Evidence about economies of scale and scope in biomedical and health research
In principle, economies of scale and/or scope in biomedical and health research could come from two main sources. They could result from infrastructure, such as expensive research equipment or costly-to-acquire skills or support services (legal, financial, managerial, administration) that can be provided more cost effectively where a large group of people are making use of that infrastructure and are in the same location. Or they could result from the interactions between researchers, enabling them to work together better and draw on advice from colleagues. Diseconomies of scale and/or scope might result from growing difficulties of coordination when research groups employ larger numbers of people or wider varieties of different kinds of researchers (with different expertise and interests and ways of working). The quantitative analyses we found did not attempt to identify the relative importance of the possible sources of any economies or diseconomies that were detected.
We found 60 papers from the literature review discussing, in at least part of their content, whether there exist economies of scale, or economies of scope, or both, in publicly funded biomedical and health research. Table
2 shows which of the 60 papers concerned economies of scale, which concerned economies of scope, and which concerned both scale and scope. Seven of the 60 papers included discussion of factors affecting the productivity of biomedical and health research, particularly the effect of co-location of researchers, but contained no empirical analysis of economies of scale or scope (last row of Table
2). Of the other 53 papers, 31 look only at economies of scale, two look only at economies of scope, and 20 cover both scale and scope (Table
2). Thus, scale was frequently analysed independently of scope, but scope was seldom analysed in isolation. Methodological approaches are for the most part quantitative, with only a few cases of qualitative, literature reviews and mixed methods studies (Table
2).
Table 2
Articles included in the literature review
Analyses of economies of scale only | Adams & Griliches [ 39], Bauer et al. [ 34], Bonaccorsi & Daraio [ 68], Bonaccorsi & Daraio [ 38], Bordons & Zulueta [ 69], Bordons et al. [ 70], Cohen [ 71], Cohen [ 72], Gomes et al. [ 73], Heale et al. [ 35], Hoare [ 43], Johnes & Johnes [ 74], Kenna & Berche [ 33], Kenna & Berche [ 44], King [ 50], Kretschmer [ 32], Mamun [ 75], Nag et al. [ 76], Rey-Rocha et al. [ 77], Sav [ 36], Schubert [ 22], Stankiewicz [ 78] | Cohen [ 79], Heinze et al. [ 80], Johnston [ 81], Rhoten [ 40], Stokols et al. [ 82] von Tunzelmann et al. [ 85] (This is a pure theoretical analysis) | Adams et al. [ 83], Carayol & Matt [ 84] |
Analyses of economies of scope only | Cherchye et al. [ 54], Glass et al. [ 55] | | |
Analyses of both economies of scale and economies of scope | Agasisti & Johnes [ 14], Chavas et al. [ 21], Cohn et al. [ 46], de Groot et al. [ 48], Dundar & Lewis [ 52], Foltz et al. [ 41], Glass et al. [ 45], Hinze et al. [ 53], Johnes & Johnes [ 15], Johnes & Salas-Velasco [ 87], Johnes & Schwarzenberger [ 16], Laband & Lentz [ 47], Mamun [ 19], Olivares & Wetzel [ 51], Olson [ 88], Sav [ 37], Seglen & Aksnes [ 89], Spanos & Vonortas [ 31], Wolszczak-Derlacz & Parteka [ 49] | | |
Analyses of topics closely related to economies of scale or scope (e.g. co-location) | Abramovsky et al. [ 91] (effect of co-location, R&D labs and university) | Atkinson et al. [ 92] (change in scientific collaboration along time) | Beise & Stahl [ 93] (effect of co-location (private and public)) |
Antonio-García et al. [ 94] (analysed the effect of size through the satisfaction with resources) | Bos et al. [ 95] (co-location of the team members) |
Chin-Tsai & Chang-Tzu [ 96] (how research performance can be assessed) |
Coen et al. [ 97] (integrating the tangible and intangible structures that underlie research centre functioning (co-location)) |
In the following sections we set out the findings of these papers in more detail, considering in turn economies of scale and then economies of scope.
Economies of scale in biomedical and health research
The results of the 51 papers evaluating economies or diseconomies of scale in biomedical and health research or in other types of research that include or overlap with biomedical and health research (listed in rows 1 and 3 of Table
2), are far from uniform, although positive economies of scale are found more often than diseconomies of scale. Table
3 summarises the findings from these 51 papers. Forty-two of the 51 papers offer in essence a single overall conclusion about (dis-)economies of scale in biomedical and health research (or offered no clear conclusion on this point). The other nine of the 51 studies were designed to be able to identify (dis-)economies of scale in a number of separate situations and all of them found different results according to the situation of interest.
