University Distance Function Sfa

download University Distance Function Sfa

of 46

Transcript of University Distance Function Sfa

  • 7/27/2019 University Distance Function Sfa

    1/46

    1

    Palma, September 2007II IBEW

    Research Efficiency Analysis in a System of

    University Technology Transfer Offices: AnEmpirical Analysis of the Spanish Case

    DEMO Student: Maria Victoria Trujillo1

    Tutor: Dr. Emili Grifell-Tatje

    Co-tutor: Dr. Pablo Arocena

    Abstract

    The phenomenon of TTOs in Spain is relatively new and there is a need of assessingtheir efficiency. In this sense, we present quantitative evidence on the level of efficiencyof 50 Spanish TTOs for the period 2003-2005. An important methodological inside ofthe paper is the introduction of SFE with multi-output approach: Licenses and Spin-off.Our results suggest that an increment of invention disclosure and TTO size entails anincrement of the commercialization in the form of patents, licenses and Spin-off. Eventhough the evidence is weak, we also point out that public and technical universitieshave higher commercialization with respect to their counterparts, finally we find neitherexperience nor industrial cluster effects in Spanish TTOs.

    Key Words: TTO, Technical efficiency, SFE, Licenses, Spin-off

    JEL Codes: C23, D23, O31, O32.

    1We would like to thanks, the REDOTRI Universidades who kindly lend us the data set of the questionnaires of the years 2003,2004 and 2005, and specially to Susana Camara. We also want to thank the TTO of the UAB and especially to Angela Serrano whogave us an interview to define the process of technology transfer in Spain.

  • 7/27/2019 University Distance Function Sfa

    2/46

    2

    1. Introduction.

    The amount of innovation commercialized from the universities has increased

    dramatically (Nelson, 2001; Mowery et al., 2001), not only in the US, also in other

    countries where Universities have the same structure of ownership and decision rights.

    This is the case of Spain (Geuna, et al., 2003). This phenomenon started with The Bayh-

    Dole Act (1980) and the 1986 Federal Technology Transfer Act which moved the

    property and commercialization rights on inventions achieved from federally funded

    research to the universities. From then, the commercial success and the

    commercialization of research is an important option to create wealth from universities

    (Etzkowitz, 1998; Shane, 2002). It also has gone with an increasing interest of

    academics about the role of University Transfer Technology Offices (TTO). This

    phenomenon started in Spain almost 20 years ago with the First National Plan.

    Consequently, most of the Spanish TTOs have achieved enough experience

    transferring technology to the market and hence their efficiency might be analyzed in

    depth.

    In order to study their efficiency two methodologies are used in the literature. Firstly, a

    non-parametric approach (Thursby and Kemp, 2002) well known as Data Envelopment

    Analysis (DEA), and secondly a parametric approach (Siegel, et al., 2003) called

    Stochastic Frontier Estimation (SFE). The second method gives statistical information

    about the impact of independent variables but has the limitation when a multi-output

    approach is required. We have made the effort of introducing a model of two outputs:

    Licenses and Spin-off. Some of the results presented in the work are consistent with

    previous research. In short, an increment of the inputs increases the amount of

    commercialization done. Additionally, public and technical universities seem to be

    more efficient (Siegel, et al., 2003). Besides, in contrast with Siegel et al., (2003) we

    found neither experience nor cluster effects.

    The order of the paper is as follows. In the next section, an extensive review of

    international literature on the topic is done. In the third section, we analyze the

    specificities of the Spanish system and develop the hypothesis. Subsequent sections

    present data and results. Finally, concluding remarks close the work.

  • 7/27/2019 University Distance Function Sfa

    3/46

    3

    2. University to Industry Technology Transfer: a global review

    In a more global view we can understand technology transfer as the process of

    innovation that is presented, as the transference of intellectual capital and know-how

    between organizations with the purpose to be used in the creation and development of

    products and services commercially viable. (Rubiralta, 2003). Technology transfer is

    usually thought of, as occurring within or across firms (Siegel, et al., 2003) such

    information transfer is from one firm or institution to another, from one employee of

    one division to another (intra-firm transfer of technology) or to one country to another.

    - Insert Figure 1 -

    The university to industry technology transfer or commercial transfer of scientific

    knowledge from universities to firms has been a topic of interest in the literature since

    the last decades (see Figure 1). In the following section we will review this literature

    that can be divided under the following subjects:

    2.1Historic perspective, growth in academic innovation transferred to the market

    before and after Bayh-Dole Act of 1980 and statistics.

    2.2Ownership and licensing.

    2.3International and Regional comparisons through case studies.

    2.4Outputs of university research. (patents, licenses, spin-offs)

    2.5Efficiency of university Technology Transfer Offices. (TTOs).

    2.1 Historic Perspective, Growth in academic innovation transferred to the market

    before and after Bayh-Dole Act of 1980.

    The University is an institution born in the XII century, with the mission to distribute

    knowledge from professor to students, until the XIV century. In the XV and the XVI

    centuries a declivity stage starts that continued all along during the XVII and XVIII

    centuries, giving the role of knowledge and research to technical societies and

    academies that made scientific research according to the needs of a society that was

    becoming more technical. But the inefficiency of this societies and academies to

    organize themselves in specialized ways, cause the reborn of the universities in the XIX

  • 7/27/2019 University Distance Function Sfa

    4/46

    4

    century, when Van Humboldt propose a new model were is combined the already

    known educational function with a new function: research (Santamaria, et al., 2007).

    The U.S research university and the organized pursuit of R&D in industry both

    originated roughly 125 years ago, have grown in parallel throughout the 20th century

    (Mowery and Rosenberg, 1998). Even though the research cooperation between

    universities and the industry has a long story, recent changes like growth of university

    patenting and licensing of new technologies to private firms, have concerned

    considerable the attention of researches.

    In 1810, the University of Berlin is founded and is the leader of change of many other

    universities in the medieval era and also the inspiration to create new ones (Geuna, et

    al., 1999). During XIX century many universities especially the American ones were

    connected with the necessity to develop the local industry, offering guidance and

    solving the practical problems of the industry. Before World War II, which supposed an

    in flex point in the history of the science and the industry when the role of the university

    starts to change, mainly in the American universities, with their relation with the

    industry. During the war, the best scientists and technologists were incentive to create,

    promoting big advances in fields like medicine and nuclear engineer. Projects like the

    Manhattan one (atomic bomb) were the result of those efforts (Santamaria, et al., 2007).

    At this point its important to recall the Vannevar Bush inform, Science: the Endless

    Frontier, (1945) because when Bush was the director of the research and development

    office in the U.S during the war; he established the need to finance the scientific

    activities after the war and the importance of the basic research in the universities. He

    established a model of the process of innovation that has been identified as the linear

    model, were the basic research is considered as the source of technological innovation.

    The increases in university patenting and licensing are frequently asserted to be directed

    consequences of the federal policy initiative known as the Bayh-Dole Act of 1980

    (Mowery, et al., 2001). This federal policy gives the ownership of commercialization of

    federal funded research to universities. This right is usually managed by the Technology

    Transfer Office (TTO). The Bayh-Dole act has increased dramatically the

    commercialization of academic innovation. Godfarb and Henrekson, (2003) give partial

    evidence to this fact. They compare the case of Sweden where the right of selling the

  • 7/27/2019 University Distance Function Sfa

    5/46

    5

    intellectual property belong to the inventor, with the case of US, where the right of

    commercialize the intellectual property belong to the university and exploit these

    resources through the figure of the TTO. They conclude that having Sweden a higher

    relative amount of researchers than US, the income generated by licenses is relatively

    larger in US, and therefore the ownership of the decision right for the University and the

    presence of the TTO bring to a more efficient technology transfer process. This process

    has not only been taking place in the US, but also in other countries where Universities

    have the same structure of ownership and decision rights about inventions, like Spain,

    Italy or UK (Geuna, et al., 2003).

    Before the Bayh-Dole act2, U.S. public universities in the period from 1900s to 1940s

    looked for the collaboration in great spread of research within the industry. The Second

    World War transformed the role of U.S. universities as research performers, as well as

    the sources of their research funding. Then during 1950s-1960s periods, the share of

    industry findings declined because of great budgets of post war. During the 1960s the

    National Science foundation allowed academic institutions to patent and license the

    results of their research under the terms of Institutional Patent Agreements (IPAs).

    Beginning in the 1970s, the share of industry funding within academic research began

    to grow again. Most of the increase in industry funding occurred during the 1980s,

    remaining roughly constant after 1990.

    The Bayh-Dole act is contemporaneous with a sharp increase in U.S university

    patenting and licensing activity (Mowery, et al., 2001). Table 1 reveals the large

    increase.

    - Insert Table 1

    2.2 Ownership and Licensing

    Universities ownership structure has two important players in the role of technology

    transfer; one is the university, which many authors had claimed to have a new role in

    society with respect to commercialization of research results, or entrepreneurial

    science (Etzkowitz, 1998; Martin, 2003); and two is the universitys scientist or

    2 To know more about USA patenting activity before 1980 see the National Science and Foundation Board Indicators. U.S.Government.

  • 7/27/2019 University Distance Function Sfa

    6/46

    6

    inventor. From the university perspective, the challenge becomes: to increase the extent

    of commercialization, to visualize the contribution to economic development, and to

    manage the relationship between commercialization and other core activities.

