Mexico - Labor Markets : New Views on Integration and Flexibility (Vol. 2 of 2) : The Technical Papers

FOR OFFICIAL USE ONLY CONFIDENTIAL Report No. 19438-ME Mexican Labor Markets: New Views on Integration and Flexibility Volume Two: Technical Papers (in Two Volumes-Green Cover Draft) December 6, 1999 Poverty Reduction and Economic Management Unit Latin America and Caribbean Region This document has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization. Document of The World Bank FILE COPY Efficiency Wage and Union Effects in Labor Demand and Wage Structure in Mexico William F. Maloney LCSPR Eduardo Pontual Ribeiro Universidade Federal do Rio Grande do Sul, Brazil Mexico Labor Markets Volume II Technical Papers Contents 1. Efficiency Wage and Union Effects in Labor Demand and Wage Structure in Mexico William F Maloney and Eduardo Pontual Ribeiro 2. The Informal Sector, Firm Dynamics and Institutional Participation Alec Levenson and William F. Maloney 3. Logit Analysis in a Rotating Panel Context and an Application to Self- Employment Decisions Patricio Aroca Gonza7ez and William F Maloney 4. Quitting and Labor Turnover: Microeconomic Evidence and Macroeconomic Consequences Tom Krebs and William F Maloney 5. Self-Employment and Labor Turnover in LDCs: Cross Country Evidence William F Maloney I. Introduction. The evidence to date suggests that both union power and efficiency wage behavior may have large effects on the structure and dynamics of labor mhrkets. The literature documenting their effect on wages, in particular, is vast. However, as Blanchflower et. al. (1991) note, the impact on employment of unions has received relatively little study, and that of efficiency wages has attracted even less. Further, as Nickell and Wadhwani (1991) argue, the existence of both phenomena simultaneously complicates efforts to distinguish between competing models of union behavior.' Their work and that of Hendricks and Kahn (1991) are among the very few that analyze employment determination in the presence of efficiency wages and two types of union bargaining: the "right to manage" (RTM) type where unions attempt to set the wage but let firms choose the level of employment, and the "efficient bargaining" (EB) type where unions bargain over both (Oswald 1985 and Layard and Nickell 1990). Using a large panel from the UK manufacturing sector, Nickell and Wadhwani find no evidence of union influence on employment and mixed evidence of efficiency wage effects. Hendricks and Kahn find EB effects in their study of the demand for police in the US. This paper builds on this work in two ways. First, it approaches these issues using quantile analysis that more completely characterize the distributions of wages and labor demanded than can be done using the conditional mean based linear regression approaches (OLS, 2SLS) that are standard. The results show the power of this technique to uncover important differences in the impact of unions and efficiency wage effects across quantiles that would ordinarily go undetected. Second, it analyses Mexico, a country with unique institutional, economic and political characteristics that make it an important case study to add to the literature for two reasons. First, we may observe union behavior that, although theoretically plausible is contrary to that commonly found. Despite a longstanding and widespread tradition of unionization, a high degree of coordination through the social pact begun in 1987 has made possible extreme downward flexibility of wages during recent crises and may have curtailed union power along this dimension. In addition, it is sometimes argued that in the absence of unemployment insurance, employment enters more heavily in union objective functions than in the industrialized countries. The particular constraints and elements of union utility may give rise to EB outcomes with implications for the structure of wages, and the overall level of employment distinct from those generally anticipated. It also may permit a test of Nickell and Wadhwani's assertion that the wage-employment relation may slope upward. Second, the Mexican labor market is well suited to testing for efficiency wage effects. Though a longstanding literature explains dualistic LDC labor markets by government or union interference in the wage setting process,' Mexico's minimum wage was not binding in I For discussions of the theory of efficiency wages see Stiglitz (1974), Krueger and Summers (1988), Phelps (1994) and Weiss (1990). For a review of the literature on union impacts, see Lewis (1986). 2 See, for example, Harris and Todaro(1970) See Esfahani and Salehi Isfahani (1989) as an example of modeling LDC dualism in an efficiency wage context. 2 the period we study (Bell 1996), and if unions focus primarily on employment, then the wage structure and whatever dualism is observed may be emerging endogenously through efficiency wage effects. The existence of a large non-unionized sector permits isolating such effects whose manifestations can sometimes be indistinguishable from the outcomes of union bargaining. Further, since the Mexican Constitution prohibits firing of workers except in extreme circumstances we are arguably testing for one particular variety of efficiency effect arising from the prevention of turnover (Stiglitz 1974). The data set we work with is exceptionally rich. It permits conditioning on numerous dimensions of firm heterogeneity as well as offering some that may conceivably be associated with efficiency wage effects. It provides evidence on the dynamics of unionized firms and their behavior in the use of three inputs: unskilled and skilled labor and, to a lesser degree, capital. Disaggregating labor promises an improvement over the vast majority of efficiency bargain papers as we may avoid a composition bias in the demand for labor due, for example, to the substitutability of types of workers. Further, we can go beyond the standard (static) labor demand literature with worker types (e.g. Hamermesh, 1993 and references therein)3 since it has not allowed for the possibility of employment decisions occurring along a contract curve as in the EB solutions instead of the standard labor demand curve. The paper is organized as follows. The following section presents the theoretical background and the testable implications. The third section discusses quantile analysis. The fourth presents the data set used in the paper and the empirical results. The last section concludes with a summary of results. IIa. Analytics: (Overview) Efficiency Wages: The extensive literature on efficiency wages provides a rational for firms to voluntarily pay wages above the market clearing level. One common variant of these models arises from the difficulty of monitoring individual workers and the lack of any penalty from being caught "shirking" - any activity, or lack thereof, that might be detrimental to the firm. If wages are market clearing, a worker fired for shirking can simply get another job at the same wage. However, if all firms pay higher than market clearing wages, unemployment will be created in the economy that creates a disincentive to being laid off and hence to shirking. 4 Since, in many Latin American countries, workers can be fired only with difficulty, the "turnover" variant of efficiency wage models is probably more appropriate: firms must Hamermesh also points out the clear advantages of using microdata and the dearth of such studies. As Marquez and Ros (1990) noted, and has been confirmed by later studies, wages of similar workers rise with firm size, much as they do in industrialized countries. Further, Marquez (1990), Abuhadba and Romaguera (1993) and Schaffner (1998) find consistent with efficiency wage effects in the patterns of wage differentials that are strong and highly correlated among Chile, Venezuela, and Brazil and the U.S.. This suggests that the conditional wage dispersion (wages adjusted for human capital) and rigidities may be emerging endogenously and are not due to either government or union intervention. 3 invest resources in workers when they are hired, perhaps through training or through the process of recruitment, that will be lost if the worker leaves. Hence, it is worthwhile for firms to pay higher wages and raise the opportunity cost of leaving. Interviews with Mexican entrepreneurs in the survey used here support this view. Roughly 30% stated that the resignation of recently trained workers was a problem. This is almost certainly an understatement for two reasons. First, "recently" may not capture the relevant period of return on the investment in the worker. Second, if the firm is already paying the optimal efficiency wage to prevent workers from leaving, it will not report excessive turnover as a problem. Of those reporting frequent resignations after training, 58% do something to raise the total well-being of the worker after training, 28% raise remuneration without promoting the worker, and 40% take measures that increases the wage of the worker, including promotions (see Appendix I). The efficiency wage argument is particularly compelling in LDCs where firms may absorb a larger share of education costs due to poorly functioning education systems. Thus, firms will be very concerned about preventing workers they train from moving to another firm. In addition, in countries where self-employment (formal or informal) are considered desirable destinations, it is possible that workers enter formal salaried work to accumulate skills and financial capital, and then quit to open their own business. Both theories imply that the offers workers can get outside the firm, (the outside wage) as well as the probability of being able to get a job at that wage (the hiring rate) should be important to determining the wage that is set in the firm, as well as to the quantity of labor hired. Union Bargaining.5 The simplest view of union behavior sees them maximizing union utility, which may be a function of both the wage received by union members and the level of employment, subject to a constraint representing combinations of the two that firms are willing to pay, the labor demand curve. In the "Right to Manage" view unions would identify the level of the wage that maximizes their utility, and firms simply set the level of employment. However, if the firm is a monopolist or oligopolist and earns excess profits, then both unions and firms may be better off by coming to a bargain that pushes them off the labor demand curve. Figure 1 traces out a series of iso-profit curves- combinations of the wage and level of employment such that the firm earns the same level of profits. A lower curve implies a higher rate of profits. The apex of each curve is necessarily cut by the labor demand curve: the firm maximizes profits subject to any given wage, that is, it chooses the level of employment that puts it on the highest iso-profit curve possible. Point P represents one such point. As the wage rises or falls, the firm's optimal level of hiring traces out the labor demand curve, the locus of all apexes of iso-profit curves. Employment either below or above the Graphs taken from and discussion based on Borjas (1996). 4 profit maximizing level (100 at Wo) necessarily implies that the firm earns fewer profits and is thus on a higher iso-profit curve. Therefore the iso-profit curves must slope downward on either side of the intersection with the labor demand curve. As point M in figure 2 shows, a better deal for both workers and firms can be negotiated than that at "Right to Manage" equilibrium at point M. Here, the union's utility curve is tangent to the demand curve, but not to the iso-profit curve of the firms. Thus, the willingness of workers and firms to trade off employment for wages is not equal, and the equilibrium is not efficient. Two alternate and more efficient bargains where the two curves are tangent can easily be seen, both of them off the labor demand curve. First, at point R, Unions reach a higher level of utility, UR compared to Um while firms are earning the same level of profits. Alternately, at Q, unions are no worse off while firm profits are higher. Which bargain, R, Q or perhaps Q', where both are better off, are "efficient bargains" and lie on the contract curve. The contract curve is the set of efficient bargains ranging along the line PZ from P, where workers have no bargaining power and take the market wage W* and the firm takes all profits, 7c*, to Z where nz represents the level of profits below which the firm would go out of business, and the union captures all of the monopoly rents. This iso- profit curve also suggests that the maximum wage workers could ever gain would be W, and then only if it cares very little about employment. These bargains along PZ, however, are clearly not efficient from a prodIction point of view: at any bargain except P, more workers are being hired than the firm would hire in the absence of a union, E*. This "featherbedding" is a way of transferring firm profits to workers through the creation of unnecessary positions, rather than wages. The final equilibrium clearly depends then on the goals of the union as captured in the shape of its utility function, that jointly with the firm's iso-profit functions determines the contract curve, and the union's relative bargaining strength, which determines the position of the final bargain along the contract curve. Both union objectives and bargaining power in Mexico may be different from those in industrialized countries for a variety of reasons. First, like much of Latin America during the 1980's and early 90's, job growth has been slow relative to population growth. Second, as is the case with most of its neighbors, Mexico has no system of unemployment insurance and employment stability may be more highly valued than wages. Third, since the Revolution, the major unions have had a longstanding and close relationship with the government. Particularly since 1987 with the inception of the Pacto- a joint agreement of labor, business and the government to promote price stability- unions have closely coordinated wage demands with pacto guidelines. These factors taken together may lead to an emphasis on employment creation, relative to pushing up wages in the union utility function. The next section details how empirically it is possible to determine whether the type of bargaining occurring as well as if efficiency wage effects are important. 5 IIb. Analytics (detail): Broadly following Nickell and Wadhwani and Layard and Nickell we postulate a firm facing a downward sloping product inverse demand curve do with shift term, a. Its real revenue function R(N,Q,e,a)= F(N,K,e)d(F(.),a) Ri> 0,R2>0,R3>0 is a function of the labor it hires, the stock of other factors including capital, management ability, technology, Q, and also efficiency wage effects on the productivity of labor, e. Among these is the ratio of the inside wage, W, to the expected alternate outside wage, E(Wa). Firms bargain with a union whose utility u=U(W,E(Wa),N) UI>O,U2<0,U3>0 depends on the wage, the expected outside wage, and employment. In the "right to manage" model the union bargains for a level of W, and lets the firm choose the level of employment. However, if the union cares about employment as well, then its utility is maximized over both N and W and the outcome is determined jointly in an "efficient bargain" with the firm. In this case, the firm moves off the demand curve it would face in the RTM scenario and onto the contract curve. The result of a standard Nash bargaining model yields a system of equations, both for employment and the wage. The firm solution is a system of equations of an (implicit) form such as N = N(W, E(W.), Z2, e, v) W = W(E(W),Z2,e,O) that reflect the compound effects of the two utility functions, as well as union bargaining power over employment, 6N .and the wage, E Z2 contains variables that determine the position of the labor demand relation, such as ,and a. The expected outside wage enters both through the union utility function, and efficiency wage effects. Several empirically testable predictions derive from this model: a. If unions bargain solely over the wage, then union power will be captured entirely in the wages paid by the firm and free-standing proxies for union power should have no effect in labor demand functions. Alternatively, if unions also bargain over the level of employment, the union proxy should enter positively in the demand equation. In the extreme case that unions do not bargain over the wage, but only employment, the union terms should be insignificant in the wage equation. 6 b. Since the workers' alternative, the expected outside real wage adjusted for the probability of getting a job, enters both in the firm's calculation of the optimal efficiency wage as well as the union utility function, its predicted sign and magnitude are ambiguous in cases where union power is present.' To avoid this problem, we Will work with both unionized and non-unionized sectors to search for efficiency wage effects. c. The sign of the employment/wage elasticity depends on whether unions have more power bargaining over employment or over wages. d. As union power over employment determination E rises, the elements of Z2 (0, and c) should lose influence in the labor demand equations. e. If unskilled workers are represented by unions more than the skilled, we may observe different union and efficiency wage effects for each group. III. Empirical Methodology Conditional mean regression estimators, such as Ordinary Least Squares, are traditionally used to estimate the relations such as those posited above. Minimizing the squared sum of errors allows estimating the values of the parameters that predict the mean of the dependent variable, conditional on a set of explanatory variables chosen. If there are large outliers, or the distribution of the disturbances is non-normal, conditional mean estimators may be inefficient and often biased. These concerns can be reduced somewhat by estimating the conditional median regression where half the errors lie above, and half below the fitted curve. Quantile analysis, introduced in Koenker and Bassett 1978, extends this analysis to estimating curves where -% of the errors will be negative and (100-t)% of the errors will be positive." If the errors are i.i.d., slicing the distribution at different quantile levels has little effect on parameter estimates and little information is lost in a single measure of the conditional central tendency, such as the parameters generated by OLS. However, figure 3 shows that asymmetries or heteroskedasticity in the distribution of errors may lead to substantially different estimates of the impact of the variables under study. In all the empirical work below, we begin with the standard conditional mean regressions, whether OLS or 2SLS and then present the results of the quantile analysis at T= 50 ( the conditional median regression), r= 10 where 10% of the deviations lie below the estimated regression, and t= 90 where 90% lie below. 