Job Creation and Destruction


Prakash Loungani and Bharat Trehan

FRBSF Economic Letter 1997-13 | May 2, 1997

National labor market conditions are a central concern for both economists and policymakers. Traditionally, macroeconomists have not paid attention to patterns of job creation and job destruction but have tried to understand the labor market in terms of the behavior of economy-wide aggregates, such as interest rates or aggregate wage levels.

National labor market conditions are a central concern for both economists and policymakers. Traditionally, macroeconomists have not paid attention to patterns of job creation and job destruction but have tried to understand the labor market in terms of the behavior of economy-wide aggregates, such as interest rates or aggregate wage levels. However, in recent years, macroeconomists have begun to pay more attention to developments at the micro level. In this Letter we discuss some recent work by a group of economists who study plant level employment in order to understand macroeconomic developments.

What the microlevel data show

Work by Davis, Haltiwanger, and Schuh (1996, henceforth DHS) has been central to the surge of interest in this area. Their empirical analysis is based on data for manufacturing plants covering the period from the early 1970s to the mid-1980s. Defining employment increases at new and growing plants as job creation, and decreases at dying and shrinking plants as job destruction, they point out a number of empirical regularities.

A striking feature is that the data are marked by a high rate of job creation and destruction. On average, close to one out of ten manufacturing jobs disappears in a given year, while the rate of new job creation is slightly lower. These changes are quite persistent: a year later, nearly seven out of ten newly created jobs were still in existence, and about eight in ten lost jobs were still lost. In addition, job creation and destruction tends to be concentrated at plants that experience large changes in employment (those associated with plant shutdowns and startups, for instance). Another finding is that job destruction varies more noticeably over the cycle than job creation. Figure 1 shows that job destruction tends to increase sharply during a recession and then fall back, while job creation does not move as much.

Some questions have been raised about these results. For instance, some economists have cautioned against relying on data for a single sector of the economy — especially manufacturing, where employment has been shrinking so noticeably. Further, the data cover a relatively limited span (the 1970s and the 1980s), and it is possible that the recessions of this period differ fundamentally from previous (or subsequent) recessions in terms of restructuring and reallocation. Though the issue is not settled yet, some of the DHS findings have been replicated elsewhere. For instance, Blanchard and Diamond (1990) rely mainly on data from the Current Population Survey, which is not restricted to manufacturing alone, and they confirm the finding about the relative volatility of job creation and destruction. For example, they find that “…booms are times of low job destruction rather than high job creation” (p. 87); similar patterns have been discovered in data for foreign countries as well.

Proponents of the use of these data suggest they offer two advantages. First, they provide a means of distinguishing among alternative models of the economy. That is, they provide details that acceptable models of the macroeconomy must match. Second, attempts to understand these data are likely to lead to new ways to think about important macroeconomic phenomena such as business cycles.

New theories on sector-specific shocks and the business cycle

These data contradict the simple characterization of recessions as a period when workers are laid off temporarily and then recalled to their jobs once demand has recovered. Blanchard and Diamond point out that these findings also rule out the Schumpeterian view of business cycles, in which expansions are periods when new inventions are implemented, leading to high job creation. Instead, the evidence suggests that recessions are times of restructuring. Based on this understanding, researchers have recently proposed a number of different models of what happens over the course of a business cycle.

For example, DHS suggest that the available evidence is consistent with two theories, which differ in the role assigned to “sector-specific” shocks as sources of business cycles. In the first, often called the sectoral shifts theory, recessions are periods that are marked by a pronounced shift, due to either “tastes or technology,” that alters expected returns across sectors, causing some of them to expand and others to contract. In subsequent periods, capital and labor have to be reallocated from the contracting sectors to the expanding ones. While beneficial for the economy as a whole in the long run, this restructuring of activity imposes short-run costs in the form of declines in economic activity. Oil shocks provide an interesting example here: while researchers typically focus on the aggregate effects of oil shocks, these shocks are also likely to cause a reallocation of jobs across sectors.

The second theory is that the economy is hit by sector-specific shocks all the time, but reallocation is concentrated at recessions, because the opportunity cost of undertaking the needed changes is low at that point. Here, the predominant sources of business cycles are assumed to be aggregate shocks, but sector-specific shocks play an important role in determining how the aggregate shocks will affect the economy. For example, a shift in monetary policy may induce a sector to carry out some restructuring that might otherwise be postponed to a later date. DHS have suggested that the pace of restructuring in the steel industry over the years from 1973 to 1988 was influenced by the stance of monetary policy over this period.

Caballero and Hammour (1994) implement some of these ideas in a model in which recessions are a time of restructuring. They motivate the model with an example from the motor vehicle industry. At the beginning of the Great Depression, a substantial segment of the industry was still based on skilled craftsmanship, even while mass-production techniques were gradually becoming more important. Over the course of the Depression, numerous plants shut down — permanently, in most cases. These shutdowns were concentrated in the older plants. Thus, the Depression led to a shakeout in the industry. In the formal model they set up, technological progress occurs at a constant rate, with the latest technology embodied in the newest firm. Technological progress pushes prices down over time, which in turn has the effect of making older plants obsolete. A decrease in demand causes prices to fall faster than usual, implying a bunching of plant shutdowns in recessions. Caballero and Hammour show that a calibrated version of their model is able to replicate the relative volatility of job destruction and creation found in the DHS data, and also does a good job of replicating the time path of the data on job destruction.

