Skip to main content

Emergent structures in faculty hiring networks, and the effects of mobility on academic performance

Abstract

This paper is about the South African job market for Ph.Ds. Ph.D. to first job mobility involves the preferences of both the hiring institution and the candidate. Both want to make the best choice and here institutional prestige plays a crucial role. A university’s prestige is an emergent property of hiring interactions, so we use a network perspective to measure it. Using this emergent ordering, we compare the subsequent scientific performance of scholars with different changes in the prestige hierarchy. We ask how movements between universities of different prestige from Ph.D. to first job correlates with academic performance. We use data of South African scholars from 1970 to 2012 and we find that those who make large movements in terms of prestige have lower research ratings than those who do not. Further, looking only those with large prestige movements, those with higher prestige Ph.Ds or first jobs have higher research ratings throughout their careers.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Notes

  1. Language regarding racial or ethnic identity can be charged. As the empirical part of this paper is about South Africa, we follow the standard terminology accepted there.

  2. No doubt this explains why almost all rankings provide integer rankings only for a given number of top places, after which institutions are grouped into rather large groups (100–199; 200–299 and so on).

  3. NRF (www.nrf.ac.za) is a state agency that has as its mission the promotion of research and the development of national research capacity.

  4. For more information on the rating system see http://www.nrf.ac.za/rating.

  5. In 2004 the university system was reformed: some universities were merged and changed names. The reform does not affect the big universities, excepting University of Johannesburg and KwaZuluNatal, but does have some effects on the lower ranked ones. We use the post-merger names because the data are more complete. We discuss this in “Appendix 1”.

  6. In South Africa there are formally four “racial” groups: black, white, Indian and coloured. The word “black” is sometimes used to refer to the aggregate of black, Indian and coloured, and this is the meaning we apply throughout this text.

  7. The same patterns, and explanation, are present in SSH (Table 7).

  8. The exceptions to this strong correlation are for SET: Tshwane University of Technology, University of Fort Hare, and University of Limpopo that have a high connections in hiring (in-degree) and a low connections in placement (out-degree); while University of KwaZulu-Natal, and University of Cape Town instead show a high connections in placement (out-degree) and a low connections hiring (in-degree). See Table 5. For SSH we find the following exceptions: Walter-Sisulu University, Nelson Mandela Metropolitan University, and University of Limpopo with a high hiring connections (in-degree) and a low placement connections (out-degree); and University of Stellenbosch, and University of Cape Town with a high placement connections (out-degree) and a low hiring connections (in-degree).

  9. The University of KwaZulu-Natal was formed in 2004 by the merger of the University of Natal (white) and the University of Durban-Westville (Indian).

  10. Checking manually, 100 iterations is almost always enough to find a local optimum.

  11. We select the university name in the address search, restricting to the period 1988–2004. For the universities that changed name we search pre and post merger names and we refine the WOS results looking at only journals publications. Last access February 2018.

  12. Per capita measures are obtained dividing total records by number of scholars in our database with a current affiliation in the particular university. This will over-estimate the per capita figures as there are (some, though few) South African authors in WOS who never apply for a rating. There is no reason to believe there is any bias in this over-estimate however.

  13. For SSH we find non statistically significant results for total number of publications and publications per capita. As discussed previously our methodology is less suited for SSH.

  14. These numbers are 48 and 74% respectively for SSH.

  15. Additionally we perform a robustness check following a bootstrap technique in “Appendix 7

  16. This is done removing the diagonal of the matrix.

  17. In spite of the 2004 reforms, this feature remains part of the SA academic world: in the period 2004 to 2012 it still the case that about two thirds of graduate receive their first appointment at their Ph.D. institution. The fraction is slightly higher in SET, slightly lower in SSH.

  18. For a discussion of this see for example Cowan et al. (2010) pp. 278–299.

  19. According to Sekhon (2011) to account for possible presence of ties we use the ks.boot command under the R library Matching.

References

  • Allison, P. D., & Long, J. S. (1987). Interuniversity mobility of academic scientists. American Sociological Review, 52, 643–652.

