VoxEU Column Productivity and Innovation

You can raise productivity through R&D, but geography matters a lot

Why do local policymakers fight so hard to attract research and development labs to their area? This column provides a possible explanation. Using patent data, it finds a strong link between R&D and growth caused by knowledge spillovers between firms.

President Obama recently proposed increasing the generosity of the US research and development (R&D) tax credit system and making it a permanent feature of the US tax code. This was justified by the idea that more R&D would lead to growth, not just worldwide but particularly in the US.

But such a bold statement raises some fundamental questions:

  • Does the location of R&D matter?
  • Will a firm be more productive if it locates in one region rather than another?
  • Do R&D spillovers, the benefit to firms other than the company spending its money on R&D, decline with distance and, if so, how quickly?

The answers are important for a number of reasons, most notably for understanding regional growth. If geographic spillovers are confined to narrow geographic markets, growth rates will diverge, poor regions will get poorer, and rich ones will get richer.

Few doubt that, in the long run, new and better products and processes are stronger determinants of firm growth than growth in demand for existing products. Given the importance of this issue, it is not surprising that economists have long studied the link between a firm's R&D and its productivity. Equally important, if not more so, is the fact that not only does a firm's own research affect its productivity, but there are also significant spillovers from the R&D efforts of other firms. This idea, which is at the heart of modern growth theory, dates at least as far back as the 1960s (e.g. Schmookler 1966) and attempts were made to quantify it in the 1970s (e.g. Terleckyj, 1974). Our research provides an added twist to the hunt for R&D spillovers, focusing on geography.

The framework

We limit attention to studies that are based on a production function framework as laid out in Griliches (1979). In that framework, output is produced by conventional inputs and is augmented by a multiplicative term that represents knowledge capital or productivity. Productivity is, in turn, a function of own knowledge and the spillover pool, which consists of a weighted average of other firms' knowledge. Finally, each firm's knowledge is augmented by R&D expenditures and depleted by depreciation.

To work out who benefits from R&D spillovers we have to figure out which firms are close "neighbours". This boils down to working out some distance "weights" – the bigger the weight the closer you are, and thus the more likely you are to benefit from a neighbour's research. Those weights thus play two roles: they determine both the set of firms that contribute to the spillover pool and the relative importance of the firms within that set. For example, in the geographic context, one might limit attention to firms that are in the same state and weight those firms by their geographic distance from the firm doing the R&D.

There are, however, factors other than geography that affect R&D spillovers. For example, firms that produce similar products might benefit from the efforts of product market rivals (horizontal spillovers) and firms that perform R&D in similar technology classes might benefit from each other’s efforts (technological spillovers). Moreover, the factors or links between R&D and productivity are not independent. For example, firms in the same geographic region might perform R&D in similar technology classes, as is the case for Silicon Valley. This means that if we were to assess only one channel, it is possible that the relationship obtained would be spurious and due entirely to the omission of a common causal factor. One remedy for this problem is to include multiple sets of weights and multiple spillover pools, one for each channel considered.

A flavour of the current evidence

Geographic links have perhaps been studied most in the trade literature. That literature uses aggregate data by country or industry and measures closeness by whether firms are in the same country (Eaton and Kortum 1999), by geographic distance between countries (Eaton and Kortum 1996), or by an exponential function of that distance (Keller 2002). In general, researchers have found that R&D performed at home is significantly more productive than that undertaken abroad.

Domestic markets have also been studied by, e.g., Adams and Jaffe (1996) and Orlando (2004). Each uses firm data and a binary distance measure (for example, A and B are close if A’s headquarters is under 50 miles from B’s headquarters). The results that they obtain are mixed, with one weak and one strong relationship. Both Adams and Jaffe and Orlando assess horizontal and geographic channels simultaneously. Interestingly, they come to different conclusions, with Adams and Jaffe favouring geographic spillovers, and Orlando favouring horizontal spillovers.

Our new evidence on geographic spillovers

In a recent paper (Lychagin et al. 2010), we assess all three channels simultaneously. Although we focus on geographic spillovers, we control for technological and horizontal channels (following Bloom et al. 2010).

