«BEYOND SHOCKS: WHAT CAUSES BUSINESS CYCLES? AN OVERVIEW Jeffrey C. Fuhrer and Scott Schuh* In the summer of 1997, when the Federal Reserve Bank of ...»
The result of Basu’s evaluation is quite discouraging for state-of-theart macroeconomic models. He ﬁnds that neither the RBC nor the sticky price model generally ﬁts the data very well. The RBC model, in particular, does not match the dynamic properties of the data, and it 20 Jeffrey C. Fuhrer and Scott Schuh cannot reproduce the essentially zero correlation that exists between the BFK technological change and output or the negative correlation between factor inputs and output. These models also fail to reﬂect the generally sluggish response of output changes in the economy. Basu reports that the sticky price model is qualitatively better because it approximately reproduces these two correlations, although it does not do so well. The prognosis for these models becomes even bleaker when he evaluates the models with both technological change and various speciﬁcations of monetary policy.
Basu concludes that the deﬁning cyclical feature of technological change is a short-run reduction in inputs and factor utilization, and that business cycle models face the challenge of reproducing that feature. At present, standard RBC and sticky price models cannot do the job, and variable factor utilization does not impart enough rigidity to generate sufﬁcient sluggishness. He projects that the sticky-price models, modiﬁed to include other sources of rigidities, “show some promise of being able to match the data, but clearly have a long way to go.” Mark Bils questions whether Basu’s technology measure adjusts too much for the positive correlation between factor utilization and output.
He hypothesizes that the proportions of capital and labor used in production are likely to be ﬁxed in the very short run. Thus, when capital utilization rises slightly, labor hours will rise in equal proportion. If so, total factor productivity should be positively correlated with output but labor productivity should be approximately uncorrelated with output.
Bils ﬁnds exactly these correlations in data on detailed manufacturing industries. Because the BFK methodology infers movements in capital utilization from movements in materials prices, and because materials prices are more positively correlated with output than labor costs, Bils believes the BFK measure makes capital utilization more positively correlated with output than labor utilization is.
Other aspects of Basu’s methodology make Bils skeptical of the results. He doubts that labor quality (effort) is positively correlated with output, as in the BFK measure, because there is evidence that workers hired during expansions are paid less and therefore of lower quality.
Moreover, he thinks the relationship between effort and hours will vary depending on the stickiness of wages and the type of shock. Bils also argues that factor utilization will vary more if shocks are transitory rather than permanent. Basu’s methodology relies more on variables associated with transitory shocks, so it may yield estimates of utilization that are too positively correlated with output.
Finally, Bils assesses the plausibility of price stickiness in two empirical exercises. One exercise is based on the theory that if prices are sticky, then ﬁrms with signiﬁcant inventory holdings should be less likely to reduce inputs and output when technology increases, because they can
BEYOND SHOCKS: WHAT CAUSES BUSINESS CYCLES? AN OVERVIEW 21inventory unsold output. He reports evidence that “labor hours are much less likely to decline for industries that hold signiﬁcant inventories,” but points out that this evidence does not conclusively determine the actual ﬂexibility of prices. So in a second exercise he provides more direct evidence from models of relative prices. Prices are signiﬁcantly negatively correlated with current total factor and labor productivity but not with past productivity, a relation Bils interprets as evidence that prices are not sticky.
Thomas Cooley is also cautious about interpreting Basu’s results as evidence against the idea that technological change is an important source of business cycle ﬂuctuations. Like Bils, Cooley has reservations about the methodology underlying the BFK technology measure, although he embraces Basu’s ﬁnding that ﬁrms do not enjoy market power from technological advantages in production. In particular, he notes that the correlation of the BFK technology measure with output is sensitive to the exact form of the econometric methodology used to construct the measure and to the identifying assumptions of the modeling framework.