'Nickell and Wadhwani argue that the appearance in the demand function of outside wages indicates the presence of efficiency wages unless unions both bargain over employment and more importantly, have a non- standard objective function, with the sign depending on the size of the standard employment-wage elasticity. The technique has generally been applied to estimating returns to education, (Buchinsky 1994)y. 7 IV. Data: We employ the Encuesta Nacional de Empleo, Salarios, Tecnologia y Capacitacion (ENESTYC), the National Survey of Training, carried out by the Mexican Official Statistics Institute (Instituto Nacional de Estadistica, Geografia e Informatica, INEGI) for the year 1992 which contains detailed information on firms specific variables relating to employment, technology, capital stock, etc. A 1995 Survey was also available that had the advantage of collecting data on share of the work force unionized at the firm level. However, it lacked information on the human capital of the work force and because the period it spanned contained the Tequila crisis in December 1994 and the beginning of the ensuing recession, we work primarily with what may be considered a more "normal" period of relative prosperity. Variables: Wages and Employment: Following Roberts and Skoufias (1997) and others the wage and labor stock of skilled (Ws and Ns, respectively) and unskilled labor (WU and NU, respectively) are derived as weighted averages of subcategories within each. The weights for constructing the labor variables are the full wage (wage, social security and other non-wage benefits) per worker. The wage is then the total payments to the sub classes of labor divided by the labor measure, namely, directivos, profesionistas, tecnicos, empleados administrativos and supervisores ("skilled") and obreros profesionales, especializados and en general ("unskilled"). The average schooling of the unskilled is about half of the skilled workers. Outside wage(Wa): The median sectoral wage, at the 4-digit industry level. Hiring rate: In the cross sectional context, the aggregate unemployment rate employed by Nickell and Wadhwani is not useful'. We instead use the sectoral hiring rate (number of hires over level of employment in the sector), as a measure of the probability of finding a job if you leave (with your skills). This is more consistent with a labor turnover view of efficiency wages. Union Density: The 1995 ENESTYC tabulates union density (ratio of firm employees affiliated to a union) by individual firm while the 1992 only tabulates a dummy for the presence of unionization in the firm. Under the assumption that union structure changes little over two years, we assign a value of zero to the union density variable if the 1992 dummy is zero and the median sectoral value from the 1995 survey if the 1992 dummy is unity. Value Added (Q): the value of total 1991 output minus the expenses in materials and energy in million Pesos. Capu: average capacity utilization as reported by the firm in 1991. ' The cross-section nature of the data and the impossibility of identification of the firm's regional location precludes the use of an regional average wage, or typical informal sector earnings. 8 Corp: dummy for firms that belong to a corporation. Forg: dummy for firms with more than 50% foreign ownership. Age: age of the plant in years. Export: dummy for firms with 10% or more of sales to other countries. Auto: percentage of capital stock value of automated machinery. Quality Control: dummy for firms that have quality control of output. Schooling: Average years of schooling of the employed workers in each skill level in the firm, where the years of schooling were obtained from 7 levels. Experience: Average tenure in the firm of workers within each sub-class of labor. R & D: dummy for firms with positive R&D expenses in 1991. Tech: dummy for firms with positive expenses in technology acquisition in 1991. Prdtrg.: dummy for firms that indicated increases in productivity after implementing training programs. Cmptl: dummy for firms that identify their product as "competitive" against imports. Observations with missing, incomplete, or zero entries for employment, output or capital stock were dropped. We also discarded micro firms (up to 15 employees), as they are extremely underrepresented in the sample and their heterogeneity cannot be captured with the sample weights provided. We have not used any sampling weights as they are less important without the micro firms in the sample.9 Table I presents the summary statistics of the variables employed. A few general observations seem important. First, mean schooling of "unskilled" workers is half of those identified as "skilled," and the mean unskilled experience is slightly smaller than that of skilled workers, possibly due to higher turnover. Second, only 17,8% of the firms do not have their workers associated with a central sindical (union) so there are roughly five times the number of observations in the union sample than the non-union sample. This can make comparisons of the significance of effects occasionally ambiguous. Within the union sample, the mean unionization rate is 55%. Third, Mann-Whitney tests show that the unconditional 9 INEGI has a specific survey of micro firms that we feel should be used to study these firms. 9 distributions of most variables differ between the unionized and unionized samples."o This suggests that the two samples differ in findamental ways and perhaps should not be combined during the analysis. Va. Empirical Results: Wage Equations By contrast to the standard competitive model where firms take wages as given, the framework above makes it clear that wages and employment are determined jointly and hence constitute a system of equations to be estimated. However, standard wage equations with employment omitted can be thought of as a reduced form and can be estimated using least squares. Broadly following Dickens and Katz(1987) and Nickell and Wadhwani we estimate the log linear approximation ws.,. = W.yw. + e + yu + h + xy + ew where w. is a vector of proxies for the expected alternative wage, e a vector of other possible efficiency wage related variables, u, the union power measure, h a vector of human capital variables, x the vector of firm related characteristics, including those in Z2. Table 2 presents the results of using OLS wage equations and tables 2.1-2.3 the results of quantile regression at the 10', 50' and 90' quantiles for a specification that includes as shift parameters in Z2 lagged productivity," whether the firm is foreign owned, whether it has multiple branches, whether it invests in R & D, whether it purchases technology, the share of automated machinery, the capital labor ratio and age of firm. The proxy for union bargaining power over wages is the union density measure. As efficiency wage measures it includes outside sectoral wages and hiring probabilities, whether the firm asserts that it maintains quality control, whether the firm engaged in training that it felt increased productivity, two size and eight sectoral dummies. As Dickens and Katz note, the most thorough test for efficiency wages are those that are able to cross individual level human capital variables with plant level characteristics and control for both. At the individual level, we are able only to control for the mean level of schooling in the plant and the mean tenure of each category of workers within the plant. Though not a good measure of individual experience, the latter is a good proxy for the accumulation of firm specific human capital. 1. General Results The OLS specifications explain 38% of the variance for skilled unionized, 29% for those unskilled, and for the non-unionized sample 14% and 11% respectively. Several of the shift variables enter with expected signs and magnitudes although others are more ambiguous. The impact of productivity on wages in both samples is of comparable size as in the literature 10 Standard equality of mean and variances tests were not employed as the unconditional distributions are clearly non-normal. " To avoid a division bias in the productivity coefficient we use past productivity, as in Borjas (1980). Dropping this variable from the regressions does not change the results noticeably in general. 10 (Wadhwani and Wall 1991). Exporters pay less to skilled workers perhaps reflecting more elastic demand curves, although the indicator of product competitiveness is never significant in any regression. Foreign firms pay skilled workers more in both sectors. R& D enters at the 10% level only for non-unionized skilled workers, technology purchase; enter strongly and of the predicted sign only in the unionized sector, indicating that union firms use more productive technology. Though we suppress many of the secondary coefficients in the interest of compactness, the conditional median regression analogous to OLS yields very similar results. Appendix figures Al.1-Al.4 show that the predicted relation between wages and employment is very similar in both cases. 2. Union Effects The union and non-union samples were run separately for two reasons. First, we are interested in isolating efficiency wage effects that, as in the case of the outside wage, are sometimes hard to disentangle from union effects. Second, there may be serious problems of selection bias in measuring union premia. In preliminary regressions, we find that a union dummy in the combined sample suggests that firms with unions pay 15.2% more to skilled workers and 9.25% than non-union firms and the continuous union density variable enters positively and significantly as well. However, it is impossible to know whether unions cause wage differentials, or whether unions are more likely to be found in certain types of firms who also pay higher wages. As an example, schooling enters significantly across all sub-samples, but experience enters only in the union sample. The average union worker with five years of experience would make roughly 12% more than his non-union counterpart and the unskilled perhaps 8%. Constraining the experience coefficient to be equal across sub-samples could give rise to differentials of the magnitudes of the union dummies. If the differing coefficients represent rigidly enforced seniority base promotions or wage hikes, then the differential might legitimately capture union power (see Borjas 1996). However, if firms use production techniques that require more on-the-job training and also make them more prone to unionization, then the differentials capture legitimate differences in human capital rather than union power. As mentioned earlier, Mann-Whitney tests show that the unconditional distributions of most variables differ between the unionized and non-unionized sample, and Chow tests for the equality of the coefficients between union and non-union firms strongly reject at the 1% level, consistent with the view that union and non-union firms may be fundamentally different. As a strategy that partially alleviates the selection bias problem, we test for the impact of union power within the sample of firms with unions. In a very surprising result, the free-standing union term never enters significantly as a determinant of unskilled wages in the OLS regressions. The labor demand regressions in the next section will cast doubt on the obvious interpretation that unions have no power. But the quantile regressions also reveal a story hidden to standard techniques. For the upper quantiles, there is no impact on unskilled wages. However, for the 10" quantile a strong and positive coefficient emerges on the union density term while the human capital variables that are important for the other quantiles largely disappear. An interpretation is that workers who 11 earn little given measured human capital, are helped by unions. If for example, a worker's unobserved characteristics, such as reliability and diligence, dictate a low wage relative to those who, on paper, appear similar, unions will push them toward the average for their class. To the degree that this measures distortion in the wage distribution, it appears to be confined to the 10th quantile. Overall, union density does not appear to have a major impact on unskilled wages. A striking result is the strong negative impact of union density on skilled wages, precisely the opposite effect found in the combined regression. This may be due to more successful unskilled worker bargaining over firm rents (distributed in forms other than wages), or it may be related to a desire to reduce the wedge between skilled and unskilled remuneration for equity reasons. 3. Efficiency Wage Effects The significance of the outside wage in both the union and non-union samples suggest that its influence is more than simply a reference for union bargaining. The fact that we cannot reject the equivalence of the effects in the two samples supports an interpretation that the dominant effect even in the unionized sector is efficiency wage related. Somewhat counterintuitively, the hiring rate enters with unexpected sign although insignificantly in all but the unionized skilled worker sample. Whether training had been undeftaken that was perceived as productivity enhancing (p.train in table) has the predicted sign although it is insignificant in the union group, suggesting, perhaps that training is not differential among firms. Inter industry wage differentials (Krueger and Summers 1988) as captured by sectoral dummies are virtually non-existent in all sectors after controlling for human capital variables in firms contrary to the literature. On the other hand, firm size effects are strongly present as in Schaffner (1998), notably so for skilled workers. For both union and non-union firms, apparently similar skilled workers in small firms of between 16 and 100 workers make roughly 50.4% less in wages and benefits than they would in a firm of over 250 and unskilled workers roughly 30%. In the event that this is, in fact, due to efficiency wage considerations arising, perhaps from difficulties of monitoring, the implicit segmentation emerging endogenously among formal enterprises is very large. A somewhat muddy story emerges from the quantile regressions, although one broadly supportive of that told by OLS. Hiring effects are, as with the OLS results, generally insignificant with the exception of skilled unionized workers at the 50' and 90' quantiles. Outside wages enter for all union worker types and quantiles and for unskilled unionized workers at the 10" quantile and skilled unionized workers at the 50'h. It is very possible that they would also enter significantly in the unskilled union sector at the 50' as well, as suggested by the OLS regressions if the number of observations were as large as for the union sample. Again, it is difficult to attribute the effect to being a reference for union bargaining even in the union sector. The magnitude of the impact for unskilled workers at the 10' quantile is double that for the union sector and almost double for skilled workers at the 50th. For both samples, at the 90' quantile, outside wage effects drop dramatically in significance and magnitude for skilled workers. A possible interpretation is that workers paid well given 12 their human capital are likely to receive a larger share of total remuneration in non-wage benefits and hence, the efficiency wage premium may appear as unobservable benefits. Vb. Empirical Results: Labor Demand The data allow the estimation of system of skilled and unskilled labor demand functions and hence the examination of unions' impact on the substitutability of factors and the allocative efficiency of firms. We estimate log-linear approximations to both equations. ns,. = w&+ Aq + w.afl. +,&u + x + &a nS,U is the log of labor demand for skilled (Ns) or unskilled labor (Nu), w=log(Ws, WU) the vector of log own and cross wages" q, log firm output, w, the vector of proxies for own and cross-expected alternative outside wages (log outside wages and the sectoral hiring rates), u the measure of union bargaining power, x the vector of other firm characteristics. Ideally, the labor demand equation would be estimated using instruments for wages due to possible measurement error and random productivity shocks." However, good instruments prove difficult to find. We are not working with panel data and, as the previous section suggests, the most complete model for the unskilled wage explains relatively little of the variance. The results, as Roberts and Skoufias found for Colombia, were counter- intuitive.14 Output should be considered endogenous also, due to measurement error (current output different from the output used in the decision making of the firm) and the presence of unobserved firm specific shocks that affect output and employment. This is confirmed by Durbin-Hausman-Wu (Davidson and MacKinnon, 1993) tests. With more success we instrument using the capital stock (and its square), capacity utilization levels and other firm specific technology variables. Chow tests for the equality of coefficients between union and non-union firms for skilled and unskilled workers respectively reject at conventional significance levels. Again, it is clearly not appropriate to combine the samples and we report separate regressions for each group. 