Daron Acemoglu, Philippe Aghion, and two of their colleagues (MIT and Harvard) posted a fascinating working paper last October that hasn’t gotten the attention it deserves in the policy world. (After I wrote this, I spotted a blog post on it over at the World Bank, which apparently hosted one of its authors for a talk last month.) It’s a fairly technical piece of economics (though the summary is nice and clear), so I’ll try to explain a bit about what it says.
I take three big points away from it: First, under most plausible circumstances, optimal climate policy is a mix of carbon pricing and government innovation subsidies (that is, support for RD&D). Second, it’s probably much more expensive to achieve the same environmental outcomes through carbon pricing alone. Third, carbon leakage (the shifting of dirty industries to unregulated economies) can have bigger negative consequences than most have assumed.
The paper starts by pointing out something that most people probably don’t realize:
“The response of technological change to environmental policy has until very recently been all but ignored by leading economic analyses of environment policy, which have mostly focused on computable general equilibrium models with exogenous technology.”
In English, this means that most economic analyses of optimal climate policy (and of carbon pricing in particular) pretend that the pace of future innovation in clean energy technology has nothing to do with the strength of environmental regulation – it’s exogenous, or imposed on the model. Suppose a carbon price of $5, and the cost of clean energy technologies will steadily fall; impose a price of $500 instead, and that cost will drop at precisely the same rate.
This, of course, is nonsense. It sharply contradicts our intuition (which is borne out by empirical study) that environmental regulation tends to spur environmentally friendly innovation. Serious modeling of environmental policy needs to incorporate regulation-driven innovation – what economists call “endogenous innovation”.
[UPDATE: Massimo Tavoni points out in the comments that’s it’s wrong to say that "almost all" models take technological progress as exogenous. He has a fair point -- there’s a lot of work that’s been done on incorporating regulation-induced change in broader climate models. He also notes that there seems to be more work done in this vein in Europe than in the United States. I’m pretty sure that’s right. All that said, most of the models that U.S. policymakers get their estimates from -- whether it’s the economy-wide models used by the EIA or EPA, the well-known DICE model developed by Nordhaus, or the model employed by Stern in his eponymous Review -- ignore regulation-induced technological change.]
Acemoglu et al attempt to do just that. Their model consists of a simple economy that produces one final good using either dirty or clean inputs. Without policy, dirty inputs (and the machines that use them) start with an edge, and only gain greater advantage; the result is environmental catastrophe. (The one exception is if the dirty inputs are sufficiently scarce as to drive production to cleaner ones; in the real world, there’s almost certainly too much coal and oil for that to happen.)
With policy, though, things change. In typical models, imposing a carbon price has one effect: it shifts some production from dirty to clean inputs because the latter becomes more cost-effective. In the model used in this paper, it has a second effect. Returns to innovation depend on the size of the market for new products; as the carbon price leads to greater deployment of machines that use clean inputs, the market for new innovations on those clean machines expands, leading to greater returns on innovation, and hence to more innovation itself. Machines that use clean inputs get better, ones that use dirty inputs don’t. As the gap closes, lower carbon prices are able to achieve the same deployment results.
That said, with a carbon price alone, innovation in clean energy is still suboptimal. That’s because clean innovators (in the model, and to a good extent in reality) don’t account sufficiently for the possibility that their products will find a much larger market in the future. (They only receive a patent for a fixed period of time.) On the other hand, dirty machines still make up most of the market for a long time, which continues to bias innovation toward polluting inputs. Government subsidies for innovation can, in principle, correct these problems without picking technological winners. That’s why when the paper introduces innovation subsidies as an option, and optimizes policy again, the result is a substantially lower carbon price combined with significant innovation subsidies. (If governments try to achieve the same results through carbon pricing alone, they end up needing to impose a much higher price, distorting the economy and leading to broader welfare losses.)
This may sound a lot like the Breakthrough Institute argument for using a low carbon price to fund clean energy innovation. It’s not. The point of the carbon price in this model is not to fund the innovation subsidies. (The subsidies are financed through lump-sum taxes.) It’s to shift production to clean industries, both to cut emissions in the short term (because cumulative emissions matter), and to expand the near-term market for clean innovations, which in turn incentivizes clean energy innovation. An extremely weak carbon price used only as a revenue-raising measure wouldn’t do that.
The last point that I find really interesting has to do with North-South climate policy differences. The authors expand their model to assume that one part of the world (the North) adopts a mix of carbon pricing and innovation subsidies, while the rest of the world (the South) does neither. (This of course is not quite what happens in reality.) They further assume that technology can diffuse from North to South. In conventional models with trade, this leads to “carbon leakage” as dirty industries shift to the South, which in turn leads to higher near-term emissions. In the long run, though, innovation in clean technology helps the South become less polluting too.
In the model with regulation-driven innovation, though, carbon leakage has a second negative impact: it keeps the market for dirty machines, and hence dirty innovations, large. As a result, major innovation in dirty machines remains substantial, making it much more difficult for clean technologies to close the gap. (If there are lots of industries that use coal in, say, China, there’s good reason for companies to spend lots of money figuring out how to use coal more effectively, which in turn makes it harder for clean energy to catch up.) In some cases, that makes environmental catastrophe inevitable. This suggests that at least some degree of international coordination is necessary for policy to work.
One caveat to all this: The paper’s results depend substantially on the clean and dirty sectors being “substitutes” rather than “complements”, which is to say, they depend on the assumption that if clean production improves, that comes at the expense of dirty production. That seems intuitively true – nuclear, for example, will replace coal rather than enhancing it. It strikes me, though, that many innovations will improve both sectors. (Innovation in computing technology is one obvious place, but I’m sure there are many others.) Government innovation supports, to work in the way that this paper suggests, would probably need to target those technologies that improved clean energy but didn’t improve dirty energy too.
Bottom line: Fascinating paper. Take a look for yourself.