* This post builds on the IFC report “Innovation in Green Technologies” and is powered by the Innovation Strategy Explorer database.

Introduction

Economic development and the green transition need to go hand in hand. Many of the most dramatic development and growth success stories are built around waves of new technology and innovation such as electronics, wireless and ICT technologies in countries such as Japan, Korea, Taiwan and China. Ukraine is poised to be the world leader in advanced - but cheap - defence technology. Of course, China has also led the way in using the Green Technology Wave as a route to development. Indeed, there is concern that China could be so far advanced in solar or battery technology that not even high-income countries are able to catch up — and that this could impose limitations on sovereignty. However, a challenge of the clean transition is that greenhouse gases are connected to so many human activities. What’s more, we are increasingly facing the challenge of having to adapt to climate change. As a consequence we need to push the innovation frontier in many different technology areas. Thus, there is likely a lot of potential for growth success stories to arise. However, not all countries are equally well placed and not all countries can succeed in the same technologies.

In this dashboard we examine which countries show potential based on their existing innovation landscape as revealed via global patent data. Throughout we assign (patented) innovation to countries on the basis of the location of the associated innovators and the location of the owner (assignee) of the underlying patent.1 Our definition of Green includes the full CPC Y02 classification, which includes technologies relevant to the mitigation of GHG emissions as well as innovation areas relevant for adaptation to climate change consequences.

The global green technology innovation landscape

The total amount of EMDE green innovation - by innovation counts - is several orders of magnitude lower2 than for HIC and indeed China, which experienced an explosion of green innovative activity over 2000-2022 (Figure 1a). By and large, the lower numbers reflect generally low levels of innovative activity in EMDE. However, even as a share of total innovation, EMDE show lower levels (Figure 1b). This conclusion changes, however, if we focus on higher-quality and higher-value multi-country innovations (i.e. innovations that seek patent protection in at least two jurisdictions; use the interactive toggle to display them). Here the EMDE share dominated the other country groups until 2008. The Chinese innovation share only exceeded that of other EMDE from 2016. This suggests EMDE are more prominent among higher-value green innovation.

Figure 1: Green technology innovations and green share per year, by income group

Universe: All innovations Multi-country patents

Note. Interactive. Panel (a) plots the number of green-technology innovations per family year (earliest filing year) for the three income groups on a logarithmic axis; panel (b) plots green technology as a share of each group’s total innovation. The faint dashed line in each panel marks that series’ 2000-2022 average (labelled with its value). Innovations are attributed to countries by the union of harmonised inventor + holder locations (patbis2025), one observation per group × family, with the tax-haven rule and the office-of-first- filing fallback for families lacking any person mapping (national offices only; HK/MO folded into CN). The Universe toggle switches between all innovations and multi-country patents.

We explore the relative quality of EMDE green innovations in Figure 2 using several measures derived from direct and indirect citations of patents. We focus on the latest 10 years for which we have data (2013-22). Our first measure is simple citation counts. The value of average spillovers reports a weighted average of the value of all direct and indirect forward patent citations of an innovation; i.e. all innovations that cite a given innovation directly or indirectly.3 Thirdly we report an index of marginal spillover returns - i.e. the value that would arise from further R&D investment in a particular technology in a country per amount invested.4

In each case we find:

  • EMDE green innovations are of higher quality and economic value than non-green EMDE innovations.
  • The quality gap between HIC and EMDE innovation is substantially narrower for green innovation. Indeed, using our preferred measures of Patent Rank and Marginal Spillover Return, we find that green EMDE innovations are of equal or higher quality and economic value than the average HIC innovation.

Figure 2: Green technology innovation — counts, citations, value and returns by income group (2013-2022)

Universe: All innovations Multi-country patents    Metric:

Note. Interactive. The donut (left) always shows the number of green-technology innovations by income group (total 2,265,708 observations for all innovations). The Metric dropdown selects the panel on the right — average global spillover value (default; expected global spillover value, $ million), average forward citations per family, or average global returns from additional R&D (the returns-to-R&D index: expected spillover value scaled by R&D cost and patent survival, over innovations with an estimated positive R&D cost, ×100). The Universe toggle switches between all innovations and multi-country patents. All panels attribute innovations to income groups by inventor and owner location, one observation per family within each group. In each metric panel the green and grey bars are green technology vs all innovations, the dotted horizontal line marks the HIC all-innovations average as a common reference, and error bars are 95% confidence intervals for the mean (mean ± 1.96·SE).