1. General Results Table 3 presents the results from a static labor demand equation for the unionized and non-unionized sectors. The regressions are broadly consistent with standard factor demand 12 As in Roberts and Skoufias we do not have a measure for the capital services price. We use the corporation dummy as well as the level of automated machinery and firm size dummies to differentiate the firms on their opportunity costs of capital services. " Hours composition bias (Hamermesh) does not seem to be a problem in our data as the majority of firms did not change the number of weekly and daily shifts across a six month period in 94-95, during the tequila crisis. 14 In both union and non-union firms, we have an upward sloping demand curve for skilled labor. In the former a flat demand curve for unskilled workers appeared with "wrong" signed outside wage effects for both (yet consistent with the wrong insignificant inside wages coefficients). The large union effect was maintained for unskilled. Hiring rates never entered significantly. The (naive) R2 for unskilled fell from .68 to .15. We attribute the results to the unavailability of good instruments (hinted by the non-rejection of Durbin-Hausman- Wu tests for unskilled workers) and present only the results with non-instrumented wages and lagged output. 13 theory (see, for example, Chambers 1988 and Hamermesh) and other empirical studies. Own- wage elasticities strongly suggest a downward labor demand relation in both sectors. Output elasticities are statistically similar in both samples, as expected by theory under the hypothesis of a homothetic technology. Their small magnitude suggests that firms are operating in the downward sloping part of the long run average cost curve, as in a monopolistic competition model. Cross-wage effects are symmetric although in some cases we cannot reject that they differ from zero." Morishima elasticities of substitution between types of labor are in the 60- 35% range, and if we assume that total labor costs are half that of the capital services (10% of the capital stock) the capital-unskilled elasticity of substitution labor seems to be around 0.5 and nearly twice that of the capital-skilled labor, as in the references in chapter 3 of Hamermesh.16 Many of the proposed demand shift variables enter significantly and of expected sign. Capacity utilization, as a measure of factor usage is positive in all regression although clearly significantly so only for unskilled suggesting skilled labor hoarding. Exporting firms hire roughly 15-20% more unskilled and 0- to 4% more skilled workers as would be expected given that 82% of these firms engage in some maquila work that is particularly labor intensive. Foreign firms tend to hire more workers, with respect to national firms and after controlling for the other firm characteristics only statistically significantly so in the union skilled sample. Older firms have more skilled workers although not significantly more unskilled. Whether the firm is part of a large corporation enters positively for all and significantly except for the non-unionized unskilled. These results are broadly supported in the quantile regressions selectively reported in tables 3.1-3.3. Perhaps unsurprisingly, firms who have a large workforce after adjusting for the variables in the regression show higher output elasticities and hence smaller scale economies across all sub-samples operating closer to the long run minimum of the average cost curve. This is most striking for the case of non-union skilled workers where the 10' quantile has an output elasticity of .22 which rises to .63 in the 90t' quantile. Figures A2.1-2.4 in the appendix illustrate what these differences imply for the employment trajectory for each quantile level. The 90' quantile employ vastly more additional workers for an additional unit of output than the firm at the conditional median and this appears particularly to be the case for unskilled labor where the distribution appears far less symmetric than is the case for skilled workers." These firms with lower economies of scale also show higher elasticities of substitution. For skilled workers, own elasticities quadruple and double for non-union and union firms respectively across the quantile range and also substantially increase for unskilled ` Both labor types seem to be substitutes for capital, with the cross price unskilled labor-capital elasticity larger in absolute value than the skilled-capital elasticity. Roberts and Skoufias results suggest also monopolistic competition and skilled and unskilled workers have negative cross price elasticities, i.e., they are complements. 16 Morishima elasticities of substitution measure how much the ratio of factor used, rather than simply the amount of one factor, changes with respect to one of the factor prices. In unionized firms, the ratio of labor types mean wage costs is close to one and about 0.85 for non-union. We use the fact that factor demands are homogeneous of degree zero in prices to obtain the capital elasticities. " It must be noted that the standard errors employed here were not corrected for the two stage estimation as they indicate, so caution must be employed when interpreting them. They are presented here for completeness. 14 workers although less dramatically. This, increase across quantiles can be seen in appendix figures A2.1-2.4. No clear pattern emerges for cross elasticities across quantiles. 2. Union Effects Although the previous section suggested that there were no union effects on wages, union density has a strong positive effect on the level of employment in all regressions and across all quantiles. This suggests a unique example of efficient bargaining where unions accept the market wage, but cause firms to move off the demand curve to hire more labor. The 2SLS regressions suggest that for each 1% of the work force that is unionized, unskilled employment appears to increase by 2.7%. Despite this, we find little reduction in the magnitude of the shift terms, such as capacity utilization, foreign ownership, and membership in a larger corporation as the analytical overview suggests should happen, and the clear downward sloping demand relation is preserved. This suggests that the position on the contract curve may not be "too far" from the demand curve. However, estimates of the impact of unions on productivity (not shown) confirm the intuition that there is a significant adverse effect arising from the additional labor hired. The OLS regressions also find a negative union effect for skilled workers that suggests that unions represent primarily unskilled workers and "crowd out" those more skilled. A 1 point rise in the percentage unionized leads to a roughly equal .56% fall in the number of skilled laborers. However, again proving their usefulness, the quantile regressions suggest that this effect is significant only for those firms at the high end of employment for their firm type and may not represent a general pattern. As might be expected, union firms have statistically significant, although not substantially smaller own wage elasticities that might suggest slightly less flexibility in the allocation of unskilled workers. However, the pattern is reversed for skilled workers and the Quantile regressions suggest that, again, the large employment firms are driving this result. For conditionally small firms, the union sample own elasticity for unskilled labor is higher than that of the non-union sector and at the median, there is no significant difference. The presence of unions has little impact on output elasticities. A firm with 10% higher output is likely to hire roughly 3% more skilled and only slightly fewer unskilled workers in either sample. This is not necessarily surprising since a greater hiring rate of workers by unionized firms occurs on top of a larger base, leaving the elasticity virtually unchanged. In sum, unions appear to concentrate their efforts on featherbedding, but it is not obvious that they reduce the flexibility of allocating factors. 3. Efficiency wage effects. There is strong evidence of efficiency wage effects. As Nickell and Wadhwani show, the expected outside wages can enter in the demand function in the unionized sector even when there are no efficiency wage effects because unions may use them for reference. 15 However, the outside wage, and the probability of hiring enters strongly in both the union and non union sectors and of the predicted signs for unskilled workers, and either the outside wage or the hiring rate enter significantly and of the right sign in all specifications. Focusing on the non-union sector, the quantile regressions suggest that the outside wage effects are fairly consistent across quantiles-unimportant for skilled workers and of similar magnitudes and significant for unskilled. No clear story emerges from the coefficients on hiring rates. VI. Conclusion Two provocative findings emerge from this paper. First, Mexican unions do not seem to show "right to manage" behavior and there is very weak evidence that they affect wages in a positive direction. Nonetheless, they do greatly increase the quantity of unskilled workers hired and appear to "crowd out" skilled workers in what appears to be an extreme form of Efficient Bargaining. The impact on productivity is, by definition negative, but the flip side may be that unions are forcing firms to use "appropriate" technology that uses relatively less capital and more unskilled workers. As Layard and Nickell have noted, the general equilibrium effects on the overall level of labor employed in the economy are clear: if unions bargain over employment as well as wages, as seems to be the case here, employment in the union sector should be higher, under a smaller than unity elasticity of substitution, as also seems to be the case here. It is therefore possible that far from reducing the level of employment by rationing workers into the informal sector, Mexican unions may be preserving low skilled jobs, at the cost of smaller wages. The second important finding is that there appears to be strong evidence of efficiency wage effects. A tentative conclusion would be therefore, that whatever segmentation is observed, defined as equivalent workers earning different wages, is more likely to be emerging from the efficiency wage effects than from union effects. This would suggest that, even in the absence of minimum wages or union power, substantial segmentation will remain in Mexico as well as other LDCs. 16 References: Abuhadba, M and P. Romaguera (1993), "Inter-Industrial Wage Differeqtials:Evidence from Latin American Countries" Journal ofDevelopment Studies 30(1), 190-205. Bell, L (1997) "The Impact of Minimum Wages in Mexico and Colombia" Journal of labor Economics, 15(3), Part 2 S102-35. Blanchflower, D.C., N. Millward and A. J. Oswald (1991), "Unionism and Employment Behaviour," The Economic Journal, 101 815-834. Borjas, G. (1980) "The Relationship Between Wages and Weekly Hours of Work: the Role of Division Bias." The Journal of Human Resources 15(3) 409-423. Borjas, G. (1996), Labor Economics, New York:McGraw-Hill. Buchinsky, M. (1994) "Changes in the US Wage Structure 1963-87: An application of quantile regression". Econometrica, 62(3), 405-458. Chambers, R.G.(1 988) Applied Production Analysis, A Dual Approach, Cambridge: Cambridge University Press. Davidson, R. and MacKinnon, J.(1993). Estimation and Inference in Econometrics. New York:Oxford University Press. Dickens, W.T and L.F. Katz (1987) "Inter-industry Wage Differences and Industry Characteristics" in K. Lang and J.S. Leonard eds., Unemployment and the Structure ofLabor Markets, New York: Blackwell. Esfahani, H. and D. Salehi-Isfahani (1989) "Effort Observability and Worker Productivity: Towards an Explanation of Economic Dualism," Economic Journal, 99 818-836. Funkhouser, E. (1998) "The Importance of Firm Wage Differentials in Explaining Hourly Earnings Variation in the Large Scale Sector of Guatemala" Journal of Development Economics 55(1), 115-131. Hamermesh, D.(1993). Labor Demand. Princeton: Princeton University Press. Harris, J.R. and M.P. Todaro (1970) "Migration, Unemployment and Development: A Two Sector Analysis," American Economic Review, 60:1, 126-142. Hendricks and Kahn (1991), "Efficiency Wages, Monopoly Unions and Efficient Bargaining," Economic Journal, 10 1(408), 1149-62. Koenker, R. and G. Bassett (1982), "Regression Quantiles", Econometrica, 46, 33-50. 17 Krueger, A.B and L. H. Summers (1988), "Efficiency Wages and the Inter-Industry Wage Structure", Econometrica 56(2), 259-293. Layard, R. and Nickell, S.(1990). Is Unemployment Lower If Unions Bargain Over Employment? Quarterly Journal ofEconomics 105(2), 773-787. Lewis G.H. (1986) "Union Relative Wage Effects" in: 0. Ashenfelter and R. Layard eds., Handbook of Labor Economics, v.2, Amsterdan:North-Holland. Marquez, G. (1990) "Wage Differentials and Labor Market Equilibrium in Venezuela," Unpublished Ph.D. Dissertation, Boston University. Marquez, C. and J. Ros(1990), "Segmentacion del Mercado de Trabajo y Desarrollo Economic en Mexico", El Trimestre Economico, Fondo de Cultura Economica, Mexico, 17:2 Nickell, S. and Wadhwani,W. (1990) "Employment Determination in British Industry: Investigations Using Micro-Data," Review ofEconomic Studies, 58(5), 955-969. Oswald, A. J.(1991) Efficient Contracts Are on the Labour Demand Curve, Labour Economics, 1(1), 85-113. Panagides, A. and H.A. Patrinos (1269), "Union-Non-Union Wage Differentials in the Developing World" World Bank Policy research working Paper 1269. Phelps, E.(1994) Structural Slumps: The Modern Theory of Unemployment, Interest, and Assets, Cambridge, MA: Harvard University Press. Roberts, M. and Skoufias, E.(1997). The Long Run Demand for Skilled and Unskilled Labor in Colombian Manufacturing Plants. Review ofEconomics and Statistics 79(1), 330-334 Schaffner, J.A.(1998) "Premiums to Employment in Larger Establishments: evidence from Peru. Journal ofDevelopment Economics 55(1), 81-113. Stiglitz, J.E.(1 974) "Alternative Theories of Wage Determination and Unemployment in LDC's: The Labor Turnover Model, " Quarterly Journal of Economics, 88(1), 194-227. Wadhwani, S. B and M. Wall (1991), "A Direct Test of the Efficiency Wage Model Using UK Micro-Data," Oxford Economic Papers, 43(2), 529-548. Weiss, A. (1990). Efficiency Wages. Princeton: Princeton University Press. 18 Appendix I: Measures to Confront the Problem of Recently Trained Workers Who Resign, % Total Large Medium Small Micro Increase wages 23.1 7.1 7.6 9.5 29.2 Increase other remunerations 4.6 17.5 17 14 0.1 Promote those trained 12.7 39.6 32.8 24.3 6.1 Reduce the # of those trained 8.4 0.2 1.3 1.3 11.5 Reduce the training offered 0.2 0.5 0 0 0.3 Give non-monetary recognition 8.5 12.8 9.8 7.3 8.4 None 33.5 15.5 26.1 35 34.7 Don't know 1.2 2.3 0.7 5.2 0 Others 7.8 3.9 4.1 3.2 9.7 Source: 1992 ENESTYC Table I - SUMMARY STATISTICS Non-Union (n=73 1) Union*(n=3,42 1) Variable Mean Std.Dev Mm Median Max Mean Std.Dev. Min. Median Max Forg 0.331 0.471 0.000 0.000 1.000 0.198 0.398 0.000 0.000 1.000 Age 17.036 13.414 1.000 12.000 99.000 25.188 16.123 1.000 23.000 99.000 Capu* 75.104 19.339 5.000 80.000 100.000 74.479 18.102 1.000 80.000 100.000 Export 0.438 0.496 0.000 0.000 1.000 0.241 0.427 0.000 0.000 1.000 R_d 0.334 0.472 0.000 0.000 1.000 0.383 0.486 0.000 0.000 1.000 Tech. 0.454 0.498 0.000 0.000 1.000 0.508 0.500 0.000 1.000 1.000 Qual.ctr. 0.988 0.110 0.000 1.000 1.000 0.997 0.057 0.000 1.000 1.000 Corp* 0.252 0.434 0.000 0.000 1.000 0.257 0.437 0.000 0.000 1.000 Union 0.680 0.074 0.477 0.703 0.854 Log(Ws) 9.422 0.907 5.991 9.573 11.352 9.827 0.726 5.849 9.909 12.110 Log(Ns) 1.587 1.011 0.000 1.444 6.685 1.973 0.956 0.000 1.890 6.472 Log(Wu) 7.765 0.754 5.758 7.744 10.134 7.983 0.688 5.659 8.032 11.160 Log(Nu) 3.449 1.342 0.000 3.466 8.964 3.842 1.135 0.000 3.809 8.606 Sch. Sk* 11.962 2.120 3.000 12.150 18.500 12.150 1.708 4.909 12.233 17.843 Sch.unsk 6.603 1.497 3.000 6.450 12.000 6.873 1.321 3.000 6.783 12.000 Exp. Sk. 5.495 3.385 0.294 4.483 25.640 6.631 4.000 0.065 5.663 32.927 Exp.unsk 4.124 3.787 0.000 3.000 38.000 5.817 4.864 0.000 4.118 40.000 Lq* 8.274 1.717 1.792 8.343 14.878 9.250 1.597 1.386 9.263 . 15.072 Prd. Sk 6.686 1.204 1.068 6.750 12.472 7.277 1.116 0.372 7.328 13.008 Prd.unsk 4.824 1.441 -0.508 4.797 10.811 5.408 1.241 -2.635 5.435 10.659 Lwa unsk 7.837 0.210 6.742 7.899 8.189 7.847 0.193 6.742 7.825 8.189 Lwa sk 9.563 0.325 7.784 9.578 10.182 9.641 0.293 7.784 9.708 10.182 Auto* 12.726 23.325 0.000 0.000 100.000 11.837 22.601 0.000 0.000 100.000 Log(K) 7.716 1.980 2.303 7.690 14.526 9.021 2.000 1.609 9.131 15.735 Log K/Nu 3.186 1.838 -2.896 3.232 9.573 4.013 1.666 -3.626 4.191 10.125 Log K/Ns 4.342 1.635 -0.984 4.397 10.978 5.040 1.514 -1.686 5.194 10.877 Hirr.unsk 7.512 3.972 0.322 6.628 15.240 6.857 3.468 0.322 6.052 15.240 Hirr sk. 1.079 0.512 0.000 1.043 2.641 1.199 0.602 0.000 1.091 2.641 p.train 0.211 0.408 0.000 0.000 1.000 0.236 0.425 0.000 0.000 1.000 Cmptl 0.512 0.500 0.000 1.000 1.000 0.631 0.483 0.000 1.000 1.000 Source: author's calculation from ENESTYC '92. Large, medium and small firms only (from 16 employees on). * - Mann-Whitney test does not reject equality of distributions between union and non-union samples. See variable definitions in text. 20 Table 2 - WAGE EQUATIONS Non-union Union Skilled Unskilled Skilled Unskilled Prd 0.036 (0.028) 0.097 (0.025) a 0.036 (0.012) a 0.039 (0.012) a Lwa 0.384 (0.156) a 0.388 (0.204) b 0.300 (0.057) a 0.433 (0.086) a Union --- --- --- --- -0.865 (0.2 10) a 0.236 (0.197) Hirr -0.002 (0.090) -0.014 (0.010) -0.065 (0.029) a -0.002 (0.004) Sch 0.424 (0.081) a 0.235 (0.103) a 0.571 (0.058) a 0.238 (0.055) a Sch2 -0.016 (0.003) a -0.017 (0.007) a -0.022 (0.002) a -0.014 (0.004) a Exp -0.021 (0.023) 0.000 (0.015) 0.027 (0.007) a 0.018 (0.006) a Exp2 0.001 (0.001) 0.000 (0.001) -0.001 (0.000) a -0.001 (0.000) a R_d 0.121 (0.065) b 0.004 (0.064) 0.010 (0.025) -0.035 (0.026) Tech. 0.030 (0.063) 0.078 (0.061) 0.097 (0.024) a 0.135 (0.026) a P.train. 