Does Excellence in Academic Research Attract Foreign R&D?

The primary issue to be discussed is the role of academic research strength in attracting research and development (R&D) activities to different countries, and whether this serves as an independent factor in attracting R&D establishment of foreign multinational firms to such countries. A secondary issue to be addressed is whether firms differ in values they attach to academic research and hence the responsiveness to academic strengths in their R&D location decisions. Less focus will be put in the presentation on the methodology and more on the results from the empirical analysis. The question posed here is important because international R&D is on the rise, and it is becoming increasingly important for firms to coordinate foreign R&D with domestic R&D activities. Much research has been conducted on location decisions of multinational firms and their impact, however little attention is given to the role of the strength of academic research in these location decisions.

Surveys of multinational firm managers have shown that the possibility of collaborating with universities and locating in close proximity to academic institutions serve as important factors in their R&D location decision process. An important policy implication in this is that by strengthening the linkages between science and industry, R&D clusters will be strengthened by attracting foreign firms to the clusters that surround universities. Strong clusters in turn can benefit domestic companies.

Literature on international R&D processes has suggested that two main motivations exist for conducting R&D abroad. The first is referred to as home-base exploiting R&D, a traditional, development-based motivation, in which existing technologies from home research bases are developed further and adapted to localities abroad. By doing so, overseas R&D follows overseas manufacturing and sales activities, and enables the firms to get a better market response abroad, or to adapt processes to the local manufacturing conditions. The second motivation is research-based and is referred to as home-base augmenting or "technology sourcing" R&D, where new technology is created abroad by utilizing specific foreign expertise in technology development. The new technology that is developed abroad can then be used in the multinational firm's home country and the world market.

Existing research suggests that the size of overseas manufacturing and sales activities of multinational firms is correlated with subsequent investments in development efforts abroad for home-base exploiting R&D. Large and sophisticated local markets attract R&D as do a country's technological strengths in fields that are relevant to the firms. On the other hand, high wage costs of scientists and engineers can discourage R&D. Countries such as China, where costs are lower, have come to attract more R&D, and an abundance of scientists and engineers facilitates in recruiting researchers for new R&D laboratories. Policy also has a large impact; specifically, intellectual property rights protection policies. Those regions which have better enforcement of patents and copyrights have been shown to attract more foreign R&D. In our research, all of these variables are controlled for, before we examine the impact of academic research strengths.

Literature exists as to the manner in which universities can impact upon a firms' innovation activities. Universities can supply firms with consultants, scientists and engineers while acting as collaboration partners. Informal and formal knowledge transfers also exist among universities and firms, where recent developments in research are shared, which then can be employed in R&D. At the regional level, literature has shown that strong universities attract more private R&D in and around their institutions. Other research also suggests that the collaboration of university and private R&D enables radical innovation. However, not all firms can benefit from collaborations with universities. Firms require a "scientific absorptive capacity" to understand and utilize scientific research results. Often firms need to hire their own scientists with close linkages with university research in order to integrate successfully scientific knowledge into their R&D process. As a result, this may enable firms to conduct a broader scope of research activities, to come up with more radical innovations, and to come up with innovations following scientific developments more quickly.

Prior literature has not been able to use adequate data on the quality of university research. Most studies have limited analysis to the aggregate country and regional levels, and have used public R&D expenditures or the number of Nobel Prize winners as a proxy for scientific strength. The goal of our own study is to improve this situation and to perform micro-level analysis of R&D location decisions. We study location decisions covering 40 countries, and we can use detailed data on publication output in the countries as an adequate measure of academic research strength by technological field of relevance. We also examine heterogeneity between firms in their response to the presence of academic research strength in countries.

