BEMPS Bozen Economics & Management Paper Series

Transcription

1 BEMPS Bozen Economcs & Management Paper Seres NO 19 / 014 The productvty of French Technology Transfer Offces after government reforms Clauda Cur, Cnza Darao, Patrck Llerena

2 The productvty of French Technology Transfer Offces after government reforms Clauda Cur Faculty of Economcs and Management, Free Unversty of Bolzano-Bozen emal: Cnza Darao Department of Computer, Control and Management Engneerng Antono Rubert, Unversty of Rome La Sapenza Patrck Llerena Unversty of Strasbourg, BETA (Bureau d'econome Théorque et Applquée) and Observatore des Scences et Technques (OST, Pars) Ths verson: May 4, 014 Abstract Ths paper assesses the productvty change of the French Technology Transfer Offces (TTOs) after the ntroducton of the July 1999 Innovaton Law and the New Publc Management Orented Reform. By usng Data Envelopment Analyss (DEA)-based Malmqust Productvty Index, we fnd an average ncrease n the short-term productvty of the French TTOs drven by both postve effcency and technology change. The bootstrap analyss reveals that these mprovements are ascrbed to 50% of the TTOs system, whle the remanng part does not show sgnfcant changes. Moreover, whle older TTOs postvely contrbute to the performance of French TTOs n the short run, young TTOs wth hosptal seems to contrbute negatvely to the effcency of the entre system. Keywords: Technology Transfer Offces (TTOs), French Unversty System, Malmqust Productvty Index, Data Envelopment Analyss (DEA), Bootstrap. JEL Classfcaton: C14; D4; I9; O3.

3 1. Introducton It s well known that unversty technology transfer plays a decsve role n the development and growth of local and natonal economc development (e.g.; Caldera and Debande, 010). Over the past two decades, almost all European countres have passed through a transton perod that has been accompaned by the nsttutonalzaton of Technology Transfer Offces (TTOs), namely those specal-purpose organzatons n charge of transferrng technology, and regulatory nterventons (European Commsson, 004, 008). As today, t s unclear whether EU Technology Transfer Offces (TTOs) have receved a boost n productvty after the ntroducton of polcy nterventons, as there s very lmted analyss on ths matter due to a lack and heterogenety of EU data (Bonaccors and Darao 007a,b; Darao et al. 011). The am of ths paper s to contrbute to the current debate of TTOs performance n Europe by analysng the productvty of the French TTO system. Contrary to other countres (.e. the UK, the US, Japan and Germany [.e. Etzkowtz, 1998; Phlpott et al. 011; Mller et al. 01]), the French natonal government has delayed reforms to formalze unversty technology transfer actvtes. However, two man reforms have been undertaken snce The July 1999 Innovaton Law consttutes the man polcy decson and s amed at enhancng technology transfer processes between unverstes and ndustry. After ts ntroducton, the number of TTOs flourshed. In 001, the New Publc Management Orented Reform was ntroduced to accelerate and mprove the qualty of the technology transfer process. Ths nterventon ncludes: () the defnton (and development) of the fundng scheme based on TTOs performance; () the dentfcaton of the key drvers of TTOs performance and () the defnton of the metrcs amed at montorng the effcency of publc spendng, partcularly n the scence and technology feld. The metrcs nclude patent applcatons as well as patent extensons and smlar ntellectual property rghts nstruments for software (Bach and Llerena 007). After the ntroducton of the reforms, all types of metrcs have regstered an ncrease n number, whch have been nterpreted as good performance ndcators of the entre French TTOs. However, the smple ncrease n these metrcs cannot support evdence of substantal productvty growth of the French TTOs technology transfer process to draw conclusons on whether the system underwent a real productvty. Ths because the French TTOs technology

4 transfer process s a complex process characterzed by anomales n the output producton and dfferences related to the TTOs afflated unversty, age and presence of a hosptal. Frst, compared to other countres such as the US and the UK, the French technology transfer process s at ts early development as characterzed by outputs whch are produced ntermttently and t does not nclude the marketng stage of the product. Second, for some unverstes, the presence of TTO nternal to unversty s not a new concept as, before 1999, some unverstes developed ther own organzatons, although not nsttutonalzed, n charge of technology transfer. For ths set of TTOs, the producton process s more mature compared to the new TTOs and outputs are produced at hgher frequency. The ncrease n output producton s not homogenous across the dfferent unversty categores the TTO belongs to as well across type of outputs. For nstance, TTOs related to Engneerng Schools ncreased ther producton of patent applcatons over tme and produced hgher amount than the Polyvalent Unverstes wth Medcal School. However, they produced less amount of patent extenson compared to UPAM. Lastly, the presence of a hosptal s not strctly related to a medcal school but, for nstance, unverstes specalsng n scence (Scence Unverstes) may have both medcal schools and unversty-related hosptals, whle Polyvalent Unversty may have only a medcal school. Unverstes wth related hosptals carry out sgnfcant ongong medcal research, whereas unverstes wth medcal schools are more nvolved n tranng actvtes. Ths paper attempts to contrbute to the scarce lterature of TTOs n Europe by analyzng the productvty growth of French TTOs after the ntroducton of the two government reforms. The am s to shed lght on how TTOs responded n the short term productvty to these polcy nterventons. We use a unque panel data avalable for the French unversty system (Bach and Llerena 008, 010) 1, whch s comprehensve enough to carry out a productvty assessment based on Malmqust Index (Färe et al. 1994). For the specfc case of TTOs, the productvty of a TTO s a measure of the best organzatonal and manageral practces that allow producng the maxmum feasble level of outputs gven avalable nput resources. Malmqust Productvty Index, computed through the Data Envelopment Analyss (DEA), allows estmatng the productvty change and ts man drvers, namely effcency and technologcal change. Whle the effcency change refers to possble mprovements (or deteroratons) n busness practces through approachng (or movng away from) the best organzatonal and manageral practces (.e.; the fronter), the 1 It s worth mentonng that these are the most recent data avalable on the French unversty TTO system as the French government has stopped collectng these data. 3

