What do models tell us?

Josh Hendrickson has a new post that defends the use of models that might in some respects be viewed as “unrealistic”. I agree with his general point about models, and also his specific defense of models that assume perfect competition. But I have a few reservations about some of his examples:

Ricardian Equivalence holds that governments should be indifferent between generating revenue from taxes or new debt issuances. This is a benchmark. The Modigliani-Miller Theorem states that the value of the firm does not depend on whether it is financing with debt or equity. Again, this is a benchmark. Regardless of what one thinks about the empirical validity of these claims, they provide useful benchmarks in the sense that they give us an understanding of when these claims are true and how to test them. By providing a benchmark for comparison, they help us to better understand the world.

With all that being said, a world without “frictions” is not always the correct counterfactual.

Taken as a whole, this statement is quite reasonable.  But I would slightly take issue with the first sentence, which is likely to mislead some readers.  Ricardian Equivalence doesn’t actually tell the government how it “should” feel about the issue of debt vs. taxes, even if Ricardian Equivalence is true.  Rather it says something like the following:

If the government believes that debt issuance is less efficient than tax financed spending because people don’t account for future tax liabilities, that belief will not be accurate if people do account for future tax liabilities.

But even if people do anticipate the future tax burden created by the national debt, heavy public borrowing may still be less efficient than tax-financed spending because taxes are distortionary, and hence tax rates should be smoothed over time.

I happen to believe Ricardian Equivalence is roughly true, but I still don’t believe the government should be indifferent between taxes and borrowing.  Similarly, I believe that rational expectations is roughly true, and yet also believe that monetary shocks have real effects due to sticky wages.  I believe that the Coase Theorem is true, but also believe that the allocation of resources depends on how legal liability is assigned (due to transactions costs).  Models generally don’t tell us what we should believe about a given issue; rather they address one aspect of highly complex problems.

Here’s Hendrickson on real business cycle theory:

Since the RBC model has competitive and complete markets, the inefficiency of business cycles can be measured by using the RBC as a benchmark. In addition, if your model does not add much insight relative to the RBC model, how valuable can it be?

[As an aside, I agree with Bennett McCallum that either the term ‘real business cycle model’ means a model where business cycles are not caused by nominal shocks interacting with sticky wages/prices, or else the term is meaningless.  There is nothing “real” about a model where nominal shocks cause business cycles.]

Do RBC models provide a useful benchmark for judging inefficiency?  Consider the following analogy:  “A model where there is no gravity provides a useful benchmark for airline industry inefficiency in a world with gravity.”  It is certainly true that airlines could be more fuel efficient in a world with no gravity, but it’s equally true that they have no way to make that happen.  I don’t believe that gravity-free models tell us much of value about airline industry efficiency.

In my view, the concept of efficiency is most useful at a policy counterfactual.  Thus monetary policy A is inefficient if monetary policy B or C produces a better outcome in terms of some plausible metric such as utility or GDP or consumption.  (I do understand that macro outcomes are hard to measure (especially utility), but unless we have some ability to measure outcomes then no one could claim that South Korea is more successful than North Korea.  I’m not that pessimistic about our knowledge of the world.)

In my view, you don’t measure inefficiency by comparing a sticky price model featuring nominal shocks against a flexible price RBC model, rather you measure efficiency by comparing two different types of monetary policies in a realistic model with sticky prices.

That’s not to say that there are not aspects of RBC models that are useful, and indeed some of those innovations might provide insights into thinking about what sort of fluctuation in GDP would be optimal.  But I don’t believe you can say anything about policy efficiency unless you first embed those RBC insights (on things like productivity shocks) into a sticky wage/price model, and then compare policy alternatives with that model.  I view sticky prices as 90% a given, much like gravity.  (The other 10% is things like minimum wage laws, which can be impacted by policy.)

PS.  Just to be clear, I agree with Hendrickson on the more important issues in his post.  My support for University of Chicago-style perfect competition models definitely puts me on “team Hendrickson”, especially when I consider the direction the broader profession is moving.


Read More