    Rasmussen, et al., (2006), says that in addition to teaching and research, universities are

    increasingly expected to take on technology transfer and commercialization as part of

    their mission. They explore the initiatives provided by the universities to promote

    commercialization of university knowledge. McAdam, et al., (2005) construct a model

    with this initiatives analyzing licensing and business building process and suggest this

    approach to innovation centers. The University Scientist or inventor has an academic

    culture and it carries with an ambiguous relationship to commercial innovation and a

    preference for basic research (Ndonzuau, et al., 2002). Most of his career recognition,

    and consequently compensation, comes from his success in the basic research, therefore

    it is a clear, and in some cases important, opportunity cost for the development of

    commercial innovation. Aghion, et al., (2005) develop a model that clarifies the

    respective advantages and disadvantages of academic and private-sector research and

    they also examine when in the process of technological transfer is optimal to make the

    transition from academia to the private sector. They found that innovative ideas can be

    recognized by the private sector in a more early-stage. More recently Lowe, (2006)

    proposes a theoretical model that shows how the inventor know-how affects the

    decision to license his invention for development or star a spin-off with his invention.

    Since the passage of the Bayh-Dole Act, proponents of this legislation argue that

    industrial use of federally funded research would be reduced without university patent

    licensing. This issue means if the commercial application and diffusion of inventions

    from federally funded research critically depends upon allowing universities to retain

    title to and license them. Jensen and Thursby, (2001), address this issue providingevidence of 62 U.S universities analyzing several related theoretical models. Behind a

    moral-hazard framework, and taking into account that the effort of the inventor is not

    observable, they stated that no development of an invention will be made unless

    inventors receive incentive such as royalties or equity. Besides, following the

    framework where firms have incomplete information of the quality of the inventions

    Macho-Stadler, et al., (2005) developed a theoretical model explaining the specific role

    of TTOs in licensing university inventions. They say that beyond the classical

    economies of scale, a university wide TTO can be an instrument to reduce the

  • 7/27/2019 University Distance Function Sfa

    7/46

    7

    asymmetric information problem found in the scientific knowledge market. They

    consider a model of technology transfer between a research institute (university) and the

    industry (firm) and in parallel they develop a reputation argument for a Technology

    Seller (TTO). Consequently TTOs exist as they reduce asymmetries of information

    between academic entrepreneurs and established firms.

    Ownership, income splits, stage of development, marketing, license policies and

    characteristics, goals of licensing and the role of the inventor in licensing are studied by

    Thursby, et al., (2001), as they describe their survey of licensing at 62 research

    universities. Their most relevant find is that additional disclosures generate smaller

    percentage increases in licenses, and those increases in licenses generate smaller

    percentage increases in royalties.

    2.3 International and Regional comparisons through case studies

    There have been also a lot of studies comparing different approaches in the international

    arena like the case of Goldfarb and Henrekson, (2003), as mentioned before, that

    compare a subset of policies of the US and Swedish innovation systems that affect the

    commercialization of university technology. Owen-Smith, et al., (2002) compared US

    and European practices in terms of university industry relations.

    Feldman, et al., (2002), explore also policy issues but in a regional perspective, using

    data from U.S Carnegie I and Carnegie II universities. They estimate a model were they

    consider equity as technology transfer mechanism and found that offers advantages to

    university in revenues and ownership interests. Bercovitz, et al., (2001) studied the

    influence of university organizational structure on technology transfer performance.

    They treat the structure of the TTO as an independent variable that accounts, in part, for

    measured inter-institutional differences in patenting, licensing, and sponsored research

    activities. Using prior theories of distinct forms of organizational structure: U-Form, M-

    Form and H-Form3; they analyze three major research universities John Hopkins

    University, Pennsylvania State University and Duke University. They found that the

    3 For more details about this three forms of organizational structure see the studies of Chandler (1990) and Williamson (1985).

  • 7/27/2019 University Distance Function Sfa

    8/46

    8

    three universities differ in their organizational structure4 and that this structure affects

    performance in a predictable manner.

    Di Gregorio and Shane, (2003) analyze cross-institutional variations in new firms

    formation rates between university licensing offices (TLOs) over the 94-98 periods, in

    other words they explore empirically why some universities generate more start-ups

    than others. They say that several major recognized corporations had their origins as

    TLO start-ups, thats why they are an important mechanism for economic activity.

    Colyvas, et al., (2002) studied cases of commercialization of technologies. Feller, et al.,

    (2002), studied ways in which academic R&D and education contribute to industrial

    innovation. Beise and Stahl, (1999), find that in Germany high-technology does not

    depend on co-location of public and private research.

    Other studies focused on individual cases to explore similar issues. Zucker, et al.,

    (2002) in a biotechnology case study looked at the efficiency of university technology

    transfer process. Goldhor and Lund, (1982) made a study case of Massachusetts

    Institute of Technology (MIT) that examines the events in the transfer on an advanced

    technology (a text-to-speech reading machine) from the university to an industrial firm

    seeking to exploit the innovation. They suggest important policy making issues and

    important implications that latter gave birth to TTOs.

    2.4 Outputs of University Research (patents, licenses, spin-offs)

    The studies that focus on the outputs of the university research are widely focused on

    spin-offs companies. The development of spin-offs is analyzed by Vohora, et al.,

    (2004) they say that spin-offs has to pass through 5 phases of development in order tobe a successful venture. Clarysse, et al., (2005) explore the different incubation

    strategies for spinning-out companies employed by European Research Institutes. They

    use a two-stage approach where three distinct incubation models of managing the spin-

    out process where identified: Low selective, Supportive and Incubator. Leitch and

    Harrison, (2005) explore the spin-offs with a study case and they also examine the role

    4 More specifically they found that John Hopkins most closely aligns with the H-Form, Duke with the U-Form and Pennsylvania hascreated an M-Form.

  • 7/27/2019 University Distance Function Sfa

    9/46

    9

    of the TTO in this context. They suggest that TTO should be assumed in a more

    economic developer role.

    Locket, et al., (2003) made a comparison of two groups of universities in the U.K. They

    identify that more successful universities tend to create new ventures were the equity is

    divided more equally between the TTO, the venture capitalist and the academic

    entrepreneur. Chukumba and Jensen, (2005) develop and empirically tested a game-

    theoretic model that explains why a university invention is commercialized in a spin-off

    rather than in an established firm. The most relevant conclusion for the literature is that

    they proof that when the invention, specially engineering, is of high quality universities

    license more. And when the quality of the inventions is low universities make spin-offs.

    In the U.S, legislative support to university patenting of federally funded research

    results was largely motivated by expectations that such a policy would increase the level

    of industry R&D; Mazzoleni, (2006) presents a theoretical model of R&D competition

    based on a university invention. Patenting and licensing are studied. They found that the

    results on patenting and licensing derive in increase of R&D investment and social

    welfare, but they suggest this should be tested empirically. Macho-Stadler, et al., (1996)

    analyze terms of contracts in licensing agreements, between Spanish and foreign firms.

    They found that royalties are relative more important in contracts that transfer know-

    how, and they explain this proposing a theoretical model where know-how is hard to

    quantify so its difficult to include in a contract.

    2.5 Efficiency of University TTOs

    There have been a group of studies focusing on the use of tangible outputs to measurethe efficiency of TTOs. Markman, et al., (2005) studied the variable speed in 95 U.S

    university technology transfer offices (UTTOs) and they find that faster UTTOs can

    commercialize patent-protected technologies, greater licensing revenues and generate

    more spin-offs. Chapple, et al., (2005) present evidence on the performance of TTOs in

    the U.K. using Data Envelopment Analysis (DEA) and Stochastic Frontier Estimation

    (SFE). They found that there is a need to increase capabilities and business skills of

    TTO managers.

  • 7/27/2019 University Distance Function Sfa

    10/46

    10

    A series of studies built model to establish efficiency metrics and measure relative

    productivity. Siegel, et al., (2003) present quantitative analysis of efficiency, measuring

    the relative productivity of TTOs in the U.S using SFE measures. Their findings suggest

    that TTO activity is characterized by constant return to scale and the variation in

    performance is explained by environmental factors they use. They also present some

    qualitative analysis. Thursby and Kemp, (2002) explore the increase in licensing

    activity of U.S universities focusing on efficiency; employing DEA combined with

    regression analysis. They find that licensing activity had increased over the years by

    others factors than the relative size of the university. Anderson, et al., (2007) use a DEA

    approach to examine efficient and inefficient TTOs within U.S. universities. As a result

    efficiency in TTOs is found in many leading universities. Siegel and Phan, (2004)

    analyze and describe the most important tools DEA and SFE as the most used technique

    tools of evaluation.

    3. Spanish legal environment and the Spanish Situation

    3.1 TTO: Origin, Concept and Nature

    The theoretical analysis of the process of technological transfer and its connection witheconomical growth is relatively new. Economic theory had always intuited the

    importance of innovation and its effects in economic growth, but its until the decades

    of the 1950 through 1960 that this variable started to be considered as exogenous.

    The first theories not only had demonstrated the significant effect of innovation in

    productivity, but also, they had demonstrated the existence of failures in the

    transference of it to the market.

    In the Spanish context, the public initiatives of promotion of innovation arrive with

    remarkable delay with respect to other countries with stronger economies5. One of the

    first laws, the Organic Law 11/1983 (LOU)6, where its principal objective is to regulate

    the emerging relationships between the university and the firms, makes the role of the

    first one, as the dynamic element of the innovative process. But the policies of R&D in

    Spain have their in-flex point until 1986 when the law of Science 13/19867, takes effect.

