Political power of Italian rectors: an analysis of recruitments in the period 2001–2021

ABSTRACT

We examine the political power exerted by Italian rectors by investigating the preferential treatment received by the organisational subunits they belong to in terms of personnel resource allocation. During the rectors’ mandate, their organisational subunits tend to grow significantly more (by ∼9%) than the others. The effect persists even after the implementation of the New public management-inspired policy – the Gelmini reform.

KEYWORDS:

• New public management
• higher education reform
• university resource allocation
• rectors
• political power

Disclosure statement

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

Notes

1 Thirty universities have a university hospital. In 27 universities with a university hospital, medicine is the largest scientific area.

2 The number of SSDs underwent minor changes in the period under scrutiny: it fluctuated between 365 and 371 from 2001 to 2010 and remained equal to 369 from 2011 on.

3 Though we collected data on non-state universities, these latter are characterised by significant differences in statutes, governance and rules.

4 We included observations with the number of faculty members equal to 0 in a university-SSD but only if the same university-SSD had at least one member in another year of our time window. Excluding 0 members’ university-SSDs made the observations drop to 224,253. In our “full sample” estimations, observations dropped to 260,646 due to the exclusion of singleton observations, resulting from the inclusion of multiple fixed effects.

5 The year 2000 was excluded because of frequent missing information about the SSDs.

6 Details of the classification system are provided here: https://www.miur.gov.it/settori-concorsuali-e-settori-scientifico-disciplinari

7 The Breusch and Pagan Lagrangian multiplier test rejects the null that the sample variance of the error term at the university-SSD level is equal to 0. In other words, a random effects model is preferable over a pooled OLS. In addition, a test of overidentifying restrictions was performed using the Stata routine xtoverid (an alternative and heteroschedastic- and cluster-robust version of the Hausman test). The test rejects the null that the overidentifying restrictions of the random effects model are valid, suggesting that the FE model is preferable.

8 Given the use of FE models, our concerns are mostly related to the time-variant component of the unobserved heterogeneity between SSDs.

9 In another estimation (not reported but available from the authors upon request), we exclude all university-SSDs without any rector, with a consequent drop in the number of SSD-university groups from 12,495 to 216. Again, the aim is to increase comparability between SSDs. The estimation results are in line with those presented in Table 2.

10 We tried including different measures related to ARWU and QS rankings. Besides being non-significant, their explanatory power seems limited.

11 As an additional check (not reported but available from the authors upon request), we ‘restrict’ the observation period for the SSDs with a rector. In particular, we include observations related to (i) the focal mandate in which one of the professors of an SSD is a rector and (ii) the mandate immediately before. Restricting the observation window for SSDs with a rector should reduce concerns about some underlying factors that may change in a 20-year window. In other words, we would expect time-invariant factors to be even less relevant and thus our FE model to capture most of the unobserved heterogeneity potentially leading to an omitted variable bias. Indeed, our findings do not significantly change in these additional estimations. The previous models estimated on a matched sample also aim to reduce endogeneity concerns, for example, about rectors’ SSDs being different from other SSDs In addition to increasing the comparability between SSDs, matching an SSD with a rector with the most similar one (in terms of the observed characteristics that affect the probability of having a group member elected as a rector) should also control for the unobserved determinants of both rector election and recruiting dynamics.