Entrepreneur: Start & Grow Your Business

Synergies or trade-offs in university life sciences research.


by Foltz, Jeremy D.^Barham, Bradford L.^Kim, Kwansoo
American Journal of Agricultural Economics • May, 2007 • increasing returns (scale and scope economies) in the production of three major life science research outputs: patents, articles, and doctorates analyzed

Major legislative, legal, and technological changes paved the way for a period of remarkable growth in the patenting of life science research by U.S. universities in the 1980s and 1990s. (1) During this time, the relative importance of life science patents granted to U.S. universities grew from 10% of all university patents in 1980 to almost 25% in 1999. (2) This dramatic expansion in the role of life sciences research occurred in a period when the annual number of patents granted to U.S. universities grew almost tenfold from 340 patents granted in 1980 to 3,274 in 1999. Similarly, funding to support research and education activities in the life sciences at major research universities nearly doubled in constant dollar terms, with an especially rapid expansion occurring in the 1990s (National Science Foundation 2000). As a leading edge technology for U.S. universities, life science patenting has clearly become a major research priority over the past two decades, but many would ask, at what cost? This article analyzes a particular aspect of this question by identifying the impact of life science patents on other university life science research outputs, namely, published articles and doctorates.

Researchers concerned about university-level trade-offs associated with the expansion of patenting tend to focus on three potential negative outcomes: (a) universities moving away from basic research to pursue commercial patents (Kennedy 2000; Dasgupta and Ray 1994; Blumenthal et al. 1996); (b) universities placing priority on the establishment of intellectual property rights instead of on knowledge generation and idea sharing and those intellectual property rights making public research more difficult (Rai and Eisenberg 2003; Campbell et al. 2002; Blumenthal et al. 1996); and (c) university research quality declining as patent activity increases (Henderson, Jaffe, and Trajtenberg 1998; Sampat, Mowery, and Ziedonis 2003). All three of these trade-offs can be translated into university research production outcomes, the first two into fewer and lower-quality journal articles and potentially fewer doctorates, and the third into lower-quality patents or articles (i.e., ones with fewer citations). Concerns about these tradeoffs are especially heightened in discussions of patenting trends in public, land grant universities, which historically have been viewed as institutions dedicated to creating public goods in their research and teaching enterprises (Atkinson et al. 2003).

Most quantitative research on the impacts of academic patenting has focused on effects outside the university. Some important examples are the Jensen and Thursby's (2001) examination of the private investment incentives associated with universities having the right to offer exclusive licensing of their patents, and the Zucker, Darby, and Brewer (1998) exploration of the synergy between top scientists and biotech firms where universities and companies are proximately located. (3) However, only with respect to the evolution of patent quality has there been any systematic empirical analysis done on the effects of increased patenting on university research performance, with the most recent evidence on patent citations suggesting no significant changes (Sampat, Mowery, and Ziedonis 2003). One case study at MIT has shown complementarities between patents and other research outputs for university scientists (Agrawal and Henderson 2002), but its results are limited to two departments at the top patenting university in the country. In the sociological literature, Owen-Smith and Powell (2003) suggest that high-quality research generates both highly cited articles and a rich pool of potential patent opportunities for enterprising technology transfer offices to exploit. This positive outcome is the main "synergy" of interest in this article; that is, the degree to which there are scope economies associated with high-quality research generating both patents and traditional research outputs (articles and trained students) in a more cost effective manner than if those research outputs were produced separately.

Using panel data for U.S. universities, this article explores the evidence for two types of increasing returns (scale and scope economies) in the production of three major life science research outputs: patents, articles, and doctorates. These measures are important indicators of potential synergies associated with the portfolios of university research outputs. While they are not, in and of themselves, direct welfare measures, they can help to identify whether more production of these university research outputs, separately or jointly, is more cost efficient. The methods used below allow the construction of both "overall" scope and scale estimates as well as a distribution across university sizes and type to investigate whether scale and scope outcomes are more prevalent in life sciences research.

