Printer Friendly

Employment and deadweight loss effects of observed nonwage labor costs.

I. INTRODUCTION

Nonwage labor costs consist of several mandated benefits, health, training, accident, housing plans, as well as several taxes, and are intended to increase workers' welfare and job security. However, these benefits tend to come at the expense of reducing employment and deadweight losses. To quantify these resulting effects, one needs a reliable measure of the employment labor cost elasticity. Most studies conducted on this subject find low estimated values for this elasticity, which contrasts with policy-makers' enthusiasm to reduce labor costs.

In this article, using a Peruvian-matched firm-workers data set, we find that nonwage labor costs reduce employment measured in total hours of work by 17% for white collars and by 53% for blue collars with associated deadweight losses of 10% and 35% of total contributions' revenues, respectively. We also compute employment losses of compliance with mandated employers' and workers' contributions of 4% for white collars and 12% for blue collars, with respective associated deadweight losses of 2% and 6% of contribution revenues.

These results come from estimating the long-run unconditional firm-level labor demand function by a procedure that corrects for endogeneity of wages. We show that unbiased estimates, even using small units such as firm-level data, require not only that wages are exogenous but also that their unobserved determinants are uncorrelated with the unobserved determinants of labor demand. This requirement is not fulfilled in the likely event, pointed out by Becker (1993), that larger firms are matched with more productive workers. Consequently, an estimation that corrects for endogeneity yields a larger labor cost elasticity of labor demand than one that does not, like ordinary least squares (OLS).

In the past two decades, there has been intensive theoretical and empirical research on the effects of labor market frictions on job creation. Western European economies, characterized by large job security provisions, have been the center of attention of this research that has modeled these frictions as adjustment costs in labor demand (Bentolila and Bertola 1990; Hamermesh 1989; Hopenhayn and Rogerson 1993; Nickell 1987; Rendon 2004). This literature finds that the effects of "eurosclerosis," that is, labor markets with high firing costs, are ambiguous: in good times, sclerotic labor markets create fewer jobs than free labor markets; however, in bad times, sclerotic labor markets reduce job destruction. Most of the research done in Latin America has been done under these guidelines, obtaining statistically significant negative effects of job security provisions on employment rates. (1) Furthermore, using country-level data for Latin America and Organization for Economic Cooperation and Development countries, Heckman and Pages-Serra (2000) found large effects of job security provisions on employment, (2) which are robust to several specifications, OLS, random, and fixed effects. The authors present their results as a strong evidence against Freeman's (2000) view that job security regulations mostly affect distribution, but not efficiency, and "advocate the substitution of job security provisions by other mechanisms that provide income security at lower efficiency and inequality costs." Under a classic labor demand framework and using worker- and firm-level data, this article provides new evidence in the same direction, extending the computation of the effects of job security provisions to deadweight losses.

Hamermesh (1993) surveys several studies on the estimation of labor demand and remarks that estimates of the labor cost employment elasticity should be interpreted and compared cautiously depending on the specification adopted. Capital is an explanatory variable in a short-run labor demand estimation; it is not in a long-run labor demand estimation. Similarly, output is included in the estimation of a conditional labor demand but excluded in the estimation of an unconditional labor demand. When estimating a system of equation and thus controlling for endogeneity of wages, estimates will be typically larger than in a uniequational estimation. In our study, we assume homogeneous labor and make two separate estimations for white- and blue-collar workers without allowing for interactions between them. Since our study is done for small units, at the firm level, it is a reasonable assumption that firms face a horizontal labor supply. However, assortative matching between firms and workers gives rise to correlation between wages and firm size, which calls for a correction of this endogeneity. Capital is not included in the estimation, nor output, which makes the estimation a long-run, unconditional labor demand estimation, that is, the scale effect is included in the total effect captured by our estimated elasticity. These features of our research explain why we find larger labor costs elasticities of employment (3) than estimates usually found for the standard specification of a conditional labor demand. (4) Our results are thus encouraging of policies for stimulating job creation by inducing movements along as well as shifts of the labor demand curve.

The remainder of this article is organized as follows. The next section details our estimation approach, which addresses the issue of endogeneity correction; Section III describes the data set used as well as descriptive statistics. Section IV discusses the resulting estimated labor cost employment elasticities obtained by OLS and by an instrumental variable (IV) estimation, both for legal and for observed nonwage labor costs. In Section V, we compute the employment reduction caused by workers' and employers' contributions. Section VI reports the calculated deadweight losses of workers' and employers' contributions as a proportion of contribution revenues. Section VII reports the employment and deadweight losses of complying fully with legal workers' and employers' contributions. Finally, Section VIII summarizes the article's main conclusions.

II. ESTIMATION APPROACH

Consider a static setup where a firm i [member of] {1, 2, ..., N} chooses inputs to maximize profits: [max.sub.K,L]{[p.sub.i] f ([K.sub.i], [L.sub.i]|[[OMEGA].sub.i]) - [r.sub.i] [K.sub.i] - [w.sup.a.sub.i] [L.sub.i]}. The resulting demand for labor is given by the function [L.sub.i]([w.sup.a.sub.i]|[X.sup.*.sub.i]), where [w.sup.a.sub.i] is the total labor cost paid by the firm and [X.sup.*.sub.i] = {[p.sub.i], [r.sub.i], [[OMEGA].sub.i]} represents output and input prices and the parameters of the production function for the firm. A log-linear approximation to this function is

(1) ln [L.sub.i] = [beta] ln [w.sup.a.sub.i] + [X.sub.i][delta] + [u.sub.i],

where [X.sub.i] are those variables contained in [X.sup.*.sub.i] that are observed by the researcher, whereas [u.sub.i] is a random variable representing the unobserved components of [X.sup.*.sub.i] assumed to be normally distributed with zero mean and variance [[sigma].sup.2.sub.u].

If wages are fully exogenous, that is, if In [w.sup.a.sub.i] is uncorrelated with [u.sub.i] then one can obtain an unbiased and consistent estimation of [beta] by OLS. For this assumption to hold, wages have to be determined by an infinitely elastic labor supply. Actually, most empirical research assumes exogenous wages and estimates the labor cost elasticity of employment by OLS. This is true even for estimations that account for individual (fixed) effects when panel data are available.

To illustrate this assertion, let the infinitely elastic labor supply of individual j [member of] {1, 2, ..., [M.sub.i]} related to firm i be:

(2) ln [w.sup.d.sub.ij] = [Z.sub.ij][gamma] + [v.sub.ij],

where [w.sup.d.sub.ij] is the take-home wage; [Z.sub.ij] is a vector of covariates that determine wages; [gamma] is its associated parameters; and [v.sub.ij] is a random variable of unobservables with zero mean, variance [[sigma].sup.2.sub.v], and covariance with [u.sub.i] equal to [[sigma].sub.uv]. This is a reduced form equation that can be thought of as a Mincer equation in which [Z.sub.ij] only includes supply-side variables, such as education, (5) tenure, experience, and individual workers' attributes. (6) Then, firm i faces a labor supply function:

(3) ln [w.sup.d.sub.i] = [Z.sub.i][gamma] + [v.sub.i],

where ln [w.sup.d.sub.i] = [M.sup.-1.sub.i] [summation] ln [w.sup.d.sub.ij], [Z.sub.i] = [M.sup.-1.sub.i] [summation] [Z.sub.ij], and [v.sub.i] = [M.sup.-1.sub.i] [summation] [v.sub.ij] is distributed with mean 0, variance [M.sup.-1.sub.i] [[sigma].sup.2.sub.v], and covariance with [u.sub.i] equal to [[sigma].sub.uv], respectively. Assume for the moment that wages paid by employers coincide with workers' take-home wages: [w.sup.a.sub.i] = [w.sup.d.sub.i]; then, as this is a recursive or limited information estimation model, OLS estimates yield unbiased estimates of the labor demand parameters. The crucial assumption for this to be true is that the firm's individual labor supply is infinitely elastic and error terms are independent, [[sigma].sub.uv] = 0.