Research was conducted on the R&D location decisions for the top five R&D spending firms from broad industries in three regions, the U.S., Europe, and Japan. Industries looked at were medium to high technology industries: chemicals, pharmaceuticals and biotechnology, IT, hardware, electronics and electrical machinery, and non-electrical machinery. These are all industries in which patent applications are imperative. Patent application data was used to establish where firms were engaging in R&D, by examining inventor address locations. One limitation is that patent application data is more representative of the research part of R&D because the application for and granting of patents occurs when novel processes and technologies are being assessed. The study distinguishes between 30 technology fields of patenting, taking into account that host countries differ in their areas of technology expertise. The patents were added up in two four-year periods, and a binary variable identifies the presence of local R&D activities during the period.

The patent data suggest that European and U.S. firms conduct 25% to 40% of their R&D activities abroad. These figures are similar to numbers reported in other studies using survey data. European firms are the most internationalized, but most of this R&D occurs within Europe. U.S. firms conduct relatively more R&D in other "TRIAD" regions. Japanese firms clearly lag behind in the importance of international R&D.

Q: Do the patents from a French firm originating in Germany count as international R&D?

Rene BELDERBOS
Yes. European countries are treated as unique locations, so if one European firm conducts R&D in another European country, this is international R&D in our analysis.

Q: The binary variable of patent application is a good idea, but it is just the outcome of R&D, and not a direct R&D activity. Why do you use the outcome of R&D instead of direct R&D activities?

Rene BELDERBOS
In a way, it can be said that the study looks at successful R&D laboratories abroad, rather than all R&D activities abroad. But I note that we pick up more R&D activities as we are using data on all patent applications, which also include unsuccessful applications not leading to patent grants. Use of patent information is the only way that a broad assessment can be made with published data. Survey data could also be looked at, but this is not available on a comparable basis for U.S., European, and Japanese firms, and usually there are issues related to access to these micro data and the quality of the survey responses on R&D related questions.

Q: Are patent applications that only protect intellectual property rights but that are unrelated to R&D very small in number?

Rene BELDERBOS
In general, the correlation between patents and R&D is very high among large firms, especially in industries where patents are important.

Up to 2002, Asia did not have substantial research activities leading to international patent applications. Domestic technological activities in regions such as Singapore, China, and India have increased rapidly but from a low base. Only a small share of the multinational firms' R&D activities are conducted in Asia.

Looking at scientific publications, the U.S. has the clear lead, with 3 million publications during 1996-2002. Japan places second with almost 1 million papers published, followed by the United Kingdom and Germany. The data shows that different countries have different publication activity concentrations in different industries. China, India, and Singapore have a strong emphasis on electrical engineering, while the U.S. and Europe share a strong concentration of publication activities in the chemical and pharmaceutical industries. This confirms that it is useful to examine strengths in the different relevant fields of academic research in host countries as a driver of foreign R&D decisions.

In terms of variables, the analysis controls for market size and market sophistication, technological strength, the intellectual property right protection level, cost of R&D personnel in terms of the yearly gross income of engineers, language similarities between host and home countries, geographic distance. As there may be a greater tendency for European inventions to be patented at the European Patent Office, the analysis also includes a dummy for Europe-invested patents. The analysis takes into account the degree to which R&D activities within each of the firms is science-oriented. This is measured by looking at the patents applied for prior to the period of the empirical analysis and by examining the references these patents make to the scientific literature. The majority of patent inventors are aware of specific scientific papers cited on patents, and the intensity of science citations can be regarded as a useful measure of "closeness to science" of the firms' R&D activities. Control variables for the firm include technological strength in the field, overall size of R&D activities, the presence of a sales or manufacturing subsidiaries in a host country, the age of the firm, international R&D experience, and country of origin.

The analysis can utilize a large number of observations due to the fact that different countries, technologies and firms have been looked at. This allows for more freedom and the ability to estimate the coefficient with more precision. The main variable, academic research strengths, has a positive and significant effect on the R&D location decision of firms. This impact is there after controlling for all other variables that were expected to affect R&D location decisions. For instance, language similarities facilitate R&D abroad, and geographic distances discourage R&D abroad. Firm's technological strength increase foreign R&D, as well as previous international R&D experience and the presence of sales or manufacturing affiliates in a country.

Q: Isn't it possible that the academic research variable reflects country or market size rather than academic research? Shouldn't you measure it independently of the size of the country?