0.197 (0.067) a 0.110 (0.065) b 0.024 (0.025) 0.001 (0.026) Corp -0.038 (0.081) -0.008 (0.079) 0.028 (0.026) 0.067 (0.028) a Forg. 0.327 (0.096) a 0.127 (0.093) 0.179 (0.030) a 0.025 (0.032) Export. -0.123 (0.074) b 0.012 (0.073) -0.108 (0.026) a 0.024 (0.028) Auto 0.001 (0.001) 0.001 (0.001) 0.001 (0.000) b 0.000 (0.001) Cmptl -0.013 (0.055) 0.026 (0.054) -0.008 (0.022) -0.003 (0.023) Qual.ctr. 0.083 (0.248) -0.193 (0.242) 0.094 (0.186) 0.238 (0.197) LogK/L 0.001 (0.019) -0.026 (0.018) 0.016 (0.008) b -0.014 (0.008) b Sect_32 0.046 (0.110) 0.072 (0.101) -0.008 (0.043) -0.028 (0.042) Sect_33 -0.018 (0.143) 0.094 (0.126) 0.063 (0.068) 0.075 (0.069) Sect_34 0.129 (0.134) 0.200 (0.156) 0.072 (0.056) 0.103 (0.063) Sect_35 0.162 (0.124) 0.059 (0.132) 0.116 (0.044) a 0.061 (0.047) Sect_36 0.142 (0.194) -0.006 (0.183) 0.079 (0.060) 0.078 (0.061) Sect_37 0.226 (0.233) -0.192 (0.227) -0.011 (0.065) -0.068 (0.064) Sect_38 0.035 (0.102) 0.107 (0.110) 0.085 (0.038) a 0.065 (0.042) Sect_38 -0.503 (0.256) b -0.045 (0.250) 0.031 (0.090) 0.117 (0.094) Medium -0.132 (0.077) b -0.085 (0.075) -0.229 (0.025) a -0.091 (0.026) a Small -0.702 (0.083) a -0.349 (0.082) a -0.676 (0.032) a -0.335 (0.033) a Const. 2.912 (1.500) b 3.943 (1.643) a 3.523 (0.680) a 3.082 (0.732) a R2 0.3987 0.1692 0.2895 0.1191 F - test 17.26 a 5. 30 a 49.43 a 16.39 a Note- Sample sizes: Non-union n=73 1, Union, n=3,422. Chow tests for equality of union and non-union coeff.(27d.f.) 86.66 a (skilled), 42.37 a (unskilled) a-significant at the 5% level, b-significant at the 10% level. 21 Table 3 - LABOR DEMAND EQUATIONS - 2SLS Non-union * Union Skilled Unskilled Skilled Unskilled Lq * 0.441 (0.039) a 0.493 (0.049) a 0.480 (0.018) a 0.405 (0.019) a Lws -0.252 (0.038) a -0.082 (0.047) b -0.379 (0.019) a -0.024 (0.020) Lwu -0.057 (0.036) -0.636 (0.045) a 0.037 (0.016) a -0.591 (0.017) a Lwas -0.044 (0.174) -1.028 (0.219) a -0.091 (0.073) 0 195 (0.079) a Lwau 0.094 (0.255) 1.423 (0.320), a -0.002 (0.121) -0.244 (0.132) b Union --- -- --- --- -0.417 (0.243) b 2.689 (0.264) a Hirrs 0.237 (0.078) a -0.154 (0.098) 0.081 (0.033) a 0.067 (0.036) b Hirru -0.008 (0.009) 0.036 (0.011) a 0.008 (0.004) a 0.001 (0.004) Capu -0.001 (0.001) 0.004 (0.002) a -0.001 (0.001) b 0.003 (0.001) a Corp 0.154 (0.068) a 0.190 (0.086) a 0.080 (0.025) a 0.059 (0.027) a Forg 0.068 (0.082) 0.124 (0.104) 0.155 (0.029) a 0.010 (0.032) Age 0.002 (0.005) -0.003 (0.006) 0.004 (0.002) a -0.003 (0.002) Age2 0.004 (0.007) 0.003 (0.009) -0.002 (0.002) 0.008 (0.003) a Export 0.056 (0.064) 0.233 (0.081) a 0.068 (0.026) a 0.234 (0.028) a Auto 0.001 (0.001) 0.001 (0.001) 0.000 (0.000) 0.000 (0.000) Qual.ctr. -0.215 (0.209) 0.253 (0.263) -0.060 (0.176) 0.048 (0.192) R_d 0.073 (0.055) -0.001 (0.070) 0.042 (0.024) b -0.039 (0.026) Tech. -0.064 (0.054) -0.020 (0.068) -0.016 (0.023) -0.027 (0.025) Sect_32 0.200 (0.106) b -0.115 (0.134) 0.075 (0.045) b 0.182 (0.049) a Sect_33 0.344 (0.120) a 0.161 (0.151) 0.228 (0.065) a 0.163 (0.071) a Sect_34 0.370 (0.148) a -0.647 (0.186) a 0.292 (0.065) a 0.010 (0.071) Sect_35 0.056 (0.114) -0.456 (0.143) a 0.283 (0.049) a -0.025 (0.053) Sect_36 0.268 (0.164) 0.166 (0.207) 0.370 (0.058) a 0.150 (0.063) a Sect_37 0.368 (0.199) b 0.064 (0.250) 0.253 (0.061) a 0.016 (0.067) Sect _38 0.361 (0.101) a -0.331 (0.128) a 0.371 (0.046) a 0.177 (0.050) a Sect_38 0.057 (0.223) -0.319 (0.280) 0.100 (0.088) 0.003 (0.096) Medium -0.396 (0.076) a -0.528 (0.095) a -0.339 (0.030) a -0.612 (0.033) a Small -0.611 (0.109) a -1.207 (0.136) a -0.524 (0.050) a -1.129 (0.054) a Const. 0.484 (1.538) a 3.789 (1.934) b 1.986 (0.690) a 3.212 (0.752) a R2 0.6518 0.6872 0.6339 0.6918 F -test 46.77 a 59.47 a 210.25a 269.56 a DHW 32.52 a 19.68 a 72.44 a 151.53 a Note. Sample sizes. Non-union n=731, Union, n=3,422. Chow tests for equality of union and non-union coeff.(26d.f.) 36 23 b (skilled), 130.89 a (unskilled). a-significant at the 5% level, b-significant at the 10% level. * Instruments for output: capital stock, its square, sector dummies and its interactions and technology variables. 22 Table 2.1 WAGE EQUATIONS (10% Quantile) Non-union , Union Skilled Unskilled Skilled Unskilled Prd 0.025 (0.062) 0.094 (0.036) a -0.004 (0.030) 0.026 (0.030) Lwa 0.219 (0.250) 0.803 (0.202) a 0.462 (0.114) a 0445 (0.178) a Union --- --- --- -0.910 (0.461) a 0.825 (0.414) a Hirr -0.138 (0.160) -0.018 (0.012) -0.101 (0.065) -0.003 (0.008) Sch 0.458 (0.128) a 0.249 (0.099) a 0.898 (0.132) a 0.212 (0.111) b Sch2 -0.016 (0.006) a -0.016 (0.007) a -0.034 (0.005) a -0.012 (0.008) Exp -0.045 (0.039) -0.019 (0.019) 0.054 (0.015) a 0.008 (0.011) Exp2 0.001 (0.002) 0.000 (0.001) -0.002 (0.001) a -0.001 (0.000) Pseudo-R2 0.2992 0.1128 0.213 0.0652 Table 2.2 WAGE EQUATIONS (50% Quantile) Non-union Union Skilled Unskilled Skilled Unskilled Prd 0.036 (0.027) 0.116 (0.031) a 0.034 (0.011) a 0.034 (0.014) a Lwa 0.566 (0.151) a 0.303 (0.255) 0.318 (0.054) a 0.486 (0.099) a Union --- --- --- --- -0.907 (0.199) a 0.259 (0.226) Hirr -0.085 (0.086) -0.009 (0.013) -0.080 (0.028) a -0.003 (0.005) Sch 0.408 (0.074) a 0.294 (0.13 1) a 0.588 (0.054) a 0.305 (0.063) a Sch2 -0.015 (0.003) a -0.021 (0.009) a -0.022 (0.002) a -0.019 (0.004) a Exp -0.029 (0.023) -0.012 (0.018) 0.024 (0.007) a 0 019 (0.007) a Exp2 0.001 (0001) 0.000 (0.001) -0.001 (0.000) a -0.001 (0 000) a Pseudo R2 0.2435 0.0929 0.1578 0.0650 Table 2.3 WAGE EQUATIONS (90 %Quantile) Non-union Union Skilled Unskilled Skilled Unskilled Prd 0.037 (0.049) 0.086 (0.033) a 0.078 (0.019) a 0.068 (0.0 18) a Lwa -0.020 (0.373) 0.306 (0.396) 0.189 (0.081) a 0.539 (0.129) a Union --- --- --- --- -0.988 (0.316) a 0.339 (0.296) Hirr 0.185 (0.151) 0.011 (0.015) -0.081 (0.044) b 0.000 (0.006) Sch 0.319 (0.135) a 0.439 (0.123) a 0.179 (0.079) a 0.235 (0.073) a Sch2 -0.010 (0.006) b -0.034 (0.008) a -0.006 (0.003) b -0.014 (0.005) a Exp 0.005 (0.046) 0.012 (0.022) 0.014 (0.012) 0.020 (0.008) a Exp2 0.000 (0.002) -0.001 (0.001) 0.000 (0.001) -0.001 (0.000) a Pseudo R2 0.1670 0.1647 0.1266 0.0834 Note: Sample sizes. non-union n=731, Union, n=3422. Chow tests for equality of unin and non-union coeff. (27d f.) 80.57a (skilled), 75 1l a (unskilled). a- significant at the 5% level, b-significant at the 10% level. Pseudo R2=1-(minimum sum of deviations over the raw sum of deviations). Variables used but not presented- r d,t ech, p.train, corp, forg, export, auto, cmprtl,qual.ctr.,Log(k/I), 8 sector dummies, 2 size dummies and the constant. See the least squares wage regressions 23 Table 3.1 - LABOR DEMAND - 2SRQ (10% Quantile) Non-union Union Skilled Unskilled Skilled * Unskilled Lq 0.223 (0.064) a 0.402 (0.089) a 0.414 (0.024) a 0.356 (0.027) a Lws -0.097 (0.052) b -0.159 (0.081) a -0.226 (0.027) a -0.038 (0.030) Lwu 0.045 (0.047) -0.469 (0.081) a 0.083 (0.021) a -0.580 (0.027) a Lwas -0.115 (0.224) -1.063 (0.449) a -0.187 (0.098) b 0.328 (0.114) a Lwau -0.004 (0.326) 1.974 (0.651) a 0.051 (0.177) -0.192 (0.207) Union --- --- --- --- -0.171 (0.319) 2.407 (0.377) a Hirr.sk 0.200 (0.105) b -0.375 (0.182) a 0.111 (0.044) a -0.089 (0.053) b Hirr.unsk -0.002 (0.011) 0.029 (0.021) 0.003 (0.005) 0.009 (0.006) Capu 0.001 (0.002) 0.005 (0.003) -0.002 (0.001) a 0.005 (0.001) a R2 0.3648 0.4241 0.3634 0.4273 Table 3.2 - LABOR DEMAND - 2SRQ (50% Quantile) Non-union Union Skilled Unskilled Skilled Unskilled Lq 0.402 (0.049) a 0.416 (0.062) a 0.460 (0.023) a 0.402 (0.021) a Lws -0.220 (0.048) a 0.000 (0.060) -0.341 (0.024) a -0.012 (0.022) Lwu -0.025 (0.046) -0.638 (0.058) a 0.022 (0.021) -0.621 (0.019) a Lwas -0.024 (0.218) -1.027 (0.271) a -0.044 (0.094) 0.071 (0.085) Lwau 0.170 (0.320) 1.594 (0.402) a -0.076 (0.156) -0.141 (0.140) Union --- --- --- --- -0.215 (0.314) 2.546 (0.282) a Hirr.sk 0.231 (0.097) a -0.116 (0.124) 0.105 (0.043) a 0.072 (0.039) b Hirr.unsk -0.021 (0.012) b 0.018 (0.015) 0.010 (0.005) b -0.002 (0.005) Capu 0.001 (0.002) 0.006 (0.002) a -0.001 (0.001) 0.003 (0.001) a R2 0.4143 0.5067 0.4154 0.4605 Table 3.3 - LABOR DEMAND - 2SRQ (90% Quantile) Non-union Union Skilled Unskilled Skilled Unskilled Ly 0.627 (0.074) a 0.573 (0.093) a 0.536 (0.034) a 0.430 (0.037) a Lws -0.394 (0.074) a -0.113 (0.093) -0.519 (0.038) a -0.037 (0.037) Lwu -0.111 (0.064) b -0.743 (0.074) a -0.043 (0.026) -0.633 (0.033) a Lwas 0.142 (0.310) -1.189 (0.397) a -0.157 (0.125) 0.235 (0.137) b Lwau 0.196 (0.433) 1.686 (0.535) a 0.117 (0.206) -0.482 (0.222) a Union --- --- --- --- -0.779 (0.408) b 2.950 (0.486) a Hirr.sk 0.164 (0.140) 0.000 (0.210), 0.075 (0.055) 0.125 (0.065) b Hirr.unsk -0.009 (0.014) 0.046 (0.018) a 0.010 (0.007) -0.002 (0.008) Capu -0.004 (0.002) b 0.002 (0.003) -0.001 (0.001) 0.002 (0.001) R2 0.4538 0.5236 0.4288 0.4615 Note. Sample sizes- non-union n=731, Union, n=3422. Chow tests for equality of union and non-union coeff. (27d.f) 83.01a (skilled), 79 45a (unskilled). a- significant at the 5% level, b-significant at the 10% level. Pseudo R2=1-(mnimum sum of deviations over the raw sum of deviations). Variables used but not presented. r _d, tech, corp, forg, export, auto, qual.ctr., age, age squared, 8 sector dummies, 2 size dummies and the constant. See the least squares labor demand regression. 24 Figure 1: Relationship Between Iso-Profit Curves and the Labor Demand Curve Dollars P Profits = $100,000 0 Profits = $150,000 Employment 0 50 100 150 Figure 2: Bargaining Solutions Between Unions and Firms Dollars WZ* z U 0* UR 0z UI U. EZ E, Employment 25 Figure 3: Quantile Regression 90h % 50,h% 10h% 26 27 и р я �, � � � .'� ,� й � я и � � д � о ^.. л •G � , � ,�,о,, � � � А " i �� w Е -' � .� �'" Q � = г� � ° ,� � 3 �� �� �� и� � и ее � � •С � е�б а � а� г Е� 1. Introduction Crude estimates suggest that the informal production sector is largd, accounting for 20 to 50 percent of employment in many developing countries (Portes, 1994). Yet progress towards consensus on the sector's origins, operations and even definition has been hampered by two problems.' First, the lack of comprehensive data has prevented accurately establishing the basic characteristics of informal production beyond conjecture and casual observation. Second, whereas the literature on informal finance -- i. e. unregulated financial intermediation -- has a broad theoretical underpinning,' the literature on informal production tends toward ad hoc characterizations and lacks a comparably broad foundation.' In general, these frameworks rely on an institutional distortion such as a binding minimum wage, evasion of government regulation and taxation, or differences between firms in worker monitoring ability to generate the informal sector.' This paper makes two contributions. First, it offers systematically collected data on a broad cross-section of urban firms in Mexico with details on compliance with or participation in a number of different societal institutions. The data are derived from a nationally representative sample of all such firms, a significant improvement over pre-existing case study data sets. We are thus able to move beyond anecdotal analysis and establish some definitive stylized facts about informal production for the first time. Second, it offers a theoretical framework to motivate the analysis of the data. The approach is unique because it assumes that informal firms behave no differently from small firms in industrialized counties and that no institutional or governmental distortions are required to generate their behavior. To this end the analysis builds on recent mainstream empirical and theoretical research on firm dynamics and extends it to incorporate a general concept of formality. The traditional view of tax and regulatory compliance is that government enforcement is the sole determinant.' In contrast, we argue that voluntary compliance may arise because the firm derives ' A large body of literature equates informality with the low-wage, low-productivity segment of a dual labor market (for example, Lewis, 1954, and Harris and Todaro, 1970). An equally sizeable literature equates informality with unregulated self-employment (for example, Hart, 1972, and de Soto, 1989). See Thomas (1992) and Portes (1994) for excellent overviews. Our approach in this paper equates informality and noncompliance with societal norms such as tax obligations, labor protections, census enumerations, business guild participation, etc. In line with both Thomas' and Portes' characterizations, we are concerned with unregulated/unmonitored activities that are ostensibly legal, not those that are truly illegal (criminal). See Besley (1995) for an excellent overview. See Thomas (1992) and Portes (1994). Exceptions include Esfahani and Salehi-Isfahani (1989), Rauch (1991), Loayza (1995), and Banerji and Jain (1996). 4 For the remainder of the paper we will use "informal" exclusively to characterize the production and distribution of goods processes. For example see Ashenfelter and Smith (1979), Fenn and Ve1janovski (1988), Cowell 0 990). either direct or complementary benefits from participating in a particular societal institution. Several appealing results emerge. First, the framework is able to gederate many of the cross- sectional patterns of firm and worker behavior addressed by existing models and those found in our data. Second, previous approaches have been static: firms are either formal or informal and none transition in equilibrium. However, empirical evidence suggests that developing country (LDC) firms share some of the evolutionary dynamics of their industrialized country counterparts. We show that these dynamics may be important when analyzing informality because they can generate firm characteristics commonly associated with the formality-informality comparison. Moreover, such dynamics imply equilibrium transitions from informality to formality. The nature of the data employed does not permit following individual firms over time and hence precludes rigorous testing of the dynamic predictions. However, we add a new dimension to the theoretical literature and the predicted cross sectional patterns are supported empirically. Finally, the framework can nest many of the existing conceptions of informality, including models that generate the sector through governmental or institutional distortions. I. Formality as participation in civic institutions Different contributions in the literature view compliance with or participation in the institutions of society in seemingly inconsistent ways. Some emphasize firms' desires to evade taxes, regulations or other state controls (for example, Loayza, 1995). Others see the inability to access institutions, such as those securing property rights, as hampering firm growth (for example, de Soto, 1989). Further there tends to be an assumption that formality is an all or nothing state. We argue that these views are valid only as special cases of a more general and continuous relation between the firm and society. We recast the question of formality as the firm's decision of how much to participate in the numerous institutions of civil society: federal and local treasuries, governmental programs such as social security (including pensions and health care), the legal system, the banking system, health inspection, firm censuses, trade organizations, civic organizations, etc. We argue that a minimal degree of participation in some institutions is a necessary input to growth for many firms, and that participation increases with the success of the business. That is, formality can be viewed as a normal input to production: q = f(L,K,P), where L is labor, K is capital, and P is participation in (a number of different) societal institutions, and all three inputs are complementary. The benefits of formality, while often overlooked, are numerous. They include, but are not limited to:' 6 See also de Soto (1989). 2 1. Enforceable/impersonal contracts and credible signaling. All entrepreneurs have access to social relationships to enforce implicit contracts among their friends and family, who form a small number of their potential customers and employees. Participation in the legal system is needlessly expensive for transactions with these individuals. Similarly, old age and health insurance may be easily handled by insuring through their mutual extended network of friends and family. Property rights secured by personal ties may be sufficient if investment is minimal. These characteristics of small scale economic transactions are commonly observed in developing countries, as well as in many ethnic enclaves in developed countries. But this mode of operation is constrained by the ability of the entrepreneur to maintain personal relations with all involved parties, a task increasingly unmanageable as firms expand. Legally recognized, enforceable contracts lend credibility to arrangements, permit entry into long term commitments, diminish risk, and can reduce monitoring costs. For example, in a world of imperfect information, certification that the firm complies with government health and safety codes may be necessary for firms to attract the largest customer base possible.' Larger investments require that property rights be secured through the legal system. 2. Access to capital. Informal capital markets (Besley, 1995) may be sufficient to fulfill the firm's external financing needs at low levels of production. However, the small scale and undiversified nature of informal capital markets makes them unsuitable for satisfying the firm's financing needs at larger scales of operation. Growing firms will turn to formal financial intermediaries such as banks. 3. Access to public risk-pooling mechanisms. In order to attract good quality workers the firm may have to offer fringe benefits such as workers compensation, health/unemployment/disability insurance, and pensions. However, uncertainty over the expected costs of these benefits is high for risk pools with limited numbers of participants, i.e. small firms. Indeed, there is evidence that United States firms backed the introduction of a workers' compensation system to decrease the risk of self- insuring against individual claims (Fishback and Kantor, 1996). Hence, even in the absence of mandatory enrollment laws, a firm may want to enroll in government programs that pool risks over a larger population than its own employees. In exchange for this participation, society imposes "taxes" such as reporting requirements,8 fiscal obligations, or social insurance payments. We can conceive of these as comprising an initial ' While we frame the empirical discussion in terms of formal versus informalfirms, the concept of informality also applies to subsets of transactions that an ostensibly formal firm may undertake. For example, Palay (1984, 1985) shows that certain transactions between rail-freight shippers and their clients in the United States can be characterized as informal because they occur outside the bounds defined by regulation, and hence are legally unenforceable. In keeping with our motivation here, we would expect such informal transactions to take place primarily between two parties that have a long-standing relationship, even if both parties are large firms and not individual people. A different perspective is offered by Portes (1994) who notes that a portion of economic activity at officially-sanctioned firms often goes unreported; that portion of transactions should be considered informal. This is particularly relevant for bank financing. The firm may have to become registered when it seeks such financing: the government may require the bank to report the identity of all its loan recipients for tax or other purposes. 