The methodological approach builds on Baumol, Panzar, and Willig's (1988) framework by constructing a university multiple-output cost function. We present panel data estimates of the multiple-output cost function from fixed-effects and random-effects models for both quantity and quality-adjusted outputs.

The panel data econometrics advance previous cost-function estimations aimed at identifying the underlying properties of university production processes, as do the quality adjustments made on output quantities. These empirical innovations are made possible by a data set that combines annual data from 1981 to 1998 for ninety-six U.S. universities on life science research expenditures, patents, journal articles, and doctorates, including citation data for the patents and journal articles that can be used to construct quality-adjusted output measures. The analysis starts with nonparametrically smoothed costs surfaces that provide visual evidence of scope and scale economies in the production of patents and articles, especially among universities with medium to large production levels. Econometric estimates from a series of panel data models provide the basis for testing systematically for the presence of economies of scale and scope. The estimates reveal economies of scale in both the quantity and quality-adjusted data, and economies of scope in only the quality-adjusted data. Comparisons across university types reveal the strongest scale and scope economies in land grant universities.

The organization of the article is as follows. In the next section, an explanation is given for how scale and scope are measured in a multiproduct cost function. The estimation strategy is presented in the third section followed by a section introducing the panel data set on U.S. university life science research, and explaining how the quality-adjusted research outputs are constructed. In the fifth section, we provide the results of the empirical analysis, which is followed by a conclusion.

Measuring Scale and Scope in a Multiproduct Cost Function

Standard analyses of patent production both in industry and at universities, (Hausman, Hall, and Griliches 1984) have used a production function approach to estimate the determinants of patent production. Building on Arora (1995), a recent piece by Graff, Rausser, and Small (2003) tests for complementarities in reduced form production function models among private firms. These techniques rest heavily on key assumptions regarding the nature of complementarities and the validity of some exclusion restrictions, which are unlikely to be satisfied in the typical university setting where output prices are difficult to measure. (4)

A more promising line of inquiry for identifying synergies or trade-offs among multiple outputs involves using the dual, i.e., a cost-minimization framework as set forth by Baumol, Panzar, and Willig (1988). Since their work on scale and scope economies first appeared, this cost function approach has been applied extensively to many sectors including universities (de Groot, McMahon, and Volkvein 1991; Cohn, Rhine, and Santos 1989). Previous university applications of the Baumol, Panzar, and Willig (1988) framework involve either cross-sectional analyses, or pooled versions of panel data.

Typical multiproduct cost function estimations are based on a version of the following equation,

(1)

C(Y, w) = [a.sub.o] + [summation over (j)] [b.sub.j][Y.sub.j] + 1/2 [summation over (j)] [summation over (k)] [C.sub.jk][Y.sub.j][Y.sub.k]

+ [summation over (l)] [d.sub.l][w.sub.l] + 1/2 [summation over (l)] [summation over (m)] [d.sub.lm][w.sub.l][w.sub.m],

where C(Y, w) is the total cost of producing a vector of outputs Y with a vector of input prices w, and [a.sub.o], [b.sub.j], [c.sub.jk], [d.sub.l], [d.sub.lm] are scalars. (5) The coefficient estimates, [b.sub.j] and [C.sub.jk], are then used as evidence for synergies and trade-offs and as arguments in the construction of estimates for ray economies of scale and economies of scope using formulas presented below. In order for the cost function to be valid it must satisfy homogeneity of degree 1 in input prices. The procedure for ensuring this is described below in the empirical implementation section.

Ray Economies of Scale and Scope

Following the standard formulas in Baumol, Panzar, and Willig (1988), we calculate ray economies of scale and scope from the cost function parameters. They are calculated as follows:

1. Ray economies of scale: The ray economies of scale for the joint production process are defined by:

(2) [S.sub.n](Y) = C(Y)/[[summation].sub.j] [Y.sub.j][partial derivative]C(Y)'/[partial derivative][Y.sub.j]

where ray economies of scale exist if [S.sub.n] (Y) is greater than 1.

