Jonah Lehrer, who I think is brilliant, writes a defense of the Solyndra investment, which I think is wrong:


A few years ago, a team of economists at MIT and UCSD analyzed the data from NIH and HHMI funded labs to see which funding strategy was more effective. The economists tried to control for every possible variable, such as outside scholarships and the quality of graduate students. Then, they compared the output of NIH researchers to HHMI investigators with similar track records.

The data was clear: In every biomedical field, the risky HHMI grants were generating the most important, innovative and influential research. Although HHMI researchers had similar qualifications to their NIH counterparts when they first applied for funding, they went on to produce twice as many highly cited research articles and win six times as many awards. They also introduced more new "keywords" into the scientific lexicon, which is a marker of highly original work.

The bad news, of course, is that all this creativity comes with a cost. This is why, according to the economists, the HHMI researchers also produced 35 percent more research papers that went largely uncited. (These papers were abject failures, the Solyndra of peer-review.) The moral is that these scientists weren't producing better research because they were smarter or more creative or had more money. Instead, they had more success because they were more willing to fail.

I am a huge fan of Tim Harford's Adapt, and indeed am now attempting to write a book of my own on a related theme.  But it isn't a brief for hog-wild speculation on any sort of project.  Scale matters--you can't take billion-dollar fliers on too many ideas.  Appropriateness matters--retirees shouldn't put their living expenses in tech stocks.  And the potential payoff matters--we should not invest billions to develop a slightly better form of white-out, or into experiments that have a 1-in-a-trillion chance of developing a low-cost way to turn lead into gold.


I think Solyndra flunks on all three counts.

In this post, let's think about the potential payoff.  Solyndra didn't invent the photovoltaic cell; they had a solar panel technology that didn't use silicon and was (by the company's account, anyway) supposed to be easier to install on the roofs of big box stores.  However, this design was tricky and very expensive to manufacture, apparently: my reading indicates that Solyndra was able to make a product for $6 that sold for $2-3 (before the market collapsed, anyway).  I take it that the idea was that Solyndra would somehow ease this disparity by getting up to scale, but the product was apparently extremely difficult to manufacture, and they never got their assembly line working properly.  In the comments to my last post, one of their vendor reps reports:

My company was a vendor for Solyndra from 2008 to 2010. We built machine components both directly for Solyndra and indirectly for their third party assembly machine builders. In total we did about $500,000 worth of business with them during this time. It's not a lot given the scale of their business, but we gained a lot of insight about their business model from the shop floor level during our visits to install and service equipment.

What we saw on the production floor was total chaos! Nobody knew what was going on. Their assembly lines were way too ambitious and complex, difficult to setup and maintain. I personally spent a day on their production floor and I don't think they got more than an hour's worth of parts out of an 8 hour shift. We built about $100,000 worth of components for them that never went into production because they kept changing the plan.

At least we got paid in full. And when they lost faith in the abilities of their US based machine builders they contracted with a European builder for an even more ambitious and complex machine. The last thing I heard was that the machine never worked and they were in litigation to get their down payment money back.

I distinctly remember a conversation that I had with a Solyndra engineer in 2009. He was complaining about how messed up (not his words) things were in his department. Somehow they had misplaced $50,000 worth of sensors, either through theft or incompetence, and that he needed to replace the entire order. I got the feeling that things like this were happening accross the board. That's how you spend so much money in such a short period of time. It's too bad for the employees. We worked with some very bright and dedicated people, but they got no support from the top. The investigations need to continue, both for their sake and for the sake of the taxpayers.

As far as I can tell, Solyndra was having these troubles with its first fab . . .  so we lent them money to build a second fab.  No one ever said, "hey, this manufacturing process is really fussy and may never work;" the idea was always that if you could just scale up and spend more money, somehow, eventually this product would break even.  Again, if they were going to discover cold fusion, maybe this would have been worth it.  But their only competitive advantage--using no silicon--was dependent on continued high prices of silicon.  No, it's worse than that; it was dependent on there being a big cost wedge between the price of silicon, and the prices of the commodities that Solyndra did use, like copper, gallium, and iridium. This was a bad bet; Lending them money seems like a very complicated and expensive way to take a large bet in the commodity markets.  And not necessarily a good bet.  Copper, gallium, etc are mined and in limited supply.  The primary ingredients in silicon are wood, charcoal/coal, and silica, aka sand.  


Moreover, say they were the cheapest solar panel on the market: as far as I can tell, they were still going to be rather expensive in competition with, say, coal.  That was going to limit their market to companies who were trying to boost their "green" footprint, either for ideological reasons, or because they thought it was good PR.  For them to move beyond that market, one of a few things was going to have to happen:

1)  The government would have to offer people a massive subsidy that made their solar panels cost-effective

2)  The government was going to have to enact some law that would massively penalize fossil fuels

3)  The world's supply of coal, oil, and natural gas were going to have to run dry.

4)  Their manufacturing economies of scale were going to have to improve massively.

Numbers one and two are properly defined not as risk, but what Frank Knight called "uncertainty": unless they get inside information about imminent passage about a law, there is no way for investors to know if and when this will happen.  (Worse, even if it does happen, it may not stick: one of the many things that hurt Solyndra was that the Italian government got into financial trouble and cut back its solar subsidies)

Number three is outside of any reasonable investment horizon, even for the US government.

Number four is possible, but you always have to be mindful of their input cost--unlike silicon, the price of their raw materials is limited by the ability to dig the stuff out of the ground, not the number of firms in the business.  (Silicon prices collapsed because persistent rising demand caused high prices which drew in a bunch of new firms, a classic manufacturing story.  By contrast, look at the market for oil:  high prices are signaling rising demand bumping up against supplies that grow more slowly than our appetite for hydrocarbons.)

In other words, it was unlikely that Solyndra was going to work, and the payoff if it did work was pretty low.  The reason that so many private sector investors turned this down was not that it was "too risky"--risk changes the price of the investment, not necessarily the willingness to make it.  Rather, the problem was that the scenarios in which Solyndra could be successful were few, and even if they were successful, their success would consist of putting a modestly different sort of solar panel on the market.

Or as Andrew Samwick said, somewhat more briefly, "If you have a risk that the private sector won't take, it is because the private sector does not see it as a profitable risk to take. It does not get more profitable if the government takes it. It only formally shifts the risk of losses to the taxpayers."

It's not enough to say that radical basic research often has unexpected payoffs. This was not a grant to a scientist; it was a loan to a company.  As such, you can't just judge whether their technology was interesting--but whether the product had a potential market, and if so, whether the company was capable of manufacturing product to meet that market demand.

More later on scale, and whether this sort of investment is appropriate for the government to make.