    5 OTRI: entre la relacin y el mercado. Available in www.redotriuniversidades.net, Biblioteca, Libros, Capitulo 2.6 Ley Orgnica de Reforma Universitaria del 11/1983.7 Ley de Fomento y Coordinacin General de la Investigacin Cientfica y Tcnica del 13/1986.

  • 7/27/2019 University Distance Function Sfa

    11/46

    11

    Until this date, we cannot talk about the existence of a scientific and technological

    policy. This law defines a new organizational framework where the most important

    instrument of planning and execution would be the National Plan of R&D8 that would

    be implemented, followed and coordinated by the Inter-ministerial Commission of

    Science and Technology (Comisin Interministerial de Ciencia y Tecnologa-CICYT-).

    The law of Science in its 5 article says: that one of the objectives of the National Plan

    of R&D is to promote research and development activities within the firms and the

    collaboration between the firms and the public centres of research (law 13/86). In this

    context, at the end of 1988, with the beginning of the first National Plan of R&D 1988

    1991, starts-up the program of creation of the TTOs9 (in Spain Oficina de Transferencia

    de Resultados de Investigacin-OTRI).

    Following the previous approach of the Spanish context and the increasing flow of new

    technologies in the recent scenario is important to view in a more general view the

    process of technology transfer in two perspectives: (Rubiralta, 2003) the transference

    that is produced between firms (horizontal transference) and the transference that is

    developed between the agents (universities and public organisms) that are generators of

    knowledge and the industry (vertical transference). The principal objectives of the TTO

    are as follows (Conesa, 1997):

    - To elaborate the databank about the infrastructure and the supply of R&D.

    - To identify the transfer of results generated by the active groups of investigation

    and to directly spread them between the firms or in collaboration with the next

    units of interface.

    - To facilitate the transference of those results to the firms, or in other case, the

    correct assimilation of foreign technologies.

    - To collaborate and participate in the negotiations of the contracts of research,

    technical assistance, consultancy, license of patents, etc., between the groups of

    research and the firms.

    - To manage, with the support of the respective administrative teams, the

    contracts to carry out.

    8 Plan Nacional de I+D Espaa.9 To see the evolution and the way that have followed the Spanish TTOs until now see the documents of CICYT of 1989.

  • 7/27/2019 University Distance Function Sfa

    12/46

    12

    - To inform about the European programs of R&D and to facilitate technically the

    elaboration of the projects, and also to manage the transaction of such a projects.

    The first empirical study about the Spanish TTOs showed that they were very good

    welcome and start to spread widely, but they well in that moment in process of

    consolidation (Fernandez de Lucio y Conesa, 1996), because of their small size (2

    technicians and 2 support persons) and they were more oriented to the university only,

    which the authors qualified as a default in the mission to consolidate the relations

    between universities and firms. Rubiralta, (2005) in his study found that the weakness in

    the productivity growth in Spanish regions has generated a low technological demand of

    universities R&D, making the TTOs to establish new strategies and goals. Macho-

    Stadler, et al., (1996) analyze the contract terms of licensing agreements between

    Spanish and foreign firms and found empirical evidence that royalties are relative more

    important in contracts that transfer know-how. The intuition behind this is that because

    know-how is hard to quantify and cannot be included in a contract, therefore, the license

    agreement would be more credible when the scientist is interested in transfer know-how

    and his profits depends on the sale of the license. Serarols, et al., (2007) analyze the

    evolution, objectives, resources and activities of a specialised unit Technological

    Trampoline and create some implications and recommendations to both university andTTOs.

    Perez-Castrillo, (2005) says that one of the weaknesses of the TTOs is that, besides the

    high enrolment of administrative employees, they should hire professionals highly

    qualified, that can be able to establish a connexion between firms and the specialized

    group of scientists researchers. The intuition behind this is that the participation of the

    researchers in the process of technology transfer is crucial, and to make this happen isimportant that the TTO foments an incentive scheme that enforce the invention

    disclosure and the collaboration of the research group with the firm that signs a license

    contract. Following this approach we can derive our first empirical hypothesis10 for our

    work:

    10 Notice that this approach is consistent with the work of Jensen and Thursby (2001) and Jensen et al. (2003) presented in theprevious section.

  • 7/27/2019 University Distance Function Sfa

    13/46

    13

    Hypothesis 1: The more the number of invention disclosures more

    commercialisations (patents, licenses and/or spin-offs) will be done.

    As we announced before Macho-Stadler, et al., (2005) suggest that the main objective of

    the TTOs is to reduce asymmetries of information between parties, it is worth to notice

    that to accomplish this objective the TTOs must achieve a critical size in order to be

    able to build a reputation11. The intuition behind this is; if the TTO has a work force big

    enough to control the flows of inventions that arrive constantly they can control the

    quality of commercial inventions they offer to the industry and this is transformed into

    higher reputation. We construct our second hypothesis from this statement:

    Hypothesis 2: The size of the TTO is important to achieve a degree of efficiency.

    Vendrell and Ortin, (2006) explore the process of technological transfer from

    universities. As a result they develop empirical implications. In particular they suggest

    that more efficient TTOs help to increase the number of commercial innovations 12.

    Following this intuition we can develop our third hypothesis:

    Hypothesis 3: More efficient TTOs help to increase the number of commercial

    innovations. We expect a positive relation between efficiency and the amount of

    inventions commercialized.

    3.2 Situation of the Spanish TTOs

    In Spain, as in other countries there has been an increase, but in a medium rate, inSpanish universities patenting and licensing activity. The actual society has been

    demanding the university a deeper commitment with the economy of the country;

    claiming a higher involvement of the institution that traditionally has only been

    academic. This new involvement is what Rubiralta, (2003) called technology transfer

    that know is part of the mission of the new roll of the universities.

    11 Extracted from Vendrell and Ortin (2006) implication 2 (page 11).12 Siegel et al. (2003) in their results support this empirical implication for the U.S.A case.

  • 7/27/2019 University Distance Function Sfa

    14/46

    14

    According to the 2006 annual Survey of RedOTRI Universidades one of the most

    important indicators to measure the degree of interaction between the university and

    industry is the number of contracts of R&D signed. We can see the evolution in Graphic

    1.

    - Insert Graphic 1 -

    One of the traditional indicators of reference to measure the degree of interaction

    between the university and the R&D firm is the quantity and the nature of the contracts.

    With the data available by the RedOTRI, we can observe a tendency of rise in the total

    volume of hiring, measure in euros, for the activities of R&D, consultancy or technical

    services and other agents as shown in Graphic 1.

    In other hand, the evolution of University patents also shows an important increase in

    the last years, with a growth of almost 50% in only three years (see Graphic 2). The

    international extensions have increasingly growth in the same interval of time which

    shows a major interest of obtaining more utilities from the inventions disclosed. But

    also shows that in spite of the effort, the bet for the international extension in the

    Spanish arena is still growing.

    - Insert Graphic 2 -

    On of the most relevant items to reflect the increase in the patenting activity in the

    Spanish university is the license agreements (see Graphic 3). In the last three years the

    number of license agreements has been almost tripled, in the year 2005, 106 license

    agreements were signed.

    - Insert Graphic 3 -

    The income generated by license agreements (see Graphic 4) have multiplied in the last

    years, but its interesting to note that the volume of income generated by these

    agreements do not evolve as expected, which indicates that it is not exploited in its

    maximum of possibilities.

    - Insert Graphic 4 -

  • 7/27/2019 University Distance Function Sfa

    15/46

    15

    Finally, another important aspect in the transference of technology is the university

    based off firms or spin-offs (see Graphic 5). In Spain this aspect is still in an embryonic

    stage in spite that there have been a lot of initiatives that are starting up from different

    approaches.

    - Insert Graphic 5 -

    According to the data of the RedOTRI Universidades, in the year 2001 only 39 spin-offs

    were created but after the 2003 the growth has been constant. There is a regular creation

    of around 90 Spin-offs per year.

    4. Determinants of the Process of University to Industry Technology Transfer

    4.1 Internal Inputs, External Outputs and Environmental Factors.

    In terms of organizational structure, the existence of a TTO inside a university is shown

    as a fundamental instrument for the development of good relations with the industry

    (Perez-Castrillo, 2005). In order to define the suitable inputs in the process oftechnology transfer from a university to a firm or entrepreneur it is important to describe

    the principal steps of this process for universities that have a TTO. First we would make

    a description of the process as follows taking into account the wisdom of specialized

    administrative personnel13 and the literature. We will focus in the transference of

    technologies through a license contract. This linear model (see figure 1) does not

    represent that all technologies are transferred in the same way in all Spanish universities

    this can be a topic for further research. The steps that follow the process of transferenceof technologies that results in a spin-off or start-up are very similar14.

    - Insert Figure 1 -

    The first step, necessary to transfer knowledge, is the research of innovations. This

    process is done in laboratories or departments that are in charge of groups that work in

    13 The interview was made in the TTO of the UAB to the person in charge of contracts and licence agreements.14 If you want a more detail specification of the steps for the creation of a spin-off or start-up that is of a result from an exploitedinnovation see Vendrell and Ortin (2006).

  • 7/27/2019 University Distance Function Sfa

    16/46

    16

    the same research line. This is generally a decentralized process where the researcher is

    free to make research in those lines of his/her interest ( more near to their knowledge

    areas) or that are considered more promising. In any case the lines of interest in research

    are influenced by the possibility of obtaining financial aids.

