Productivity and Models

I wrote a post on BloombergView yesterday, arguing that there are good reasons to think that demand stimulus would lead to an increase in labor productivity.  Here, I want to dig a little more into the numbers and the issues.  They are, I think, an interesting illustration of the key intellectual problem that we face in macro.

 I use two data sets: the CBO forecasts of potential GDP from August 2007 and the 2015 data from the most recent total factor productivity (TFP) data release from the San Francisco Fed.[1]  Both center on the non-farm business sector. In 2015:

  • Real output was about 15% lower than forecast.
  • Labor input was about 6% lower than forecast.
  • Capital input was about 18% lower than forecast.
  • Total factor productivity was 5% lower than forecast.

(By the way, I found these numbers to be completely shocking – but perhaps that’s just me.)

I’ll now sketch a couple of stories (models?) that are consistent with these facts, but have differing implications for policy and future growth. 

The Exogenous Story 

Suppose we view:

  • the fall in labor input as primarily due to the CBO’s missing key (exogenous) birth-cohort effects in labor supply[2]
  •  the fall in TFP is due to a permanent shock to an exogenous process that is a near unit root around a trend.
  • TFP is expected to grow less rapidly than in 2007, leading to a fall in potential growth of about a percentage point. 

If you put all these together, you would conclude that capital is not much below its BGP.  This is the “it’s about as good as it gets” view of the economy.   

The Endogenous Story

In this story, I think of labor input as endogenous, and ignore the cohort effects mentioned above.   I think of TFP as being a stock variable, influenced by investment in R&D, in implementation of ideas, and perhaps public capital. (See NBER WP 22005 for a model along these lines.Under this interpretation, TFP has a BGP, just like capital.  

The heart of this story is that, because monetary policy has been unduly tight over the past few years, there has been underinvestment in both innovation and implementation.   Because fiscal policy has been too tight, we have too little public capital. The result is that TFP is now 5% below its BGP. 

At the same time, physical investment has been too low (because of overly tight monetary policy).  As a result, physical capital is also well below its BGP.

Conditional on these state variables, we might well be close to full employment.  Suppose capital is 18% below its BGP and TFP is 5% below its BGP.   Ignoring income effects on labor supply, and assuming a Frisch elasticity of about 1/2, these changes in TFP and capital would imply that (potential) employment should be about 5-6% lower than was expected in 2007. But, even though we’re close to full employment, there’s a lot of room for super-normal growth. Both capital and TFP are well below their BGPs.  The full-employment growth rate is going to be well above its long-run level for several years.  We can’t conclude the economy is overheating just because it is growing quickly.

Of course, even given these state variables, we might not be close to full employment.  Consumption is unexpectedly low relative to the CBO’s 2007 forecast.  Income effects on labor supply would push up on the willingness of people to work.  This would imply that employment is also below potential (even given the low values of TFP and capital).  

I should note too that in many models of endogenous TFP, it has an external component or a component that depends on public capital.  That aspect would imply that the government should facilitate even faster TFP growth. 

The Intellectual Problem

As my writings on BloombergView suggest, I’m a fan of the second story.  But let’s forget about that issue for the moment.   We have two stories.  In one, much is exogenous and beyond the control of policy.  In the second, much is endogenous and shapeable by policy.  The more interesting intellectual question is:

What advice should macroeconomists be giving policymakers, given these two (of many?) stories and given the limited data available to distinguish them? 

In my view, we’d arrive at a much better macroeconomics if the field were much more centered around these kinds of questions.

You might be waiting for me to provide an answer to the last emboldened question.  I don’t have one yet.  But here are some very tentative thoughts. 

First, I'd be skeptical of the following approach. 

  • Figure out which model fits the data better. 
  • Use that model to figure out the best policy. 
I'd be skeptical of trying to rule out one model in favor of the other.

Second, I’d try to use a range of data/methods to update my prior beliefs about the two models, rather than use one data set and/or one method. But I’d be surprised if that process ever resulted in eliminating either model completely. 

Third, given this residual uncertainty, we need to take into account the model-based gains/losses associated with different policy choices.   Suppose that, in model 1, a lot of choices end up being largely irrelevant for outcomes but in model 2, those same choices matter a lot.  Surely, we should put more weight on model 2?

Anyways, these thoughts are highly tentative.  The more important point is that macro should be much more about confronting policy questions in the face of limited data and model uncertainty.

N. Kocherlakota
University of Rochester
August 12, 2016



[1] A couple of side-comments: I’m using a measure of TFP that’s unadjusted for underutilization of resources.  According to the San Francisco Fed data, about a percentage point of the TFP shortfall is due to underutilization.   I’ve folded the SF Fed’s measure of labor quality into TFP. Neither of these issues matter all that much for what I’m going to say.

[2] See https://www.federalreserve.gov/pubs/feds/2014/201464/201464pap.pdf for interesting recent work along these lines.