Who Benefits from Labor Market Regulations?1 Chile 1960-1998 Claudio E. Montenegro The World Bank and Carmen Pagés Inter-American Development Bank Abstract2 Economists have examined the impact of labor market regulations on the level of employment. However, there are many reasons to suspect that the impact of regulations differs across types of workers. In this paper we take advantage of the unusually large variance in labor policy in Chile to examine how different labor market regulations affect the distribution of employment and the employment rates across age, gender and skill levels. To this effect, we use a sample of repeated cross-section household surveys spanning the period 1960-1998 and measures of the evolution of job security provisions and minimum wages across time. Our results suggest large distribution effects. We find that employment security provisions and minimum wages reduce the share of youth and unskilled employment as well as their employment rates. We also find large effects on the distribution of employment between women and men. JEL Classification Codes: E24, J23, J65 Keywords: Employment, Employment Regulations, Chile. World Bank Policy Research Working Paper 3143, October 2003 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. 1We thank the University of Chile for giving us permission to use their data. Introduction The economic literature has devoted considerable attention to studying the impact of labor market regulations on labor market outcomes. However, the issue of whether some sub-groups of workers bear the brunt or enjoy the benefits of such regulations has been much less studied.3 One notable exception has been the burgeoning literature studying the effect of statutory minimum wages on youth employment. Although this subject remains controversial, many studies have found negative effects of minimum wages on teenagers and young workers.4 Less attention has been paid to the issue of whether minimum wages particularly affect women or men or unskilled versus skilled workers. Similarly, very little attention has been paid to the effect that job security provisions may have on particular sub-groups of the labor force. In this paper, we take advantage of the unusual variance in labor market policies in Chile to examine how minimum wages and job security provisions affect different types of workers. To this effect, we use a sample of repeated household surveys spanning the period 1960-1998 and several measures of labor market regulations across time. We make use of cross-section and time-series methods to estimate the effect that these policies have on the distribution of employment and on particular sub-groups' employment rates. To assess whether our estimates are reflecting the effect of regulations instead of the effect of some unobservable correlates, we also estimate the effect of labor policy on sectors not covered by regulations. We find large and statistically significant effects on the covered sectors and no effects, or effects going in the opposite direction, on the uncovered sectors. Our results indicate that labor market regulations are far from neutral. We find that job security provisions and minimum wages reduce the employment rates of youth and the unskilled at the benefit of older and skilled workers. We also find opposite effects of these policies on women's and men's employment shares and rates. Job security provisions tend to benefit men at the expense of women, while the reverse seems to be true for an increase in the minimum wage. We then explore some explanations for these regularities and, while we cannot fully discriminate among all of them, we are at least able to reject some hypotheses. There is little 3One reference in this literature is the paper by Bertola, Blau and Kahn (2002) on the effect of unions' involvement in wage setting on the relative employment of youth, women and older individuals. 4Among the most recent studies, Williams and Mills (2001), Partridge and Partridge (1998) and Bazen and Skourias (1997) find a negative relation between minimum wages and youth employment, while Katz and Krueger (1992), Card, Katz and Krueger (1994) and Card and Krueger (2000) find no evidence of such an effect. 2 evidence that these differential effects are driven by differences in labor supply elasticities or wage adjustments across sub-groups. Instead, our findings suggest that labor market regulations produce unequal shifts in labor demand across groups of workers. The rest of the paper is organized as follows. Section 2 reviews the arguments that predict non-neutral effects of regulations. Section 3 describes the evolution of job security and minimum wage regulations in Chile. Section 4 describes the data used in our empirical section. Section 5 describes the methodology implemented to estimate the effects of regulations on the distribution of employment. Section 6 describes our results for both the distribution of employment and the overall effect on employment rates. Finally, Section 7 concludes. 1. Why Regulations May Affect Some Workers Differently There are a number of reasons to suspect that labor market regulations alter the distribution of employment across sub-groups. In the next two subsections we review the theoretical arguments that predict differential effects of job security provisions and minimum wages across workers of different age, skill level and gender. 1.1 Job Security Job security provisions are introduced to discourage firms from adjusting their labor forces in the face of adverse economic conditions. However, job security provisions also alter hiring decisions. In good times, firms hire fewer workers because they take into account that these workers may have to be laid off in the future, and that is costly. The overall impact of job security provisions on employment rates is undetermined because it depends on whether the negative effect on layoffs is offset by the reduction in hiring rates.5 Job security provisions will have differential effects across sub-groups of workers if changes in legislation bring changes in hiring and layoff rates that have a larger impact on some sub-populations than on others. Lazear (1990) conjectured that an increase in job security might act as a barrier preventing the entry of young workers into the labor market. This is because job security reduces job creation, and entry rates are especially high among youth. This argument, however, does not consider that the effect of lower job creation rates can be offset by lower job 3 destruction rates--which also tend to be large among youth. Pagés and Montenegro (1999) suggest an argument whereby job security provisions may actually increase young workers' layoff rates. Their argument is related to the regularity that, across countries, job security is positively related with a worker's tenure. Mandatory severance payments that increase with tenure change the cost of dismissing workers with short tenures relative to workers with more seniority at the firm. In this context, it is expected that job security concentrates layoffs among youth because, other things being equal, young workers tend to have lower average tenures than older workers. If severance pay increases substantially with tenure and this effect is important, job security simultaneously reduces entry and increases layoffs among youth, resulting in a lower employment share and lower employment rates for this group of workers. Instead, the share of older workers in employment tends to increase due to their relatively lower layoff rates. Similar reasoning can be used to predict the effect of job security provisions across gender. To the extent that women experience higher rotation and therefore have lower average tenure than males at every age, high job security will tend to concentrate layoffs among women. This effect will tend to reduce their employment share relative to men. However, higher turnover rates also imply that stringent job security may be less of an issue when hiring female workers because employers expect them to quit prior to attaining high job security.6 In this case, employers might be more willing to hire women relative to men, but also more likely to lay them off should bad times arise. The overall effect on female versus male employment rates is undetermined and remains an empirical issue. It is tempting to extend the former argument to unskilled and skilled workers. If unskilled workers have higher rotation and lower tenures than skilled workers, the same reasoning applies. However, while higher female turnover rates may be motivated by life-cycle decisions exogenous to the employer, such exogeneity is more difficult to claim when explaining the higher rotation of unskilled workers. The insider-outsider literature provides further arguments for why job security may have a differential effect on the employment rates of different sub-populations.7 According to this 5See Bertola (1990), Bentolila and Bertola (1990), Bertola (1991), Bentolila and Saint-Paul (1994), Hopenhayn and Rogerson (1993) and Risager and Sorensen (1997) among others for a theoretical discussion of the effects of job security on employment rates. 6See Pagés and Montenegro (1999) for a more formal development of this argument in the context of a partial equilibrium model. 