    The second stage of the model is the scientific discovery. When a group of researchers

    find an innovation is fundamental that the university have knowledge that the

    innovation exists. In this second stage, the TTO staff must encourage the scientific

    members to disclose the inventions. Besides the decentralization in the process of

    research, the knowledge that the innovation exist cannot be supposed. There should be a

    system of information transmissionfrom the research group to the TTO, and this is what

    we know as invention disclosure.

    Once the invention is formally disclosed, the TTO in the third stage with a specialized

    team evaluates the potentiality of the technology and decides whether to patentor not

    the innovation. If the TTO considers that the innovation represents a step forward in the

    scientific arena and its possible it has a commercial value, they will start the transaction

    to obtain a patent that protects the innovation. The origin of the regulations of the

    universities technology transfer starts with the approval of Bayh-Dole Act in the United

    States. Its prime objective was to stronger the interaction between the universities and

    the industry, the result of this law was that the universities, by the figure of the TTO,

    could retain the property of the technologies and grant licenses on them to companies,

    always giving preference to the PYMES. In Spain the Foundation University-Industry

    (FUE Fundacin Universidad Empresa) and the TTO have been contributing for years

    to facilitate the collaboration in this sector. The Organism that regulates in Spain the

    FUE and the TTO is the RED OTRI. The RED OTRI has its roots when the National

    Plan of Scientific Research and Technological Development (PNID) start activities

    (1988).

    Its very important to take into account that to request and maintain a patent is very

    expensive and the TTO have limited resources to fill; thats why only potential

    innovations or the interest of the industry in the technology are the only criteria that are

    taken into account. Then the university decides to apply to domestic or internationalpatent protection.

  • 7/27/2019 University Distance Function Sfa

    17/46

    17

    If the patent is awarded, then in the fourth stage of the process the TTO would often

    attempt to market the technology, and the scientific members are involved in the

    marketing process because their technical knowledge makes them a natural partner for

    the firms. The TTO would search for potential buyers to license the technology. The

    most active and wide experienced TTOs that usually are the big ones, making the size of

    the TTO an important input, usually are the ones that have a portfolio of possible

    clients. But the researcher is usually a very important source of information of the

    industry because usually they know firms that work and commercialise with these

    products. In the webpages of the Spanish TTOs there is only information of the

    available patents. Jensen and Thursby, (2001) report in their study, and Siegel et al

    (2003) confirmed in their field research, that many firms will license a technology

    before it is patented. This means that a key input of the university to industry transfer is

    invention disclosure, because is the portfolio of available technologies for licensing.

    In the final stage of the model if a firm or individual entrepreneurs are interested in the

    patent, also the TTO is the one in charge to negotiate and redact a contract or license

    agreement for the transference of technology. This process can also derive in the

    constitution of a spin-off or start-up but this is sub stage (Chukumba and Jensen, 2005).

    In resume we assume that the following internal factors are inputs of the university to

    industry technology transfer: invention disclosure, and the size of the TTO (labour

    employed by the TTO). And that the following external factors are outputs: Licence

    agreements, and patents. This would be our TTO production function where the

    inputs are under the control of the producer in this case the TTO director.

    The technology transfer activity may also depend on some institutional factors. For

    example, being near an industrial zone may facilitate the commercialization of

    innovations. For instance we can recall the case of Standford University being near an

    industrial cluster like Silicon Valley15 or for the case of Spain the great industrial

    clusters are located in the big cities like Barcelona, Madrid, etc. Another important

    factor is the public status of the university. Private universities may have less rigid

    policies and public universities may have more financial aids from the state. The year of

    15 For more information about Silicon Valley see Hayes (1989).

  • 7/27/2019 University Distance Function Sfa

    18/46

    18

    creation of the TTO or its age can be a relevant factor too. We can assume that TTO

    with more experience may be more efficient that the young ones.

    5. Data Construction

    Our data is based on the RedOTRI survey for the years 2003, 2004 and 2005. We

    complement this information from data of a survey made by DGPYME and ICO16 to the

    different to several TTOs and some Webpage information of some TTOs we needed to

    complement information17. We could cover a complete and balanced panel for 50

    Spanish universities18 over the period 2003-2005. We also conduct an in depth

    interview with a responsible of the TTO of the Autonomous University of Barcelona,

    our objective was to understand the objectives of the Spanish TTOs.

    We consider university outputs19 of the TTOs those that imply the commercialisation

    of one invention previously developed in the University. In this sense, we differentiate

    in two different outputs. The Licenses are the sum of patents and licenses and are

    considered the commercialisation through the market that gives an immediate cash-flow

    to the TTO and the inventors. We also consider the number of university Spin-offscreated as an output. Those firms need time to get positive profits and sometimes these

    returns do not go to the TTO and remain to the academic entrepreneurs and/or investors

    (usually, equity holders as venture capitalists or debt-holders as Banks). We define two

    different inputs. In this sense, we have Inv.disclosure that is the inventions received to

    the TTO that has potential commercialisation to the market. We also have the Size of the

    TTO, which are the number of employees in each office. Finally we define 4

    environmental variables consistent with the Spanish model. In this sense, we havecreated four dummy variables. Technical measures if the university is based on

    technological studies (16%), Public determines the property of the university being

    either public or private (84%), we also differentiate between new and old TTOs. The

    16 For more detail on the survey or the institutions see Ortin et al. (2007).17 We check the webpage of the following universities: U. Alcala de Henares, U. Politecnica de Madrid, U. de las Islas Baleares, U.de Salamanca, U. de Sevilla, U. de Valencia y U. de Zaragoza.18

    In Spain there are 61 universities. So we have information of 82% of the population. Moreover, we have information of theuniversities that traditionally has been more active in innovation and commercialisation, remaining out of our sample thoseuniversities with minor impact on innovation. The universities are listed in table 5.19 As we have some zeros we might sum 1 to all the inputs and outputs.

  • 7/27/2019 University Distance Function Sfa

    19/46

    19

    new (52%) ones are those created after the National Plan (1988). We also define those

    universities that are near to industrial areas 20(44%).

    Table 2 shows the mean of all those variables for the period 2003-2005. We observe

    that Inv.Disclosure has dramatically increased between 2004 and 2005. Probably this

    fact explains also the increment of licenses and patents21. We see that the size of the

    TTOs remaining constant in the range of 13-14 employees. The number of Spin-off per

    year is around 65 in 2004 and 100 in 200322.

    - Insert Table 2 -

    6. Methodology and Results

    6.1 The treatment of Efficiency

    In the previous section, we identified a set of potential determinants of the process of

    university to industry technology transfer; which includes internal inputs, external

    outputs and environmental/institutional factors.

    The microeconomic literature defines efficiency of a production unit as the comparison

    between observed and optimal values of its output and inputs. The technical efficiency

    is defined as the ability to obtain the maximum potential output obtainable from the

    given inputs or the ratio of minimum potential to observe inputs required to produce the

    given output.

    In the economic theory two approaches are recognized to construct frontiers: DEA and

    SFE. In this work is proposed an application of models of the distance function

    proposed by Shephard (1953) and Coelli and Perelman (1996) to analyze technical

    efficiency. The notion of distance to the frontier proposed by Shephard (1953) can be

    used to calculate the efficiency of a set of units of production in a scene of multiple

    outputs. DEA first developed by Charnes et al. (1978) is a non-parametric estimation

    20

    We consider those universities near Barcelona, Madrid, Sevilla, Bilbao and Valencia as universities near an industrial centre.21 Notice the patents are on average 75% of the sum of patents and licenses.22 For more detail in the evolution of Spanish Spin-offs see Ortin et al. (2007). Our data is consistent with their results. Theyestimate a creation of 90 university Spin-offs per year.

  • 7/27/2019 University Distance Function Sfa

    20/46

    20

    technique that has been used extensively to compute relative productivity in services

    industries. SFE method first developed by Aigner, et al., (1977) and Meussen and Van

    den Broeck (1977), are extensions of the tradition regression model, based in the

    microeconomic premise that the production function represent an ideal: the maximum

    output that can be obtained with a set of inputs. This interpretation as pointed by Green

    (1993 (b)) recalls naturally under an econometric analysis, in which the inefficiency is

    identified with the errors of the regression model.

    6.2 Multi-output approach

    Thursby and Kemp, (2002) use DEA to assess the relative efficiency of TTOs using a

    multi-output approach. Multi-output approach is used when is assumed that producer

    use multiple inputs to produce multiple outputs. DEA is a mathematical programming

    approach that does not require the specification of a functional form for the production

    function; but DEA doesnt allow the study of the relations of causality between

    variables of resource and production. SFE, introduces this last approach but introduces

    the problem that it is only design to incorporate only one endogenous variable, that

    represents the output obtained by the productive system and is explained by a set of

    production factors.

    One possible solution to the problem is to add the outputs in a single one dependent

    variable by the calculation of an indicator that gathers the set of outputs by their prices

    (Shadow Prices) or by a set of assigned weights subjectively. In the public

    administration arenas, that is the case of the TTOs, is not advisable to use shadow prices

    because the sell of the services is not observed so its hard to get the shadow prices to

    obtain a structure to weight.