However, it may well be the case that [[sigma].sub.uv] [not equal to] 0, which implies that ln [w.sup.a.sub.i] and [u.sub.i] are correlated and the estimation by OLS generates biased estimates. Moreover, if [[sigma].sub.uv] > 0, that is, if unobservables that increase wages are positively correlated with unobservables that increase labor demand then [beta] estimated by OLS will exhibit an upward bias. One can think about this positive correlation as evidence for positive assortative matching between firms and workers: more productive workers are matched to larger firms. Or, in terms of the variables that are unobserved to the researcher, workers of higher ability may work in firms of higher total factor productivity. On the contrary, [[sigma].sub.uv] < 0 is associated with unobservables that increase wages negatively correlated with unobservables that increase labor demand, in which case OLS leads to underestimate [beta]. Therefore, that is the case of negative assortative matching between firms and workers: more productive workers are matched to smaller firms. (7)

Figure 1 depicts the supply and demand for labor and illustrates this matching effect: if [D.sub.1] is matched to [S.sub.1], [D.sub.2] to [S.sub.2], and [D.sub.3] to [S.sub.3] then the resulting equilibrium points describe a positive relationship or a negative relationship that is steeper than the labor demand (overestimation of [beta]); if matching between supply and demand is the other way round, negative, the relationship resulting from the equilibrium points is flatter than the labor demand (underestimation of [beta]).

[FIGURE 1 OMITTED]

Now, let us allow labor costs paid for by employers differ from workers' take-home wages. Suppose that employers' contributions expressed as a percentage of wages are [a.sub.i], so that employers' labor costs are [w.sup.a.sub.i] = [w.sub.i](1 + [a.sub.i]). On the other hand, let [d.sub.i] be workers' contributions as a percentage of wages and workers' take-home wage is then [w.sup.d.sub.i] = w(1 - [d.sub.i]). Under the assumption that both employer- and worker-paid contributions are unrelated to unobservables that determine firms' size (8) one can relate employers' labor costs to workers' earnings simply by:

(4) ln [w.sup.a.sub.i] = ln [w.sup.d.sub.i] - ln(1 - [d.sub.i)] + ln(1 + [a.sub.i]).

Notice that if wages are fully exogenous, (9) under no circumstance, workers earn below ln [w.sup.d.sub.i], which implies that workers' contributions are actually paid for by employers. In particular, if workers' contributions were reduced, employer-paid wages would adjust, so that workers' take-home wages are left unchanged. This feature of the labor market will be important when analyzing the employment effects of removing workers' and employers' contributions.

Under this setup, we can propose the following estimation procedure:

(1) First stage: Estimate Equation (2), predict the workers' take-home wage, aggregate (10) them to [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and use it to predict the total labor cost paid by the employer [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] Equation (4).

(2) Second stage: Estimate Equation (1) using the predicted employers' labor cost [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. This estimation procedure removes the correlation between the labor costs paid for by employers and the disturbance term in Equation (1) and yields unbiased estimates of [beta] and [delta].

To implement this estimation, we need firm-level data matched with data on individuals working at the firm. In the next section, we describe the data used in the estimation.

III. DATA

The data used in this estimation come from the Wage and Salary National Survey carried out by the Ministry of Labor of Peru. This is a biquarterly survey applied in June and December that comprises private firms of ten and more workers and is representative for the main cities (Metropolitan Lima and urban areas of 24 main cities in the country), economic sectors and activities, and firm sizes in Peru. The information for this survey is gathered by qualified interviewers from the Labor Ministry who review firms' payrolls and are specialized in labor costs in Peru; thus, this survey does not consist of self-reported data.

The survey is organized in three sections. Section A aggregates firm-level information such as the total number of workers, wages by occupational category, total hours worked, legal workers' deductions, and employers' contributions by occupational category. Section B contains information on a sample of individual workers inside the firm, with variables such as age and gender of the worker, hours worked, basic wage or salary, legal workers' deductions and employers' contributions, and other nonpermanent payments. Finally, Section C provides information on collective bargaining and unionization of workers.

We use the survey for June 2004, which consists of a sample of 1,772 firms, for which we have 19,770 workers. Because we concentrate on the demand for white- and blue-collar workers, we select two subsamples of firms that hire at least one of these two types of workers. Thus, the resulting samples contain respectively 1,714 firms with 13,097 white-collar workers and 692 firms with 5,413 blue-collar workers. Table 1 shows the descriptive statistics for all variables in the final sample divided by type of worker and by section of the survey. For several variables, there are both firm-level and individual information. Understandably, there is more dispersion for information in the workers' sample (Section B). Both white- and blue-collar workers work on average more than 40 h a week. On average, blue-collar workers work in firms that are 30% larger than white collars and earn half as much as white collars.

In this study, gross wages are defined as payments made by employers to their employees for their work that end up in employees' paychecks, including any mandatory benefits, (11) before deductions for income taxes and pension contributions are applied. Workers' contributions are mandatory payments deducted from workers' paychecks, that is, pension contributions and income taxes. Employers' contributions are payments associated with workers' wages that employers pay and workers do not take home nor are deducted from workers' gross wages, but go to several funds, such as health or training systems, and payroll taxes. (12)

We distinguish between observed and legal contributions: the former comes from the information in our sample, while the latter is calculated from the current labor laws in the country. We will speak of undercompliance when observed fall below legal contributions and of overcompliance if it is the other way round. (13) Both observed employers' and workers' contributions are on average lower than legal ones, that is, undercompliance is predominant. (14) Observed employers' contribution are between 10% and 11% of wages for all samples, while legal contributions are established at 14.5%. For white collars, observed workers' contributions are between 15% and 16% of wages, while the legal contributions are around 18%; for blue collars, observed workers' contributions are around 12% of wages, while legal contributions are set at around 13%. Figure 2 shows the employers' contributions and workers' deductions as a function of wages, both for white- and blue-collar workers. For employers' contributions, one can distinguish a dispersion around horizontal lines because legal contributions are a fixed percentage that does not depend on wage levels. However, the dispersion of workers' deductions occurs around both increasing curves and horizontal lines, as some deductions are increasing in wages, income taxes, while others are fixed, pension contributions. Moreover, there is compliance with some contributions and not with others, which explains why one can distinguish several patterns in these graphs, which are illustrative of the important differences between legal and observed nonwage labor costs.

[FIGURE 2 OMITTED]

In the subsample of white collars, around 70% of firms and 77% of individuals are occupied in the service sector, whereas in the subsample of blue collars, around 50% of firms and 57% of individuals are classified as part of the industrial sector. For both subsamples, most firms are small, more than 50% employing 50 workers or less; however, more than 50% of workers work in firms with 100 or more workers. Around half of firms are located in Lima, the capital city of the country. Unionization is higher for blue-collar workers with rates between 15% and 20% of workers, while for white collars, unionization rates are between 8% and 14%.

In terms of individual data, we find that 38% are females among the white collars, while only 14% are among the blue collars. On average, both white and blue collars are around 38 yr old; however, among blue collars, there is more age dispersion especially at the lower tail: around 12% of blue collars, against only 4% of white collars, are 24 yr old or younger. In the samples, both white and blue collars have on average around 6 yr of tenure; however, in tenure, it is blue collars who also exhibit more dispersion, especially at the lower tails. While 23% of white collars have around less than 1 yr of tenure, 29% of blue collars have tenure of less than 1 yr.

