There’s a widespread quick-and-dirty version of the relationship between theory and empiricism in economics, which is that one first creates theories, tests those theories with data, and then iterates with new theories and empirical tests. But in the 21st century, I’m not sure anyone really believes this. It’s well-known that you can create an internally consistent theory to reach pretty much any conclusion you want, as long as you tinker with the underlying assumptions. Moreover, it’s well-known that when doing empirical work, one can try out a bunch of different statistical tests until you find one that reaches the conclusion you want. To make matters worse, there’s no particular reason to believe that if some particular economic theory is validated by some particular empirical estimate in one context that it will also hold true in all other times and places. These concerns prove the case that a social science is not a natural science, but it would be as severe overreaction to hype them up into a claim that social sciences can’t lead to meaningful knowledge.
In my new book, I argue that economics is not a science. I say that we deal primarily in non-falsifiable frameworks of interpretation, rather than non-falsifiable hypotheses. Taylor’s comments speak to some of the reasons that this is the case.
The problem becomes how to evaluate competing frameworks if scientific epistemology (i.e., falsificationism) does not apply. Taylor is discussing essays by Harrod and Keynes, who, each in his own way, seems to argue for an “I’ll know it when I see it” approach to evaluation. However, I think we should try harder to spell out the criteria that are most helpful.
Does this apply equally to macro and micro? I would tend to trust a well defined micro study more than some macro model that tries to tie total economic growth to one per variable
Micro deals with things like “people do things in their interest” and maybe we test for some biased. Macro deals with “house prices are going up, let’s assume they go up forever and then what happens.”
What are the axioms of macro? What keeps them from changing?
Maybe one can draw a distinction between theory falsification and prediction falsification. The problem with macroeconomic theories is that they are always salvageable no matter what happens. If employment following a stimulus was higher than the model predicted for both the stimulated and unstimulated scenarios, then proponents can always say, “other stuff happened too, it still would have been worse without the stimulus.” Like climate change or astronomical scale gravity, one can only observe and not perform double-blind, controlled experiments, and there is always the possibility of adding in another degree of freedom to ‘excuse’ inconsistencies.
But it’s logically impossible to claim that a theory is a better guide to real world policy vs rival theories without that claim being reducible to competing predictions which can be compared for accuracy after the fact, which can accumulate track records, and which can be bet against each other in a prediction market.
The details of those markets, like liquidity, volatility, bet size, and magnitude of participation in the market, the necessity for subsidy (and whether the usual loud prognosticators put their own money where their mouths are) can be used to infer a measure of confidence or humility one sound have about the nature of the whole field of inquiry. That is, it is the market assessment for how valid theories in the field can be, and how much any particular one can be trusted enough to form a justifiable basis for policy intervention.
For example, if no good theory of future NGDP is actually achievable, then no amount of subsidy can make Sumner’s prediction market work well except in the manner of self-validating expectations.
I don’t see this as a problem. Theories are frameworks of interpretation. Hypotheses are generated and tested. Theories then have to adapt to those facts. It is when the adaptation becomes so complex, the assumptions so ridiculous, theories can no longer adapt or fail to adapt, and a better theory comes along that progress is made. If a theory is incapable of generating testable hypotheses, it isn’t a theory at all, only a just so story. A just so story may be developed into a theory, but without testable hypotheses it will never be. The testing may only be indicative rather than deductive, suggestive rather than conclusive, retrospective rather than prospective, it may never be complete, but it must be done, or it is only speculative hot air.
Of course when people can reject objective science when it fails to conform to their desires, what hope is there for any knowledge? Even more so in areas less definitive.
Ludwig Von Mises chuckles knowingly from the grave, saying “It took you long enough to get here, but welcome.”
I’m still not sure they’re at where Mises was, (rather, I’m pretty sure they’re not) but that is where they need to end up, yeah.
Yes, everyone just needs to get religion. What progress!
I can deliberately misconstrue people and be snarky about it, too, Lord.