    Lets assume that and represent the input and

    output vectors at time

    ),,...,( 1t

    M

    ttyyy =

    Tt ,...,1=

    { }tttt xfromobtainableisyxP =)(

    += KtKtt Rxxx ),,...,( 1

  • 7/27/2019 University Distance Function Sfa

    21/46

    21

    Shephards distance function avoids having to establish a priori shadow prices. The

    output Distance function is defined by Shephard, (1970) as:

    (1)

    The parametric approach of the distance function parts from the theory of the

    homogeneous functions to introduce the assumption of multiple outputs obtained by

    multiple inputs. The method was introduced by Aigner and Chu, (1968) for the

    assumption of one single output and multiple inputs and a Cobb-Douglas function. Here

    is used that approximation to adapt it to the assumption of a translogaritmic function,

    that is based in a Cobb-Douglas function, but more flexible than this because its partial

    derivatives are not constant. The expression proposed of the distance translog function

    for the assumption of M output and K inputs is the following:

    )2(,...,2,1lnln2

    1

    lnln2

    1lnlnln

    2

    1lnln

    1 1

    ,

    1 11 1 11

    00

    = =

    = == = ==

    =+

    ++++=

    K

    K

    M

    m

    mikikm

    K

    k

    K

    l

    likiki

    M

    n

    M

    m

    ki

    K

    k

    knimimnmi

    M

    m

    mni

    Niyx

    xxyyyD

    WheremiY is the production of output m and kiX the quantity of input k for the

    productive unit i. ,, are the parameters to estimate and iD0ln is the term of

    inefficiency of the evaluated unit.

    Additionally two constraints of homogeneity are required. Homogeneity of degree +1 in

    outputs and homogeneity of degree +1 in inputs:

    =

    =

    ==

    ==

    M

    n

    km

    M

    n

    mn

    Kk

    Mm

    1

    1

    ,...2,1,0

    ,...,2,1,0

    Homogeneity of degree +1 in outputs is imposed in order to obtain an output oriented

    radial distance function. Homogeneity of degree +1 in inputs implies constant returns

    to scale technology, an assumption necessary to accurately measure productivitychange.

    )}()/(:min{),( xPyyxD o =

  • 7/27/2019 University Distance Function Sfa

    22/46

    22

    The third constraint is of symmetry of the parameters:

    .,...,2,1,.,...2,1, klkandMnm lkklnmmn ====

    A case of a homogeneous function of degree +1, where is accomplished:

    ,0),(*)*,(0 >= gaforyxDgygxD o

    If we select arbitrarily one of the M outputs and we consider Myg /1= , then:

    MM yyxDyyxD /),()/,( 00 =

    And the translogaritmic function would be transformed into:

    = = ==

    =

    =

    = =

    =+

    ++++=

    K

    k

    K

    k

    M

    m

    mikikm

    K

    l

    kiki

    M

    m

    M

    n

    M

    m

    K

    k

    kiknimimnmimnMioi

    Niyxxlix

    xyyyyD

    1 1 11

    1

    1

    1

    1

    1

    1 1

    0

    ,...,2,1,*lnln2

    1lnln

    2

    1

    lnln*ln2

    1*ln)/ln(

    )3(

    Where Mim yyy /* = , this means, the ratio between each one of the outputs and the

    output selected as a reference to transform g.

    The translog can be expressed as follows:

    .,...,2,1),,,/,()/ln( NiyyxfyD miiimio ==

    And this is equal to:

    ),,,/,()ln()ln( miiimio yyxfyD =

    Or

    imiiimi UDwhereDyyxfy == )ln(),ln(),,,/,()ln( 00

  • 7/27/2019 University Distance Function Sfa

    23/46

    23

    This is equivalent to identify the term of error with the logarithm of the distancefunction.

    The stochastic approximation in this case adapts the translog form to the assumption

    that the decomposition of the error term of the model of two stochastic terms: itV , which

    is assigned a normal distribution ),0( N , with mean equal 0 and constant standard

    deviation, that represents the deviations of the values of production with respect to the

    frontier by factors affected by uncertainty; not controlled by the participants in the

    productive process. A second component:

    itUD = )ln( 0 , that represents the deviations of the observations in the sample with

    respect to the efficient frontier, because of the inefficiency of the economic agents

    represented by them. Coelli and Perelman, (1996) considered this second component as

    the product of two terms: a term that is a deterministic function dependent of time23.

    This function represents the change experimented with the pass of time of the economic

    agents represented by the sample used. The other term is a random variable that has

    been assigned a normal truncated distribution in the value24 0. Beside is allowed that the

    mean of this second component is zero and that the standard deviation to be constant,

    but different from the term itV .

    It should be noticed that the value of 0D is not directly observable because it forms part

    of the composed error termititit UVE += . The estimation of that value is made by the

    expected value of the errors due to inefficiencies conditioned to the composed error:

    [ ]ititoi EUED /)exp(=

    That is the expected value of the degree of inefficiency of the correspondingobservation, obtained by the comparison of the value of production of this with the

    value of efficient production, for its levels of available productive resources.

    For the translog function we use the program FRONTIER 4.1 (Coelli, 1996).

    23 It can be exponential, linear or quadratic depending of witch has been the temporal evolution of the degree of inefficiency of theobservations.24 With this the level of efficiency is not negative.

  • 7/27/2019 University Distance Function Sfa

    24/46

    24

    Our equation (4) as indicated before, to construct the distance function is precisely to

    choose one of the outputs to normalize the function, and the expression of the adjusted

    function is:

    )4(

    6.3 Single output approach

    In order to asses relative productivity in the process of technology transfer and using a

    single output approach, following the methodology of Siegel, et al., (2003) we use SFE.

    SFE creates a production frontier (Aigner et al., 1977, & Meussen and Van den Broeck,

    (1977)) with a stochastic error term that is composed by: a conventional random error

    and term that represents relative inefficiency (deviations from the frontier).

    Following Aigner et al. a production function for a given university, say the ith, its

    estimated:

    );( ii xfy = ii Xy = ( )5

    Here i denotes the ith university, iy is the maximum output obtainable from ix , a vector

    of (non-stochastic) inputs and is an unknown parameter.

    In order to characterize differences in output among universities with identical input

    vectors or to explain how a given universitys output lies below the frontier, );( ixf

    a disturbance term is assumed.

    In an attempt to give them a statistical basis, Schmidt (1976) explicitly added a one-

    sided disturbance which yields the model,

    iii xfy += );( Ni ,......,1= Where 0i .

    ,iii UV += Ni ,......,1=

    )ln(

    )ln(

    )/ln(

    )(2/1)(2/1)(2/1

    2

    1

    1

    122111112112

    2222

    2111

    2111221111

    sizeX

    isclosureinventiondX

    licencesspinoffY

    YXYXXX

    XXYXXYLnLicences

    =

    =

    =

    ++++++++++=

  • 7/27/2019 University Distance Function Sfa

    25/46

    25

    i is an error term with two components. The error component

    iv represents the

    symmetric disturbance: iV are assumed to be independently and identically distributed

    as ).,0(

    2

    vN The error component iU is assumed to be distributed independently of iv ,

    and to satisfy .0iu The non-positive disturbance iu reflects the fact that each

    university output must lie on or below its frontier.

    To estimate the technical efficiency of a producer, distributional assumptions are

    required: The Normal-Exponential Model, the Normal Truncated Normal Model, and

    the Normal-Gamma Model which are the ones that can be considered. In this model we

    would use The Normal-Half Normal Model, considering the stochastic production

    frontier model we make the following distributional assumptions:

    regressorstheofandothereachoftlyindependenddistributeareuandviii

    normalhalfenonnegativaasisthatNiiduii

    Niidvi

    ii

    ui

    vi

    ,)(

    ),,0()(

    ),0()(

    2

    2

    +

    The inefficiency term iU is assumed to have a half normal distribution. The log

    likelihood function for a sample of ith universities is:

    +=

    i

    i

    i

    iItconsL2

    22

    1lnlntanln

    (6)

    Subsequent with some SFE models25, within current time, have been created to allow

    that the technical inefficiency term can be expressed as a function of a vector of

    environmental and organizational variables. This is consistent with our assumption that

    relative inefficiency is related to environmental/institutional factors. So following the

    studies of Reifschneider and Stevenson (1991) and Siegel et al. (2003), we presume

    that26 the inefficiency disturbance is composed of two factors, a factor reflecting

    systematic influences and a random factor:

    iiii wZgU += )( ( )7

    25 For more details see Reifschneider and Stevenson (1991).26

    iU (universities on or below the frontier) are independently distributed as truncations at zero of the ),( 2uimN

  • 7/27/2019 University Distance Function Sfa

    26/46

    26

    Where Zis a vector of firm specific inefficiency explanatory variables and is a

    parameter vector. iw is the unexplained component of inefficiency error and has the

    same assumed normal distribution.

    Using the program FRONTIER 4.1 (Coelli, 1996), we obtain maximum likelihood

    estimates27 of the parameter vectors and from the estimation of the production

    function and inefficiency term equations.

    iiii UVSizedisclosureInvLicenseLn +++= )ln().ln()( 210 ( )8

    Our equation (8) is based on the model28of Siegel et al. (2003) usingLicenses as a proxy

    of the process of technology transfer output and relating to two inputs: Invention

    Disclosure and Size; assuming a three-factor log-linear Cobb-Douglas production

    function.