In the next sections, we present the results of estimating a labor demand model both by simple OLS and by controlling for the endogeneity of wages.

IV. ESTIMATION RESULTS

In this section, we present the elasticities estimated by the procedure described in Section II. We estimate several versions of the model, the results of which are presented in Table 2. For the total hours of work and for the number of workers, we report in the first column, an OLS estimation using the average wage reported in the firm-level information; in the second column, an OLS estimation using an average wage constructed using the individual information; and in the third column, an estimation that accounts for endogeneity, an OLS estimation using the average of a predicted wage. For individual hours, we report OLS and IV results in the first and second columns, respectively.

In these regressions, explanatory variables besides labor costs are dummy variables indicating location, whether there is a union in the firm, and sector of activity. These variables capture differences in capital prices across regions, labor relations across firms, and technologies across industrial sectors. Exogenous sources of variation for endogeneity correction are workers' age, tenure, and gender. Further details on the first-stage wage regressions, Equation (2), and their explanatory variables are given in Appendix A3.

Both for white and blue collars, when employment is firms' employment measured by hours of work or by the number of workers, an estimation that accounts for endogeneity yields a larger labor cost elasticity of labor demand than one that is done by simple OLS, which suggests the existence of positive assortative matching between firms and workers. When employment is measured by individual hours of work, correcting for endogeneity reduces the labor cost elasticity of labor demand, implying negative assortative matching of workers and firms' individual hours of work. In sum, more productive workers are matched to firms that are larger, in terms of total hours worked and number of employees, and in which working time is shorter.

It is also noteworthy that the labor cost elasticity of the firm-level labor demand is larger for blue-collar workers than for white-collar workers: measured by total hours, it is -0.65 for white collars and -2.31 for blue collars and measured by the number of workers, it is -0.51 for white collars and -2.31 for blue collars. In contrast, the individual hours labor demand has an elasticity of -0.05 for white collars and 0.00 for blue collars.

For white-collar workers, there is no big difference in estimating by OLS the total labor cost elasticity with the reported or the constructed firm average wage (ln [w.sub.m] and [bar.ln w]). The sign, however, in both of these estimations is wrong and only becomes negative once endogeneity is corrected for. Using legal rather than observed nonlabor costs produces an underestimation of the labor cost elasticity of employment as measured by total hours of work or number of workers, although this underestimation is lower once a correction for endogeneity is introduced. For individual hours of work, using legal rather than observed labor costs leads to a slight overestimation of this elasticity, though both yield very low values.

For blue-collar workers, estimating the labor cost elasticity by OLS with the reported or the constructed firm average wage (ln [w.sub.m] and [bar.ln w]) yields substantially different results. Estimating the total labor cost elasticity correcting for endogeneity yields very high values for employment measured by total hours and number of workers. The labor cost employment elasticity measured by total hours is underestimated when legal instead of observed nonwage labor costs are used; however, it is slightly overestimated when employment is measured by the number of workers. The labor cost elasticity of employment measured by individual hours has the wrong sign; its value is very small and in many cases nonsignificant. As with white collars, using legal rather than observed labor costs implies a slight overestimation of the labor elasticity of demand.

In sum, estimations of the firm-level labor demand that are corrected for endogeneity generate large labor costs employment elasticities, especially for blue-collar workers. The opposite is true for the corresponding individual labor demand estimations, where corrections for endogeneity lower labor cost employment elasticities, especially for blue collars. Furthermore, as a means to check for the sensitivity of the instruments, we perform an alternative specification for the first-stage wage regression omitting tenure as a regressor (Appendix A4).

In the next section, we use these estimated elasticities to forecast the employment and deadweight loss effects of nonwage labor costs.

V. EMPLOYMENT EFFECTS OF NONWAGE LABOR COSTS

In this section, we predict the percentage employment variation produced by removing nonwage labor costs fully, that is, the effects of eliminating:

(1) employer-paid nonlabor costs: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(2) worker-paid nonlabor costs: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(3) both employer- and worker-paid nonlabor costs: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

We perform these exercises (15) using both observed and legal nonwage labor costs and their associated estimated elasticities computed in the previous section for the three measures of employment by total hours, number of workers, and individual hours. The results are reported in Table 3 and show that the largest employment effects are obtained from reducing workers' nonwage labor costs, especially for white-collar workers. For white-collar workers, removing employer-paid nonwage labor costs increases firm-level total hours of work by around 6% and the number of workers by 5%, while removing workers' paid nonwage labor costs increases hours of work by 11% and the number of workers by 8%. The elimination of both nonwage labor costs increases total hours by 17% and the number of workers by 13%. Thus, employment effects for white collars are larger for total hours than for number of workers. For individual hours, effects are small but significant and positive: around 1% increase.

For blue-collar workers, employment effects are similar for total hours and for the number of workers and much larger than for white collars: 24% for removing employers' contributions and 29% for removing workers' contributions, so that the employment effect of removing both contributions is 53%. For blue-collar workers, individual hours effects are not significantly different from zero.

Using legal rather than observed nonwage labor costs introduces an important overestimation of the employment effects for blue-collar workers, measured by total hours, number of workers, or individual hours. For white-collar workers, the differences between legal and observed employment effects are less pronounced. There is also an overestimation of employment variations when using legal nonwage labor costs for total and individual hours, but an underestimation for the number of workers.

Hence, the employment losses provoked by both employers' and workers' contributions are shown to be substantial, especially for blue-collar workers. Estimations that do not correct for endogeneity would have found positive significant employment increases for white collars and negligible increases for blue collars, concluding thereby that nonwage labor costs had very little effect in stimulating employment, as is the case in most of the literature on this subject. In the next section, we compute the deadweight losses associated with these employment effects.

VI. DEADWEIGHT LOSSES OF NONWAGE LABOR COSTS

In this section, we compute the deadweight loss effects of employers' and workers' contributions. We estimated the model [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], where [A.sub.i] = [X.sub.i][delta] are the employment effects of all other regressors; that is, the level of employment is L = [A.sup.[beta].sub.w]. We have two wage levels, [W.sub.1] and [w.sub.0] ([w.sub.i] > [w.sub.0]), which imply employment levels [L.sub.1] = [A.sup.[beta].sub.w1] and [L.sub.0] = [A.sup.[beta].sub.w0], respectively. Contribution revenues are then R = ([w.sub.1] - [w.sub.0]) [L.sub.1] > 0, and the deadweight loss area comes from integrating:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

In Appendix A5, we give further details on this computation (see also Auerbach and Hines 2002). One can also approximate the deadweight loss, as it is usually done, by a (Harberger) triangle:

T = ([w.sub.i] - [w.sub.0])([L.sub.0] - [L.sub.1])/2.

Figure 3 illustrates the deadweight loss area T and the contribution revenues R when there are both employers' contributions, which shift the labor demand downward, and workers' contributions, which shift the labor supply upward.

Then, wage levels [w.sub.1] and [w.sub.0] are defined for:

(1) employers' contributions: [w.sub.1] = w(1 + a), [w.sub.0] = w

(2) workers' contributions: [w.sub.1] = w, [w.sub.0] = w(1 - d)

(3) both workers' and employers' contributions: [w.sub.1] = w(1 + a), [w.sub.0] = w(1 - d).