2. Economies of scope: The economies of scope for a product set t relative to the product set of all other n products not including t: (n - t), can be computed from following function:

(3) [SC.sub.t](Y) = [C([Y.sub.t]) + C([Y.sub.n-t])

- C(Y)]/C(Y),

where C([Y.sub.t]) is the cost of producing only the product set t and C([Y.sub.n-t]) is the cost of producing the other n products except those in set t. Economies of scope exist when [SC.sub.t](Y) > 0.

In this application, we analyze economies of scope that compare producing patents as a separate operation from articles and doctorates and producing all three together as a single operation.

Econometric Specification

In the case of university research output in the life sciences, the vector of outputs Y is measured by journal publications, patents, and doctorates, while the costs are measured by the total expenditures on life sciences research in a given year. To control for the presence of university-specific effects in the error structure, panel data are used to estimate a panel data model, such as that presented in equation (4).

(4) [C.sub.it] = [alpha] + [x.sub.it][beta] + [u.sub.it], where [u.sub.it] = [v.sub.i] + [[epsilon].sub.it],

where [C.sub.it] are costs, [x.sub.it] represent the independent variables (Y, w), [beta] is a vector of parameters to be estimated, [v.sub.i] is a university-specific residual estimated as either a fixed or random effect, while [[epsilon].sub.it] is the "usual" residual which contains both a time-specific element and a standard equation residual (Wooldbridge 2002).

We estimate two versions of the econometric specification under different assumptions on the two error terms. A fixed effects model, which estimates [v.sub.i] separately for each university, is presented first. A random effects specification is then estimated to accommodate a number of regressors that change infrequently and to include a number of indicator variables that parameterize the differences between universities in ways that we predict will affect the cost of research production, e.g., presence of a longstanding tech transfer office and medical school, and land grant status. The random effects estimator imposes the assumption that [x.sub.it] and the random effects, [v.sub.i], are uncorrelated. The results of a Hausman test of this assumption along with tests for significance of the random effects and for heteroskedasticity are described in the results section.

In terms of the functional form of the multiple-output cost framework, the literature presents a number of variants, including the generalized quadratic and the translog forms. Since a key independent variable, patents, is zero for nearly a quarter of the university-year combinations the translog formulation would be undefined for a large part of our sample unless we added an ad hoc small number to each of these data points. Given that the literature provides no specific guidance on the optimality of one functional form over another; we therefore have chosen the generalized quadratic because it minimizes the number of ad hoc assumptions necessary for implementation. None of the key results about economies of scale or scope presented in this article are sensitive to our choice of functional form. (6)

In order to meet the theoretical requirements of a cost function, we impose homogeneity in input prices in the manner suggested by Chambers (1988):

(5) C = [w.sub.l] f (y, [w.sub.m]/w.sub.l]),

where [w.sub.l] and [w.sub.m] are two different input prices. The interpretation and estimation are facilitated if one divides through by [w.sub.l] to get the equation to be estimated as:

(6) C/[w.sub.l] = f (y, [w.sub.m]/[w.sub.l]).

The choice of normalizing input price is discussed below in the data section.

Finally, the econometric models estimated below use both strict quantity measures for research output and quality measures in which citations of articles and patents are used to control for quality of those two research outputs. The specifics of this citation adjustment are discussed next in the data section.

Data on University Life Science Research, 1981-1998

The data set combines information on life science research inputs and outputs for ninety-six U.S. universities over an eighteen-year period, spanning an era of remarkable growth in the role of life sciences in universities and the global economy. We focus on the segment of life sciences--biological and agricultural sciences--that has been most affected by recent court rulings in the United States that allow patenting of life forms. Following the National Science Foundation's (NSF) definition of "life sciences," these categories include departments that produce most biotechnologies and agricultural science research, but exclude departments that are primarily engaged in clinical medicine (see Appendix for a complete listing). This choice is consistent with a historical division within most universities, where biological and agricultural life sciences are contained in distinct administrative units from medical and pharmaceutical schools. The ninety-six U.S. universities roughly correspond to the Carnegie classification of "Research I" universities, and they are responsible for the vast majority of U.S. university production of articles and patents in life sciences. (7) The exact choice was driven in large part by the availability of accurate article and cost data.