    Article  Google Scholar 

  • Araki, S., Kawaguchi, D., & Onozuka, Y. (2016). University prestige, performance evaluation, and promotion: Estimating the employer learning model using personnel datasets. Labour Economics, 41, 135–148.

    Article  Google Scholar 

  • Bair, J. H. (2003). Hiring practices in finance education: Linkages among top-ranked graduate programs. American Journal of Economics and Sociology, 62(2), 429–433.

    Article  Google Scholar 

  • Bair, J. H., & Bair, R. K. (1998). Linkages among top-ranked graduate programs in special education: A brief report. Mental Retardation, 36(1), 52–54.

    Article  Google Scholar 

  • Bair, J. H., & Boor, M. (1991). The academic elite in law: Linkages among top-ranked law schools. Psychological Reports, 68(3), 891–894.

    Article  Google Scholar 

  • Baldi, S. (1995). Prestige determinants of first academic job for new sociology Ph.Ds 1985–1992. The Sociological Quarterly, 36(4), 777–789.

    Article  Google Scholar 

  • Barnard, H., Cowan, R., & Müller, M. (2012). Global excellence at the expense of local diffusion, or a bridge between two worlds? Research in science and technology in the developing world. Research Policy, 41(4), 756–769.

    Article  Google Scholar 

  • Barnard, H., Cowan, R. A., Kirman, A. P., & Müller, M. (2016). Including excluded groups: The slow racial transformation of the south african university system. Working Paper Series in Economics, Karlsruher Institut für Technologie (KIT).

  • Barnett, G. A., Danowski, J. A., Feeley, T. H., & Stalker, J. (2010). Measuring quality in communication doctoral education using network analysis of faculty-hiring patterns. Journal of Communication, 60(2), 388–411.

    Article  Google Scholar 

  • Bedeian, A. G., & Feild, H. S. (1980). Academic stratification in graduate management programs: Departmental prestige and faculty hiring patterns. Journal of Management, 6(2), 99–115.

    Article  Google Scholar 

  • Burris, V. (2004). The academic caste system: Prestige hierarchies in Ph.D. exchange networks. American Sociological Review, 69(2), 239–264.

    Article  Google Scholar 

  • Clauset, A., Arbesman, S., & Larremore, D. B. (2015). Systematic inequality and hierarchy in faculty hiring networks. Science Advances, 1(1), e1400005.

    Article  Google Scholar 

  • Connelly, B. L., Certo, S. T., Ireland, R. D., & Reutzel, C. R. (2011). Signaling theory: A review and assessment. Journal of Management, 37(1), 39–67.

    Article  Google Scholar 

  • Conti, A., & Visentin, F. (2015). A revealed preference analysis of Ph.D. students choices over employment outcomes. Research Policy, 44(10), 1931–1947.

    Article  Google Scholar 

  • Cowan, W. B., Cowan, R., & Llerena, P. (2010). Running the marathon. In M. McKelvey, M. Holmén (Eds.), Learning to Compete in European UniversitiesFrom Social Institution to Knowledge Business, Chapter 11, pp. 278–299. Edward Elgar Publishing.

  • Hagstrom, W. O. (1971). Inputs, outputs, and the prestige of university science departments. Sociology of Education, 44, 375–397.

    Article  Google Scholar 

  • Hurlbert, J. S., & Rosenfeld, R. A. (1992). Getting a good job: Rank and institutional prestige in academic psychologists’ careers. Sociology of Education, 65, 188–207.

    Article  Google Scholar 

  • Judge, T. A., Kammayer-Mueller, J., & Bretz, R. D. (2004). A longitudinal model of sponsorship and career success: A study of industrial-organizational psychologists. Personnel Psychology, 57(2), 271–303.

    Article  Google Scholar 

  • Jung, J., & Lee, S. J. (2016). Influence of university prestige on graduate wage and job satisfaction: The case of South Korea. Journal of Higher Education Policy and Management, 38(3), 297–315.

    Article  Google Scholar 

  • León, L. R., Cowan, R., & Müller, M. (2016). Formation and persistence of research communities in middle income countries: The case of South Africa. UNU-MERIT Working paper.

  • Long, J. S. (1978). Productivity and academic position in the scientific career. American Sociological Review, 43, 889–908.

