An Industrial Organization Perspective on Productivity

An Industrial Organization Perspective on Productivity. Jan De Loecker & Chad Syverson. NBER Working Paper 29229, Sep 2021. https://www.nber.org/papers/w29229

Abstract: This chapter overviews productivity research from an industrial organization perspective. We focus on what is known and what still needs to be learned about the productivity levels and dynamics of individual producers, but also how these interact through markets and industries to influence productivity aggregates. We overview productivity concepts, facts, data, measurement, analysis, and open questions.

---

This chapter focuses on the implications and applications of productivity analysis within Industrial Organization. There is a long tradition in IO of studying productivityrelated topics like allocative eciency, technological change, regulatory effects, cost effciencies in merger analysis, and returns to scale, to name a few. The term productivity is, however, more often than not used in a fairly loose sense, usually referring to a measure of performance. In this chapter we make an explicit distinction between productivity in a strict production eciency sense, on the one hand, and performance on the other. The study of production eciency, the rate at which a producer can convert a bundle of inputs into a unit of output, is in essence about the technical relationship between output and inputs. Performance captures a variety of measures, but as this chapter will highlight, it is intimately related to eciency. However, the distinction can be crucial when analyzing the very topics listed above.

1.1 Background and Focus

While having a long history in economics, the past few decades have seen the productivity analysis of individual producers and the corresponding industry- and economy-wide aggregates become a central topic in both academic and policy circles. This renewed interest has paralleled at least three main developments.

First, over the last two decades the access to micro data has exploded. At the turn of the century, only a few large-scale producer-level datasets existed, with limited access to researchers. In contrast, the current list of countries for which micro census data is available (in manufacturing, at least) contains a rather large share of the world. In addition, private data providers have emerged offering comprehensive accounting data capturing typical variables used in productivity analysis, through the reporting of balance sheets and income and loss statements.

Second, accompanying the increased access to micro data has been a renewed interest in the estimation and identification of production functions. These are of course key objects of interest for most productivity analyses, both for their own sake as well as supplyside inputs into equilibrium analysis. This research has focused on obtaining reliable productivity measures for sets of producers when, as is the case, the researcher cannot directly observe productivity but producers can. This leads to two well-known biases, the simultaneity and selection biases, that researchers must face.

Third is the prominent role of productivity analysis in forming and executing economic policy. While policymakers still mostly focus on industry- or economy-wide aggregates, there is increasing recognition that this analysis is often most informative when built from the ground up using micro data. The melding of data, methods, and economically oriented policy analysis has spurred informative interactions among microeconomists, macroeconomists, and policymakers that have created many insights into productivity.  Our intent in this chapter is to organize and review the intellectual underlayment of this burgeoning literature. There are many facets. Our coverage includes key conceptual issues, facts about micro-level productivity, models of markets with heterogeneous productivity producers, measurement and data, productivity estimation, the positive and normative implications of the static and dynamic allocation of activity across heterogeneous producers, and an overview of what we expect to be active areas of work in the near future. We do this while taking stock of several decades of empirical work on productivity using micro data. We will unavoidably miss certain dimensions, and simple space constraints mean we cannot do justice to many contributions and insights from this extensive literature. We do hope, however, to offer a structured view on the field of productivity and how Industrial Organization scholars have contributed.

1.2 Productivity Conceptualized

Productivity is conceptualized in a number of related ways. All productivity metrics in one form or another measure how much output producers obtain from a given set of inputs. As such they are measures of the ecacy of the supply side of the economy (though \ecacy" need not always be synonymous with social welfare).  An interpretation of productivity as an economic primitive is as a factor-neutral (aka Hicks-neutral) shifter of the production function.1 Consider the general production function Q = Omega * F(.), where Q is output and F() is a function of observable inputs capital such as capital, labor and intermediate inputs.  Omega is productivity, the factor-neutral shifter.  It reflects variations in output not explained by shifts in the observable inputs that act through F(). A higher value of Omega implies the producer will obtain more output from a given set of inputs. That is, it denotes a shift in the production function's isoquants down and to the left.

A second conceptualization is empirical: productivity as a ratio of output to inputs.  This is tightly related to the production-function-shifter interpretation above. This can be seen by isolating productivity from the production function: Omega = Q / F(.) . A is clearly an output-to-input ratio. Here, where output is divided by a combination of observable inputs, the productivity concept is named total factor productivity (TFP) (it is also sometimes called multifactor productivity, MFP). There are also single-factor productivity measures, where output is divided by the amount of a single input, most commonly labor; i.e., labor productivity. Because single-factor productivity measures can be affected not just by shifts in TFP but factor intensity decisions as well, TFP measures are often conceptually preferable. On the other hand, labor productivity is often easier to measure than TFP.

A third conceptualization is of productivity as a shifter of the producer's cost curve.  Higher productivity shifts down the cost curve; that is, at the producer's cost-minimizing combination of inputs, its total cost of producing a given quantity is lower, the higher is its TFP level. This productivity conceptualization is related to the other two because the cost function is the value function of the producer's cost minimization problem, which takes the production function as its constraint. This cost function shifts down when productivity rises.

Because it plays such an important role in the producer's production technology, measuring productivity and its influence on outcomes is the subject of an enormous literature.

This work has found in many disparate settings that, as an empirical matter, productivity is hugely important in explaining the fortunes of producers, their workers, their suppliers, and their customers. We survey work on both the measurement and effects of productivity below.