5 organsatonal and techncal change refers to the possble mprovements (or deteroraton) of the best organzatonal and manageral practces through output mx expanson (or decreases) (.e. expanson or reducton of the technologcal fronter) gven the resources (nputs) used. In ths study we extend the use of Malmqust Index, already appled n several felds and also n the TTOs effcency estmaton (Thursby and Thursby 00; Km 011), to the bootstrapped Malmqust Productvty Index (Smar and Wlson 1999). By usng the bootstrappng technque, we are able to uncover nformaton on the actual changes occurred n the French TTOs system and excludng those nducted by samplng varaton, whch would lead to a based estmaton. In other terms, our approach provdes a productvty and effcency assessment based only on the real changes n productvty. Summng up, we attempt to fll the gap exstng n the lterature by provdng the frst country-level assessment of performance of the young system of TTOs operatng n Europe based on output ndcators brought about by the governmental reforms of the early 000s. We address the followng man research questons: 1) Dd French TTOs experence postve effcency and technologcal gans rght after the ntroducton of the government reforms?; ) Is there any group of TTOs (n terms of unversty category, age and presence of a hosptal) able to boost the productvty of the entre system n the short term? The paper s organzed as follows. Secton provdes a selected overvew of the lterature, and Secton 3 descrbes the methodology appled. Secton 4 descrbes the data, and Secton 5 dscusses the man results. Secton 6 concludes.. A selected revew of the lterature The emprcal lterature related to the analyss of the producton process productvty of TTOs s mnmal and focuses on US data and covers dfferent tme spans. The poneer paper (Thursby and Thursby, 00) nvestgates the man drvers of the productvty growth of 65 US unverstes over the perod by usng a lnear three-stage process and DEA Malmqust Productvty Index. They found that ncreased lcensng was due prmarly to an ncreased wllngness of faculty and admnstrators to lcense and ncreased busness relance on external R&D rather than a shft n faculty research. The second paper, by Km (011), extends the prevous analyss to a sample of 90 US TTOs by trackng the year-to-year productvty growth over the perod The author consders the technology transfer process as one-stage process and fnds that there was More specfcally, nventon dsclosure s the output of the frst stage, patents applcaton s the output of the second stage and lcenses and opton agreements are the outputs of the thrd stage). 4

6 substantal competence demonstrated n the productvty growth of US unversty technology transfer actvtes for the perod , contrary to the prevous research fndngs. These postve gans n productvty are ascrbed to the ncreasng wllngness of unverstes to mprove commercalzaton to gan a hgh return on nvestment. However, the author does not dsentangle the sources of productvty growth. There s a broader lterature, however, on the factors whch affect the TTO s effcency. Wth regards to specfc characterstcs of dfferent unversty categores, several papers (e.g. Thursby and Kemp, 00; Cur et al. 01) fnd effcency dspartes across TTO groups attrbutable to unversty producton specalzaton (such as basc research vs. appled research vs. teachng) rather than to competences n lcensng. More specfcally, t s generally assumed that unverstes wth medcal schools are lkely to be more effcent than those wthout because t s easer for them to conduct clncal trals and produce a large number of unversty lcenses related to bomedcal nventons. Thursby and Thursby (00) fnd that the presence of a medcal school postvely affects the performance of US TTOs at the nventon dsclosures and patent applcatons stages but negatvely affects performance at the last stage of lcense agreements, whereas Km (011) shows that the presence of a medcal school does not by tself guarantee hgh productvty n technology transfer. It s also generally assumed that a TTO could beneft from a learnng by dong experence whch s drectly proportonal to ts senorty. However, the emprcal evdence shows that t mght play a dual role, dependng on the strateges pursued by the TTO s management. Segel et al. (003) and Cur et al. (01) show that older US and French TTOs, are more effcent. However, Segel et al. (008) fnd opposte results. Thursby and Thursby (00) do not fnd any evdence on the mportance of learnng by dong. Table 1 summarzes the man results of nne selected studes based on both effcency and productvty assessment that are related to the lterature dscussed n ths paper. The present paper attempts to assess the productvty growth of EU TTOs usng France as a specal case wth the am to uncover evdences on the statstcally sgnfcant changes of Malmqust Productvty Index (and relatve components) n terms of TTOs groups and TTO s specfc characterstcs, namely afflated unversty, age and presence of a hosptal. Because ths more reslent methodology allows dsentanglng the heterogenety across TTOs groups, t should be effectve for clarfyng the actual effect of effcency change and/or technologcal change on productvty and how they have responded to the French polcy nterventon n the short-run. 5

