Abstract
Despite being a well-known concept in neo-institutional organization theory, isomorphism (or structural similarity) has been conceptualized and empirically examined by very few studies. This paper examines isomorphism quantitatively in German higher education, using the Relative Specialization Index. Drawing on a comprehensive data set that includes information on professorial staff and students, as well as basic and grant funding, our paper shows, first, that most variables show an isomorphic pattern, and second, that German universities have become more isomorphic over time. We discuss possible explanations of this result and avenues for future research.
PEER REVIEW
1. INTRODUCTION
This paper examines isomorphism among public universities in Germany. Isomorphism means that units of an organizational field—in this case, the country’s technical and nontechnical universities—are structurally similar or are becoming more similar over time. Possible structural characteristics include university staff, university finances, and student numbers.
Why should we be concerned with this topic? Quantitative analyses in higher education studies have been influenced by Birnbaum’s (1983) concept and measurement of institutional diversity which is anchored in population ecology theory (Aldrich, 1979; Hannan & Freeman, 1977). Many studies on institutional diversity are not part of the research program of population ecology, yet they continue to use the ecology-oriented diversity measure. At the same time, both conceptualization and empirical measurement for isomorphism, a pivotal concept of neo-institutional organization theory, population ecology’s main theoretical contender in the higher education literature (DiMaggio & Powell, 1983; Meyer & Rowan, 1977), is empirically underdeveloped. Few studies have engaged in constructing measures, and even fewer have put theoretical concepts to an empirical test (Croucher & Woelert, 2016; Schofer & Meyer, 2005; Scott & Biag, 2016; Seeber, Lepori et al., 2015; Woelert & Croucher, 2018).
Hence, we start from the observation that there is very little theory-based methodological discussion about how to measure structural similarity in the quantitative higher education and science studies literatures. Based on an extensive literature review (Section 2) and using an exemplary density function of the Relative Specialization Index (RESP), we discuss how isomorphism could be conceptualized (Section 3). This conceptualization is then applied to a comprehensive data set on all German universities, spanning the years 1995 to 2020 (Section 4).
We find that none of the density functions directly mirrors the exemplary isomorphic density functions based on a beta distribution. However, density curves for three (out of four) variables exhibit an isomorphic pattern, and isomorphism in German universities has increased across all variables between 1995 and 2020.
The results of this paper are relevant, first, because there is considerable debate whether universities in Europe have become structurally more similar to each other (Hüther & Krücken, 2016; Zapp, Marques, & Powell, 2021). Second, our findings are relevant for the argument that national research systems characterized by a high level of isomorphism have lower capabilities for scientific breakthroughs. Hollingsworth (2006) concluded, based on an analysis of some 300 biomedical discoveries, that research breakthroughs occurred mostly in the United States and the United Kingdom, two countries that are believed to have low levels of isomorphism. In contrast, countries where the level of isomorphism is believed to be high, such as Germany and France, have much smaller numbers of scientific breakthroughs. While Hollingsworth’s findings regarding different capabilities for breakthrough research have been validated (Heinze, Heyden, & Pithan, 2020), the pervasiveness of isomorphism in higher education has not yet been established for these countries (France, Germany, United States, United Kingdom). Therefore, our paper is an attempt to characterize isomorphism in the German university system and thus create a basis for future comparative analyses.
The paper is organized as follows. In Section 2, we review the literature on diversity and isomorphism in higher education and science studies with special focus on the two dominant theories: population ecology and neo-institutionalism. In Section 3, we introduce an exemplary density function that displays what isomorphic distributions could look like. After a short introduction to our data set and some characterization of public universities in Germany (Section 4), we present our empirical results (Section 5), followed by a short discussion (Section 6).
2. LITERATURE REVIEW
Recent literature reviews (for example, Huisman, 2016; Huisman, Boer et al., 2016) give credit to Birnbaum (1983) for establishing a research tradition in higher education studies that is informed by the ecology perspective in organizational theory (Aldrich, 1979; Hannan & Freeman, 1977; Hawley, 1968). This perspective postulates that higher education systems are populated by forms of organizations (species) in certain environmental constraint spaces (niches). Authors in the ecology tradition are interested in whether or not the institutional ecosystem of higher education is populated by diverse species that offer different educational services to audiences with different needs. Insofar as diversity helps to cater educational needs across a broad spectrum, it is regarded as good and desirable (Birnbaum, 1983; Huisman, 2016). The literature’s focus on institutional diversity is particularly strong regarding phases of growth, in which ecosystems expand. Although one might expect that in such phases, the availability of additional resources would increase both the number of educational niches and species, the ecological perspective predicts the opposite. New resources stimulate competition between existing species in the ecosystem, and because some species are more successful than others in exploiting these resources, not all of them survive (selection), and hence a loss of diversity may be observed.
The prediction of diversity loss during phases of growth has received empirical support. Beginning with Birnbaum’s (1983) study, a consistent finding is that ecosystem growth goes hand in hand with declining institutional diversity. Birnbaum compared the years 1960 and 1980, Morphew (2009) the years 1972 and 2002, and Harris and Ellis (2020) the years 1989 and 2014. All three studies show that while the total number of higher education organizations (population) grew considerably during these (partly overlapping) periods of observation, some types of higher education institutions (species) were more successful than others and became more densely populated over time, while other types of higher education institutions either became less populated or withered. Hence, institutional diversity declined.
