mercoledì 19 maggio 2010

Economic data pro and con.

Commenting on these questions posed by Prof. Rizzo on TM I've shown the world I suffer multiple personality disorder. Here's the evidence.

I’m on two minds on the issue because although I recognize that empirical evidence and operational definitions are important, problems in these endeavors are not specific to Austrian economics, as they should be recognized as problems by almost everyone: I know, for instance, of no econometric evidence of monetary non-neutrality (cfr King & Plosser), although I believe that money is always non-neutral.

Anyway, my being on two minds has begotten to contrasting comments.

Comment #1

ABCT has not been proposed as a model so it can’t be directly compared with the data. Only highly aggregate fully specified DSGE models can, and I wouldn’t bet that this comparison with the data is better than playing curve fitting.

However, to look for evidence, I would first look at financial data (there is always plenty of financial data: the credit channel literature is full with econometrics) to check the prociclicality of all financial fragility proxies: financial leverage, maturity mismatch and risk taking. This is the credit creation (and destruction) process at work and it can take millions of different forms (pure bank credit or shadow banking, for instance).

However, credit must be linked to money in order to have ABCT, otherwise it’s Minsky, not Mises. So I would check for the effect of interest rates and monetary aggregates on the financial intermediation proxies. This I think has already been done in the credit channel literature, which I haven’t checked.

Then I would check the structure of production. Unfortunately, I know of only a few papers investigating something similar: Mike Montgomery has shown that capital complementarity explains the lags in the economy’s response to shocks, Mulligan, Wainhouse and Keeler have found something else but I don’t remember the details. It was all about correlations having the expected behavior.

However, it appears that not much evidence is needed to prove that interest-rate sensitive markets are more prone to crises than others, which is all that is required by ABCT: this is a well recognized stylized fact.

Upper and lower turning points cannot be predicted. I would guess that ABCT should predict several years of boom and several months of recessions in standard conditions: that’s what usually happen. There are lots of complicating factors, however: the Japanese ZIRP and Hoover can make things last much longer. I don’t know of anyone capable of predicting turning points, however.

For what concerns the links between monetary, financial and productive data (the third ones are quite scant, because only Austrians are interested in them, while the others are plentiful), data can only show correlations, not causality. So none of these data amounts to corroboration or refutation for any theory.

In few words, Austrians can free ride on the credit channel and look for patterns there. I normally don’t read econometrics papers (it’s deadly boring) and I’ve only seen a few of them, but there are hundreds.

Comment #2

I doubt that the notion of intertemporal disequilibrium can ever be operationalized: it’s not a matter of aggregates but of structures, and structures imply knowledge problems.

Also the notion of monetary non-neutrality has no operational content. All economists can do is to play vector autoregressions and hope noone notices that they prove nothing. The King/Plosser explanation of observed correlations between money and output is as good as any other. There is no empirical evidence that money is non-neutral or neutral, otherwise real business cyclers and new keynesians would have ended their squirmishes decades ago. The problem is empirically undecidable.

While the latter is a problem for all schools of thoughts, the former is a problem only for Austrian economics because others disregard structure. So, let’s disregard money too and all problems are solved. :-)

Efficiency is another notion I don’t know how to apply to reality: it is a property of fully specified models of artifical economies, not something which can be predicated of reality. When I see something, I can never now whether it could have been better without knowing all the alternatives and all the brute data.

Another example is the natural rate of interest: no one knows it. It’s a theoretical construct with no empirical counterpart. This is a problem both with Austrian and New-Keynesian economics. More than a problem, however, I would say it’s an epistemic property of markets.

Another example is taken from my terrible introductory macro textbook (Blanchard), which wanted me to believe that the 1980-1983 recession falsified the rational expectation theory of stagflations. It falsified nothing, as usual with economic data: newclassicals resorted to the makeshift of distinguishing between credible and non-credible disinflationary attempts and saved their theory. It is a blunder to consider ratex a falsifiable hypothesis.

Before concluding, I would add that new keynesian models have been criticized for excessive plasticity: any contrarian data was rationalized by changing some detail in the model. This is what I call “curve fitting theorizing”, which of course it is an oximoron. David Romer had some issue with this in his textbook, too.

Data collection reveals problems, but offers no solutions. Theories which are so rigid in predictions to yield falsifiable results are usually very poor theories. DSGE models are, on the contrary, very rich theories from this point of view. So rich that they can predict everything, by properly specifying the details.

One problem I’ve had is that the Greenspan era has lasted for so long: ABCT without chinese savers and technological innovation would have predicted a crunch in the ’90s, not 20 years of inflationary booms with some minor slowdowns. This is how data can improve theorizing, by showing expectations requiring second thoughts and additions.

Economic theory is not a theory in the sense of natural sciences, but it is like a language. Languages enable understanding, but are never falsified: they are enriched by interaction with new problems.

The distinction between theory (a mere inquiry into logical structure with limited predicted power) and history (the understanding of complexity) will sooner or later become necessary also among the “mainstream”.

Nessun commento:

Posta un commento