Table 3
Numbers of papers finding (dis-)economies of scale in biomedical and health research – by type of research group whose productivity is being analysed
Diminishing returns to scale/Diseconomies of scale | 2 | 3 | 1 | 3 | – | 9 |
Increasing returns to scale/Economies of scale | 1 | 3 | 2 | 9 | – | 15 |
Constant returns to scale | 1 | 3 | 1 | – | 2 | 7 |
Inverse U-shape | – | 2 | – | 1 | – | 3 |
Not clear | 2 | 1 | 3 | – | 2 | 8 |
Sub-total | 6 | 12 | 7 | 13 | 4 | 42 |
Articles that suggest both diseconomies and economies of scale |
Multi-product cost functions applied | – | – | – | 4 | – | 4 |
Other methodology applied | – | 1 | – | 4 | – | 5 |
Total | 6 | 13 | 7 | 21 | 4 | 51 |
There is no dominant overall conclusion; both economies and diseconomies of scale are reported in the literature. Rather more papers report positive economies of scale (15/42) than report diseconomies (9/42) or constant returns to scale (7/42). Three papers find that, while there are positive economies initially as scale increases, these eventually turn to diseconomies after a certain scale is reached. Spanos and Vonortas [
31] detect economies of scale, measured as the number of partners in European research projects, up to 25 such partners but diseconomies of scale thereafter. While Kretschmer [
32] found that diseconomies might set in for research group sizes above 6–12 members, Kenna and Berche [
33] suggested the turning point might be around 41 group members for medical sciences, 21 for biology and 11 for economics and statistics.
The empirical literature includes studies that focus on a variety of levels of aggregation of research groups, from individual research projects under a principal investigator to whole universities or large research institutes. In Table
3, the results are broken down according to the types of ‘research group’ studied. We use the term ‘research group’ to apply to any team of people working together on research, or to an agglomeration of such teams. We have defined four ‘research group’ types according to the type of unit whose productivity is being analysed in a study. For instance, those papers classified under ‘University/Research Institute’ are analysing differences in productivity between universities and/or research institutes. Productivity might be measured, for example, as the total number of publications from each university/institute in a particular period of time, while the scale measure could be the average size of the research teams within the university/institute. Similarly, in the case of the ‘Individual level (principal investigator)’ category, those studies are measuring the productivity of a particular principal investigator while the scale measure could be the size of their research team.
Among the 51 papers studying economies of scale, six present analyses in which the unit explored is the researcher (Table
3). The number of studies in this category is small and reveals a spread of findings across all of decreasing, constant or increasing returns to scale. In these six studies, neither multi-product cost functions nor multi-product production functions were used. Bauer et al. [
34] predict the number of publications per researcher as a function of academic status, sex, department size, and quota of senior researchers, and they predict the number of citations by publication count, status, department size, and quota of senior researchers. Similarly, Heale et al. [
35], using information on publications and funding per individual principal investigator, analyse the determinants of scientific knowledge output in biomedical research. They find diminishing returns in relation to the amount of funding available and that a reasonably stable network of external collaborators is key for high productivity.
The 13 studies focused at the ‘Research Team’ and the seven studies focused at the ‘Department’ (sub-university/institute) level reveal a similar spread of results across decreasing, constant and increasing returns to scale (Table
3).
Whole universities or research institutes were the focus in 21 of the 51 studies looking at economies of scale. At that level, nine out of 21 papers show positive economies of scale in research, compared with just three finding diseconomies of scale. For instance, Sav [
36], by applying DEA, found that “
productivity regress was accompanied by decreasing returns to scale that prevailed among more than half of the USA universities”. Nine other studies focused at the university/research institute level find a mixed picture with both increasing and decreasing returns evident in different circumstances.
We focus now specifically on the nine articles studying economies of scale that suggest the existence of both economies and diseconomies of scale, depending on the specific context (Table
3). Four of these nine articles used a multi-product cost function approach and estimated both ray economies of scale and product-specific economies of scale [
15,
16,
19,
37]. Johnes and Schwarzenberger [
16] used panel data that allowed the estimation of a random parameter stochastic frontier model, meaning that parameters could vary across institutions. Based on this methodology the authors found product-specific economies of scale for research in German universities (i.e. where research activity increases but teaching activity remains unchanged the average cost of the research falls) but minimal ray diseconomies of scale (i.e. when research and teaching increase by the same proportion, average costs of both rise slightly).
