Difficulty: Intermediate | Prerequisites: Parts 1 and 2 of these study notes.
Part 3 covers the geographic dimension of technology diffusion, the role of human capital and R&D as determinants of successful adoption, the quantitative split between domestic and foreign sources of productivity growth, and the policy implications drawn from the literature.
Technology is geographically localised: diffusion within countries is far stronger than across countries, and it decays sharply with distance (a 50% drop every 1,200 km). However, localisation has weakened over time. Human capital and R&D serve as absorptive capacity, enabling countries to benefit from foreign technology. For most countries, foreign technology accounts for 90%+ of productivity growth, but benefiting from it requires deliberate domestic effort. Policy should favour openness, but simply being open is not enough.
Geographic localisation of technology diffusion
The empirical finding that technology spillovers are stronger between geographically proximate agents and weaken with distance. Keller (2002a) estimates a 50% drop in diffusion for every additional 1,200 kilometres.
In simple terms, the further away you are from a source of innovation, the less you benefit from it.
Distance-decay parameter (delta)
In Keller's exponential decay specification, delta measures how rapidly the weight given to foreign R&D declines with geographic distance. delta > 0 means localisation; delta = 0 means geography does not matter.
Absorptive capacity
The skills and investments a country or firm needs in order to identify, assimilate, and use external knowledge. Two forms: human capital (Nelson and Phelps 1966) and R&D spending (Cohen and Levinthal 1989). Without absorptive capacity, foreign technology is available but unusable.
Think of it as the prerequisite knowledge you need before you can learn from someone else's research.
Convergence versus divergence
Global technology spillovers favour income convergence (poorer countries catch up). Local spillovers favour divergence (rich countries benefit more, pulling further ahead). Whether the world converges depends on how geographically broad and how accessible technology diffusion is.
Gerschenkron's advantage of backwardness
The idea (Gerschenkron 1962) that latecomers to industrialisation can benefit from adopting technologies already developed by frontier countries, leapfrogging some stages of development. The evidence on whether this advantage is realised in practice is mixed.
Domestic versus foreign sources of productivity growth
The share of a country's TFP growth attributable to its own R&D versus foreign R&D. For the United States, roughly 60% of productivity growth is domestic in origin. For small OECD countries, the domestic share may be as low as 10%.
The evidence generally supports the hypothesis that technology diffusion is stronger within countries than across them:
Jaffe, Trajtenberg, and Henderson (1993) find that U.S. patents are significantly more often cited by other U.S. patents than by foreign patents, even controlling for other characteristics.
Branstetter (2001b) confirms this using R&D and patenting data on U.S. and Japanese firms: within-country spillovers are much stronger than between-country spillovers.
Eaton and Kortum (1999) estimate that the rate of domestic technology diffusion for the G-5 countries is about 200 times the average rate of international diffusion.
An exception: Irwin and Klenow (1994) do not find stronger within-country spillovers in the semiconductor industry, but this may reflect the particular characteristics of that industry (a small number of firms concentrated in the US and Japan, limiting the scope for identification).
Keller (2002a) goes beyond the within/across distinction by estimating spillovers conditional on continuous geographic distance. Using industry-level productivity in nine OECD countries and R&D in the G-5, he fits an exponential decay function. The key finding: with every additional 1,200 kilometres of distance, technology diffusion drops by 50%. For a geographically remote country like Australia, this implies extremely limited benefit from technology created in the G-5.
Bottazzi and Peri (2003) find a similarly strong geographic decay in technology diffusion between European regions.
Keller (2002a) estimates separate decay parameters for the late 1970s and the early 1990s. The decay parameter has shrunk substantially over time, suggesting that the degree of geographic localisation has weakened. This is consistent with falling transport costs, improvements in information and communication technology, and increased multinational activity.
A concern is that the estimated geography effect may be spurious, driven by unobserved heterogeneity across locations. While the proposed solutions are imperfect, the overall evidence suggests geography is a genuine determinant. The deeper question is what geography proxies for in economic terms: trade costs, the cost of moving people for face-to-face interaction, or something else. Keller (2001) is a first attempt to decompose the geography effect into imports, FDI patterns, and person-to-person communication, but more work is needed.
Evidence from several studies confirms that human capital helps countries absorb foreign technology:
Eaton and Kortum (1996): inward technology diffusion increases with a country's human capital level.
Caselli and Coleman (2001): computer imports (a proxy for inward technology diffusion) are positively correlated with human capital.
Xu (2000): rich countries benefit from hosting U.S. MNE subsidiaries while poorer countries do not, and this appears related to a threshold level of human capital in the host country.