For the dependent variable in our estimation, [C.sub.it], we use university life science research costs as measured by the NSE The university's outputs, the elements of the vector [Y.sub.it], are measured as life sciences patents, articles, and doctorates. (8) Life science patent assignee and citation information were extracted from the NBER patent database (Hall, Jaffe, and Trajtenberg 2003), while the Science Citation Index (ISI Web of Science 2002) provided the life science article and citation counts by year for each university. Patents are credited by application year rather than by grant date in order to measure them as close as possible to the date research costs were involved. In addition, although our cost measure does not include teaching costs, we include the university's undergraduate student to faculty ratio as a method of controlling for differences in teaching loads that might influence the costs of research.

We use three input costs, [w.sub.it], in the cost function estimation: average faculty salary, average wage rate in the university's town as a measure of the cost of support personnel, and an index of overall costs in agricultural research compiled by Huffman and Evenson (2005). In order to preserve homogeneity of degree 1 in prices we divide all input prices and our dependent variable, research costs, by the research cost index. Since this also has the effect of deflating our cost variables, we otherwise use nominal values of the variables.

Also to capture the university's level of technology transfer infrastructure we include two indicator variables: one captures whether the university has a technology transfer office, while the other captures whether a university had a technology transfer office before the promulgation of the Bayh-Dole act in 1980. Finally, in the random effects models, we include three variables to control for missions of universities that may be poorly measured in the outputs variables we use. They are (a) LGU an indicator variable for whether a university has land grant status, (b) "Extension FTE" a measure of the number of extension personnel (measured in FTE) in the state (Ahearn, Lee, and Bottom 2002), and (c) "Med School" for whether a university has a medical school. We expect that land grant universities will have higher base costs (positive coefficient) because of their multiple outreach missions of providing public goods to the state. These outreach missions may be poorly measured by the research outputs we are including. We therefore include the extension FTE measure, which we expect to be positive since it represents a major component of the outreach mission and we expect that servicing a larger set of extension personnel could raise the costs of research. (9) The medical school dummy variable should be negative if we have counted outputs from medical programs without accounting for their costs, or insignificant otherwise.

We estimate two sets of regressions, one using quantity output measures and another adjusting the quantity of articles and patents by their citation counts as a measure of quality. (10) Citation adjustments were sought because in the case of research output, quality is likely to matter significantly to the implicit value of the research and also to the potential synergies between patents and articles. In the first case, highly cited articles and patents are likely to generate flows of additional research or licensing funds to the author or assignee, while in the latter research that gives rise, for example, to an article that is highly cited may also be more likely to generate a patent than would a larger number of uncited articles. Empirically, studies of patent citations have shown that they provide a proxy for both the quality of a patent and knowledge spillovers from patents, because each time a new patent uses a piece of research from another patent it is obligated to cite the previous patent (Henderson, Jaffe, and Trajtenberg 1998). Article citations are also commonly used as measures of quality in studies of departmental or university quality, e.g., Adams (1998).

Using citations requires attending to the time and subject dependency of the counts, namely, the truncation problem associated with more recent articles or patents that may not have had time to generate many citations as well as different citation rates across disciplines (Sampat, Mowery, and Ziedonis 2003). The citation adjustment measure constructed here for each life science article/patent is the deviation from the average citation rate of an article/patent in the same broad patent class or disciplinary category published in the same year. For example, a 1995 biochemistry article with ten citations is compared to the average level of citations of all biochemistry articles produced in 1995. For a given year, the average article within a disciplinary category has a citation rate of 1, with higher-quality articles then having a measure greater than 1 and lower-quality articles receiving a measure between zero and 1. This relative citation approach minimizes a truncation bias that would be introduced using an absolute citation count. Further details on the citation measure are in the appendix.