Two of the nine articles use DEA analysis [
21,
38]. In these cases, for each university/institute analysed, a different measure of economies of scale is estimated, and some of the institutions show diseconomies while others show economies of scale. For instance, of 52 American universities analysed by Chavas et al. [
21], 69% (36/52) showed economies of scale and 31% (16/52) diseconomies. The findings of Bonaccorsi and Daraio [
38] suggest diseconomies of scale in the French university system and a U-shape pattern (diseconomies in smaller and economies in bigger research institutes) in the case of Italian universities.
Two other articles from these nine applied different proxy variables for scale and estimated more than one equation, and found different views of (dis-)economies of scale according to which proxy for research output was used. For instance, Adams and Griliches [
39] found diseconomies of scale in the production of research publications but positive economies of scale in citations.
Finally, the qualitative analysis by Rhoten [
40] indicates that an increase in centre size from small to medium could increase the number of information sharing ties but not knowledge creating ties, while an increase from medium to large does not appear to increase either of the two and so would not imply economies of scale over that range.
Taken as a whole across the group of 51 articles looking at economies of scale, number of publications was the most common proxy measure of research output used – in 26 of the 51 papers – reflecting the comparatively ready availability of publication counts. Patents were used as an output measure in just two papers [
21,
41,
42]. Out of the 26 articles based on publication counts, two included international publications and 10 consider only publications in indexes of refereed journals. Moreover, 11 of the 26 articles took into account the quality of the publications through the numbers of citations.
An additional proxy measure of research output used in the literature is the quality of research, which applied in four of the 51 articles and was measured using results from the United Kingdom’s Research Assessment Exercise (RAE). The RAE classified university research teams into five quality levels by means of peer review [
33,
43‐
45].
Numbers of students (postgraduate and undergraduate usually counted separately) were also quite frequently used as output of the work of a university or department, alongside research. For instance, Mamun [
19] includes as output measures the full-time equivalent undergraduate students and the full-time equivalent masters (MPhil) and PhD students. He also included the research expenditure level as an approximation of the research output. In this case, ray economies of scale were found for the totality of outputs while research and undergraduate teaching show product-specific diseconomies of scale. Postgraduate student numbers are arguably a weak proxy for research output. They may be correlated with research but that correlation might be negative or positive – negative because time spent teaching students is time not spent researching and positive because larger universities, viewed as multi-product firms, may produce both more teaching and more research. Numbers of PhD students may alternatively be seen as an indicator of the scale of inputs to the research process, rather than of the outputs from it. So the use of student numbers as an output is problematic when trying to assess (dis-)economies of scale, or scope, in research.
The extent of grants and other external research funding obtained was used as a proxy for the scale of research output in 11 of the 51 papers. From the perspective of the recipient research group, the ability to win external funding is an indicator of success, but that funding is an input to the research process rather than an output from it when viewed from a societal perspective. When comparing two research groups, A and B, if A wins more research funding than B, we know that it is using more inputs than B but we do not know if that greater funding is being turned into proportionally more or less research output unless we have another proxy measure for output.
Results by type of model used
As shown in Table
2 (second and fourth columns), of the 51 articles found that studied economies of scale (either or alone or along with economies of scope), 45 applied a quantitative (41) or mixed (4) methodology for the analysis of economies of scale. We have classified them according to the three main types of models for assessing economies of scale and/or scope found in our review of the general econometric literature, and found 20 articles did so through selection of proxy variables, 13 through multi-product cost functions, 7 through multi-product production functions, and 5 where the method was unclear or other quantitative approaches had been used.