    And the technical inefficiency ( )iU term expressed as:

    ++=k

    iki INSTENVU /0

    Where ENV/INST is a vector of environmental and institutional factors, and is a

    disturbance term. Thus, our equation (8) is:

    iiiiii CLUSTERNATIONALPUBLICTECHU +++++=

    43210 )9(

    Consistently with our interview in depth and according to Chukumba and Jensen (2005)

    licenses and patents are important outputs for universities. In this sense, these outputs

    generate immediate cash-flows, and universities do not have to pay the opportunity cost

    of renouncing to academics. Moreover, Chukumba and Jensen suggest that TTOs try to

    commercialise the projects through established firms and consider Spin-offs as a second

    27See Battese and Coelli (1995).28Siegel et al. developed their equation (page 32) based on the knowledge production function framework developed by Griliches(1979).

  • 7/27/2019 University Distance Function Sfa

    27/46

    27

    option. Consequently in this section we analyze the efficiency of TTOs considering on

    licenses as an output (Siegel, Waldman and Link, 2003). Besides, we also consider spin-

    off as a single-output to make the analysis more robust (See equation (10)).

    iiiiUVSizedisclosureInvoffSpinLn +++= )ln().ln()( 210 )10(

    6.4 Results and Comments

    Table 3 contains two sets of parameters estimates of the Multi-output distance function

    outlined in the previous section (equation (4)) for the dependent variable licences

    (licences + patents). Models

    29

    1 and 2, (with and without environmental factorsrespectively), are presented in the first two columns. Across all variants, the estimated

    elasticity of Y2/Y1 (Y1) with respect to invention disclosure is positive and significant

    in model 1. This means that if more new innovations are disclosed then more licences

    agreements are commercialized. The estimated elasticity of Y1 with respect to size is

    positive and significant in model 1 and 2. It appears that hiring additional staff for the

    TTO increments the commercialization of licences agreements. Consequently, the extra

    information that we can extract from this analysis come from the relation of the outputs.

    We see that there is a quadratic effect between the ratio Y2/Y1 (Y1) and the amount of

    licenses (Y1). In this sense from Model 1 there is an optimal ratio that maximises the

    level of licenses commercialised. Operating30 this optimal ratio is Y2/Y1 = 0,7639.

    - Insert Table 3 -

    Removing logarithms we get that in the optimal situation Spin-offs equal to

    Licenses0,7639. Finally, from the positive sign and significance of the parameters

    2 and

    11 of the Models 1 and 2 we can accept for Hypothesis 1 and 2.

    In the model 3 (licences) and 4 (spin-off) of the table 3 we show the results for panel

    data analysis. We can see that invention disclosures affect positively and highly

    significant the number of licenses and the number of spin-off created. In this sense,

    29 Notice that variables x1, x2 and y2 are divided by y1. For homogeneity reasons this fact is not mentioned in the Table 3.30 We make a first order approach to get such result (Y2/Y1 = Y1 = 2,24/2,93 = 0,7639)

  • 7/27/2019 University Distance Function Sfa

    28/46

    28

    from the coefficient of Model 3 and 4, if we double the number of invention disclosures

    the licenses would increase in 82%, and the spin-off would increase in 15%. The

    relation between size of the TTOs is positive and highly significant for model 3 this

    means that if we double the size of the TTO licences would increase in a 27%.

    It is important to notice, that for the Model 1, we have significant results for the

    environmental variables. So we can say that universities that have a technical profile

    and public universities have higher licence activity. But the ones that are near an

    industrial cluster reduce the licence activity. This last result might come because of the

    high competition, in this sense, universities operating in small markets or cluster work

    better as they have less competence. Besides, we do not find experience effect as 3

    is

    not significant. Notice that the results of Model 3 are consistent with the results

    outlined.

    The mean technical efficiency is consistent with Siegel et al. (2003). In particular, they

    found that the mean of technical efficiency for US are closed to 0,75 very similar to the

    ones found for model 1 (0,72) model 2 (0,68) and model 3 (0,64).

    A set of aspects that are interesting and that can be studied in an analysis of this type of

    models, are the ones referring to the structure of the term of error, dispersion of

    efficiencies and the distributions of probabilities of the density function component of

    the error term representing the degree of inefficiency. In the models 1, 2 and 4 results

    highly significant the parameter bounded to sigma squared, that gathers the total

    variance of the error term. Also the parameter associated to gamma, are significant in

    the model 1 and 3, which represents the proportion of variance of the stochastic term of

    inefficiency with respect to the total variance.

    Besides, we check for specificities of technology based31 universities (UPC; UPV;

    UPM). In order to see if technological based are more efficient a One-way ANOVA was

    run. We use the technical efficiency from model 1, 3 and 4. We cannot reject the null

    hypothesis that the mean are equal (see table 4) in model 1 and 3 at the traditional levels

    of significance. Even though for model 4 (spin-off) there is a significant difference (this

    31 In Spain they are called Polytechnics (Politcnica in Spanish).

  • 7/27/2019 University Distance Function Sfa

    29/46

    29

    means that for the spin-off there a difference between the TTOs) we consider that this

    fact does not justify a separate analysis. Therefore, we can say that technology based

    universities are not more efficient than their counterparts.

    - Insert Table 4 -

    Besides, the efficiency analysis at university level has also special interest; in particular,

    the study of the efficiency of the technological based universities. To do so we use the

    technical efficiency for each one of the university considering the three years. We have

    the source of technical efficiency coming from Panel Data analysis: SFE with multi-

    output (from Model 1) and SFE with single output (from Model 3). The results are

    presented in Table 5.

    - Insert Table 5 -

    We can observe that the polytechnic universities are ranked in both models below the

    mean, and this is consistent with the results we obtained from the ANOVA that we

    explained before. We can extract from Table 5 that the universities that maintain the

    first positions constant for the three years are University of Navarra or University of

    Zaragoza; and the universities that maintain the worst position are the University of La

    Corua.

    6.4 What affects efficiency?

    It is of special interest to explain which variables determine efficiency. In this sense, our

    Hypothesis 3 states that there is a positive relation between efficiency and the amount of

    inventions commercialized (Vendrell and Ortin, 2006). In a first attempt to see this

    relation we use a one way ANOVA. First we divided the sample of licences in three

    groups32. From table 6 we can observe that the mean efficiency is significantly different

    between groups, being higher for the universities that have a greater number of licences.

    From this result we can accept Hypothesis 3.

    32 Group 0= between 0 and 1 licence, Group 1= between 2 and 9 licences, Group 2= 10 or more licences.

  • 7/27/2019 University Distance Function Sfa

    30/46

    30

    - Insert Table 6 -

    In order to make the analysis more robust we propose a panel data with fixed effects

    (Green, 1983; pp. 560-566). Our dependent variables are the scores presented in Table

    5. Consequently, we run two different models taking into account the mean efficiency

    of the models 1 and 3. The results are shown in the Models 5 and 6 of the Table 7.

    - Insert Table 7

    The results indicate that an increment of one license or patent entails a growth of 2% of

    the technical efficiency (it ranges from 1,5% to 2,2%). Similarly creating a new spin-off

    increase the technical efficiency around 1%. Ceteris Paribus, the expected effect of

    increasing an input is a reduction of the technical efficiency. This fact explains the sign

    of invention disclosure. A new invention disclosure produces a reduction of 2% in the

    technical efficiency. Its important to recall that the effect of TTO size is diffuse

    because it has a positive and not signficant impact on efficiency. Consequently TTO

    size does not have the effect predicted by the theory on technical efficiency.

    7. Conclusions, limitations and further research.

    The efficiency of the university TTOs is an important discussion in the academic

    literature (Siegel et al., 2003; Thursby and Kemp, 2002). In this sense, this paper fills

    two existent gaps of the previous literature. First, set up evidence on Spanish TTOs.

    Second, introduce an important methodological tool that has not been used in the

    previous works. This important methodological inside of the paper is the use of SFE

    with a multi-output approach.

    From our analysis we shed light into several issues. The mean technical efficiency for

    the Spanish TTOs ranges from 0,72 (SFE with multi output) to 0,64 (SFE with single

    output). These results are consistent with the evidence found by Siegel et al. (2003) for

    the case of US (around 0,75).Invention disclosure and TTO size increases the amount

    of commercialisation done and hence the efficiency of the TTO. From this result we can

    state an important advice for policy makers. TTO might increase their size and amountof invention disclosure. For this second variable it is important to look for good

  • 7/27/2019 University Distance Function Sfa

    31/46

    31

    incentives and information. In technical efficiency terms they should try to avoid

    inventions without potential commercialisation. This is probably the case of technology

    based universities (UPC, UPM and UPV).

    We also look for the impact on environmental factors. Even the evidence is weak, the

    results indicate that Public and technical universities are more efficient than their

    counterparts. Additionally, we could find neither experience nor industrial cluster

    effects.

    The work has two important limitations related to the data base. First, our sample is

    small what difficult the introduction of several independent variables required by SFE.

    Second, we do not have information of a relevant output that would enrich our analysis.

    In this sense, the introduction of research contracts could modify some of the results. As

    a further research we recommend the extension of the sample with the TTOs of some

    South-European countries similar to Spain such as Portugal, France or Italy. Moreover,

    apart from the introduction of the amount of research contracts we also think that it is

    important controlling for the quality of research as an environmental factor. A possible

    proxy would be a relative amount of the papers published in top scientific journals.

  • 7/27/2019 University Distance Function Sfa

    32/46

    32

    References

    Aigner, D., and Chu, S.F., 1968. On Estimating the Indistry Production Function.American Economic Review, 58, pp. 826-839.

    Aigner, D., Lovell, C.A.K., Schmidt, P., 1977. Formulation and Estimation ofStochastic Frontier Production Function Models. Journal of Econometrics, 1, pp. 21-36.