Two basic conceptualizations for diversity are common in the ecology-oriented higher education literature: first, the number and growth of organizational forms, also called institutional types (species), in relation to the number and growth of higher education organizations (population); and second, the distribution of higher education organizations across institutional types. Regarding the first conceptualization, Birnbaum (1983) defined a multidimensional matrix, using categorical variables that comprised 768 institutional types, 141 of which were populated in 1960 and 138 in 1980; hence his conclusion of declining institutional diversity. Harris and Ellis (2020) found that the number of higher education organizations increased at a greater rate than the number of unique institutional types and concluded from that finding that institutional diversity had declined. In comparison, while Birnbaum (1983) examined absolute growth of institutional types, Harris and Ellis (2020) used relative growth of institutional types as a manifestation of institutional diversity.
Regarding the second conceptualization, Huisman, Meek, and Wood (2007, p. 569) argue that “the more different types of institutions that are found and the more evenly the institutions are distributed across the types, the more diverse the system. A system is maximally diverse when all institutions belong to different types. A system is minimally diverse when all institutions belong to the same type.” In accordance with this characterization, empirical studies have used concentration measures. Birnbaum (1983) examined Lorenz curves, a relative measure of concentration, whereas the more recent literature tends to use the Herfindahl-Index (or variations of it), an absolute concentration measure (Huisman, Lepori et al., 2015; Huisman et al., 2007; Lepori, 2022; Lepori, Huisman, & Seeber, 2012; Seeber et al., 2015; Rossi, 2009). In addition, some authors examine the concentration of higher education organizations in the most populated institutional type (Morphew, 2009) or the top 10% of institutional types (Harris & Ellis, 2020).
As mentioned, several (mostly recent) studies on institutional diversity in higher education systems are not part of the research program of population ecology but have kept the ecology-oriented conceptualization of diversity. For example, Rossi (2009), Lepori et al. (2012), Huisman et al. (2015), and Lepori (2022) use, among other indicators, the Herfindahl-Index (including its inverse) to examine institutional diversity in the Italian (Rossi, 2009), Swiss (Lepori et al., 2012) and European (Huisman et al., 2015; Lepori, 2022) higher education systems. These studies are not informed by the ecology framework; none of them puts ecology-derived hypotheses to an empirical test. Rather, the Herfindahl-Index (and other statistical measures) are used in these studies
to examine whether predefined institutional categories are empirically robust: Lepori et al. (2012) examine the binary distinction between universities and colleges of applied science in Switzerland;
to identify groups of countries with regard to institutional diversity, using several variables, such as students or staff: Huisman et al. (2015) find that Switzerland, Austria, Germany, Denmark, Belgium, and Hungary have diverse higher education systems, whereas Lithuania, Spain, Greece, Ireland, the Czech Republic, and Portugal show low institutional diversity scores; and
to generate typologies for large heterogeneous higher education systems, such as the European one, which is classified by Lepori (2022) using the following typology: research universities, generalist colleges and universities, technical universities, social sciences universities, and specialized colleges.
Given the fact that institutional diversity in higher education has been studied primarily from an ecology-oriented perspective, we argue that higher education studies need also proper conceptualization and measurement for theoretical concepts in the neo-institutional research program. In contrast to the ecology-oriented research program, there are only very few quantitative studies with reference to theoretical concepts in the neo-institutional program. Three of them refer to isomorphism (Croucher & Woelert, 2016; Schofer & Meyer, 2005; Woelert & Croucher, 2018) and one of them refers to complete organizations, a concept developed by Brunsson and Sahlin-Andersson (2000) that postulates three criteria (identity, hierarchy, and rationality) for full organizational actorhood (Seeber et al., 2015). We will return to these studies later in this section, after we have reviewed what is more common in the neo-institutional literature on higher education systems: conceptual papers focused on causal mechanisms.
To begin with, Hüther and Krücken (2016) argue that the literature on institutional change in European higher education has identified a general trend towards structural similarity (i.e., isomorphism) among higher education institutions. First, the authors summarize the literature on “New Public Management” (NPM) that diagnoses, for all national higher education systems: an increase in internal hierarchical decision-making with a simultaneous decrease in collegial, academic self-control; an increase in external control by state-appointed actors (e.g., accreditation agencies) with a simultaneous decrease in bureaucratic control by the state itself; and an increase in competition at all levels. It is important to note that isomorphism as an empirical phenomenon was first addressed in the late 1990s, primarily in the context of NPM. The first studies in this regard used expert interviews to assess the extent of isomorphism, focusing on Australian and Nordic higher education systems where NPM-inspired reforms had been implemented early on (Stensaker & Norgard, 2001; Townley, 1997). In addition, a few studies were published using longitudinal descriptive data to trace the professionalization of administrative staff in higher education systems (Krücken, Blümel, & Kloke, 2013; Stage, 2020).
Second, Hüther and Krücken (2016) discuss the extent to which higher education institutions have become complete organizations. Yet, and most importantly, they reject the central conclusion from these two literatures, namely an increase in isomorphism in the European higher education system. Rather, they claim that the purported increase pertains to formal structures only (which are decoupled from practice to a considerable extent), while the traditional, state-centered structures of higher education systems have retained their nationally defined peculiarities. Even if increasingly similar formal organizational structures were observed, the authors argue, institutional diversity would continue to exist on the behavioral/practice level.
