Entrepreneurial universities and overt opportunism

Abstract

Entrepreneurial universities are a catalyst for regional economic development and growth. Yet, the entrepreneurial university may not always capture the full gains made by the broader economy due to opportunistic behavior by its faculty scientists. We expand agency theory to address conditions when opportunistic behavior persists in the face of substantial information symmetry and where principals appear to tolerate opportunism despite the authority to sanction their agents. Using a sample of 105 US research universities and 73,603 scientists, we demonstrate that some scientists privately leak discoveries invented while working for their universities. Counter to prior prediction, we find that universities acting as principals often do not retaliate even when having knowledge of such opportunistic behavior and the capability to punish non-complying agents. To extend theory, we examined contexts that exacerbate and alleviate such overt opportunism: High-value discoveries and sizable entrepreneurial activity are associated with greater opportunistic behavior, whereas efforts to empower those who have close professional ties with scientists—their departments and technology commercialization offices—are related to reduced opportunism. To complement our empirical analyses, we provide a stylized mathematical model that shows how agency theory can improve its predictive power once it formally recognizes certain conditions under which agents wield substantial bargaining power over their principals, who in turn seem to tolerate agents’ overt opportunistic behavior.

Appendix: Mathematical model that simulates the licensing strategies of a scientist

Broadly, a scientist can pursue two strategies; she can either license her invention privately (via bypassing), or through a TCO (via disclosure). For brevity, we henceforth refer to these strategies as bypassing and disclosure, respectively, and we present a model that assesses these strategies by focusing on the effort that the TCO and scientist must exert to match the invention to the licensing partner that garners the most value added. Our aim is in delineating the payoffs behind each strategy by accounting for the parameters that we have introduced. This approach allows for precise comparisons that indirectly substantiate the causal links between the main variables and the hypotheses. We find that: (I) more autonomous TCOs are associated with less bypassing; (II) higher royalty payments to scientist’s are associated with less bypassing; (III) highly professional TCOs can offer a lower share of royalties to their scientists; (IV) entrepreneurial environments are associated with greater bypassing; (V) inventions of great value are not revealed to the TCO.

Assume a rational, forward-looking, profit-maximizing scientist and a TCO of similar characteristics. The scientist has created a prototype of an invention whose full potential is difficult to assess. The profits from licensing the invention depend not only on the prototype’s initial value, but also on the value added that a licensing firm can affix to the original. However, in finding the firm that can affix the most value-added one must exert some effort.

With that in mind, allow a scientist, denoted by i, to have created a working prototype of an invention. As discoveries are initially quite rudimentary, they require further development, without which a scientist can expect only a discounted stream of future royalties R if any. However, when a discovery is licensed to a firm and developed into a usable component or final product, it can yield a greater stream of future royalties V, where V > R, to the licensor; be it the scientist or the TCO. Because firms are endowed with varying capabilities to convert discoveries into products licensors exert effort e to find suitable licensees, where the more effort licensors put into finding a suitable firm the greater the return V to the licensor.⁸

Focusing on inventions for which R is greater than any transaction cost that the TCO/scientist faces; if a scientist pursues a bypassing strategy she fully appropriates the licensing proceeds π because the TCO and the university are not a party in the deal. If, on the other hand, she discloses the innovation to the TCO she garners only a share α of π; the remaining rent 1 − α goes to the TCO/university.

Since there may exist asymmetric information as to the prototype’s prospects, we denote the invention value (as viewed by the TCO) as γR, where γ ∊ R ⁺ models how proficient (compared to the scientist) the TCO is in finding a licensee that can increase the invention’s value added. The more autonomous the TCO is, and the stronger its ties with the business and scientific community, the greater the value of γ. Furthermore, regardless of licensing strategy licensors incur a search cost (denoted by c) for finding a licensee. This cost is a negative function of R (i.e. it is easier to license attractive inventions) and a positive function of e.

Focusing on bypassing, i a scientist’s profits from privately licensing her invention are,

$$\pi_{i} = V_{i} - c_{i} ,$$

(1)

Equation (1) expresses π _i in terms of i’s search cost (c _i ) and future stream of royalties (V _i ) where V _i is expressed as,

$$V_{i} = e_{i}^{\zeta } R^{1 - \zeta } .$$

(2)

Equation (2) captures V _i as an increasing function of the effort e _i that a scientist i exerts in finding a licensee, and R. Since ζ ∊ (0, 1) is the share of effort in the final discounted stream of royalties V, we use ζ to capture the extent to which the market is entrepreneurial. Highly entrepreneurial markets allow scientists to privately license their invention with less effort, where the optimal effort can be found as, $e_{i}^{*} = \arg \,\max_{{e_{i} }} \pi_{i}$

We model a scientist’s search cost for finding a licensee as,

$$c_{i} = e_{i} - R^{0.5}$$

(3)

In (3) c _i is a positive function of e _i , correspondingly decreasing with R, albeit in a diminishing fashion. In symmetric terms, for the disclosure strategy Eqs. (1)–(3) become

$$\pi_{\text{TLO}} = \left( {1 - \alpha } \right)\left( {V_{\text{TLO}} - c_{\text{TLO}} } \right),$$