7 [TABLE 1 AROUND HERE] 3. The methodology We examne productvty change through a Malmqust Productvty Index (Färe et al. 1994) approach based on effcency measures computed through Data Envelopment Analyss (DEA). DEA s a non-parametrc estmator of the effcent fronter and determnes the benchmark aganst whch to compare the performance of each TTO. The dstance of a TTO from the fronter ndcates ts neffcency score (Farrell 1957). The productvty change (hereafter ndcated by M whch stands for Malmqust) can be decomposed n effcency change (EC) and organsatonal and techncal change (TC). The effcency change captures how close (or far) the TTOs move to (or away from) the fronter and nforms about the possble catchng-up of TTOs. The organsatonal and techncal change captures the shft upward (or downward) of the fronter and nforms about possble new busness practces adopted by TTOs. From a practcal vewpont, the DEA-based Malmqust Index s very useful to carry out our analyss as: () t does not mpose any assumpton on the functonal form of the producton fronter, () t allows for the smultaneous use of multple nputs and outputs and () t does not requre the knowledge of fnancal nformaton, whch s not avalable for the French TTO database. Untl recently (see, e.g., Thursby and Thursby 00; Km 011), DEA has been appled as a smple lnear programmng approach to estmate effcency scores wthout any statstcal nvestgaton on ther sgnfcance. The novelty of ths study from a methodologcal vewpont s to perform the TTO productvty assessment wthn an nferental settng, adoptng the bootstrap procedure suggested by Smar and Wlson (1999). The crucal advantage of ths approach s the removal of bas from each estmate so that we are able to derve bas-corrected estmates (BC-M, BC-EC, and BC-TC), as well as the dentfcaton of confdence ntervals for each estmate so to derve the statstcal sgnfcance of each (at a gven level of sgnfcance %). The producton model s based on the assumpton that French TTOs share a common producton fronter and promote the maxmzaton of ther commercal outputs and s n lne wth the polcy objectve ntroduced by the two man reforms (see the Introducton secton). We compute a Malmqust output-orented ndex (M) for each TTO, between tme t 1 and t as follows: 6

8 M 0.5 t, t t, t1 t1, t1 t1, t D D D t1, t t1, t EC t1, t1 t, t TC t1, t D D D (1) t1, t1 t1, t1 t1 t1 where D D ( x, y ) s the Farrell output dstance (.e. the maxmum proportonal expanson of the outputs gven the nputs used, see Farrell, 1957) of each TTO at tme t 1 t1, t t1, t t1 t1 relatve to the technology at tme t 1, and D D ( x, y ) s the dstance relatve to the t t1 t, t1 t t, t t, t t t technology at tme t. D D ( x, y ) and D D ( x, y ) are respectvely the, t dstance of each TTO at tme t to the technology at tme t 1 and t. The two components of the Malmqust ndces are t 1 EC,t and t 1 TC,t, whch represent the effcency change and technologcal change, respectvely. t 1,t M, t 1 EC,t and t 1 TC,t greater (less) than unty ndcate regress (progress) between tmes t 1 and t. Values equal to unty ndcate no change. However, relaton (1) does not allow determnng whether changes n productvty, effcency and technology are real or merely artfacts of the fact that we have to estmate the producton fronters because we do not know the true ones (Smar and Wlson, 1999). Thus, we employ a consstent bootstrap estmaton procedure for correctng for the bas and obtanng the confdence ntervals on t 1 M,t, t 1 EC,t and t 1 TC,t. The dea underlyng ths approach s to generate a sequence of pseudo datasets to obtan bootstrap estmates for each estmate by usng the bvarate kernel densty estmaton-where the bandwdth s selected followng the normal rule of thumb- and the adaptaton of the reflecton method. From the bootstrap sample assocate to each estmate, we compute the bas-correcton terms and the confdence ntervals by selectng the approprate percentles 3. Now, we test whether changes are statstcally sgnfcant. In partcular, f a confdence nterval contans the unty (that means no change), then the change turns out to be not statstcally sgnfcant; whereas f the confdence nterval does not contan the unty, then the change turns out to be statstcally sgnfcant. Only estmates that fall nto the latter category are consdered n the evaluaton of the productvty growth because they carry statstcal sgnfcance and are not the result of samplng varaton (for techncal detals, see Smar and Wlson, 1999). 3 More detals on the bootstrap procedure are avalable upon request. 7

9 4. Data descrpton Our sample conssts of a balanced data panel of French TTOs collected by BETA (Bureau d Econome Théorque et Applquée, UMR UdS-CNRS 75, Strasbourg) wth the support of the natonal French TTO network (CURIE), the Conference of Unversty Rectors (CPU) and the Assocaton of Engneerng Schools Drectors (CDEFI) from the years 005 and We evaluate the French TTO system adoptng the operatonal perspectve as ndcated by French law. We thus dstngush between core output, whch ncludes patent applcatons (PAT_APP), and ancllary output, whch ncludes number of patents wth submtted extenson requests (PAT_EXT), number of extensons requred (Nb_PAT_EXT) and software applcatons (SW_APP). As nputs, we select labour, proxy by the number of fulltme equvalent employees (ETP), as the man means whch enables a TTO to operate. In addton, we select the knowledge, proxed by the scentfc publcatons (PUB), produced by the afflated unversty, whch s consdered as some knd of raw materal at the TTO s dsposal. We model the technology transfer process of French TTOs at ther early stage of development by accountng for the msmatchng n the producton process due to outputs produced at a very slow pace (.e. dscrete output flow), whle nputs used at a contnuous flow. By assumng a two-year producton cycle, that s two years from the start of nput usage to delver the output, we construct the producton model as follows. On the nput sde, we select the average nputs used over two consecutve years n lne wth prevous papers (e.g. Thursby and Kemp 00; Anderson et al. 007; Cur et al. 01), whle on the output sde we select the cumulatve output producton over two consecutve years. In ths way, we reduce the error from tme lags. From an emprcal pont of vew, ths model wpes out two changes, lmtng the analyss to one productvty change, namely the perod vs. the perod In terms of sample composton, we nclude TTOs actve n the producton of at least patent applcaton actvty (the core output) and one of the remanng actvtes. We end up wth a database of 30 TTOs coverng all the French unversty categores 5, except the socal scence and humantes, law and economcs unverstes because ther producton s close to zero n all the technology transfer actvtes. Our sample s well representatve of the entre 4 See Bach and Llerena (006, 008, 010). 5 They are: ) Engneerng School (ING), () Polyvalent Unversty wth Medcal School (UPAM); () Polyvalent Unversty wthout Medcal School (UPSM) and (v) Scence Unverstes (USC). 8