7See for instance, Lindbeck and Snower (1988). 4 literature, more stringent job security reduces the elasticity of wages to changes in the unemployment rate. When employed workers know their jobs are insured against demand fluctuations, they may be less willing to accept the wage adjustments necessary to reduce unemployment rates. This situation may help to create two kinds of workers: insiders, who hold their jobs and have high wages, and outsiders, who either are unemployed or hold temporary, part-time or fixed terms jobs without job security.8 If women, the young and the unskilled are more likely to be outsiders, then job security (through this wage effect) will bias employment against these groups. Finally, differences in labor supply elasticity may contribute to differential effects across sub-populations even if job security brings a uniform change in labor demand across groups. Let us assume that an increase in job security reduces labor demand. If women, the young, and the unskilled have higher labor supply elasticity than the average worker, higher job security would bring a higher decline in employment for these workers than for other groups with a lower elasticity of labor supply.9 Summarizing, the arguments put forth in this section suggest that youth, and possibly women and the unskilled, bear the brunt of job security regulations. 1.2 Minimum Wages The effect of minimum wages on employment remains a controversial topic. In the competitive model, workers are paid their marginal product, and any artificial increase in the price of labor above the marginal product therefore prices the worker out of the labor market. Conversely, models based on some form of imperfect competition predict wages lower than the marginal product, and thus, an increase in minimum wages can increase wages without reducing employment rates.10 On average, youth, women, and the unskilled tend to have lower wages than older, male or skilled workers. Therefore, since minimum wages are more likely to be binding among these workers, the competitive model predicts larger unemployment effects for the first group. In the 8 The insider-outsider argument requires a strong union fixing wages for new entrants. Otherwise, firms could always pay very low wages at the beginning of the employment relationship to compensate for higher wages in the future. See Bertola (1990) for an analytical study of this issue. 9See Hamermesh (1993). 10There are many situations that give raise to imperfect competition in the labor market, such as monopolistic power on the part of employees, incomplete information or imperfectly mobile workers. 5 imperfect competition model, however, the effects are less clear-cut. In principle, the magnitude and sign of the minimum wage effect will depend on how far wages are from their respective marginal products in each sub-population. If that gap is larger in some groups than in others, an increase in minimum wages may have "competitive" effects on some groups and "non- competitive" effects on others. Given this ambiguity, the sign and magnitude of the effects become an empirical question. 2. Labor Market Regulations in Chile Chile has experienced a very wide range in labor market policies, providing a privileged case scenario for analyzing the impact of regulations on labor market outcomes. We distinguish between job security provisions and statutory minimum wages.11 2.1 Job Security Provisions Among the most interesting aspects of the Chilean experience is that, in the 39 years covered by our sample, Chile has gone from a situation of dismissal at will to a rigid labor market by OECD standards.12 Since their inception in 1966, job security provisions have favored full-time indefinite employment over part-time, fixed-term or temporary contractual relationships. To this end, in case of a firm-initiated separation, labor codes regulate (1) compulsory advance notice periods, (2) the causes for which a dismissal is considered justified or unjustified and (3) severance pay related to the tenure of a worker and the cause of dismissal. While the minimum period of advance noticed has always been kept constant and equal to one month, the formula for computing severance pay and the causes for just or unjust dismissal have varied widely over the years. This is the variance that we exploit in our empirical work. Table 1 summarizes the changes in legislation that took place in the 1960-1998 period. From 1960 to mid-1966, firms had to provide a one-month advance notice (or pay the equivalent of one month of salary) but otherwise "employment at will" was the norm. In 1966, the Congress approved a new law under which firms had to pay compensation equal to one month's wage per year of work to all workers dismissed without just cause. The economic needs of the firm were considered a just cause in the law and therefore a worker dismissed for this reason 11See Edwards and Cox-Edwards (2000) for an excellent summary of labor market reforms in Chile during the 1960-2000 period. 6 would not qualify for severance pay. In practice, however, workers would appeal to courts, and judges tended to consider these dismissals unjustified.13 In that event, the employer could choose between paying the mandatory compensation--plus wages foregone during trial--or reinstate the worker in his/her old post. This reform substantially increased the difficulty and the cost of labor force adjustments. After 1973, a violent change in political regime brought about a de facto liberalization. Although job security provisions were not modified in the law, in practice, it was more likely that judges ruled against workers, effectively reducing dismissal costs. In 1989 and 1981, successive modifications reduced the cost of dismissal under the law. In 1981, the maximum amount to be awarded to a worker dismissed without just cause was reduced to the equivalent of five months' pay. This reform substantially reduced the cost of dismissal, particularly for workers with long tenures, although it only applied to newly hired workers. After 1984, the tide shifted and job security provisions became progressively stricter. In December of that year, the law was modified to exclude economic needs of the firm as a justified cause of dismissal. However, the maximum amount payable to a worker was kept at five months of pay. In 1990, after the return of democracy, a new labor reform still in force further increased the cost of dismissal. This law considers dismissals motivated by the economic needs of the firm justified, but employers are still liable to pay compensation equal to one month's pay per year of work with a maximum amount of 11 months of pay. It is the responsibility of the firm to prove just cause. If such causality cannot be demonstrated, there is a 20 percent surcharge in the amount of compensation. We summarize this variance in law and court practice by means of a job security measure derived in Pagés and Montenegro (1999).14 This measure is computed as follows: T JSt = i (1- )(bt+i + atSPt+i + (1- at )SPt+i ) i-1 jc uc i=1 where is the probability of remaining in a job, is the discount factor, T is the maximum tenure that a worker can attain in a firm, bt+i is the advance notice to a worker that has been i 12Heckman and Pagés (2000). 13Romaguera, Echavarría and González (1995). 14See the mentioned paper and Heckman and Pagés (2000) for a complete description of the methodology used, how it is applied across time and countries and the relative advantages and costs of using this measure versus other measures of job security. 7 years with a firm, at is the probability that the economic difficulties of the firm are considered a justified cause of dismissal, SPt+1 is the mandated severance pay in that event to a worker that jc has been i years at the firm, and finally, SPt+1 denotes the payment to be awarded to a worker uc with tenure i in case of unjustified dismissal.15 This measure computes the expected cost, at the time a worker is hired, of dismissing this worker in the future. This cost is measured in terms of monthly wages. The advantage of this measure in respect to other measures that compute the cost conditional on having achieved a certain tenure is that our job security measure captures the whole profile of severance pay at each level of tenure. The assumption is that firms evaluate future dismissal costs based on current law. Higher values of this variable indicate periods of relatively high job security, whereas lower values characterize periods in which dismissals were less costly. Based on the legal information summarized in Table 1 and assumptions regarding , , a, and T, we obtain a measure of JS. We take to be a constant value such that the average real interest is equal to 8.4 percent, which corresponds to the average real interest rate in Chile during the 1960-1998 period. The discount rate is computed based on the assumption that without job security, turnover rates in Chile would be comparable to those observed in the US.16 Davis and Haltiwanger (1992) report an average annual turnover rate of 12 percent. The probability that a dismissal originated by the economic needs of the firm will be considered just depends on whether the law says so and whether labor judges rule so if workers take firms to court. For the period 1966-1984, although economic needs of the firm were considered just cause in the law, we assume a to be larger than zero and determined by the position taken by labor courts. Finally, we assume T = 25. See Table 2 for a complete description of the parameters used in the computation of the JS measure. The evolution of this variable over time is depicted in Figure 1. After some years of relatively low employment protection, JS increases eight-fold after the introduction of compulsory severance pay in the law. Expected dismissal costs decline markedly in 1973 and then successively in 1978 and 1981. Subsequently, employment protection increases again but without reaching the levels attained during the late 1960s. 16Although turnover rates can be measured, this measure is itself affected by labor law. Given this endogeneity, we choose instead to use the U.S. turnover rate, since it is well established that dismissal costs in the U.S. are very small. 