However, granting the validity of Basu’s results, Cooley directs his critique at the logic of Basu’s inferences about the implications for macroeconomic models. First, he questions Basu’s conclusion that the results necessarily rule out RBC-type models. He argues that RBC models no longer rely on artiﬁcially sluggish technology shocks to obtain sluggish output responses. Sluggishness can arise from factor utilization as well as ﬁnancial market imperfections, differences among ﬁrms, and other features. As for the RBC model’s inability to generate a negative correlation between technology and factor inputs, he suspects that this result is not robust.
Cooley also questions whether the evidence should lead one to conclude that prices are sticky. Basu provides no direct evidence of sticky prices, and economic theory does not make clear predictions about the direction in which capital and labor should respond to technology changes. The response will depend, among other things, on the nature of the technology change, market structure, and the sensitivity of demand to prices. This point calls into question Basu’s assertion that he does not need to consider the behavior of proﬁts and product markets.
Cooley thinks Basu’s results suggest that technological change is embodied in new capital investment—a characteristic absent from the BFK methodology. With technology embodied in capital, the short-run responses of output and factor inputs to technological change are different from those of a standard RBC model and are capable of yielding the patterns Basu ﬁnds in the data. Moreover, in this case the nature of depreciation matters for interpreting the effects of cyclical factor utilization.
22 Jeffrey C. Fuhrer and Scott Schuh
JOB REALLOCATION BUSINESS CYCLEAND THE Scott Schuh and Robert Triest investigate the idea that business cycles might be caused by the shufﬂing of jobs as ﬁrms restructure the way they do business. New data produced during the past decade show that ﬁrms are continuously changing. Some expand and create jobs while others contract and destroy jobs. The pace of change is rapid; one in 10 jobs is newly created and one in 10 jobs newly destroyed in manufacturing each year. The sources of these ups and downs of particular ﬁrms include product demand and innovation, prices and wages, regional economic conditions, technological change, and other factors idiosyncratic to each ﬁrm, rather than factors common across all ﬁrms. Job creation and destruction together represent job reallocation, a measure of job turnover or churning in the economy.
Traditionally, macroeconomists looking at the labor market have ignored job reallocation and have focused solely on total employment growth (or the total unemployment rate). However, Schuh and Triest point out that a given rate of employment growth can occur with either low or high rates of job reallocation. More important, the intensity of job reallocation has signiﬁcant consequences for unemployment, wage growth, and productivity growth.
For example, if changes alter the desired distribution of jobs across ﬁrms, industries, and regions, job reallocation must intensify to keep productive efﬁciency high. More intense reallocation usually means higher job destruction that forces many workers into unemployment.
These unemployed workers lose any skills they had that were unique to their previous job (such as knowledge of ﬁrm operating procedures), have a hard time ﬁnding a comparable new job, and stay unemployed longer. Eventually they may have to accept a job entailing sizable reductions in their wages. Such issues are linked inherently to the determination of aggregate unemployment, wage growth, and productivity.
Schuh and Triest point out that job reallocation and the pace of restructuring rise markedly during recessions. Traditional macroeconomic models cannot explain why because they do not incorporate the phenomenon of job reallocation. But in light of the potentially negative economic consequences of job reallocation, it is important to know whether an identiﬁable connection exists between reallocation and business cycles, and whether the correlation between them is of no consequence and can continue to be ignored.
Schuh and Triest ask the following fundamental question: Does job reallocation cause business cycles, or do business cycles cause job reallocation? Evidence on job reallocation has sparked an interest in building theoretical models capable of explaining the observed patterns in the data, and they classify these theories into two types. One type stresses the role of factors that primarily determine the desired allocation
BEYOND SHOCKS: WHAT CAUSES BUSINESS CYCLES? AN OVERVIEW 23of economic resources, such as workers, across ﬁrms. The other type stresses the role of aggregate factors, such as monetary policy, that primarily determine the overall level of economic activity. Both types of theories aim to explain why job reallocation rises during recessions. Yet both types of theories tend to rely on vaguely deﬁned aggregate and allocative “shocks” rather than observable variables.