Empirical Results

Descriptive Statistics

A useful starting point for considering the issue of trade-offs or synergies between university life science article, patent, and doctorate production is an aggregate view of the recent trends in those outputs. Table 1 demonstrates the tremendous takeoff in life science patent production at U.S. universities in the 1990s, with the number of accepted patents in 1998 at sixteen times the level of 1981. Table 1 also shows the approximately 50% growth in published life sciences articles from 1981 to 1998 and the 33% growth in life science doctorates at a time in which life science R&D expenditures grew 88% in real terms. These growth rates in life science production are quite remarkable when contrasted with Foltz et al.'s (2005) estimates of only a 1% yearly rate of technical change in the university production process. (11)

The growth in life science article production shows steady growth over the entire period averaging about 2.4% per year, with the most rapid growth period being between 1984 and 1992. Patents show short growth spurts in the 1980s and then stable growth until 1995 when three years of exponential growth occurred. Doctorates, meanwhile, grew most in the early 1990s. While the boom in life science patenting in the late 1990s may have been fueled by the growth in life science article production in the earlier period, the leveling off of all three research outputs at much higher levels at the end of the 1990s suggests that at least strict tradeoffs among articles, doctorates, and patents during the boom era of life science patenting did not occur. It is possible, nonetheless, that the explosion in patent activity in the latter part of the decade may have dampened the other forms of research production.

Cost Surfaces

While table 1 demonstrates the growth of university life science outputs, it does not give evidence on potential complementarities between outputs. Descriptive evidence of economies of scale and scope can be seen in the realized cost surfaces of university production choices. A cost surface (or region) with cost complementarities will be convex with respect to costs across the two outputs, higher along the edges where more of a single product is produced and lower in the middle where both products are produced. A cost surface exhibiting returns to scale in a single product will be concave to the origin along one output axis.

Descriptive evidence on the shapes of university life sciences research cost surfaces is presented in figures 1 and 2 using a nonparametric Lowess smoothing estimation procedure and the pooled data set. (12) The relationship between articles and patents in quantity space is shown in figure 1. With its strong concavity along the article axis, it suggests significant returns to scale in article production, and with both convex and concave regions in the article-patent plane it also appears to show some regions of cost complementarities along with some regions of trade-offs. For example, one major convex region appears between thirteen and twenty-two patents and 660 to 1,080 articles. Also noteworthy is the plateau at the upper end of the article distribution, above 1,700 articles per year, where increases in either articles or patents appear relatively costless. This provides some suggestion that returns to scale and economies of scope may exist for the most productive/largest universities.

[FIGURES 1-2 OMITTED]

The second nonparametric cost surface (figure 2) depicts the citation-adjusted cost relationship between articles and patents. Along the article axis, the initial slope of this surface shows much steeper costs than did the quantity version, suggesting that quality research articles do not come cheaply. At higher levels of quality-adjusted article output, however, economies of scale do appear and persist. The cost surface also shows approximately the same inflection points for the region of convexity between articles and patents, but overall this surface is less suggestive of cost complementarities than was the surface in quantity space.

Econometric Estimates

The nonparametric cost surfaces obviously do not control for other factors and therefore only provide suggestive evidence about scale and scope economies. The next step in the analysis is to estimate for both quantity and citation-adjusted outputs the life science research cost function using panel data methods. Estimates are presented in tables 2 and 3 using a panel of 1,563 data points from eighty-seven universities over eighteen years (1981-1998). Each table presents two regression models: a fixed effects and a random effects regression. The tables also show the chi-square ([chi square]) statistic from the Hausman test of random versus fixed effects, the Breusch Pagan test that the random effects parameter [v.sub.i] is different than zero and two t-tests for the most likely types of heterskedasticity, that the estimated variance is different (a) by university and (b) by year.

