Yesterday, I mentioned that, looking back historically, pollution regulations often end up costing much less than initially projected. When Congress created a cap-and-trade program for sulfur-dioxide emissions back in the 1990s, for instance, the EPA predicted that permit prices would hover around $750 per ton. Instead, by 1997, they were selling for around $100 per ton. All told, the program ended up costing about one-fourth of what was predicted. The EPA's original forecast was way off. And there are plenty of similar examples.

So when the CBO recently found that the cap on carbon emissions in the Waxman-Markey climate bill would only cost the United States about $70 per person per year by 2020, there's decent reason to think even this modest amount might be an overestimate. But Jim Manzi isn't buying it:

Presumably the same awareness of the track record of asserted prior under-estimation of environmental costs was available to both the EPA and CBO as they prepared their cost estimates. Unless we wish to assert that they are biased or simply irrational, why would we assume they failed to incorporate this information into their (very similar) forecasts of costs by 2020?

This is a good question. If the EPA and CBO and other analysts consistently keep overestimating the cost of environmental rules, why can't they just factor this into their models? Why do they keep making the same mistake? One possible explanation, given by economists Eban Goodstein and Hart Hodges back in 1997, is that these forecasters have a hard time predicting the sorts of technological and market innovations that enable companies to comply with pollution controls so cheaply:

So, for example, the much lower than expected costs for the acid rain program can be explained in retrospect by the increased flexibility that firms were given to achieve their mandated reductions in sulfur dioxide emissions. Rather than install expensive scrubbers (or buy extra permits), many more firms than expected have met their sulfur dioxide targets by switching to low-sulfur coal, or developing new fuel-blending techniques. Railroad deregulation, along with economies of scale, led to an unexpected decline in low-sulfur-coal prices. And with the increased competition from coal, scrubber prices fell by half from 1990 to 1995.

All this is easy to see after the fact, but would have been very hard to predict. Hahn got his 20,000 lost jobs from air toxics regulation following this same practice—ignoring innovative market responses. While parenthetically noting that "technological improvements could reduce the direct economic impacts," the study explicitly ignores the possibility "because of the difficulties in predicting how technology will evolve." Because in the mid- to late 1980s, available control technologies for coke ovens seemed to be quite expensive, Hahn assumed that regulating air toxics would simply shut down much of the industry. ...

Peering into the future is hard work. It is, in fact, close to impossible for economists to predict the specifics of how technology will evolve. This is especially true since much of the information about potential innovations consists of closely held trade secrets, which industry has little incentive to reveal. But basing cost predictions on scenarios that assume no technical evolution is guaranteed to produce gross overestimates. Innovation is indeed something at which markets are very good. When given a narrowly defined task—to produce commodity x emitting less of pollutant y—short-term substitutions and long-term shifts in technology guarantee large cost reductions over current practice.

The basic point is that markets almost always tend to be smarter than forecasters, and adjust in ways that no one expected. (As such, this can be difficult to account for in an economic model.) So, while we have to make policy using the best estimates we have available—and the CBO's approach is highly, highly respected—some additional optimism may, in fact, be warranted. Yesterday, Megan McArdle argued that, "[u]nless Waxman-Markey serendipitously leads to the development of some clean technology that makes carbon obsolete, I'm pessimistic about how much it will accomplish." But that's the whole point of putting a price on carbon. And I wouldn't call it serendipity. If past examples are any guide, there's very good reason to think a climate bill will accomplish exactly that.

P.S. To be clear, the CBO and the EPA do go to great lengths to try to model how markets and technology will react to a price on carbon, and the models keep improving. The CBO explains its methodology here. It's just that, in the past, these predictions have tended to err on the cautious side.

--Bradford Plumer