Cohen and Levinthal (1989) argued that R&D is not just about creating new knowledge; it also builds the capacity to understand and use knowledge created elsewhere. Griffith, Redding, and Van Reenen (2000) find that among 12 OECD countries, industries with higher R&D expenditures see faster productivity growth conditional on a given gap to the frontier. This is consistent with R&D playing an absorptive-capacity role.
Whether technology diffusion leads to convergence or divergence depends on the distribution of absorptive capacity:
If today's richer countries are better at adopting foreign technology (because they have more human capital and R&D), greater diffusion could lead to divergence.
On the other hand, several FDI studies find that less productive firms benefit most from foreign technology (Haskel et al. 2001, Griffith et al. 2003, Keller and Yeaple 2003), which favours convergence.
The evidence is mixed, and more work is needed to resolve this question.
For most countries, foreign sources of technology account for 90% or more of domestic productivity growth. Only a handful of the largest R&D-performing countries derive a substantial share from their own research.
Coe and Helpman (1995) estimate:
G-7 countries: domestic R&D elasticity of about 23%, foreign R&D elasticity of about 6%. The foreign share of the total effect is roughly 21%.
Smaller OECD countries: domestic R&D elasticity of about 8%, foreign R&D elasticity of about 12%. The ratio of foreign to domestic effect is about 3:2.
Eaton and Kortum (1999) estimate that the domestic share of productivity growth is 60% for the United States, 35% for Japan, and between 11% and 16% for France, Germany, and the United Kingdom. Keller (2002a) estimates a domestic share of about 10% for nine countries smaller than the UK.
Smaller countries are more open, which means the foreign R&D effect (which is proportional to the import share in the Coe-Helpman specification) is larger relative to the domestic effect. Controlling for size using GDP, the United States has a domestic-technology-to-GDP ratio of about 1.24, while the Netherlands has a ratio of about 2.18. This suggests that relative to GDP, the United States relies less on domestic technology sources than the Netherlands does, which may reflect a distance-to-the-frontier dynamic: even poor countries that receive almost all their technology from abroad depend critically on domestic efforts to make use of it.
Foreign sources of technology are of dominant importance (90%+) for most countries' productivity growth.
International technology diffusion is neither inevitable nor automatic. Domestic absorptive capacity is necessary.
Imports are associated with technology spillovers, though the precise quantitative importance is debated.
Learning-by-exporting effects are weak in the econometric evidence.
FDI spillovers exist but are concentrated in high-technology sectors.
Technology is geographically localised, though localisation has weakened over time.
How strong is embodied versus disembodied technology diffusion through trade?
How large are vertical compared to horizontal FDI spillovers?
Do FDI spillovers in richer and poorer countries differ, and are they large enough to justify government subsidies to attract MNEs?
What exactly does the geography effect represent in economic terms?
Does greater international technology diffusion lead to convergence or divergence?
The evidence is not yet strong enough to support specific policy measures (such as particular subsidies for MNE location). The broader lesson is that the international dimension of technological change is of key importance, and a closed international economic regime must have detrimental welfare consequences.
The performance advantage of outward-oriented economies over inward-oriented economies is likely to grow as international technology diffusion increases with economic integration. However, the success of countries like Hong Kong and South Korea came not from simply producing high-tech goods, but from a deliberate commitment to matching international performance standards through ongoing interaction with foreign firms and customers. Local efforts are clearly necessary for successful technology adoption. Goods characteristics are only part of what matters; what counted was the quality and standards achieved, not the sophistication of the product per se.
"Geography doesn't matter because we have the internet." The internet reduces the cost of transmitting codified knowledge, but a large share of technology is tacit and requires face-to-face interaction. The empirical evidence shows a strong distance-decay effect even in recent data, though it has weakened over time.
"Poorer countries automatically benefit more from technology diffusion (the advantage of backwardness)." Gerschenkron's idea is appealing but the evidence is mixed. If absorptive capacity (human capital, R&D) is concentrated in richer countries, greater diffusion could actually widen income gaps. Benefiting from foreign technology requires deliberate domestic investment.
"A country that imports 90% of its technology from abroad should focus on openness rather than domestic R&D." Keller's evidence suggests the opposite: even countries that receive almost all their technology from abroad depend critically on domestic technology investments to absorb and use that foreign knowledge. Openness and domestic effort are complements, not substitutes.
⚠️ The 1,200 km / 50% distance-decay estimate is a memorable and frequently cited figure. Know the specification (exponential decay of R&D weights in distance) and what delta > 0 implies.
