By Daniel S. Hamermesh*

*Edward Everett Hale Centennial Professor, University of Texas at Austin and research associate, National Bureau of Economic Research and Institut für die Zukunft der Arbeit.

I. Introduction

The central questions in labor demand deal with the responsiveness of employers’ use of various components of their inputs of labor to changes in their costs. This general rubric includes employment-wage elasticities for labor as a whole and for various labor subaggregates; elasticities of relative employment in different groups in response to changes in their relative costs; the patterns of employment change as scale expands and as capital deepens and/or improves; the paths of employment as old equilibria are shocked and new ones are approached; and these same things for measures of labor utilization, such as hours per time period.
All of these have been very extensively studied; and the existing literature would seem to give a convincing degree of agreement on at least some of the central issues (see Hamermesh,
1993). Nonetheless, not everyone is convinced about how much we know on even the simplest question—the constant-output own-price elasticity of demand for aggregate labor (Topel, 1998). Since this parameter is fundamentally important for understanding the impacts of such diverse policies as payroll taxation, subsidies for employment growth, and others, one wonders whether there is any hope of convincing skeptics that something can be known. Part of the reason for the skepticism may be the fact that most empirical research is based on labor markets in industrialized countries. Such studies suffer from the problem that exogenous changes in labor costs or restrictions on employment demand are very rare in those countries; and, when they do occur they are typically small. This means that researchers trying to identify structural parameters using these data must either rely on models that go to great lengths to establish the exogeneity of the labor-cost measures, or they must search for tiny changes in employment and/or hours in response to the few tiny exogenous changes in labor costs. Neither approach is particularly satisfying.
So long as one believes that the underlying technologies are the same in developed and developing economies, data from the latter provide an ideal way to infer the sizes of important structural parameters. Broad swings in the political viewpoints of succeeding governments lead

to broad changes in labor-market policy. The major vicissitudes in policy can be exploited to allow inferences about these parameters that are based on large exogenous changes in labor costs. It is true that in many cases the data on labor markets in developing countries are not as complete as in developed economies; but in some cases they are, and in those instances the availability of good data that cover periods of widespread and substantial policy changes allows us to make inferences about labor demand that should be useful for students of labor-market behavior generally.
In some instances Latin America and the Caribbean meet both criteria: Some of the policy changes are much larger, in terms of the size of the shocks, than we typically see in developed economies; and in many cases the data, especially establishment data, are very well suited to studying labor demand. In what follows I examine what we can learn about these central issues in labor demand from recent studies of the economies of this region.
II. Evidence on the Overall Demand for Labor

The central parameter in the study of labor demand is the constant-output own-wage elasticity of employment demand. Hundreds of estimates of this parameter have been produced using a variety of methods and types of data (Hamermesh, 1993). A large number of the recent Latin/Caribbean studies generate estimates of precisely this parameter. Moreover, while most of the studies in the literature use aggregate or industry data, many of the studies in this recent group produce their estimates using firm-level data, thus avoiding the aggregation biases that are likely to be severe in what are surely nonlinear economic relationships. Assuming measurement errors in these micro data are not a serious problem, even apart from the greater exogenous variation in labor costs that is likely in developing countries the spatial disaggregation of the data gives these studies an advantage over most earlier studies.
In Table 1 I summarize estimates of these elasticities from a number of recent Latin/Caribbean studies. The methodologies of the studies are fairly closely comparable: All estimate equations describing the logarithm of employment as a function of wages and output,