[FIGURE 3 OMITTED]

Table 4 reports the estimated deadweight losses of eliminating workers' and employers' contributions, using different measures of employment, of contributions, and of deadweight losses, for white and blue-collar workers, respectively. As with employment effects, deadweight losses are larger for workers' than for employers' contributions for both occupational categories and larger for blue-collar than for white-collar workers. For white collars, deadweight losses of observed employers' contributions are 5.6% of contribution revenues; of observed workers' contributions, they are 3.2% of contribution revenues when measured by total hours. For blue collars, deadweight losses of observed employers' contributions are 12.5% of contribution revenues; of observed workers' contributions, they are 16.1% of contribution revenues when measured by total hours. For blue-collar workers, deadweight losses are practically identical when employment is measured by the number of workers. For white-collar workers, deadweight losses are 2.5% of contribution revenues for employers' contributions and 4% for workers' contributions. For individual hours, effects are very small for white-collar workers and, as employment effects, appear with the wrong sign for blue-collar workers.

For white-collar workers, deadweight losses, in terms of both total hours and the number of workers, are somewhat larger when measured by observed rather than by legal nonwage labor costs. For blue-collar workers, they are substantially smaller for the observed nonwage labor costs when using both the total hours and the number of workers.

Thus, there are not only substantial employment losses but also large deadweight losses of mandated employers' and workers' contributions, especially for blue-collar workers.

VII. EMPLOYMENT AND DEADWEIGHT LOSSES OF COMPLYING WITH LEGAL CONTRIBUTIONS

In this section, we analyze the employment and deadweight loss effects of adjusting observed nonwage labor costs to their legal level.

What is the employment effect of firms' not complying with paying employers' and workers' contributions fully? Because of undercompliance, the actual labor cost incurred by firms is lower than the stipulated legal one. Table 5 reports the effects of not complying with legal contributions on employment measured by total hours worked at the firm level, the number of workers by firm, and individual hours. For white-collar workers, the effect of not complying with legal workers' contributions is about the same as the effect of not complying with employers' contributions, 2.42% for total hours, 1.90% for the number of workers, and 0.17% for individual hours. The total effect of not complying with these two contributions amounts to 4.83% for total hours, 3.81% for the number of workers, and 0.34% for individual hours.

For blue-collar workers, the picture is somewhat different, as the employment effect of not complying with the employers' contributions is larger than the effect of not complying with the workers' contributions. Moreover, since the labor cost elasticity is large, the employment effects are much larger than for white-collar workers. As one would expect because of the low labor costs individual hours elasticities, for blue-collar workers, the effect in individual hours is negligible. Both for employment measured as total hours of work or the number of workers, employment effects of not complying with employers' contributions are around 7.5% and are around 4.1% for not complying with workers' contributions. Thus, the effect of not complying with both contributions is around 11.6%. As shown in the previous sections, variations in the individual hours margin induced by labor costs variations is somewhat important only for white-collar workers, not for blue collars.

In Table 6, we show the deadweight losses resulting from complying fully with the legal level of nonwage labor costs. (16) As in the previous section, we compute them both as an integral and simply as a triangle. For white-collar workers, the deadweight loss of both employers' and workers' contributions measured as total hours or as the number of workers are around 2% of contribution revenues, split almost evenly between the two. For blue-collar workers, the deadweight loss represents around 6% of contribution revenues, again for both contributions and measured as total hours or the number of workers. Once again, individual hours of blue-collar workers are not reactive to labor costs variations; consequently, deadweight losses are smaller than for white collars.

In sum, judging from its implied employment and deadweight losses, undercompliance is substantial. Employment losses of compliance with mandated employers' and workers' contributions are 4% for white collars and 12% for blue collars, with respective associated deadweight losses of 2% and 6% of contribution revenues.

VIII. CONCLUSIONS

Using a matched firm-workers data set, we have shown that an estimation that accounts for endogeneity of wages yields a larger labor cost elasticity of a long-run unconditional labor demand than one obtained by OLS. We explain that this result is evidence for positive assortative matching between firms and workers: larger firms are matched with more productive workers. We find that employers' and workers' paid nonwage labor costs reduce employment by 17% for white collars and by 53% for blue collars. The associated deadweight loss of these nonwage labor costs are 9% of contribution revenues for white-collar workers and 31% of contribution revenues for blue collars. Significant increases of individual hours only occur for white-collar workers, not for blue-collar workers, that is, white collars exhibit a larger labor costs elasticity of demand for individual hours than blue collars.

On the other hand, estimating labor costs employment elasticities using nonwage labor costs measured in the available data sets rather using legally established rules yields substantially different results only for white-collar workers, for which undercompliance with legal contributions is larger than for blue collars. Furthermore, we compute the employment effects of undercomplying with the mandated employers' and workers' contributions. Because of undercompliance employment is 4% larger for white collars and 12% larger for blue collars. The deadweight loss of complying with mandated contributions is 2% for white collars and 6% of contribution revenues for blue collars.

These results show large employment and, often ignored, deadweight losses of both mandated employers' and workers' contributions and are thus encouraging of policies to increase job creation by lowering nonwage labor costs.

ABBREVIATIONS

IV: Instrumental Variable

OLS: Ordinary Least Squares

doi: 10.1111/j.1465-7295.2009.00190.x

APPENDIX A1. LABOR REFORMS IN PERU IN THE 1990S

According to Saavedra (2000), labor laws were very restrictive, protectionist, and cumbersome. In the early 1990s, Peru went through a process of "structural" reforms that were intended to make labor markets more flexible.

Blue Collars and White Collars

Before the reforms, there was a strong distinction between white- and blue-collar workers, so that firms had to have different payrolls with different payment frequencies for these two types of workers: blue collars were paid on a weekly basis, while white collars on a monthly basis. Blue collars had more benefits than white collars, which reduced the relative hiring of blue collars (Chacaltana 1999). The reforms eliminated the strong distinction between them, so that both are considered workers with same severance payments and other benefits. Firms are also free to choose the frequency of payment to their workers.

Firing Costs

Up to the 1990s workers in Peru enjoyed absolute stability at the workplace, a right that was protected by the Constitution. The labor reform changed this completely by introducing the "unfair" firing, that is, workers can be fired without any justification, just receiving a severance payment. In 1996, after several changes, firing costs for unfair dismissals were established at one and a half monthly wages for every year employed at the firm, with a ceiling of 12 wages. The reform also extended "fair" dismissals to include workers' bad conduct and low productivity and introduced technological, economic, and structural reasons as valid causes for collective layoffs, that is, dismissals of not less than 10% of the workforce.

Temporary Contracts

Before the reforms, temporary contracts required written authorization by the Ministry of Labor, had a maximum duration of 1 yr, and were renewable only for 1 yr. The labor reforms allowed temporary contracts of several durations and without any authorization by the government. Workers under these contracts have the same benefits as workers with contracts of undetermined duration; however, if the employer fires a worker before the term of the contract, the firing cost of the permanent contract applies. These contracts can be of 1 yr, with a maximum renewal of 5 yr. The law also allowed temporary contracts for a specific work or service of a determined duration and with different frequencies.

These labor reforms made the labor markets more flexible by eliminating absolute job stability, reducing firing costs, and allowing temporary contracts without any duration restriction. They also simplified payroll management by equalizing white- and blue-collar workers.
TABLE A1
Additional Wages as a Percentage of Basic Wages
(June 2004)

Additional Concepts
of Wages % of Basic Wage

Compulsory weekly rest 13.30
Nonworking holidays 3.33
Family assignments 2.70
Two monthly wages 16.67
Vacations 8.33
Tenure bonus 9.72
Total additional wage 54.05


APPENDIX A2. LEGAL NONWAGE LABOR COSTS

On top of the basic wage, workers receive several additional bonuses, which are subject to employers' contributions and workers' deductions. In this study, the additional wage is already included as part of the total wage. These additional concepts of wages add up to 54.05% of the basic wage, as can be seen in Table A1.