Schuh and Triest argue that these theories do not and cannot answer their fundamental question, for two reasons. First, although the two-way classiﬁcation of factors may be conceptually sensible, in practice the deﬁnitions of allocative and aggregate factors become hopelessly muddled. Second, these theories have little to say about what causes business cycles—that is, why they occur— because they focus more on how they occur.
Schuh and Triest present results from three empirical exercises that extend research by Schuh with Steven Davis and John Haltiwanger on job creation, destruction, and reallocation (henceforth referred to as DHS).
One exercise analyzes the behavior of job reallocation during the 1990s using newly available data. A second exercise attempts to learn what kinds of plants destroy and reallocate jobs and how, in hope of discovering clues about the causes of recessions. The third exercise looks for evidence of causal relationships between job reallocation, the fundamental determinants of reallocation, and the business cycle. Each of these exercises uses data from the U.S. Bureau of the Census on individual manufacturing plants (the Longitudinal Research Database (LRD)).
The new data show that the 1990-91 recession was much less severe in manufacturing than preceding recessions, as evidenced by a relatively modest decline in employment. Nevertheless, job destruction and job reallocation both increased in a manner similar to that in previous recessions. The ensuing expansion was unusual in that job destruction and reallocation remained above average, rather than declining quickly after the recession. In addition, job creation experienced two large surges that were not preceded by surges in job destruction, as creation surges typically are. The authors interpret these surges as evidence of favorable allocative shocks, in contrast to the unfavorable allocative shocks of the 1970s and 1980s.
Regarding the nature of job creation and destruction, Schuh and Triest take a deeper look at two areas: (1) the magnitude, permanence, concentration, and cyclicality of job ﬂows; and (2) the differences in job ﬂows between larger, older, and higher-wage plants (henceforth, simply “large”) and smaller, younger, lower-wage plants (henceforth, simply “small”). Previous DHS research concluded that job ﬂows are large, permanent, and concentrated in a minority of plants with large employment changes. Also, large plants account for most of the increases in job destruction and reallocation during recessions. Together these DHS ﬁndings suggest that during recessions only a small fraction of really 24 Jeffrey C. Fuhrer and Scott Schuh large plants experience really large and permanent rates of job destruction, and thus they imply that the cause of job destruction and recessions is related to large plants.
The Schuh and Triest ﬁndings signiﬁcantly reﬁne this DHS view.
They ﬁnd that small plants tend to have much higher rates of job creation and destruction than large plants, and that high rates of job creation and destruction— especially plant start-ups and shutdowns—are much more likely to be permanent. Thus, even though large plants account for most of the increase in job destruction during recessions, these large-plant job destruction rates are likely to be much smaller in percentage terms and less permanent. In fact, Schuh and Triest ﬁnd that almost one-half of all jobs destroyed by plants experiencing relatively mild contractions are ultimately restored within ﬁve years. In other words, all plants are adversely affected by recessions but large plants appear to be more resilient than small plants, which expand and contract more dramatically and permanently.
Finally, Schuh and Triest uncover some evidence that suggests allocative factors cause business cycles. Their evidence is based on the premise that there are observable determinants of the allocation of jobs across ﬁrms, industries, and regions—prices, productivity, and investment—and that changes in those determinants cause job reallocation to increase, which in turn causes recessions. One key ﬁnding is that when relative prices and productivity growth across detailed industries change dramatically, job destruction and job reallocation also increase dramatically shortly afterward. Another key ﬁnding is that increases in job reallocation generally are not associated with increases in trend productivity and investment growth, as some recent theoretical models seem to imply.
Ricardo Caballero regards some of the Schuh-Triest results as “potentially promising,” but he challenges two fundamental tenets. He questions the central premise that job reallocation is countercyclical, and he doubts that reallocation shocks actually cause ﬂuctuations. In addition, he objects to the authors’ characterization and testing of theories of job reallocation.