In contrast to this line of reasoning, Zapp et al. (2021) consider the institutional diversity of national higher education systems simply as ceremonial façade under which fundamental processes of structural alignment (i.e., isomorphism) have taken place. In line with the neo-institutional literature mentioned, the authors regard the transformation of higher education institutions into complete organizations with internal and external agency capabilities as the central result of a long-lasting process to which they attribute greater structural importance than the reproduction of national structures in higher education.
Despite their considerable interpretative differences regarding the empirical significance of isomorphism in the European higher education system, both Hüther and Krücken (2016) and Zapp et al. (2021) focus on the three causal mechanisms that have been theorized as generating isomorphism: (state) coercion, normative pressure, and imitation. However, while Zapp et al. (2021) adhere to the tenet of neo-institutional theory that these three mechanisms generate isomorphism, Hüther and Krücken (2016) argue that they could also generate diversity.
Although it might seem surprising that the three mechanisms should explain both isomorphism and its opposite, namely, diversity, Hüther and Krücken’s (2016) claim rests on the idea that the multiple institutional embeddings in European, national, and regional contexts (so-called “nested organizational fields”) could possibly lead to isomorphism on one level (national), but diversity on another level (Europe). This point had already been made on a general level by Beckert (2010), according to whom the three mechanisms (coercion, normative pressure, imitation) are conceptualized to explain isomorphism, but that it would be desirable to more closely investigate the precise mode of operation of the three social mechanisms with regard to diversity.
The considerable disagreement among authors in the neo-institutional research program with regard to whether there is an overall isomorphic trend in higher education systems (and thus, loss of diversity) and the relevant mechanisms is indicative, at least in part, of the lack of conclusive quantitative-empirical research. Our literature search retrieved the four articles mentioned above: three papers examine isomorphism in the Australian (Croucher & Woelert, 2016; Woelert & Croucher, 2018) and global higher education contexts (Schofer & Meyer, 2005), and another paper quantifies the extent to which 26 European universities (in eight European countries) have become complete organizations (Seeber et al., 2015).
The first three papers use the coefficient of variation, with Schofer and Meyer (2005) showing a cross-national increase in isomorphism between 1900 and 2000 with respect to student enrollment; and Croucher and Woelert (2016) as well as Woelert and Croucher (2018) finding a rise in isomorphism with respect to Australian universities’ institutional structures, student enrollment, and the ratio of teaching to research from 1987–1991. The paper on European universities operationalizes actorhood (identity, hierarchy, and rationality) using cross-sectional survey data, primarily on an ordinal scale, and the Herfindahl-Index for disciplinary specialization. However, they find no consistent empirical support for increased actorhood of European universities (Seeber et al., 2015).
The general paucity of quantitative studies on isomorphism is true for not only higher education research but the neo-institutional research program as a whole. Given the appreciable age of the isomorphism concept (Meyer & Rowan, 1977), the three mechanisms (DiMaggio & Powell, 1983), and the complete organizations concept (Brunsson & Sahlin-Andersson, 2000), it is surprising that there have been so few attempts to conceptualize and measure these concepts quantitatively. This is also true for “decoupling,” another key concept in neo-institutional theory. DiMaggio, one of the founding theoreticians of neo-institutionalism, introduced the conformity index (DiMaggio & Stenberg, 1985), but this measure has not been taken up outside the sociology of music and culture; moreover, recent work in this area has shifted the focus from the organizational field to the actor level, using conventionality and alignment as measures for conformity (Durand & Kremp, 2016; Tamburri, Munn, & Pompe, 2015).
Summing up, we have argued that quantitative analyses in higher education studies examine institutional diversity primarily in the population ecology framework. Yet, while many recent studies have kept ecology-oriented measurements of diversity, they have not contributed to this research program. Rather, many quantitative studies of higher education systems, particularly those with focus on (countries in) Europe tend to discuss methodological questions, in particular with regard to classifications that have little connection to broad theoretical postulates (in population ecology or neo-institutional organization theory). Furthermore, there exist very few efforts to put central theoretical propositions of the neo-institutional research program to an empirical test. In the following section, we outline a possible conceptualization for isomorphism.
3. ISOMORPHISM: METHODOLOGY
The initial theoretical statements of both population ecology (Hannan & Freeman, 1977) and neo-institutional organization theory (DiMaggio & Powell, 1983; Meyer & Rowan, 1977) refer to isomorphism as defined by Hawley (1968, p. 334): “units subject to the same environmental conditions (…) acquire a similar form of organization,” and each unit “tends to become a replica of every other unit.” Hannan and Freeman (1977, p. 939) agree with this characterization when they write: “the diversity of organizational forms is isomorphic to the diversity of environments. (…) Each unit experiences constraints which force it to resemble other units with the same set of constraints.” Similarly, DiMaggio and Powell (1983, p. 149) argue that “isomorphism is a constraining process that forces one unit in a population to resemble other units that face the same set of environmental conditions.”