thus providing a direct estimate of this crucial parameter. All except the Barbadian study include at least one lag in employment as an additional regressor. Clearly and necessarily the results differ across the studies. Partly these differences arise from slight variations in the methods of estimation and in the data (their frequency, their level of spatial aggregation, and their definitions of the measures of labor costs). Part too may be due to true differences in the nature of the technologies used in the different countries, perhaps differences in the output mix, perhaps true differences in the means of producing the same product.
Despite the obvious differences the results are remarkable for their apparent consistency. Taking the results for the four countries—Barbados, Brazil, Peru and Uruguay—for which the estimates have been produced covering all employment, the average constant-output own-wage elasticity is -.30. The estimates by Fajnzylber and Maloney are somewhat larger than the other estimates, but one must remember that they based on employment disaggregated by a measure of skill, so it is unsurprising that they are bigger (more below). Taking all the estimated elasticities together, one must infer that they are essentially identical to the consensus estimate, -.30, that I identified (Hamermesh, 1993) from the many studies covering mainly industrialized economies and based mainly on more highly aggregated data.
That the estimates replicate the previous literature should in part lay to rest concerns that our knowledge of the size of this crucial parameter is too uncertain to allow us to make predictions about the likely impact of imposed changes in labor costs on the level of employment. The sizes of the shocks to labor costs in these countries, and their source in the sharp political changes that have swept over the region, suggest that concerns about identification that have led critics to question studies based on developed economies should be less severe here. As in Angrist’s (1996) study of the West Bank/Gaza Strip, these additional estimates for developing economies suggest that our inferences from developed economies may be applied broadly elsewhere.

Although estimates of the constant-output own-wage elasticity for homogeneous labor are the most common in the set of recent Latin/Caribbean studies, they contain a variety of other implications for static labor demand that merit attention. The Argentine study (Mondino- Montoya) estimates the demand for employment and hours separately, including as regressors lagged values of each component of the labor input. The long-run wage elasticities generated in the study are thus measures of the responsiveness of the employment-hours ratio to changes in payroll costs. The long-run constant-output elasticity of employment to a change in labor costs, holding hours constant, is -.94. This implies a substantial degree of worker-hours substitution and indicates that a rise in wages induces employers to work a smaller labor force more intensely each hour. Since the same elasticity based on the equation for hours is only -.03, by inference a rise in labor costs does not alter the length of the workweek.
Because it disaggregates employment by (a broad measure of) skill, the three-country Fajnzylber-Maloney study allows us to examine whether there are differences in the long-run responsiveness to shocks to labor demand by skill. This depends on the nature of the underlying possibilities for multifactor substitution among the types of labor and capital; but past studies suggest that there is an inverse relationship between the amount of skill embodied in a group of workers and the (absolute value of) the elasticity of demand for their labor. Except for the estimates for Colombia, this is not the case in this study. This apparent inconsistency with the literature may not be as disturbing as it seems at first glance, however. The level of skill in even the lower-skilled groups examined in studies of developed economies may be as high as the average in the white-collar groups summarized in Table 1. Because the definition of skill is so fluid, the results here may simply not be comparable to those found elsewhere in the literature.
In their estimates for the Caribbean Downes et al recognize that a well-enforced minimum wage that bites high into the distribution of earnings will have major negative effects on employment. Indeed, even though the estimates are based on aggregate data and thus include many workers whose wage and employment are unlikely to be affected by changes in minimum

wages, they find this effect for Jamaica, where they specify a coverage-weighted minimum wage index (although not for Trinidad and Tobago where a less complex index is used). The exact sizes of the estimated elasticities tell us nothing about the underlying structure of demand, since they depend on both the labor-demand elasticity and the fraction of the workforce affected by the minimum.
While not included in the summary Table 1, most of the studies other than Fajnzylber— Maloney include measures of the costs of employment regulations in addition to the direct measure of labor cost. Most specify these as representing the effect of the changing stringency of administrative regulations rather than the impact of endogenous changes in the actual costs of the programs. Indeed, perhaps the strongest point in some of these studies and their biggest innovation is the careful construction of these measures in an area where the literature is fairly sparse (Hamermesh, 1993, Chapter 8) and most studies simply look at before-after changes. That in their study of Peru Saavedra—Torero find significant reductions of employment in response to increases in the generosity of severance pay suggests that, as with the wage-cost measures that are included in standard labor-demand equations, so too do other exogenous cost increases reduce employment demand. A similar result is found for Argentina by Mondino—Montoya, who also carefully constructed a regulatory-cost measure.
Taken together the Latin American evidence should add considerably to economists’ and policy advisors’ assurance in emphasizing the static economic costs of so-called job-protection policies. They should also underline the essential irrelevance of a recent spate of mathematically clever theoretical models based essentially on monopsony arguments that claim that such policies may actually increase employment (e.g., Bertola, 1992); and they should generate additional skepticism about empirical results from cross-country comparisons that examine the estimated impacts of quickly-constructed indexes of regulatory stringency in the labor market (e.g., OECD,
1999).