Heterogeneity across countries

The country groupings combine a range of highly diverse countries. Figure 3 explores this heterogeneity with a scatter that compares each country’s green innovation against two benchmarks. The vertical axis plots the country’s green average relative to the global green average; the horizontal axis plots the country’s green average relative to its own average across all technologies. In other words, the vertical axis tells us whether it makes sense to invest in a particular country conditional on wanting to invest in green technology; the horizontal axis tells us whether it makes sense to focus on green technology conditional on focusing on a particular country. The Measure dropdown switches the underlying quantity between the average value of global spillovers per innovation (default) and the average global returns to additional R&D (both as in Figure 2); the Universe toggle switches between all and multi-country patents. The three donuts report the share of green innovation that sits in countries above average on the vertical axis, on the horizontal axis, and on both.

A number of developing countries are above one (i.e. above average) on both dimensions — under the returns measure: India, Turkey, Malaysia, Thailand, Vietnam, Iran, Serbia and Indonesia — and a fairly large number more (all with the spillover measure) are above average in terms of local options (horizontal axis).

Figure 3: Green gap scatter with above-average shares

Universe: All innovations Multi-country patents    Measure:

(a) Green gap scatter — hover a dot for country and values

Note. Interactive. Panel (a) is a scatter over countries with at least 25 green-technology families, each bubble sized by log10(number of green families) — the size legend is labelled with the corresponding actual counts: x = the green gap (the country’s green average ÷ its own all-technology average), y = the country’s green average relative to the global green average (dashed lines mark parity). To keep the plot legible only a subset of countries is labelled, prioritising the largest EMDE ex China countries by green-family count (plus the largest HIC and China); hover any dot to read its country name, green-innovation count and both axis values. Use the Measure dropdown to choose the underlying quantity — average global spillover value (default) or average global returns to R&D — and the Universe toggle for all vs multi-country patents. Bubbles are coloured by country group (HIC, EMDE ex China, China). Panels (b)-(d) are donuts giving the share of green innovations (weighted by green families) that sit in countries above average on the vertical axis (above the global green average), above average on the horizontal axis (above their own all-tech average), and above average on both axes. Panels (e)-(g) name the EMDE ex China countries in each donut’s category: word clouds with font size scaled to log10(green families) (as the scatter bubbles).

Figure 4 shows the location of the inventors and associated firms for innovations with the highest spillover values. To give readers an impression of the exact innovation activities we provide links to the official patent documents in each case. The figures reported above already suggested that the bulk of activity is in HICs and China. However, there are clusters of activity across all continents, although there is a clear under-representation of Africa. That said, there are clusters of high-value green innovative activity in Kenya, Morocco, Egypt and South Africa.

Figure 4: Where the top green innovations happen

Note. For every country, its top green innovations that have a genuine city-level geocode in that country (inventor location preferred, holder location as fallback; innovations whose only coordinates are capital-city fallbacks are excluded). Each popup links the country-verified family document — a publication of the same family on which an inventor or applicant carries this country’s code — so the Espacenet page substantiates the mapping; in the rare case no such document exists, the patent number appears without a link. Innovations are ranked by global spillover value; only those above the whole-sample green average (≈ $4.8 million) qualify, capped at 100 per country. Dot colour encodes the spillover value on a log scale (green → yellow). A family attributed to several countries appears once per country, always at its location in that country. Same-city innovations are spread in a small circular pattern whose size is fixed in screen pixels at city-level zoom — so they remain individually visible as you zoom in and converge on the true location; at wider zooms the spread instead freezes in geographic size, so country- and world-level views are not over-spread; colour encodes the global spillover value ($ million, log scale). Click a dot for the patent number (linked to Espacenet), place and value.