Differences between population ecology and neo-institutionalism occur not so much on the definition of isomorphism but more with regard to the abovementioned mechanisms that generate it. Whereas Hannan and Freeman (1977, p. 939) focus on competition, by which “nonoptimal forms are selected out of a community of organizations,” DiMaggio and Powell (1983, p. 149) argue that competitive isomorphism “must be supplemented by an institutional view” that includes “1) coercive isomorphism that stems from political influence and the problem of legitimacy; 2) mimetic isomorphism resulting from standard responses to uncertainty; and 3) normative isomorphism, associated with professionalization.”
In addition, population ecology differs from neo-institutionalism in that it builds on the idea of ecosystems (institutional environment) in which niches (resource and constraint spaces) are populated by species (organizational forms) that compete against each other. The equivalent concepts in neo-institutional theory are organizational field (ecosystem), structural adaptation (competition), and legitimacy (niche fitness). In such fields, organizations observe each other and tend to become alike (or isomorphic) over time. In other words, while population ecology focuses on competition between distinct organizational forms to survive in their resource environment, neo-institutionalism centers around structural adaptation between organizations to obtain legitimacy within an institutionally defined organizational field (DiMaggio & Powell, 1983; Hannan & Freeman, 1977).
These theoretical differences are reflected in different methods. On the population ecology side, Birnbaum (1983) examined how many of the 768 possible niches were populated by distinct organizational forms in the United States higher education system in 1960. He found that 141 niches were filled, with some niches more populated than others. Because the number of populated niches decreased from 141 to 138 between 1960 and 1980, although the number of higher education organizations had substantially grown, he concluded that the system had become more isomorphic: “It appears that the higher education system has used the vast increase in resources primarily to replicate existing forms (…) rather than to create new ones” (Birnbaum, 1983, p. 144). Based on this method, it needs just one value change in one variable (such as enrollment size, changing from “small” to “midsized”) for new niches to emerge. Had all 768 possible niches been populated by one organizational form each, diversity would have been at its maximum, but if only one niche had been populated by all higher education entities, diversity would have been minimal (Birnbaum, 1983).
On the neo-institutional theory side, the method used (so far) is different. The abovementioned studies (Croucher & Woelert, 2016; Schofer & Meyer, 2005; Woelert & Croucher, 2018) employed the coefficient of variation (CoV) which scales the standard deviation of a variable using its arithmetic mean. The CoV captures differences between organizations in dimensional terms: It measures the degree to which organizations diverge from the average of the organizational field. In other words, the CoV measures the degree of structural similarity (isomorphism) between entities in higher education systems (organizational fields).
The organizational field of higher education may include not only organizations that offer educational services but also those that provide funding (public agencies, private foundations) and/or regulate the field (accreditation agencies or parliaments). Hence, field boundaries can be defined flexibly depending on the particular research question. In the case of Lepori (2022), who is interested in entities that offer educational services in Europe, the boundaries are defined narrowly in institutional terms. As we will show later, our focus is narrow as well, focusing on structural similarity within two groups of education providers: nontechnical and technical universities in Germany.
The CoV is clearly not the only possible conceptualization of how gradual differences between entities in an organizational field can be measured. The higher education literature has suggested other measures, yet mostly without explicit reference to neo-institutional organization theory, and thus without the potential to engage in a theory-guided methodological discussion about measuring isomorphism in higher education. One such measure is the Activity Index (AI) and its normalized cousin, the Relative Specialization Index (RESP). Both measures were used by Rossi (2009), who found that Italian universities are underrepresented in the natural, technical, and medical sciences but overrepresented in the arts and humanities, as well as in the social sciences. Similarly, these measures were used by studies on disciplinary profiles of universities in the Nordic countries and Germany (Heinze, Tunger et al., 2019; Piro, Aldberg et al., 2017).
The example just described focuses on a specific discipline (e.g., mathematics) at one university and compares it with the average of the discipline at all universities. However, if one wants to illustrate the distribution of a discipline (e.g., mathematics) across all universities, density plots (Figure 1) are needed.
Isomorphic distribution (one discipline, 50 universities). Exemplary density functions of a given variable, such as number of professors (left) and corresponding RESP values (right). y-axis: density; x-axis: share (left) and RESP values (right). Density functions follow a beta distribution.
Isomorphic distribution (one discipline, 50 universities). Exemplary density functions of a given variable, such as number of professors (left) and corresponding RESP values (right). y-axis: density; x-axis: share (left) and RESP values (right). Density functions follow a beta distribution.
The left side of Figure 1 shows the density of a beta distribution with parameters α = 15 and β = 85. The expected value of such a distribution is equal to α/(α + β) = 0.15. Let us assume that the share of a particular discipline (e.g., mathematics) for a variable (e.g., number of professors) is independent and identically beta-distributed in this way at 50 universities (e.g., nontechnical universities). This means that most shares of this variable for the specific discipline are close to 15%. How would the associated RESP values then be distributed? Assuming that all universities have the same size for the discipline, a Monte Carlo method (4,000 iterations) was used to generate 50 random numbers for the shares that follow the above beta distribution in order to calculate the corresponding RESP values. The result is shown on the right side of Figure 1, where the density function of the RESP values looks similar to a normal distribution but tapers off somewhat more steeply to the right. This result could be termed “isomorphic distribution,” because many observations are found around the RESP value zero, with decreasing frequency of observations to the left and right, producing what could be called an inverted V-shape.
