Determinants of the Pace of Global Innovation in Energy Technologies

Renewable energy patents surged in the two thousands. Research and development funding barely moved. Those two facts sit side by side and refuse to agree. The question they generate is the one Luís Bettencourt, Jessika Trancik, and Jasleen Kaur set out to answer: if research money didn't drive the boom, what did? The historical baseline makes the puzzle sharper. After the oil crises of the nineteen seventies, public investment in energy research rose sharply. Then, as oil prices fell in the mid-1980s, it stalled. Political and commercial interest in alternatives faded. Patenting in energy technologies stagnated from roughly 1980 to 2000. Analysts at the time attributed that stagnation directly to disinvestment. The story seemed tidy: fund research, get patents; cut funding, get fewer patents. To test whether that story still held, Bettencourt, Trancik, and Kaur built something that hadn't existed before: a comprehensive global database of energy technology patents covering 1970 to 2009. This included roughly 73,000 published patents and applications pulled from major patent offices — Japan, the United States, Europe, Germany, and the World Intellectual Property Organization. The records were organized by technology: fossil fuels including petroleum, natural gas, and coal; renewables including solar, wind, hydroelectric, geothermal, and biofuels; and nuclear. That database, paired with production data and public research and development funding time series from the International Energy Agency, is what made the reversal visible. And the reversal is striking. Solar patent filings grew at an average annual rate of thirteen percent between 2004 and 2009. Wind grew at nineteen percent. All energy technologies together grew at eleven point nine percent. For context, the World Intellectual Property Organization reported worldwide energy patent growth of four point six percent between 2001 and 2005 — itself comparable to semiconductors at four point nine percent and digital communications at three percent. The recent renewable surge runs well ahead of all of that. Crucially, patent citations, a standard proxy for patent quality and impact, stayed steady or grew slightly over this period. The boom wasn't inflation. Something real was happening. The regional picture adds texture. Japan leads in cumulative patents across most energy technologies. But China is the story of the two thousands: it surpassed the European Patent Office in annual energy patent filings and is growing faster than any other nation in the dataset. China files more coal-technology patents than any other country and ranks a close second to Japan in cumulative wind patents. Meanwhile, the European Patent Office has seen a downturn in fossil-fuel patenting over the last decade, particularly in coal. Nuclear fission, despite receiving sustained high levels of public investment over decades, shows low patenting throughout. That last point alone punctures the simple research and development drives patents story — if public funding were sufficient, nuclear should be thriving. Public research and development funding, as the International Energy Agency tracks it for member countries, rose dramatically in the nineteen seventies and early nineteen eighties, and then remained relatively flat. Private research and development, where survey data exist, stayed low. And yet solar and wind patenting accelerated precisely during the period when those funding levels had plateaued. The old correlation between research and development and patents, which had seemed solid for United States data through 2003, was no longer holding globally in the two thousands. So the team built a model to explain what they were seeing. The logic starts with a simple observation: patents are a measurable output of an unobservable stock of knowledge, and that knowledge stock is built up by two observable inputs — cumulative public research and development investment and cumulative market size, approximated by production or deployment figures. Bettencourt, Trancik, and Kaur write a multiplicative relationship: cumulative patents equal a baseline productivity term multiplied by cumulative public research and development raised to one power and cumulative production raised to another. Verbally, patents are a function of both inputs together, not either one alone. What makes the model powerful is the nonlinearity. Because the two inputs multiply rather than add, simultaneous growth in both research and development and markets produces disproportionately large increases in patenting. If the cumulative stock of public research and development is already large, a given expansion of the market yields more inventive activity than it would have at an earlier stage of development. The reverse is equally true. The authors also build in what they call knowledge persistence — past investments don't disappear from the system; they compound, continuing to amplify the value of new investment in either input. That persistence, combined with the multiplicative interaction, is what can turn modest simultaneous increases in research and market growth into sharp patenting surges. The model's fit to actual data is remarkably tight. For solar photovoltaics, the best-fit elasticities, the exponents on research and development and production, are zero point two two and zero point four one respectively, with an R-squared of approximately zero point nine nine seven. For wind, the elasticities are zero point five one and zero point three zero, with an R-squared of roughly zero point nine eight three. For United States coal, the fit gives elasticities of zero point one eight and zero point four three, again with an R-squared of approximately zero point nine nine seven. Those are not merely acceptable fits; they suggest the model is capturing something structural about how innovation works across very different technologies over four decades. The contrast between technologies is itself informative. Solar and coal both show greater sensitivity to market growth than to public research and development — the market elasticity exceeds the research and development elasticity in both cases. Wind is the exception: its patenting shows greater sensitivity to public research and development than to market size. Wind also has a timing peculiarity — the model predicts wind patents best when production data are taken with a negative lag of about three to four years, which the authors interpret as a signal of advanced planning in large-scale wind installation projects. The same basic mechanism operates across all three technologies, but the weights differ. These findings reframe a debate that energy policy has been having for decades. The standard argument has been that governments must choose: invest in basic research, or use policy to build markets through incentives, mandates, and procurement. Bettencourt, Trancik, and Kaur's evidence says that framing is wrong. The two levers don't substitute for each other; they multiply each other. Because markets for low-carbon technologies have themselves depended heavily on public policy — feed-in tariffs, renewable portfolio standards, public procurement — market growth is not a purely private-sector phenomenon. It is, to a significant degree, a policy output that feeds back into the innovation system. The paper is explicit about the implication: policies are likely needed to fund research and incentivize market growth further until these technologies become cost-competitive. Not one or the other — both, simultaneously, because each makes the other more productive. The analysis also operates at a global scale, which means that national action contributes to a shared international innovation dynamic. Knowledge, once created, doesn't respect borders. The practical takeaway is concrete. The recent surge in solar and wind patenting — thirteen and nineteen percent annual growth during 2004 to 2009 — happened not because research and development budgets suddenly expanded, but because growing markets and a persistent accumulated knowledge base interacted to make innovation productive. That interaction is what the model captures, and it is what governments can deliberately reinforce. Fund the science. Build the market. Not as alternatives, but as complements — because the data, across 73,000 patents and nearly four decades, show that each one is what makes the other work. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.