III. Dynamic Labor Demand in Latin America and the Caribbean

Most of the recent Latin/Caribbean studies allow us to examine the path of employment as the labor market moves between old and new equilibria in response to cost or output shocks. All of the studies specify smooth symmetric adjustment by including one or two lagged dependent variables in the labor-demand equations. Thus while they ignore the innovations in demand dynamics that were pointed out in the 1990s (see Hamermesh—Pfann, 1996, for a summary), their firm rooting in the standard literature on factor demand allows for ready comparisons.
Table 2 summarizes the speeds of adjustment of labor demand that have been estimated in the recent Latin studies. In each case I have calculated the speed as the number of time periods, t*, for half the gap between old and new equilibria to be traversed. In these equations with a single lagged dependent variable the calculation is:
(1) t* = ln(.5)/ln λ ,

where λ is the coefficient on the lagged dependent variable. To facilitate comparisons across studies the table expresses t* in quarters. That there may be temporal aggregation biases is well known and may explain why the estimates here and in the literature generally indicate slower adjustment the more highly temporally aggregated are the underlying data. Acknowledging this, however, except for the Paes de Barros—Corseuil study of Brazil all of the estimated speeds of adjustment are slow relative to the typical estimates for developed economies. In the literature generally the best estimate is that the half-life of the lag in adjustment of employment demand is around two quarters (Hamermesh, Chapter 7), below all but the one exception listed in Table 2.
We cannot determine whether this striking difference between these Latin estimates and the estimates in the general literature arises because adjustment is truly slower in Latin America, or because the dynamic specifications of the estimating equations are incorrect. For example, it may be that the assumption of smooth symmetric adjustment is incorrect, and any one of a large variety of alternative specifications would describe the dynamics in the data sets better. These would include lumpy adjustment, linear adjustment costs and any kind of asymmetry.

Findings in the literature for developed economies have also established some regularities in the estimated relative speeds of adjustment of the demand for different components of the labor aggregate and for different groups of workers. Mondino—Montoya’s finding for Argentina that the demand for hours adjusts more rapidly than the demand for employees is consistent with both the large empirical literature for developed economies and with the observation that adjusting hours initially in response to what may be perceived as a temporary shock allows employers to avoid incurring the fixed costs of hiring/firing workers.
A comparison of the relative speeds of adjustment of the demand for blue- and white- collar employees in the Fajnzylber—Maloney estimates is less encouraging. In estimates for two of the three countries that they examine adjustment is slower for blue- than for white-collar employees. This result is inconsistent with the widespread finding in developed countries’ labor markets that adjustment speeds decrease with the skill of the work group (presumably because the fixed costs of hiring/firing are greater among more skilled workers). Why the results should be opposite this is unclear. Perhaps unions and other institutional arrangements restrict the adjustment of blue-collar employment more than that of white-collar employment; perhaps misspecification of the dynamics due to the temporal aggregation of the data is more serious in the equations describing blue-collar employment.
Beginning with Nadiri—Rosen (1969) a small literature has developed examining the linkages among the adjustment paths of several factors. The issue is whether greater speed in the adjustment of one productive input raises the rate at which another adjusts—whether the inputs are dynamic complements or not. There is no consensus in the literature on this point for any pair of inputs; and the results in the one Latin study that provides evidence on this issue do not help to pin down the answer to this question. Mondino—Montoya do include lags in hours per worker (employment) in the dynamic equations describing employment (hours). Unfortunately, the coefficients on lagged hours in the employment equation sum to .15, while those on lagged employment in the equation describing hours sum to -.06. Perhaps the best conclusion is that