Isomorphic distribution (20 disciplines, 50 universities). Exemplary density functions of a given variable, such as number of professors (left) and corresponding RESP values (right). y-axis: density; x-axis: share (left) and RESP values (right). The common density function follows a Dirichlet distribution.
Isomorphic distribution (20 disciplines, 50 universities). Exemplary density functions of a given variable, such as number of professors (left) and corresponding RESP values (right). y-axis: density; x-axis: share (left) and RESP values (right). The common density function follows a Dirichlet distribution.
For both Figure 1 and Figure 2, the RESP density functions represent an “isomorphic distribution” because observations peak around the RESP value zero and then taper off with smaller frequencies to the left (RESP values below zero) and right (RESP values above zero). These density functions look like inverted V-shapes with decreasing frequencies to the left and right. First, this inverted V-shape can vary in height and width. Second, when we focus on selected value intervals, we can measure the frequency of observations that fall within each interval. Consequently, when examining such V-shapes over time, it becomes possible to make statements about whether isomorphism has increased or decreased. For example, consider the symmetric RESP interval of [−50, 50] in Figure 2’s inverted V-shape. If 60% of all empirical observations fall within this interval in t1, but 70% in t2, isomorphism had increased. Why is this so? As more observations cluster within the symmetric interval of the inverted V-shape in t2, more observations are closer to the expected value of the RESP distribution, and thus closer to the average of the organizational field.
Based on this consideration, a methodological rule can be formulated for how differences in the degree of isomorphism of a given variable (e.g., number of professors, students, etc.) can be empirically determined. The rule states
When the density curve of RESP values approximates an inverted V-shape and the number of observations within a symmetrically defined interval (such as [−25, 25] or [−50, 50]) around the RESP expectation value increases between t1 and t2, the distribution has become more isomorphic. The higher the density of the V-shaped curve within a given symmetric interval, the more observations are closer to the field average (isomorphism).
Proposition 1: German universities exhibit isomorphic disciplinary structures.
Proposition 2: Disciplinary isomorphism in German universities increases over time.
4. UNITS OF ANALYSIS, VARIABLES, AND OBSERVATION PERIOD
Our analysis focuses on German public universities. According to Germany’s Federal Statistical Office (StBA), 102 universities are permitted to award doctoral degrees (82 public, 20 private). However, private universities play only a minor role given their low share of all enrolled students, the very limited range of disciplinary fields offered, and the low level of research activities (Hüther & Krücken, 2018). They are therefore not considered further here. There are eight public universities with an extremely narrow disciplinary profile which are clearly not suitable for comparison with universities that offer a broad range of disciplines; in addition, five other public universities exhibit significant data gaps and could not be considered further (for details, see Heinze et al. (2019)). Our analysis thus covers 68 public universities. Of these, 17 were classified as technical (TUs, Table S1 in the Supplementary material), either because they are members of the TU-9 association or because they include the terms “Technische Universität” or “Technische Hochschule” in their name. We consider the other 51 universities to be nontechnical universities (NTUs, Table S2 in the Supplementary material).
Data on staff, finances, and student numbers were obtained directly from the StBA (1992–2016, 2022). They correspond to the published data from the reports in Series 11: Education and Culture, Sections 4.1 (Students at Universities), 4.4 (Personnel at Universities), and 4.5 (Finances at Universities). These data were processed at the level of universities and their teaching and research departments. Medicine was excluded because separating hospital units from their affiliated university institutions was not possible for all years (Text S1 in the Supplementary material). For students, the StBA records the target degrees grouped into accumulated higher-level categories, of which we use the categories university degrees, bachelor’s and master’s degrees, and teaching examinations (Table S3 in the Supplementary material).
The data from the StBA are available differentiated by teaching and research areas. However, these are too finely differentiated to determine the disciplinary orientation of a university, as between 30 and 55 subject fields are found per university. We therefore use the somewhat coarser classification system that was introduced by Fuchs and Heinze (2023), which comprises 17 subject categories for NTUs and 14 subject categories for TUs. In sum, we consider 68 universities (51 NTUs, 17 TUs), 17 (NTUs), and 14 (TUs) subject categories, and the time period from 1995 to 2020.
Statistically reliable data on the entire German higher education system, combining finances, students, and staff by universities and disciplines, have been available since 1992. Comparisons for the period before reunification are not possible due to the lack of official data. The year 1995 was chosen because it followed the last East German university founding: University of Erfurt (1994). Data collection was completed in early 2023, so data on staff, finances, and students were available up to 2020, and partially up to 2021. To ensure full comparability of the four variables (Professors, Students, Basic Funding, Grant Funding), 2020 was chosen as the final year. The Excellence Initiative, which ran from 2005 to 2018, was not a significant financial event for the German university landscape, as university third-party funds doubled between 2005 and 2018. The financial volume of the Excellence Initiative, at €4.639 billion (including medicine), represents a modest portion (5.8%) of the total university third-party funds (€79.985 billion, including medicine).
Basic descriptive characterizations of the variables are provided in Tables 1 and 2). First, the average number of professors across all subject categories has slightly increased in TUs, and somewhat more in NTUs (Table 1), with some variation across subject categories (Table 2). Second, the number of students has increased substantially, more in TUs than NTUs (Table 1), again with considerable variation across subject fields (Table 2).