10 sector as: () PAT_APP consttutes 94.3% and 94.56% of the total PAT_APP over the two perods, respectvely; and () the remanng outputs represent 94.15% and 96.67% of the total producton of the ancllary outputs over the two perods, respectvely. Table presents the summary statstcs of the nput and output varables assocated wth the pooled sample over the two perods and , accordng to the categores of the dscplnary felds of the related unverstes. We dscuss the ancllary output broken down nto ts components to better understand how French TTOs dversfed ther outputs durng the perods under analyss. These statstcs enlghten a general ncrease n SW_APP for all TTO categores-except for UPSM and a decrease n PAT_APP for some unversty categores 6. Only Engneerng School (ING) shows a consderable postve growth (.e %) n PAT_APP. Decrease n PAT_EXT and NB_PAT_EXT are also regstered n most of unversty categores, expect for UPSM. We observe a consderable ncrease n the average amount of ETP 7 along wth a modest ncrease n the amount of stock of PUB. [TABLE AROUND HERE] In terms of dstrbutons of nput and output usage, not surprsngly, USC unverstes employ, on average, the largest amount of ETP and have the largest amount of technology stock. USC unverstes produce, on average, the largest amount of patent applcatons, as well as the largest amount of the remanng actvtes. ING and UPAM, on average, seem to have smlar patent applcaton producton over the perod but not n the followng perod. Evdence n output varablty s partcularly found across USC TTOs. We classfy TTOs nto TTOs wth and wthout hosptals as well as TTOs that are young versus those that are mature. In contrast to other studes, we do not consder the TTO s ownershp (prvate vs. publc) as all French TTOs are related to publc unverstes. On the other hand, the TTOs senorty, measured as the length of tme that has passed snce the creaton of the technology transfer offce, accounts for possble learnng by dong effects n technology transfer actvty. In fact, older TTOs could beneft from ther experence compared to younger TTOs. We could, thus, expect a postve mpact on TTOs effcency f, for nstance, TTOs learn to focus ther strateges of technology transfer on transformng the best nventons nto patents and hence produce a source of ncome afterward. In addton, n France, there are USC unverstes wth both medcal schools and unversty-related hosptals, whle UPAM unverstes may have only medcal schools. It s, therefore, more nformatve 6 Namely, UPAM s the category that suffers the most (-38.1%), followed by USC (-0.71%) and UPSM ( %). 7 Namely, by category, UPSM (+63.64%), UPAM (+56.47), ING (+35.36%) and USC (+30.47%). 9

11 n the French case to control for the presence of a unversty-related hosptal. The presence of a unversty-related hosptal mght guarantee sgnfcant ongong medcal research, whereas a smple medcal school mght have only tranng actvty. It s usually thought that medcal research s an mportant source of technology transfer. However, t mght happen that nsttutons, unverstes and unversty-related hosptals, whch are legally ndependent enttes, try to capture the potental technology transfer comng from the lfe and medcal scences. For ths reason, the mpact of the presence of unversty-related hosptals s ambguous and mght depend on the competton between the two nsttutons (namely the unversty and the unversty-related hosptal). Accordng to the lterature, n fact, results are uncertan and, n the French case, Cur et al. (01) fnd that the presence of a unverstyrelated hosptal s detrmental to effcency. [TABLE 3 AROUND HERE] Table 3 shows a descrptve analyss of the varables AGE and HOSPITAL across TTOs grouped n dfferent dscplnary felds. It appears that ING (.e. engneerng unverstes) are the most experenced (mean: years), whle UPSM unverstes (.e. polyvalent unverstes wthout medcal schools) are the youngest. The presence of a hosptal, n contrast, s a specfc characterstc assocated wth UPAM and USC unverstes. These evdences show the heterogenety wthn the French TTOs system and renforce the use of our methodologcal approach rather than the use of an analyss based on smple nomnal growth rate, whch fals to capture the multple dmensons of TTO producton. 5. Emprcal results Before carryng out the TTOs productvty assessment, we formally test f the fronter globally exerts constant, non-ncrease or varable returns to scale (Smar and Wlson 00), acknowledgng that f the technology exhbts varable returns to scale (VRS), then the Malmqust Productvty Index s a based estmate of productvty change (Grfell-Tatjé and Lovell, 1995) and alternatve ndces have to be used. Snce our analyss s based on the evaluaton of two fronters (one for the perod and the other one for the perod ), we perform two tests. We fal to reject H0 hypothess at 5% confdence level for both fronters (p-values are equals to 0.19 and respectvely, thus both greater than 10