⚠️ The declining localisation over time is an important finding. Be prepared to explain what might be driving it (transport costs, ICT, multinational activity) and what it implies for convergence.
⚠️ The 90% figure (foreign sources account for 90%+ of most countries' productivity growth) is central to the paper's argument for why international technology diffusion matters. Know the contrast between large and small countries.
⚠️ Absorptive capacity is a concept that bridges this literature with human capital theory and the economics of R&D. Know both forms (human capital per Nelson-Phelps, R&D per Cohen-Levinthal) and be able to explain why they matter for whether diffusion leads to convergence or divergence.
⚠️ The paper's overall policy message, that openness and domestic effort are complements rather than substitutes, is a strong candidate for essay questions.
Eaton and Kortum (1999) estimate that domestic technology diffusion in the G-5 is approximately _____ times the average rate of international diffusion.
True or false: Keller (2002a) finds that the degree of geographic localisation of technology diffusion has increased over time. _____
The domestic share of U.S. productivity growth is estimated at roughly _____%, compared to roughly _____% for countries smaller than the United Kingdom.
True or false: Cohen and Levinthal (1989) argue that R&D serves only to create new knowledge, not to absorb existing knowledge. _____
Fill in the blank: Keller's evidence suggests that openness and domestic technology effort are _____ (substitutes/complements).
Answers: 1. 200. 2. False (it has decreased). 3. 60%, 10%. 4. False (R&D also builds absorptive capacity). 5. Complements.
Q: Explain the Keller (2002a) distance-decay specification and its main finding.
A: Keller regresses industry-level TFP in nine OECD countries on a composite R&D variable that weights G-5 countries' R&D by an exponential decay function of geographic distance: the weight on foreign R&D falls as e^(-delta * D), where D is bilateral distance and delta is estimated. He finds delta > 0 and substantial: every additional 1,200 km of distance reduces the weight of foreign R&D by 50%. This means technology diffusion is strongly geographically localised. He also estimates delta separately for the late 1970s and early 1990s and finds it has shrunk, indicating localisation has weakened over time.
Q: Why might greater international technology diffusion lead to divergence rather than convergence? Under what conditions would it favour convergence?
A: Divergence can occur if the ability to absorb foreign technology (absorptive capacity) is concentrated in richer countries. If richer countries have more human capital and R&D, they will benefit disproportionately from the global pool of technology, pulling further ahead. Convergence is favoured if poorer countries or less productive firms benefit more from exposure to foreign technology, which some FDI studies suggest (less productive domestic firms gain more from MNE presence). The outcome depends on the distribution of absorptive capacity and on whether the benefits of backwardness (a larger gap to the frontier) outweigh the disadvantages.
Q: Keller argues that for most countries, foreign technology accounts for 90% or more of productivity growth. What are the implications for policy?
A: The dominance of foreign technology sources means the international dimension of technological change is critical. A closed international economic regime cuts a country off from the main driver of its productivity growth, with detrimental welfare consequences. However, openness alone is not sufficient: countries also need domestic absorptive capacity (human capital and R&D) to benefit from foreign technology. Policy should therefore promote openness and domestic investment in skills and research as complements.
Q: Contrast the evidence on within-country versus across-country technology spillovers. What might explain the difference?
A: The evidence generally finds within-country spillovers to be much stronger (Jaffe et al. 1993, Branstetter 2001b, Eaton and Kortum 1999 estimate domestic diffusion is 200 times the international rate). The most likely explanation is the tacit component of technology: knowledge that cannot be fully codified is best transferred through face-to-face interaction, which is costly over long distances. Trade costs, language barriers, and institutional differences between countries may also contribute.
The geographic localisation results connect to economic geography (the gravity model of trade, agglomeration economics) and to the literature on the economics of cities and innovation clusters. The absorptive capacity material links to human capital theory and the economics of education, as well as to the R&D and innovation literature. The convergence/divergence question connects to the broader growth theory debate (Solow convergence, conditional convergence, club convergence). The policy discussion connects to international trade policy (openness vs protectionism), development economics (import substitution vs export orientation), and the political economy of FDI subsidies.
geographic localisation, distance decay, spatial decay, technology spillovers geography, within-country spillovers, Jaffe Trajtenberg Henderson, Keller 2002a, Bottazzi Peri, absorptive capacity, human capital technology adoption, Cohen Levinthal, Nelson Phelps, convergence divergence, Gerschenkron backwardness, domestic versus foreign R&D, G-7 productivity growth, OECD R&D, country size technology, openness and growth, outward-oriented economies, import substitution, export orientation, FDI subsidies policy, tacit knowledge geography, 1200 km distance decay, Eaton Kortum 1999