Monday, October 14, 2013

GMM and the structural vs. reduced form debate

stolen from Wikipedia
Among this year's economics Nobel laureates is Lars Hansen, who pioneered the generalized method of moments approach for estimating economic models. Tyler Cowen gives a nice summary of Hansen's contributions to economics and finance, and Alex Tabarrok attempts to explain GMM to the layman.

I think that one underappreciated aspect of GMM is that is illustrates how silly the structural vs. reduced form debate is. GMM allows us to estimate equations derived from structural models at low computational cost and with minimal assumptions. We don't have to assume that the entire structural model is "true;" we only have to assume that the functional form of the estimated equation is meaningful relative to the parameters being estimated.

Vocal opponents of structural approaches seem to think that estimating a structural model requires far more heroic assumptions than estimating the typical linear model used in reduced form work. GMM shows that this is not necessarily the case.

Ultimately the difference between reduced form work and a lot of structural estimation work boils down to functional form. Structural approaches choose the functional form of the estimated equation based on a derivation from a structural model. Reduced form approaches choose based on treatment effect concerns, and they typically choose from within the universe of linear or nearly linear functional forms. They both fit the equation to data by minimizing error. I fail to see how one approach is more realistic than the other. It's not immediately obvious that a linear model of anything is a more or less accurate representation of the real world than any other functional form; rather, it likely depends on the research question and the items being measured. It's nice to approach the discipline with a variety of tools so we can find the right tool for each job.

Monday, October 7, 2013

The recent decline in employment dynamics

That is the title of an excellent paper by Henry Hyatt and James Spletzer, recently published here (ungated!). The abstract reads,

We document and attempt to explain the recent decline in employment dynamics in the U.S. We have four major empirical findings. First, each measure exhibits a “stair step” pattern, with the declines concentrated in recessions and little increase during subsequent expansions. Second, changes in the composition of workers and businesses can explain only a small amount of the decline. Third, any explanation for the decline in job creation and job destruction will account for no more than one-third of the decline in hires and separations. Fourth, the decline in hires and separations is driven by the disappearance of short-duration jobs.

A major contribution of this paper is the documentation of decline in measures of dynamics from four different datasets:

  • LEHD: A quarterly longitudinal household dataset based on state unemployment insurance records and the Quarterly Census of Employment and Wages
  • BED: The quarterly establishment-level administrative dataset based on the BLS business register (I used this dataset here)
  • JOLTS: The monthly BLS establishment survey data on hires and separations
  • CPS: The monthly BLS household survey on which monthly headline unemployment rate estimates are based

I've discussed the decline in job creation and job destruction rates many times. This paper also looks at hires and separations and job-to-job flows (see chart, which is Figure 3 in the paper). Note the stair-step pattern.

click for larger image

Composition effects do not explain the declines. The aging and increasingly educated workforce and the aging of the firm distribution have some explanatory power but are insufficient. In a related working paper, coauthors and I conduct a thorough investigation of job creation and destruction in the BDS (which I've described here and elsewhere). We likewise find insufficient explanatory power from composition effects. Factors such as the evolving sectoral and regional composition of employment or the changing racial makeup of the population do not have explanatory power and, if anything, act in the "wrong" direction. In short: the composition of the US economy has changed, but these changes cannot explain the decline in dynamics.

Further, H&S provide evidence that "the explanation for the decline in hires and separations will be different than the explanation for the decline in job creation and job destruction." So the puzzle is multifaceted.

What could be causing these trends? The authors discuss a few ideas, none of which provide a smoking gun:

  • Changes in employment adjustment costs, either technological or legal, could conceivably work in either direction. These could include the decline in unionization rates, discharge laws, cost of vacancy postings and job search, and others.
  • Changes in the nature of job matches could extend employment duration and have similar effects on other dynamics measures.
  • Changes in the level of uncertainty about match quality, productivity, future profits, etc., could matter. But this can be a tricky idea, and developing a theory of uncertainty that can explain both the secular and cyclical path of dynamics measures would be difficult.
  • Changes in the production process, such as an "outsourcing" of volatility to international labor markets or the rise of temporary help agencies, may have some explanatory power.
  • Job- and house-lock, in principle, could drive declining dynamics, but the authors note that evidence linking the two adequately has yet to be found.
Each of these potential explanations represent avenues for future research. It is nice to see that this topic is getting some attention, as it may provide clues to the arrival of "jobless recoveries" and other labor market anomalies. Ben Casselman at the Wall Street Journal is more aware of the issue than any other journalist and has a nice new note on the topic (gated; try Google News).