12 0.05), acceptng global CRS 8. Gven these results, we can assume constant returns to scale fronters and thus use the Malmqust Productvty Index n our productvty assessment. Regardng the peculartes and heterogenety between the groups of the French TTOs, presented n the prevous secton, we report the geometrc mean of the effcency change, the techncal change and the Malmqust Productvty Index, aggregated accordng to the entre system, unversty category, presence of a hosptal and relatve age. Followng our am of uncoverng the actual productvty (effcency and technologcal) change, we compute the geometrc mean (.e.; Geom.Mean-all) ncludng all the estmates (both statstcally sgnfcant and not) from the set of TTOs under nvestgaton and compare aganst the geometrc mean of only those estmates wth statstcal sgnfcance (.e. Geom.Mean-subset) whch carry economc meanng. Estmates are reported bas-corrected and expressed à la Farrell so that scores greater than one mples a declne whle scores less than one mply a growth. Changes n productvty expressed n percentages are computed as the dfference between one and the Farrell scores. [TABLE 4 AROUND HERE] Table 4 shows an overall ncrease (6.6%) n productvty over the perod of the French TTOs system as a whole. Ths fndng s consstent wth the productvty growth assocated wth each TTO unversty category, except for USC TTOs whch experence an average declne of -7.1%. Ths can be ascrbed to that fact that USC TTOs are the only category whch experence declne n both outputs (PAT_APP and ANC_OUT) of the technology transfer process. In lne wth Thursby and Thursby (00) for US TTOs at ther early stage of development, the overall productvty gan seems to be drven by postve organzatonal and technologcal change (3.99%) whch s, however, dampened by negatve effcency change (-15.8%). Although there s evdence of overall postve productvty growth assocated to each group, effcency and technologcal changes seem to have played dfference roles between groups. ING and UPSM unverstes experenced postve change n both effcency and technology. On the other hand, TTOs assocated wth UPAM and USC experence, on average, postve technologcal change but negatve effcency change leadng to a gan n productvty for UPAM and a declne n productvty for USC. Followng Thursby and 8 Ths fndng s dfferent from the fndng presented by Cur et al. (01) because of the dfferent TTOs sample composton. For nstance, n ths paper we exclude TTOs assocated to socal scence and humantes, law and economcs unverstes whle they are ncluded n Cur et al. (01). 11

13 Thursby (00), we dstngush between unverstes wth and wthout hosptals 9 (Table 5) as possble source of productvty heterogenety mght be found between these two groups due to dfferences n the commercalzaton process strateges. Our fndngs show that French TTOs wthout hosptals regster postve changes n effcency and technology whle TTOs wth hosptals regster postve change only n technology. The last nvestgaton s wth respect to TTOs senorty (Table 6). Mature (older than seven years old 10 ) and young TTOs seem to perform accordng to dfferent patterns. Young TTOs deterorate n effcency but mprove n technology, whereas older TTOs exhbt postve changes n both effcency and technology. Thus, t seems that, after the ntroducton of the polcy reforms, the French TTOs system responded wth postve productvty change but suffered from deteroraton n effcency rather than n technology. Ths mght be ascrbed to the fact that there are TTOs, mostly young TTOs, laggng behnd and fndng dffcultes n catchng-up the best busness practces of the older TTOs. Also, the presence of a hosptal hnders the adopton of best busness practces. On the other hand, the ntroducton of new legslaton encourages TTOs to become more compettve n mprovng ther busness models n terms of organsaton and technology. Turnng to the results for the bas-corrected actual values (estmaton based on confdence ntervals), we more precsely fnd that 14 TTOs (47% of the sample) do not experence sgnfcant change n effcency and 3 TTOs (77%) do not experence technologcal change. These results suggest that contrary to the smple pont estmates used below, only few TTOs postvely mprove ther busness models n terms of organsaton and technology whle the majorty of French TTOs are subjected to effcency change (both postve and negatve). More specfcally, the majorty of ING and UPSM TTOs do not show change ether n effcency or n technology. Among UPAM TTOs, fve and seven out of 1 do not show sgnfcant changes n effcency and technology, respectvely. Among USC TTOs, four out of 1 do not change n ther effcency and any of USC TTOs experence technology change. Ths analyss based on the statstcal sgnfcance of changes uncovers that the entre system as a whole s not senstve to productvty change but t could be dentfy a small group of TTOs that experenced real postve technology change and a larger one senstve to 9 Ths because n France there are Scence Unverstes wth both medcal schools and unversty-related hosptals, whle a Polyvalent Unversty wth a Medcal School may have a medcal school wthout a hosptal (see Table 5). 10 We assume seven years as the mnmum age for TTOs to be consdered mature. We compute senstvty analyss on our estmates assumng age rangng between 5 and 9. Results do not change. 1

