Strategic timing of academic commercialism: evidence from technology transfer

Abstract

While the markets for technology have received considerable attention because of the contribution to management and economy, universities and government research institutes have risen as important providers of technology. Their early licensing agreements may contribute to enhancing the licensor’s productivity, the licensee’s competency, and the efficiency of national innovation system. However, later licensing agreements enhance the licensor’s bargaining power. Thus, the timing of licensing is not only a policy consideration at the national level but also a key strategic consideration at the R&D entity level. We first theoretically claim that the ability to “invent around” determines the impact of uncertainty attenuation on the timing of licensing on the condition of market friction. Based on this claim, the paper argues that technology transfers from public research organizations differ from inter-firm licensing transactions in regard to their timing patterns. Using the data of commercialization activity through national R&D programs in South Korea, we empirically find that resolving uncertainties rather delays the licensing time for technology transfers, as opposed to inter-firm transactions. In addition, our findings provide evidence of frictions related to search costs associated with the unique nature of R&D processes in public research organizations.

Appendices

Appendix 1: The case of frictionless market

We present a refined analytic model based on the Gans et al. (2008) framework, to be suitable for public research organizations, and explain how it differs for firms. It describes an analytic model that explains why market friction influences the timing of licensing and impedes market efficiency. Although their model does not specify every element required to understand the nature of licensing contracts, their study not only describes how formal intellectual property rights affect the timing of licensing in the presence of market friction but also empirically demonstrates the effects in the technology transaction records of start-up firms. Like their work, our model neither aims to provide a comprehensive mechanism for the determination of license timing nor seeks to find the optimal timing for a licensing contract. Rather, we briefly show the changing role of patent allowance in determining the timing of technology licensing according to the level of imitability.

In an ideal condition where no market friction exists, a potential licensor (hereafter defined as R) fully recognizes the consumer (hereafter defined as C) who needs technology (i.e., no search costs exist) and both parties wholly assimilate the technology (i.e., information is symmetric). At time 0 (i.e., before the patent allowance notification), the scope of technology is uncertain. With probability p, the scope of technology may be determined to be broad, and the value of the patent may turn out to be v. Otherwise, with probability $1 - p$, the examined scope may be identified to be narrow, limiting the technology value to w (i.e., $v > w$). Accordingly, the expected value of the technology can be expressed as follows: $E(v) \equiv pv + \left( {1 - p} \right)w$. Under the assumptions that R and C are risk neutral and that C pays a flat rate fee to R, C may get $\left( {1 - \delta^{T} } \right)v + \delta^{T} E(v)$ for a fee t ₀ at time 0 where $\delta^{T}$ is the productivity if the technology is transferred to C. If the technology is not licensed until the time of patent allowance (point (C) in Fig. 1), the technology may be negotiated at time T. If the technological value v is confirmed, C may license in the technology at a fee $t_{T} = \frac{v}{2}$ under the premise of cooperative equilibrium.¹¹

What matters is the other case of confirmed technological value, w. In this case, C may “invent around” rather than license in the technology from R under the primary assumption that C owns the required knowledge and resources to realize the innovation value. This assumption might work for inter-firm technology negotiation. However, it may not do so for technology transfer from public research organizations to firms, which generally have access to knowledge and resources different from those of public research organizations, as discussed in Sect. 2.2.

According to the above discussion, we introduce $m$, a dichotomous value that takes 0 when firms are able to invent around and 1 otherwise. If the value of patent scope turns out to be $w$, C may license in the technology at a fee, $t_{T} = m\frac{w}{2}$. That is, if $m = 0$, C can invent around the patent and thus does not license in the technology from R, solely gaining the value $w$. Therefore, if both parties wait, R gets an expected return $\delta^{T} \left( {p\frac{v}{2} + \left( {1 - p} \right)m\frac{w}{2}} \right)$, while C receives $\delta^{T} \left( {p\frac{v}{2} + \left( {1 - p} \right)\frac{1}{2}^{m} w} \right)$. The joint expected returns are higher when the parties do not wait because they are $(1 - \delta^{T} )v + \delta^{T} E(v)$ at time 0 but $\delta^{T} E(v)$ at time T; this discrepancy means that it is more beneficial for both parties to enter into a licensing contract earlier regardless of whether C can invent around and how the patent scope turns out.

Including the price at the time may show the distinct impact of patent scope on R and C. Under a cooperative condition, the parties share the surplus evenly:

R’s surplus from an early contract = C’s surplus from an early contract

$$ t_{0} {-} \delta^{T} \left( {p\frac{v}{2} + \left( {1 - p} \right)\frac{mw}{2}} \right) = \left( {1 - \delta^{T} } \right)v + \delta^{T} E(v) - t_{0} - \delta^{T} \left( {p\frac{v}{2} + \left( {1 - p} \right)\frac{1}{2}^{m} w} \right) $$

$$ = \frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)v + \delta^{T} \left( {1 - p} \right)\left( {1 - \frac{1}{2}^{m} } \right)w $$

(1)

In this setting, because C can earn additional rewards when $m = 0$ and the scope is $w$, we can set C’s expected return to include $\delta^{T} \left( {1 - m} \right)\left( {1 - p} \right)w$, the additional return from inventing around a patent, as follows.