    Article  Google Scholar 

  • Mai, B., Liu, J., & González-Bailón, S. (2015). Network effects in the academic market: Mechanisms for hiring and placing Ph.Ds in communication (2007–2014). Journal of Communication, 65(3), 558–583.

    Article  Google Scholar 

  • Mas-Colell, A., Whinston, M. D., Green, J. R., et al. (1995). Microeconomic theory (Vol. 1). New York: Oxford University Press.

    MATH  Google Scholar 

  • Merton, R. K., et al. (1968). The matthew effect in science. Science, 159(3810), 56–63.

    Article  Google Scholar 

  • Miller, C. C., Glick, W. H., & Cardinal, L. B. (2005). The allocation of prestigious positions in organizational science: Accumulative advantage, sponsored mobility, and contest mobility. Journal of Organizational Behavior, 26(5), 489–516.

    Article  Google Scholar 

  • Reskin, B. F. (1977). Scientific productivity and the reward structure of science. American Sociological Review, 42, 491–504.

    Article  Google Scholar 

  • Sekhon, J. S. (2011). Multivariate and propensity score matching software with automated balance optimization: The Matching package for R. Journal of Statistical Software, 42(7), 1–52.

    Article  Google Scholar 

  • Shavit, Y., et al. (2007). Stratification in higher education: A comparative study. Palo Alto: Stanford University Press.

    Google Scholar 

  • Vries, H. d. (1998). Finding a dominance order most consistent with a linear hierarchy: A new procedure and review. Animal Behaviour, 55(4), 827–843.

    Article  Google Scholar 

  • Webster, D. S., Conrad, C. F., & Jensen, E. L. (1991). Objective and reputational rankings of Ph.D.—granting departments of sociology, 1965–1982. Sociological Focus, 21(2), 177–198.

    Article  Google Scholar 

  • Williamson, I. O., & Cable, D. M. (2003). Predicting early career research productivity: The case of management faculty. Journal of Organizational Behavior, 24(1), 25–44.

    Article  Google Scholar 

Download references

Acknowledgements

Financial support was provided through the Institut Universitaire de France. Much of this work was done when Rossello and Cowan were visiting scholars at the Centre for Research on Evaluation, Science and Technology (CREST) at Stellenbosch University, South Africa. We gratefully acknowledge the comments and suggestions of participants of the seminar at CREST, as well as those of Prof. Pierre Mohnen, Dr. Moritz Müller, and Prof. Giorgio Fagiolo. We would like also to thank an anonymous referee for the helpful comments.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Giulia Rossello.

Appendices

Appendix 1: University reform in 2004

The South African university system saw a major reform in 2004. The reform merged and split university departments in the spirit of a geographical rationalization and racial integration. In our analysis we use post-merger names mostly because the data are more complete. More precisely, it is possible to make an accurate translation from pre- to post-merger names, but not from post- to pre-merger names, so by using pre-merger names we would lose a significant number of observations. Moreover, the use of post-merger names represents a value added of the work. It is a way to produce the prestige ranking of the South African universities that can be compared to the actual system. From the point of view of the analysis we note the following. The University of Johannesburg came into existence as the result of a merger between Rand Afrikaans University, Technikon Witwatersrand, and Vista University, where the latter two have almost no Ph.Ds (3 in total) in the period. So using University of Johannesburg instead of its disaggregation pre-merger would not make much difference. Similarly, Nelson Mandela Metropolitan University was created by the merger of Port Elizabeth Technikon, University of Port Elizabeth, and Vista University where the sample is dominated by Ph.Ds from Port Elizabeth. NorthWest University is a merger of University of the North West and Potchefstroom University, and the latter dominates Ph.D. production, particularly if we restrict attention to SET where 32 Ph.Ds are from Potchefstroom versus 6 from the University of the North West. The only possible problem could arise for the case of University of KwaZulu Natal which is the merger of University of Durban West Ville and University of Natal. Though, restricting to SET, Natal dominates with 32 Ph.Ds versus 6 in Durban West Ville.