Heterogeneity across technologies

Green technology comprises a wide range of different sub-technologies. What exactly are EMDE innovators doing in this space, and does it differ from what HIC innovators do? We sample 500 of the highest-valued innovations for each group (HIC, EMDE, CN) and divide them into 12 clusters using an unsupervised clustering procedure based on embeddings of the patent abstracts. In Figure 5 this is projected onto a two-dimensional space separately for the country groupings. It is clear that there are differences in focus between the country groupings. Figure 6 shows the distribution across technology clusters for each country grouping. While there is some broad overlap in the frequency of different technologies — e.g. power conversion is a smaller category everywhere and Advanced Materials and Energy Storage is a bigger category throughout — there are some distinct patterns: HICs seem particularly focused on IoT and smart systems. China has a focus on Deep Learning and Image Processing. Other EMDEs diverge with a focus on Catalytic Processes and Materials and Waste Management and Recycling. Figure 6 can also display the distributions for the top 25 innovations across all countries with more than 25 green innovations. This reveals substantial heterogeneity: For instance, Turkey has an unusually strong focus on Advanced Robotics and Automation. The biggest category in India is Advanced Materials for Energy Storage. In Brazil Catalytic Processes and Materials is the biggest category.

Box 1 — What does EMDE green innovation look like on the ground? The various figures provided here include many links to the patent documents of high quality green EMDE innovations. Here we select three examples from three different countries.

India — Council of Scientific and Industrial Research (CSIR) (WO2014188454). A simple, easily scalable process that turns graphene into two value-added materials at once: nanoporous graphene and graphene quantum dots. The resulting nitrogen-doped porous graphene shows excellent electrochemical activity, making it a low-cost candidate material for fuel cells and energy storage. This is the highest-valued verified EMDE example in our samples (global spillover value ≈ $318 million).

Brazil — Braskem S.A. (US2019315948). A bio-based ethylene-vinyl-acetate (EVA) copolymer in which the ethylene is obtained at least in part from renewable carbon sources rather than fossil feedstock. The material matches the flexible, rubber-like hardness range of conventional EVA (Shore A 60–100) while carrying certified bio-based carbon content — a drop-in greener plastic for footwear and other moulded articles (global spillover value ≈ $119 million).

Kazakhstan — Nazarbayev University Research and Innovation System (US2020335285). A method for preparing perovskite light-absorber material for solar cells that decouples nucleation from crystallisation, so each step can be controlled independently. This yields highly uniform nucleation sites and hence more uniform perovskite films — addressing one of the main manufacturing hurdles on the road to efficient, low-cost perovskite photovoltaics (global spillover value ≈ $55 million).

Figure 5: Innovation clusters — top green innovations by global spillover value

Note. Each panel places a group’s top green innovations by global spillover value in the shared embedding/cluster space: (a) HIC and (c) China show their top-500 samples; (b) EMDE ex China pools the developing-country samples and keeps the top 500 by global spillover value. Titles and abstracts are fetched from EPO OPS, embedded (OpenAI text-embedding-3-small), clustered into 12 themes (k-means) and projected with UMAP once, so coordinates and cluster identities are directly comparable across panels; the legend lists each cluster’s GPT-generated theme and its patent count in the pooled space. Hover a point to highlight its cluster within the panel and read its theme and panel count; click to open an Espacenet search of that panel’s cluster patents, ranked by global spillover value (only family documents verified to name a party in the panel’s country group are listed).

Figure 6: Cluster-theme mix of the top green innovations

Groupings: ▾
Countries: ▾

Note. Interactive: the share of each selected series’ green innovations across the 12 shared cluster themes of Figure 5, drawn as tightly packed grouped bars — clusters on the vertical axis (ordered by the “All” share), shares on the horizontal; bar thickness adapts to the number of selected series so the chart stays compact, and hovering a bar shows the exact value. The selection dropdowns collapse to a single line after use. Click any bar to open an Espacenet search of that series’ top-25 innovations by global spillover value (country-verified family documents for country and group series; the “All” series links the pooled top innovations directly). Groupings use the Figure 5 samples — top-500 innovations by global spillover value per group, “All” pooling them. Countries (all 74 countries with more than 25 green innovations) use each country’s top-25 green innovations by global spillover value: those already in the cluster sample keep their cluster, the rest are embedded with the same model and assigned to the nearest cluster centroid. Shares sum to 100% within each series; note the different sample depths (top-500 vs top-25) when comparing groupings with countries. Default selection: the three country groups (HIC, EMDE ex China, China), in the same colours used throughout the report; add or remove groupings and countries via the dropdowns.

