Learning by Doing in Solar, Even if Proven, Would Not Imply Solar Was Subsidized Too Little, Too Late

The estimable Francis Menton has noted repeatedly that the “energy transition” has set loose upon the world a host of innumerates who assure us that they know best when in fact they know less than nothing. And perhaps to coin a phrase, I would add that they are ineconimate, i.e., know nothing about economics. Arrogance and ignorance is a lethal combination. Such people will make us poor, and likely shivering in the dark.

Alas, the mind eating virus has even infected many who were once sensible, or at least periodically sensate or sentient. Such as the FT’s Tim Harford. (I am guessing that the brain eating ameobae at the FT have finally gotten to him.)

Harford wrote that instead of saying “I could’ve had a V8” 40 years ago, we should have said “I could’ve subsidized solar and then all our energy and climate problems would have been already solved.”

In a nutshell, Harford invokes learning by doing, which he refers to as Wright’s Law in honor of an aeronautical engineer in the 1930s who first identified this phenomenon. It has since been documented in numerous other areas, starting probably with Liberty Ships in WWII.

Yes, LBD is a thing. It has been part of the theory of economic growth since at least the 1970s, starting most notably with Paul David’s work on the antebellum cotton spinning industry in the US, and earlier than that even with work by Kenneth Arrow. Robert Lucas taught about it in the economic growth undergraduate (!) course I took as a small child at Chicago in 1981, FFS. (What a privilege and experience it was to be in that course.) I was so taken by the subject that my paper for George Stigler’s economic policy course in 1982 (when he won the Nobel Prize) examined empirically learning by doing at the Springfield and Harpers Ferry Arsenals prior to 1860. It’s hardly a new idea, or an unexplored one.

Alas, there are numerous problems with Harford’s application of LBD/Wright’s Law to solar.

One issue is: who is to say that the highly touted reductions in the cost of solar aren’t due to LBD?

That is, since cumulative output in solar panels has indeed increased dramatically over the years, learning would presumably have taken place and that plausibly accounts for some of the cost reductions. Harford himself says “PV is now so cheap that the question is moot.” So, perhaps LBD did its work.

Empirical evidence would be nice. And at most what Harford is saying is that we could have learned earlier. But if the learning has taken place (as evidenced by it being “so cheap”), albeit belatedly, solar should be taking over the world now, without subsidies, right?

Further, if Harford really means that too little learning has taken place, or it has occurred too late, then that would require (a) externalities/spillovers in learning, (b) the large subsidies to solar (which Harford pooh-poohs) were in fact too small and/or too late to generate the right amount of learning at the right time, and (c) market participants were unaware of the spillovers and did not take obvious steps to internalize them. He provides support for none of these.

With respect to externalities, it is not obvious that LBD effects are largely external to firms. Firms may be able to keep the benefits of their experience largely to themselves. To the extent they are internalized, there is no rationale for subsidies, and competitive firms will treat current production in part as an investment in future lower costs and expand output accordingly without need for government support or protection.

(NB. Non-compete agreements may be one way firms attempt to keep the benefits of experience internalized. I will soon write a post on the idiocy of the FTC’s ban of such agreements.)

I have analyzed LBD in the US shale sector in detail. I have found extensive learning effects, but the evidence for learning spillovers is weak. A firm’s own experience contributes more to its productivity than collective industry experience. This is evidence that learning is internalized.

Further, firms respond rationally to spillovers–by trying to internalize them. Mergers, consolidation, and concentration are means of internalizing learning. I note that consolidation is coming to shale only after more than a decade after the industry dramatically increased output and drilling experience.

In shale, it is plausible that much of the learning is done by service firms who internalize the benefits. Thus, even to the extent that industry experience explains productivity improvement, to the extent that service firms who, well, service the industry are the ones who generate this learning, these industry experience effects may be internalized as well.

That is, just because there is learning by doing, doesn’t necessarily mean that there are learning spillovers of the type that justify subsidies (or tariffs) to increase output (and hence learning). And if there are, there are strong economic incentives to internalize them. And if there aren’t, there’s no justification for subsidization. (David’s work on the cotton industry addressed the question of whether tariffs to stimulate domestic cotton cloth output were justified because of learning spillovers.)

All of these factors undercut the argument that the PV industry learned too little, too late. Where is the evidence that PV is unlike shale, and characterized by large learning spillovers which industry participants did not attempt to internalize through merger or other means? (I would also like to highlight the irony that Harford’s argument would imply that shale, for which there is actual evidence of LBD, should have been subsidized decades ago.)

Harford also has a myopic focus on PV cost, and fails to consider the total cost of renewables, including solar. Like other renewables, solar has intermittency and diffusiveness problems. Moreover, it has large and predictable output fluctuations (e.g., the “duck curve” problem in which solar output plunges when the sun starts to set). Due to these inherent features, useful solar will require beyond revolutionary innovations in battery technology (something Menton has analyzed in detail) that are not anywhere on the horizon.

(I note that battery technology has been the subject of massive research. It has also experienced tremendous growth in cumulative output, which has presumably contributed to learning. Yet it is nowhere even close to being an economical way to address output variability for renewables.)

Word to the wise: we are not going to be able to learn our way out of the sun rising, and more importantly setting. Or out of rain, clouds, and hailstorms. Or out of voracious needs for land to site renewables. Or out of the difficulties of disposing defunct panels.

Solar is part of a complex energy system. The cost of solar panels is actually among the least important aspects of the cost of relying on solar as a source of energy.

And talk about hindsight. Harford laments our failure to gaze into the distant future and foresee with precision the obsession with CO2 and climate change, and supersize solar panel output in time to provide cheap solar power when those obsessions became manifest. Yeah, and I should have invested in Apple when Harford says we should have subsidized solar. Or Bitcoin in 2013.

In sum, Harford’s woulda, coulda, shoulda lament in the FT is yet another example–as if more were needed–of the intellectual vacuity of those hyping the “energy transition.” Harford invokes a respectable economic concept–learning by doing–but does so in a superficial way that betrays a complete lack of understanding of it. And in a way that also betrays a lack of understanding of the real challenges of transforming an extremely complex energy system. Cheap solar panels may be a necessary condition for a cheap transition, but it’s hardly a sufficient one, or indeed, even likely an important one.

Alas, learning by doing doesn’t appear to apply in the writing of newspaper columns.