As a robustness check we redo the prestige ranking in SET with those observations for which we have full data using pre-merger names. Table 4 shows the results which are quite consistent to those of the full sample. We must also be cognizant of the fact that in the period there was to some extent a language divide which appears mitigated using post-merger names. Indeed, looking separately at English and Afrikaans language universities in Table 4 we can observe a more informative pattern. English universities have the same rankings here as in the main analysis. The exception being KZN: University of Natal is ranked 4th in the analysis using old names whereas KZN was ranked second among english universities in the main analysis. Afrikaans universities have almost the same ranking here as they do in the main analysis; and if UNW (NWO and Potchefstroom) are excluded.

We have also repeated the matched pairs analysis with pre-merger names, and notwithstanding the reduced sample size, results are in line with our main findings.

Table 4 Prestige ranking for SET 1970–2004 using pre-merger names. The prestige ranking is in ascending order from the highest prestige which correspond to one. The number is the average of the orders with the maximum scores of the 10,000 repetitions. The algorithm is run on the adjacency matrix of pre-merger university names of the hiring network. Universities with fewer than 5 Ph.Ds are excluded

Appendix 2: Faculty hiring network

See Table 5.

Table 5 Indegree and outdegree SET hiring network for the years 1970–2004. Network statistics are compute without considering self-loops

Appendix 3: Prestige ranking aggregating SET and SSH

See Fig. 6.

Fig. 6
figure 6

Prestige Ranking 1970–2004 without distinction of fields. The frequency scores are in ordered left to right from the highest prestige which corresponds to one. The Black dots represent the average placement of each university in the maintained orderings, red dots and green dots are respectively one and two standard deviation from the average. Our algorithm runs on the adjacency matrix of the hiring network. Universities with fewer than 5 Ph.Ds are excluded. (Color figure online)

Appendix 4: Cohort effects

See Figs. 7, 8 and 9.

In this “Appendix” we present the results of the matched pair analysis but restrict the sample to those who received a Ph.D. prior to 1992. This way we reduce significantly any possible cohort effects.

Fig. 7
figure 7

Up versus stay comparison, with the sample restricted to those who received their Ph.D. degrees prior to 1992. The black curves are cumulative distribution functions of the proportion of observations in which \(R_{\mathrm{up}}>R_{\mathrm{stay}}\) was the case for \(p\%\) of the matched pairs. Grey curves are the CDFs for the \(R_{\mathrm{stay}}>R_{\mathrm{up}}\) proportions. From top to bottom 5, 10, 15, and 20 years after Ph.D. Pairs matched using gender, race, Ph.D. obtained years and first job university (left column) or Ph.D. institution (right column). a First job, 5 years after Ph.D. b Ph.D., 5 years after Ph.D. c First job, 10 years after Ph.D. d Ph.D., 10 years after Ph.D. e First job, 15 years after Ph.D. f Ph.D., 15 years after Ph.D. g First job, 20 years after Ph.D. h Ph.D., 20 years after Ph.D.

Fig. 8
figure 8

Down versus stay comparison, with the sample restricted to those who received their Ph.D. degrees prior to 1992. The black curves are cumulative distribution functions of the proportion of observations in which \(R_{\mathrm{down}}>R_{\mathrm{stay}}\) was the case for \(p\%\) of the matched pairs. Grey curves are the CDFs for the \(R_{\mathrm{stay}}>R_{\mathrm{down}}\) proportions. From top to bottom 5, 10, 15, and 20 years after Ph.D. Pairs matched using gender, race, Ph.D. obtained years and first job university (left column) or Ph.D. institution (right column). a First job, 5 years after Ph.D. b Ph.D., 5 years after Ph.D. c First Job, 10 years after Ph.D. d Ph.D., 10 years after Ph.D. e First Job, 15 years after Ph.D. f Ph.D., 15 years after Ph.D. g First Job, 20 years after Ph.D. h Ph.D., 20 years after Ph.D.