Innovating firms and institutions

To understand how to invest in innovation we need to understand who is doing it. Figure 7 reports word clouds of the assignees of the highest-value innovations in the various country groupings. Several results emerge:

  • In HICs the top owners are dominated by large — often multinational — firms.
  • In China, leading universities and state-affiliated institutions feature prominently alongside firms.
  • The list for the developing groups is dominated more by private firms — in some instances firms headquartered in developing countries, but more often than not large multinational firms from HICs.

Figure 7: Innovation owners — top green innovations by global spillover value

(a) HIC
(b) EMDE ex China
(c) China
State Grid Intel Southeast University Univ Southeast Ping An Technology China Electric Power Research Institute Beijing University of Technology Tencent Tencent Technology Univ Beijing Technology Apple Broad Institute Guangdong University of Technology Univ Guangdong Technology Univ Northeastern Hefei University MIT Massachusetts Institute of Technology North China Electric Power University South China University of Technology Univ Hefei Technology Univ South China Tech Harbin Institute of Technology Harvard College Jiangnan University Nanjing University of Posts and Telecommunications Tsinghua University Univ Beijing Univ Jiangnan Univ Nanjing Univ North China Electric Power Univ Tsinghua Beijing Institute of Technology Beijing University of Posts and Telecommunications DJI Technology Dalian Institute of Chemical Physics Hohai University Northeast University Shanghai Jiao Tong University Soochow University
Owner type: ● company   ● university   ● research_institute   ● government   ● other

Note. Harmonized patent owners of the top green innovations by global spillover value, for (a) HIC (top-500 sample), (b) EMDE ex China (the two developing-group samples pooled, top 500 by global spillover value) and (c) China (top-500 sample). The dropdown menus above panels (a) and (b) switch the cloud to an individual HIC or EMDE ex China country, built from that country’s top-25 green-innovation sample (the same samples behind Figure 6), with owner names harmonised by the same procedure as the pooled panels. Name variants are merged to canonical firms/institutions, individuals dropped, words coloured by entity type and sized by the log of the number of patents mapped to the owner (so size differences understate count differences). Each word is a hyperlink — clicking it opens an Espacenet search (pn= publication numbers) of the top patents by global spillover value mapped to that owner within the panel — listing only family documents verified to name a party in the panel’s country group (for an individual-country cloud, in that specific country). The corresponding innovation topics are in the appendix (Figure A1).

Collaboration in green-technology innovation

The prominent presence of HIC MNEs in EMDE green technology innovation might lead to two opposing concerns.

  • Firstly, could this re-direct EMDE innovation capabilities away from EMDE benefits and needs?

  • Secondly, while potentially beneficial for EMDE, could this re-direct R&D resources away from more productive outcomes in HICs?

In this section we explore this by looking at the relative quality of innovations by collaboration type between HIC and EMDE institutions and inventors.

How prevalent are these collaborations? From the perspective of HICs they are tiny. About 1% of HIC green innovations are with EMDE other than China (Figure 8). This contrasts with 9% of innovations that involve collaborations with Chinese inventors and firms and 11% with inventors in other HICs. For Chinese innovations the share of international collaboration of any kind is even smaller, at around 4%. By contrast, for EMDE excluding China collaborations with HICs are a defining feature: 38% of their green innovations involve HIC partners, plus another 7% that also involve China, while 54% are purely national. International collaboration with non-HIC-based inventors and firms remains a rarity (2% or less in every group).

Figure 8: How collaborative is innovation, by group

Note. Share of each group’s green-technology innovations by inventor-team type (National only, and the four international types by partner income mix — with the North–South mix split into HIC × China and HIC × non-China EMDE). A family is attributed to a group if it has an inventor in one of that group’s countries (all inventor countries, classified by income — no minimum-size cutoff), so a mixed family appears in several groups and the same international team type has identical mean returns in each group it belongs to; inventor countries come from the geocoded countrymap (full coverage, as in the LMICinnovation analysis). For the HIC group this separates its developing-world collaboration into the part running through China (red) versus the rest of the developing world (orange). Some types are structurally absent for a group: the HIC group has no non-HIC only teams, and the China / developing groups have no HIC only (or, for China, no HIC × non-China EMDE) teams. (Green here is the iseapp “Green Technology” tech group.)