    Aghion, P., Dewatripont, M., Stein, J.C., 2005. Academic Freedom, Private-SectorFocus, and the Process of Innovation. NBER Working Paper Series. 11542.

    Anderson, T.R., Daim, T.U., Lavoie, F.F., 2007. Measuring the Efficiency of UniversityTechnology Transfer. Technovation, doi:10.1016/j.technovation.2006.10.003.

    Beise, M., and Stahl, H., 1999. Public Research and Industrial Innovations in Germany.

    Research Policy, 28, pp. 397-422.

    Bercovitz, J., Feldman, M., Feller, I., Burton, R., 2001. Organizational Structure as aDeterminant of Academic Patent and Licensing Behavior: An Exploratory Study ofDuke, Johns Hopkins, and Pennsylvania State Universities. Journal of TechnologyTransfer, 26, pp. 21-35.

    Chandler, A., 1990. Scale and Scope: The Dynamics of Industrial Capitalism.Cambridge, MA: The Belknap Press of Harvard University Press.

    Chapple, W., Lockett, A., Siegel, D., Wright, M., 2005. Assessing the relativeperformance of U.K. University Technology Transfer Offices: Parametric and non-parametric evidence. Research Policy, 34, pp. 369-384.

    Charnes, A., Cooper, W.W., Lewin, A., Seiford, L.M., 1995. Data EnvelopmentAnalysis: Theory, Methodology and Applications. Kluver Nijhoff Publishing Boston.

    Chukumba, C., and Jensen, R., 2005. University Invention, Entrepreneurship and Start-ups. NBER Working Series, 11475.

    Clarysse, B., Wright, M., Lockett, A., Van de Velde, E., Vohora, A., 2005. Spinning

    Out New Ventures: a typology of incubation strategies from European researchinstitutions. Journal of Business Venturing, 20, pp. 183-216.

    Coelli, T.J., 1996A. A guide to Frontier Version 4.1: A Computer Program forStochastic Frontier Production and Cost Function Estimation. CEPA Working Papers.http://www.une.edu.au/econometrics/cepawp.htm

    Coelli, T.J, and Perelman S., 1996. Efficiency Measurement, Mult-ioutput and DistanceFunctions: with application to European railways. CREPP WP 96/05. Centre deRecherche en Economie Publique et Economie de la Population, Universit de Lige.

  • 7/27/2019 University Distance Function Sfa

    33/46

    33

    Colyvas, J., Crow, M., Gelijns, A., Mazzoleni, R., Nelson, R.R., Rosenberg, N.,Sampat, B. N., 2002. How do University Inventions Get into Practice? ManagementScience, 48, pp. 61-72.

    Conesa, F. 1997. Las Oficinas de Transferencia de Resultados de Investigacin en el

    Sistema Espaol de Innovacin. Tesis doctoral. Universidad Politcnica de Valencia.

    Di Gregorio, D., and Shane, S., 2003. Why do some Universities Generate more Start-ups than Others? Research Policy, 32, pp. 209-227.

    Etzkowitz, H., 1998. The Norms of Entrepreneurial Science: cognitive effects of thenew university-industry linkages. Research Policy 27, pp. 823-833.

    Fre, R.S., Grosskopf, M., Lovell, C.K., Yaisawarng, S., 1993. Derivation of ShadowPrices for Undesirable Outputs: A distance function approach. Review of Economicsand Statistics, 75, pp. 374-380.

    Feldman, M., Feller, I., Bercovitz, J., Burton, R., 2002. Equity and the TechnologyTransfer Strategies of American Research Universities. Management Science, 48, pp.105-121.

    Feller, I., Ailes, C.P., Roessner, D., 2002. Impacts of Research Universities onTechnological Innovation in industry: Evidence from Engineering Research Centers.Research Policy, 31, pp. 457-474.

    Fernndez de Lucio, I and Conesa F., 1996. Estructura de Interfaz en el Sistema Espaolde Innovacin. Su papel en la difusin de Tecnologa. Centro de Transferencia deTecnologa. Universidad Politcnica de Valencia.

    Friedman, J., and Silberman, J., 2003. University Technology Transfer: Do incentivesManagement, and Location Matter? Journal of Technology Transfer 28, pp. 17-30.

    Geuna, A., 1999. Determinants of University Participation in EU-funded R&DCooperative Projects. Research Policy 26, pp. 677-687.

    Geuna, A., Salter, A.J., Steinmueller, W.E., 2003. Science and Innovation Rethinkingthe Rationales for Public Funding. Edward Elgar. Cheltenham, UK.

    Godfarb, B., Henrekson, M., 2003. Bottom-up versus top down policies towards thecommercialization of university intellectual property. Research Policy 32, pp. 639-658.

    Goldhor, R.S., and Lund, R.T., 1982. University-to-Industry Advanced TechnologyTransfer: A case study. Research Policy, 12, pp. 121-152.

    Greene, W.H. 1993 (a). Econometric Analysis. Prentice Hall (4th edition)

    Greene, W., 1993 (b). The econometric approach to efficiency analysis, in Lovell K.and Schmidt S. (Eds.). The Measurement of Productive Efficiency: Techniques and

    Applications. Oxford University Press, Oxford, 68-119.

  • 7/27/2019 University Distance Function Sfa

    34/46

    34

    Griliches, Z., 1979. Issues in Assessing the Contribution of R&D to ProductivityGrowth. Bell Journal of Economics 10, pp. 92-116.

    Hayes, D., 1989. Behind the Silicon Curtain: the seductions of work in a lonely era.London: Free Association Books.

    Jensen, R., Thursby M., 2001. Proofs and Prototypes for Sale: The licensing ofuniversity inventions. The American Economic Review, 91 (1), pp. 240-259.

    Leitch, C.M., and Harrison, R.T., 2005. Maximizing the Potential of University Spin-Outs: the development of second-order commercialization activities. R&DManagement, 35, pp. 257-272.

    Locket A., Wright M., Franklin S., 2003. Technology Transfer and Universities Spin-out Strategies. Small Business Economics, 20, pp. 185-200.

    Lowe, R., 2006. Who Develops a University Invention? The impact of tacit knowledgeand licensing policies. The Journal of Technology Transfer, 31, pp. 415-429.

    MacAdam, R., Keogh, W., Galbraith, B., Laurie, D., 2005. Defining and ImprovingTechnology Transfer Business and Management Process in University InnovationCentres. Technovation, 25, pp. 1418-1429.

    Macho-Stadler, I., Martnez-Giralt, X., Prez-Castrillo, J.D., 1996. The Role ofInformation in Licensing Contract Design. Research Policy, 25, pp. 43-57.

    Macho-Stadler, I., Castrillo-Prez, D., Veugelers, R., 2005. Licensing of UniversityInventions: The role of a technology transfer office. January 19. BBVA Working Paper;forthcoming in International Journal of Industrial Organization (2007).

    Markman, G.D., Giadionis, P.T., Phan, P.H., Balkin, D.B., 2005. Innovation Speed:Transferring University Technology to Market. Research Policy, 34, pp. 1058-1075.

    Martin, B.R., 2003. The Changing Social Contract for Science and the Evolution of theUniversity, in:

    Mazzoleni, R. 2006. The effects of University Patenting and Licensing on Downstream

    R&D Investment and Social Welfare. Journal of Technology Transfer, 31, pp. 431-441.

    Meussen, W., and Van den Broeck, J., 1977. Efficiency Estimation from Cobb-DouglasProduction Function with Composed Error. International Economic Review, 18, pp.435-455.

    Mowery, D.C., Nelson, R.R., Sampat, B.N., 2001. The Growth of Patenting andLicensing by U.S Universities: an assessment of the effects of the Bayh-Dole Act of1980. Research Policy, 30, pp. 99-119.

    Mowery, D.C, Rosenberg, N., 1998. Paths of Innovation: Technological Change in 20 th

    Century America. Cambridge University Press, New York.

  • 7/27/2019 University Distance Function Sfa

    35/46

    35

    Ndonzuau, F.N., Pirnay, F., Surlemont, B., 2002. A stage model of academic spin-offcreation. Technovation, 22, pp. 281-289.

    Nelson, R., 2001. Observations on the Post-Bayh-Dole rise in patenting at American

    universities. The Journal of Technology Transfer 26, 13-19.

    Ortn, P., Salas, V., Trujillo, M.V., Vendrell, F. 2007. El spin-off universitario en Espaa comomodelo de creacin de empresas intensivas en tecnologa. Estudio DGPYME.http://www.ipyme.org/IPYME/es-ES/Publicaciones/estudios/

    Owen-Smith, J., Riccaboni, M., Pammolli, F., Powell, W.W., 2002. A Comparison ofU.S and European university-industry Relations in the Life Science. ManagementScience, 48, pp. 24-43.

    Perez-Castrillo, D., (Coordinadora Isabel Bussom) 2005. La innovacin en Catalua;

    Las Oficinas de Transferencia de Tecnologia. Coleccin estudios CIDEM. Generalitatde Catalunya.

    Rasmussen, E., Moen, O., Gulbrandsen, M., 2006. Initiatives to PromoteCommercialization of University Knowledge. Technovation, 26, pp. 518-533.

    Reifschneider, D., and Stevenson, R., 1991. Systematic Departures from the Frontier: aframework for the analysis of firm inefficiency. International Economic Review 32, pp.715-723.

    Rubiralta, M., 2003. Transferencia a las empresas de la Investigacin Universitaria.

    Academia Europea de Ciencias y Artes. Espaa.

