Chapter 6. LESSONS FROM JAPANESE DATA

TIME PATH OF BENEFITS FROM CONSORTIA

Our analytical framework for analysis of U.S. data was developed and pre-tested on Japanese data.¹ Japanese government involvement in publicly supported research consortia dates back to the late 1950’s; examining this data will enable us to observe the long-run impact of consortia on patenting outcomes.

Using a model similar to that presented in Table 4, column 2, we estimated the time path benefits of consortia participation up to 13 years after inception of the consortia. Figures 3 and 4 graph the coefficients from the project duration dummy variables against the time since inception of the project. The regression coefficients represent the percentage increase in patenting in the targeted area associated with that year. Figure 3 shows a linear specification along with three alternative negative binomial specifications. Figure 4 shows one of the negative binomial specifications and the 95% confidence interval associated with each estimated coefficient.

Figure 3. Alternative Specifications of the Time Path of Benefits from Japanese Data

Figure 4. Time Path of Benefits from Japanese Data with Confidence Interval

Results from Japanese data indicate that the effect of consortia on patenting outcomes tends to persist for relatively long periods of time. In fact, patenting in the targeted area seems to increase a bit after the cessation of the consortium, before leveling off again in later years. This may be due to the rules under which subsidies were disbursed to firms in Japan. Any idea conceived as a direct result of the consortia was supposed to be patented in the name of the consortium itself rather than in the name of the participating firms. This created an incentive for firms to delay patenting some of their most useful ideas until after the official end of the consortium. For our purposes, the important point to keep in mind is that the effect of consortia can be quite long lasting. This suggests that our estimates of the impact of ATP-funded consortia, based on only four years of data, may underestimate the total impact of research consortia on patenting outcomes of the firms that were involved.

CONSORTIA CHARACTERISTICS

Using Japanese data, we examined the impact of two consortia characteristics on patenting outcomes: spillover potential and product market proximity. Economic theory predicts a positive association between spillover potential and patenting outcomes, and a negative association between product market proximity and patenting outcomes. Research consortia may intensify competition in the industry, in turn lowering profits. Firms that are direct competitors might conduct less R&D in a consortium than they would individually.

Spillover potential is assessed using the measure of technological proximity defined previously. Product market proximity measures “competitive distance” between each pair of firms in a consortium by dividing the number of product markets in which two firms in a consortium “meet” by the total number of product markets in which each firm is active. Two firms that meet one another in a large number of product markets are presumed to be more proximate to each another than firms with few or no overlapping products.² Based on data from 591 distinct product markets, we constructed an average measure of proximity for firms within a consortium.³

Many of the variables on consortium characteristics do not change over time. Including them in a panel regression creates statistical problems (see Moulton, 1986). For that reason, we collapsed the time series dimension of the data. Consortium outcomes are measured as the cumulative sum of patenting in the targeted classes, taken over a fifteen-year horizon from the official inception of the project (or for as long as the data allowed). This sum was regressed on summed measures of direct and indirect research inputs, pre-consortium technological strength, and time-invariant consortium characteristics.

The first two columns of Table 8 illustrate the impact of technological_proximity and product_market_proximity on patenting outcomes.⁴ Consistent with theoretical predictions, the first variable has a positive impact on consortium outcomes, whereas the second variable has a negative impact, albeit one that is only marginally significant in a statistical sense.

These results have important implications for our previous analysis of U.S. data. First, it confirms the importance and robustness of technological proximity as a predictor of consortium success in a data set with a much longer, more complete time series dimension.

Table 8. Firm-Consortium Level Analysis Using Japanese Data
              Negative binomial models
              Dependent variable: Patenting in the targeted area

Variables	(1)	(2)	(3)
Budget	.068 (.039)	.196 (.045)	—
Pre-project patenting	.799 (.026)	.756 (.030 )	—
Real indirect inputs	.028 (.011)	.020) (.013)	—
Technological proximity	.497 (.250)	.020 (.284)	1.16 (.41)
Product market proximity	-.323 (.298)	-.552 (.341)	-.164 (.524)
Technological goal	—	— (.121)	.254
Cumulative total patents	.103 (.042)	.142 (.051)	—
Year	-.114 (.014)	—	—

Second, the negative effect of product market proximity on consortium outcomes suggests that bringing product market rivals into a consortium is unlikely to produce a successful pattern. Few ATP-funded consortia have this structure, and that is probably a good thing.

The third column adds an additional variable to the regression equation that measures participating firms’ perceptions concerning the degree to which the technological goals of a given consortium are “close to commercialization” versus “pre-commercial” or “basic.” The positive, statistically significant coefficient on this variable suggests that projects focusing on pre-commercial research are likely to yield better outcomes. Thus, Japanese data lend support to ATP’s focus on pre-commercial research.

CAVEATS IN APPLYING JAPANESE LESSONS TO U.S. CONSORTIA

When comparing results from Japanese consortia to the U.S. experience, a number of caveats apply. Research consortia may have a larger impact on Japanese research productivity than they would in the United States because of the very different structure of the labor market in Japan. Most scientists and engineers employed in Japanese corporations are part of the so-called “lifetime employment system,” spending most of their careers with a single company. In contrast, U.S. scientific labor tends to be quite mobile across firms. The movement of scientists and engineers between U.S. companies is an important mechanism by which new technology diffuses in manufacturing industries. Because the same mechanism is less operative in Japan, research consortia may play a particularly important role in enabling new technological innovation to flow across firm boundaries.

There is also greater involvement of the Japanese government in establishing consortia, selecting members, directing research, and ensuring that the results are widely diffused throughout Japanese industry than in the United States. The ATP model of consortium governance is essentially a “bottom-up” model, in which the burden of organization falls upon the participating firms—in particular, the joint venture lead participant. By contrast, Japanese consortia are run according to a more “top-down” approach, in which the sponsoring ministry is generally more involved in directing the activities of the consortium. Our preliminary investigations suggest that the degree of centralization is negatively correlated with good outcomes, which, again, validates ATP’s approach to consortium management.

Return to Contents or go to next section.