Specifically, whereas earlier studies have used distance between headquarters, we postulate that inventors are more apt to be sources of spillovers than top management. We therefore build a distribution of each firm's inventor locations from patent data. From that we construct a measure of match between two firms' geographic locational distributions, which we multiply by a function of the distance between each pair of locations. Finally, we estimate the distance function semi-parametrically, which is much more flexible than has been done in previous studies.

Although for many small firms the locations of headquarters and research labs are highly correlated, many large firms have several labs in different locations. In Figure 1, the triangle is the location of Eaton Corporation’s headquarters, whereas the dark circles indicate the locations of its inventors. It is clear from this that only taking the headquarters into account could severely underestimate the importance of inventors learning from neighbouring inventors.

We find that:

  • Locations of researchers are more important than locations of headquarters but both have explanatory power.
  • The effects of R&D do fall with distance.
  • Geographic R&D markets are very local.
  • Although the effects of technological and geographical proximity are strong in all specifications, horizontal proximity does not matter much for total factor productivity.

Figure 1. Eaton Corporation; An example of the location of corporate headquarters and the distribution of inventors

Notes: The triangle is the headquarters. The centre of a red dot corresponds to a county where Eaton has some iventors; the size of the red dot is proportional to the number of inventors. Grey dots indicate US counties.

Location, location, location

We conclude that location does matter. There is a strong link between R&D and growth through knowledge "spilling over" between firms – this means that research will generally be under-provided by the market. But this process has an important geographic element – having your inventors close to where the R&D is occurring means that you benefit a lot more from the new ideas. This is why local policymakers like to attract R&D facilities into their areas, but it is also why regional convergence, if it occurs, is often so slow.

Our findings are complementary to those of Greenstone et al. (2010) who find that locating a large new plant in a region increases the productivity of other plants in that region. Moreover, our research provides one explanation for the findings of Wilson (2008), who documents that local policymakers invest substantial sums in the form of tax incentives to attract R&D labs to their regions.

References:

Adams, James D, and Adam B Jaffe (1996), “Bounding the Effects of R&D: An Investigation Using Matched Establishment-Firm Data”, The RAND Journal of Economics, 27(4):700–721.

Bloom, Nick, Mark Schankerman, and John Van Reenen (2010), “Technology Spillovers and Product Market rivalry”, LSE/Stanford mimeo, revised version of Centre for Economic Performance Discussion Paper No. 675

Eaton, Jonathan and Samuel Kortum (1996), “Trade in Ideas Patenting and Productivity in OECD”, Journal of International Economics, 40(3-4):251-278.

Eaton, Jonathan and Samuel Kortum (1999), “International Technology Diffusion: Theory and Measurement”, International Economic Review, 40(3):537-570.

Greenstone, Michael, Richard Hornbeck, and Enrico Moretti (2010), “Identifying Agglomeration Spillovers: Evidence from Winners and Losers of Large Plant Openings”, Forthcoming in The Journal of Political Economy.

Griliches, Zvi (1979), “Issues in Assessing the Contribution of Research and Development to Productivity Growth”, The Bell Journal of Economics, 10(1):92¬¬-116.

Keller, Wolfgang (2002), “Geographic Localization of International Technology Diffusion”, American Economic Review, 92(1):120-142.

Lychagin, Sergey, Joris Pinkse, Margaret Slade, and John Van Reenen (2010), “Spillovers in Space: Does Geography Matter?”, CEPR Working Paper 7929.

Orlando, Michael J (2004), “Measuring Spillovers from Industrial R&D: On the Importance of Geographic and Technological Proximity”, The RAND Journal of Economics, 35(4):777-786.

Schmookler, Jacob (1966), Invention and Economic Growth, Harvard University Press.

Terleckyj, Nestor (1974), Effects of R&D on the Productivity Growth of Industries, National Planning Association.

Wilson, Daniel J (2009), “Beggar thy Neighbor? The In-State, Out-of-State and Aggregate Effects of R&D Tax Credits”, The Review of Economics and Statistics, 91(2):431-436.

2,625 Reads