Ezra links, in a spirit I think we can charitably call self-flattering, to Dani Roderik’s taxonomy of dismal scientists. “First-best economists” (we learn) decide public policy questions by “knee-jerk” application of simplistic, idealized models. “Second best economists”—who understand that the world is, like, complicated and stuff—are more amenable to targeted regulation. (Is there some kind of unintentional-irony prize we can award for casting your opponents as adherents of reductive, oversimplified models within a reductive, oversimplified model of your disagreement? The Gödel Award, perhaps? Or the Chait?) Ezra is, of course, a “second-best” type, while his libertarian friends (I suppose that’s me) tend to be “first-best” sorts.

Naturally, there are some difficulties with the simple model: As self-described “better-than-first-best” economist Tyler Cowen notes, libertarians too are usually prepared to endorse government provision of certain public goods. And I don’t think anyone literally believes real markets are as frictionless as blackboard models. As Rodrik himself suggests, the real disagreements are not so much about whether market imperfections exist, but about how pervasive and serious they are and, more importantly, how feasible it is for the political process to reliably identify and target those imperfections without causing other problems. But from this perspective, the difference here is less about how the world is than about which model is relevant to the policy task—which may explain Rodrik’s observation that some people who construct complex models in their academic research nevertheless apply simpler ones when discussing policy. Euclidian geometry may not accurately model our universe—spacetime is curved, after all—but if you’re renovating your house, it will do. And, indeed, if you overhear your contractor muttering about Riemannian manifolds as he looks over the plans, you’re apt to get a little nervous.

Also, as I suggested in a previous post on the old “Econ 101” sneer, the way the more complex models are invoked does not always inspire enormous confidence. Often, it looks suspiciously like what’s going on is something like this: We want very badly to improve situation X. Unfortunately, basic economic theory suggests that the big, crude, obvious intervention that would fix everything will have unwanted side-effects. (Think minimum wages depressing employment.) Uh-oh! But wait, there are specific theoretical conditions under which this doesn’t hold, and the intervention works. (Monopsony, say.) Therefore (in some appropriately loose sense of that term) those conditions must obtain here, because otherwise (sad!) we couldn’t act to improve things. We may then trust that the textbook optimizing intervention will be precisely identified by the democratic magic of the visible hand.

That’s a caricature too, of course. Any sane person realizes that the Civics 101 model is also too simple. What real, reasonably sophisticated people instead believe is that relative to the magnitude of the potential gains from intervention, the problems of public or bureaucratic ignorance, of bureaucratic inertia or self-dealing, of interest-group influence on the regulatory process, and so on, are unlikely to distort policy enough to change the basic balance of considerations. Perhaps they want to consider the broad question of whether to intervene on the assumption that some way can be found to reduce those complications to tolerable levels. Which, as Rodrik might have it, “allows each expert in a field to propose first-best solutions in that field, leaving complications elsewhere to be dealt with by others.”

There’s a neat symmetry here. The more closely you think the “first-best” model comes to approximating reality, the more concerned you’ll be with how political failure might muck things up. If real markets are just horribly riddled with huge imperfections, even a suboptimal, clumsy political process has a good shot at making things better, and so at least for the purposes of the decision to intervene, we can adopt the “first-best” Civics 101 model. And just so in the other direction: If political failure is significant and endemic, there’s no point modeling markets in more detail than the political process can resolve; if the first-best model is pretty close, we’ll want to investigate every helpful little tweak we can make. Since the adequacy of each model is, in part, a function of the adequacy of the other, it’s easy to imagine a feedback loop: Start by focusing on a simpler, more idealized model of markets, which gets you thinking about why interventions might misfire, causing you to pay more attention to the second-best aspects of politics, reinforcing your preference for a simpler economic model, and so on. Or vice-versa.