Summary
Multinational corporations are among the most innovative firms in the world. Although they are defined by their geographically fragmented production processes, the innovation that they carry out is comparatively concentrated, with a large share of them pursuing innovation investment mainly in one ‘headquarters’ country.
A new study analyzes data on the global operations of US-based multinationals to quantify the impact of innovation investments by parent firms on their affiliates located abroad. These data have been collected by the Bureau of Economic Analysis and cover the period from 1989 to 2008.
The results suggest that ‘headquarters innovation’ has a positive impact on the productivity of foreign affiliates. Quantitatively, the research shows that the median firm generates approximately 20% of the overall returns from its US research and development (R&D) investment abroad. This suggests that estimates based only on domestic operations understate the effect of innovation for US firms’ gains.
Furthermore, the productivity of an affiliate firm increases with both its own innovation investment and in the investment of its parent. Parent and affiliate R&D investments are complementary in their impact on affiliates’ productivity – and the returns to headquarters innovation investment is larger in those affiliates that also perform R&D at their own site. This suggests that firms that make innovation a priority are able to enjoy productivity benefits both directly and indirectly within their affiliates.
The results also suggest that ‘parent innovation’ is a key determinant of long-run affiliate productivity: eliminating the effect of parent R&D would imply an average reduction in affiliate productivity of 36%. This implies that the median productivity reduction from eliminating the effect of parent R&D is approximately four times larger among innovating affiliates (compared with those that are not investing in R&D).
Combining these results with the established dominance of US parent innovation within American research gives support to the idea that multinational production is an important determinant of the diffusion of technology and new ideas between countries. Although R&D spending is concentrated in the United States, the positive effects can be transmitted to overseas firms through the multinational network of parents and affiliates.
While further work is needed to determine the impact of aggregate US R&D investment on outcomes abroad, this study suggest that US innovation subsidies could have a positive effect on foreign countries.
Beyond subsidies, a deeper international coordination of innovation policy that helps governments to capture the positive impact that their innovation policies may have on foreign countries may provide further productivity gains.
Main article
How global are the gains from innovation? When firms operate production plants in multiple countries, technological improvements developed in one location may be shared with foreign affiliates for productivity gains. This column presents analysis of this cross-border transfer of innovation, measuring the private returns to R&D investments for US multinationals over the period 1989-2008. The research finds that such investments boost firms’ performance beyond the initial innovation site. The median US multinational realizes approximately 20% of the returns to its US investment from overseas operations, which suggests that estimates based only on domestic operations understate the overall gains from R&D. The study reveals a ‘spatial disconnect’ between the costs and potential gains of policies that encourage firms to innovate.
Multinational corporations are among the most innovative firms – and they account for the majority of the investment in innovation made around the world. Yet compared with the fragmentation of multinationals’ production across countries, investment in innovation by these firms is typically concentrated in one ‘headquarters’ country.
The potential geographical mismatch between where knowledge is created and where it is used implies that the returns to investments in research and development (R&D) made by US multinationals depend on the degree to which these firms operate in foreign countries.
Isolating the US economy from other major markets (such as China or the European Union) could therefore reduce the overall returns to R&D investments made by US multinationals. This could then reduce the level of innovation investment that these firms make at their headquarters. As multinationals account for 91% of the innovation investment by US firms, this could reduce overall R&D investment and have a potentially negative impact on US productivity growth.
Domestic innovation can have positive effects on the productivity of foreign affiliates of US multinationals
In Bilir and Morales (2020), we use data on the global operations of US-based multinationals to quantify the impact of their innovation investments on their affiliate firms located abroad. These data have been collected by the Bureau of Economic Analysis and they cover the period from 1989 to 2008.
The results provide evidence to show that ‘headquarters innovation’ has a positive impact on the productivity of foreign affiliates. Innovation by affiliates is also important in that it substantially increases the productivity impact of headquarters innovation for the innovating affiliate. But affiliate innovation does not affect productivity at other firm sites.
Quantitatively, the results show that the median firm realizes approximately 20% of the returns to its US R&D investment abroad, which suggests that estimates based only on domestic operations understate US firms’ overall gains from investing in R&D.
How does the allocation of innovation vary across different sites within the same firm?
A key feature of our data is that they include information on the allocation of innovation investment across sites within the same firm. Access to this kind of information is rarely provided in standard datasets, and it is essential for our analysis.
The returns to headquarters innovation are larger in affiliates that perform R&D at their own site
We first use these data to describe innovation and production activities within multinationals. Innovation investment is more clearly concentrated at the US headquarters in comparison with production. Affiliate innovation nevertheless accounts for approximately 15% of firm-wide R&D spending.
Our data also reveal a ‘reduced-form’ relationship between affiliate ‘value added’ and parent company innovation, which suggests that the productivity growth of an affiliate is affected by the R&D investment of its parent.
The results also indicate low volumes of intra-firm trade among affiliates within multinationals. This suggests that the output of innovation investments performed at the headquarters of US multinationals is passed onto their affiliates through channels other than the supply of specialized intermediate inputs.