The dependent variable is university life science research and development expenditures measured in thousands of dollars. In addition to the quadratic formulation for the three research outputs and the two input costs we include a number of regressors to control for possible unmeasured differences. We include the university-wide undergraduate student-to-faculty ratio in order to control for potentially higher research costs for places with higher undergraduate teaching responsibilities. The regression also includes an indicator variable for whether the university is a land grant (LGU) institution, a medical school dummy variable, and the two technology transfer variables described above.

Table 2 presents fixed effects and random effects parameter estimates for the quantity model, while table 3 presents the citation-adjusted parameters. The tables also show the results of the tests of the error terms, while table 4 presents the estimates of scale and scope that are derived from inserting the parameters into equations (2) and (3) from the regression estimates. In terms of regression diagnostics, all equations have reasonably high [R.sup.2]'s. For the random effects models, we cannot reject the null hypothesis of homoskedasticity, while the Breusch Pagan test shows that we can reject the null hypothesis that [v.sub.i] = 0. For the quantity regressions, the insignificant Hausman test implies that we cannot reject the random-effects model as the correct model, while we can reject the random-effects model specification for the citation-adjusted results.

All of our model specifications provide similar and highly significant results for most of the regressors. The coefficient estimates for articles and patents in all regressions show that their production increases costs, but at a decreasing rate. These significant estimates provide supportive evidence of the necessary conditions for scale economies in those outputs. The parameter estimates on graduate student production have more ambiguous but statistically insignificant effects. The interaction terms between outputs show significant trade-offs between PhD's and both articles and patents, although in the citation-adjusted regressions the significance of the PhD/article trade-off disappears. The negative, though insignificant, coefficient on the article/patent interaction term suggests some possibility of synergies between these outputs.

As anticipated, both the faculty salary and LGU variables positively and significantly increase research costs, while staff wage is positive but not significant. The insignificant parameters on the undergraduate-to-faculty ratio suggest that, at an aggregate level, undergraduate teaching responsibilities do not spill over to a great degree onto research costs. Schools with medical schools did not have significantly different costs than those without, which supports the division we have imposed between the life sciences and other related parts of the university We find no significant effect of extension personnel on overall research costs, suggesting that the higher base costs at land grant universities in life sciences research come from sources other than the extension mission. The technology transfer office variables provide some surprising results, with the existence of a technology transfer office causing an increase in overall research costs. This effect is partially muted by the negative estimated parameter for those with technology transfer offices in existence before 1980, but that estimated parameter is not significant. Overall, this result is suggestive of trade-offs between increased technology transfer activities and overall research costs. (13)

While the regression coefficients provide suggestive evidence of scale and scope by output, estimates of ray economies of scale and scope derived using equations (2) and (3) provide the global measures of interest. These are estimated using the regression coefficient estimates and values of the independent variables in the formulas for ray economies of scale and scope. These are presented two ways: in table 4 using the mean of the independent variables and the regression estimates from tables 2 and 3, while table 5 presents median scope estimates for different types of universities (public/private and large/small) using the random effects parameters and independent variables for each of the universities to generate a distribution of scope and scale estimates. In table 4, the mean scale and scope estimates are tested using nonlinear Wald tests, which takes into account the variance of the estimated parameters and tests whether scale = 1 or scope = 0. Significance tests in table 4 are denoted by asterisks on the coefficients.

Table 4 shows significant estimates of increasing returns to scale, with the citation-adjusted regressions exhibiting larger scale economies. Table 5 demonstrates that these returns to scale are greatest at land grant universities, while nonland grant universities show lower-scale economies that approach constant returns to scale for the quantity regressions.

The scope estimates have more varied patterns in tables 4 and 5. The mean estimates from the quantity regressions show no significant evidence of economies of scope, while the citation adjusted models do exhibit significant economies of scope. This suggests that synergies between patents and other research outputs are most pronounced in the production of high-quality outputs. In the median estimates presented in table 5, the estimates of economies of scope are larger, especially for land grant universities, and are greatest for the small land grant universities.

