Economists dating back to Alfred Marshall in the late 19th century have emphasised the importance of geographical distance for the flow of ideas and the diffusion of technologies. Inventors find it hard to capture knowledge and like water it “spills over” to other people. Geographic proximity facilitates this process if face-to-face interaction is important, for example, where knowledge is tacit so that it is not codified in writing or in standard industry practices. Workers and managers meeting at formal and informal meetings and moving between firms will help to transmit new ideas between organisations.
The idea that closeness matters for the diffusion of knowledge lies behind a number of government polices. Government subsidies to research and development (R&D) generally target activities that are geographically located within the country, based on the premise that these activities are the ones that have the highest positive externalities. There are a range of subsidies that seek to attract shiny high-technology firms to locate in specific areas, in order to reap positive benefits not only for the investors, but also for firms and workers near to them. Other polices aim to encourage clusters of high-technology firms, again, because of the belief that the sum of benefits from clusters of firms is greater than the constituent parts.
In contrast to the idea that geographic proximity is important for technology transfer, the notion of the "death of distance" has recently flourished in the popular imagination (e.g. Friedman, 2005, Cairncross, 1997; Coyle, 1997). Messiahs of globalisation claim that information travels around the globe at rapid speed from California to Calcutta or Coventry via through the internet, conferences, telephone and other communication devices. Under this view, geography plays little role.
There are plenty of anecdotes, but what is the hard evidence that technology spillovers continue to increase with geographic proximity? And to what extent has it changed over time, as new technologies that reduce communication and travel costs have become more pervasive?
Existing research: distance matters
Answering these questions turns out to be difficult for a number of reasons, the main two are: how do we trace knowledge flows? And how can we distinguish the importance of geographic proximity for knowledge flows from other factors that may be associated with geography?
Patent citations have become an important source of information on the way that knowledge flows between firms and countries. When an inventor takes out a patent they have to provide citations to the prior technology from which their ideas are drawn. This is a pretty direct measure of knowledge flows. A prominent paper using patent citations in this way is Jaffe, Trajtenberg and Henderson (1993). They show that inventors were far more likely to cite other inventors living in geographic proximity, when compared to inventors in other locations. Several papers have followed this approach, and the consensus emerged that knowledge really is subject to a significant degree of "home bias". In Griffith, Harrison and Van Reenen (2006) we also showed that British firms who locate R&D labs in the US where able to better able to tap into American knowledge than those that did not.
New research: distance starting to die
But much of this research uses older data and maybe things have changed recently with innovations such as the internet. In recent research, we have used patent citations to address whether the rumours of the death of distance have been somewhat exaggerated.. In particular, we look at the speed of knowledge flows between countries (see Griffith, Lee and Van Reenen (2007)). A simple look at the raw data suggests substantial evidence of "home bias" in the way that knowledge is transmitted; in the sense that being geographically close makes knowledge transfer easier. But we also see evidence that distance has become less important over time for the international transmission of ideas.
In the figure above we plot the relative speed of patent citations across countries and over time. For example, in the top left panel we look at all successful applications to the US Patent Office for inventors living in Germany in an "early" period (1975-1989) on the left and then in a "later" period (1990-1999) on the right. Looking first at the early period, the height of each bar indicates how much slower foreign inventors were in being first to cite Germans relative to other German inventors. So American inventors were about 14% slower in citing Germans than the Germans themselves, and the French were about 4% slower.
The fact that the bars are almost all positive suggests the well-known phenomenon of home bias in ideas is alive and well – German are quicker at citing other Germans, British quicker at citing other British, and so on. What is more interesting is how home bias has changed over time – on average the bars in the later period are lower than the bars in the earlier period. This suggests that home bias in ideas has fallen. In the later, post-1990 period, the French are only about 1% slower in citing Germans, and the Americans only about 5% slower in citing Germans inventors than the Germans themselves.
This looks promising, but there are many reasons why the simple patterns in the raw data might be misleading. In particular, knowledge may spread more or less quickly due to many patent characteristics that may be poorly captured by observable characteristics, and may be associated with geography. For example, if high quality patents are cited more quickly than lower quality patents, and if high quality patents are geographically clustered for some other reason, then this could give the impression of "home bias" whereas in fact it is to do with the higher average quality of inventors in one location.
The traditional way to deal with this sort of problem in econometric research is to control for unobserved and correlated “fixed effects” (things which we don’t measure, but are pretty fixed over time). Lee (2007) develops an econometric estimator that allows us to do this basically by taking quasi-differences over time between multiple citations on the same patent.
Our research suggests that controlling for these unobserved characteristics makes an important difference to estimates of the importance of home bias in innovative activity. First, the evidence for home bias is much weaker once we control for fixed effects. The non-fixed effects models (which are standard in the literature) suggest home bias in a majority of cases, whereas our preferred models indicate home bias in only a minority of cases. Second, home bias is much stronger in the "traditional sectors" (such as chemicals and mechanical engineering) than in more "modern" sectors (such as computing), consistent with the idea that information diffuses faster internationally in these sectors. Finally, and most interestingly, we find evidence that home bias has declined over time, being much stronger in the pre-1990 period than the post-1990 period. We interpret this as suggesting that information flows more easily across national boundaries as the cost of international communication and travel has fallen.
Conclusions: don’t bury economic geography just yet
So the bottom line is that we find the first hard evidence that distance has become less important over time. When it comes to the flow of new ideas distance is now dying. But, for many sectors, especially in the more traditional and mature technologies, it is not yet dead.
Cairncross, Francis (1997) The Death of Distance: How the Communications Revolution Will Change Our Lives, Harvard: Harvard Business School Press
Coyle, Diane (1997) Weightless World, Princeton: Princeton University Press
Friedman, Thomas (2005) The World is Flat, New York: Farrar, Strauss and Giroux
Griffith, Rachel, Harrison, Rupert and John Van Reenen (2006) "How special is the special relationship: Using the impact of US R&D spillovers on British firms as a test of technology sourcing" American Economic Review (2006) 96(5) 1859-1875
Griffith, Rachel, Sokbae Lee and John Van Reenen (2007) “Is distance dying at last? Falling home bias in fixed effects models of patent citations” CEPR Discussion Paper 6435
Jaffe, Adam, Manuel Trajtenberg and Rebecca Henderson (1993), "Geographic localization of knowledge spillovers as evidenced by patent citations", Quarterly Journal of Economics 108 (3), pp 577-598
Lee, Sokbae (2007) "Estimating Panel Data Duration Models with Censored Data" Econometric Theory, forthcoming.
Marshall, Alfred (1890) Principles of Economics, London: Macmillan