The importance of HIC-EMDE collaborations for EMDE becomes even more pronounced when looking at the spillover values of innovations (Figure 9a). Collaborations between HIC and EMDE innovators are over 4 times more valuable than entirely domestic EMDE innovations (centre grouping in Figure 9a). Collaborations between Chinese and other EMDE innovators are even more valuable. A similar ranking emerges when looking at spillover values where the recipients of spillovers are other EMDE innovators. This would suggest that such collaborations are not only beneficial to HIC partners or other HIC firms but also to EMDE by increasing the flows of knowledge spillovers in EMDE economies as a whole.

Similarly, we can compare HIC-EMDE collaborations to international collaborations between innovators in different HIC countries, or to domestic HIC innovations. We find that HIC-EMDE collaborations are of similar value to HIC-HIC collaborations and of higher value than purely domestic HIC innovations. In other words, there is no evidence of the second concern raised above. Looking at this in terms of marginal returns (using the dropdown menu) leads to a similar assessment.

Figure 9: Innovation quality by collaboration type for Green Technology innovation

Outcome:

  1. Global spillovers
  1. Spillovers to EMDE ex China only

Note. Interactive: the Outcome dropdown switches between the mean spillover value per innovation (ev, $ million; default) and the mean returns to additional R&D (×100, %), for green-technology innovation by group and inventor-team type. Panel (a) values spillovers to the whole world (ev_global / istrax_global); panel (b) counts only the spillovers received by EMDE ex China countries (ev_emdenocn / istrax_emdenocn), i.e. how much each team type’s innovation benefits the non-China developing world. Error bars are 95% confidence intervals for the mean (mean ± 1.96·SE, lower bound clipped at 0). Same family attribution and team types as Figure 8 (the North–South mix split into HIC × China and HIC × non-China EMDE). Across all groups, internationally-collaborative innovations — especially mixed HIC × developing teams — carry markedly higher returns than national-only innovations (the wider intervals for the smaller team types reflect their sample sizes).

Conclusion

Is it a good idea for EMDE to engage in green technology innovation? In terms of volume, EMDE (excluding China) today account for less than 1% of global green innovation. The quality picture, however, is considerably brighter: green innovations are cited above the all-technology average in every country group. The gap is particularly stark for EMDE innovations, so much so that the average quality of green EMDE innovations is similar to that of HIC innovations. Estimates of returns from additional investments also suggest that further R&D investment in Green Tech in EMDE would be economically beneficial. There is substantial heterogeneity across countries however. We suggest that the strongest case can be made in countries where green innovations are of above-average value (relative to global benchmarks) and where, within a country, green innovations are of higher value than other technologies. By that account, perhaps the strongest case can be made for further investments in green technologies in India, Indonesia, Serbia, Lebanon, Panama, Samoa, Bosnia & Herzegovina, Thailand, Armenia, Vietnam, Iran and Malaysia, where we find above-average performance on both criteria and across multiple variations of our outcome measures. However, green technology is a broad field with a wide variety of underlying capabilities. We document this variation in interactive figures that allow zooming into particular countries and specific innovations. We also find that the most valuable green innovations in EMDE emerge in collaboration with HIC (as well as Chinese) partners — often well-known MNE firms and sometimes universities and other research institutions. We also document that this does imply diverting EMDE innovation activities towards areas that might be less valuable for EMDE themselves (using within-EMDE knowledge spillovers as benchmark). This suggests that a strategy for scaling up EMDE R&D is to encourage more HIC firms and institutions to seek such partnerships.

Appendix: Innovation topics

The figures below report the harmonised innovation topics (the GPT-consolidated plain-language themes) for the same samples whose owners appear in the main text (Figure 7). Topics are coloured by group and sized by frequency; each word links to an Espacenet search of the top-20 patents by global spillover value on that topic.

Figure A1: Innovation topics — top green innovations by global spillover value

(a) HIC
(b) EMDE ex China
(c) China
image processing catalyst health communication battery detection energy materials power system electric vehicle energy management microgrid monitoring machine learning scheduling data environment solar energy material neural network power systems robotics 3d modeling drone tracking automation data management management mapping network management power management sensing technology ai antibodies batteries biotechnology catalysis charging cloud computing

Note. Topic counterpart to the owners in Figure 7 — the same three samples (HIC and China top-500 by global spillover value; EMDE ex China pooled top-500), with the GPT-harmonised plain-language topic per innovation, sized by frequency and coloured by group. As in Figure 7, the dropdown menus above panels (a) and (b) switch the cloud to an individual HIC or EMDE ex China country, built from that country’s top-25 green-innovation sample with topic labels kept consistent with the pooled vocabulary. Each word links to an Espacenet search of the top patents by global spillover value on that topic (country-verified family documents only; for an individual-country cloud, verified for that specific country).