14 effcency change. Taken together, t suggests that the almost half of the French TTOs system (54%) mproves the productvty n the short run whle the remanng part s dvded nto a group that shows any productvty change (16%) and the other group (30%) deterorates ts productvty. Wth respect to the presence of a hosptal, the analyss based on the confdence nterval reveals that effcency and technologcal changes assocated wth TTOs wthout hosptals are not statstcally sgnfcant. On the contrary, TTOs wth hosptals regster statstcally sgnfcant changes n effcency, both postve and negatve and postve technologcal changes for some of them. These fndngs confrm the dfferent nature of busness models adopted the two forms of TTOs. [TABLE 5 AROUND HERE] Wth regards to the senorty of TTOs, the statstcal analyss of effcency and technologcal change shows that wthn mature and young TTOs, t s possble to dentfy a subset of TTOs whch exhbt sgnfcant changes. However, whle young TTOs regster a negatve effcency change, the older TTOs show effcency gans. Regardng the technologcal change, mature and young TTOs exert postve gans n technology. [TABLE 6 AROUND HERE] Overall, t seems that the French TTO system responded postvely to the ntroducton of the government nterventons as 50% of TTOs mproved ther productvty through mprovements n technology and effcency. Among the dfferent TTO categores, UPAM TTOs are the most actve n mprovng ther busness models as well as operatonal effcency, along wth USC TTOs. However, ING and UPSM TTOs show stable productvty behavour. Our fndng hghlghts the presence wthn the French TTOs system of a subset of TTOs whch proactvely reacted to the natonal reforms. Ths mples that further polcy measures are desrable to enhance the productvty of those TTOs facng problems n changng up wth best practse and those whch need more stmulas to make process. To nvestgate the equalty n productvty change, we look at the dstrbutons of DEA estmates of Malmqust Index (M) effcency change and organsatonal and techncal change assocated wth each French TTO. Fg. 1 dstngushes between unversty categores and accordng to the presence of a hosptal. Lookng at Fg. 1, mxed evdence s found wth respect to effcency change (top panel). ING TTOs perform statstcally dfferently, whle wthn UPSM we fnd no dfferences. However, wthn the UPAM and USC categores, TTOs are clustered nto four groups, whch perform smlarly wthn each group, respectvely. 13

15 Wth respect to the presence of a hosptal, we observe heterogenety n performance wthn the UPAM and USC groups. In addton, f we compare UPAM TTOs wth hosptals to USC TTOs wth hosptals, we do not fnd statstcally sgnfcant dfferences. Thus, we cannot statstcally conclude that the dfferences between TTOs effcences can be entrely attrbutable to the presence of a hosptal. On the other hand, the technologcal change dstrbutons show that, to a large extent, TTOs perform smlarly. Ths mght suggest that n the short term, technologcal change does not depend strctly on the category nor on the specfc characterstcs. [FIGURE 1 AROUND HERE] Fg. reveals other nterestng aspects. It shows a lack of systematc dfferences between young and older TTOs; n the same way, there s a lack of systematc dfferences between TTOs wth and wthout hosptals. For some categores, such as ING and UPSM, older and young TTOs seem to perform smlarly to a large extent. However, f we compare Fgs 1 and, we can observe that young TTOs assocated wth hosptals exert the hghest negatve effcency change. Thus, we cannot statstcally conclude that the dfferences n TTOs effcences can be entrely attrbutable to the age of the TTOs. The dfference n effcency change we found mght pck up some specfc TTOs characterstcs that go beyond category, age and the presence of a hosptal. For nstance, dfference n effcency change mght be due to specfc characterstcs of the related unversty, such as whether or not the unversty s well establshed or young. Ths needs further research. [FIGURE AROUND HERE] 6. Conclusons Ths paper s ntended prmarly to examne the productvty of European TTOs after the ntroducton of the major reforms of the early 000s. Emprcal evdence s provded by usng a unque and orgnal dataset on the French TTOs system collected by BETA (Unversty of Strasbourg) over the perod The novelty s twofold: frst, ths s the frst productvty assessment of the French TTOs system; second, ths s the frst study wheren productvty changes and ts components are assessed n an nferental settng. We focus our analyss on a short tme perod, namely The use of bascorrected ndces revealed a postve productvty change, drven by effcency and technology changes. Moreover, the analyss of the confdence ntervals suggest that the 14

16 productvty have been changed sgnfcantly only for half of the TTOs. In addton, TTOs seems more senstve to effcency changes than to technologcal changes, as most TTOs dd not mprove ther busness models. Furthermore, TTOs wth hosptals seem to contrbute the most to performance changes. Whle older TTOs postvely contrbute to the performance of French TTOs n the short term, young TTOs wth hosptal seem to contrbute negatvely to the effcency of the entre system. Ths study confrms that the productvty of TTOs s more affected by ther operatonal effcency rather than ther ablty n mprovng ther busness models. However, the stmulus provded by the reforms enabled some TTOs to push upward the fronter. The combnaton of senorty and presence of a hosptal consttute an obstacle to reap productvty gan n the short run. More polcy actons are needed n ths regard. A possble area for future research could be to provde a wder analyss across countres of the determnants of productvty growth by examnng the relatonshp between government polcy and TTOs strategy on the man components of productvty change. 15

