The Problems With Cost-Benefit Analysis

Kevin Drum has a smart post on the difficulty of applying cost-benefit analysis to long-term issues like, but not limited to, climate change :

I didn't get involved in this conversation for a simple reason: I've been on both the producing and receiving end of too many cost benefit analyses to trust them. If you're being relatively honest and if you're dealing with fairly concrete, short-term issues, they're useful tools, but even then it's still the case that you can manufacture strikingly divergent conclusions by manipulating your assumptions and inputs by surprisingly small amounts. Cost-benefits usually look like they're grounded in hardheaded thinking simply because they're numerically based, but quite often they're nothing of the kind.

And that's in the best case. Climate change is far worse. Not only are we decidedly not talking about concrete, short-term issues, but there's a huge asymmetry in what we can say about the cost side and the benefit side of fighting global warming.

On the one hand, you have the actual science of climate change. And although climate models are enormously complex and subject to considerable uncertainty, they're fundamentally based on physics, chemistry, and thermodynamics. We know how much CO2 we're pumping into the atmosphere and we can project with pretty good confidence how much that's going to increase over the next century if we do nothing to stop it. We know how the greenhouse effect works, we have pretty good historical records of how CO2 concentration correlates with global temperatures, and we have a pretty good sense of the feedback loops involved in things like melting icecaps and saturation of the ocean sinks. Basically, our level of uncertainty is within tolerable bounds here. And what we know is that if we do nothing, global temps are absolutely certain to rise 2°C over the next century, fairly likely to rise by 4-5°C, and at least somewhat likely to rise by 6-7°C. The lower number would be bad but, just possibly, manageable. You could at least make an arguable case, as Manzi does, that the cost of preventing an additional 2°C is higher than it's worth. The two bigger numbers, however, would be catastrophic. Unfortunately, the science increasingly suggests that these higher numbers are considerably more likely than we thought even a few years ago, and any serious cost-benefit analysis needs to address that. Using only the lower number avoids tackling the real problem we're up against.

So that's the climate analysis in a nutshell. On the opposite hand you have the economic analysis. And that's simply hopeless. An economic analysis that goes even ten or twenty years into the future is as much guesswork as anything else. One that goes a hundred years into the future is just voodoo. It looks like economics, but you might as well be throwing darts. Compounded over a century, even minuscule changes in assumptions and operating parameters produce enormous changes in your conclusions, and the result is that you end up deep in the weeds arguing over tiny differences in those assumptions instead of simply admitting that they're flatly impossible to forecast. That's good for slowing down the debate, but not much else.

That last point is important. Numbers are a language. And you can lie, or mislead, or miscommunicate, in any language. The neat thing about using numbers is that relatively few people can understand you and many more will simply assume your analysis is sound. After all, if it wasn't, how would there be all these numbers? That was, as far as I can tell, the principle underlying the financial industry between 2003 and 2007.

The whole situation leads to really banal conclusions like "sources are important" and "make sure you know the uncertainty of the estimate." Not exactly the most exciting close to a blog post, but true nevertheless.