8 2.2 Minimum Wages Columns (2) and (3) in Table 3 present the hourly real minimum wage in 1998 pesos; these indices were constructed using Chile's Central Bank Bulletins.17 It is interesting to note that since 1989 there has been a lower minimum wage for workers 18 years old or younger. This wage has been fixed at a level between 15 and 20 percent of the adult wage. Figure 2 summarizes the evolution of the minimum wage in relation to the average wage for teen and adult workers. The figure shows that minimum wages are much higher, relative to each group average rate, for teen than for adult workers. It also shows that the level of teen minimum wages has been quite volatile relative to the average wage. Between 1960 and 1998, adult real minimum wages increased by 186 percent and teen minimum wages by 104 percent. However, because average ages rose by more, minimum wages lost ground in relation to the average wage. Despite this long-term secular trend, Chile experienced a wide range of fluctuations in minimum wages, both in its rate of growth (in real terms) and in its level in relation to the average wage. During the 1960s, the real value of minimum wages was held constant, but since real wages increased, the ratio of the minimum to the average real wage declined. In the early 1970s, minimum wages increased substantially, surpassing the growth rate of average wages. In consequence, the ratio of the minimum to the average real wage increased sharply in that period. From 1975 to 1980 minimum wages lost ground relative to the average wage. After the return to democracy in 1990, real minimum wages increased steadily, but they continued declining relative to the average wage. The decline was particularly sharp for the teen group, whose minimum to average real wage rate fell from 1.80 in 1975 to 0.50 in 1998. It is interesting to note that while there are several studies in the Chilean case that suggest that the minimum wage is binding, others such as Bravo and Vial (1997) suggest that it is not.18 17Per hour minimum wages are constructed as monthly minimum wages divided by 4.2*40 hours. 18 See, for instance, Castañeda (1983), Paredes and Riveros (1989), Montenegro (2002) and Cowan, Micco, Mizala et al. (2003). An excellent review of the impact of minimum wages in the case of the United States can be found in Kosters (1996). A more recent survey on the international evidence of minimum wages can be found in Dowrick and Quiggin (2003). 9 4. Data The household surveys used in this study were obtained from the University of Chile's Economics Department. The Economics Department's Survey monitors the employment- unemployment status in the metropolitan area of Santiago de Chile four times a year. Unfortunately, only the surveys taken in June of each year contain information about wages and other employment status variables. Therefore, these are the surveys used in this study. The format of the survey and the definition of the variables have been kept constant since 1957, when the survey started, and so the information contained in them is comparable across years.19 During the period 1960 to 1998, the surveys interviewed between 10,000 and 16,000 people, and around 3,700 and 5,400 active labor force participants each year. During this period, the Metropolitan Area of Santiago de Chile represented about one third of Chile's total population, and a higher proportion of GDP.20 The data set is formed by stacked cross-sectional data sets, which means that individuals are not followed over time. The only restriction applied to our sample is that the people included in the estimates must be at least 15 years old and no older than 65. We merge labor policy and macro variables taken at the annual frequency with our individual-level annual data. We include the job security index and the minimum wage data described in Section 3. We also include a measure of wage bargaining to control for changes in union activity that can be correlated to our variables and to employment. While perhaps the best measure of the influence of unions on wage determination is union coverage, that is, the share of workers whose wages are affected by collective bargaining, a time series of this nature does not exist in Chile. Since union membership is also not available for all years covered in our sample, we measure unions' bargaining power by means of an index that reflects the degree of centralization of collective bargaining constructed by Edwards and Cox-Edwards (2000). This variable takes values from 1 (total decentralization) to 4 (total centralization). The use of this measure is based on the observation that union coverage tends to be larger in countries where collective bargaining is centralized. Finally, we include as a measure of economic activity deviations with respect to potential GDP. To obtain this variable, we use GDP data from the World Bank and apply a Hodrick-Prescott filter to obtain trend GDP. 19In this study we use data from 1960 on, because the previous years (1957-1959) do not have reliable data. 20According to the 1992 Census, the metropolitan area accounted for 39 percent of the total population. 10 Table 3 summarizes some basic statistics of our sample, by year. The first three columns display the value of the job security index and the real minimum wage for people 18 or younger and for adult workers. The next two columns summarize the index of bargaining (column (4) presents the original index, and column (5) presents the smoothed index). The evolution of these variables over time is depicted in Figure 5. Higher values of this measure, like those registered from 1960 to 1970, reflect periods of higher union centralization.21 The next seven columns summarize the average hourly wage broken down by sex (columns (6) and (7)); skill level (columns (8) and (9)) and age group (columns (10), (11) and (12)). Column (13) summarizes the deviation of the GDP from its potential or trend value. Finally, columns (14), (15) and (16) present the percentage of total people employed, the percentage of people that work for someone else (wage employment), and the percentage of people self-employed as a proportion of total population between 15 and 65 years old. These three rates are also depicted in Figure 3, which jointly with Figure 4 (which shows GDP deviations from its trend), illustrates the violent swings experienced by the Chilean economy during the 1960-1998 period, and in particular between 1970 and 1985.22 Some additional indicators describing the performance of the Chilean economy are summarized in Table 4. 4. Methodology To estimate the differential impact of labor market regulations across sub-populations we assume that the employment status of an individual is characterized by y*ijt = Xit*1+X'it*Zt*2+ t + ijt (1) where yijt =1 if y*ijt > 0 yijt =0 otherwise and y*ijt is an unobservable variable that determines whether an individual i, in sub-population j, at time t will be employed or not, and yijt is the observable employment status of this individual. 21Although not shown in the results, we checked the robustness of our results using the strikes index constructed by Edwards and Cox-Edwards (2000) instead of the centralization index. The results were invariant to different specifications. 11 In addition, Xit is a vector of variables that summarizes the personal characteristics of the individual i at time t, Zt is a vector of variables that vary with t, t is a year fixed effect and ijt is an error term. Among the personal characteristics we include age, gender, skill level, number of children and number of children interacted with gender. In some specifications, we also include age interacted with gender, and age interacted with skill to capture differential effects of age across gender and skill groups. Given the number of observations available, we divided the data into three age groups (15-24, 25-50, and 51-65) and two skill levels (9 years of education or less, and more than 9 years). Adding the skill and the age groups to the gender division, we have 12 different sub-populations, j=1,...12 In the vector of aggregate variables Zt we include the index of job security, deviations from GDP trend and the union centralization variable (all in logarithms). We also include the minimum wage index (also in logarithms), but we let it change for individuals 18 and younger. By construction, the vector of coefficients on the interaction of Xit and Zt,, 2, gives the sign of the differential effect. In addition, assuming that the Prob(y*ijt > 0) is distributed as a standard normal distribution, the size of the marginal differential effect is given by (.)Xit2, where (. ) is the normal density function. Our original intention was to estimate y*ijt = Xit*1+X'it*Zt*2 + Zt*3 + ijt (1') With such a specification we could recover the total marginal effect of a labor policy on sub-population j as (.)