Among the 20 papers based on the proxy variables approach, the findings were varied – two found positive economies of scale, four found diseconomies of scale, five found constant returns to scale, three found evidence of both economies and diseconomies of scale, and three cases showed no clear results. The results of the remaining three cases indicate an inverse U-shaped relationship between scale and productivity of research – as research group size increases, there are positive economies of scale initially, but eventually they cease and further increases in scale are then associated with diseconomies. For example, Spanos and Vonortas [
31] found an inverse U-shape between number of partners and networking impacts and a U-shaped effect of budget on goal achievement. They measure networking impacts thorough a construct that reflects strengthening links with other research organisations/businesses and dissemination of research results, while goal achievement is measured through three Likert-type scales of the degree to which the project achieved its scientific, technical and commercial objectives. Spanos and Vonortas [
31] found that, with a budget around €1,795,000, the expected value of goal achievement reaches its minimum level, and that networking impacts achieve a maximum point around 24.5 partners. They conclude that the effect of the scale is curvilinear, meaning that, at high levels of scale, the positive returns will begin to diminish. Kenna and Berche [
44] find an inverse U-shape between research quality (measured through the RAE) and the number of researchers in the group. They identify two critical masses, a lower limit, which is the minimum number of researchers needed to ensure stability in the team, and an upper limit after which a greater number of members increasingly obstructs meaningful communication and negatively affects the quality of research. Kenna and Berche [
44] suggest that these upper and lower limits depend on the field of research; for instance, they found an upper limit for the number of individuals in a research group equal to 41 (±8) for medical science, 21(±4) for biology and 11 (±3) for economics and statistics.
Additionally, 13 articles are based on analysis of multi-product cost functions. Two of these studies found diseconomies of scale and seven found positive economies of scale. It is worth mentioning that the results presented by Cohn et al. [
46] have been classified as evidence of positive economies of scale overall, even though, for private universities, the product-specific indicator shows diseconomies of scale in research. This is because the results for the more numerous (and larger on average) public universities show product-specific economies of scale and, moreover, global (ray) economies of scale were found in both public and private universities [
46].
The other four of the 13 multi-product cost function articles show a clear difference between the ray economies of scale indicator and the product-specific indicators. Three of them suggest the existence of global (ray) diseconomies of scale together with positive product-specific economies of scale. However, Mamun [
19] finds the opposite in the case of public universities in a low-income economy, Bangladesh, namely, that there are positive global (ray) economies of scale but product-specific diseconomies of scale for the output ‘research’. However, as research output is measured by research expenditure (which is a measure of inputs, when seen from a societal perspective), it is difficult to interpret this result.
It is also worth noting that all of the 13 multi-product cost function studies use numbers of students (undergraduate and postgraduate students separately) alongside a measure of research as the outputs of universities/institutes of higher education. Eight out of the 13 articles use ‘grants and other research funding/research expenditures’ as a measure of research output and four use the number of publications. For instance, Laband and Lentz [
47] based the estimation of the multi-cost product function on three outputs, (1) the total sum of the federal, state, local and private research grants in millions of dollars, (2) the total number of undergraduate students enrolled in the institution, and (3) the total number of postgraduate students. They found that private as well as public universities are characterised by ray economies of scale and product-specific economies of scale in research and undergraduate education. Patents are used as a research output measure by Foltz et al. [
41] and the quality of the research (as measured by the RAE) is used as an output measure by Glass et al. [
45].
An important part of multi-product cost function estimation is the choice of control variables used. A common practice is to include academic wages (e.g. average faculty salary) as a proxy for the level of input prices [
19,
37,
41,
46,
47]. Likewise, and given the particularly high costs of a medical department, some studies include a dummy variable for those universities or institutions that include a medical school [
14,
37,
41,
48].
Regarding the seven articles that were based on analysis of multi-product production functions, five use DEA [
21,
22,
36,
38,
49]. A particular case in this group is the study conducted by Wolszczak-Derlacz and Parteka [
49], who analysed 259 European universities. They estimated first the efficiency levels of the universities using DEA and then, in a second step, they regressed these efficiency levels on a group of covariates. Among the covariates they included the number of different faculties. Their findings suggest that a university with a higher number of different faculties is more efficient. They consider this result as a possible signal of both economies of scope and economies of scale. Another example is the study by Chavas et al. [
21], who broke down the economies of scope effects into three different parts (complementary, convexity and economies of scale effects). They found that the scale effect is particularly important for the small United States research universities. Only one [
22] out of the five articles that applied DEA has a unit of analysis different from the university. He analysed German research groups from three scientific fields and found that the efficiency curve estimated by DEA shows increasing returns to scale and constant returns to scale. The two articles that do not use DEA are King [
50], who applied a multi-input multi-output model, and Olivares and Wetzel [
51], who used an input distance function approach.
Types of biomedical and health research
It might be expected that the likelihood of economies or diseconomies of scale might differ according to the particular type of biomedical and health research that is being conducted. However, the studies we found provide no basis for drawing conclusions about different sub-headings of research within ‘biomedical and health’ such as basic research versus clinical research, or pharmaceutical versus non-pharmaceutical. Most studies looked at research at a high level of aggregation.