there is little evidence for or against the dynamic complementarity of employment and hours per worker.
Increased product-market competition, such as would arise from greater openness to international trade, might be expected to spur employers to adjust more rapidly in response to shocks to costs or output. With competitors from other countries introducing new products and technologies, domestic producers are uncompetitive if they maintain antiquated staffing structures. Evidence for Mexico based on monthly data covering 1987-1995 (Robertson, 1997) supports this hypothesis; and Cassoni et al provide some weak evidence in favor of it too, since there was a slight rise in the speed of adjustment of employment demand as the Uruguayan economy became more open.
One of the central purposes of the recent group of Latin/Caribbean studies was to examine how changing employment regulation, particularly job-security laws such as those governing severance pay, affected the speed of adjustment of employment demand. (Heckman and Pagés, 2000, discuss this at length.) The evidence on this issue in the literature (Hamermesh,
1993, Chapter 8) is sparse and conflicting. Much of it, like Paes de Barros—Corseuil’s study for Brazil, is based on comparing speeds of adjustment pre- and post-regulatory change and is inconclusive. Like those other very few studies (e.g., Burgess—Dolado, 1989) that actually try to measure the severity of regulations, Saavedra—Torero find in their panel of establishments that increases in a carefully constructed measure of the cost of severance pay slow the speed of adjustment of employment. Their results suggest that, if we hope uncover their dynamic effects, there is a large payoff in empirical research on labor demand to specifying the details of labor- market regulations.
III. Conclusions and Implications

No single study of an economic phenomenon, or even several studies, is ever highly convincing, since one can worry about the representativeness of the example and particular problems with the research design. The results of recent Latin American and Caribbean studies, however, based as they are labor markets that have not been thoroughly examined using modern econometric techniques and that have been characterized by relatively large shocks, should reinforce our certainty about the negative impact on employment of higher labor costs, both payroll costs and job-market regulations. They should remind policy-makers that in developing economies, as in developed ones, policies that may be socially desirable, but that raise labor costs and/or increase labor-market rigidity, have negative consequences for the level of employment.
In many ways the strengths and weakness of these Latin/Caribbean studies mirror those of the vast literature of labor demand. The estimates of static labor-demand parameters seem generally reasonable—reasonably tightly estimated and consistent with an underlying theory that is fairly closely linked to the estimating equations. The estimates of the dynamics of adjustment are much less convincing, both in these studies and in the larger literature. This may be because the dynamic specifications are much less loosely linked to theory than are the static estimates. This difference suggests that further work on Latin America and the Caribbean should focus on more careful specification of labor-market dynamics if we are to be able to draw inferences about the impact of shocks and labor-market regulations on fluctuations in the demand for workers and hours.

REFERENCES

Joshua Angrist, “Short-Run Demand for Palestinian Labor,” Journal of Labor Economics, 14 (July 1996), pp. 425-453.

Giuseppe Bertola, “Labor Turnover Costs and Average Labor Demand,” Journal of Labor
Economics, 10, (October 1992), pp.389-411.

Simon Burgess and Juan Dolado, “Intertemporal Rules with Variable Speed of Adjustment: An Application to U.K. Manufacturing Employment,” Economic Journal, 99 (June 1989), pp. 347-65

Adriana Cassoni, Steven Allen and Gaston Labadie, “Unions and Employment in Uruguay,” Unpublished paper, GEOPS—Montevideo, May 1999.

Andrew Downes, Nlandu Mamingi and Rose-Marie Belle Antoine, “Labor Market Regulation and Employment in the Caribbean,” Inter-American Development Bank, Research Network Working Paper R-388, 2000.

Pablo Fajnzylber and William Maloney, “Labor Demand in Colombia, Chile and Mexico,” Unpublished Paper, World Bank, March 2000.