(Xit2 +3). However, despite finding robust estimates for the differential effects, our estimates for the level effect (3) proved to be extremely sensitive to the set of variables included in Zt., suggesting that our time variables did not properly account for the time variation of the series. In view of these results, we opted for estimating specification (1). This estimation still allows us to compute marginal effects, but the total effects are now absorbed by the constant term. Therefore, we can measure the impact of labor market regulations on the distribution but not on the level of employment. 22Chilean economic performance has been extensively documented by Edwards and Cox-Edwards (1991, 2000), de la Cuadra and Hachette (1994), Wisecarver (1992), Bosworth, Dornbusch and Laban (1994), Hudson (1994), Soto (1995), and Cortazar and Vial (1998). 12 Although specification (1) is a reduced form equation, in some cases, it will be useful to add a measure of wages. To construct this variable, wijt, we assign to all workers i j, j=1,..,12, at period t, the average wage of all employed workers in group j at period t. We minimize the risk of omitted variable biases and spurious correlations in five ways: First, by using individual data from a series of stacked household surveys to estimate specification (1), we can control for changes in the relative size of the population of each group and changes in fertility which, if omitted, could bias our estimates. Second, by introducing time dummies, we control for macroeconomic trends and cycles as well as policy changes that affect the overall population. Third, by controlling for effects of changes in the business cycle (using GDP deviations from its trend) across individuals (that is, including X'it* Zt , where Zt contains the business cycle variable) we can partially control for changes in policy and institutions that are endogenous to changes in relative employment. This is because such movements are likely to be correlated with changes in the business cycle. Fourth, by estimating the differential effect of policy while including contemporary labor market policies and institutions, we make sure that our measured effects are not biased by the correlation between these variables and the distribution of employment. Lastly, by comparing the estimated effects on the probability of wage employment (which is covered by labor policy) with the results on self-employment (which is not covered), we assess whether we are capturing the effect of policy, or instead, the effect of some unobservable correlate. 6. Empirical Results 6.1 The Effect of Job Security on the Distribution of Employment Our results indicate that job security provisions have a differential impact across demographic sub-groups. In Table 5, we report the results of estimating our empirical specification (1) assuming normality in the distribution of errors. The reported numbers correspond to the coefficients of the probit model, while the marginal effects for selected sub-populations of workers are reported in Table 6. The t-tests, reported next to the coefficients, are robust to the presence of heteroskedasticity of unknown kind using the White (1980) method. Most coefficients on the individual characteristic variables exhibit the expected patterns: female and older workers are less likely to be employed than prime-age (26-50) men. Additionally, the number of children per father increases the probability of being employed, and the number of 13 children per mother decreases the probability of being employed. Instead, the coefficients on the variable young and unskilled change signs across specifications. In column (1) we report the results of interacting the JS measure with dummies for age (young and older), gender (women) and skill level. A negative (positive) sign indicates that periods of more stringent JS provisions are associated with a decline (increase) in the probability of employment of a particular sub-population relative to the omitted category. We find strong age effects. The coefficient on the young-JS interaction is negative and statistically significant, while the coefficient on the older-JS interaction is positive although not statistically significant. Our results suggest that high job security tends to bias the distribution of employment against younger workers. We also find significant effects across the skill divide. The coefficient on the unskilled-JS interaction is negative and statistically significant, suggesting that JS provisions reduce the probability of employment of unskilled workers relative to skilled ones. Lastly, the coefficient on the female-JS interaction suggests a negative effect of JS on the probability of employment of women relative to men. Column (2) shows the results once we control for the evolution of the minimum wage, union activity and deviations of GDP with respect to its trend, as well as interaction of these variables with age, gender and skill dummies. The only difference with respect to column (2) is that the coefficient on the dummy for older workers is now somewhat larger and statistically significant at the 10 percent level, suggesting that job security provisions benefit older workers relative to prime-age ones. In columns (3) and (4) we report the coefficients resulting from estimating the same specification for wage-employment and self-employment separately. Our results are encouraging since they suggest that our findings are driven by policy changes instead of by some unobservable factors correlated with labor policy and employment. The signs and magnitudes of the coefficients for total and wage-employment are very similar, except for the coefficients on women. Instead, for self-employment the coefficients are either not statistically different from zero or going in the opposite direction than for wage-employment. This is the case with the coefficients on the gender and unskilled variables, which suggests that more stringent JS regulations increase the probability that women and the unskilled are employed in the self- employment sector relative to men and the skilled. Column (5) exhibits the results once we allow for further interactions between age, skill and gender groups. With this finer level of disaggregation we can examine whether the impact of 14 job security is the same across young men and young women, or across young skilled and unskilled workers. These additional variables not only provide a more complete description of the effects of JS on the distribution of employment, but also help to infer the channels through which JS affects that distribution. The coefficients for these additional interaction variables are all statistically significant, and a test for their joint significance strongly rejects the null hypothesis of all the coefficients being zero. The estimates in column (5) contain some interesting additional information relative to the estimates in column (1)-(4) We find that an increase in JS tends to reduce the employment probabilities of young men relative to those of young women. However, we also find that this effect is reversed at older ages. Thus, JS provisions seemingly reduce the probabilities of employment of middle-aged and older women relative to those of men in that same age group. Our estimates also suggest that an increase in JS provisions reduces the probability of employment of both skilled and unskilled youth, but the effect is larger for unskilled youth. Finally, column (6) reports the results of estimating the same specification as in column (5) but controlling in addition by the average wage of each sub-population group, in period t. Controlling for the wage level of each group allows us to assess whether some of the observed effects are driven by differences in wage adjustment across sub-populations. Yet, the results should be taken with caution because some wage movements may be endogenous to the probability of employment. Overall we find that holding wages constant does not affect our main results. The only coefficient that changes size and significance is the interaction between young unskilled and job security. Holding wages constant reduces the coefficient and the significance of the effect on unskilled youth (relative to more skilled youth). Instead, most of the other coefficients become larger (in absolute value) than the ones reported in column (5). This suggests that more stringent regulations are partly paid by workers in the form of lower wages. In light of the different theories described in Section 2, how do we explain the results presented above? Although we cannot totally discriminate among different theories, we are at least able to reject some hypotheses. The fact that most of our results remain unchanged when wages are included suggests that the differential effects