3 fixed cost po that may include information or initial registration costs such as those documented by de Soto (1989), and per period costs, p,, such as taxation that we assume for simplicity are the same for all firms.' We initially assume that the market for formality is voluntar (society levies no costs on firms that choose not to participate in an institution) and that non-payers are perfectly excluded (no free riders). While extreme, these assumptions are consistent with voluntary health or social security programs, and business associations. For example, Chile's self-employed are offered the choice of whether to participate in the state social security program (The Economist, 1996). Just as importantly, our approach highlights an important effect that is not considered by the standard approach in the literatures on tax evasion and regulatory compliance (for example, Cowell, 1990, Fenn and Veljanovski, 1988). These assume that enforcement is the only determinant of compliance because no private benefit is derived from participation: the institution is treated as a strict public good. However, there may be private benefits that make compliance in many public institutions voluntary. In the mandatory workers' compensation system example cited above, the private benefit of participation outweighed the private cost for many, if not all, firms. De Soto claimed that Peruvian sidewalk vendors sought, not to avoid but, to pay taxes as a way to establish property rights over their precarious business locations. In reality, though the direct private benefit from paying taxes may be zero (again, assuming no enforcement penalties), there may be ancillary benefits that make compliance worthwhile.0 This very stylized concept of participation can now be embedded in a model of firm dynamics that has become popular in the industrial organization literature." A number of the existing models of the informal sector (e.g. Rauch, 1991) are motivated by Lucas' (1978) model of the size distribution of firms: Lucas argued that there is a distribution of entrepreneurial ability in the population: Those with a sufficiently high level of proficiency become entrepreneurs, while the rest become wage workers. Among the entrepreneurs, those who are more proficient have firms that are larger and/or more successful. However, the model is static: firms do not grow or fail, nor are they born; no one transitions between wage work and self-employment in equilibrium. 9 p, could increase with firm size, i.e. p, = r(q)*q. So long as d T/dq <0, the basic conclusions about participation and firm size and age would not change. '0 Even in cases where the private benefit of participation does not exceed the private cost, the net private cost may differ substantially, leading to different probabilities of compliance conditional on a given level of enforcement resources. For example, it may be quite difficult for a firm to undo the effects of a binding minimum wage if the compensation package does not include fringe benefits that can be reduced when the wage is raised. In contrast, it may be easier for the firm to comply with mandated health, pension or other benefits programs by adjusting the wage without significantly altering labor input (for example, Gruber, 1994). Our general point is that the probability of compliance is a positive function of the relative private benefit of participation (net of private costs). " See also Lippman and Rumelt (1982) and Ericson and Pakes (1995). 4 Jovanovic (1982) addressed these limitations by further assuming that entrepreneurs have uncertainty over their firms' true costs of production: Their precise entrepreneurial ability initially is unknown and can only be learned gradually over time by actually operatIng a business. Potential entrepreneurs' idiosyncratic entrepreneurial ability, 0, affects their costs, c(q)x,, through a multiplier x,(6+c,), where c(q) is convex,12 q is output, and e, are random firm specific shocks that prevent certain knowledge of 0. Entrepreneurs make their best guess of X e (the expectation of x, conditional on information received prior to time t), pay a one time fixed cost of entry, and thereafter choose a level of output qt to maximize expected profits: maxq [P,q, - c(q)x,e] where P, is the (price-taking) firms' output price. Each period firms get new information on their cost structure from the level of profits. Firms that realize profits above their expected level revise downward their estimate, xte, because 7 - e= -c(q)(x, - xte) (2) This yields two important predictions. First, Jovanovic showed from equation (1) that 8t_ C/ < 0 (3) 8X e x e // I I which, given the properties of the cost function, implies that a lower cost multiplier raises the level of output. Thus, longstanding firms differ in size because some firms discover that they are more efficient than others. Since participation is a normal input in the production process, the distribution of formality among established firms reflects the underlying distribution of 0. Second, this learning process broadly defines firms' trajectories of growth and formality over time. Unexpectedly good information on profits leads to a downward revision in xe,, and a rise in qt, above q,; i.e. the firm grows. It also permits more precise estimates of 0, making viable firms more confident that they will survive. Both elements influence the choice of the degree of participation: A firm will choose to become formal if the discounted benefit net of pt across the expected lifetime of the firm exceeds the fixed costs, po. 12 That is, c'(q)>O, c"(q)>O, c(O)=O, c'(O)=O. 5 Figure 1 presents three highly stylized alternate firm trajectories." A new small firm that realizes profits that suggest a high xc,,, will stop growing at a relatively small size. These "Type 2" firms -- the small survivors -- include businesses such as corner grocery sfores, push cart vendors, and door-to-door sales operations with relatively high 0. Given the relatively low benefits of formality for small firms, the expected discounted present value of participation may not exceed po until the firm is very confident about its long run viability, if ever. In contrast, a firm realizing large unexpected profits will sharply revise downward its x0,1 and set q, I much higher than q, These "Type 1" businesses in Figure 1 -- the large survivors -- also start small but rapidly expand to a large long-run size. Examples of this type of firm are medium- to large-scale manufacturing plants and wholesale trade warehouses. Finally, Type 3 firms are the false starters that quickly learn that they are unprofitable, and fail.'" The population of young firms contains a disproportionate number of such firms that have not yet received enough signals on 0 to figure out that they are not viable. The combination of their small size and uncertainty about being able to recoup po over their expected lifetime makes them unlikely to choose to become formal. Figure 2 presents these relations in a very stylized fashion. It shows alternate expansion paths -- with and without participation in a societal institution -- for the types of firm from Figure 1. The expansion paths with participation are net of the variable costs of participating, p; and have been drawn so that the percentage increase in revenue is approximately the same for firm types 1 and 2. Comparable proportionate increases in net revenue for small and large firms is a reasonable assumption given that participation is a complement to the other inputs to production. However, as explained below, it is not crucial for the key conclusions to be drawn from the analysis. One feature of the Jovanovic model is that there is a common failure bound for all firms in an industry, a size below which no firm can profitably operate. Large firms are farther from the failure bound, so they have a higher survival probability. This translates into a longer expected lifetime at any given age. Suppose D is the length of expected firm life -- measured from the curreAt period forward, not from the date of firm formation -- at which the discounted present value of the net benefits of formality (net of pt) exactly equals po. Those firms with expected lifetimes greater than D -- the larger firms -- would choose to participate at an early age, e.g. TI; the smaller firms with shorter expected lifetimes would defer until a later age, e.g. T2. Realistically, as shown in Figure 2, the benefits of participation are likely to be greater for larger firms. This simply accentuates the positive relationship between size and participation: larger firms realize greater per period benefits from formality and they expect to reap those benefits over a longer period. " Those shown in Figure I are for illustrative purposes and do not exhaust the range of possible firm types. 14 Jovanovic showed that there exists a maximum level of the cost multiplier, or "failure bound," x*,,,: firms that realize xe, > x*,,, shut down.. 6 Similarly, there is a positive relationship between firm age and participation. Older firms have greater expected lifetimes because the increasingly precise estimate of their costs makes it less and less likely that they will fail as time goes on. Consequently, older firms are also larger on average. However, the positive relationship between age and participation is not an artifact of larger size alone. Conditional on size, older firms have longer expected lifetimes" and thus greater potential for realizing the benefits of participation. So both firm size and age are positively correlated with participation: among the youngest firms, only the largest choose to become formal; over time they are joined by smaller firms. To summarize the predictions of our framework: 1. There is heterogeneity in the degree offormality. The benefits and costs of participation undoubtedly vary across societal institutions, and vary for firms of different size and expected lifetime. While there are potential complementarities between different societal institutions, a large number of firms will choose to participate in only a subset of institutions at any point in time. For example, the legal system and bank financing are complements, but a firm may have to register legally before seeking external financing. Thus informality is not an all-or-nothing state and the degree varies by firm. This is not addressed by the other theoretical approaches -- including models of regulatory and tax compliance that typically consider only one dimension of participation -- but it accords with Tokman's observations (1992). 2a. Small firms are disproportionately informal. They benefit least from participation because of the small scope of their dealings with the public and hired employees (relative to the total volume of transactions undertaken by the firm). This has the corollary that: 2b. "Inefficient "firms are disproportionately informal. This implication is in line with many characterizations of the informal sector (Thomas, 1992; Portes, 1994). However, in contrast to other formulations, in this case the causality is not necessarily from informality to inefficiency. High 0 -- i.e. high cost -- firms choose less formality because it benefits them less than more efficient firms that produce at higher volumes for longer lengths of time. 2c. Youngfirms are disproportionately informal. This is partly because young firms are more likely to be small. Conditional on size, the population of young firms contains a disproportionate number that have not received enough signals to figure out whether paying the costs of formality are worthwhile; many eventually will go out of business. 3. Mode of operation (type of work site) and formality are jointly determined Small firms range in mode of operation from ambulatory hawkers to more settled establishments. One dimension of mode of operation, work site permanence, is not addressed by the other theoretical models. ' Their more precise cost estimates mean they are less likely to realize unexpectedly bad profits that would cause them to reach the failure bound. 7 However, a number of ad hoc characterizations -- most notably de Soto's (1989) -- draw a strong link with informality: informal firms operate out of temporary/makeshift buildings or stalls, or even door- to-door. Firm expansion involving greater capital outlays, K, requires greater permanent work sites and, simultaneously, greater formality to establish property rights or formalize contracts." As a second example, firms of different sizes (at different stages of growth) may have different degrees of interaction with the public. Because implicit contracts over product quality are cheaper and feasible to enforce with friends and family, the entrepreneur may find it most cost effective to primarily serve such customers when faced with small sales volumes. At larger volumes (later in the firm's life cycle), friends and family cannot necessarily buy all the firm's output, so sales to the general public and other firms should increase. 4a. Underlying patterns offirm dynamics should be comparable in both developing and industrialized countries. If the distribution of entrepreneurial ability and the learning process are similar across countries, then so should be the patterns of firm entry and exit. This also implies similar firm age distributions and overall firm dynamics (assuming comparable economic environments). 4b. Informal sector firms have relatively high mortality rates. The high turnover rate of informal firms that might appear as evidence of the inferiority of informal employment reflects the high mortality among small firms observed everywhere. The high turnover rate of such firms and jobs is not necessarily related to being informalper se. Although many informal firms will be small mature firms with high costs (but not so high that they eventually go out of business), many will be the "false starters" with imprecise estimates of their profitability that eventually fail. 5. Firms participate in an increasing number ofsocietal institutions as they grow. As firms with a low 0 grow to their equilibrium size, the depth of participation -- measured by the fraction of all institutions in which the firm participates or by the degree of participation with each individual institution -- increases as well. The implications for standard models of tax evasion and regulatory compliance are straightforward. Traditional approaches assume that enforcement agencies try to maximize social benefit (minimize social harm) subject to a binding budget constraint. Both these approaches and ours predict that large firms (the biggest violators on a per unit output basis) are more likely to participate. The difference between approaches lies in the determinants of compliance: traditional approaches assume that enforcement solely determines compliance; we model the (relative) net benefit to the firm. The actual importance of gross benefits versus gross (penalty) costs is an empirical matter, one that, unfortunately, we cannot test with our data. However, our approach shows 6 Our assumption that the government can perfectly exclude firms that do not voluntarily pay the full costs of participation undoubtedly is too restrictive. Hence larger businesses that have more permanent work sites are easier for the government to detect. So participation -- as measured by tax compliance and public registry -- will be greater for such firms. 8 that both costs and benefits to the firm should be accounted for when attempting to identify the importance of enforcement efforts. Moreover, our approach indicates that the duration of an economic activity should be considered when modeling participation. Traditional approaches to tax evasion and regulatory compliance typically ignore this issue, in part because they consider long-lived economic agents; in particular, firms are viewed as infinitely long-lived. However, we have shown that if firm dynamics play an important role in the economy -- as they appear to do -- then they should be factored into participation considerations. III. Empirical results 1992 National Micro Enterprises Survey (ENAMIN) from Mexico, offers the first comprehensive survey to date on compliance with or participation in several distinct markers of formality including registration with the tax authorities, tax payment, labor protection, participation in guilds or trade associations, and enumeration in the census, as well as other relevant characteristics. It thus, permits us to generate a reliable picture of the nature of informality, as well as to test the consistency of our framework with reality. The sample was generated by selecting approximately 11,000 individuals from the 1991:4 National Urban Employment Survey who declared that they were self-employed or heads of firms of five workers or fewer (fifteen or fewer in manufacturing). They were reinterviewed in the next quarter to generate "S more detailed accounting of income, capital stock, costs, employment patterns, and a variety of details related to participation in societal institutions. Ofthe sample of individuals reinterviewed in early 1992, a total of 9,036 were still operating businesses. Our empirical approach is to seek patterns of participation that accord with the predictions detailed in the previous section. However it should be emphasized (again) that there is a fundamental identification problem faced both by our methodological approach and by other approaches that assume enforcement efforts are the sole determinants of participation. We are aware of no data set with the requisite information on both costs and benefits of participation to evaluate the relative importance of each approach." Our limited goal in this section is to document empirically the heterogeneity and depth of participation; show the importance of firm size, age, and mode of production as correlates of participation; and (partially) establish a role for firm dynamics and life cycle considerations as key concerns for modeling participation (in both developing and industrialized countries). " Such a data set would have to identify exogenous variation in government policy that is independent of firms' decisions over formality. This identification is extremely difficult in practice because most policies are implemented nationwide, confounding the effect of policy changes with business cycle and macroeconomic forces that also influence firm behavior. A differences-in-differences approach that utilized between state (or province or region) variation in policies would work in principle. But the existence of multi-establishment firms that cross state lines would make assignment into the proper treatment groups problematic. 