Additional wages include payments that should cover for weekends (DSO, Descanso semanal obligatorio), nonworking holidays (FNL, Feriados no laborables), and bonuses that are related to the number of family members (family assignments). Besides, workers receive two extra monthly wages every year as Christmas and National Holiday bonuses called Gratificaciones. Vacations are paid holidays that give 30 d per year worked for the same employer. The tenure bonus (CTS, Compensacion por Tiempo de Servicios ) is an additional wage payment for every tenure year of the worker.

Employers' contributions in Peru amount to 14.45% of the basic wage (see Table A2) and consists of the following concepts. Health plan payments represent 9% of the basic wage, of which 6.75% is for the public system (ESSALUD, Seguro Social de Salud) and 2.25% goes to the private system (EPS, Empresas Presmdoras de Salud) if the worker has private health insurance. Otherwise, the whole contribution goes to the public system.

The solidarity extraordinary tax (IES, Impuesto Extraordinario de Solidaridad) was created in 1998 to replace mandatory contributions to finance housing, establishing a national housing fund (FONAVI, Fondo National de Vivienda). Initially, it amounted to 2% of the basic wage, but then went down to 1.7%.

Manufacturing training contributions (SENATI, Servicio Nacional de Adiestramiento en Trabajo) only apply to some industrial firms (Category D of SIC). From 1994 onward, they have been going down from 1.5% to become, in 1997, 0.75% of the basic wage. Accident insurance (SCTR, Seguro Complementario de Trabajo de Riesgo ) is on average 3% of the basic wage.

Workers' deductions consist of income taxes and social security contributions and vary depending on the wage level and on whether the pension system is private or public (see Table A3). Employers retain income taxes and mandatory social security contributions from workers' wages. For wage levels below 7 tax units (UIT, Unidad Impositiva Tributaria), there is no retention; tax units are monetary amounts fixed by the government and updated from time to time. On June 2004, the UIT was S/. 3,200. From 7 UIT onward, employers have to make income tax deductions on workers' payments: the lowest rate is 15%, while the highest reach 30%. The amount of social security contributions differs depending on the type of system. In the public pension system, the contribution is 13%. In the private pension system, the mandatory fixed contribution rate is 8%, the maintenance fee is on average 2.27%, and the insurance fee is on average 0.92%, totaling to 11.19%.
TABLE A2
Employers' Contribution as a Percentage of Basic Wages
(June 2004)

Employers' Contributions % of Wage

Health plan payments 9.00
Solidarity extraordinary tax 1.70
Manufacturing training fund 0.75
Accident insurance 3.00
Total Employers' Contributions 14.45


APPENDIX A3. WAGE REGRESSIONS

In Table A4, we report the wage regressions, the first stage of our estimation of the labor demand elasticity. These are Mincer regressions done for white and blue collars, both for weekly and for hourly wages, and for take-home wages using observed and legal workers' deductions. Unfortunately, in the data set, we do not have workers' education, which would make our regression a typical Mincer regression. However, we have age (which proxies potential experience) and tenure as well as gender, union status, city of residence (Lima vs. other), and industrial sector.

Both returns to age and to tenure are larger for white collars than for blue collars. However, while returns to age are larger than returns to tenure for white collars, the opposite is true for blue collars; returns to tenure are larger than returns to age. Among blue-collars gender wage differences are more pronounced than among white collars.

For blue-collars male workers earn around 20% more than their female counterparts. Among white collars males earn around 7% more in weekly wages, but 3% more in hourly wage than female workers. This difference may be due to the lower amount of hours worked by female white-collar workers. The effect of being unionized is to increase wages by around 25% for white collars and by around 21% for blue collars. Working in Lima, the capital of Peru, means a differential of more than 50% in white-collar wages and of 20% in blue-collar wages over working in other cities.
TABLE A3
Workers' Deductions, Income Tax and Social Security,
as a Percentage of Basic Wages (June 2004)

 Pension System
Workers' Deductions Contributions (%)

Wage Bracket Tax (%) Private 11.19 Public 13.00
(UIT) Tax (%)

0-7 0.00 11.19 13.00
7-27 11.00 22.19 24.00
27-54 16.00 27.19 29.00
+54 +16.00 +27.19 +29.00


APPENDIX A4. SENSITIVITY OF RESULTS: EXCLUDING TENURE

In Table A5, we check for the sensitivity of the wage elasticities of demand when tenure is excluded from the first-stage wage estimation. There is a slight decrease in absolute value for white-collar workers' total hours and number of workers elasticities. The variation is especially large when legal employers' contributions are used. For individual hours of work, it is the opposite: excluding tenure increases labor demand elasticities, especially for legal employers' contributions. For blue-collar workers excluding tenure produces an increase in the absolute value of the labor demand elasticities measured by total hours and number of workers. Measured by individual hours of work, these elasticites are nonsignificant with or without tenure as a regressor in the first stage.

APPENDIX A5. COMPUTATION OF DEADWEIGHT LOSSES

The computation of the deadweight areas proceeds in the following way:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].

Notice that for any value of [beta], if [w.sub.1] > [w.sub.0] then A/[beta] + 1 [[w.sup.[beta] + 1.sub.1] - [w.sup.[beta] + 1.sub.0]] > 0.

APPENDIX A6. DEADWEIGHT LOSSES OF UNDERCOMPLYING WITH LEGAL CONTRIBUTIONS

For computing variations from observed to legal contributions, denoted, respectively, with the subscript "obs" and "legal," wage levels [w.sub.1] and [w.sub.0] are then defined for:

(1) employers' contributions: [w.sub.1] = w(1 + [a.sub.legal]), [w.sub.0] = w(1 + [a.sub.obs])

(2) workers' contributions: [w.sub.1] = w[1 - [d.sub.obs]/1 - [d.sub.legal], [w.sub.0] = w

(3) both workers' and employers' contributions: [w.sub.1] = w[1 - [d.sub.obs]/[1 - [d.dub.legal] (1 + [a.sub.obs]), [w.sub.0] = w(1 + [a.sub.obs]).

These wage levels are used both in the computation of deadweight losses as integrals and as triangles.
TABLE A4
Wage Regressions

 White Collars

 Wage Hourly Wage

 Observed Legal Observed Legal

Age 0.0775 0.0761 0.0805 0.0791
 0.0047 0.0044 0.0047 0.0045
[Age.sup.2] -0.0008 -0.0008 -0.0008 -0.0008
 0.0001 0.0001 0.0001 0.0001
Tenure 0.0348 0.0348 0.0349 0.0349
 0.0024 0.0023 0.0025 0.0024
[Tenure.sup.2] -0.0006 -0.0006 -0.0007 -0.0007
 0.0001 0.0001 0.0001 0.0001
Male 0.0681 0.0663 0.0294 0.0276
 0.0118 0.0115 0.0123 0.0120
Union 0.2291 0.2246 0.2597 0.2552
 0.0180 0.0175 0.0184 0.0180
Lima 0.5300 0.5160 0.5683 0.5543
 0.0114 0.0111 0.0119 0.0116
Constant 3.5407 3.5485 -0.3719 -0.3640
 0.0968 0.0914 0.0975 0.0925
[R.sup.2] 0.311 0.313 0.314 0.316

 Blue Collars

 Wage Hourly Wage

 Observed Legal Observed Legal

Age 0.0180 0.0178 0.0180 0.0180
 0.0045 0.0042 0.0044 0.0044
[Age.sup.2] -0.0001 -0.0001 -0.0001 -0.0001
 0.0001 0.0001 0.0001 0.0001
Tenure 0.0231 0.0226 0.0212 0.0212
 0.0026 0.0024 0.0025 0.0025
[Tenure.sup.2] -0.0003 -0.0002 -0.0002 -0.0002
 0.0001 0.0001 0.0001 0.0001
Male 0.2285 0.2216 0.2030 0.2030
 0.0138 0.0134 0.0126 0.0126
Union 0.1939 0.1910 0.2330 0.2330
 0.0166 0.0160 0.0164 0.0164
Lima 0.2226 0.2100 0.1966 0.1966
 0.0132 0.0126 0.0126 0.0126
Constant 4.3410 4.3331 0.6051 0.6051
 0.0847 0.0806 0.0829 0.0829
[R.sup.2] 0.300 0.309 0.315 0.315

Notes: Dummies for industrial sectors are used, but not
reported; values given in italics are standard errors.