Fig. 9
figure 9

Up versus down comparison, with the sample restricted to those who received their Ph.D. degrees prior to 1992. The black curves are cumulative distribution functions of the proportion of observations in which \(R_{\mathrm{up}}>R_{\mathrm{down}}\) was the case for \(p\%\) of the matched pairs. Grey curves are the CDFs for the \(R_{\mathrm{down}}>R_{\mathrm{up}}\) proportions. From top to bottom 5, 10, 15, and 20 years after Ph.D. Pairs matched using gender, race, Ph.D. obtained years and first job university (left column) or Ph.D. institution (right column). a First job, 5 years after Ph.D. b Ph.D., 5 years after Ph.D. c First job, 10 years after Ph.D. d Ph.D., 10 years after Ph.D. e First job, 15 years after Ph.D. f Ph.D., 15 years after Ph.D. g First job, 20 years after Ph.D. h Ph.D., 20 years after Ph.D.

Appendix 5: SSH results

Table 6 gives the adjacency matrix of Ph.D. to first job transition in SSH.

Table 6 Adjacency matrix of the hiring network for the years 1970–2004 in SSH, rows are Ph.D. institutions and columns are first job institutions. Each entry represents the number for people with a Ph.D. in university i hired as first job in university j

See Figs. 10 and 11.

Fig. 10
figure 10

Hiring network 1970–2004 SSH. The vertex are the South African Universities, plotted according to their geographical coordinates (for the institutions located in the same area we separated manually). Vertex size in-degree, vertex colour out-degree. Where the correlation between in-degree and out-degree 0.53

Fig. 11
figure 11

Prestige Ranking for SSH 1970–2004. The frequency scores are in ascending order from the highest prestige which correspond to one. The black dots is the average of the orders with the maximum scores under 10,000 repetition, red dots and green dots are respectively one and two standard deviation from the average. Our algorithm runs on the adjacency matrix of the hiring network. Universities with fewer than 5 Ph.Ds are excluded. (Color figure online)

Table 7 contains summary statistics of the SSH hiring network.

See Tables 7 and 8.

Table 7 Summary statistics SSH hiring network for the years 1970–2004. Network statistics are compute without considering self-loops
Table 8 Ph.Ds hired from the top 5 prestigious universities in SSH, according to our prestige ranking. The total of SSH Ph.Ds Hired in the period is 542

Appendix 6: KS test matched pairs

See Tables 9 and 10.

Table 9 Results of KS test relative to Fig. 3, 4, and 5. For two tailed test the null hypothesis is that cdf are different, for the less/greater test the first cdf lies below/above the other
Table 10 Results of KS test relative to Figs. 7, 8, and 9. Sample restricted to those who received their Ph.D. degrees prior to 1992. For two tailed test the null hypothesis is that cdf are different, for the less/greater test the first cdf lies below/above the other

Appendix 7: Robustness check of KS test

See Tables 11 and 12.

To address the well known problem of excess of sensitivity of KS test with large samples, we perform a robustness check, following a bootstrap technique. For each comparison of the distributions of proportions obtained with our matched pairs technique, we sample with replacement 100 samples of size 100 and each time we compute the KS test,Footnote 19 then we store the obtained p values and we count how many times the p values are lower than 0.05. Under the null hypothesis, p values are distributed as a uniform, so if the fraction of p values under 0.05 is larger that 0.05, then we can conclude in a more consistent way that the two distributions are different. Table 11 shows the results of this procedure for Figs. 3, 4, and 5. Observing the table the results of KS tests are confirmed.

Table 11 Proportion of p values lower than 0.05 of Bootstrap KS test repeated 100 times on repeated re-samples of size \(=\) 100 for p%
Table 12 Proportion of p values lower than 0.05 of Bootstrap KS test repeated 100 times on repeated re-samples of size \(=\) 100 for p%. Sample restricted to those who received their Ph.D. degrees prior than 1992

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Cowan, R., Rossello, G. Emergent structures in faculty hiring networks, and the effects of mobility on academic performance. Scientometrics 117, 527–562 (2018). https://doi.org/10.1007/s11192-018-2858-8

Download citation

  • Received:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11192-018-2858-8

Keywords

  • Academia
  • South Africa
  • Faculty hiring network
  • Institutional prestige
  • Institutional stratification
  • Scholars research performance
  • University system
  • Matched pair analysis

JEL Classification

  • D7
  • I2
  • J15
  • O3
  • Z13