Rene BELDERBOS
The absolute amount rather than the intensity was included because from a theoretical viewpoint we expect the critical mass of scientific research to attract R&D, rather than intensity. Other variables included are also affected by country size. The estimation strategy here was to control for the relevant size aspects by including market size and countries' patent strengths. This ensures that what is being measured is what is intended to be measured: the effect of critical academic research strengths.

The data shows that firms based in smaller countries with traditionally very highly internationalized multinational firms conduct more R&D abroad compared to U.S. firms. For Japanese firms, there is still a strong negative "home country" effect after many other factors are controlled for. Firms with a stronger science orientation in their R&D activities have a greater impact on academic research in their R&D locations, compared to firms with a limited science orientation. Many firms with a high science orientation are active in the biotechnology and electronic sectors. Firms with a high science orientation have a very low probability to conduct foreign R&D in countries with little academic research strength. Conversely, those highly science-oriented firms show very high probabilities to invest in countries with high academic research strengths.

Spilt sample test confirmed looking at firms with more than median and less than median science orientation, confirmed the results. A significant impact could be seen for the sample including higher science orientation firms, with no significant impact for low science orientation firms. Other sensitivity tests, such as estimation of a count model or including a lagged dependent variable for firms' prior R&D in a country, showed roughly similar results.

The role of firm heterogeneity relating to technology leadership was also examined. Here the idea is that firms that are leaders in technology are more attracted to public R&D than private R&D due to concerns about losing their technological lead to local competitors in the host country. Controlling for technological leadership, however, we see that even firms that are not technological leaders are responsive to academic research strengths. Although having a technological lead plays an important role in location decisions, it is not a necessary condition for academic research to play a role in location decisions. Rather, the organization of firms' R&D that includes scientific development and orientation with strong linkages with universities is the factor that is driving firms' decisions on international R&D location patterns where academic research plays an important role. A final sensitivity check at a more aggregate level showed robust results.

In conclusion, there exists a significant and large impact of academic research strengths of host countries on the R&D location decisions of large, R&D intensive multinational firms. This impact was found robust throughout many specifications. The elasticity of the probability of conducting foreign R&D with regard to academic research exceeds the elasticity for market size and GDP per capita in the second period. Countries with greater research strengths and academic publications attract firms with a greater science orientation in their research activities.

In terms of policy, the results indicate that strengthening basic research at universities can attract strong technology leading firms and firms with a strong science orientation to host countries. In order to make a better assessment, however, more detailed analysis is required in the future. One such analysis would be to look at the regional level, because academic research spillovers are strongest at the local level. Global R&D location analysis needs to be combined with more detailed data on U.S. states, European regions, and Japanese prefectures. In this way, the impact of regional strengths in academic research on the location of R&D can be examined. Also, the type of publications analyzed should be looked at as to whether they are basic or applied. By distinguishing between basic and applied journals, the type of research that is most relevant to attract firms can be determined. Prefectures and regions with strong universities that collaborate with firms may also show a stronger attraction to foreign firms. Once such issues are examined, further implications can be derived as to the most appropriate types of research management for universities in order to create benefits to the local economy, such as attracting foreign firm R&D.

Q: Does private firm R&D have an impact on academic research? Can you include more features of industry science links in the analysis?

Rene BELDERBOS
Indeed there may be a reverse effect of private R&D on university research. Not much research has been done on this area, but one paper studied U.S. firms in a number of states in the biotech industry. Because such firms can become sponsors of academic research, they have a certain amount of power in the direction of university academic research. However, the study found this effect to be rather limited compared to the larger effect of academic research and the strengths of departments attracting foreign and domestic firms. Moreover, in most cases, the linkages are stronger between domestic firms and universities, rather than international firms. So it is not likely that such a reverse linkage would influence the results of the analysis.

