Tuesday, February 08, 2005
He asks for an answer to this question: Where are the models? No, not those.
It�s likely that I�m just missing something, but I often wonder something when politicians and pundits are talking about economic issues: where are the models? They throw out a lot of numbers, but rarely do they explain their sources, let alone the assumptions involved, and I often get the impression that the sources of the numbers didn�t spend a lot of time generating them.
Meanwhile, I keep getting the feeling that a basic spreadsheet model of some key variables in the US economy would help me�and, presumably, a lot of other well-informed but economically clueless people�come to a conclusion on issues like Social Security reform. I�ve seen models online for the entire US economy, and I�m sure policies like this would have an effect on all levels of the economy, but to begin with I�d like to see a well-explained model on just how some of these plans might work. (And inevitably, I�d have to see the competing model�s claims about the effects of the same policy and make up my mind which model to trust.) But sadly, I have almost no clue what to model, what parameters to use, and how to model any of the effects. And absolutely no time to learn any of that.
So maybe I�m being over-optimistic that simple models of some of these economic ideas would be useful, but then I�d have to ask: if V = I*R works for 99% of the problems in electronics (and more physics-based�read: complex�equations can be found for most of the rest of the cases), why can�t simple models be equally useful in economics?
Part of the answer is just that: You are being over-optimistic. Modeling the behavior of human beings is different than modeling the behavior of particles or waves. In the post just below we had to ask how much people would value being given the option of getting out of Social Security and going into private accounts. That means we have to predict how people behave. Since I don't believe in psychohistory, models end up having an inherent level of uncertainty about a model's forecasts. And I mean an uncertainty that does not lend itself to statistical measurement like variance. I mean uncertainty in terms of a process that has random events drawn from a jar, if you will, the contents of which are unknowable.
Second, the types of models that pundits and politicians use are often policy prescriptions about macroeconomics, the economics of the aggregation of millions of individuals and firms into such data as GDP. As such we don't have good laboratory experiments to test models in. Macroeconomists -- the area in which I work -- take a good deal of time coming to closure on whether a model is predictive (I resist "accurate" because models are not reality but abstractions from reality to make analysis easier).
The argument over Social Security is in part an argument over consumption theory. The crabbing about Krugman's piece last week comes in no small part from Krugman trying to cram some rather contentious parts of consumption theory into a NYT op-ed. And, as he admits himself, it didn't work. The details of that theoretical argument would make your eyes glaze over (a small illustration will be in a post I wrote last night and will post later this morning); a big reason why it's hard is that the arguments occur increasingly in terms of mathematical formulations rather than good analytics written in English, and it takes a real special skill to move the math into the analytics. The problem, in short, is rhetorical. We don't talk to non-economists in a convincing way by and large. (I have a short list of exceptions to this, but that deserves a post of its own.)
The problem therefore isn't that we don't have a model but that we cannot explain it very well.
*--"To make economics an accurate science."