TABLE A5
Sensitivity of Wage Elasticities of Demand: 0 with and without
Tenure as an Explanatory Variable in the Wage Regression

 Total Hours Number of Workers

 ln [w.sup.a.sub.p] ln [w.sup.a.sub.p]

Tenure With Without With Without

White-collar workers
 Observed -0.6451 -0.6245 -0.4982 -0.4615
 0.1882* 0.2185* 0.1949* 0.2244*
 Legal -0.5198 -0.4539 -0.3628 -0.2732
 0.1882* 0.2169* 0.1911* 0.2223*
Blue-collar workers
 Observed -2.2821 -2.8897 -2.3043 -2.8914
 0.388* 0.4629* 0.3806* 0.4555*
 Legal -2.2682 -2.9268 -2.3206 -2.9702
 0.3984* 0.4754* 0.3911* 0.4667*

 Individual Hours

 ln [w.sup.a.sub.p]

Tenure With Without

White-collar workers
 Observed -0.0448 -0.0561
 0.0046* 0.0054*
 Legal -0.0482 -0.0600
 0.0046* 0.0054*
Blue-collar workers
 Observed 0.0004 0.0043
 0.0118* 0.0137*
 Legal 0.0110 0.0190
 0.0124* 0.0147*

Notes: Values given in italics are standard errors.

Notes: Values given in italics are standard errors indicated with *.


REFERENCES

Auerbach, A. J., and J. J. Hines. "Taxation and Economic Efficiency," in Handbook of Public Economics, Vol. 3, Chapter 21, edited by A. J. Auerbach and M. Feldstein. Amsterdam: Elsevier, 2002, 1347-421.

Becker, G. A Treatise on the Family. Cambridge: Harvard University Press, 1993.

Bentolila, S., and G. Bertola. "Firing Costs and Labor Demand: How Bad is Eurosclerosis?" Review of Economic Studies, 57, 1990, 381-402.

Chacaltana, J. "Los Costos Laborales en el Pert]," in Inseguridad Laboral y Competitividad: modalidades de contratacion. Lima, Peru: Oficina Internacional del Trabajo, 1999, 205-84.

Clark, K. B., and R. B. Freeman. "How Elastic is the Demand for Labor?" Review of Economics and Statistics, 62, 1980, 509-20.

Freeman, R. B. "Single Peaked vs. Diversified Capitalism: The Relation Between Economic Institutions and Outcomes." NBER Working Paper No. 7556, National Bureau of Economic Research, 2000.

Hamermesh, D. "Labor Demand and the Structure of Adjustment Costs." American Economic Review, 79, 1989, 674-89.

--. Labor Demand. Princeton, NJ: Princeton University Press, 1993.

Heckman, J., and C Pages-Serra. "The Cost of Job Security Regulation: Evidence from Latin American Labor Markets." Economia, 1(1), 2000, 109-144.

Hopenhayn, H., and R. Rogerson. "Job Turnover and Policy Evaluation." Journal of Political Economy, 101, 1993, 915-38.

IPE. Peru: Costos no Salariales y Competitividad (estudio de actualizacion). Serie de Estudios, Estudio 1998-042. Lima, Peru: Instituto Peruano de Economia (IPE), 1998.

Jaramillo, M. La Regulacion del Mercado Laboral en el Peru. Informe de Consultoria. Lima, Peru: Grupo de Analisis para el Desarrollo, 2004.

MTPE. "Costos Laborales, Competitividad y Empleo en el Perth," in Boletin de Economia Laboral (BEL) No 11. Lima, Peru: Ministerio de Trabajo y Promocion del Empleo (MTPE), 1999, 2-15.

--. "Costos laborales en el Peril," in Boletin de Economia Laboral (BEL) No 28-29. Lima, Peru: Ministerio de Trabajo y Promocion del Empleo (MTPE), 2004, 1-32.

Marquez, G., and C. Pages-Serra. "Ties that Bind: Employment Protection and Labor Market Outcomes in Latin America." Working Paper No. 373, Research Department of Inter-American Development Bank, Washington, DC, 1998.

Nickell, S. "Dynamic Models of Labour Demand," in Handbook of Labor Economics, Vol. 1, Chapter 9, edited by O. Ashenfelter and E. R. Layard. Amsterdam: Elsevier, 1987, 473-522.

Rendon, S. "Job Creation and Investment in Imperfect Capital and Labor Markets." Working Paper No. E2004/35, Fundacion Centro de Estudios Andaluces, 2004.

Rendon, S., and R. Barreto. "La demanda laboral en la manufactura peruana." Documento de Trabajo No. 1, ADEC-ATC, 1992.

Roberts, M. J., and E. Skoufias. "The Long-Run Demand for Skilled and Unskilled Labor in Colombian Manufacturing Plants." Review of Economics and Statistics, 79, 1997, 330-34.

Saavedra, J. "La flexibilizacion del mercado laboral," in La Reforma Incompleta, edited by R. Abusada et al. Tomo I. Lima: Centro de Investigacion de la Universidad del Pacifico e Instituto Peruano de Economia, 2000.

Saavedra, J., and M. Torero. "Labor Market Reforms and Their Impact on the Formal Labor Demand and Job Market Turnover: The Case of Peru." Research Network Working Paper No. R-394, Inter-American Development Bank, Washington, DC, 2000.

GIOVANNA AGUILAR and SILVIO RENDON *

* We thank Juan Chacaltana, Cecilia Garavito, Daniel Hamermesh, Miguel Jaramillo, and participants of the LACEA-LAMES Meetings in Mexico, the Second Meetings of Labor Economics in Lima, the Seventh IZA/SOLE Transatlantic Meeting of Labor Economists at Buch am Ammersee, and seminars at the Universidad Catolica of Petal, GRADE, and ITAM for their comments and suggestions. We also thank Ramon Diaz for outstanding research assistance. Financial support of the Universidad Catolica of Peru and of the Asociacion Mexicana de Cultura is gratefully acknowledged. All errors and omissions are only ours.

Aguilar: Associate Professor, Departamento de Economia, Pontificia Universidad Cato1ica del Peru, Av. Universitaria cuadra 18 s/n, Lima 32, Per& Phone 511-626-2000--Anexo 4952, Fax 511- 626-2874, E-mail gaguila@pucp.edu.pe

Rendon: Assistant Professor, Economics Department, stony Brook University, stony Brook, NY 11794-0001. Phone 1-631-632-1422, Fax 1-631-632-7516, E-mail srendon@ms.cc.sunysb.edu

(1.) As referred by Heckman and Pages-Serra (2000, table 2 and footnote 6), many of the research projects that provide this empirical evidence (for instance, Marquez and Pages-Serra 1998 or Saavedra and Torero 2000) were evaluations of the labor reforms that reduced nonwage labor costs in Latin America and were sponsored by an Inter-American Bank's research network coordinated by Heckman and Pages-Serra.

(2.) In particular, for Peru, where our data come from, research done so far on the demand for labor finds very low labor costs elasticities of labor demand: -[[eta].sub.L,w] [member of] [0.10, 0.65] (Chacaltana 1999; IPE 1998; Jaramillo 2004; MTPE 1999, 2004; Rendon and Barreto 1992).