presented above cannot be explained by differences in the elasticity of labor supply across demographic groups. The only exception is the larger effect on young unskilled workers, which seems to be driven by a higher labor supply 15 elasticity of this group.23 Our results also suggest that these differential effects cannot be explained by insider-outsider theories, since in that case the effect would also be through wages. Instead, our results suggest that the differential effects on employment are demand driven: Changes in job security provisions bring about changes in hiring and firing rates that selectively affect different types of workers. A barrier-of-entry effect can explain the negative impact of job security on the employment rates of young workers relative to other demographic groups. However, it cannot account for the estimated differences in impact between young women and young men. One possible way to explain these findings is to consider differences in turnover rates across groups. As discussed in Section 2, a higher exogenous turnover rate can bring about two effects. On the one hand, workers with a higher propensity to rotate have lower average tenures and therefore are more likely to be laid off in bad times. On the other hand, higher rotation reduces expected severance payments and therefore increases the incentives to hire these workers. Consequently, higher rotation among women can explain why JS provisions affect young women less than young men. It can also explain why middle-aged and older women benefit less from JS than men of the same age. Differences among turnover rates could also partially explain the results for skilled and unskilled workers. Higher rotation among the unskilled would imply lower tenure rates and higher probabilities of dismissal for middle-aged and older unskilled workers relative to more skilled ones. This is consistent with the deleterious effect of job security on the employment rates of middle-aged and older unskilled workers, relative to skilled ones. Of course, the higher turnover rates among unskilled workers are less likely to be exogenous to the decisions of employers than female turnover rates. In consequence, a complete discussion of this effect requires a model that explains why turnover rates are different in the first place. The model does not seem to able to explain why the effect on employment appears more negative on the unskilled than on skilled youth, but as we have seen, this effect seems to be driven by the more elastic labor supply of this group. 23Cowan, Micco, Mizala et al. (2003) find that, in Chile, seemingly high transitions between schooling and the labor market lead to a very elastic labor supply for the young unskilled. 16 6.2 Distribution of the Effect of Minimum Wages Table 5 also reports the results of interacting personal characteristic dummies with the evolution of minimum wages over time. An increase in the statutory wage has qualitative effects on the distribution of employment across age and skill that are similar to the qualitative effects of stricter job security provisions. To account for contemporary employment policies and economic conditions we include measures of union activity, job security provisions and GDP deviations, interacted with demographic dummies in all specifications in columns (2) to (6) but not in column (7). As in other studies for developed countries, the results in column (7) suggest that an increase in the minimum wage reduces the employment prospects of young workers relative to older ones. We also find a negative effect on the unskilled. Instead, our results also indicate that minimum wages hikes may increase the probability of employment for women relative to men. Controlling for the sub-group effects of contemporary changes in policy and the business cycle does not alter the results reported in column (7).24 The comparison between the results obtained from the wage employment and the self-employment specifications (column (3) and (4)) is also encouraging. As with the coefficients associated with job security provisions, we find that the coefficients on wage employment are very similar to the ones obtained for total employment, while the coefficients on self-employment are not statistically significant. All in all, these results suggest that the effects we are capturing are indeed associated with changes in policy rather than with some unobservable correlate of employment. In column (5) we present our results once we allow for differential effects across age- skill and age-gender categories and control for contemporaneous changes in policy and economic conditions. As in column (7), we find a negative effect of minimum wages on the employment rates of unskilled workers, particularly for middle-aged ones. The effect of minimum wages is negative for young unskilled workers and not statistically significant for young skilled ones. Instead, higher minimum wages tend to shift employment towards older workers. Lastly, we find that women, and in particular the young, tend to benefit from minimum wage policies. The former specification assumes that the effect of raising the minimum wage is unrelated to the level of the going wage. However, it is plausible that the effect may be positively related to the distance between the statutory and the going wage. To account for this possibility, 24See column (3) as well. 17 we include average wages, computed as described in Section 5.25 The results reported in column (6) indicate that controlling for the time evolution of the average wage of sub-population j = 1,...,12 does not alter the results reported in columns (3) to (5). While most of our findings are consistent with the competitive model, some are difficult to explain with this paradigm. For instance, this model cannot explain why minimum wages tend to shift employment towards women. Assuming that women have higher marginal products than men and adding worker heterogeneity to the simple competitive model, this shift can be explained as a "flight to quality" effect. To see that, assume a population of heterogeneous workers that prior to the minimum wage increase were each paid their marginal value. After an increase in minimum wages, all workers with a marginal value below the new minimum wage cease to be employed. Assuming a perfectly elastic supply of all types of workers, firms replace lower marginal value workers with higher value ones. This explanation, however, is at odds with the widespread observation that women's wages are lower than men's. Another possible interpretation is that while men are able to obtain wages that are close to the competitive ones, women's wages are below their marginal products. This would be consistent with the systematic wage gaps found between observationally identical men and women and with the asymmetric gender effects of minimum wages. If wage-gaps are explained by imperfect competition in female labor markets, employers are supply constrained when hiring women. Therefore, an increase in minimum wages can expand both labor supply and employment rates. 6.3 Total Effects In our previous results, all the estimated coefficients measured the effects of labor regulations on each particular sub-population relative to the omitted category, but they did not provide information on whether the employment probabilities of the different sub-groups increased or declined in absolute terms after changes in policy. In this section, we attempt to gauge the total effects of labor market policies on the probability of employment by estimating their effect on the aggregate employment rates of prime-aged skilled men (the omitted category in the specifications reported in Table 5). To do so, we estimate the following error correction specification: 25Including such variables is tantamount to including a set of non-coverage adjusted, demographic group-specific Kaitz ratios. However, we are not imposing the constraint that the coefficient on the minimum wage is the same as the coefficient on the group-specific average wage. 18 Nt=c-(t-1-) + 1(yt­ yt*)+ +B2Log(Wt )+ 3-L+t (1) where = 0 + 1Log(JSt)+ 2 Log(MWt)+3Log(Uniont) (2) and where Nt denotes the employment rate--i.e., the employment to population ratio--of prime- aged male skilled workers in period t, denotes long-run equilibrium employment, yt ­yt* denotes GDP deviations from its trend (in logs), Wt denotes average wages for prime-age skilled male workers, JSt denotes the measure of Job Security, MWt denotes minimum wages, Uniont denotes the index of wage bargaining and L is the length of the maximum lag. In expression (1), employment changes in function of: previous period deviations from long-run equilibrium employment; GDP deviations from its trend; changes in wages and short run dynamics. Expression (2) assumes that, in the long run, employment rates are a function of labor market policies and the structure of wage bargaining. Using aggregate time series techniques to estimate the effect of policies on the reference group allows us to model short and long-run employment dynamics. The first step in the estimation of expression (1) and (2) is to test whether the variables are stationary. The first panel in Table 7 reports the results of testing for the presence of unit roots using the Augmented Dickey-Fuller test (ADF). The tests are specified with three lags. In those cases in which the plot of the series indicated the presence of a time trend we included a constant and a time trend in the specification, in the other cases, we included only a constant. While we can reject the unit root hypothesis for GDP deviations from its trend and for changes in hourly wages, we cannot reject non-stationarity for the lagged employment rate, the logarithm of minimum wages, the logarithm of the job security index and the logarithm of union centralization. However, ADF tests on the first differences of these four series indicate that the hypothesis that these series are integrated of order one, I(1), is not rejected. Given the non-stationarity of the employment rate, expression (1) is well defined only if lagged employment deviations with respect to the long-run equilibrium rate are stationary. This is equivalent to saying that the series t has to cointegrate with t-1. The second panel in Table * 7 reports the results of the Johansen cointegration test between and t-1. The likelihood ratio test indicates the presence of three cointegrating equations indicating that the error correction model is well defined. 