Descriptive characterization of NTUs and Tus. Mean values of all subject categories of a given year, for NTUs and TUs separately, along with respective significance levels of two-sample T-test (between 1995 and 2020). Basic funding and grants are given in thousands of euros; they are adjusted for inflation (base year: 2010)
. | 1995 . | 2020 . | p-value . |
---|---|---|---|
NTUs | |||
Professors | 16.7 | 18.7 | 0.010 |
Students | 1,045.3 | 1,325.5 | 0.000 |
Basic funding | 6,118.1 | 6,794.3 | 0.071 |
Grants | 1,370.2 | 3,733.6 | 0.000 |
TUs | |||
Professors | 20.1 | 20.3 | 0.842 |
Students | 1,165.5 | 1,719.3 | 0.000 |
Basic funding | 9,414.5 | 10,404.4 | 0.348 |
Grants | 3,826.8 | 10,868.5 | 0.000 |
. | 1995 . | 2020 . | p-value . |
---|---|---|---|
NTUs | |||
Professors | 16.7 | 18.7 | 0.010 |
Students | 1,045.3 | 1,325.5 | 0.000 |
Basic funding | 6,118.1 | 6,794.3 | 0.071 |
Grants | 1,370.2 | 3,733.6 | 0.000 |
TUs | |||
Professors | 20.1 | 20.3 | 0.842 |
Students | 1,165.5 | 1,719.3 | 0.000 |
Basic funding | 9,414.5 | 10,404.4 | 0.348 |
Grants | 3,826.8 | 10,868.5 | 0.000 |
Descriptive characterization of selected subject categories at NTUs. Mean values of selected subject categories at NTUs, along with respective significance levels of two-sample T-tests (between 1995 and 2020). Basic funding and grants are given in thousands of euros; they are adjusted for inflation (base year: 2010)
. | 1995 . | 2020 . | p-value . |
---|---|---|---|
Computer Science, Electrical Engineering | |||
Professors | 11.9 | 19.8 | 0.002 |
Students | 712.2 | 1,707.8 | 0.000 |
Basic funding | 5,089.2 | 6,755.9 | 0.163 |
Grants | 1,498.7 | 6,893.9 | 0.000 |
Mathematics | |||
Professors | 18.1 | 17.4 | 0.690 |
Students | 589.5 | 905.2 | 0.008 |
Basic funding | 4,225.3 | 4,890.6 | 0.157 |
Grants | 492.6 | 1,884.6 | 0.000 |
Social Sciences | |||
Professors | 14.4 | 20.2 | 0.044 |
Students | 990.5 | 1,533.3 | 0.008 |
Basic funding | 3,530.3 | 5,231.4 | 0.022 |
Grants | 675.2 | 2,723.3 | 0.001 |
Economics, Engineering Economics | |||
Professors | 20.6 | 29.8 | 0.002 |
Students | 2,544.2 | 2,572.2 | 0.938 |
Basic funding | 7,686.3 | 9,377.3 | 0.148 |
Grants | 678.5 | 3,051.3 | 0.000 |
Physics, Astronomy, Geosciences | |||
Professors | 31.2 | 33.6 | 0.586 |
Students | 993.1 | 1,367.2 | 0.056 |
Basic funding | 12,553.7 | 11,350.7 | 0.522 |
Grants | 5,786.1 | 12,960.6 | 0.000 |
Biology, Chemistry, Pharmacy | |||
Professors | 36.4 | 41.5 | 0.269 |
Students | 1,345.1 | 1,968.9 | 0.006 |
Basic funding | 19,997.9 | 21,588.1 | 0.569 |
Grants | 5,885.5 | 14,255.7 | 0.000 |
. | 1995 . | 2020 . | p-value . |
---|---|---|---|
Computer Science, Electrical Engineering | |||
Professors | 11.9 | 19.8 | 0.002 |
Students | 712.2 | 1,707.8 | 0.000 |
Basic funding | 5,089.2 | 6,755.9 | 0.163 |
Grants | 1,498.7 | 6,893.9 | 0.000 |
Mathematics | |||
Professors | 18.1 | 17.4 | 0.690 |
Students | 589.5 | 905.2 | 0.008 |
Basic funding | 4,225.3 | 4,890.6 | 0.157 |
Grants | 492.6 | 1,884.6 | 0.000 |
Social Sciences | |||
Professors | 14.4 | 20.2 | 0.044 |
Students | 990.5 | 1,533.3 | 0.008 |
Basic funding | 3,530.3 | 5,231.4 | 0.022 |
Grants | 675.2 | 2,723.3 | 0.001 |
Economics, Engineering Economics | |||
Professors | 20.6 | 29.8 | 0.002 |
Students | 2,544.2 | 2,572.2 | 0.938 |
Basic funding | 7,686.3 | 9,377.3 | 0.148 |
Grants | 678.5 | 3,051.3 | 0.000 |
Physics, Astronomy, Geosciences | |||
Professors | 31.2 | 33.6 | 0.586 |
Students | 993.1 | 1,367.2 | 0.056 |
Basic funding | 12,553.7 | 11,350.7 | 0.522 |
Grants | 5,786.1 | 12,960.6 | 0.000 |
Biology, Chemistry, Pharmacy | |||
Professors | 36.4 | 41.5 | 0.269 |
Students | 1,345.1 | 1,968.9 | 0.006 |
Basic funding | 19,997.9 | 21,588.1 | 0.569 |
Grants | 5,885.5 | 14,255.7 | 0.000 |
Third, the highest growth is observed in grant funding, which almost tripled during the observation period (Table 1), with some subject categories having even higher growth, including Social Sciences (quadrupled) (Table 2). Finally, the category with least growth is basic funding, reaching roughly 10% over 25 years in both NTUs and TUs (Table 1) with some subject categories even undergoing decline, such as Physics, Astronomy, Geosciences (−10 %).