Countries studied
The majority of the studies that analysed economies of scale (33/51) took an individual country as the focus of analysis and most of the rest studied two or more countries internationally. There were few purely sub-national analyses – six of the papers focused on local level analyses of individual research institutes or departments and it was at this level that diseconomies of scale were most likely to be identified (3/6), whereas positive economies of scale were found just once, constant returns to scale were found once and no clear result was found once. Kretschmer [
32] analysed research activity in molecular biology by some 450 scientists in 56 research groups, but the source of data and the country to which the analysis refers are not clearly stated in the paper.
Most of the papers we reviewed in detail looked at research in one or more countries in Europe (24/51) or North America (16/51) or both (3/51). The remaining eight studies were either multinational beyond those regions or concerned individual countries in other regions (e.g. Taiwan), with the exception of Kretschmer [
32], in which it was not possible to identify the country conclusively. Once we exclude those articles whose results indicate both economies and diseconomies of scale, the remaining 43 articles show a pattern of a slightly higher probability of finding economies of scale than diseconomies of scale regardless of geography.
Summary of findings about economies of scale
In summary, the empirical literature is varied both in method and findings, but taken as a whole, implies a greater likelihood of finding positive economies of scale in biomedical and health research than diseconomies, or constant returns to scale, although all three kinds of findings are common. This literature is largely focused at the whole-university or whole-institute level and usually takes a national perspective. Results differ across the wide range of institutions analysed as well as the outputs selected. Taken together, the literature does not permit conclusions to be drawn about different kinds of research within the overall heading of biomedical and health research.
Economies of scope in biomedical and health research
Having discussed economies of scale we now turn to economies of scope. We found 22 papers that presented evidence on (dis-)economies of scope in biomedical and health research. Nearly half of these studies (10/22) found evidence of positive economies of scope. In other words, they found that the total cost of producing multiple research outputs in one place would be lower than the total cost of producing the same volume and mix of outputs in multiple places with each place only concentrating on a single type of output. However, five of the 22 studies found evidence of diseconomies of scope – implying that it would be more efficient to concentrate research in any one place on a narrower range. The remaining seven studies found evidence of both economies and diseconomies of scope in research (Table
4).
Table 4
Numbers of papers finding (dis-)economies of scope in biomedical and health research – by type of research group whose productivity is being analysed
Diseconomies of scope | – | 1 | 4 | 5 |
Economies of scope | 2 | 1 | 7 | 10 |
Evidence for both | – | 1 | 6 | 7 |
Total | 2 | 3 | 17 | 22 |
Studies of economies of scope were most commonly focused at the level of whole universities or research institutes (17/22) (Table
4). Seven of the 17 papers analysing at this level found positive economies of scope in research, but four found diseconomies of scope and the other six showed evidence for both economies and diseconomies. Thus, overall, there is no dominant picture emerging at this level of research group. There were only three studies focused at the level of departments within universities and institutes, and taken together they present mixed results – one study finding economies of scope [
52], one finding diseconomies [
53] and one finding evidence for both [
54]. The two studies focused at the ‘Research team’ level support economies of scope at that level.
Twenty of the 22 studies looking at scope had also assessed economies of scale. Consequently, it is no surprise that the same range of measures of research outputs was evident in studies of scope as in studies of scale. However, when scope was analysed, it was less common for publications to have been used to measure research output. Whereas half of the papers analysing economies of scale (26/51) had included publications as an output, a slightly smaller proportion (9/22) of papers looking at economies of scope did so, two of which considered citations. Contrastingly, 15 of the studies that included scope in their analysis used postgraduate and undergraduate student numbers, separately, as outputs. Nine of the times when student numbers were among the outputs (and not on any other occasions), the amount of research funding raised was also included as an output, on the basis that more productive research groups will attract more funding. Postgraduate student numbers, specifically PhDs and other research-based degrees, might justifiably be seen as research (plus teaching) outputs, though undergraduate student numbers are a proxy for teaching output alone. However, PhD students are also an input to research, as well as a partial proxy for output. Grants and other external funding received are inputs to biomedical and health research as well as outputs from it, so the direction of causation runs both ways.