(1′)

$$V_{\text{TLO}} = e_{\text{TLO}}^{\zeta } \left( {\gamma R} \right)^{1 - \zeta } ,$$

(2′)

$$c_{\text{TLO}} = e_{\text{TLO}} - \left( {\gamma R} \right)^{0.5}$$

(3′)

While $e_{\text{TLO}}^{*} = \arg \max_{{e_{\text{TLO}} }} \pi_{\text{TLO}}$, solving the model, we find,

$$e_{i}^{*} = R\zeta^{{\frac{1}{1 - \zeta }}} ,\quad e_{TLO}^{*} = R\zeta^{{\frac{1}{1 - \zeta }}} .$$

(4)

Equation (4) suggests that increasing R, ζ increases the e of both agents. The same holds true for the TCO in terms of γ, i.e. autonomous TCOs exert more effort. We can use (4) in deriving π _i and the payoff from the disclosure strategy, which is α(V _TLO − c _TLO). In turn, we can subtract the one from the other and examine the conditions under which disclosure provides a greater payoff than bypassing. In the parlance of game theory the difference between the two payoffs delineates when one strategy dominates the other. This difference, which is equal to

$$R^{0.5} \left( {\left( {1 - \alpha \gamma } \right)\zeta^{{\frac{1}{1 - \zeta }}} \left( {1 - \zeta^{ - 1} } \right)} \right) - \left( {1 - \alpha \gamma^{0.5} } \right),$$

(5)

can outline the conditions that lead the scientist to disclose her technology to the TCO or not. In simple terms, when (5) is positive disclosure is the dominant strategy. A first look at (5) reveals that the derivatives of Eq. (5) with respect to α and γ are in both cases strictly positive. This positive relationship seems to validate claims (I) and (II), i.e. TCOs that are more autonomous are associated with less bypassing, and higher payments of royalties to scientists are equally associated with less bypassing.

Furthermore, on account of the positive derivatives of (5) with respect to α and γ, via the implicit function theorem we can immediately note that there exists a negative relationship between γ and α. This negative relationship accords with claim (III), which suggests that highly professional TCOs can offer a lower share of royalties to their scientists.

Notwithstanding the above, the derivative of Eq. (5) with respect to R is not straightforward to assess. For this reason, we rearrange (5) as to determine the R for which disclosure is optimal, i.e.

$$R > \left( {\frac{{1 - \alpha \gamma^{0.5} }}{{\left( {1 - \alpha \gamma } \right)\zeta^{{\frac{1}{1 - \zeta }}} \left( {1 - \zeta^{ - 1} } \right)}}} \right)^{2} .$$

(6)

In order to uncover the relationship between the value of the invention and disclosure we plot Eq. (6) in Fig. 1. Without loss of generality, and in order to focus on the area where disclosure is dominant, Fig. 1 is plotted for a limited range of the main exogenous variables i.e. a ∊ (0, 1), γ ∊ (.9, 2), and ζ = .1.

Fig. 1

Depiction of dominant disclosure strategy. This graph denotes the value of the licensing proceeds (R), as given by Eq. 6, which will induce disclosure, in terms of the share of the proceeds that accrue to the faculty scientist (α) and the degree of TCO autonomy (γ). Axis labels: Axis R denotes the discounted stream of future royalties; axis α denotes the share of the licensing proceeds that accrue to the scientist; axis γ denotes the degree of TCO autonomy

How changes in ζ affect Fig. 1: In Fig. 1 disclosure takes place in the area that has been labeled “This is the area in which we see faculty disclosure”. Only in this encircled parameter space Eq. (5) is positive, which is the prerequisite for disclosure being the dominant strategy. Increasing ζ diminishes the area for which disclosure is dominant by shifting it to the right. Hence, as claim (IV) suggests, entrepreneurial environments are associated with greater bypassing.

Two things are immediately apparent from Fig. 1. First, disclosure is possible only if γ > 1. Second, the area under which disclosure is dominant is not really affected by the size of R. Thus, when γ > 1 (depending on the α) inventions of all types of R will be disclosed. The latter point is due to the symmetry that we have (for simplicity) imposed on the model by assuming that both agents face similarly structured costs. In reality, there exist transaction costs that only the scientist has to incur. For example, when bypassing the scientist has to acquire a costly new set of skills (that the TCO already has) e.g. learning how to patent and how to find a licensee.

Even without replicating the math, the assumption of asymmetric transaction costs suggests that (in Fig. 1) the encircled area for which disclosure is dominant must shift to the left. In short, the scientist should be willing to disclose her invention to TCOs that have a γ less than one, and are willing to offer an α that is lower than the one depicted by the encircled part of Fig. 1. As in this region there exists a positive relationship between R and α, the more valuable an invention is the greater the α that is required to induce the scientist to reveal such an invention to the TCO. Therein lays the problem. As α faces an upper bound (because faculty scientists can appropriate only part of the licensing proceeds) for very valuable inventions, there is no α that can induce disclosure. Hence, on par with claim (V), inventions of great value are not revealed to the TCO.