How does innovation at headquarters affect the productivity of foreign affiliates within the same firm?
Our main results are based on estimation of a structural model of innovation and production in multinationals. The key novelty in our framework is explicit consideration of the impact of headquarters innovation on foreign affiliates’ productivity within the same firm.
To estimate our model, we build on work by Aw et al (2011) and Doraszelski and Jaumandreu (2013). Our estimates show that the productivity of an affiliate increases with both its own innovation investment and the investment of its headquarters firm.
Not all foreign affiliates of multinationals should be seen as ‘pure recipients’ of technological improvements
Furthermore, parent and affiliate R&D investments are complementary in their impact on affiliates’ productivity. The returns to headquarters innovation investment are larger in those affiliates that also perform R&D at their own site. This second result is a partial contradiction of the idea that in the case of multinationals, all affiliates should be seen as ‘pure recipients’ rather than producers of technology.
In addition, our results reveal that the returns to US parent R&D is understated significantly when only the impact on the firm’s US operations is taken into account. Specifically, the returns to headquarters R&D investment exceed the parent-level returns by approximately 20% for the median multinational.
How can parent innovation affect the long-run productivity of affiliates?
Our results imply that parent innovation is a key determinant of long-run affiliate productivity: eliminating the effect of parent R&D would imply an average reduction in affiliate productivity of 36%. The results also imply that the median productivity reduction from eliminating the effect of parent R&D is approximately four times larger among innovating affiliates (compared with those that are not investing in R&D).
The productivity of an affiliate increases with its own innovation and with investment by headquarters
The evidence also suggests that countries hosting high levels of US affiliate activity would observe a decline in industry-level productivity if affiliates were suddenly unable to benefit from US parent innovation. This effect may be magnified further by technology spillovers from affiliates to domestic firms (Javorcik, 2004) and by links with domestic input producers (Rodríguez-Clare, 1996).
Combining our estimates with the dominance of US parent innovation within aggregate US R&D lends support to the idea that multinational production is an important determinant of technology diffusion between countries (Keller, 2004).
What about unobserved characteristics of firms?
Despite these implications, a potential concern with our estimates is the possibility that they are affected by omitted variable bias caused by some unobserved characteristics of the multinationals. There may be some hidden characteristics that affect both the R&D investment of a headquarters firm as well as the productivity of all its affiliates (for example, the quality of the firm’s management).
To address this concern, we follow several empirical approaches:
- First, we focus on estimating the impact of lagged measures of R&D investment on current productivity.
- Second, we control for fixed effects that vary by year and by the affiliate’s host country.
- Third, we control for lagged values of affiliates’ productivity, ensuring that our estimates do not reflect a spurious correlation between lagged R&D and current productivity.
- Fourth, we control for permanent unobserved affiliate-specific determinants of productivity.
- Finally, we estimate the parameters of specifications that replace our measure of headquarters R&D investment with their physical capital investment.
While both R&D investments and investments in physical capital by headquarters are likely to respond similarly to unobserved firm-specific characteristics that determine the firm’s propensity to invest, only R&D investments should logically have an impact on affiliates’ productivity. Our estimates indicate that headquarters investment in physical capital has no impact on affiliate productivity, which suggests that it is unlikely that our estimates are affected by an omitted variable bias problem.
What about the accuracy of sales revenue reports for affiliate firms?
A second possible concern with our estimates stems from the idea that multinationals could misreport affiliate-specific measures of sales revenue or innovation investment for tax purposes. For example, firms may intentionally over-report R&D spending by affiliates located in high-tax countries with the aim of under-reporting profits.
US innovation subsidies could have a positive effect on foreign countries
To account for this possibility, we provide estimates that exclude from our analysis any affiliates that are located in tax havens. Furthermore, we replace as the key variable of interest a continuous measure of affiliate R&D investment with a dummy variable that takes value ‘one’ as long as an affiliate has some positive investment in innovation in a given year.
While firms may misreport the exact amount of innovation investment in response to the host country’s tax level, it seems unlikely that affiliates will report positive R&D investment when no actual investments in innovation have taken place. When tax havens are excluded from the analysis, our results for parent innovation remain consistent with our main estimates.
Deeper international coordination of innovation policy may provide productivity gains
Finally, consistent with work on the existence of international technology transfers across plants within multinationals (Branstetter et al, 2006), we find that the payment of royalties and technology license fees is indeed positively correlated with affiliate productivity growth. We do find, however, that parent R&D has an independent effect on affiliate productivity.
Conclusions and policy implications
While further work is needed to determine the precise impact of aggregate US R&D investment on outcomes abroad, our results suggest that US innovation subsidies could have a positive effect on foreign countries.
It is likely that this positive effect is not internalized by policy-makers responsible for determining the level of subsidies in their own constituency. Adjusting the tax treatment of multinationals (so that they pay taxes in the United States for the gains accrued by their foreign affiliates) is one potential remedy.
Alternatively, a deeper international coordination of innovation policy that helps governments to internalize the positive impact that their innovation policies may have on foreign countries through multinational networks may provide further efficiency gains.
