Co-authored with Nat Keohane.
Last week Nat Keohane and I tried to shed some light on the inner workings of economic models when it comes to jobs. Among other more specific statements around climate policy, we also said that,
many macroeconomic models don’t actually attempt to model jobs. In fact, they generally assume full employment no matter what happens, which doesn’t leave any room for estimating increases or decreases in jobs as a result of specific policies.
We should have been clearer here. If you think of employment as “total units of labor employed, given equilibrium wages and household labor-supply decisions,” then yes macroeconomic models do model employment—just as they model the equilibrium values of other inputs and outputs in the economy.
That’s just not how you or I think about jobs, and that has some major implications.
Economic modeling versus the real world
Computable general equilibrium models of the economy literally assume full employment in the sense that everyone who wants to work works.
When we think about unemployment in the real world, it is due to the fact that people actually lose their jobs. More technically, there are market frictions that keep wages high in recessions. As a result, demand for workers goes down.
Not so in the world of general equilibrium economic models. There, wages rise and fall with the fate of the economy. In a recession, wages fall and people simply choose to work less. The technical term is the household “labor-leisure tradeoff.” People work less because they supply less labor to the economy.
Most people, of course, would argue that when wages fall you have to work more to make up the gap. That intuition can’t be true in an economic equilibrium, which is what the models are trying to capture; hence the discrepancy between models and the real world.
Not all models are created equal
So how then do economic modelers estimate employment impacts?
First, not all models are of the “general equilibrium” type that have the full-employment assumption built into them. For example, the Department of Energy’s Energy Information Administration relies on a macroeconomic model called National Energy Modeling System (NEMS), which has a different structure. Partly as a result, the NEMS model does produce employment estimates.
Second, you sometimes see even general equilibrium models being used to derive numbers on jobs. In those cases, however, the jobs impacts are computed after the fact: the modelers take the estimated effects on output and then back out employment impacts using rather arbitrary rules of thumb that assume a high degree of friction in labor markets even over long periods of time, often decades. This is problematic, to say the least, because it goes directly against the grain of the underlying models used to produce those results.
The Peterson Institute study mentioned in our last post, which uses the NEMS model, is in the first camp. It is also among the first to take a more realistic look at the jobs impacts of climate policy given the current recession. Not surprisingly, it comes up with a very different answer on the jobs question.
Yesterday’s New York Times published three terrific op-eds worth a second thought.