19 Table 8 presents the results of estimating the error correction model (ECM) once expression (2) has been substituted into (1). We use the results of the AIC test to determine the optimal length of the lagged endogenous variable and determine that L=1. We estimate the ECM with and without wages to see whether introducing wages alters our results and find the results to be very similar in both cases. Essentially, we find that while job security provisions increase the long-run equilibrium rate of prime-aged skilled male employment. This is not totally surprising. As mentioned in Section 2, job security provisions increase the cost of dismissing workers with long tenure relative to the costs of dismissing less tenured workers, reducing the layoff rate of the first relative to the layoff rate of the latter. Since prime-age skilled workers tend to have longer tenures than other, younger, less skilled workers, job security provisions reduce the layoff rates of prime-age skilled workers relative to the layoff rate of other demographic groups. The positive sign in the ECM suggests that this effect on the layoff rate more than compensates for the negative effect of JS on employment creation. Instead, we do not reject the hypothesis that an increase in the minimum wage does not affect the employment rate of prime-aged, skilled male workers regardless of whether we control for the evolution of wages. The estimated effect of job security provisions and minimum wages on the employment rate can be used to infer the total effect of these regulations on the employment probabilities of other demographic groups. In order to do so, the coefficients on job security provisions and minimum wages, reported in Table 8, should be divided by (minus) the coefficient on the lagged employment variable, to obtain the coefficients in expression (2). They reflect the magnitude of the long-run effect of regulations on prime-age skilled male employment. The third and fourth columns of Table 6 present our estimates for the total effects. They are obtained by adding the marginal effect reported in the first and second columns of Table 6 to the long-run elasticities obtained from specification (1) in Table 8.26 The total effects reported in columns (3) and (4) suggest that job security provisions not only shift the distribution of employment towards older and skilled workers, but also increase their employment rates. Instead, more stringent job security provisions reduce the employment rates of young workers. Moreover, job security provisions reduce employment opportunities for women while increasing those of men. The magnitudes of these estimated effects are substantial. 26The long-run effect of job security on the employment rates of middle age skilled workers is computed as 0.011 divided by 0.63, which is equal to 0.017. 20 According to them, the 1990 labor reform, which increased our measure of job security by about one third, reduced the employment rates of young unskilled male workers by 1.6 percentage points of the population. We also find non-neutral effects of minimum wage spikes. Our estimates suggest that a 10 percent increase in minimum wages reduces the probability of employment for young unskilled male workers by 0.51 percentage points. Lastly, we find that a 10 percent increase in the minimum wage raises the employment rates of women by 0.46 percentage points. 7. Conclusions The effect of regulations is far from neutral across demographic sub-groups. Paradoxically, job security and minimum wage regulations appear to be detrimental to the very workers that they are supposed to help. Our results suggest that both minimum wages and job security regulations reduce the employment opportunities of the young and the unskilled--and particularly unskilled youth--while promoting the employment rates of skilled and older workers. We have also found indications that job security regulations may force some workers, particularly women and the unskilled, out of wage employment and into self-employment. There is an ongoing debate on whether raising minimum wages and job security provisions have any effects on aggregate employment rates. However, even if researchers concluded that job security provisions or minimum wages do not have an effect in the aggregate, it is important to carefully consider these distributional effects when evaluating their desirability. At best, these policies will help some disadvantaged workers at the expense of other poor, young or low-skilled workers. At worse, they distribute jobs from less advantaged to better-off workers. 21 References Banco Central de Chile. 2001. "Indicadores Económicos y Sociales de Chile: 1960-2000." Banco Central de Chile, Santiago de Chile. Bazen, S., and N. Skourias. 1997. "Is There a Negative Effect of Minimum Wages on Youth Unemployment in France?" European Economic Review 41(3-5): 723-32. Bentolila, S., and G. Bertola. 1990. "Firing Costs and Labour Demand: How Bad is Eurosclerosis?" Review of Economic Studies 57: 381-402. Bentolila, S., and G. Saint-Paul. 1994. "A Model of Labor Demand With Linear Adjustment Costs." Labour Economics (1): 303-26. Bertola, G. 1990. "Job Security, Employment and Wages." European Economic Review 34: 851- 86. ----. 1991. "Labor Turnover Costs and Average Labor Demand." NBER Working Paper 3866. Cambridge, United States: National Bureau of Economic Research. Bertola, G., F. Blau and L. Kahn. 2002. "Labor Market Institutions and Demographic Employment Patterns." NBER Working Paper 9043. Cambridge, United States: National Bureau of Economic Research. Bosworth, B., R. Dornbusch and R. Laban, editors. 1994. The Chilean Economy, Policy Lessons and Challenges. Washington, DC, United States: Brooking Institution. Bravo, D., and J. Vial. 1997. "La Fijación del Salario Mínimo en Chile: Elementos para una Discusión." Colección de Estudios CIEPLAN 43: 117-151. Card, D., L.F. Katz and A.B. Krueger. 1994. "Employment Effects of Minimum Wages: Panel Data on State Minimum Wages Laws: Comment." Industrial and Labor Relations Review 47(3): 487-97. Card, D., and A. Krueger. 2000. "Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania: Reply." American Economic Review 90: 1397-1420. Castañeda, T. 1983. "Salarios mínimos y empleo en el Gran Santiago: 1978 y 1981." Cuadernos de Economía 61. Cortazar, R., and J. Vial. 1998. Construyendo Opciones: Propuestas Economicas y Sociales para el Cambio de Siglo. Santiago, Chile: CIEPLAN and DOMEN Editions. 22 Cowan, K., A. Micco, A. Mizala, C. Pagés, and P. Romaguera. 2003. "Un Diagnóstico del Desempleo en Chile." Washington, DC, United States and Santiago, Chile: Inter- American Development Bank and Departamento Ingeniería Aplicada de la Universidad de Chile. Mimeographed document. Davis, S., and J. Haltiwanger. 1992. "Gross Job Creation, Gross Job Destruction, and Employment Reallocation." Quarterly Journal of Economics 107(3): 819-63. De la Cuadra, S., and D. Hachette. 1994. "Lessons from the Chilean Stabilization and Recovery." In: B. Bosworth, R. Dornbusch and R. Laban, editors. The Chilean Economy, Policy Lessons and Challenges. Washington, DC, United States: Brooking Institution. Dowrick, S., and J. Quiggin. 2003. "A Survey of the Literature on Minimum Wages." Canberra, Australia and Brisbane, Australia: Australian National University and University of Queensland. Mimeographed document. Edwards, S., and A. Cox-Edwards. 1991. Monetarism and Liberalization: The Chilean Experiment. Second edition. Chicago, United States: University of Chicago Press. ----. 2000. "Economic Reforms and Labour Markets: Policy Issues and Lessons from Chile." Economic Policy 15(30): 181-230. Hamermesh, D.S. 1993. Labor Demand. Princeton, United States: Princeton University Press. Heckman, J., and C. Pagés. 2000. "The Cost of Job Security Regulation: Evidence from Latin American Labor Markets." Economía 1(1): 147-151. Hopenhayn, H., and R. Rogerson. 1993. "Job Turnover and Policy Evaluation: A General Equilibrium Analysis." Journal of Political Economy 101(5): 915-38. Hudson, R. 1994. Chile: A Country Study. Washington, DC, United States: Library of Congress. Katz, L., and A. Krueger. 