9 1. Heterogeneity ofParticipation Though the data set is bounded above at five workers (fifteen in mantifacturing), even within this narrow firm size range informality is clearly not an all or nothing proposition. The summary statistics in Table 1 show that there are high participation rates in societal institutions for even these small firms: 41.7 percent are registered with the federal treasury, 25.2 percent are registered with the local treasury (including Mexico City), 34.6 percent pay some taxes to one or both treasuries, 34.6 percent of firms with paid workers have them registered with IMSS (Mexico's social security administration), 22.5 percent are members of a business guild or association, 15.6 percent pay dues to a business organization, and 33.1 percent of firms that existed in 1989 were enumerated in the Census of that year. Table 2 presents cross tabulations along several dimensions of participation and shows that participation along one dimension need not imply participation along others. For example, the bottom left panel contains all the firms that have paid workers and are greater than three years old, which means that they should be registered with the federal treasury, should have their (paid) workers registered with IMSS, and should have been enumerated in the Census. However, of this group only 72.9 percent are registered with the federal treasury, 63.8 percent pay taxes, 31.8 percent participated in the Census but not IMSS, while 7.2 percent participate in IMSS but not the Census. Clearly, participation is a question of degree and spans many dimensions. This suggests that previous research that lumped together all small firms as representing the informal sector (e.g. Rauch, 1991) obscured important differences among them. In the interest of avoiding some of the conceptual confusion that surrounds the topic, it may therefore be preferable that future analysis employ the term "informal" to exclusively refer to the issues of participation discussed here. This would leave considerations of firm size, wages/productivity, labor market segmentation, etc. to be addressed under labels that correspond more precisely to the phenomena being studied. " More generally, our evidence indicates that models of regulatory and tax compliance may need to consider possible complementarities between different institutions when modeling participation and enforcement for individual institutions. 2. Distribution of Formality Across Firm Characteristics Points 2a-2c above argue that participation decreases with 0 and increases with the probability of long run success. Although we cannot observe either, the framework shows that they are monotonically related, respectively, to firm revenue/size and to firm age (conditional on size). " This is consistent with many characterizations of the informal sector (Thomas, 1992; Portes, 1994). Strictly speaking, in our framework firms choose between different institutional arrangements (Lin and Nugent, 1995). Informality encompasses a set of institutional arrangements including enforcement of contracts through social networks and self-insurance against employee health problems. Formality encompasses a different, complementary set of institutional arrangements including compliance with government reporting requirements. See also Peattie (1987) who critiques usage of the term "informal sector." 10 This leads to the following empirical specification: Pr(Participation) = Io Constant + P,Revenue + P2Age *- e (4) where revenue is total firm revenue, age is the number of years the firm has been in business (or the number of years the current proprietor has been operating it), and e is an idiosyncratic error term. We measure the probability of participation a number of different ways: (a) as an indicator variable for any participation in an individual institution, (b) as the degree of participation within a particular institution, and (c) as the degree of participation among a range of potential institutions. In each case the null hypothesis is that: (a) P,>0, (b) p2>0, that is participation should be an increasing function of both firm size and firm age. Table 3 shows the rate of registration with the federal treasury, the rate of registration of firms' paid workers with the social security administration (IMSS), and the rate of enumeration in the 1989 Census by firm size and by firm age. As predicted, there is a very strong positive relationship between participation and firm size/revenue. The relationship between participation and firm age is also positive, though much weaker. Table 4 reports the results from fitting probit regressions for the seven different types of institutions. In each case the coefficients on both firm size and age are positive and significant at better than the 1 percent level of confidence. (The standard errors were corrected for arbitrary forms of heteroskedasticity.) The estimated changes in the probability of participation for a unit change in each regressor are reported in Table 4.A. The relationship between size and participation is very strong: each point increase in log revenue corresponds to, for example, a 20 percent greater rate of tax compliance, a 23 percent greater rate of social security compliance, and an 11 percent greater rate of business guild registration.9 The relationship between age and participation is more marginally significant: a ten year difference in age increases participation in the various institutions by 2 to 4 percent.20 One potential concern is that the benefits of participation undoubtedly vary by industry. Given systematic differences in average firm size and age across industry, the positive relationships between size and participation and between age and participation in Table 4 may be spurious. To test this we tried alternate specifications (not reported), both including industry dummies and running separate regressions by industry. The results including industry dummies were virtually identical to those in Table 4. The industry-specific regressions, despite the markedly reduced degrees of freedom, also yielded comparable results. 3. Mode of operation. 11 Log revenue was used to avoid giving undue weight to the small number of firms with extremely high levels of revenue. 20 As a specification check, we tried substituting the two sets of dummy variables for the revenue and age classes (from Table 2) for the linear terms. The results were qualitatively the same both in these and the subsequent regressions. 11 Table 5 reports the results from fitting equation (4) to two other measures of formality that capture the nature of the production process jointly determined with the level of formality: the permanence of the firm's work site and whether individuals and families Are the firm's only main customers. Permanent work site is a dummy variable equal to one for those firms that operate out of a fixed site in a public marketplace, a factory, a variety/grocery store, or a retail service establishment.21 Changes in probabilities are reported in Table 5.A. As expected, firm size and age are positively (and significantly) related to work site permanence: older firms and those producing at larger volumes require more permanent work sites. The indicator for firms whose only main customers are individuals and families potentially is an inverse measure of formality. The ideal measure would include only close friends and family of the proprietor. The survey's measure is more broad but may still provide evidence in favor of our framework, so long as the measure is most accurate for smaller firms. Smaller firms should market more exclusively to close acquaintances because larger volumes of production require firms to seek customers among the general public. The measure may be negatively related to firm age for the same reason The second row of Table 5 shows that, as predicted, firm size is negatively related to whether the firm sells primarily to only individuals and families. However, the relationship with firm age is positive. To investigate the source of that positive relationship, the bottom row of the table reports the same regression including industry dummies. The results show that the positive relationship disappears when industry dummies are included, indicating a spurious effect in the previous regression. However, the strong negative relationship with firm size persists, providing evidence in favor of our framework. Separate regressions by industry (not reported) yielded similar results. 4. Firm dynamics To fully test the dynamics of our framework would require longitudinal data. Though the ENAMIN is the most comprehensive data source available to date, it lacks this dimension. Nonetheless, the cross sectional evidence it does offer is consistent with our framework. First, though the data is truncated at a firm size of five employees (fifteen in manufacturing), long lived firms exist across the revenue distribution, reflecting the underlying distribution of 0. Remarkably, average firm age is roughly the same across all deciles of the revenue distribution, ranging from a low of 7.9 years for the seventh decile to a high of 9.0 years for the second decile; the first and last 2 Only those firms operating out of temporary work sites that might serve as launching pads for more permanent work sites, and those firms operating out of permanent work sites that could have transitioned from a less permanent work site, were included in the regression. Excluded firms included those operating out of unspecified non-permanent or permanent work sites. In addition, those whose business is the transportation of people or merchandise, and hotels/taverns/inns/hostels were not included in the regression. Specifications that, in turn, (a) included these firms, and (b) limited the definition more narrowly, yielded comparable results. 12 deciles have respective means of 8.8 and 8.7 years.22 Moreover, the relatively uniform revenue-age distribution is not an artifact of the upper limit on number of employees: average firm revenue in the last revenue decile is more than eighty times larger than the first revenue decile. This suggests that this population of firms may be in a steady state, with entry and exit rates roughly uniform across the revenue distribution.23 Second, the observed patterns of firm entry and exit are consistent with those predicted by our framework and with those observed in the U.S. and in other developing countries. Numerous studies have documented high entry and failure rates among startups that decline with size and age of the firm.24 Evans and Leighton's (1989) study of self-employment dynamics in the U.S. provides the most comparable benchmark for our analysis. They find that inflows into self-employment over the previous year account for about 20 percent of self-employment for men over 35, with an even greater proportion for younger men. This is consistent with a constant rate of entry and older men running more established firms that are less likely to fail. Evans and Leighton also document a sharply decreasing exit rate from self-employment for the U.S., with the probability of failure ranging from 15 percent for the oldest of the self-employed to over 50 percent for the youngest of the self-employed. The overall patterns of firm age by age of the owner for Mexico in Table 6 are comparable. The last two sets of columns in Table 6 show the fraction and number of firms at each age range that are no more than one year and two years old, respectively. Consistent with Evans and Leighton's estimates, the number of entries is relatively flat throughout the life cycle. Yet the fraction of the self employed comprised of new entrants declines steadily, commensurate with a sharp increase in average firm age. Together, these patterns suggest that declining exit rates are probably partially responsible for the sharp increase in average firm age in these data.25 These broad similarities in self- 2 The difference in average firm age between the second and seventh deciles is statistically significant at a five percent level of confidence. The difference between the first and last deciles is not. 3 Note that "exit" could happen for two reasons. The upper size limit on number of employees means that firms would leave the sample frame either if they failed or if they added too many employees. Smaller firms undoubtedly are more likely to fail; whereas larger firms are more likely to grow their way out of the sample frame. 24 Mansfield (1962) shows that smaller firms have higher and more variable growth rates. Dunne, et al (1989) demonstrate that U.S. manufacturing plant failure rates decline steadily with the age of the plant. Davis, et al (1994) find that net job creation in small U.S. manufacturing firms is not high relative to large businesses, despite inordinately high rates of gross job creation, because of their disproportionately high rates of job destruction. Roberts and Tybout (forthcoming) find that in Mexico, Colombia and Morocco business births and failures are even more frequent and numerous in those countries than in the U.S., accounting for much larger shares of total employment adjustment. New plants are much smaller and less productive than the industry average and the failure rate is highest at young ages. 2 Evans and Leighton do not report average firm age by age of the owner, making direct comparison with the numbers in Table 6 difficult. However, if we assume that the exit probabilities in their data apply disproportionately to the very young firms within each age cohort, a likely phenomenon given the learning process, 13 employment dynamics between Mexico and the U.S. suggest that common determinants of self- employment may be as important as differing institutional factors in explaining the observed patterns of participation. 5. Depth ofparticipation Our approach predicts that firms participate in an increasing number of institutions as they grow. Again, because the ENAMIN lacks a longitudinal dimension, we cannot directly test the time series implications. However, the cross sectional implications are supported by the data. The depth of participation -- as measured by degree of compliance -- is analyzed in Table 7. Two measures are used: the fraction of the firm's paid workers registered with IMSS (for the subset of firms with any paid workers), and the fraction of all institutions in which the firm participates. Both variables are bounded below by zero and above by one, so the estimation used double-censored tobits.26 The second and third rows of the table report two different specifications for the fraction of all institutions in which the firm participates. The first encompasses all seven institutions in Table 4. The second excludes business guild registration and dues payment because not all firms may have access to such institutions; i.e. differences between firms in participation along this dimension may simply represent cross-industry differences in production technology or market structure. The patterns in Table 7 again are consistent with our prediction that the depth of participation is an increasing function of both firm size and age. Excluding business guild registration and dues payment in the third row makes these relationships stronger. The inclusion of industry dummies (not reported) leaves the results largely unchanged. IV. Relation to Previous Informal Sector Research The framework offered here departs from the premise that the small scale firms found in developing countries are fundamentally different from those in industrialized countries. Thus, it is solidly in the spirit of Hart (1972) and de Soto who stressed the intrinsic dynamism of the sector. In conceiving of formality as an input into the production function and that firms choose the optimal level along a continuum, we provide a theoretical underpinning both for Tokman's grey areas of partial compliance, and for de Soto's view that a lack of access to institutions is a binding constraint on firm growth. De Soto claimed that onerous compliance costs prevent firms from becoming formal. This concern was echoed by Porter (1995) who cited high and uncertain regulatory costs as barriers to firm growth in United States inner cities. Both of these perspectives are easily nested in our then average firm age must rise with age of the owner, as in Table 6. 26 Ordinary least squares regressions yielded comparable results. 