    Rubiralta, M., 2005. Transferencia a las Empresas de la Investigacin UniversitariaDescripcin de Modelos Europeos. COTEC.

    Serarols, C., Urbano, D., Vaillant, Y., 2007. Technological Trampolines for newventure creation in Catalonia: The case of the University of Girona. Eighteenth IRMAInternational Conference (Information Resources Management Association) VancouverCanada.

    Santamara, L., Barge, A., Modrego, A., 2007. Anlisis del Proceso de Transferencia

    Tecnolgica Universidad-Empresa.http://otri.uc3m.es/docweb/pct/2007/N7-comercializacion-abril07-estudiotecnologico.pdf.

    Shane, S., 2002. Selling University Technology: Patterns from MIT. Management Science 48,

    122-137.

    Shephard, R., W., 1953. Cost and Production Functions. Princeton University Press.

    Shephard, R.W., 1970. Theory of Cost and Production Functions. Princeton UniversityPress.

  • 7/27/2019 University Distance Function Sfa

    36/46

    36

    Siegel, D.S., Waldman, D., Link, A., 2003. Assessing the Impact of OrganizationalPractices on the Relative Productivity of University Technology Transfer Offices: anexploratory study. Research Policy, 32, pp. 27-48.

    Siegel, D.S., and Phan, P.H., 2004. Analyzing the Effectiveness of University

    Technology Transfer: Implications for Entrepreneurship Education (No. 0426).Rensselaer Polytechnic Institute, Troy.

    Thursby, J., Jensen, R., Thursby, M.C., 2001. Objectives, Characteristics and Outcomesof University Licensing: A Survey of Major U.S. Universities. Journal of TechnologyTransfer, 26, pp. 59-72.

    Thursby, J.G., and Kemp, S., 2002. Growth and Productive Efficiency of UniversityIntellectual Property Licensing. Research Policy, 31, pp. 109-124.

    Vendrell, F., and Ortin P., 2006. Technological Transfer from Universities: A

    theoretical review and an empirical analysis of Spin-Offs in Spain. Working Paperpresented in I International Business Economics WorkshopUIB. Palma de Mallorca, September 7 and 8th 2006. demo.uib.es

    Vohora, A., Wright, M., Lockett, A., 2004. Critical Junctures in the Development ofUniversity high-tech spin-out Companies. Research Policy 33, pp. 147-175.

    Williamson, O., 1985. The Economic Institutions of Capitalism, New York Press, NY:The Free Press.

    Zucker, L.G., Darby, M.R., Armstrong, J.S., 2002. Commercializing Knowledge:university science, knowledge capture, and firm performance in biotechnology.Management Science 48, pp. 138-153.

  • 7/27/2019 University Distance Function Sfa

    37/46

    37

    Figure 1: Linear model of the process of technology transfer from universities to firms.

    Source: Siegel et al. 2003; Friedman and Silberman 2003.

    Invention

    Disclosure

    Evaluation ofthe inventionfor patenting

    Patent

    Commercialization of the

    technology tothe firm

    Negotiation of

    a license

    License to thefirm (existent

    or start-up

    TTO

    Group ofInvestigation

    FIRM

  • 7/27/2019 University Distance Function Sfa

    38/46

    38

    Graphic 1. Evolution of the volume of R&D contracts signed*(in millions of Euros)

    *Contracts of R&D and Consultancy (art. 83), services and other activities. Data of 51 of the 60 universities.Source: REDOTRI Universidades

    Graphic 2. Evolution of the activity of Intellectual Property

    Data of 52 of the 60 universities.Source: REDOTRI Universidades

  • 7/27/2019 University Distance Function Sfa

    39/46

    39

    Graphic 3. Evolution of the number of license agreements

    Data of 48 of the 60 universities.

    Source: REDOTRI Universidades

    Graphic 4. Evolution of the income generated by license agreements

    Data of 41 of the 60 universities.Source: REDOTRI Universidades

  • 7/27/2019 University Distance Function Sfa

    40/46

    40

    Graphic 5. Evolution of the Spin-off created

    Data of 40 of the 60 universities.Source: REDOTRI Universidades

  • 7/27/2019 University Distance Function Sfa

    41/46

    41

    Table 1Utility Patents Issued to U.S Universities and Colleges, 1969-1997 year issue

    Year Number of U.S patents

    1969 1881974 2491979 264

    Bayh-Dole Act 1980

    1984 5511989 17801997 2436

    Source: Extracted from Mowery et al. 2001

    Table 2: Descriptive Statistics

    Variables 2003 2004 2005

    Inputs

    Inv. Disclosure 10,8 10,1 32,6TTO Size 12,9 13,4 14

    Outputs

    Licenses 4,4 9,5 30,2Spin-off 2 1,3 1,4

    Environmental factors

    Technical 16%Public 84%Posterior National

    Plan

    52%

    Near a Industrial

    center

    44%

  • 7/27/2019 University Distance Function Sfa

    42/46

    42

    Table 3: Results SFE with Panel Data

    Inputs:Invention disclosure (x1), size (x2)

    Outputs:Licenses (y1), Spin-off (y2)Environmental factors: Technical (z1), Public, Posterior National Plan (z3), Near an Industrial Centre (z4)

    Variables and ParametersModel 1

    Licenses &

    Spin-off(lny1&lny2)

    Model 2Licenses &

    Spin-off(lny1&lny2)

    Model 3Licenses

    (lny1)

    Model 4Spin-off

    (lny2)

    Intercept 0

    -0,736 -0,596 -0,112 -0,141

    Inputs ln x1

    1 0,215* 0,036 0,820*** 0,153***

    ln x2

    2

    0,785* 0,804* 0,271*** 0,188

    (ln x1)2

    11 0,361*** 0,374***

    (ln x2)2

    22 -0,218 -0,232

    (ln x1)(ln x2)

    12 -0,069 -0,048

    Outputs ln y2

    1 2,240*** 2,523***

    (ln y2)2

    11 -2,932*** -0,345***

    Inputs-Outputs

    (ln x1)(ln y2)

    12 -0,165 0,064

    (ln x2)(ln y2)

    22 0,132 -0,014

    Environmental z0

    0

    -9,692* -21,014* 0,014

    z1

    1 1,762** 3,692* 0,258

    z2

    2

    6,620** 7,720** -0,215

    z3

    3 0,195 2,443* 0,275

    z4

    4 -1,182** -1,920** -0,271

    Other ML

    Parameter

    2 0,192** 0,512*** 0,174* 0,474***

    0,915*** 0,561** 0,403*** 0,698

    Log-

    Likelihood-123,51 -128,86 -165,14 -149,68

    MeanEfficiency

    0,72 0,68 0,64 0,94

    Level of Statistical significance: *** 1%, ** 5%, * 10%

  • 7/27/2019 University Distance Function Sfa

    43/46

    43

    Table 4. Results of technology based universities vs. non-technology based universities.

    From Model 1

    Mean Efficiency Observations SignificanceProb>F

    Technology Baseduniversity

    0,648 9

    Other 0,721 141

    Total 0,717 150

    0,1782

    From Model 3

    Mean Efficiency Observations SignificanceProb>F

    Technology Baseduniversity

    0,623 9

    Other 0,646 141

    Total 0,644 150

    0,7218

    From Model 4

    Mean Efficiency Observations SignificanceProb>F

    Technology Baseduniversity

    0,917 9

    Other 0,834 141

    Total 0,839 150

    0,0018

  • 7/27/2019 University Distance Function Sfa

    44/46

    44

    Table 5. Technical Efficiency in SFE

    Technical Efficiency SFE Multi

    Output (from Model 1)

    Technical Efficiency SFE Multi

    Output (from Model 3)

    Year

    2003

    Year

    2004

    Yeat

    2005

    Rank

    Mean

    Year

    2003

    Year

    2004

    Yeat

    2005

    Rank

    Mean

    U. de Alcala de Henares 0,832 0,769 0,747 13 0,559 0,724 0,708 28

    U. de Alicante 0,821 0,816 0,770 11 0,713 0,744 0,772 10

    U. de Almera 0,703 0,715 0,768 26 0,543 0,679 0,719 29

    U. Autnoma de 0,756 0,754 0,816 14 0,714 0,674 0,765 13

    U. Autnoma de Madrid 0,722 0,783 0,708 23 0,663 0,736 0,655 21

    U. de Barcelona 0,326 0,877 0,819 35 0,299 0,841 0,718 33

    U. de Burgos 0,742 0,748 0,716 25 0,577 0,699 0,732 26

    U. de Cdiz 0,697 0,787 0,252 44 0,662 0,723 0,247 44

    U. de Cantabria 0,606 0,725 0,802 30 0,365 0,583 0,820 38

    U. Carlos III de Madrid 0,581 0,481 0,801 41 0,240 0,505 0,658 48

    U. de Castilla-La 0,876 0,622 0,451 390,750 0,799 0,489 23

    U. Complutense de 0,690 0,776 0,698 27 0,690 0,727 0,649 20

    U. de Crdoba 0,724 0,780 0,779 17 0,472 0,740 0,786 27

    U. de da Corua 0,480 0,689 0,353 49 0,250 0,665 0,167 49

    U. de Deusto 0,826 0,817 0,817 9 0,719 0,707 0,707 15

    U. Europea de Madrid 0,911 0,877 0,842 3 0,798 0,806 0,637 9

    U. de Extremadura 0,882 0,692 0,815 12 0,774 0,692 0,777 8

    U. de