Overall, the results for the median university in table 5 suggest that economies of scale and scope are the strongest for land grant institutions. Moreover, the finding in table 5 that these economies are even stronger in the citation-adjusted measures suggests that quantity and quality of articles and patents go hand in hand. The cost advantages that these increasing returns may provide the leading universities could cause divergence in productivity and overall performance even among Research I universities.

Conclusions

This work has estimated cost functions for university life science research using panel data methods in order to investigate economies of scale and scope. In contrast to much of the literature on academic patenting, the dual formulation used here allows an explicit estimate of cost complementarities and obviates the need to specify prices for research outputs. The results demonstrate the benefits of using panel data to take into account time- and university-specific effects as well as the importance of taking into account quality in measuring university outputs.

In contrast to a literature that has worried about both the declining quality of university patenting and an increased commercialization of the academic enterprise due to patenting especially in the life sciences, the results show evidence of economies of scope between patents and other missions of research universities in the life sciences. Once one adjusts for the quality of the output, our data suggest significant synergies between patents and other research outputs. This implies that rather than declining patent and article quality due to the increase in university patenting, we find evidence of lowered costs for producing high-quality outputs simultaneously.

The synergies between patents and traditional research outputs are especially evident for land grant universities. They exhibit the highest levels of economies of scale and scope, although they also have higher base costs as evident in the large and positive coefficient on the LGU dummy variable. We find that these higher base costs are not directly related to their extension mission though they may come from the expanded mandate land grant universities have to provide public goods to their states. The efficiency in the production process evident in higher-scale and scope economies for land grant universities may come from the discipline imposed by two decades of shrinking state budgets and legislative oversight, or may be due to different internal organizational structures. Whatever the cause, the strong economies of scale and scope in life science research among land grant institutions suggest that these universities have a distinct cost advantage in the production of high-quality life sciences outputs.

Our results leave some key issues on the effects of patenting on university life science research open for further research and analysis. The advent of technology transfer offices appears to increase costs in the life sciences rather than reduce them. While this effect may be due to the relative immaturity of the technology transfer process during our study period, this effect is significant and robust to alternative specifications. It suggests that there is a long learning curve to the operation of an effective technology transfer office before it generates positive synergies to a university.

In addition, the estimations show evidence of trade-offs between graduate student training and both patent and article production. The fact that this effect is stronger for patents than for articles suggests some trade-offs with respect to the long-term effects of the Bayh-Dole act, if research productivity in articles and patents comes in part at the expense of training the next generation of scientist. Future research with university level cost functions, perhaps at the level of all university outputs, might be able to shed more light on the potential trade-offs between training graduate students and other outputs.

While this work has found some synergies at the university level in the production of life science outputs at the university level, it leaves open a number of questions on how far reaching these results are. Do these synergies exist for all scientific outputs? Are they the product of aggregating to the university level or are they also present within individual labs or even faculty members? We plan to investigate these issues in future research.

Data Appendix

Academic Departments

We follow the National Science Foundation's NCES classification of disciplines for the agricultural and biological sciences. This definition includes what are generally the life science departments that do most research, but excludes clinical medical departments. The following broad department groups are included in the NSF definition of agricultural and biological sciences:

Agricultural: agricultural chemistry, agronomy, animal science, fish and wildlife, forestry, horticulture, plant sciences, aquaculture, soil sciences, landscape architecture, conservation, renewable natural resources.

Biological: anatomy, cellular, and developmental biology; biochemistry/chemistry; biostatistics and epidemiology; ecology and organismal biology; foods and nutrition; general biology/bioscience; genetics and molecular biology; microbiology and immunology; pathology; pharmacology and toxicology; physiology and biophysics; veterinary sciences.

Patents

Patent data were culled from the NBER patent database, where they were identified as having a university assignee. Patents assigned to the University of California system were associated with a campus (Berkeley, Davis, Los Angeles, etc.) by the location of their authors through searches of campus directories.

Patents were categorized as life sciences based on the categories and subcategories in Hall, Jaffe, and Trajtenberg (2003, pp. 452-53). Patents were chosen in the NBER subcategories 33 (biotechnology as part of the drugs and medical cat