Subsidies are hated by most every economist. Economists believe subsidies distort markets and reward dollars to inefficient projects. The 30% solar Investment Tax Credit (ITC) is often decried by oil and gas think tanks, libertarians, and pure market economists.

Read this quote from what is clearly an industry-backed thinktank, the “Institute for Energy Research”:

“The U.S. Treasury estimates that the Production Tax Credit will cost taxpayers $40.12 billion from 2018 to 2027, making it the most expensive energy subsidy under current tax law.”

This statement may be true in a vacuum, but it ignores the positive externalities associated with a manufacturing subsidy, and it is wrong on a couple of fronts.

First, sunshine is a non-rival good. The materials and man hours needed for solar panel production are rival in the short term-someone making a solar panel cannot also be making shoes. But long term, if subsidies truly do push the solar cost curve down, the rival marginal production cost of each solar panel is almost nothing. And, if that’s the case, there is a key element here that is different from ethanol subsidies.

Ethanol subsidies affect corn, which, no matter how efficiently produced, comes from limited land with limited nutrients. Corn has a law of diminishing returns on input dollars-over time, to eke one more bushel of corn out of an acre of land requires more and more input dollars and man hours.

Solar though, has nearly infinite power available. It is not a production subsidy, but rather a manufacturing subsidy. That is, similar to Moore’s Law in semi-conductors, for every doubling in manufacturing capacity of solar panels, we see cost reductions in the total Levelized Cost of Energy (LCEO) of 20%. This is known as Swanson’s Law or the Swanson Effect. Below is an old graph. Costs per watt in 2020 on solar panels are closer to 30 cents a watt!

The costs come down because as we build things, we learn to build them better. T P Wright’s seminal 1936 paper on airplane construction was one of the first formalizations of this relationship. Build something over time, and initially you get way better at it really fast, and then over time you continue to get better at building the item, but you need more inputs for the same rate of improvement. Think of a concert violinist, for instance. They may achieve 98% mastery by 2,000 hours of practice, but take another 2,000 for the final 2%. Building things work the same way, but there is never mastery. The only asymptote is the availability of raw goods.

Solar used to be insanely expensive. California bought a whole bunch of it when they shouldn’t have. Now they have too much solar, at a cost 5x the market, because they signed long term contracts. In retrospect, this was not wise. It was shorting the price of all other power sources, and natural gas became capable of providing clean-ish energy at 1/5th the price.

But, at the time California signed their ill-fated Power Purchase Agreements (PPAs), it seemed that the only way to start making panels so that solar became more affordable so that we could make more solar was for energy prices to rise. Now, with energy prices falling, we were faced with a conundrum. The eventual marginal price of solar should be close to the raw inputs, assuming we’re going to make a whole bunch of it. That’s definitely good. But solar was super expensive, so you would need hundreds of loss leader.

That’s where the 30% solar subsidy came in. We still had hundreds of loss leaders, but many were losing because the subsidy worked even better than expected. Solar has gone from 30+ cents a kiloWatt hour (kWh) to 2 cents. That is a dramatic decrease in cost that beat even optimistic projections from the past.

The short term losses from subsidies are real. It is no different than a tax in support of solar. There are key distinctions however.

First, the subsidy was on the total cost of the project rather than at a fixed rate or a fixed procured amount. This incentivized total costs of production to decline dramatically.

Second, when faced with an a threat to the human species, subsidies alter the temporal position of energy production. CO2 produced now vs CO2 produced in the future has a different position on the (quantity, marginal potential harm) graph. By shifting CO2 production into the future, you are not only creating a net positive for each marginal CO2 shift, you are also buying more time for other innovations (maybe actually learning to build nuclear again), and avoiding non linear responses. Perhaps there is a certain amount of CO2 we can pump into the atmosphere with no changes ,and adding a single molecule more of CO2 will have a massive negative effect. Non linear relationships dominate in complex systems, and, when manufacturing subsidies produced Gigawatts of Solar, they enabled us to reduce our chance of hitting an unknowable threshold.

Now, where the United States likely went wrong is not requiring the 30% subsidy to be used on parts from Americans or close trade allies (read: not China). The inverse happened. China heavily subsidized their solar panel production, then sold it into a subsidized US market. China absorbed even more up front losses than the United States on the solar production curve, but now they have knowledge that puts them so far ahead of the US that solar panel manufacturing is almost extinct in the United States.

Ironically, this means the United States is facing a similar question to that which it faced before: for the nation to learn how to manufacture solar again would require that we take an even larger short term hit to solar panel prices than hit the United States’ imposition of a 30% tariff in 2018 caused. But, without an imposition of a large import tariff (which will cause less solar to be built in the US in the short term), we cannot ever learn to build solar panels ourselves.

So now the relative CO2 position must reverse in the short term. Manufacturing subsidies worked and worked well. Solar’s price per kilowatt hour is now among the cheapest forms of energy in the history of humankind, and before long the marginal cost will be nearly zero. But if the United States can’t manufacture the needed technology, we will have shifted the ability to produce zero marginal cost power to a nation that puts millions of people in concentration camps. And, when China inevitably introduces export controls to the United States in 10 years or so, if the United States has never had the guts to learn how to make solar panels, a country that cares not at all about pollution, like China, will continue to burn coal, simply because they have plentiful supplies, cheap, and near power generation stations.

Manufacturing subsidies work, and likely produce substantially more benefits than harms in areas that we know have plentiful to limitless natural material available. They shift potential harm into the future and reduce non-linear complex responses. However, the original manufacturing subsidies on solar didn’t go far enough. We are now faced with a choice. To best lower future energy costs in the United States to zero, as well as prevent long term spikes in CO2 production, it’s time to again produce solar panels in the United States.