Tuesday, September 11, 2012

Austrian Theory and Historical and Empirical Analysis

This month's Cato Unbound is a bit "inside baseball" among Austrians and post-Austrians on Austrian economics and empirical analysis. Steve Horwitz starts things off with an argument that Austrian economists are in fact more empirical, because they consider more factors to be important for what contributes to the features of the economy than do other schools. At least among the Hayekian scholars I mostly associate with, there is little question this is true. Hayek said you can't be a good economist without being an interdisciplinary scholar, and I (and Horwitz) concur.

George Selgin gives a very clear rebuttal -- so clear that it seems clear where the problems are in his analysis. To some degree, there is little one can argue with in regards to what he says about economics analysis and about how many (let's say, half) Austrian economists do economics. However, I think he is missing the real critique of historical and empirical analysis.

Let us consider the minimum wage. If you look at the historical/empirical evidence, you will see that most of the time in the United States after the federal government raised the minimum wage, unemployment went down. Conclusion based on history and empiricism alone: raising the minimum wage causes unemployment to decrease. However, economic theory says increasing the minimum wage causes unemployment to go up. This clearly contradicts the empirics.

Or does it?

The problem is that we are dealing with a complex system. In a complex system, there is not a simple cause-and-effect relation. Whether or not there is a minimum wage, and whether or not one raises it if there is one, will affect unemployment -- but it is hardly the only factor affecting unemployment. If there are other factors acting in the economy to decrease unemployment more than the increase in minimum wage would increase it, you will of course see unemployment go down.

By understanding that the minimum wage increases unemployment, you can understand that without the minimum wage, unemployment would have gone down even more than it would have. Could a regression analysis help us understand how much a particular increase in the minimum wage would increase unemployment? Well, that would require that you include all the factors contributing to employment. But how can you do that? Here are a few of the things that will contribute to employment levels:
  •  the property rights regime
  • the institutional composition of the economy
  • trade regulations
  • economic regulations and enforcement
  • immigration regluations and enforcement
  • features of the black market for banned or regulated goods
  • rate of innovation and entrepreneurship
  • demographics
  • general racial/ethnic attitudes of society at large
  • child labor laws -- whether or not they exist, and their features if they do exist
  • tax policy and changes
  • monetary policy and changes
There are no doubt more. Please note that some of these are measurable, but some are not. You cannot actually quantify a quality (like racial attitudes), so this cannot actually be factored into a regression analysis. (The fact that there are people who think you can accurately measure such things as "happiness" such that you can mathematize it rather than create simple, inaccurate rankings does not disprove this point, but rather is a condemnation of those social scientists who are engaging in the logical fallacy of quantifying a quality.)

But even if there were nothing but measurable factors, you would have to make sure you are including all the factors involved -- and the ways in which they interact -- to do the analysis. What we are faced with is an incomputable problem. You might be able to discover general trends, but I doubt you would ever be able to say that if we raise the minimum wage from $6.00 to 7.15 per hour you would necessarily get an X% increase in unemployment.

What you can know is that an increase in the minimum wage will price out low-skilled workers. It can do so in a variety of ways. It may be cheaper to automate; it may be cheaper to hire one skilled worker rather than three unskilled workers; it may be cheaper to relocate to a place with a lower minimum wage; etc. The minimum wage may also drive some businesses out of business. Which of these will occur? In what ratios? What will the resulting effects be?

One resulting effect may be that, over time, as the minimum wage continues to go up, low-skill jobs are run out of a region, meaning nothing but high-skill, high-wage jobs are available such that a subsequent increase in the minimum wage will have absolutely no effect on unemployment. When, if ever, will this stage be reached? What conditions would have to exist for it to be reached at all?

So where does this leave us? It leaves us with Mises' praxeology. We have to understand the underlying principles of human action to understand what is happening in the economy. If we look at the historical and empirical data, we are seeing the surface results of an underlying set of complex actors engaging in complex interactions, reacting to complex factors.

Actually, it leaves us with both praxeology and contemporary computer modeling. But even the models have to have complex actors acting as humans would act. Which brings us right back to . . . praxeology. The models, too, only show us what will result over time -- the emergent patterns, or surface results. The computers can only run the underlying complex interactions, not make them transparent to the humans trying to understand the results.

And even with computer models, you have to make sure you have all the relevant parameters. One can never tell what was the relevant parameter in a historical situation that made this downturn better or worse than the last one. Most of the time, we think we know that this or that factor was relevant, but too often we are looking at surface issues and missing the underlying issues. We stop the cause-and-effect chain too soon, and have no real idea what caused the current conditions to be as they are or were.

A good example the Great Recession. What caused it? If you were to rely only on what mainstream economists and the news media have been saying, it is the fault of the mortgage companies who created the bundles with the toxic mortgages in them. But that is in fact pretty far down the road from the primary causes -- some of which are the factors explained by ABCT. Most people fail to see that it was government policy, government regulations, and cronyism that caused the Great Recession -- because few follow the causal chain far enough. Austrians do.

And few understand the complexities involved. How many economists consider entrepreneurship and innovation and technology to be "external shocks" to the economy? But such an attitude is the height of absurdity. Economic growth is caused by the above mentioned external shocks. Which means understanding the causes of economic growth is outside mainstream economics. Which is absurd. Interdisciplinary economists -- which include almost all the Austrian economists I know -- understand that there are a lot of complex factors at work. And the understand that economic analysis is at best incomplete unless you include all of these factors, that most of the things that "shock" the economy are not really external to it, but are necessarily part of the economy. This means economic analysis is necessarily extremely complex. And the more complex a phenomenon is, the less it is able to be reduced to mathematical analysis. Especially, as Hayek observed, if it is more complex than the entities studying it.

Finally, Julian Baggini and Lawrence Krauss debate some of these same issues in a somewhat different context. Are all questions that are "answerable" necessarily "empirical"? I think this is the real debate.

1 comment:

Andrea Clark said...

I've been following this debate a bit in my not-so-spare time and I hadn't yet decided what to think of Selgin's piece. This a beautiful response. Thank you.