(3.) Another important source of underestimating the labor cost elasticity of employment is the presence of measurement error combined with the assumption of homogeneity of labor (Clark and Freeman 1980; Roberts and Skoufias 1997).

(4.) Unlike most studies, however, we use observed rather than legal nonwage labor costs and avoid thereby an overestimation of the labor cost employment elasticity.

(5.) Unfortunately, our data set does not include any variable that could proxy workers' education.

(6.) If [Z.sub.ij] in Equation (2) includes labor demand variables that are excluded in Equation (1) then it becomes a typical reduced form and one can attempt to identify not only the labor demand labor cost elasticity but also the labor supply labor cost elasticity. We leave this extension for future research.

(7.) Or, when employment is measured by individual hours of work, it is likely to find negative matching: more productive workers are matched to firms with fewer individual hours of work, as shown in Section IV.

(8.) In this article, we take undercompliance as given and only analyze its effects on employment and measurement of the labor cost elasticity; however, strictly speaking, undercomplying is also a decision made by employers, which therefore would require a specific theoretical and empirical analysis.

(9.) This is an important identification assumption; the labor supply is horizontal to the market, which makes feasible the determination of the labor cost elasticity of the labor demand. This assumption does not hold if the labor supply is horizontal to the firm, but not to the market, as aggregation does not preserve the horizontal labor supply.

(10.) When employment is measured by individual hours of work and, consequently, one is estimating an individual labor demand, there is no need to perform this aggregation.

(11.) These definitions are important because many other studies confuse mandatory benefits that workers take home, such as vacations payments, which here are considered as part of wages, with nonwage labor costs paid for by employers.

(12.) Thus, employers' and workers' contributions are nonwage labor costs. The former displace the demand and the latter the supply of labor. Unlike other studies, which only focus on employers' nonwage labor costs, both types of nonwage labor costs matter in reducing employment. Interestingly, from an economic point of view, as said before, if the labor supply is fully elastic, all nonwage labor costs are paid for by employers.

(13.) Actually, labor laws in Peru allow employers delays in paying workers' and employers' contributions and even payments in installments. Thus, the discrepancy between observed and legal contributions is only partially a matter of compliance. We follow the approach of most studies that just use the established legal percentages of contributions in the analysis of the effects of nonwage labor costs on employment.

(14.) The interested reader will find a brief explanation of the labor reforms in Peru in the 1990s in Appendix A1 and a detailed description of nonwage labor costs in Appendix A2.

(15.) Notice that these predicted employment variations assume a constant elasticity of demand, that is, they are a linear approximation of the employment effects and are thus less accurate, the larger the labor cost variation.

(16.) In Appendix A6, we provide some details on the computation of the deadweight losses of undercompliance with legal contributions.
TABLE 1
Descriptive Statistics for White- and
Blue-Collar Workers Firm and Worker Samples

Workers White Collar Blue Collar
Survey Section Firm Worker Firm Worker

Hours of work 44.5 46.1 42.9 45.8
 6.4 7.5 7.7 9.6
Employment 99.8 128.4
 300.9 297.5
Wages 542.2 605.0 232.9 246.2
 477.0 668.5 188.3 386.1
Employers' contributions (% of wage)
 Observed 10.2 10.3 10.7 10.9
 1.4 1.4 3.6 4.8
 Legal 14.5 14.5 14.5 14.5
 0.0 0.0 0.0 0.0
Workers' contributions (% of wage)
Observed 15.7 14.9 12.0 11.8
 4.8 5.5 2.3 2.7
Legal 17.8 18.0 13.1 13.3
 6.0 6.5 3.2 3.6
Economic sector
 Primary 6.4 4.8 10.2 13.7
 Industry 23.4 18.6 48.3 56.4
 Services 70.2 76.6 41.5 29.9
Firm size
 [less than or equal to] 50 62.2 30.9 50.9 19.7
 51-99 13.1 19.2 13.4 13.4
 [greater than or equal to] 100 24.7 49.9 35.7 66.9
Lima 48.6 55.6 42.8 46.9
 Metropolitan
Union 8.2 13.6 15.0 20.0
Women 37.4 14.0
Age 37.8 37.4
 10.2 11.1
 [less than or equal to] 24 4.1 8.4
 25-45 72.3 66.7
 >45 23.7 25.0
Tenure 6.1 6.0
 7.4 8.1
 <3 46.9 53.6
 3-8 29.2 22.7
 > 8 23.9 23.7
Number of 1,714 13,097 692 5,413
observations

Note: Values given in italics are standard errors.

TABLE 2
Estimated Wage Elasticities of Demand for Labor
Measured as Total Hours, Workers, and Individual Hours

 Total Hours

 [bar.ln
 ln [w.sup. [bar.ln [w.sup.a
 a.sub.m] [w.sup.a]] .sub.p]]

White-collar workers
 Observed
 [beta] 0.181 0.1659 -0.6451
 0.0519* 0.0532* 0.1882*
 [R.sup.2] 0.179 0.178 0.179
 Legal
 [beta] 0.1827 0.1683 -0.5198
 0.0519* 0.0532* 0.1882*
 [R.sup.2] 0.179 0.178 0.177
Number of observations 1714
Blue-collar workers
 Observed
 [beta] -0.13 -0.0555 -2.2821
 0.1217* 0.1326* 0.388*
 [R.sup.2] 0.183 0.181 0.23
 Legal
 [beta] -0.1322 -0.0553 -2.2682
 0.1249* 0.1382* 0.3984*
 [R.sup.2] 0.183 0.181 0.227
Number of observations 692

 Number of Workers

 [bar.ln
 ln [w.sup. [bar.ln [w.sup.a
 a.sub.m] [w.sup.a]] .sub.p]]

White-collar workers
 Observed
 [beta] 0.3233 0.2255 -0.4982
 0.0518* 0.0521* 0.1949*
 [R.sup.2] 0.212 0.202 0.196
 Legal
 [beta] 0.3258 0.2282 -0.3628
 0.0519* 0.0521* 0.1911*
 [R.sup.2] 0.212 0.202 0.194
Number of observations 1714
Blue-collar workers
 Observed
 [beta] 0.1406 -0.0319 -2.3043
 0.1184* 0.1314* 0.3806*
 [R.sup.2] 0.1838 0.182 0.234
 Legal
 [beta] 0.1511 -0.0311 -2.3206
 0.1215* 0.1368* 0.3911*
 [R.sup.2] 0.184 0.182 0.232
Number of observations 692

 Individual Hours

 ln [w.sup. ln [w.sup.
 a.sub.i] a.sub.ip]

White-collar workers
 Observed
 [beta] -0.0567 -0.0448
 0.0018* 0.0046*
 [R.sup.2] 0.118 0.047
 Legal
 [beta] -0.0569 -0.0482
 0.0018* 0.0046*
 [R.sup.2] 0.119 0.048
Number of observations 13097
Blue-collar workers
 Observed
 [beta] -0.0381 0.0004
 0.0041* 0.0118*
 [R.sup.2] 0.073 0.063
 Legal
 [beta] -0.0408 0.011
 0.0043* 0.0124*
 [R.sup.2] 0.073 0.063
Number of observations 5,413

Notes: Values given in italics are standard errors. Ln [w.sub.m],
log of the average firm-level wage (firms' sample); [bar.ln w],
firm-level average of log wage (workers' sample); [bar.ln
[w.sub.p]], average firm-level predicted log wage (workers'
sample); ln [w.sub.i], log of the individual wage (workers'
sample); ln [w.sub.ip], log of the predicted individual wage
(workers' sample).