Appendix: How the value, cost and spillover-return indicators are built

The value and return indicators used throughout this note — the expected global spillover value, the expected spillover value to EMDE (excl. China), and the returns to additional R&D — are produced by the Patent Rank methodology of Guillard et al. (2021), applied to the 2025 PATSTAT patent universe. The same construction underlies the green-technology analysis of Martin & Verhoeven (2023) and the IFC (2025) green-technology report (the methodology corresponds to Box 3.1 and Appendix 2 of that report; see also CEP Special Paper 54). The patent-level private value is from Kogan et al. (2017). This appendix sketches the various steps:

  1. Inferring the private values of innovation from event studies and interpolating them to non-stock listed firms
  2. Computing a centrality measure with an economic interpretation by combining the PageRank approach with individual private values obtained in step 1, which we call Patent Rank
  3. We use the distribution of private value to identify the parameters of a simple model of the innovation process, with parameters \(K_c\) and \(\alpha_c\) capturing the cost of innovation steps and the skewness of the idea-generation function in technology area c
  4. We combine 2 and 3 to compute estimates of the marginal return of further R&D in different technology and country cells.

Private patent value

Every patent \(i\) carries a private value \(v_i\) (USD millions) following Kogan et al. (2017): for patents granted to publicly listed firms, \(v_i\) is read off the abnormal stock-market return of the patentee in a narrow window around the grant date — the market’s estimate of the present value of the cash flows attributable to that single patent, net of the firm-level common factor. For patents whose owners are not listed, \(v_i\) is imputed from observable features (4-digit CPC class, claim count, patent-family size, …) estimated on the listed-firm subset, giving a \(v_i\) for every patent in the universe.

Patent Rank

Raw forward-citation counts treat a citation from a low-value patent the same as one from a high-value patent and ignore chains of citations across generations. Patent Rank corrects both by adapting the PageRank algorithm to the patent citation network. The rank \(PR_i\) solves the linear system

\[ PR_i \;=\; v_i \;+\; d \sum_{j \in F(i)} w_{ji}\,PR_j , \]

where \(F(i)\) is the set of patents that directly cite \(i\); \(w_{ji} = 1/n_j\), with \(n_j\) the number of backward citations of the citing patent \(j\) (each citer spreads its rank uniformly over what it builds on); and \(d \in (0,1)\) is a damping factor, set to \(d = 0.5\). Solved iteratively across the millions of patents in the network, \(PR_i\) equals the inventor’s own private value \(v_i\) plus a discounted share of the private values of every patent that builds on \(i\) directly or indirectly — i.e. the total social value that patent \(i\) ultimately seeds along the subsequent innovation chain.

Spillover flows by group - To isolate the social value that reaches a given inventor-country group \(G\), the same system is re-solved with \(v_j = 0\) for every patent \(j\) whose inventors are outside \(G\). The group-restricted rank \(PR_i^{G}\) then measures only the spillovers from \(i\) that land on group-\(G\) inventors. Averaging per innovation gives this note’s value indicators: the expected global spillover value keeps every downstream patent; the expected spillover value to EMDE (excl. China) keeps only the EMDE-ex-China downstream patents (HIC and other group-to-group flows are constructed the same way).

A simple model of the innovation process

Patent values vary not only because some ideas are better, but because technology fields differ in how expensive a patentable innovation is and how skewed the quality of ideas is. A simple structural model fitted at the 4-digit CPC level separates the two.