$$ \frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)v + \delta^{T} \left( {1 - p} \right)\left( {1 - \frac{1}{2}^{m} } \right)w + \delta^{T} \left( {1 - m} \right)\left( {1 - p} \right)w $$

$$ = \frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)v + \delta^{T} \left( {1 - p} \right)\left( {2 - \frac{1}{2}^{m} - {\text{m}}} \right)w $$

(2)

In sum, if the technology can be invented around (i.e., $m = 0$), R’s surplus becomes $\frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)v$, as in Gans et al. (2008); otherwise (i.e., $m = 1$), the surplus is equal to $\frac{1}{2}\left( {v - \delta^{T} \left( {1 - p} \right)\left( {v - w} \right)} \right)$. Though varying by $v$, $w$, $\delta^{T}$, and $p$, these surplus values imply that uncertainty does not affect license timing irrespective of whether $m = 0$ or $m = 1$ in a frictionless environment.¹²

Appendix 2: The case of market with frictions

Suppose R needs to spend sunk cost $f$ to find C. The appropriate consumer C may value the invention by $\Delta$ more than others do: for C, the value of the invention is $v + \Delta$ with a broad scope and $w + \Delta$ with a narrow scope. Therefore, the surplus from a search for C at time 0 can increase by:

$$ \frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)\Delta + \delta^{T} \left( {1 - p} \right)\left( {1 - \frac{1}{2}^{m} } \right)\Delta $$

$$ = \left( {\frac{1}{2} + \delta^{T} \left( {1 - p} \right)\left( {\frac{1}{2} - \frac{1}{2}^{m} } \right)} \right)\Delta $$

(3)

If R waits until time T, the expected return may depend on $m$, the inability to invent around. In case the technology has a narrow scope and can be easily invented around, R may not search for C. For inventions granted, R’s return from searching at time T increases by $\frac{1}{2}\delta^{T} \Delta$. Meanwhile, in case the technology has a broad scope and cannot be easily invented around, the return again increases by $\frac{1}{2}\delta^{T} \Delta$, showing that the increase in R’s return at time T is $\frac{1}{2}\delta^{T} \Delta$ in all cases except for a technology with a narrow scope that can be easily invented around.

In this setting, if $\left( {\frac{1}{2} + \delta^{T} \left( {1 - p} \right)\left( {\frac{1}{2} - \frac{1}{2}^{m} } \right)} \right)\Delta <\, f \le \frac{1}{2}\delta^{T} \Delta$, R would be willing to find C at time T. In fact, for $m = 0$, the above inequality turns out to be $\frac{1}{2}\left( {1 - \delta^{T} \left( {1 - p} \right)} \right)\Delta <\, f \le \frac{1}{2}\delta^{T} \Delta$ as shown in Gans et al. (2008), providing the underlying condition of firms willing to make a licensing agreement after patent grant. On the contrary, for $m = 1$ (i.e., the assumed situation between public research organizations and firms), the above inequality turns to be $\frac{1}{2}\delta^{T} \Delta < \,f \le \frac{1}{2}\delta^{T} \Delta$. Since no $f$ value satisfies this condition, R would receive no incentive to license at time T rather than at time 0, even in the presence of search costs. This inference implies that the inability to invent around determines its licensing time: when a consumer can invent around, uncertainty can delay the licensing contract, whereas it cannot do so when a consumer is incapable of inventing around a patent in the presence of market friction.

In sum, the insight obtained from our modified analytic model with a term for market frictions is that the inability to invent around determines its licensing timing. The model shows that when a potential licensee can invent around, uncertainty can delay the licensing contract; this result is consistent with Gans et al. (2008). However, it shows that when the technology is not easy to invent around in the presence of market friction, waiting until the uncertainty of intellectual property disappears by patent allowance would not bring more surplus to the technology provider than earlier licensing does.

A change in search costs due to the extent of scope may reinforce R’s incentive to license at time 0 compared to time T. Suppose the search cost is $f_{v}$ with a broad scope and $f_{w}$ with a narrow scope—this setting works only when $m = 1$. Since a broader scope is associated with more technological opportunities, technology with a broader scope receives more attention from potential consumers and thus attracts more firms in the markets for technology (Gambardella et al. 2007; Merges and Nelson 1990; Palomeras 2007); that is, $f_{v} < f_{w}$. At time 0, the patent has a broad scope as originally intended; thus, at time 0 the search cost would be $f_{v}$. Accordingly, R’s expected returns, including gain or loss from searching, is $\frac{1}{2}\delta^{T} \Delta - \delta^{T} f_{v}$ at time 0 and $\frac{1}{2}\delta^{T} \Delta - \delta^{T} \left( {pf_{v} + \left( {1 - p} \right)f_{w} } \right)$ at time T. Because $f_{v} <\, f_{w}$, $\delta^{T} f_{v} <\, \delta^{T} \left( {pf_{v} + \left( {1 - p} \right)f_{w} } \right)$. In other words, searching for consumers after patent allowance can entail more search costs, increasing the gap in expected returns between searching at time 0 and T. With the introduction of search costs, R’s expected return at time 0 relative to time T increases by the gap between the search costs at time 0 and T, i.e., $\delta^{T} \left( {1 - p)(f_{w} - f_{v} } \right)$.