17 References Anderson, T.R., Dam, T.U., Lavoe F.F. (007) Measurng the effcency of unversty technology transfer, Technovaton 7, Bach, L. and Llerena. P. (006) Les actvtés de recherche contractuelle et de transfert de technologe dans les établssements franças d ensegnement supéreur, Rapport Fnal, Bach L, Llerena P, pour le Mnstère de l Ensegnement Supéreur et de la Recherche (MESR), Pars. Bach, L. and Llerena, P. (007) Indcators of hgher-educaton nsttutes and publc-research organzatons technology transfer actvtes: Insghts from France, Scence and Publc Polcy, 34, 10, Bach, L. and Llerena, P. (008) Les actvtés de recherche contractuelle et de transfert de technologe dans les établssements franças d ensegnement supéreur, Rapport Fnal, Bach L, Llerena P, pour le Mnstère de l Ensegnement Supéreur et de la Recherche (MESR), Pars. Bach, L. and Llerena, P. (010) Les actvtés de recherche contractuelle et de transfert de technologe dans les établssements franças d ensegnement supéreur, Rapport Fnal, Bach L, Llerena P, pour le Mnstère de l Ensegnement Supéreur et de la Recherche (MESR), Pars. Bonaccors, A. and Darao. C. (007a) Theoretcal perspectves on unversty strategy, n Unverstes and strategc knowledge creaton. Specalzaton and performance n Europe. (Eds) A. Bonaccors and C. Darao, Edward Elgar, Cheltenham, pp Bonaccors, A. and Darao. C. (007b) Unverstes as strategc knowledge creators. Some prelmnary evdence, n Unverstes and strategc knowledge creaton. Specalzaton and performance n Europe. (Eds) A. Bonaccors and C. Darao, Edward Elgar, Cheltenham, pp Caldera, A. and Debande, O. (010) Performance of Spansh unverstes n technology transfer: An emprcal analyss, Research Polcy, 39, 9, Chapple, W., Lockett, A., Segel, D.S., Wrght, M. (005) Assessng the relatve performance of unversty technology transfer offce n UK: Parametrc and non-parametrc evdence, Research Polcy 34, 3, Cur, C., Darao, C., Llerena, P. (01) Unversty technology transfer: How (n)effcent are French unverstes? Cambrdge Journal of Economcs 36, 3, Darao, C. et al (011) The European unversty landscape: A mcro characterzaton based on evdence from the Aquameth project, Research Polcy, 40,

18 European Commsson (004) Improvng Insttutons for the Transfer of Technology from Scence to Enterprse. Best Project ITTE1.11/00. DG Enterprse, European Commsson. European Commsson (008) Commsson Recommendaton on the Management of Intellectual Property n Knowledge Transfer Actvtes and Code of Practce for Unverstes and other Publc Organzatons. Brussels, , C(008) 139. Etzkovtz, H. (00) Incubaton of ncubators: Innovaton as a trple helx of unverstyndustry-government networks, Scence and Publc Polcy, 9, Färe, R., Grosskopf, S., Norrs, M., Zhang, Z. (1994) Productvty growth, techncal progress and effcency change n ndustralzed countres, Amercan Economc Revew, 84, Farrell, M.J. (1957) The measurement of productve effcency, Journal of the Royal Statstcal Socety Seres A, 1, 0, Grfell-Tajé, E., Lovell, C. (1995) A note on the Malmqust productvty ndex, Economcs Letters, 47, Km, Y. (011) The vory tower approach to entrepreneural lnkage: Productvty changes n unversty technology transfer, Journal of Technology Transfer do: /s Mller, K., McAdam, M., McAdam, R., Teague, S. (01) The development of unversty technology transfer stakeholder relatonshps at a regonal level: Lessons for the future, Technovaton, 3, 1, OST (006, 008, 010) Rapport de l Observatore des Scences et des Technques, Economa, Pars, Edtons 006, 008, and 010. Phlpott, K., Dooley, L., O Relly, C., Lupton, G. (011) The entrepreneural unversty: Examnng the underlyng academc tenson, Technovaton, 3, Segel, D. (007) Technology transfer offces and commercalzaton of unversty ntellectual property: Performance and polcy mplcatons, Oxford Revew of Economc Polcy, 3, 4, Segel, D., Waldman, D., Lnk, A. (003) Assessng the mpact of organzatonal practces on the relatve productvty of unversty technology transfer offces: An exploratory study, Research Polcy, 3, Segel, D., Wrght, M., Chapple, W., Lockett, A. (008) Assessng the relatve performance of unversty technology transfer n the US and UK: A stochastc dstance functon approach, Economcs of Innovaton and New Technology, 17,

19 Smar, L. and Wlson, P.W. (1999) Estmatng and bootstrappng Malmqust ndces, European Journal of Operatonal Research, 115, Smar, L. and Wlson, P.W. (00) Non parametrc test of return to scale, European Journal of Operatonal Research, 139, Thursby, J.G. and Kemp, S. (00) Growth and productve effcency of unversty ntellectual property lcensng, Research Polcy, 31, Thursby, J.G. and Thursby, M.C. (00) Who s sellng the vory tower? Sources of growth n unversty lcensng, Management Scence, 48, 1,

20 Tables of the paper Table 1: Selected results from the lterature on TTOs effcency and productvty TTOs Number of outputs Authors Output used MED AGE PUB GROUP a Effcency analyss France Multple Cur et al. 01 Patent and software applcatons, patents wth submtted extenson requests, number of extenson requred - + yes US One Segel et al. 003 Number of lcenses Anderson et al. 007 ne ne ne Lcensng ncome ne + ne Multple Thursby and Kemp 00 Number of lcenses, patent applcatons, nventon dsclosures, amount of royaltes and ndustry sponsored research - + b yes UK One Chapple et al. 005 Number of lcenses ne + Lcensng ncome - ne Span One Caldera and Number of lcenses Debande 010 ne ne - ne Lcensng ncome ne + ne + US & UK US Multple Segel et al. 008 Number of lcenses, ncome from lcenses, unversty start-ups One Thursby and Thursby 00 Productvty analyss Inventon dsclosures, Patent applcatons, Lcenses executed ne Multple Km 01 Number of patents, Number of lcenses, Lcense ncome ne Notes: MED = medcal school; AGE = TTO age; PUB = publc unversty; GROUP = group characterstcs. ne ndcates no effect; + ndcates postve effects; - ndcates negatve effects; and yes ndcates that there s an mpact, and t depends on what we control for. a Ths effect s controlled usng dfferent approaches. b In ths case, the authors test whether (or not) a prvate unversty affects TTO performance.