14 framework as a case where the government sets po so high that for most firms the discounted net present value of participation never exceeds its costs. But our model also implies that it may never be possible to induce all firms to participate simply by streamlining conpliance procedures: for many very small firms, the benefits of participation may not exceed even modest costs.27 By relaxing our stylized view of the well functioning "market" for formality, we can encompass both Rauch's and Loayza's views. Clearly, reality is more complex than our extreme assumption that firms get only the participation they pay for and pay for none they do not want. Many institutions of civic society are public goods and the government imposes universal fiscal levies, making tax evasion attractive. Moreover, an enforcement agency seeking to maximize social benefit could easily choose to focus its efforts on longstanding, large firms, leaving the door open for small, low productivity firms to avoid taxation and regulation. Rauch, in fact, defines the informal sector as those firms of a size below which the government chooses not to enforce minimum wages. Incorporating this into our perspective, a growing firm may willingly choose to comply with such size-based regulations in order to get unimpeded access to necessary institutions.28 The logic behind Esfahani and Salehi-Isfahani's model (1989) is also consistent with the view presented here. Larger firms use more complex production technologies as they grow, making worker monitoring more difficult. They thus voluntarily pay efficiency wages that, in practice, may include health care or other benefits, to reduce shirking. Again, more efficient firms would become increasingly formal as they grow. But formality of the firm is an independent consideration from the wages it pays its workers: a small, longstanding firm that does not pay efficiency wages likely would participate in formal institutions such as the legal and banking systems.29 Finally, there is nothing in the static nature of previous models that makes them inconsistent with the approach described here. For instance, both Rauch and Loayza require an exogenous change in government policy to induce transitions between the formal and informal sectors. But this arises purely because they were not concerned with modeling firm dynamics. Within the context of our approach, such a policy change is equivalent to altering po and/or pt, which leads to comparable comparative static results as those described in both Rauch and Loayza. Conclusion Using a unique data set from Mexico, we have provided a more detailed characterization of the nature of informal production than previously possible. Beyond the empirical regularities, we 2 See Ozorio de Almeida, et al. (1994) for a discussion of deregulating the informal sector. 2 In this example, the cost of complying with the minimum wage is part of po and p,. 29 Both Schaffner (1996) and Velenchik (1996) have documented a positive firm-size wage effect for Peru and Zimbabwe. Such an effect is a critical component of Esfahani and Salehi-Isfahani's efficiency wage model. However, both of the former authors argue that such evidence is not supportive of that class of model. 15 also have offered a new approach to analyzing the informal sector. This approach assumes that informal firms in developing countries behave similarly to those in the industrialized countries, and is based on a model of firm dynamics frequently used in the industrial o*rganization literature. It offers an alternative motivation for informality which, unlike much of the literature on tax evasion and regulatory compliance, asserts that participation in societal institutions may be essential to growth, and therefore at least partially voluntary. It also can nest many existing models that base existence of the informal sector solely on institutional distortions, market failures, or excessive government regulation. Though the data do not permit a definitive test of competing models, they are consistent with the predictions of our approach. This suggests that our framework is an important benchmark to be considered when analyzing the informal sector and regulatory compliance. References Ashenfelter, Orley and Robert S. Smith (1979), "Compliance with the Minimum Wage Law," Journal ofPolitical Economy, 87:2, 333-350. Banerji, Arup and Sanjay Jain (1996), "Quality Dualism and the Informal Sector," mimeo. Besley, Timothy (1995), "Savings, Credit and Insurance," in J. Behrman and T.N. Srinivasan, eds., Handbook ofDevelopment Economics, Volume III, Amsterdam: Elsevier. V, Cowell, Frank A. (1990), Cheating the Government, Cambridge, Massachusetts: MIT Press. Davis, Steven J., John Haltiwanger, and Scott Schuh (1994), "Small Business and Job Creation: Dissecting the Myth and Reassessing the Facts," in Lewis C. Solmon and Alec R. Levenson, eds., Labor Markets, Employment Policy, and Job Creation, Boulder, Colorado: Westview Press. Dunne, Timothy, Mark J. Roberts, and Larry Samuelson (1989), "The Growth and Failure of U.S. Manufacturing Plants," Quarterly Journal ofEconomics, 104:4, 671-698. The Economist (1996), "Not-so-wondrous pensions," December 14, p. 46. Esfahani, Hadi S. and Djavad Salehi-Isfahani (1989), "Effort Observability and Worker Productivity: Towards an Explanation of Economic Dualism," The Economic Journal, 99, 818-836. Evans, David S. and Linda S. Leighton (1989), "Some Empirical Aspects of Entrepreneurship," American Economic Review, 79:3, 519-535. Fenn, P. and C. G. Veljanovski (1988), "A Positive Economic Theory of Regulatory Enforcement," The Economic Journal, 98, 1055-1070. Fishback, Price V. and Shawn Everett Kantor (1996), "The Adoption of Workers' Compensation in 16 the United States, 1900-1930," National Bureau of Economic Research Working Paper 5840. Gruber, Jonathan (1994), "The Incidence of Mandated Maternity Benefits," American Economic Review, 84:3, 622-641. Harris, John R. and Michael P. Todaro (1970), "Migration, Unemployment, and Development: A Two-Sector Analysis," American Economic Review, 60:1, 126-142. Hart, Kieth (1972), Employment, Income and Inequality: A Strategy for Increasing Productive Employment in Kenya, Geneva: ILO. Jovanovic, Boyan (1982), "Selection and Evolution of Industry," Econometrica, 649-670. Lewis, W.A. (1954), "Economic Development with Unlimited Supplies of Labour," Manchester School ofEconomics and Social Studies, 20, 139-191. Lin, Justin Yifu and Jeffrey B. Nugent, "Institutions and Economic Development," in J. Behrman and T.N. Srinivasan, eds., Handbook ofDevelopment Economics, Volume III, Amsterdam: Elsevier. Lippman, S.A. and R.P. Rumelt (1992), "Uncertain Imitability: An Analysis of Interfirm Differences in Efficiency under Competition," Bell Journal ofEconomics, 13:2, 418-438. Loayza, Norman V. (1995), "The Economics of the Informal Sector: A Simple Model and Some Empirical Evidence from Latin America," mimeo. Lucas, Robert E., Jr. (1978), "On the Size Distribution of Business Firms," Bell Journal of Economics, 9:2, 508-23. Ozorio de Almeida, A.L. and S.E.M Graham and L.F. Alves (1994). "Poverty, Deregulation and Informal Employment in Mexico," ESP Working Paper, World Bank. Palay, Thomas M. (1984), "Comparative Institutional Economics: The Governance of Rail Freight Contracting," Journal ofLegal Studies, 13, 265-287. Palay, Thomas M. (1985), "Avoiding Regulatory Constraints: Contracting Safeguards and the Role of Informal Agreements," Journal ofLaw, Economics, and Organization, 1, 155-175. Peattie, Lisa (1987), "An Idea in Good Currency and How It Grew: The Informal Sector," World Development, 15:7 851-860. Porter, Michael E. (1995), "The Competitive Advantage of the Inner City," Harvard Business Review, 73:3, 55-71. 17 Portes, Alejandro (1994), "The Informal Economy and Its Paradoxes," in Neil J. Smelser and Richard Swedberg, eds., The Handbook of Economic Sociology, Princeton, N.J.: Princeton University Press. Rauch, James E. (1991), "Modeling the Informal Sector Formally," Journal of Development Economics, 35:1, 33-47. Roberts, Mark J. and James R. Tybout (forthcoming), Industrial Evolution in Developing Countries: Micro Patterns of Turnover, Productivity and Market Structure, New York: Oxford University Press. Schaffner, Julie Anderson (1996), "Premiums to Employment in Larger Establishments: Evidence from Peru," mimeo, forthcoming Journal of Development Economics. de Soto, Hernando (1989), The Other Path: The Invisible Revolution in the Third World, New York: Harper and Row. Thomas, J.J. (1992), Informal Economic Activity, Ann Arbor, University of Michigan Press. Tokman, Victor E. (1992), "The Informal Sector in Latin America: From Underground to Legality," in Victor E. Tokman, ed., Beyond Regulation: The Informal Economy in Latin America, Boulder, Colorado: Lynne Rienner Publishers. Velenchik, Ann D. (1996), "Government Intervention, Efficiency Wages, and the Employer Size Wage Effect in Zimbabwe," mimeo, forthcoming Journal of Development Economics. 18 Revenue Firm type # I Firm type # 2 Firm type # 3 Time in Business Figure I 19 Revenue Firm type # I With participation Without participation Firm type # 2 With participation Without participation Firm type# 3 T2 Time in Business Figure 2 20 Table 1: Summary statistics Variable Mean S.D. Min Max N Total Revenue 3668 8145 2- 200000 8873 Log(total revenue) 7.42 1.19 .693 12.2 8873 Net Income 1260 3138 -78498 126300 8807 Years in Business 8.58 9.37 .167 51 9033 Number of paid workers .384 1.08 0.00 15 9036 Any paid workers .193 .395 0.00 1 9036 Number of unpaid workers .232 .601 0.00 5 9036 Any unpaid workers .164 .370 0.00 1 9036 Registered with federal treasury .417 .493 0.00 1 9036 Registered with local treasury .252 .434 0.00 1 9036 Amount of taxes paid 78.5 552 0.00 30000 8401 Paid any taxes .346 .476 0.00 1 9036 Any paid workers registered with IMSS .346 .476 0.00 1 1748 Number of paid workers registered with IMSS .318 .451 0.00 1 1748 Business registered with guild or association .225 .418 0.00 1 9036 Amount of business organization dues paid 7.04 43.8 0.00 1500 9019 Any business organization dues paid .156 .363 0.00 1 9036 Enumerated in the 1989 Census .331 .471 0.00 1 5220 Fraction of all institutions .279 .301 0.00 1 9036 Fraction of all institutions (excluding guild, dues) .325 .371 0.00 1 9036 Types of work site: Transportation: people or merchandise .064 .245 0.00 1 9036 Makeshift stand in a public road .041 .199 0.00 1 9036 Fixed stand in a public road .017 .130 0.00 1 9036 Makeshift stand inside a marketplace .023 .149 0.00 1 9036 Fixed stand forming part of a marketplace .019 .138 0.00 1 9036 Door-to-door or street vendor .045 .207 0.00 1 9036 Services offered via vehicle or cart .034 .181 0.00 1 9036 Own home: without specialized equipment .123 .328 0.00 1 9036 Own home: with specialized equipment .051 .220 0.00 1 9036 Other non-permanent work site .008 .088 0.00 1 9036 Fixed work site in a public marketplace .032 .176 0.00 1 9036 Hotel, tavern, inn or hostel .0006 .024 0.00 1 9036 Factory: production or repair services .095 .293 0.00 1 9036 Variety or grocery store .075 .263 0.00 1 9036 Retail service establishment .131 .338 0.00 1 9036 Other permanent work site .008 .091 0.00 1 9036 Home of customer or client .233 .423 0.00 1 9036 Permanent work site .485 .500 0.00 1 6194 Family/friends only main customers .384 .486 0.00 1 9036 21 Table 2: The Varying Degrees of Formality Incidence of Registration with Federal Treasury and Payment of Taxes by Various Firm Characteristics Firms registered Firms registered All firms with local treasury with business guild or organization # firms % firms % regis % % % % pay in this in this with fed pay # % fed pay # % % fed % pay guild/ group group treasury taxes firms firms treas taxes firms firms treas taxes org dues 1. All firms 9036 100.0 41.7 34.6 2273 100.0 78.8 74.0 2035 100.0 67.7 54.0 61.4 2. Paid workers>0 1748 100.0 73.1 61.3 690 100.0 91.6 79.6 573 100.0 88.0 72.1 57.4 2.A. and IMSS=0 1144 65.4 61.5 50.9 396 57.4 87.1 74.7 282 49.2 97.9 82.5 58.8 2.B. and IMSS>0 604 34.6 95.0 81.1 294 42.6 97.6 86.1 291 50.8 77.7 61.3 56.0 3. In business>3 years 5220 100.0 42.8 37.1 1365 100.0 79.1 76.3 1284 100.0 68.8 56.9 62.6 3.A. not in 1989 Census 3493 66.9 20.1 18.5 482 35.3 55.6 63.3 637 49.6 46.3 37.4 64.2 3.B. and in 1989 Census 1727 33.1 88.8 74.8 883 64.7 92.0 83.4 647 50.4 91.0 76.0 61.1 4. Pd wkrs>0, >3 years 1096 100.0 72.9 63.8 440 100.0 92.7 82.7 378 100.0 89.2 75.1 58.7 4.A. IMSS no, Census no 348 31.8 31.0 28.4 68 15.5 77.9 75.0 62 16.4 61.3 48.4 62.9 4.B. IMSS yes, Census no 79 7.20 86.1 74.7 26 5.91 100 80.8 34 8.99 97.1 73.5 50.0 4.C. IMSS no, Census yes 337 30.7 88.4 77.2 168 38.2 92.9 81.0 110 29.1 88.2 76.4 54.5 4.D. IMSS yes, Census yes 332 30.3 97.9 84.6 178 40.5 97.2 87.6 172 45.5 98.3 84.3 61.6 Only those firms in business for more than three years could have been enumerated in the 1989 Census of businesses. IMSS refers to registration of the firm's paid workers with the Mexican Social Security administration. Only paid workers have to be registered with IMSS. 22 Table 3: Summary Statistics The Relationship between Formality and Firm Size and Age Registered with Any workers registered -Enumerated in federal treasury with IMSS 1989 Census Revenue decile Mean Ohs Mean Obs Mean Obs 1st decile .080 889 0 6 .087 472 2nd decile .125 910 0 11 .109 512 3rd decile .184 917 0 19 .136 523 4th decile .257 1197 .024 84 .196 730 5th decile .401 558 .075 67 .299 341 6th decile .444 1024 .102 167 .317 590 7th decile .581 816 .219 206 .444 441 8th decile .600 788 .279 219 .442 443 9th decile .710 909 .331 381 .547 537 10th decile .855 865 .606 538 .751 535 Registered with Any workers registered Enumerated in federal treasury with IMSS 1989 Census Years in business Mean Obs Mean Obs Mean Obs less than one year .323 876 .237 131 0 0 1 year .374 838 .221 131 0 0 2 years .416 1094 .343 181 0 0 3 years .455 857 .345 171 0 0 4 years .463 559 .391 110 .309 537 5 years .432 562 .336 125 .309 538 6 to 7 years .411 672 .397 156 .316 648 8 to 9 years .401 504 .290 100 .287 492 10 to 12 years .441 975 .415 193 .312 952 13 to 15 years .416 551 .418 122 .353 541 16 to 19 years .467 304 .333 69 .391 297 20 to 29 years .438 765 .321 159 .359 744 30 years or more .435 476 .414 99 .376 468 23 Table 4. Participation in societal institutions Dependent variable Log revenue Years in business Pseudo R' # obs Registered with federal treasury .661 .006 0.200 8870 (37.7) (3.95) Registered with local treasury .371 .006 0.083 8870 (26.9) (3.57) Pays any taxes .556 .008 0.158 8870 (35.6) (5.35) Any paid workers registered with .634 .010 0.161 1697 IMSS (17.2) (2.95) Enumerated in 1989 census .538 .013 0.154 5121 (26.5) (6.26) Registered with business .383 .006 0.090 8870 guild/organization (26.8) (3.87) Pays any dues .312 .007 0.066 8870 (21.6) (4.02) Probit regressions. Absolute values of z statistics in parentheses. The standard errors have been corrected for arbitrary forms of heteroskedasticity. Table 4.A. Participation in societal institutions (difference in probabilities from a one unit change in the regressor) Dependent variable Log revenue Years in business Registered with federal treasury .254 .002 Registered with local treasury .113 .002 Pays any taxes .199 .003 Any paid workers registered with .226 .004 IMSS Enumerated in 1989 census .188 .004 Registered with business .108 .002 guild/organization Pays any dues .070 .002 24 Table 5. Other measures of formality Dependent variable Log revenue Years in business Pseudo R' # obs Permanence of work site .538 .009 0.162 6066 (31.3) (4.97) Families/individuals only main -.277 .007 0.048 8870 clients (22.9) (4.63) Families/individuals only main -.208 -.001 0.176 8868 clients (adding industry (15.1) (0.63) dummies) Probit regressions. Absolute values of z statistics in parentheses. The standard errors have been corrected for arbitrary forms of heteroskedasticity. Table 5.A. Other measures of formality (difference in probabilities from a one unit change in the regressor) Dependent variable Log revenue Years in business Permanence of work site .214 .004 Families/individuals only main -.105 .003 clients Families/individuals only main -.078 -.0004 clients (adding industry dummies) 25 Table 6: Distribution of Firm Age by Age of the Owner Years in Years in Business: Years in Business: Business One year or less Two years or less Age of Owner Mean Mean Frequency Mean Frequency 19 or younger 2.61 .524 99 .693 131 20 to 24 2.86 .366 191 .573 299 25 to 29 3.96 .277 253 .462 421 30 to 34 4.96 .233 268 .380 437 35 to 39 6.41 .193 250 .325 420 40 to 44 7.90 .168 196 .283 329 45 to 49 9.53 .138 138 .251 251 50 to 54 11.5 .128 117 .191 174 55 to 59 12.8 .116 75 .199 129 60 to 69 15.8 .107 91 .180 153 70 or older 17.8 .086 33 .153 59 Table 7. Depth of participation in societal institutions Dependent variable Log revenue Years in business Pseudo R' # obs Fraction of paid workers 2.84 .051 0.115 1697 registered with IMSS (9.15) (2.86) Fraction of all institutions in .245 .004 0.211 8870 which the firm participates (52.2) (7.17) Fraction of all institutions, .386 .005 0.145 8870 excluding business guild (44.6) (5.53) registration and dues Tobit regressions. Absolute values of t statistics in parentheses. 