Notes: Values given in italics are standard errors indicated with *.

TABLE 3
Employment Effects of Removing Employers',
Workers', and Both's Contributions

 Total Hours

 [bar.ln
 ln [w.sup. [bar.ln [w.sup.a
 a.sub.m] [w.sup.a]] .sub.p]]
White-collar workers
 Employer
 Observed -1.76 -1.62 6.26
 0.00* 0.00* 0.04*
 Legal -2.47 -2.27 6.99
 0.00* 0.01* 0.06*
 Worker
 Observed -3.29 -2.72 10.51
 0.02* 0.02* 0.29*
 Legal -3.85 -3.39 10.42
 0.03* 0.02* 0.24*
 Employer and worker
 Observed -5.04 -4.34 16.76
 0.02* 0.02* 0.29*
 Legal -6.31 -5.66 17.42
 0.03* 0.02* 0.25*
Blue-collar workers
 Employer
 Observed 1.33 0.57 23.45
 0.02* 0.02* 0.98*
 Legal 1.78 0.75 30.61
 0.03* 0.03* 0.29*
 Worker
 Observed 1.67 0.7 28.84
 0.03* 0.03* 0.81*
 Legal 1.87 0.8 32.69
 0.03* 0.04* 0.92*
 Employer and worker
 Observed 2.99 1.27 52.29
 0.03* 0.03* 1.27*
 Legal 3.66 1.54 63.3
 0.13* 0.15* 2.06*

 Number of Workers

 [bar.ln
 ln [w.sup. [bar.ln [w.sup.a
 a.sub.m] [w.sup.a]] .sub.p]]
White-collar workers
 Employer
 Observed -3.14 -2.2 4.86
 0.00* 0.00* 0.04*
 Legal -4.4 -3.08 4.9
 0.00* 0.00* 0.07*
 Worker
 Observed -5.87 -3.69 8.16
 0.05* 0.03* 0.21*
 Legal -6.86 -4.59 7.3
 0.07* 0.03* 0.20*
 Employer and worker
 Observed -9.01 -5.89 13.02
 0.05* 0.03* 0.23*
 Legal -11.26 -7.67 12.19
 0.07* 0.03* 0.21*
Blue-collar workers
 Employer
 Observed -1.43 0.33 23.68
 0.02* 0.02* 0.99*
 Legal -2.04 0.42 31.32
 0.03* 0.03* 0.28*
 Worker
 Observed -1.8 0.4 29.12
 0.02* 0.03* 0.82*
 Legal -2.14 0.45 33.44
 0.03* 0.04* 0.93*
 Employer and worker
 Observed -3.24 0.73 52.8
 0.03* 0.03* 1.28*
 Legal -4.18 0.87 64.76
 0.12* 0.14* 2.06*

 Individual Hours

 ln [w.sup. ln [w.sup.
 a.sub.i] a.sub.ip]
White-collar workers
 Employer
 Observed 0.55 0.44
 0.00* 0.00*
 Legal 0.77 0.65
 0.00* 0.00*
 Worker
 Observed 0.93 0.73
 0.00* 0.00*
 Legal 1.15 0.97
 0.00* 0.00*
 Employer and worker
 Observed 1.48 1.17
 0.00* 0.00*
 Legal 1.91 1.62
 0.00* 0.00*
Blue-collar workers
 Employer
 Observed 0.39 0
 0.00* 0.00*
 Legal 0.55 -0.15
 0.00* 0.00*
 Worker
 Observed 0.48 0
 0.00* 0.00*
 Legal 0.59 -0.16
 0.00* 0.00*
 Employer and worker
 Observed 0.87 -0.01
 0.00* 0.00*
 Legal 1.14 -0.31
 0.00* 0.00*

Notes: Values given in italics are standard errors. Ln [w.sub.m],
log of the average firm-level wage (firms' sample); [bar.ln w],
firm-level average of log wage (workers' sample); [bar.ln
[w.sub.p]], average firm-level predicted log wage (workers'
sample); ln [w.sub.i], log of the individual wage (workers'
sample); ln [w.sub.ip], log of the predicted individual wage
(workers' sample).

Notes: Values given in italics are standard errors indicated with *.

TABLE 4
Estimated Deadweight Loss of Employers', Workers', and Both's
Contributions as a Percentage of Contribution Revenues

 Total Hours Number of Workers

Contribution Integral Triangle Integral Triangle

White-collar workers
 Employer
 Observed 3.14 3.23 2.44 2.50
 Legal 3.50 3.62 2.43 2.51
 Worker
 Observed 5.59 5.90 3.88 4.06
 Legal 5.15 5.47 3.41 3.58
 Employer and worker
 Observed 8.77 9.51 6.31 6.76
 Legal 8.64 9.49 5.79 6.27
Blue-collar workers
 Employer
 Observed 12.51 13.35 12.33 13.13
 Legal 17.02 18.39 16.59 17.91
 Worker
 Observed 15.77 17.02 15.85 17.12
 Legal 18.26 19.96 18.65 19.19
 Employer and worker
 Observed 30.37 34.94 30.25 34.84
 Legal 38.55 45.69 37.21 43.96

 Individual Hours

Contribution Integral Triangle

White-collar workers
 Employer
 Observed 0.22 0.22
 Legal 0.32 0.33
 Worker
 Observed 0.36 0.37
 Legal 0.47 0.49
 Employer and worker
 Observed 0.56 0.59
 Legal 0.77 0.82
Blue-collar workers
 Employer
 Observed 0.00 0.00
 Legal -0.01 -0.01
 Worker
 Observed 0.00 0.00
 Legal -0.08 -0.08
 Employer and worker
 Observed 0.00 0.00
 Legal -0.15 -0.15

TABLE 5
Employment Effects of Undercomplying with
Legal Nonwage Labor Costs

 Total Number Individual
Contribution Hours of Workers Hours

White-collar workers
 Employer 2.40 1.86 0.17
 0.01 0.01 0.00
 Worker 2.40 1.86 0.17
 0.13 0.08 0.00
 Employer and worker 4.79 3.72 0.33
 0.13 0.08 0.00
Blue-collar workers
 Employer 7.35 7.42 0.00
 0.70 0.71 0.00
 Worker 4.05 4.09 0.00
 0.81 0.82 0.00
 Employer and worker 11.40 11.51 0.00
 1.07 1.09 0.00

Note: Values given in italics are standard errors.

TABLE 6
Deadweight Loss of Complying with Employers', Workers', and
Both's Legal Contributions as a Percentage of Contribution
Revenues

 Total Hours Number of Workers

Contribution Integral Triangle Integral Triangle

White-collar workers
 Employer 1.21 1.24 0.89 0.91
 Worker 0.93 0.98 1.16 1.19
 Employer and worker 2.13 2.20 2.05 2.11
Blue-collar workers
 Employer 4.02 4.20 3.94 4.10
 Worker 1.98 2.08 2.22 2.31
 Employer and worker 6.01 6.30 6.22 6.50

 Individual Hours

Contribution Integral Triangle

White-collar workers
 Employer 0.08 0.08
 Worker 0.08 0.08
 Employer and worker 0.16 0.17
Blue-collar workers
 Employer 0.00 0.00
 Worker 0.00 0.00
 Employer and worker 0.00 0.00
COPYRIGHT 2010 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Aguilar, Giovanna; Rendon, Silvio
Publication:Economic Inquiry
Geographic Code:1USA
Date:Jul 1, 2010
Words:10721
Previous Article:Home computers and educational outcomes: evidence from the NLSY97 and CPS.
Next Article:The economic roots of the American "zigzag": knives, forks, and British mercantilism.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters