Professional Documents
Culture Documents
Metrics of R&D Performance and Management of Public Research Labs
Metrics of R&D Performance and Management of Public Research Labs
net/publication/4048675
CITATIONS READS
56 267
1 author:
Mario Coccia
Italian National Research Council & Yale University
316 PUBLICATIONS 8,781 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mario Coccia on 12 January 2015.
National Research Council of Italy, Institute for Economic Research on Firms and Growth (Ceris-CNR),
Torino (Italy)
Nowadays, the Italian science sector is undergoing a strategic reform due to budget cuts and
there is a need for measuring and evaluating research performance of public research institutes.
This research presents a new measure to assess the scientific research performance of public
research institutes. The new model is successfully applied to 108 public research institutes
belonging to the Italian National Research Council, using data from year 2003 and displays the
laboratories with high/low performance. The results are substantially stronger and quicker to
obtain than those calculated by using conventional indicators. This model supports the policy-
makers, who must decide about the level and direction of public funding for research and
technology transfer.
Introduction
The debate on the reform of the research sector in many European countries has
recently become more important, both due to the reduction in public funding and due to
the domination of the United States and Japan in the field of new technology (SENKER,
2001; HERBST, 2004). Research is discussed both in terms of the method of public
financing, and in terms of scientific production. With their theory of the triple helix,
ETZKOWITZ & LEYDESDORFF (2000) state that nowadays universities and public
research bodies play a fundamental role in the production of inventions and innovations,
necessary for the development of a competitive industrial system, in a society that is
increasingly knowledge-based. The public sector research is, according to SENKER
(2001), defined as civil research in institutions for which the major source of funds is
public, which are in public ownership or control and which aim to disseminate the
results of their research, i.e. the defense research is excluded. The studies on these
institutions in many countries, including Italy (COCCIA & ROLFO, 2002; COCCIA, 2004),
the United Kingdom (HARRIS & KAINE, 1994; SENKER, 2001) and Finland (LUWEL et
al., 1999), show a growing interest in evaluating scientific performance (STAINER &
0138–9130/US $ 20.00
Copyright © 2005 Akadémiai Kiadó, Budapest
All rights reserved
M. COCCIA: Scientific research performance within public institutes
NIXON, 1997). The metrics and evaluation of scientific research (NAGPAUL & ROY,
2003; LEE, 2003) may reflect the interest of the Government in assigning clear
objectives to research laboratories so that they are managed efficiently (MORENO &
TADEPALLI, 2002) in the light of diminishing public funds. In this process of
rationalization, the Government as principal, using the terminology of the theory of
contracts, often has objectives that are in conflict with those of research institutes
(agents), above all due to imperfect information on their activities and behavior
(RADNER, 1987; HART & HOLMSTROM, 1985; LEVINTHAL, 1988). Within the research
system, the Government could have as its objective function the maximization of the
added value of less productive institutions, seeking to level out performances. This
objective can be reached by identifying high performance institutes (HPIs) and low
performance institutes (LPIs). The purpose of this research is to construct a model for
measuring the scientific research performance of public research institutes, using a
systemic approach, which considers financial, scientific and technological indices. This
model is based on a multivariate approach. The output of the model is the research
score, which is a global indicator of the scientific performance of research laboratories.
This model supports the policy-makers, who must decide about the level and direction
of public funding for research and technology transfer.
The next section describes the theoretical framework of the research, followed by
the introduction of a scientometric model. We continue with the analysis of the results
on the institutes of the largest Italian public research body, the Italian National Research
Council (CNR), using data from year 2003. Some management implications conclude
the paper.
Theory
where i = number of the indicator; n(i) = number of the measures of the i-th indicator.
Thus, for each indicator the value is the sum of the weighted measures, so that the index
value for the i-th indicators is:
n (i )
Index Value(IV)= d ia wia
a =1
where wia = weight of the a-th measure of the indicator i; n(i) = number of the measures
of the i-th indicator; dia = value of a-th measure of indicator i. So, the index for each
stage of output is:
n
Index s = IV wis
i =1
where s = stage of the output; n = number of the index values; wis=0 if indicator i does
not bear on stage s. Such an overall index for stage of outputs (e.g. immediate) may also
be named factor. GEISLER (2000) points out:
• Alpha Factor = index of immediate outputs (level of technical expertise,
attractiveness of the S&T Organization);
• Beta Factor = index of intermediate outputs (level of investment in
exploitation of S&T; level of importance of S&T outcomes; climate, and
leadership);
• Gamma Factor = index of pre-ultimate outputs (investment in adoption of
S&T; structure of the industry; strategy and life cycle);
• Omega Factor = index of ultimate outputs (role and importance of S&T;
S&T level of population).
These factors are tools for managers and policy makers when evaluating Science
and Technology (S&T) at different levels. The Beta Factor, for example, describes how
firms absorb the output of science and transform it within their internal processes, while
the Omega Factor describes the total impact of science on society.
Several scholars (HARE & WYATT, 1988) measure the scientific performance of
research bodies by a series of indices, which consider all the organizational aspects of
Research & Development (R&D) bodies, from financial to technological and scientific
points of view. COCCIA (2001) evaluates the performance of public research bodies
using metrics based on the measurement of k indices, which represent the main
activities of the institutes. The model called RELEV (REsearch Laboratory EValuation)
is simple in order to bound the subjectivity. This model considers the following seven
indices for n R&D Institutes:
• = Index of public funding attributed to the research bodies = ( 1, 2, …, n);
• B = Index of self-financing = ( 1, 2, …, n );
• X = Index of personnel in training = ( 1, 2, …, n );
If
X1 = (1 / max ) × i;
X2 = (1 / max ) × i; and so on,
then:
RELEV (i) = 3 – X1 + X2 + X3 + X4 + X5 + 2 X6 +X7
A model (0 - 1) is applied to vector , this has the value 1 if the number of patents is
at least 1, the value 0 if there are no patents. The reason for this is to avoid penalizing
research bodies operating in the field of social or mathematical sciences which do not
produce patents, as instead do other sciences (physics, chemistry...). The RELEV Model
(COCCIA, 2001) measures R&D performance on various dimensions and gives a single
output: the R&D performance score. In fact, the values deriving from the evaluation
function RELEV = ( 1, 2, …, n) are the research performances of the laboratories.
Each i synthesizes the financial, technological, and scientific aspects of the research
institute i-th. The weakness of the RELEV model is that almost all operators have the
same weighting in the function and therefore a second RELEV model is elaborated by
COCCIA (2004), in order to improve the first one. The latter uses the discriminant
analysis with a direct method, i.e., all the output variables. The model is based on the
following discriminant function:
Since international publications generate greater diffusion of knowledge, they are given double
weighting with respect to those published in a domestic context.
The application of Model II on 200 Italian public research institutes, with data from
year 2001, shows that 22.5 per cent of the total institutes are efficient (i.e. they have
high research performance). The discriminant analysis has several applications in
researches on academic institutions, such as the study carried out by HARRIS & KAINE
(1994). They investigate whether it is possible to predict the position of a researcher
group on the basis of independent variables represented by preferences and perceptions
concerning their research and environment.
The research presents the RELEV Model II adjusted, which uses new samples and
the discriminant analysis with the Wilks Method. This model reduces the correlation
among some explanatory variables and represents a more reliable scientometric tool to
assess the research performance of institutes.
The first methodological step of Model II adjusted is to identify two groups of
institutes represented by:
1. A B=
2. x A\B x A, x B.
Once the two sets or groups of the institutions are fixed, the research investigates
whether it is possible to predict the location of an institute, taken from a given
population, within one of the above subsets, A or B, on the basis of key variables.
The following ten variables (Table 1), which concern the main inputs and outputs in
absolute numbers of public research laboratories, are considered independent variables
of the RELEV Model II adj.
and the maximum level of F to remove is 2.71. The complexity and abundance of
calculations, due to the high number of variables, is overcome thanks to the application
of the SPSS statistical package, which provides all the results described and analyzed in
the following section.
Results
The RELEV model II adj. is constructed using data from the Italian National
Research Council (CNR) institutes. The CNR is the body that promotes, coordinates
and disciplines scientific research in Italy, in order to increase the scientific and
technological progress of the country. The institutional scientific activity is carried out
by 108 research institutes, operative bodies whose general aim is to do research in
relation to the programmatic objectives of the Italian government and of the European
Union.
The organizational structure of the Italian CNR is based on laboratories operating in
five scientific fields:
1. Basic Sciences, with 28 new institutes in the fields of mathematics,
physics, and chemistry;
2. Life Sciences, 33 new laboratories in the fields of medicine and biology,
agriculture and molecular biology;
3. Earth and Environmental Sciences, concerning geology, environment, and
habitat, with 10 laboratories in total;
4. Social and Human Sciences, with 19 laboratories in the fields of history,
philosophy, and philology; juridical and political sciences; economics,
sociology, and statistics; cultural heritage;
5. Finally, there is the field of Engineering and Information-Communication
Technology (ICT) Sciences, made up of 18 laboratories.
All data analyzed are taken from the official document of the Italian National
Research Council: Report, year 2003. Set A or HPI of the new methodology has 4
institutes of basic sciences, 4 institutes of life sciences, 1 laboratory of earth sciences,
and 4 institutes of engineering and ICT sciences. Set B or LPI has: 1 institute of basic
sciences, 3 institutes of life sciences, 4 institutes of social sciences and 2 institutes of
engineering and ICT sciences. The selection rule of the RELEV model II adj., i.e.
minimizing Wilks’ lambda, chooses the following five explicative variables:
a) Public funds;
b) Trainees;
c) Number of courses held by researchers in universities and other institutions;
d) International publications;
e) International proceedings.
The prior probability of group A (or 1) is equal to 0.565, while the prior probability
of group B (or 2) is 0.435 (the total probability is 1). The canonic correlation of the
discriminant function (Wilks method) is high and equal to R 2c1=0.9596 (Table 2).
Function 1
Stepwise method of Wilks
Table 4. Structure matrix: pooled within-groups correlations between discriminating variables and canonical
discriminant function (variables ordered by size of correlation within function)
Function 1
Stepwise method of Wilks
This table shows the discriminant variables classified by size of correlation within
the function.
The unstardardized canonical discriminant function is given by:
F1 = 5.1038744 + 0.0874307(Teaching courses) + 0.00000350498206(Public funds)
0.0370964(Trainees) + 0.0171910(International proceedings) + 0.0210310(International pub.)
The appendix shows the discriminant scores and the all-groups stacked histogram of
the analysis.
permanent (JOHNES, 1992), since within the Italian science sectors the gap of scientific
performance between the two sets is low. Moreover, LPIs within the Italian national
system of innovation (LUNDVALL, 1992) are characterized by poor performance due to:
a) environmental causes (reduction in public funding, changes in the technological
trajectories, etc.), b) organizational problems, such as internal conflicts in the choice of
management. These two main causes increase the so-called “X inefficiency”
(LEIBENSTEIN, 1966).
Besides the concentration measures already undertaken by the Italian Government,
another research policy to increase the performance of LPIs could be the re-location of
researchers so that they can choose in which laboratory to work according to their
scientific preferences. Moreover, to achieve the same result, it could be necessary to
perform specific actions that introduce greater incentives, for example: i) awards to
researchers for articles published in international journals (i.e. according to their
scientific quality), ii) high percentages of profit to researchers who have generated new
technology that can be transferred, iii) certain career paths and promotions linked to the
scientific recognition attained, iv) high salary to attract international scientists and to
reduce the migration of the best scholars towards foreign firms and/or universities.
HARRIS & KAINE (1994) show that, in the presence of high motivations and stimulating
environments, researchers continue to publish and are considered high performers, even
after they have reached the peak of their career.
This research confirms the concern that the Italian national system of innovation is
not working satisfactorily and that financial resources are insufficient to strengthen the
Italian scientific network in terms of production and diffusion of scientific researches
and technology in the economic system. The past and present Italian research policy,
based on cuts to public funds, is pushing towards a dangerous reduction of
competitiveness of firms and of the economic system due to low technology transfer. In
order to increase the future economic growth and wealth of the nation, the Italian
government should use the suggestions of the new theory of endogenous growth
(ROMER, 1990) i.e. increasing: 1) Gross Domestic Expenditure on R&D (GERD) as a
percentage of Gross Domestic Product (GDP); 2) GERD per capita population; 3) total
researchers per thousand labor force. However, this research also shows as the
percentage of HPIs (33 per cent) is higher than the percentage in 2001 (COCCIA, 2004),
when HPIs were 22.5 per cent of the total. This result is due to a new Italian research
policy based on the merger of research institutes within the same scientific fields, in
order to increase the economies of scale and to reduce the costs. In fact, the
reorganization of the Italian national research council (since 2002) and the reduction (as
of the mid-nineties) of the public funds assigned to research institutes is pushing the
Italian public research Institutes towards the efficient use of public resources, reducing
the X-Inefficiency. Nowadays, the CNR institutes have a different approach towards the
environment and the market, which are seen as an important financial resource
necessary to carry out research activities, but this phenomenon can transform research
laboratories into organizations focused on consultancy and applied research, with
negative repercussions on basic and fundamental research and, therefore, on the long-
term economic growth of the country.
I wish to thank Prof. Secondo Rolfo and the conference participants at the Indian Institute of Technology
(New Delhi) for helpful discussions and suggestions to a preliminary version of this paper. Particular thanks
go to Prof. Luca Gnan of the Università Commerciale Luigi Bocconi in Milan (Italy) for his useful tuition in
SPSS in company modelling, and to Frabrizio Tuzi (DAST-CNR) for the data set of Cnr. I am grateful to
Diego Margon and Silvana Zelli (Ceris-Cnr) for the research assistance. Thanks also to an anonymous referee
for helpful comments. Any errors are my sole responsibility.
References
ALTMAN, E. I. (1978), Financial application of discriminant analysis: a clarification, Journal of Financial and
Quantitative Analysis, 13 (1) : 185–195.
BALL, R., WILKINSON, R. (1994), The use and abuse of performance indicators in UK higher education,
Higher Education, 27 : 417–427.
COCCIA, M. (2001), A basic model for evaluating R&D performance: theory and application in Italy, R&D
Management, 31 (4) : 453–464.
COCCIA, M. (2004), Models for measuring the research performance and identifying the productivity of public
research institutes, R&D Management, 34 (3) : 267–280.
COCCIA, M., ROLFO, S. (2002), Technology transfer analysis in the Italian National Research Council,
Technovation, 22 : 291–299.
DAST CNR (2003), CNR Report, D’Anselmi (Ed.), Rome.
DIETZ J. S., BOZEMAN, B. (2005), Academic carrers, patents, and productivity: industry experience as
scientific and technical human capital, Research Policy, 34 (3) : 349–367.
ETZKOWITZ, H., LEYDESDORFF, L. (2000), The dynamics of innovation: from National Systems and
“Mode 2” to a Triple Helix of university-industry-government relations, Research Policy, 29 (2) :
109–123.
GEISLER, E. (1995), An integrated cost-performance model of research and development evaluation, Omega,
23 (3) : 281–294.
GEISLER, E. (2000), The Metrics of Science and Technology. Westport, Connecticut, London: Quorum Books.
GUAN, J., WANG, J. (2004), Evaluation and interpretation of knowledge production efficiency, Scientometrics,
59 : 131–155.
HARE, P. G., WYATT, G. J. (1988), Modelling the determinant of research output in British universities.
Research Policy, 17 : 315–28.
HARRIS, G., KAINE, G. (1994), The determinants of research performance: a study of Australian university
economists, Higher Education, 27 : 191–201.
HART, O., HOLMSTROM, B. (1985), Theory of contracts. In: T. BEWLEY (Ed.), Advances in Economic Theory.
Cambridge: Cambridge University Press.
HERBST M. (2004), Governance and Management of Research Universities: Funding and Budgeting as
Instruments of Change. Center for science and technology studies, 4, Bern, CH.
JOHNES, G. (1992), Performance indicators in higher education: A survey of recent work, Oxford Review of
Economic Policy, 8 (2) : 19–34.
LEE, C. K. (2003) A scientometric study of the research performance of the Institute of Molecular and Cell
Biology in Singapore, Scientometrics, 56 : 95–110.
LEIBENSTEIN, H. (1966), Allocative efficiency versus «X-Efficiency», The American Economic Review,
56 : 392–415.
LEVINTHAL, D. (1988) A survey of agency models of Organizations, Journal Economic Behavior and
Organization, 153–186.
LUNDVALL, B. (1992) National Systems of Innovation, Pinter Publishers, London.
LUWEL, M., NOYONS, C. M., MOED, F. (1999), Bibliometric assessment of research performance in Flanders:
policy background and implications, R&D Management, 29 (2) : 133–141.
MASSART, D. L., VANDEGINSTE, B. G. M., DEMING, S. N., MICHOTTE, Y., KAUFMAN, L. (1988),
Chemometrics: A Textbook, Elsevier.
MORENO A. A., TADEPALLI, R. (2002), Assessing academic department efficiency at a public university.
Managerial and Decision Economics, 23 : 385–397.
Morris Report (1990), Performance Indicators, Report of a Committee Enquiry chaired by Mr. Alfred Morris,
PCFC.
NAGPAUL, P. S., ROY, S. (2003), Constructing a multi-objective measure of research performance,
Scientometrics, 56 : 383–402.
NARIN, F., HAMILTON, K. S. (1996), Bibliometric performance measures, Scientometrics, 36 : 293–310.
PAGE, E. S. (1989), Management statistics and performance indicators in British universities. Paper presented
at the 11th European Association for Institutional Research Forum. Trier, 27–30 August.
RADNER, R. (1987), Decentralization and incentives. In: T. GROVES, R. RADNER, S. REITER (Eds),
Information, Incentives and Economic Mechanisms. Oxford: Blackwell.
ROMER, P. M. (1990), Endogenous technological change, Journal of Political Economy, 98 (5) : 71–102.
RUBENSTEIN, A., GEISLER, E. (1991), Evaluating the Outputs and Impacts of R&D/Innovation, International
Journal of Technology Management, Special publication on the Role of Technology in Corporate Policy,
181–204.
SENKER, J. (2001), Changing organisation of public sector research in Europe – implications for
benchmarking human resources in RTD, Paper prepared for Human resources in RTD session of the
“The contribution of European socio-economic research to the benchmarking of RTD policies in
Europe”. Conference: Brussels, March 15–16.
SEXTON, T. R. (1986), The methodology of data envelopment analysis. In: R. H. SILKMAN (Ed.), Measuring
Efficiency: An Assessment of Data Envelopment Analysis. San Francisco: Jossey Bass.
STAINER, A., NIXON, B. (1997), Productivity and performance measurement in R&D, International Journal
of Technology Management, 13 (5-6) : 486–496.
Appendix
Table A1. Discriminant Scores
Case Actual Highest probability group 2nd Highest Discriminant
number group Group P (D/G) P (G/D) Group P (G/D) scores
1 1 1 0.0020 0.829 2 0.171 -0.22830
2 1 1 0.4069 1.000 2 0.000 2.02770
3 1 1 0.0546 1.000 2 0.000 4.77880
4 1 1 0.5454 1.000 2 0.000 2.25230
5 1 1 0.3606 1.000 2 0.000 3.77130
6 1 1 0.8100 1.000 2 0.000 3.09740
7 1 1 0.7513 1.000 2 0.000 3.17400
8 1 1 0.9766 1.000 2 0.000 2.82760
9 1 1 0.4608 1.000 2 0.000 3.59460
10 1 1 0.8837 1.000 2 0.000 3.00330
11 1 1 0.2987 1.000 2 0.000 1.81780
12 1 1 0.9667 1.000 2 0.000 2.89880
13 1 1 0.2045 1.000 2 0.000 4.12590
14 2 2 0.7758 1.000 1 0.000 -3.42940
15 2 2 0.5116 1.000 1 0.000 -4.37050
16 2 2 0.6681 1.000 1 0.000 -3.28540
17 2 2 0.4634 1.000 1 0.000 -2.98090
18 2 2 0.7797 1.000 1 0.000 -3.99390
19 2 2 0.5061 1.000 1 0.000 -4.37900
20 2 2 0.7763 1.000 1 0.000 -3.99830
21 2 2 0.9081 1.000 1 0.000 -3.59870
22 2 2 0.4832 1.000 1 0.000 -3.01300
23 2 2 0.7054 1.000 1 0.000 -4.09220