In G. Libecap (Ed.), Spanning boundaries and disciplines: University technology commercialization in the idea age (Vol. 21, pp. 1–25). Bingley: Emerald Group Publishing.
Frolich, M. (2007). Propensity score matching without conditional independence assumption—With an application to the gender wage gap in the United Kingdom. Econometric Journal, 10, 359–407.
FWO. (2013). Ph.D. Fellowship. Retrieved 27 Jan, 2013, from www.fwo.be/Aspirant.aspx.
Gerfin, M., & Lechner, M. (2002). A microeconometric evaluation of the active labour market policy in Switzerland. Economic Journal, 112, 854–893.
Heckman, J. (1976). The common structure of statistical models for truncation, sample selection and limited dependent variables and a simple estimator for such models. Annals of Economic and Social Measurement, 5(4), 475–492.
Heckman, J. J., Ichimura, H., & Todd, P. (1998). Matching as an econometric evaluation estimator. Review of Economic Studies, 65(2), 261–294.
Imbens, G., & Wooldridge, J. (2009). Recent developments in the econometrics of program evaluation. Journal of Economic Literature, 47, 5–86.
Konrad, A., & Pfeffer, J. (1990). Do you get what you deserve? Factors affecting the relationship between productivity and pay. Administrative Science Quarterly, 35, 258–285.
Lacetera, N. (2009). Different missions and commitment power in RD organizations: Theory and evidence on industry–university alliances. Organization Science, 20, 565–582.
Laitner, J., & Stafford, F. (1985). The academic labor market: Has compensation diverged from other professions?. Washington, DC: Econometric Society Meetings.
Lechner, M. (2001). Identification and estimation of causal effects of multiple treatments under the conditional independence assumption. In M. Lechner & F. Pfeiffer (Eds.), Econometric evaluation of active labor market policies (pp. 43–58). Heidelberg: Physica-Verlag.
Lillard, L., & Weiss, Y. (1979). Components of variation in panel earnings data: American scientists. Econometrica, 47, 437–454.
McNabb, R., & Wass, V. (1997). Male–female salary differentials in British universities. Oxford Economic Papers, 49, 328–343.
Merton, R. (1973). The sociology of science: Theoretical and empirical investigations. Chicago: University of Chicago Press.
Moore, W., Newman, R., & Turnball, G. (1998). Do academic wages decrease with seniority? Journal of Labour Economics, 16, 352–366.
Moortgat, P., & Van Mellaert, G. (2011). CDH (Careers of Doctorate Holders). Onderzoek, ontwikkeling en innovatie in België studiereeks 12. http://www.belspo.be/belspo/organisation/Publ/pub_ostc/ind/ind12_nl.pdf. Brussels: Belgian Science Policy Office.
National Science Board. (2012). Science and engineering indicators 2012. Arlington VA: National Science Foundation.
Nopo, H. (2008). Matching as a tool to decompose wage gaps. Review of Economics and Statistics, 90, 290–299.
OECD. (2013). OECD/UNESCO Institute for Statistics/Eurostat Careers of Doctorate Holders (CDH) project. Retrieved June 08, 2013, from http://www.oecd.org/innovation/inno/oecdunescoinstituteforstatisticseurostatcareersofdoctorateholderscdhproject.htm.
Ong, L., & Mitchell, J. (2000). Professors and hamburgers: An international comparison of relative academic salaries. Applied Economics, 32, 869–876.
Roach, M., & Sauermann, H. (2010). A taste for science? PhD scientists’ academic orientation and self-selection into research careers in industry. Research Policy, 39, 422–434.
Rosenbaum, P., & Rubin, D. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41–55.
Rosenberg, N. (1990). Why do firms do basic research (with their own money)? Research Policy, 19, 165–174.
Rubin, D. (1977). Assignment to a treatment group on the basis of a covariate. Journal of Educational Statistics, 2, 1–26.
Sauermann, H., Roach, M. (2012). Taste for science, taste for commercialization, and hybrid scientists. Paper presented at DRUID conference, June 19–21. Denmark: Copenhagen Business School.
Sauermann, H., & Roach, M. (2014). Not all scientists pay to be scientists: Heterogeneous preferences for publishing in industrial employment. Research Policy, 43(1), 32–47.
Sauermann, H., & Stephan, P. (2013). Conflicting logics? A multidimensional view of industrial and academic science. Organization Science, 24(3), 889–909.
Stephan, P. E. (1996). The economics of science. Journal of Economic Literature, 34, 1199–1235.
Stephan, P., & Levin, S. (1992). Striking the mother lode in science: The importance of age, place, and time. Oxford: Oxford University Press.
Stern, S. (2004). Do scientists pay to be scientists? Management Science, 50, 835–853.
Stevens, P. (2004). Academic salaries in the UK and US. National Institute Economic Review, 190, 104–113.