21 Table : Descrptve statstcs by unversty categores Unversty category Statstcs ETP PUB PAT_APP SW_APP PAT_EXT Nb_PAT_EXT ETP PUB PAT_APP SW_APP PAT_EXT Nb_PAT_EXT ING Mean SD Mn Max UPAM Mean SD Mn Max UPSM Mean SD Mn Max USC Mean SD Mn Max Source: Authors calculatons

22 Table 3: Descrptve statstcs by categores of unverstes Unversty category Varable N.Obs Mean SD Mn Max ING 3 Hosptal Age UPAM 1 Hosptal Age UPSM 3 Hosptal Age USC 1 Hosptal Age Sources: Bach and Llerena 006, 008, 010. Notes: Age has been computed startng from 006.

23 Table 4: Summary of results for French TTOs between 003 and 006 by unversty category Unversty category BC-EC BC-TC BC-M ING ** ING ** ** ING ** Geom.Mean-all Geom.Mean-subset Improvement 1 1 No change 1 Declne UPAM UPAM ** UPAM ** ** ** UPAM ** 0.777** 1.763** UPAM UPMA ** UPAM ** 0.79** ** UPAM ** ** UPAM ** ** UPAM ** 0.65** UPAM ** 0.736** ** UPAM ** Geom.Mean-all Geom.Mean-subset Improvement No change Declne UPSM ** UPSM ** ** UPSM ** Geom.Mean-all Geom.Mean-subset Improvement No change 3 0 Declne USC-5.571** ** USC ** ** USC ** ** USC ** USC ** ** USC ** ** USC ** USC USC ** ** USC ** ** USC ** USC ** **

24 Unversty category BC-EC BC-TC BC-M Geom.Mean-all Geom.Mean-subset Improvement No change Declne Geom.Mean-all Geom.Mean-subset Improvement No change Declne Notes: EC = Effcency Change, TC = Techncal Change, M = Malmqust Index, BC- = Bas-Corrected. ** Statstcally sgnfcant at 5%. Geom.Mean-all s calculated as the mean of changes assocated wth the entre sample, whereas Geom.Mean-subset s calculated as the mean of only statstcally sgnfcant changes. Improvement, No change and Declne ndcate the number of TTOs whch experenced statstcally sgnfcant progress, any sgnfcant change and statstcally sgnfcant regress, respectvely. 3

25 Table 5: Summary of results for French TTOs between 003 and 006 dependng on the presence of a hosptal Hosptal Unversty category BC-EC BC-TC BC-M Yes UPAM Yes USC-5.571** **.041** Yes USC ** ** Yes UPAM ** Yes UPAM ** ** ** Yes UPAM ** 0.777** 1.763** Yes UPAM ** Yes USC ** ** Yes UPAM ** 0.79** ** Yes USC ** ** Yes UPAM ** ** Yes USC ** ** Yes UPAM ** ** Yes UPAM ** 0.65** Yes UPAM ** 0.736** ** Yes UPAM ** Yes USC Yes USC ** ** Yes USC ** ** Yes USC ** Yes USC ** ** Geom.Mean-all Geom.Mean-subset Improvement No change Declne No UPSM ** No ING ** No UPSM ** ** No UPAM No ING ** ** No UPSM ** No ING ** No USC ** No USC ** Geom.Mean-all Geom.Mean-subset Improvement 0 6 No change 8 7 Declne

26 Table 6: Summary of results for French TTOs between 003 and 006 by age Young TTO Unversty category BC-EC BC-TC BC-M Yes UPSM ** Yes USC ** ** Yes UPSM ** ** Yes UPAM Yes UPAM ** Yes UPSM ** Yes UPAM ** 0.79** ** Yes USC ** ** Yes UPAM ** ** Yes USC ** ** Geom.Mean-all Geom.Mean-subset Improvement 1 4 No change Declne No UPAM No USC-5.571** ** No ING ** No UPAM ** No UPAM ** ** ** No UPAM ** 0.777** 1.763** No ING ** ** No USC ** ** No ING ** No USC ** No USC ** ** No UPAM ** ** No UPAM ** 0.65** No USC ** No UPAM ** 0.736** ** No UPAM ** No USC No USC ** ** No USC ** ** No USC ** Geom.Mean-all Geom.Mean-subset Improvement No change Declne

27 Fgures Fg. 1: Boxplot of TTOs effcency changes, technologcal changes and Malmqust Index by unversty category and presence of a hosptal. TTOs wthout hosptals are represented n blue and TTOs wth hosptals n green 6 5 ING UPAM UPSM USC 4 EC TTO 1.5 ING UPAM UPSM USC 1 TC TTO ING UPAM UPSM USC M TTO 6

28 Fg. : Boxplot of TTOs effcency changes, technologcal changes and Malmqust Index by unversty category and senorty. Mature TTOs are represented n blue and young TTOs n green 6 5 ING UPAM UPSM USC 4 EC TTO 1.5 ING UPAM UPSM USC 1 TC TTO ING UPAM UPSM USC M TTO 7