1992. "The Effect of the Minimum Wage on the Fast-Food Industry." Industrial and Labor Relations Review 46(1): 6-21. Kosters, M.H., editor. 1996. Effects of the Minimum Wage on Employment. Washington, DC, United States: AEI Press. Lazear, E. 1990. "Job Security Provisions and Employment." Quarterly Journal of Economics 105(3): 699-726. Lindbeck, A., and D.J. Snower. 1988. The Insider Outsider Theory of Employment and Unemployment. Cambridge, United States: MIT Press. 23 Montenegro, C.E. 1998. "The Structure of Wages in Chile: 1960-1996: An Application of Quantile Regression." Estudios de Economía 25(1): 71-98. ----. 2002. "Unemployment, Job Security, and Minimum Wages in Chile: 1960-2001." Washington, DC, United States: World Bank. Mimeographed document. Pagés, C., and C.E. Montenegro. 1999. "Job Security and the Age Composition of Employment: Evidence from Chile." Research Department Working Paper 398. Washington, DC, United States: Inter-American Development Bank. Paredes, R., and L. Riveros. 1989. "Sesgo de Selección y el Efecto de los Salarios Mínimos." Cuadernos de Economía 79. Partridge, M., and J. Partridge. 1998. "Are Teen Unemployment Rates Influenced by State Minimum Wage Laws?" Growth and Change 29(4): 359-82. Risager, O., and J.R. Sorensen. 1997. "On the Effects of Firing Costs When Investment Is Endogenous: An Extension of a Model by Bertola." European Economic Review 41(7): 1343-53. Romaguera, P., C. Echevarría and P. González. 1995. "Chile." In: G. Márquez, editor. Reforming the Labor Market in a Liberalized Economy. Washington, DC, and Baltimore, United States: Inter-American Development Bank/Johns Hopkins University Press. Soto, R. 1995. "Trade Liberalization in Chile: Lessons for Hemispheric Integration." In: E.L. Echeverri-Carroll, editor. NAFTA and Trade Liberalization in the Americas. Austin: University of Texas, Graduate School of Business, Bureau of Business Research. White, H. 1980. "A Heterokedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heterokedasticity." Econometrica 48: 817-838. Williams, N., and J. Mills. 2001. "The Minimum Wage and Teenage Employment: Evidence from Time Series." Applied Economics 33(3): 285-300. Wisecarver, D, editor. 1992. El Modelo Económico Chileno. Santiago, Chile: Centro Internacional para el Desarrollo Económico (CINDE), Instituto de Economía de la Pontificia Universidad Católica de Chile. . 24 Figure 1. Job Security (in monthly wages) 4.5 4 3.5 3 wages 2.5 hly 2 Mont 1.5 1 0.5 0 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 Source: Pagés and Montenegro (1999). Figure 2. Minimum to Average Real Wages 2.00 1.80 e 1.60 Wag 1.40 ealR 1.20 1.00 Average 0.80 0.60 Minimum/ 0.40 0.20 0.00 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 For people under 18 y/o For people over 18 y/o Source: Authors' calculations (see data section). 25 Figure 3. Employment and Dependent Rates 60.0% 14.0% 55.0% 13.0% 50.0% 12.0% Rate Dependent s and te 45.0% 11.0% Ra 40.0% 10.0% Self-Employment Employment 35.0% 9.0% 30.0% 8.0% 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 Years Employment Rate Dependent Rate Self-Employment Rate Figure 4. GDP Deviation from Trend 20.00% 15.00% 10.00% 5.00% 0.00% Percentage -5.00% -10.00% -15.00% -20.00% 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98 Source: Authors' calculations (see data section). 26 Figure 5. Bargaining Index 4 3 2 1 0 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 Year Bargaining Index measures the degree of centralization of wage bargaining. It takes values from 1 to 4 of collective bargaining. Source: Edwards and Cox-Edwards (2000). Bargaining Index Smoothed Bargaining Index 27 Table 1. Employment Protection Provisions in Chile: 1960 ­ 1998 Periods Prior Economic Compensation Compensation for To whom Notice reasons just for dismissal in dismissal in case of unjust do changes Period cause for case of just cause apply? dismissal on the cause law?/ in the courts? 1960 ­1966 1 month Dismissals at will Dismissals at Dismissals at will Dismissals will at will 1966-1973 Economic The law does One month's pay per year All workers 1 month reasons were just not mandate any of work at the firm plus cause in the law. compensation in foregone wages during trial. Firms could not In practice labor this case. Trials could last at most 6 dismiss workers courts considered months. without a just most dismissals There is no maximum in cause. unjustified. the amount to be awarded 1973-1978 1 month Labor courts were Same as Same as previous period All workers much more pro- previous period firms. Workers' claims were weaker. 1978-1980 1 month Economic needs zero 1 month per year of work, Only to were considered without maximum limit. workers June 15, 1978 just cause. hired after Decree 2,200 June 1978 1981-1984 1 month Economic needs zero 1 month' wage per year of Only to Law 18,018 were considered work with a maximum of workers (August 14, just cause. 150 days hired after 1981) August 1981 1984-1990 1 month Economic needs zero 1 month's wage per year of All workers Law 18,372 were no longer work with a maximum of (Dec, 1984) considered just 150 days cause for dismissal. 1990- today 1 month Firms have to Economic 1.2-1.5 months per year of All workers (Nov. 1990) justify dismissals reasons: work hired after Firms need to but economic 1 month' wage August 1981 justify needs are per year of work dismissals considered just with a maximum cause for of 11 months' dismissal. pay 28 Table 2. Parameters Used To Compute Index b a SPfc SPuc 1960-65 0.92 0.88 1 1 0 0 1966-73 0.92 0.88 1 0.2 0 (1) 1974-77 0.92 0.88 1 0.5 0 (2) 1978-80 0.92 0.88 1 0.8 0 (2) 1981-84 0.92 0.88 1 0.8 0 (3) 1985-90 0.92 0.88 1 0 0 (3) 1991- 0.92 0.88 1 0.9 (4) (5) Notes: To compute we use the fact that the average real interest from 1960-1998 was 8.4 percent. To compute we assume that the average Chilean turnover rate without employment protection would be similar to the US rate. According to Davis and Haltiwanger (1995) average turnover rates average 12 percent a year in the United States. (1) Corresponds to one month's pay per year of work augmented by three months to capture the average payments in foregone wages during trial. (2) One month's pay per year of work without upper limit. (3) One month's pay per year of work with an upper limit of five months' pay. (4) One month' s pay per year of work with an upper limit of 11 months' pay. (5) 1.2 months of pay per year of work with 11 months upper limit. We assume the maximum tenure a worker can attain at a firm is 25 years. 29 Table 3. Basic Statistics of the Sample Minimum Wage Bargaining Index Average Wage Wage Self Job By Sex By Skill Level By Age Group GDP Employ Employ- Security Under Over 25- deviation Employ- ment ment Index 18 y/o 18 y/o Original Smoothed Male Female Low High 15-24 49 50-65 from Trend ment Rate Rate Rate Col. Col. Col. Col. Col. Col. Year Col. (1) Col. (2) Col. (3) Col. (4) Col. (5) (6) Col. (7) (8) (9) (10) (11) (12) Col. (13) Col. (14) Col. (15) Col. (16) 60 0.5199 119 119 3.33333 3.33333 302 152 157 475 133 283 306 -0.86% 52.5% 39.8% 12.7% 61 0.5199 114 114 3.33333 3.33333 370 179 171 554 164 331 435 -1.41% 52.2% 41.1% 11.1% 62 0.5199 126 126 3.33333 3.33333 373 203 181 615 162 361 418 -1.37% 53.2% 41.2% 11.9% 63 0.5199 109 109 3.33333 3.33333 376 206 n.a. 311 219 342 395 0.20% 53.0% 41.4% 11.5% 64 0.5199 107 107 3.33333 3.33333 268 160 n.a. 230 133 272 296 -2.15% 52.9% 42.3% 10.6% 65 0.5199 114 114 3.33333 3.33333 n.a. n.a. n.a. n.a. n.a. n.a. n.a. -5.23% 54.4% 43.3% 11.2% 66 3.9090 118 118 3.33333 3.33333 380 211 187 591 179 376 434 1.50% 53.0% 42.2% 10.8% 67 3.9090 116 116 3.33333 3.34724 427 268 222 648 217 420 539 1.50% 54.0% 43.2% 10.8% 68 3.9090 111 111 3.33333 3.39543 466 278 224 699 251 450 502 1.79% 53.2% 41.9% 11.4% 69 3.9090 107 107 3.33333 3.46403 475 279 231 709 218 470 560 2.79% 52.4% 41.2% 11.2% 70 3.9090 133 133 3.66667 3.53596 549 351 256 804 248 536 693 2.97% 52.3% 41.4% 10.9% 71 3.9090 183 183 3.66667 3.57675 689 437 302 957 307 660 779 9.67% 53.7% 42.1% 11.5% 72 3.9090 195 195 3.66667 3.52856 712 457 342 929 359 698 729 7.28% 52.7% 41.3% 11.4% 73 3.9090 108 108 3.66667 3.40525 525 332 279 671 280 512 553 0.37% 51.4% 39.6% 11.8% 74 1.8642 204 204 3 3.26140 435 310 275 561 255 436 496 0.12% 49.0% 37.1% 11.8% 75 1.8642 245 245 3 3.12419 376 277 225 483 214 376 420 -14.58% 45.0% 34.7% 10.4% 76 1.8642 259 259 3 3.01390 486 352 249 635 280 474 542 -12.67% 45.8% 34.5% 11.2% 77 1.8642 269 269 3 2.88227 692 512 320 953 357 696 786 -5.01% 48.3% 38.1% 10.1% 78 1.0599 346 346 3 2.62090 868 517 360 1090 400 799 1072 0.87% 48.0% 37.1% 10.9% 79 1.0599 345 345 2.66667 2.27455 913 640 432 1150 496 904 1009 6.66% 47.8% 36.8% 10.9% 80 1.0599 354 354 1.33333 1.90434 890 611 424 1120 476 881 932 11.83% 47.4% 36.6% 10.7% 81 0.8772 334 334 1.33333 1.53353 1057 799 510 1338 590 1099 1016 15.64% 50.9% 39.3% 11.6% 82 0.8772 365 365 1.33333 1.25825 1235 852 508 1499 618 1206 1295 -1.15% 41.8% 33.0% 8.8% 83 0.8772 276 276 1 1.13070 842 622 345 1056 416 872 721 -6.79% 43.5% 34.4% 9.1% 84 0.8772 243 243 1 1.06209 843 573 355 1028 371 845 780 -4.19% 46.1% 35.8% 10.3% 85 2.2915 220 220 1 1.01390 699 480 312 808 323 683 725 -6.19% 46.4% 36.6% 9.8% 86 2.2915 215 215 1 1 653 471 301 742 314 634 731 -5.35% 47.0% 37.3% 9.7% 87 2.2915 199 199 1 1 796 539 288 932 355 764 907 -4.05% 50.1% 39.5% 10.5% 88 2.2915 222 222 1 1.02781 766 542 316 902 376 751 799 -2.93% 50.9% 38.6% 12.2% 89 2.2915 293 340 1 1.12419 869 679 376 981 434 868 973 0.41% 53.1% 41.6% 11.5% 90 2.2915 298 346 1 1.26140 1003 682 390 1074 462 960 1011 -2.83% 52.0% 40.5% 11.4% 91 3.0598 278 327 1.66667 1.40525 971 694 401 1046 470 951 949 -2.47% 53.2% 41.2% 11.9% 92 3.0598 293 340 1.66667 1.54247 904 726 455 998 503 914 900 1.47% 55.7% 43.6% 12.1% 93 3.0598 294 341 1.66667 1.63885 1072 832 496 1158 627 1054 1093 0.98% 55.9% 44.0% 11.9% 94 3.0598 294 342 1.66667 1.66667 1141 840 535 1194 624 1101 1163 -1.22% 55.4% 42.5% 12.9% 95 3.0598 302 351 1.66667 1.66667 1230 919 566 1310 657 1215 1199 0.81% 55.5% 42.8% 12.7% 96 3.0598 279 324 1.66667 1.66667 1329 1047 621 1412 725 1283 1465 1.59% 55.8% 43.7% 12.0% 97 3.0598 248 333 1.66667 1.66667 1392 1100 613 1505 775 1380 1335 2.79% 56.7% 44.1% 12.6% 98 3.0598 243 341 1.66667 1.66667 1356 1136 759 1427 792 1325 1500 0.70% 56.8% 43.6% 13.2% Source: Authors' calculations (see data section). 