The transient and persistent efficiency of Italian and German universities: a stochastic frontier analysis

ABSTRACT

Despite measures on the European level to increase the compatibility between the higher education sectors, the recent literature exposes variations in their efficiencies. To gain insights into these differences, we split the efficiency term according to the two management levels each university is confronted with. We separate short-term and long-term efficiency while controlling for unobserved institution-specific heterogeneity. We argue that the first term reflects the efficiency of the individual universities working within the country, while the second term echoes the influence of the overall country-specific higher education structure. The cross-country comparison displays whether efficiency differences between countries are related to the individual performance of their universities or their higher education structure. This allows more purposeful policy recommendations and expands the literature regarding the efficiency of universities in a fundamental way. Choosing Italy and Germany as two important illustrative examples, we show that the Italian higher education sector exhibits a higher overall efficiency value. With the individual universities working at the upper bound of efficiency in both countries, the remaining inefficiency and the gap between the countries are caused by persistent, structural inefficiency. Future measures should hence aim at the country-specific structure and not solely at the activities of single universities.

KEYWORDS:

• Stochastic frontier analysis
• Persistent inefficiency
• higher education
• Italy
• Germany

JEL CLASSIFICATION:

• C23
• D61
• I22
• H52

Acknowledgments

We presented earlier versions of this paper in several academic conferences: the “XV European Workshop on Efficiency and Productivity Analyses”, the “5^th Workshop on Efficiency in Education”, the “4th LEER Conference on Education Economics”, the “X North American Productivity Workshop (NAPW), and we thank the participants for valuable comments and intensive discussion. We are also grateful to the Editor and to an anonymous reviewer; their suggestions and comments helped us in improving the paper substantially.

Funding: Financial support was provided by the German Academic Exchange Service Program “IPID4all”, mediated by the TU Dresden Graduate Academy, for the preparation of this manuscript.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The term HE sector refers to the whole system of the respective country, including the HE structure as well as the HE institutions.

² Within the present essay, the terms persistent and structural, as well as transient and institutional efficiency, are used interchangeably.

³ For a review of empirical studies utilizing frontier efficiency measurement techniques in education, see Rhaiem (2017), De Witte and López-Torres (2017) and Gralka (2018a).

⁴ Along with Kumbhakar, Lien, and Hardaker (2014) similar models were developed simultaneously by Colombi et al. (2014) and Filippini and Greene (2016).

⁵ A skewness test on the OLS residuals was conducted and provided support for the cost frontier specification of the model.

⁶ For a comprehensive review of the literature and the considered variables see De Witte and López-Torres (2017) and Gralka (2018a).

⁷ Non-science subjects are courses related to art, economics, law, sport and culture. General science contains mathematics, natural sciences, agricultural, forest sciences and engineering. Medicine includes human and health science and veterinary medicine.

⁸ The inclusion of the subject group medicine could lead to a bias of the efficiency results because they are part of the general health provision and therefore exhibit inflated cost. We account for the matter by implementing the subject as a separated group.

⁹ It can be argued that the multi-step method is inefficient relative to a simulated ML estimation method (for a discussion, see Heshmati, Kumbhakar, and Kim 2018). We deliberately choose the former method because of its relatively straight-forward estimation procedure, compared to the simulated ML method, and the possibility to verify the estimation result in every step.

¹⁰ General tuitions fees were introduced in seven of the 16 German states in 2005 but were abolished shortly after, particularly as a result of great public pressure and changes in government. In addition, several states levy special fees for students whose studies are taking longer than the required time and second degrees which are independent of the first degree.

¹¹ The following universities are excluded, mainly because of merges within the time frame: U Duisburg-Essen, Brand. TU Cottbus-Senftenberg, HafenCity U Hamburg, U Kiel, U Lübeck and Università di Camerino, Stranieri di Perugia, Stranieri die Siena, Università di Trento, Sissa Trieste, and Università degli Studi di Urbino Carlo Bo.

¹² Both datasets have been used previously in separate efficiency analyses and were merged for the following analysis; see Agasisti (2011) and Gralka (2018b).

¹³ The differences in the displayed costs compared to the study by Agasisti and Pohl (2012) are due to different definitions of cost. The present study assumes that third-party funding should be excluded from the overall cost because it represents an output.

¹⁴ The choice to cluster into three groups is thereby deliberate. In an analysis regarding the horizontal differentiation of the German HE sector, Erhardt and von Kotzebue (2016) identify three main groups of universities. In line with our results, they also ascertain one large, homogeneous group of institutions, a second smaller one and third containing mainly outliers.

¹⁵ Because of its great size, the university Roma La Sapienza was excluded from the figure but belongs to the first group.

¹⁶ This could be driven, among other factors, by the central government in Italy. Diversity efforts of the federal states in Germany presumably lead to more evenly distributed funding, probably because of risk aversion.

¹⁷ The cluster analysis originally assigned six technical universities to the third group. For the later use of the clusters in the interpretation of the results, we choose to allocate all technical universities to this group.

¹⁸ In fact, the German HE sector contains 15 technical institutions, but the Brand. TU Cottbus-Senftenberg had to be excluded from the analysis due to a substantial merger within the evaluated time period.

¹⁹ The efficiency values for each university are not shown due to space limitations but will be provided by the authors upon request.

²⁰ The study by Badunenko and Kumbhakar (2017) in which the banking sector is assessed, could serve as an example.

²¹ We deliberately included this variation as a sensitivity analysis, for we prefer to keep our baseline model as straightforward as possible to emphasize the advantages and implications of the new efficiency specification.

Additional information

Funding

Financial support was provided by the German Academic Exchange Service Program ‘IPID4all’, mediated by the TU Dresden Graduate Academy, for the preparation of this manusscript.