D. Felsensteincateg<strong>or</strong>y <strong>of</strong> education, rU_ = effect attributableto the jth categ<strong>or</strong>y <strong>of</strong> university interaction,yZ, = effect attributable to the kth categ<strong>or</strong>y <strong>of</strong>innovation level, y,?& = effect attributable to theinteraction between the ith categ<strong>or</strong>y <strong>of</strong> educationand the kth categ<strong>or</strong>y <strong>of</strong> innovation level, andyEVij = effect attributable to the interactionbetween the ith categ<strong>or</strong>y <strong>of</strong> education and the jthcateg<strong>or</strong>y <strong>of</strong> university interaction.Figures in parentheses are chi-square statisticsand degrees <strong>of</strong> freedom, all significant at thep < O-05 level, except f<strong>or</strong> the TEZik term that wasf<strong>or</strong>ced into the model.Likelihood ratio chi-square value f<strong>or</strong> estimatedmodel = 4.64 (p = O-0106, 2 degrees <strong>of</strong> freedom).This is the most parsimonious model fitted tothe data. While it shows that education, universitylinkage and innovation are all significant in theirown right, f<strong>or</strong> our purposes the interaction effectsare <strong>of</strong> greater interest. Although the second-<strong>or</strong>derrelationship between university interaction andeducation level <strong>of</strong> the entrepreneur/manager issignificant, this does not necessarily lead to innovativeactivity (the E*I relationship is not significant,but was f<strong>or</strong>ced into the model). The third-<strong>or</strong>derinteraction E*Z*U is also not significant. All thiswould seem to suggest that the ability to realizethe commercial potential <strong>of</strong> innovations (throughestablishing an innovative firm) is not necessarily<strong>related</strong> to academic education <strong>or</strong> university linkage.Success in innovation and its commercial exploitation(i.e. the ability to sell the innovation andkeep the firm going on this basis) are probably<strong>related</strong> to other supply conditions (such as theentrepreneur’s w<strong>or</strong>k hist<strong>or</strong>y and experience) andresult from demand fact<strong>or</strong>s such as market structure(f<strong>or</strong>eign <strong>or</strong> local, barriers to entry and so on).An exhaustive examination <strong>of</strong> these fact<strong>or</strong>s isbeyond the scope <strong>of</strong> this analysis, but a curs<strong>or</strong>yexamination <strong>of</strong> some <strong>of</strong> these fact<strong>or</strong>s does notyield any significant relationships. Supply fact<strong>or</strong>s,taken here as entrepreneur’s w<strong>or</strong>k experience asmeasured by the relationship between presentproduct and previous product experience (P),show no significant relationship to innovation.This holds true even when this relationship isstratified by previous employment position (R&Dvs. sales/administration/production). This wasexpected to add a further dimension to the depth<strong>of</strong> w<strong>or</strong>k experience, but the x2 statistics are allinsignificant. Market fact<strong>or</strong>s, as measured by thelocation <strong>of</strong> main market (f<strong>or</strong>eign <strong>or</strong> local) (M),where f<strong>or</strong>eign markets are expected to be m<strong>or</strong>einnovative, competitive and with higher entrybarriers, also yield no significant relationship toinnovation. However, there is a confounding effecthere with innovation and location, which will bediscussed below.6.2. Seedbed effects and <strong>science</strong> parklocationNearly half the firms surveyed here are locatedon <strong>science</strong> <strong>parks</strong>. This begs the question as towhether there is any significant difference ininnovative activity between on- and <strong>of</strong>f-park firms( i.e., does the park have a seedbed effect?) andwhether this relationship is confounded by anyother fact<strong>or</strong>s. In view <strong>of</strong> the popular perception <strong>of</strong>the <strong>science</strong> park as facilitating university-industryinteraction, and in the light <strong>of</strong> the paucity <strong>of</strong>empirical evidence supp<strong>or</strong>ting this claim, it isimp<strong>or</strong>tant to try to gauge the imp<strong>or</strong>tance <strong>of</strong> <strong>science</strong>park location in the innovation process.When turning to the features associated withlocation on a <strong>science</strong> park (and presumed toenhance innovative activity), the most obviousstarting-point is university interaction. In commonwith the many studies cited earlier, this is foundto be low amongst all <strong>science</strong> park firms surveyed.High-level interactions (joint research and industryfunding <strong>of</strong> university research) were rep<strong>or</strong>ted by13% and 9% <strong>of</strong> firms respectively. Mid-levelinteractions are not much m<strong>or</strong>e prevalent, withreceipt <strong>of</strong> university consultancy services rep<strong>or</strong>tedby less than 20% <strong>of</strong> <strong>science</strong> park firms, and keyemployees holding faculty positions rep<strong>or</strong>ted byonly 8%. As expected, low-level interactions basedon recruitment <strong>of</strong> local university graduates (28%)and use <strong>of</strong> university facilities (24%) were m<strong>or</strong>euniversal. This pattern, while not illustrating particularlyhigh-level interactions, was nevertheless104 Technovation Vol. 14 No. 2
Science <strong>parks</strong> - seedbeds <strong>or</strong> enclaves <strong>of</strong> innovation?significantly different to that observed f<strong>or</strong> non<strong>science</strong>park firms (x2 = 3.947, p = O-047).If <strong>science</strong> park firms have higher level interactionswith universities, does this result in technologytransfer and, as a consequence, high levels <strong>of</strong>innovation (the ‘seedbed’ hypothesis)? This causalrelationship is not particularly significant(x2 = 2.438, p = 0.118). However, the possibilitydoes exist that the relationship is mediated throughthe effect <strong>of</strong> some other fact<strong>or</strong>. As illustratedabove, innovation is inter<strong>related</strong> with inf<strong>or</strong>mationand much <strong>of</strong> this flows through channels that aregrounded in w<strong>or</strong>k experience, academic educationand the like. In this instance, theref<strong>or</strong>e, we testf<strong>or</strong> the interrelationship between the fact<strong>or</strong>s <strong>science</strong>park location (L), innovation level <strong>of</strong> the firm (Z)and w<strong>or</strong>k experience <strong>of</strong> the entrepreneur/manager(w).These relationships are depicted in Fig. 2. Ascan be seen, no clear pattern can be observed f<strong>or</strong>the relationship between innovation and <strong>science</strong>park agglomeration. When adding the w<strong>or</strong>k experiencedimension, we arrive at a series <strong>of</strong> pr<strong>of</strong>iles.The conventional path is represented by pr<strong>of</strong>ileA; an entrepreneur with a background in R&Dsets up a high-tech firm producing unique productson a <strong>science</strong> park. This development traject<strong>or</strong>y,however, accounts f<strong>or</strong> only 5% <strong>of</strong> all firmssurveyed. The maj<strong>or</strong>ity <strong>of</strong> <strong>science</strong> park firms(nearly 70%) fall into pr<strong>of</strong>ile G, which representsthe <strong>science</strong> park firm engaged in the productionand modification <strong>of</strong> existing products and foundedby an entrepreneur from a non-R&D background.When stratifying the relationship between <strong>science</strong>park and innovation by w<strong>or</strong>k experience, wefind that the seedbed hypothesis can be upheldindependently <strong>of</strong> w<strong>or</strong>k experience. Thus, f<strong>or</strong>firm founders with an ‘R&D background, thisrelationship is marginally significant (x’ = 2.93,p = 0.083). F<strong>or</strong> entrepreneurs with technical andproduction backgrounds this relationship is slightlystronger (x2 = 5.99, p = O-013). This suggests thatw<strong>or</strong>k experience might have a direct input intothe innovation capabilities <strong>of</strong> the firm (i.e. through‘learning by doing’ [48]). If this experience istechnical and managerial, this could lead to m<strong>or</strong>ecommercially viable innovative products than thoseproduced by firms where the main entrepreneurshave an R&D <strong>or</strong>ientation. In other w<strong>or</strong>ds, commerciallyexploitable innovations call f<strong>or</strong> m<strong>or</strong>e than justPROFILEABCDEFGHFig. 2. Firm pr<strong>of</strong>iles based on entrepreneur’s wbrk background (w), innovation level <strong>of</strong> the firm (l) and <strong>science</strong> park location (IL.).Technovation Vol. 14 No. 2 105