Uebersax, J. (2000). Estimating a latent trait model by factor analysis of tetrachoric correlations. Retrieved March 16, 2011, from http://john-uebersax.com/stat/irt.htm.
Unesco. (2012). Tracking the careers of doctorate holders. Retrieved May 07, 2012, from http://www.uis.unesco.org/ScienceTechnology/Pages/doctorate-degree-holders.aspx.
Walker, J., Vignoles, A., & Collins, M. (2010). Higher education academic salaries in the UK. Oxford Economic Papers, 62, 13–35.
Weiss, Y., & Lillard, L. (1978). Experience, vintage, and time effects in the growth of earnings: American scientists. Journal of Political Economy, 86, 427–474.
Wetenschapsbeleid, Federaal. (2006). Careers of Doctorate Holders Survey [Database]. Gent: ECOOM UGent.
Acknowledgments
Balsmeier gratefully acknowledges financial support from the Flemish Science Foundation. Pellens gratefully acknowledges financial support from the National Bank of Belgium. Both authors thank participants in the LEI & BRICK workshop on the ‘Organization, Economics, and Policy of Scientific Research’, the Technology Transfer Society Conference, the DRUID Society Conference, a lunch seminar at KU Leuven, as well as Dirk Czarnitzki, Lee Fleming, Christoph Grimpe, Reinhilde Veugelers, Stijn Kelchtermans, Henry Sauermann, Toby Stuart, and Scott Stern for their insightful comments.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Data
Starting from the initial dataset of the CDH survey we discarded a number of observations for various reasons. Of the 7,160 responses, 5,601 were employed at the time of the survey. 1,973 of those employed were employed in sectors other than academe or the private sector (such as the public sector or private nonprofits). Of the 3,628 remaining observations, 1,713 were active in fields other than natural sciences or engineering.
The remaining 1,915 observations were further scrutinized for inconsistencies.Footnote 29 53 academic scientists were removed from the sample because they reported distinctly non-academic job descriptions, including ‘webmaster’, ‘secretary’, ‘manager’, or ‘coordinator’. Respondents who reported being employed in multiple positions were removed from the sample. These are problematic because job characteristics are only reported for the main position, and because it is not obvious to assign scientists with academic and industrial positions to one of the two groups. 38 academic scientists and 28 industrial researchers were thus dropped from the sample. Note that the academic scientists were mainly removed because of industry involvement, while the industrial researchers were employed in multiple jobs in industry. Also note that respondents (especially academic scientists) might still hold multiple affiliations on paper if they did not report them. As such, this restriction serves to remove respondents who in practice hold multiple jobs.
Those who did not report wage or reported zero wages (224 observations), wages below the Belgian annual minimum wage (20 observations) or above the 99th percentile (€200.000, 16 observations) were not included in the analysis. Wages below the minimum wage are either misreported (caused, for instance, by reporting monthly income instead of annual income), or the result of part-time employment, which is removed in the next step. The last group is so rare and unique in the Belgian academic labor market that it would be meaningless to search for comparable scientists. They also have unwarrantedly strong influence on the average wages. These limitations are however not likely to exclude specific groups of scientists (for instance, lower-earning females) from the analysis.
In order to further ensure a reasonable comparability of the scientists included in the analysis, the sample was further restricted to scientists and researchers in the first 30 years after graduation (removing 232 scientists) who were older than 18 and younger than 65 at the moment the survey was carried out (removing 9 scientists). We also removed scientists and researchers who reported working less than full-time (38 h, 86 observations) or more than 75 h a week (the 99th percentile, 21 observations). After further removing missing values on the variables of interest, this resulted in a total sample of 1,245 scientists, 486 of which were employed in academe and 759 in industry.
Lastly, scientists and researchers who reported spending in total more than 100 % of their time on research and development were re-scaled to reflect a maximum time allocation of 100 %.
Appendix 2: Exploratory factor analysis used in tastes calculation
Since the items are all binary, we employ the Tetrachoric correlation matrix for this analysis (Uebersax 2000).Footnote 30 The analysis returns a two factor solution, explaining 87 % of total variation. Several rotations were applied from which we select the Varimax solution for interpretability. The first factor correlates with motivation through salary, extralegal benefits, career prospects, job security, and weakly with work circumstances. We name this factor “Taste for Business”. The second factor correlates with motivation through intellectual challenge, independence, contribution to society, and, more strongly, with work circumstances. We name this factor “Taste for Science”. Both were normalized for interpretability (Tables 7, 8).
Appendix 3
Appendix 4: Hourly wage results
Rights and permissions
About this article
Cite this article
Balsmeier, B., Pellens, M. How much does it cost to be a scientist?. J Technol Transf 41, 469–505 (2016). https://doi.org/10.1007/s10961-014-9388-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10961-014-9388-1