The use of publication data to look at co-publications between universities and firms as a measure of industry science links is something that we will start looking at from next month. In Europe, there is an ongoing discourse on the European paradox, where many scientists think that Europe is good at science, but bad at innovation, meaning that the links between science and firms are not well developed. The questions that can be posed are: is this a result of the science being good, but not crossing over to firms in terms of innovative, productive activity, or is it because strengths in the relevant science base do not exist? We hope that our ongoing research allows us to say something on this issue.

Q: Do you see some differences in firms between industries? Could other host country variables be considered, such as market environment and regulation? How does foreign R&D investment affect the firm's actual market performance, as well as the local R&D in the host country?

Rene BELDERBOS
There is limited research on the impact of the R&D of foreign firms on host country economies. This could in principle be studied with patent data. Gathering all of the data on the host country and the productivity of domestic firms in relevant sectors requires a lot of work and this may be why this type of research has not been conducted yet. I would like to conduct further research to see if developing countries benefit from R&D activities, or if this is mostly a developed country phenomenon.

As to the question on R&D and performance, it is clear that firms' home country activities only benefit if they engage in foreign technology sourcing and transfer of knowledge back to the home country. In order for this transfer to be effective, they should have some prior knowledge of the technology they wish to access abroad. A reduction in R&D efficiency may occur with too much geographic dispersion, leading to higher costs. A recent paper about the biotech industry shows that the number of countries in which biotech firms engage in R&D reduces innovation output if it increases too much. By concentrating R&D in established biotech clusters such as Germany, UK and the U.S., performance is increased. What matters then is investing in the correct location.

It is difficult to have systemic data on regulatory frameworks affecting foreign R&D location decisions, and I welcome any ideas on how to do this. For industry differences, an attempt was made to take out the science-oriented biotechnology and pharmaceutical industries to see if the results still held for the rest of the industries. The results were still found to be there, and showed that it is not only a biotech and pharmaceutical phenomenon, but is occurring across all of the high-tech industries that have been examined.

HOSHINO Mitsuhide
It is obvious that your comprehensive analysis has required a lot of work and effort. I am redeveloping the micro-theoretical model of the impact of diversity and culture around worker knowledge, and how this affects productivity. For incremental improvement or innovations in the case of Japan, there exists no requirement for diverse knowledgeable workers, however for frontier-exploiting research; diverse types of people are required in order for research productivity to increase.

A point of worry for me is the fact that Japan has a large, negative dummy variable. How can this negative dummy be explained? In Japan, the percentage of those inventing patents for private firms holding Ph.Ds is very low at 10%, while in the U.S., the percentage of innovators who take out patents having Ph.Ds is very high at 40%. This may partially explain the negative dummy variable, because there is the need to hold a Ph.D-level of scientific knowledge in order to conduct productive research.

Another major difference in researchers between the U.S. and Japan is that Japanese inventors do not move much further than the firm in terms of moving around the country or universities, let alone to international locations. This Japanese immobility and research done by non-Ph.D holders might partly explain the negative dummy variable for Japan. Please provide additional information in regards to this negative dummy variable.

Rene BELDERBOS
Two things must be distinguished. First, Japanese firms conduct less R&D in comparison to firms based in other countries. This indeed has to do with firm organization in terms of the effectiveness of concentrated activities, and the Japanese system of innovation involving intensive communication between marketing, R&D laboratories and manufacturing. This communication and integration leading to quick adjustment of products and processes tends to be more difficult to achieve with global dispersion. So this is the effect of the dummy variable. But your suggestion is that Japanese firms are less science oriented or possess weaker "scientific absorptive capacity", or that they may respond less to overseas academic research strength. To investigate this, we should estimate the model for Japanese firms separately to see whether there are systematic differences in the determinants of R&D location and the extent and role of science orientation. In one of our earlier sensitivity checks where we estimated separate models for U.S., European, and Japanese firms, indeed our results were in line with your suggestion, as we found that academic research was not a significant attractor of R&D in the Japanese firm sample. In contrast, the basic survey results of foreign activities in recent years of Japanese affiliates show a strong trend in basic research abilities abroad, so recent patterns may be changing. This is another interesting area for future research.