5. EMPIRICAL RESULTS
Based on considerations regarding the exemplary isomorphic distribution (Figures 1 and 2), we examine four variables to determine the extent to which subject categories at German universities are isomorphic (Figure 3) and to explore changes in RESP distributions over time (Figure 4). It is important to note that density functions display all subject categories combined. In other words, for professors in 1995, the graph lines for NTUs and TUs represent all observations across all subject categories (Figure 3).
Density functions of RESP values. Density functions shown are obtained using Gaussian kernel density estimations for the four variables Professors, Students, Basic funding, and Grants, separately for the two university groups and comparing the years 1995 and 2020; y-axis: density; x-axis: RESP values. Values for 1995 are derived from the average of 1994, 1995, and 1996, while the values for 2020 are derived from the average of 2019, 2020, and 2021.
Density functions of RESP values. Density functions shown are obtained using Gaussian kernel density estimations for the four variables Professors, Students, Basic funding, and Grants, separately for the two university groups and comparing the years 1995 and 2020; y-axis: density; x-axis: RESP values. Values for 1995 are derived from the average of 1994, 1995, and 1996, while the values for 2020 are derived from the average of 2019, 2020, and 2021.
Three results are noteworthy. First, none of the density functions strictly follows the exemplary density function based on a beta distribution (Figure 2). In addition, we conducted (for each variable and year) Kolmogorov-Smirnov tests for beta distribution, all of whom were rejected (Table S6 in the Supplementary material). The deviation from the exemplary density function arises because there are numerous subject categories with very low RESP values at the left end of the density functions, (i.e., in the RESP value range of [−100, −75]).
A second result is that, despite this deviation in RESP’s lower value range, the density curves for professors, students, and basic funding exhibit a pattern somewhat similar to the exemplary distribution (Figure 2), as most observations are clustered near the RESP value zero with decreasing density toward lower [−100, 0] and higher [0, 100] RESP values. Specifically, the density curves for professors and basic funding are closest to the exemplary distribution. In contrast, density functions for grants significantly deviate from those of the other three variables. Here, observations are distributed in a nonisomorphic fashion with considerable fluctuations across the entire RESP value range. Disciplinary shares and their respective RESP values can also be examined at the subject field level, such as for Physics, Astronomy, and Geosciences (Figure S1 in the Supplementary material).
A third result can be derived from comparing RESP values within specific value ranges (Figure 3). For this purpose, we consider two center intervals [−25, 25] and [−50, 50], as well as two complementary intervals at outer values ranges: [−100, −75] and [75, 100]. Both in NTUs and TUs, the density within the center intervals increased by several percentage points for all variables between 1995 and 2020 (Figure 3). For example, the density for NTU professors in the interval [−50, 50] increased from 66.9% in 1995 to 73.2% in 2020, while for TU professors it rose from 64.1% to 73.5%. Correspondingly, the density in both the lower and upper value ranges decreased over time: Fewer observations fall into these outer areas of the distribution (Table S7 in the Supplementary material).
Overall, these findings suggest an increase in isomorphism across all variables: observations that were in the outer areas of the distribution in 1995 moved closer to the field average. This result is corroborated by a set of linear regressions (for each interval and all variables see Table S8 in the Supplementary material), confirming the trends for both the outer intervals and the inner intervals, as shown in Figure 3. Even for grants, which, as previously explained, do not exhibit an isomorphic distribution overall, this trend can be observed: the density for NTU grants in the interval [–50, 50] increased from 34.1% in 1995 to 43.4% in 2020, while for TU grants, it rose from 38.3% to 51.2%. Grants appear to be more isomorphic in 2020 than they were in 1995.
6. DISCUSSION
In this final section, we discuss the paper’s findings in the light of the literature review, with emphasis on how the paper contributes to a theory-based methodological discussion on diversity and isomorphism in higher education and science studies.
Section 2 concluded that quantitative analyses in higher education studies examine institutional diversity primarily using the ecology-oriented framework, yet typically with little direct empirical tests that would contribute to the population ecology research program. Similarly, very few quantitative studies have put the isomorphism proposition, formulated by neo-institutional theory, to a quantitative-empirical test. This situation makes it difficult to appraise their explanatory power with regard to structural change in higher education systems. Section 3 addressed this issue: It builds on an existing set of specialization indices (AI, RESP) and uses exemplary density functions to display an isomorphic configuration. Section 5 presented relevant descriptive findings.
Our paper is a first (and small) step to closing the gap between theoretical postulates, such as those on isomorphism put forward by neo-institutional organization theory, and full-scale quantitative assessments of such claims. More specifically, we probe German universities for whether isomorphism can be detected. Based on exemplary density functions and longitudinal analysis of value intervals, we find that three out of four variables show a pattern with some resemblance to the isomorphic configuration, as specified in Figures 1 and 2. In addition, we find that the number of observations around the average of the organizational field has increased in all four variables, suggesting that isomorphism has grown during our observation period.