Daniel Hamermesh, Labor Demand. Princeton, NJ: Princeton University Press, 1993. (Published in Spanish as La Demanda de Trabajo. Madrid: Ministerio de Trabajo y Seguridad Social, 1995.)

———————– and Gerard Pfann, “Adjustment Costs in Factor Demand,” Journal of
Economic Literature, 34 (September 1996), pp. 1264-92.

James Heckman and Carmen Pages, “The Cost of Job Security Regulation: Evidence from Latin
American Labor Markets,” National Bureau of Economic Research, Working Paper No.
7773, 2000.

Guillermo Mondino and Silvia Montoya, “The Effects of Labor Market Regulations on Employment Decisions by Firms: Empirical Evidence for Argentina,” Inter-American Development Bank, Research Network Working Paper R-391, 2000.

M. Ishaq Nadiri and Sherwin Rosen, “Interrelated Factor Demand Functions,” American
Economic Review, 59 (September 1969), pp. 457-71.

Organization for Economic Cooperation and Development, OECD Employment Outlook, June
1999. Paris: OECD, 1999.

Ricardo Paes de Barros and Carlos Henrique Corseuil, “The Impact of Regulations on Brazilian
Labor Market Performance,” Unpublished paper, IPEA—Rio de Janeiro, March 2000.

Raymond Robertson, Labor Market Consequences of International Economic Integration.
Unpublished Ph.D. Dissertation, University of Texas at Austin, 1997.

Jaime Saavedra and Maximo Torero, “Labor Market Reforms and Their Impact over Formal Labor Demand and Job Market Turnover: The Case of Peru,” Inter-American Development Bank, Research Network Working Paper R-394, 2000.

Robert Topel, “Analytical Needs and Empirical Knowledge in Labor Economics,” in John Haltiwanger, Marilyn Manser and Robert Topel, eds. Labor Statistics Measurement Issues. Chicago: University of Chicago Press, 1998.

Table 1. Latin/Caribbean Estimates of Constant-Output Own-Wage Labor Demand Elasticities

Country/Study Data Frequency and Time Estimated Elasticity
Period

Barbados/Downes et al Aggregate Annual, 1970-96 -.17

Brazil/Paes de
Barros—Corseuil
Establishments
Monthly, 1986-97
-.40

Blue-collar White-collar

Chile/Fajnzylber— Firms Annual, 1981-86 -.32 -.48
Maloney

Colombia/Fajnzylber— Firms Annual, 1980-91 -1.37 -.59
Maloney

Mexico/Fajnzylber— Firms Annual, 1986-90 -.42 -.44
Maloney

Peru/Saavedra— Sectors Quarterly, 1987-97 -.19
Torero

Uruguay/Cassoni 2-digit industries Quarterly 1975-84 1985-97
et al
-.69 -.22

Table 2. Latin/Caribbean Estimates of Speed of Adjustment of Labor Demand (in Quarters)

Country/Study Data Frequency and Time Half-life of Adjust-
Period ment

Argentina/Mondino- Montoya Firms Annual, 1970-96 Employment Hours
5.4 0.4

Brazil/Paes de Establishments Monthly, 1986-97 1.0
Barros—Corseuil

Blue-collar White-collar

Chile/Fajnzylber— Firms Annual, 1981-86 8.8 4.0
Maloney

Colombia/Fajnzylber— Firms Annual, 1980-91 20.8 5.6
Maloney

Mexico/Fajnzylber— Firms Annual, 1986-90 14.4 19.2
Maloney

Peru/Saavedra— Establishments Quarterly, 1987-97 5.1
Torero

Uruguay/Cassoni 2-digit industries Quarterly 1975-84 1985-97
et al
5.9 5.0

SOURCE: https://athens.src.uchicago.edu/jenni/dvmaster/FILES/IADB/Ch12.pdf

IMAGE: mrski-apecon-2008.wikispaces.com