Saturday, July 14, 2012

Why models cannibalized economic thinking

Paul Krugman did not win the Nobel prize because of his blogging, he won it because of academic research making an extensive use of sophisticated models. Yet, in a reply to Simon Wren-Lewis, Krugman reduces the usefulness of models to the mere function of gadgets. This is not just a casual observation, it is a very strong statement that ought to trigger a much-awaited debate on the way economists think about the economy.
What makes the statement very strong is that it goes much further than the views of Wren-Lewis, who is already quite critical towards the economic profession.

What Wren-Lewis seems to be saying is that models are useful, but economists should pay more attention to the context (depressed demand, for example) when choosing a model to apply. He also believes that more sophistication (such as intertemporal optimization) helps solving questions.
Krugman considers instead that models should remain as simple as possible and don't even necessarily need to make an attempt at reflecting the reality. He sees them as tools for uncovering non-obvious mechanisms, and he names them gadgets.

"But I guess not everyone even on the sensible side of macro sees it that way. And that is a problem. A gadget is only a gadget, and you should not let it define your field."

In other words, economic thinking may be built upon sophisticated models, simple models or just logic. But logic should prevail over models, which far too often is not the case (especially if what you read between the lines is that most people on the non-sensible side of macro see it the other way).

Let's take the example of inflation. Along with growth and unemployment, this is one of the core areas of macroeconomics. Yet, this an area where models say little and economists of different schools disagree the most. Neoclassical models usually point to the undesirability of inflation, a trend that culminates in the Friedman rule of deflation. There is however a strong logical argument in favor of inflation, even when there's no need for deleveraging (don't count on logic for quantifying how much inflation is desirable though).
It derives from the empirical observation that prices and wages are sticky. They may go up, but not down. Of course, in the neo-classical framework, prices and wages are supposed to be flexible both ways, but in the real world, it looks very difficult to argue that they are (even if we're still waiting for someone to come with an undebatable explanation about that).
Now, if prices cannot go down, and inflation is nil, it means that there are no price movements at all, which implies that there can't be any relative price adjustment in the economy. As productivity increases differently across sectors, and consumers' tastes keep evolving, price adjustments are necessary for an optimal allocation of resources. Without inflation, the equilibrium is thus sub-optimal.

This leaves us with a question: why did models cannibalize economic thinking?
  • There is of course the explanation of the ideological conspiracy (I have yet to read a bit more about the Mont-Pèlerin Society, but I'm quite skeptical of conspiracy theories in general). Models served well the neo-classical agenda. So, economists of that school supposedly first provided mathematical descriptions of economic mechanisms and then used the resulting framework to suppress any ideological dissent. The maths erased history in the economic reflection, eliminating in the process all those historical facts (such as the Great Depression) that did not get along very well with the theory.
  • A more tempting explanation is to keep in mind that economists are human beings, and the key driver of human behavior is to spend the least possible energy in order to achieve any given result. In particular, humans will avoid thinking unless it is completely necessary. And here is the big advantage of models: you don't have to think, you only need to run the model (you would be supposed to think, but no one would notice if you don't). You might say that it takes some energy to understand the maths, but this is probably nothing compared to what is actually required from thinking, especially in a counter-intuitive discipline like economics. Besides, learning maths can be seen as an initial investment with a return over the entire academic career.
  • Beware of the model invasion
  • But maybe models invaded economics because they fulfill a very specific function: they allow for the discrimination of economists. Models help synthesizing what an economist thinks and understands. It would be awfully difficult for a professor to assess students if he or she could not rely on their understanding of models, and their answers  to exercises and questions based on models. Sophisticated models are taught at higher educational levels not because they deliver a better economic insight, but because they help filtering the more sophisticated students (which might be what Wren-Lewis calls "the inevitable focus at the masters level on the recent macroeconomic literature"). And it does not stop at students. Sophisticated models are usually the reason why economists receive prizes, including Nobel ones.

2 comments:

  1. >There is however a strong logical argument in favor of inflation

    There is a strong argument that inflation increases spending, but only because it helps the financial markets at the expense of people's savings. http://jobcreationplan.blogspot.com/2012/06/inflation-scam.html

    Also see http://www.politonomist.com/gdp-deflator-and-measuring-inflation-00491/ ... most economists like inflation, most other people don't.

    ReplyDelete
  2. I'm not really against the use of models in economic study; if anything, they give a pretty decent snapshot of economic conditions and they can sometimes be a baseline with which prevailing scenarios can be compared to (but that's just my opinion). Still, I think there should be greater emphasis on the need to have a more thorough approach to studying economical data, something that trimmed-down models cannot provide.

    ReplyDelete