What would be possible explanations for the differences we observe between professors and basic funding on the one hand, and grants on the other hand, with regard to the degree of isomorphism detected here?
First, one possible explanation could be that the two variables, professors and basic funding, are exclusively financed by the federal states and thus by public authorities. In contrast to third-party funds, which are financed by both public and private sources, the state influence on professors and basic funding is greater. In neo-institutional organizational theory, state influence is one of the three mechanisms that create isomorphism: “(state) coercion.” This means that the more pronounced isomorphism in professors/basic funding compared to third-party funds could indicate that state influence creates such isomorphism in the former variables, while its influence on third-party funds is less pronounced and thus causes less isomorphism.
However, such a theoretically grounded consideration only becomes an empirical explanation when the mechanism of “(state) coercion” is operationalized as a variable, measured, and then tested for its explanatory power, for example, in a (multivariate) regression analysis. The dependent variable in such a regression analysis could be the percentage value within the [−50, +50] or [−25, +25] interval, that is, the range of values that includes particularly many isomorphic units of analysis. The independent variable could be “(state) coercion” as the extent of state regulation. So far, however, such data to directly measure “(state) coercion” is not readily available. A possible example of how to measure the level of state influence could be the European University Association’s (EUA) scoreboard on organizational, financial, staffing, and academic autonomy, which is available for the federal states of North Rhine-Westphalia, Hesse, and Brandenburg, leaving most Länder states uncovered (EUA, 2023; Pruvot, Estermann, & Popkhadze, 2023).
Second, our paper addresses another issue that is important for future research: a direct confrontation between population ecology and neo-institutional theory using the same data. Based on our findings, one may argue that because grants are allocated in a competitive fashion and population ecology claims that competition produces isomorphism (equivalent to loss of diversity), our findings prove population ecology wrong. We have no final answer for this debate here, but we would like to encourage future studies to address this issue.
The more general point here is that there should be more discussion about the empirical validity of particular theoretical claims, either of population ecology or neo-institutionalism, in particular with regard to institutional diversity and isomorphism. Clearly, we do not argue here that our finding on grants, in a strict sense, proves population ecology’s claim wrong that competition produces isomorphism. Such a conclusion would require much deeper investigations, based on sophisticated analyses with longitudinal regression analyses. Yet our findings could be used to probe whether competitive grants follow an isomorphic pattern in other national higher education systems. If such data were collated and examined, population ecology’s claim about the role of competition in generating isomorphism could be put under a broader and thus more conclusive empirical test. In addition, such results could be squared with findings from Aghion, Dewatripont et al. (2010) that organizational autonomy and competition are key elements in explaining scientific productivity.
Third, comparative analyses would also be useful for probing claims of the neo-institutional research program. In the light of what we have argued in the introduction of this paper, two research strategies could be pursued. Comparative analyses could be conducted for those countries, such as the United Kingdom, the United States, and France that have produced most scientific breakthroughs (yet on very different levels) during the 20th and 21st centuries (Heinze et al., 2020), which, in turn, may be related to different levels of isomorphism in their national higher education systems (Hollingsworth, 2006). In addition, this comparison could be extended to smaller countries, such as Switzerland, the Netherlands, Norway, or Denmark (Aagaard & Schneider, 2016; CFA, Technopolis, & NIFU, 2016; van Dijck & van Saarloos, 2017), and thus probe the extent to which isomorphism impinges upon capabilities for breakthrough research.
Finally, the exemplary density functions are helpful in identifying structural patterns in organizational fields of higher education. Hence, we believe that we have outlined a useful heuristic tool for future analyses. The value of this tool is not diminished by the fact that variables do not directly mirror exemplary functions. Rather, our data suggest that university fields should be examined in a multivariate fashion. In addition, this tool could also be applied to characterizing and to comparing different institutional sectors within national higher education systems, including universities of applied science (Fachhochschulen) and publicly funded nonuniversity research centers (e.g., Max Planck Society, Fraunhofer Society, Leibniz Association, or Helmholtz Association).
Our empirical analysis builds on an aggregated disciplinary classification (Fuchs & Heinze, 2023). Although this classification might appear coarse, we believe it is better suited than the classification of Germany’s Federal Statistical Office (Tables S4 and S5 in the Supplementary material). If classifications are too fine-grained, they produce many “zeros” or “missing values” in the matrix, and thus distort the calculation of RESP values. Although we are confident that our classification is well suited to the purpose of our paper, sensitivity and robustness checks are generally recommendable for future analyses on this topic.
ACKNOWLEDGMENTS
We are grateful to Joel E. Fuchs for his contributions to an earlier version of this paper. In addition, we thank the reviewers and editors for helpful suggestions.
AUTHOR CONTRIBUTIONS
Thomas Heinze: Conceptualization, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing—original draft, Writing—review & editing. Rafael Josek: Data curation, Formal analysis, Software, Visualization.
COMPETING INTERESTS
The authors have no competing interests.
FUNDING INFORMATION
No funding has been received for this research.
DATA AVAILABILITY
The data analyzed in the manuscript are available in randomized form: https://osf.io/9pqrc/?view_only=baaf1ae5d45f45589138bed58cec512d.
REFERENCES
Author notes
Handling Editor: Vincent Larivière