While this line of research has some interesting implications in a number of other fields as well, for reasons of space we will focus on two areas here: what these studies tell us about the way in which different kinds of shocks affect the economy, and what they tell us about the cyclical behavior of labor productivity.

Distinguishing among different kinds of shocks

Davis and Haltiwanger (1997) show that a “typical” oil price increase leads to the destruction of over 275,000 manufacturing production jobs over the first eight quarters following the increase. A very significant finding is that there is actually ‘creation’ of nearly 30,000 new jobs over this period. Over the subsequent eight quarters, the cumulative job creation continues to build to over 175,000 jobs. The fact that oil price increases raise both creation and destruction of jobs suggests that oil shocks alter the pattern of comparative advantage away from some manufacturing establishments and toward others. As an example, Davis and Haltiwanger note that the OPEC oil price shock of 1973 shifted demand away from large cars toward small, fuel-efficient cars. Since American automobile companies were, by and large, poorly situated to take advantage of this demand shift, the industry’s capacity utilization and output fell in the wake of the oil shock. However, as documented by other researchers, “a handful of plants equipped to produce small cars operated at peak capacity …” (p. 2).

Employment reacts noticeably differently to a monetary (or credit) shock. Using a variety of different measures — including the federal funds rate and the spread between long- and short-term Treasury bills — Davis and Haltiwanger show that job creation and destruction move in opposite directions in response to such shocks. When the “risk spread” (that is, the spread between the commercial paper and Treasury bill rates) goes up, for example, job creation actually falls by more than 40,000 jobs after four quarters, which is more than half the increase in job destruction over the same period. Two years after the shock, job creation is still 10,000 jobs less than would otherwise be the case.

Understanding the cyclical behavior of labor productivity

Baily, Bartelsman, and Haltiwanger (1996) (BBH hereafter) used micro data to explore the cyclical behavior of labor productivity. It has long been noted that average labor productivity declines during recessions and increases during booms. A common explanation for this finding is that it reflects “labor hoarding” by firms. During a recession, output declines, but workers are retained on the payroll since the firm expects to need them once the recession is over. As noted by BBH, an implicit part of the labor hoarding explanation is that firms believe that the decline in output will be short-lived.

The micro data do not offer much support to this explanation for procyclical productivity. It turns out that much of the decline in labor productivity during recessions in the 1970s and 1980s was in plants that were permanently downsizing. It is more likely, therefore, that the decline in productivity reflected the difficulties associated with restructuring the production process in response to a permanent reduction in the scale of operations. “Over time, if the lower level of output is sustained, productivity can recover (and perhaps ultimately increase) as the production process is adjusted” (BBH, p. 3).


The line of research pioneered by DHS already has made a number of significant contributions to our understanding of the macroeconomy. Their analysis of the data provides a basis on which to judge alternative theories about the macroeconomy. And it has spurred the development of models that provide new ways of thinking about macroeconomic phenomena such as business cycles. Finally, the issues raised by this line of research highlight implications of policy actions that are usually not considered in discussions about policy. For instance, when policymakers act to stabilize the economy, they may need to take into account the effect of their actions on the pace of reallocation in the economy. Note that this is not the same thing as saying that very high rates of job destruction are desirable per se, nor that high unemployment rates are not wasteful. In any event, this literature has not yet reached the point of providing clear guidelines for policy; not all the implications of this way of looking at the economy have been worked out, nor the various debates resolved. Regardless of how various issues are resolved, however, it seems fair to say that this work has enriched our understanding of how the economy functions.

Prakash Loungani
Board of Governors

Bharat Trehan
Research Officer


Baily, Martin Neil, Eric J. Bartelsman, and John Haltiwanger. 1996. “Labor Productivity: Structural Change and Cyclical Dynamics.” NBER Working Paper No. 5503 (March).

Blanchard, Olivier Jean, and Peter Diamond. 1990. “The Cyclical Behavior of the Gross Flows of U.S. Workers.” Brookings Papers on Economic Activity 2, pp. 85-143.

Caballero, Ricardo J., and Mohamad L. Hammour. 1994. “The Cleansing Effect of Recessions.” American Economic Review (December) pp. 1350-1368.

Davis, Steven J., and John Haltiwanger. 1997. “Sectoral Job Creation and Destruction Responses to Oil Price Shocks and Other Shocks.” Mimeo. University of Chicago.

_____, _____, and Scott Schuh. 1996. Job Creation and Destruction. Cambridge: MIT Press.

Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Anita Todd and Karen Barnes. Permission to reprint portions of articles or whole articles must be obtained in writing. Please send editorial comments and requests for reprint permission to