30 Table 4. General Economic Indicators: Chile 1960-1998 National National Inflation, National Unemployment, Unemployment, Gran Santiago GDP per consumer Unemployment, female (% of youth total (% of Unemployment, Series capita growth prices (annual total (% of total female labor total labor force ages total (% of total Gini Name (annual %) %) labor force) force) 15-24) labor force) Coefficient 1960 n.a. n.a. n.a. n.a. n.a. 8.0 42.5 1961 1.5 7.7 n.a. n.a. n.a. 7.1 45.2 1962 2.7 14.0 n.a. n.a. n.a. 5.7 45.5 1963 3.6 44.1 n.a. n.a. n.a. 5.2 n.a. 1964 0.3 46.0 n.a. n.a. n.a. 4.9 n.a. 1965 -1.8 28.8 n.a. n.a. n.a. 5.0 n.a. 1966 7.6 23.1 n.a. n.a. n.a. 6.0 45.2 1967 1.5 18.8 n.a. n.a. n.a. 5.9 45.8 1968 1.6 26.3 n.a. n.a. n.a. 6.4 48.1 1969 1.5 30.4 n.a. n.a. n.a. 7.1 48.0 1970 0.2 32.5 n.a. n.a. n.a. 7.0 47.5 1971 7.1 20.0 n.a. n.a. n.a. 5.2 47.7 1972 -2.5 74.8 n.a. n.a. n.a. 3.7 43.1 1973 -6.5 361.5 n.a. n.a. n.a. 3.1 44.1 1974 0.8 504.7 n.a. n.a. n.a. 10.3 40.7 1975 -12.8 374.7 n.a. n.a. n.a. 16.1 41.1 1976 1.8 211.8 n.a. n.a. n.a. 18.0 47.2 1977 7.1 91.9 n.a. n.a. n.a. 13.0 48.4 1978 5.9 40.1 n.a. n.a. n.a. 12.8 49.8 1979 7.1 33.4 n.a. n.a. n.a. 12.5 49.4 1980 6.5 35.1 10.4 10.0 20.8 11.7 49.1 1981 3.2 19.7 11.3 9.9 21.5 9.0 47.3 1982 -11.7 9.9 19.6 18.3 30.5 23.2 51.2 1983 -5.3 27.3 14.6 14.7 24.7 22.7 52.7 1984 6.3 19.9 13.9 n.a. 25.2 18.4 54.2 1985 5.4 29.5 12.1 13.4 22.7 16.2 51.5 1986 3.9 20.6 8.8 9.7 17.3 15.4 48.7 1987 4.9 19.9 7.9 9.3 n.a. 13.5 57.6 1988 5.5 14.7 6.3 7.8 14.3 11.2 53.7 1989 8.7 17.0 5.3 6.1 13.2 9.3 50.8 1990 1.9 26.0 5.7 5.7 13.1 9.7 53.9 1991 6.2 21.8 5.3 5.8 12.7 8.3 52.4 1992 10.4 15.4 4.4 5.6 10.9 6.0 47.4 1993 5.2 12.7 4.5 5.1 11.0 6.4 45.4 1994 4.0 11.4 5.9 6.8 13.2 6.3 45.9 1995 8.9 8.2 4.7 5.3 11.5 6.1 46.3 1996 5.7 7.4 5.4 6.7 12.8 7.2 45.4 1997 6.0 6.1 5.3 6.6 13.0 6.7 n.a. 1998 2.5 5.1 7.2 7.6 16.7 6.9 n.a. Sources: World Bank World Development Indicators Data Base and Gini coefficient from background data, Montenegro (1998). 31 Table 5. The Effect of Job Security and Minimum Wages, Probit Results (1) (2) (3) (4) (5) (6) (7) Wage Self Dependent variable: Employed Employed Employment Employment Employed Employed Employed t-test t-test t-test t-test t-test t-test t-test Dummy young -0.8954 -104.2 0.4921 2.6 0.9189 5.0 -0.4202 -1.4 -1.1703 -6.1 -0.9651 -4.9 1.2757 9.1 Dummy old -0.6709 -66.8 -1.6509 -7.3 -1.6967 -7.5 0.4176 1.7 -2.0996 -9.1 -2.1226 -9.0 -1.4101 -8.6 Dummy women -0.5461 -66.7 -2.0260 -12.2 -1.8595 -11.6 -0.3632 -1.7 -2.4113 -14.2 -1.9622 -11.3 -2.7873 -22.7 Dummy unskilled 0.0007 0.1 1.8635 10.9 1.8843 11.2 -0.3281 -1.5 1.4867 8.6 1.8356 10.3 2.2867 18.1 Children per father 0.1570 45.0 0.1569 44.6 0.0594 25.7 0.0273 11.3 0.1152 32.0 0.1152 31.5 0.1562 44.6 Children per mather -0.3931 -93.9 -0.3921 -92.7 -0.3147 -86.9 -0.0196 -5.4 -0.3179 -70.1 -0.3160 -68.5 -0.3919 -93.1 Dummy young -0.0935 -10.8 -0.1112 -12.7 -0.0826 -9.7 -0.0161 -1.2 -0.0913 -5.6 -0.1163 -6.7 Dummy old 0.0124 1.2 0.0196 1.8 0.0292 2.7 0.0173 1.5 0.0253 1.2 0.0123 0.6 Dummy women -0.0468 -6.1 -0.0266 -3.4 -0.0021 -0.3 0.0267 2.7 -0.0546 -4.5 -0.0873 -6.8 thiwde boJfo Dummy unskilled -0.0334 -4.2 -0.0563 -7.0 -0.0733 -9.3 0.0344 3.4 -0.0382 -3.3 -0.0596 -4.8 Dummy young*dummy women 0.0835 4.7 0.1033 5.4 Dummy old*dummy women -0.0035 -0.2 0.0064 0.3 act mthi ytir ternI gar Dummy young*dummy unskilled -0.0381 -2.2 -0.0164 -0.9 Lo ecuS Dummy old*dummy unskilled 0.0033 0.2 0.0146 0.6 Dummy young -0.1406 -8.2 -0.1557 -9.3 -0.0366 -1.3 -0.0111 -0.6 -0.0215 -1.2 -0.2129 -16.0 Dummy old 0.0913 4.4 0.0911 4.4 -0.0286 -1.3 0.1301 6.2 0.1301 6.1 0.0715 4.6 Dummy women 0.1455 9.6 0.1551 10.7 -0.0299 -1.5 0.1677 10.8 0.1303 8.2 0.2097 18.0 thiwde fo gea W Dummy unskilled -0.1811 -11.6 -0.1811 -11.9 0.0304 1.5 -0.1587 -10.1 -0.1810 -11.2 -0.2196 -18.3 Dummy young*dummy women 0.0248 11.0 0.0223 9.8 Dummy old*dummy women -0.0035 -1.3 -0.0019 -0.7 act mthi mu ternI gar nim Dummy young*dummy unskilled 0.0393 17.4 0.0346 15.2 Lo MiDummy old*dummy unskilled 0.0133 4.9 0.0145 5.3 nionU Dummy young 0.1320 8.2 0.1422 9.2 0.0800 3.0 -0.3006 -13.1 -0.2785 -11.9 Dummy old 0.0272 1.4 0.0241 1.2 0.0152 0.7 -0.0966 -3.2 -0.0854 -2.8 Dummy women -0.0968 -6.8 -0.1222 -8.9 0.0802 4.2 -0.2494 -13.5 -0.2177 -11.6 thiwde noti Dummy unskilled 0.0756 5.2 0.0480 3.4 0.0358 1.9 -0.0843 -4.6 -0.0599 -3.3 Dummy young*dummy women 0.2957 12.3 0.2712 10.9 Dummy old*dummy women 0.1530 5.2 0.1359 4.5 act ternI zaliartneC Dummy young*dummy unskilled 0.3485 14.1 0.3306 13.0 Dummy old*dummy unskilled 0.0265 0.9 0.0249 0.8 Dummy young -0.0852 -0.9 0.2102 2.2 0.0208 0.1 -0.2928 -1.7 -0.3618 -2.1 PDG htaP Dummy old -0.3872 -3.1 -0.2161 -1.7 -0.0041 0.0 -0.7902 -3.4 -0.8027 -3.4 Dummy women -0.4917 -5.5 -0.3108 -3.6 0.3153 2.7 -0.8047 -6.0 -0.8958 -6.7 thiwde Dummy unskilled 0.4345 4.8 0.3467 3.9 0.0777 0.7 0.4079 3.2 0.4152 3.2 Dummy young*dummy women 0.3973 2.0 0.5022 2.5 Dummy old*dummy women 0.3863 1.6 0.4749 1.9 act ternI omrfnoitaiveD Dummy young*dummy unskilled -0.2455 -1.3 -0.1571 -0.8 Dummy old*dummy unskilled 0.1912 0.8 0.1761 0.7 Logarithm of hourly wage 0.1520 16.9 Number of Observations 303945 303945 303945 303945 303945 295318 303945 Pseudo R2 0.196 0.168 0.11 0.08 0.211 0.210 0.197 Notes: Besides the control variables mentioned in the table, all specifications include yearly dummies (not reported). Standard errors are robust to the presence of heteroskedasticity. The employed dummy variable is defined as 1 if the person is employed and 0 otherwise (unemployed or inactive). The wage employment dummy variable is defined as 1 if the person is a dependent employee and 0 otherwise (independent, unemployed or inactive). The self-employed dummy variable is defined as 1 if the person is an employer or if the person works as an independent worker and 0 otherwise (dependent, unemployed or inactive). 32 Table 6. Marginal and Total Effects of Labor Market Regulations Marginal Effects Total Effects Job Security Min. Wage Job Security Min. Wage (1) (2) (3) (4) Men, 15-25, unskilled -0.066 -0.0516 -0.049 -0.0516 [0.000] [0.000] Men, 15-25, skilled -0.0351 -0.004 -0.0181 -0.004 [0.000] [0.52] Men, 26-50, unskilled -0.008 -0.036 0.009 -0.036 [0.001] [0.000] Men, 51-65, unskilled -0.0035 -0.005 0.0135 -0.005 [0.620] [0.54] Men, 51-65, skilled 0.008 0.045 0.025 0.045 [0.22] [0.000] Unskilled -0.0343 -0.012 -0.0173 -0.012 [0.000] [0.09] Skilled -0.015 0.044 0.002 0.044 [0.000] [0.000] Women -0.0278 0.0463 -0.0108 0.0463 [0.000] [0.000] Men -0.0151 -0.017 0.0019 -0.017 [0.000] [0.000] Young -0.0394 0.0134 -0.0224 0.0134 [0.000] [0.08] Older -0.008 0.0596 0.009 0.0596 [0.14] [0.000] Note: P-values of the test that the marginal effects are equal to zero are reported in square brackets. 33 Table 7. Unit Root and Cointegration Tests Name of the Series Symbol Specification ADF Test Statistic 5% Critical Value GDP deviation from its trend y-y* Constant -4.8412 -2.9472 Wage Growth (logW) Constant -3.8514 -2.9705 Logarithm Minimum Wage L(Minwage) Trend -1.4709 -3.5426 Logarithm Job Security L(JS) Constant -2.43 -2.9472 Logarithm Union Centralization L(Union) Trend -2.7568 -3.5426 Lagged Employment Rate Nt-1 Constant -1.6736 -2.9472 First diff. Lagged Emp. Rate Nt-1 Constant -3.0433 -2.9499 Change in Log Minimum Wage L(Minwage) Constant -2.5591 -2.9499 Change in Log JS L(Index) Constant -2.655 -2.9499 Change in Log Union L(Union) Constant -2.3443 -2.9499 Panel 2: Johansen Cointegration Test Series: Nt-1 L(Minwage) L(JS) L(Union) Likelihood Ratio 5% critical Value Hypothesized number of CE 108.64 53.12 None** 60.35 34.91 At most 1 ** 24.64 19.96 At most 2 * 5.26 9.24 At most 3 * (**) denotes rejection of the hypothesis at 5% (1%) significance level 34 Table 8. Level Effects on Male Prime-Age Employment (1) (2) Independent Variables: Nt-1 -0.63 -0.66 (-3.05) (-3.24) Deviations GDPt 0.08 0.10 (1.21) (1.48) log Wt - 0.018 (0.84) Log (JS) 0.011 0.015 (1.80) (2.23) Log (Minwage) -0.01 -0.014 (-0.93) (-1.13) Log (Union) 0.03 0.029 (1.54) (1.45) Constant 0.61 0.651 (3.55) (3.92) Nt-1 0.277 0.239 (1.48) (1.30) N obs. 37 35 Adj. R squared 0.16 0.23 Long term Effect of 0.017 0.023 JS Long term Effect of 0 0 Minwage Note: t-statistics shown in parenthesis. 35