05-19-17

Why the Last Ten Years of Data Mean that Macro Must Change

In this post, I discuss the radical consequences of the past ten years of data for macroeconomic models.

There were a host of quantitative macroeconomic models available at the end of 2006 – in the private sector, in the public sector, and in the academe. Pick any of those models that you like. Now ask yourself: given that model, what probability did it assign (as of 2006) to either of these events:

• The employment-population ratio falls by over 4 full percentage points, and so is below 58.5%, three years later.
• Cumulative per capita real GDP growth is less than 6% over the next decade

I don’t have a detailed knowledge of all models that were available in 2006. Despite that, I’m going to assert that virtually all of the 2006 models would have assigned a probability of essentially zero to either of these events and, a fortiori, would have assigned a probability of essentially zero to their union. (I’d certainly be interested in seeing an explicit quantitative (as opposed to verbal) counterexample to my assertion.)

Assuming my claim is true, how should a macroeconomist respond to this observation? I’ve seen three basic approaches in the academe.

• Treat the past ten years of data as “outliers” driven by an unusual one-time event and throw them out (by using data up through 2007 or 2008). It’s interesting to note that this continues to be how many/most macroeconomists treat the Great Depression.
• Re-estimate one’s favored model on the extended sample that includes the past ten years of data. This response sees the past ten years of data as an additional source of information about the parameters or shock processes of interest.
• Retrofit one’s favored model with a new feature or two or three that make A (and maybe B) somewhat more likely, conditional on the what was known in 2006.

None of these responses strike me as being at all commensurate with the evidence. An event that was deemed essentially impossible, based on our best collective understanding of the macro-economy in 2006, has occurred. According to any model selection principle that I know, that’s very bad news for the 2006 universe of macro models. The evidence is telling us – in A VERY LOUD VOICE - that macroeconomists (academic, public, and private sector) need to undertake a fundamental reconstruction of what we took to be true in 2006.

How should that reconstruction begin to proceed? For myself, I’m pursuing modelling approaches in which:

• The failure or inability of policymakers to react sufficiently strongly to changes in financial and economic conditions can cause considerable short-term damage to the health of the economy.
• And that short-term damage (OR the putative cures for that short-term damage) have consequences for the long-run behavior of the economy

But I’m sure that there are many other ways to move forward. The only approach that seems completely unacceptable is to cling to the models of 2006 despite their dramatic empirical failure.

N. Kocherlakota

Rochester, NY

May 19, 2017

CORRECTION: A couple of people have pointed out to me that my proposed research agenda doesn't really address the problem raised in my post. That's completely spot on. A complete agenda needs to incorporate (at least) one more component:

• How do the short-term and long-term risks to the macroeconomy evolve in response to financial and economic conditions?

I'm very interested in this issue, but I have to admit that I haven't done much on it yet.

N. Kocherlakota

Rochester, NY

May 20, 2017

Kocherlakota Main Page