  • Model. In field \(c\), ideas are drawn from a skewed (Pareto) quality distribution with shape parameter \(\alpha_c\) — a larger \(\alpha_c\) means more mass near zero (fewer “superstar” ideas, a thinner right tail). Turning an idea into a patentable innovation requires a sunk R&D cost \(K_c\), paid before the realised value is known.
  • The kink. If values were known in advance, only ideas worth more than \(K_c\) would ever be developed, and the value distribution would be sharply truncated at \(K_c\). Because outcomes are uncertain, some ex-post low-value innovations get developed anyway. The combination of skewed idea generation and a sunk-cost cutoff produces a characteristic kink in the empirical patent-value distribution.
  • Identification. Fitting the model to the \(v_i\) distribution within each field recovers both parameters from two distinct features of the data: the location of the kink identifies \(K_c\) (the further right the kink, the higher the cost), and the curvature of the distribution to the right of the kink identifies \(\alpha_c\). Estimates differ sharply across fields — e.g. organic fine chemistry has \(K_c \approx \$25\)M per innovation with a relatively thick right tail, whereas semiconductors have \(K_c \approx \$4.4\)M with very few innovations valued above $60M.

From private value to social returns and spillover flows

The fitted model and the resulting parameters, combined with the Patent-Rank construction, yield a marginal-social-return index. At the field level it is \(\mathrm{ISTRAX}_c = (\alpha_c / K_c)\,\overline{SR}_c\), where \(\overline{SR}_c\) is the average social component of Patent Rank (\(PR_i - v_i\), the spillover left once the inventor’s own appropriation is netted out) across innovations in field \(c\). A smaller cost \(K_c\) (each marginal R&D dollar buys more projects) and a larger \(\alpha_c\) (more idea mass ready to develop at the margin) both raise the return. In this note we use the patent-level counterpart, averaged within each group,

\[ \text{return}_i \;=\; \frac{\alpha_c + 1}{K_c}\,\cdot\, \mathrm{ev}_i \,\cdot\, \mathbf{1}\!\left[\,v_i \le 2 K_c\,\right] , \]

where \(\mathrm{ev}_i\) is the patent’s expected spillover value (its group-restricted Patent-Rank social component) and the indicator restricts attention to viable projects (private value within twice the development cost). Using the global spillover value for \(\mathrm{ev}_i\) gives the global returns to R&D; using the EMDE-ex-China spillover value gives the returns within EMDE (excl. China). The gap between the two is precisely the cross-border spillover to developing economies that motivates this note’s focus on EMDE-relevant green-technology innovation.

References

Andres, P., Cavaglia, C., Fankhauser, S., Ilin, I., Lafond, F., Marotta, F., Martin, R., McNally, S., Nguyen-Tien, V., Owen, A., Ravigne, E., Read, M., Ren, X., Sato, M., Saussay, A., Shah, A., Tran, T. T., Valero, A., Ventura, G., Verhoeven, D., Vona, F., & de Rochambeau, G. (2026). Pathways to a Productive and Inclusive Net Zero. CEP Special Paper No. 54. Centre for Economic Performance, London School of Economics. https://cep.lse.ac.uk/pubs/download/special/cepsp54.pdf

Guillard, C., Martin, R., Mohnen, P., Thomas, C., & Verhoeven, D. (2021). Efficient Industrial Policy for Innovation: Standing on the Shoulders of Hidden Giants. CEP Discussion Paper No. 1813. Centre for Economic Performance, London School of Economics. https://cep.lse.ac.uk/pubs/download/dp1813.pdf

International Finance Corporation (2025). Innovation in Green Technologies. Insights Report. International Finance Corporation, World Bank Group, Washington, DC. https://www.ifc.org/en/insights-reports/2025/innovation-in-green-technologies

Kogan, L., Papanikolaou, D., Seru, A., & Stoffman, N. (2017). Technological Innovation, Resource Allocation, and Growth. The Quarterly Journal of Economics, 132(2), 665–712. https://doi.org/10.1093/qje/qjw040

Martin, R., & Verhoeven, D. (2023). Knowledge Spillovers from Clean Innovation: A Tradeoff between Growth and Climate? CEP Discussion Paper No. 1933. Centre for Economic Performance, London School of Economics. https://cep.lse.ac.uk/pubs/download/dp1933.pdf

  1. One exception: attributions to well-known offshore locations (the Cayman Islands, British Virgin Islands, Bermuda, the Bahamas, Panama and Barbados) are dropped when they arise only from the holder’s registered location and no inventor is based there.↩︎

  2. Throughout we exclude China from the EMDE group.↩︎

  3. This builds on Guillard et al (2021). Further details are in an appendix below.↩︎

  4. This also builds on Guillard et al (2021) with further details in the appendix.↩︎