Although they are crude measures, and are far from representing the full fruits of biomedical and health research, counts of publications and patents are at least outputs from research, and are arguably even better proxies if quality-adjusted in some way. Nine of the 22 studies of economies of scope used publications as their measure of research output, and two of those also used patents (no study used patents without also using publications as a research output). The two studies using both publications and patents as outputs were those of Chavas et al. [
21] and Foltz et al. [
41], both of which are of United States universities and found significant economies of scope in producing publications and patents. Foltz et al. [
41] estimated two regressions, one using only quantitative output measures and another adjusting the quantity of articles and patents according to their quality by considering the number of citations. Significant evidence of economies of scope was found only for the case in which the outputs where adjusted by the number of citations.
The nine studies where publications, alone or with patents, were used as a measure of research output were more likely than the others to show the presence of positive economies of scope – six of the nine did so (including both papers that also used patents as a research output), compared with two showing diseconomies of scope, and one showing evidence of both economies and diseconomies of scope.
Of the 13 (out of 22) studies of scope that did not use publications or patents to measure output, two used instead the quality of research as measured by the United Kingdom Research Assessment Exercise and one [
53] used another quality index. Most of the remainder used postgraduate (and, separately, undergraduate) student numbers, usually in combination with a measure of research funding. Undergraduate student numbers were used as a proxy for teaching output in studies of universities as multi-product firms with the products ‘teaching’ and ‘research’. Postgraduate student numbers were used in a similar way as a proxy for graduate teaching, where that was distinguished from undergraduate teaching. However, as discussed earlier, PhD students can be seen as an input to research activity as well as a proxy for the output of graduate teaching. This makes those findings problematic to interpret, e.g. a research group with more PhD students that produces more publications is not necessarily being more productive per unit of research labour if PhD students are considered in part to contribute to the research labour force.
Results by type of model used
All three of the types of econometric methods described earlier were once again in evidence. Six of the 22 papers concerned with scope used a proxy variables approach, 11 used the cost function method, and five used production functions (Table
5). Overall, the studies were more likely to find positive economies of scope (10/22) than diseconomies of scope (5/22), with seven studies finding evidence of both economies and diseconomies.
Table 5
Numbers of papers finding (dis-)economies of scope in biomedical and health research
Diseconomies of scope | 1 | 3 | 1 | 5 |
Economies of scope | 4 | 5 | 1 | 10 |
Evidence for both | 1 | 3 | 3 | 7 |
Total | 6 | 11 | 5 | 22 |
Types of biomedical and health research
The likelihood of economies or diseconomies of scope might be expected to differ according to the particular types of biomedical and health research that are being conducted. However, the studies we found in the literature are mostly at too high a level of aggregation to permit distinct conclusions to be drawn about the presence of economies or diseconomies of scope for different sub-headings of research within ‘biomedical and health’, such as basic research versus clinical research, or pharmaceutical versus non-pharmaceutical.
Countries studied
As in the case of the literature about economies of scale, the settings for the papers considering economies of scope were mainly Europe (12/22), North America (8/22) or both (1/22), with just one having a setting elsewhere, namely in Bangladesh [
19]. While in Europe the number of articles suggesting economies and diseconomies of scope is the same, in the case of North America no article suggests purely diseconomies of scope. Five of the articles with a focus in North America indicate positive economies of scope and three show evidence for both economies and diseconomies of scope. The focus of studies looking at scope was never at a local, sub-national level.
As mentioned above, only two studies analysed solely economies of scope without mentioning economies of scale. Cherchye et al. [
54] assumed that a multi-output production process is closely related with the concept of economies of scope, since the process depends on the joint use of input to produce a set of outputs. Using a non-parametric characterisation of cost-efficient behaviour, their results suggest that, while the best universities perform well in all areas in which they are involved, less efficient universities can also perform very well in their fields of specialisation. In the case of Glass et al. [
55], they compare more- and less-specialised universities to explore which system yields relatively higher performance. Through the use of non-parametric DEA-based models incorporating financial ratios, the findings indicate that more-specialised university production results in better performance on average than less-specialised production.
Summary of findings about economies of scope
The effect on research outputs of varying the scope of activities by a research group was less often reported in empirical studies than the effect of scale. Taken overall, the empirical literature more often suggests the existence of positive economies of scope in research than diseconomies, but the picture is mixed, perhaps owing to the variety of settings focused on, and the methods used, in the quantitative analyses. The use of postgraduate student numbers and/or external research funding won is problematic as they are inputs to the research process as well as partial proxies for research outputs. Economies of scope were predominantly tested at the whole university/institute level.