26 Logit Analysis in a Rotating Panel Context and an Application to Self-Employment Decisions Patricio Aroca GonzAlez Universidad Cat6lica Del Norte, Chile William F. Maloney LCSPR I. Introduction A substantial literature exists on limited dependent models inl panel context (see or Maddala 1983 or Baltagi, 1996 for overviews) and continuous dependent variables in an incomplete or rotating panel context (Bjorn and Jansen 1983, Nijman, Hsiao 1986, Verbeek and van Soest 1991). This paper derives a methodology for estimating logit models in a rotating panel. It then uses the technique to examine an unresolved problem in development economics: the role of self-employed workers unprotected by labor legislation in the LDC labor force. In particular, we are interested in the determinants of the decision to leave protected (formal) work to enter self-employment. An alternative theoretical model to the dualistic view generally accepted is offered. Both views are tested using rotating panel data set from Mexico and the alternate view supported. II. An Alternate View of Informal Self-Employment Much of the literature on the informal self-employed sector in LDCs beginning with Harris and Todaro(1970) has seen self-employed workers unprotected by labor legislation as those rationed out of protected or "formal" salaried jobs sector jobs by above market clearing remuneration in the protected sector. Transitions should be largely unidirectional, from the informal and presumably very low capitalized micro-enterprises, to the formal sector except in the event of downturns in which case laid off workers will be thrown back on the informal safety net. However, there is little reason to suppose that the expanding literature on self- employment in the industrialized world that views self-employment as a desirable and more flexible alternative to wage work may not also be relevant in LDCs. In particular, the debate over the dynamics underlying patterns of worker transitions into self- employment is likely to be relevant. Johnson (1978), Jovanovic (1979) and Miller(1984) argue that younger individuals are better able to bear the risk involved and hence should be heavily represented among entrants into self-employment. However, as Evans and Jovanovic note, this is inconsistent with Evans and Leighton's (1989) finding of the hazard into self-employment being constant in age which they attribute to liquidity constraints that dictate that workers require time to build up the capital needed to start a business. We argue that this phenomenon may be exacerbated in the developing world where credit markets are poorly developed. The problem can be seen as a Stopped Markovian Decision Process (SMDP)' where workers, faced with uncertainty about future streams of income as salaried and self-employed workers must decide the optimal savings and switching strategies. Their behavior can be seen as similar to that of workers who, perhaps with the idea of opening a business upon their return, migrate to a country that offers the possibility of accumulating wealth more quickly, and return home only ' See Eckstein and Wolpin(1989) for a review of the specification and estimation of dynamic stochastic discrete choice models. when they reach their target level of savings (See Piore 1979). This problem has been analyzed in detail by Berninghaus and Seifert-Vogt (1993) and we adapt their work to our problem as a way of generating predictions to be tested in the empirichl work. We assume the worker will open his own business at time r and plans on operate it for T-T years where T is the end of his planning horizon. He has subjective expectations 71 on the return to his invested accumulated real wealth x, in the business. Upon starting his business, the worker will choose a sequence of consumption bundles, c, such as to maximize T Zu (c.) t=r s.t. O<c,<x, x,= 71(xt- cJ) t =,... ,T, where u(.) represent the continuous per period utility function. We abstract from the discount factor since it will be the same in both sectors and we assume it unchanged. This reduces to a standard dynamic programming problem where V, (x) is the value function at switch time r, the maximal value of the sum of per period utility from being an entrepreneur from -r to T. While in the salaried sector from t = 1, ..., (T-1) the worker earns y-where {Y,}, is a stochastic process whose probability law is known to the worker. In each period, the worker chooses a consumption bundle c, subject to the condition c, <; (x, + y,). Any surplus can be saved at a real interest rate, i,, which is the realized value of a stochastic process {I,}, of real interest rates. In each period, the worker must decide whether to work in the salaried sector for another period, or start his business and receive V,(xt). The optimal policy for this problem is-a sequence of consumption strategies (c(.) and the stopping (switching) time T such that the total expected reward E E(xo.,o. yO) u(c,(.)) +V,(x,) 1.0 is maximized given the initial state(xo, i0,y0). The optimal stopping time is associated with each "state history" (xi, i,yt) and because these are realizations of a stochastic process, T too, is a random variable. From the framework, several predictions emerge. First, there exists a critical level of target savings below which the worker will prefer to continue to stay salaried. For the case of a two period model with logarithmic utility, Berninghaus and Seifert-Vogt show that the target level of savings falls with a rise in the subjective return to self- employment, n, rises with an increase in the opportunity cost of savings, i, and rises if a higher wage in the salaried sector raises the required comparable stream of income resulting from self-employment, and hence the start up capital required. The first two also have a predictable effect on the switch time, t. In the last case, however, the overall impact of current income on r is ambiguous since higher incomes both increase the level 2 of target savings as well as increase the possible rate of savings accumulation. Given two workers with identical savings, the one with higher income may find himself below the target rate of income and stay one more period to earn another period twage.2 In sum, the probability of a move into self-employment at a particular moment: Pr(move) = P( 7c,i,y) a>,ap<O ap=, 8ir ai ay It is worth comparing these predictions to those from the standard dualistic view where an above market clearing formal sector remuneration, y, rations workers into the informal sector where the returns fall to absorb those in the queue. A fall in relative returns of self-employment for salaried work occurs in the context of economic downturns where the informal return must fall to absorb displaced workers. Similarly, to the degree that increased interest rates are associated with recession and the loss of salaried jobs, again, we may expect more movement of the displaced into the informal sector: movement into self-employment would be counter-cyclical. In both cases, the predicted signs would be the opposite of those postulated by the model above.' The next section offers a method for using logit methods in a rotating panel context to estimate the determinants of the worker's decision to move, and hence to test between these two views. III. Logit Analysis in a Rotating Panel Context Selection of Individuals t In the relatively common case that we address, individuals are selected according to a "rotating" scheme in the following manner. In period 1 of a total of T periods, the first sample is selected of N individuals who will remain in the sample for z periods : y, , Y21.**-*** YNI In the second period, the first m= N/z individuals are retired and the first place until the Nth place are occupied by the individuals who follow individual m: y12, Y22 ,.... YN2 The process of retiring and replacing continues for each period t with a new sample: y1, y2t *I . Nt The combination of data obtained by this process is called a Rotating Panel and we 2 By the same token, a worker who suddenly loses his job, y goes to 0, will suddenly see the target level of saving decline and is more likely to move. In this way, the common vision of the informal sector as the reserve army of the unemployed can be seen in somewhat different light. ' See Maloney (1997) for a discussion of the relative merits of formal vs informal work and the procyclicality of the latter in Mexico. 3 can considered it ordered as: Y111Y21s*Y.mb Y(m+)1)s **-**N Y121Y22 ..**Ym2 Y(m+1)25 ..* * YN2 Y 13, ..........,Ym3-. -J In this manner, H = (T-1)m + N individuals are partially observed across T periods. In our example, y represents whether the worker moves in that period (y=1) or stays in salaried employment (y=O). It is useful to reframe the problem as a T X H fixed panel: Y y 21 ............ y HI Y12 ..Y... Y IT Y2T.********YHT where Yj, denotes the position of individual j in time t, whether there is an entry or not. For example: j 12 22.- * j7 m2 are positions that do not have entries. Observation: Each position in the fixed panel corresponds to one in the rotating sample: Yj3t: Yj7-(t-])m t ( 1 ) However, an individual is included in the rotating sample, and has an entry in the fixed panel in time t only if: ] : j - (t - I)m < N In the analysis, we will only be concerned with individuals in the sample for a full z periods which can be shown to be the case for individuals entering the panel in t ( {1,2,...,T-z+1 }. Several results pertaining to this group are described in appendix I. Definition: For t E { 1,2,...,T-z+1 }, if individual j enters the sample of size N in period t, we define Y as a vector representing the sequence of the z consecutive entries. =i (Yit Yj(t+1), Yi(t+z-] )) In the present application, this is the sequence of moves that individual j is observed to make across the z periods in the sample. In theory, there could be multiple moves or none. 4 Probability Function We are interested in understanding what determines the timin* of the individual's decision to change state. To be consistent with the theory above, we assume the individual moves only once into self-employment, and that the decision to move in each period is independent of the previous decisions. The vector e'k =(0,...1, 0.....0) e R L-- position k permits us to identify the period in which the individual changes state. We define the probability that an individual j that has been k periods in the sample, changes its state in period t, and that it changes only once as: p,(k) = p y ,I= ek EY(t=-) k=1,2,3,...,z k=1 To calculate pjt(k), we first find an expression for: p (Y, = e, ) k = 1,2,...z For example: p(YI = el) = p =)p , = 0)... p(7,+- = 0) We assume that the probability that an individual changes state follows a logit distribution, and this probability is a function of a set of the environment and individual's characteristics (X). This can be shown equal to: exp[,X,_,_,,]1 1 (1 + exp[f6X,-(,-,)., ]) (1 + exp[,8X,-,.i]) ( + ) ex p [feX X-(I-2) +] e- )X M~(1- ,8) where M() is the product of the denominators above. Generalizing, we obtain that the probability of changing states exactly once in the kth period of observation is: ( eexp [fXJ-(+k-2)m,+k-1] The probability of moving in the kth period of observation, given that the individual will 5 move exactly once in one of the k periods is: PJ, (k) =, = ep, =e)+ p(Y, = e2 )+.....+ = e.) exp[)6XY-(t+k-2)m,t+k-1 exp[ ,,-om.,+exp[ ,m;]....+exp[, -(1+z-2)m,t+z-1 And finally, the conditional probability that individual j who enters in the sample of size N in period t and remains z periods counting from t and who changes state in exactly one of the z periods is: pi, = P, (1) "' P,,(2)"'..,(z) "- As Y,, takes a value of 0 or 1, and in our example there is uniquely one non-zero entry, this expression effectively selects which Pjt determines pjt. Likelihood Function Before writing the likelihood function we establish some definitions to simplify the presentation. Definition: Let M(K*J) be the set of the matrix of rank (K*J). If 15 u5 K, we can define: 7,: M(K * J) + R' i,,(A) = 7U(ak)I1kSK = ul u2,***, uJ which states that 7r, projects the row u of the matrix A. This ensures the condition that each individual moves only once or: Definition: Let B = {j such that ;,Y,, + ;r2y,,+.*+*r*Yjt 1 }, that is the set of j where the individual moves only once. Definition: Let B = {Bt such that 1 5 t T - z + 1 }, that is, the set of Bt such that the individual is in the sample for exactly five periods. Definition: Let X, be the z*p matrix of independent explanatory variables for each individual (See appendix for more detail). 6 With these definitions, we can rewrite P,, (k) as: P, (k) = 1 V l k z Eexp[(,1-r)X,] 1= 1 and finally: pi, =1 P,(k)"Yk Yt k=1 Therefore the log of the likelihood function for all individuals in the panel through all periods will be: A = 1 Jnp, = -1xxZ k ,,nexpqrl -)T Jyj k teB pEB, teB pEB, k=1 (1=1 Employing the Newton-Raphson algorithm, the first and second derivative of the likelihood function with respect to the parameters are calculated as: VA663) = - : ZZ ky,l Pfl(' zexpq1T, - Ik]f ?'}7,-rT X," IeB jEB, k=1 \I=1 = z zz ~~jPpexp({ 7r + Ir 2,kWXj,,6)* J;r, - ;r, W,I [ f 1 i] 2'(8 lkyl jt2,Xp ,+ -2;kr I rg I rv tc:B jeB, k=1 1=1 r>I which permits us to estimate f, and V IV. Data: The National Urban Employment Survey (NUES) conducts extensive quarterly household interviews in the major metropolitan areas and is available from 1987 to 1993. It is structured as a rotating panel where in each quarter, a fifth of the sample is dropped and replaced by individuals who will be interviewed for each of the next five quarters. In 24 overlapping panels spanning 1987-1993, individual workers can be followed as the move among sectors of work. Individuals are matched by position in an identified household, sex, level of education, and age to ensure against generating spurious transitions. The analysis restricts itself to men aged 16-65 with a high school education or less. It also focuses on formal salaried workers and the "informal" self-employed, including owners of firms under 16 employees who do not have social security or medical benefits and are therefore not protected.' Only those who begin in formal salaried 4 It is often the case that the informal sector is defined as firms with five or less workers. As we are ocusing on informality defined as being unprotected by social security or other legislation, we loosen the 7 employment and move only once over five quarters into self-employment are retained, yielding a sample of 1087 workers. In the estimations, we employ predicted earnings in each sector as a measure of the "own" and "alternate" earnings, giventhe standard human capital variables, experience, experience squared, education, education squared. The return to accumulated capital (the opportunity cost of using savings to open a business) is the real 30-60 day deposit rate as calculated from the International Financial Statistics of the IMF deflated by growth of the consumer price index. We also test state dependence through introducing the lag of the independent variables in the regressions. V. Results: Table 1 presents the results of the estimation of the model set out in section II. Table 1: Results from Rotating Panel Logit Regression COEF. S.E COEF S.E. Wage (Salaried) -.347 1.34 -.357 1.28 Wage (-1) .132 1.41 Earnings (Self) 4.02 .249 4.03 .235 Earnings (-) -.115 .354 Interest Rate -3.39 e-3 2.45 e-5 -2.71 e-3 1.92 e-5 Interest Rate(-1) -2.68 e-3 1.93 e-5 -3.83 e-3 1.65 e-5 Nobs=1078, Sample includes 24 complete panels of 5 quarters each spanning 1987-1993 The results are supportive of the model. The first and second columns present the complete specification and show that for only the interest rate are lagged values significant. This suggests the absence of state dependence. The second specification presents only the significant coefficients. Here, self-employed earnings appear very strongly and of the correct sign reflecting that as opportunities improve in the informal sector, workers are more likely to open their own businesses. The current wage in the formal sector still enters ambiguously, again, as predicted, and is not significant. This is to be expected given that a rise both increases the attractiveness of formal sector employment, and raises the savings rate making a move into self-employment possible. Finally, the interest rate is strongly significant and of the predicted sign suggesting that a rise in the opportunity cost of the capital used for start up discourages opening up a business. In all cases, the sign is the opposite of that predicted by conventional dualistic views of informal self-employment. Appendix II derives the cross section marginal effects and Table 2 calculates them for the regression above. In each panel of the table, k represents the period in which the individual moved and h the period corresponding to the variables observed. Of greatest importance, the diagonals of the tables are both relatively stable and of the sign found in table 1. Calculating the marginal effects has not reversed the effect as is sometimes size limit to the next category tabulated. In practice, the vast majority of firms are under 3 workers. 8 found and the theoretical framework remains supported. The off-diagonal elements (symmetric) are less intuitive. In every case the impact of the variable one period forward or backward has the reverse impact of the contemporaneous effect. Table 3 derives the marginal effects over time, which are calculated by taking the difference between the maximum and the minimum value of each variable. As with the cross sectional marginal effects, the signs are those predicted and expected self-employed 9 Table 2: Cross Section Marginal Effects Self-Employed Earnings dPjt(k)/dX'h k \ h 1 2 3 4 1 0.6978200 -0.1989130 -0.2283948 -0.2705122 2 0.6957832 -0.2274624 -0.2694078 3 0.7651952 -0.3093380 4 0.8492580 Formal Sector Wage dPjt(k)/dX'h k \ h 1 2 3 4 1 -0.0617104 0.0175905 0.0201977 0.0239222 2 -0.0615302 0.0201152 0.0238246 3 -0.0676686 0.0273557 4 -0.0751025 Interest Rate dPjt(k)/dX'h k \ h 1 2 3 4 1 -0.0004693 0.0001338 0.0001536 '0.0001819 2 -0.0004680 0.0001530 0.0001812 3 -0.0005147 0.0002081 4 -0.0005712 Interest Rate Lagged dPjt(k)/dXh k \ h 1 2 3 4 1 -0.0006628 0.0001889 0.0002169 0.0002569 2 -0.0006609 0.0002160 0.0002559 3 -0.0007268 0.0002938 4 -0.0008066 10 earnings is the most important variable to explaining the transitions change from formal to informal sector. Table 3: Marginal Effects Over Time Variable Variation Probability Variation Formal Sector Wage 0,3654 -0,0315 Self-Employed Earnings 0,2411 0,2913 Interest Rate 57,83 -0,0377 Interest Rate Lagged 64,03 -0,0429 VI. Conclusion The paper has derived a methodology for analyzing logit models in a rotating panel context. Using data from Mexico, it then applied the technique to test between two theories of why salaried workers enter the informal self-employed sector. The evidence supports a view that self-employment is a desirable destination, but one that in the presence of credit constraints requires accumulated capital before the business is opened, over the more traditional view of self-employment as a safety net for those losing preferred formal sector jobs. 11 References: Baltagi, B. H. (1985), "Pooling Cross Sections with Unequal Time Series Lengths" Economics Letters, 18p. 133-136. Baltagi, B.H. (1996) Econometric Analysis of Panel Data, John Wiley and Sons (New York). Baltagi, B. H. and B. Raj (1992), "A Survey of Recent Theoretical Developments in the Econometrics of Panel Data," in Bjorn, Erik,(1981) Estimating Economic Relation from Incomplete Cross Sectional/Time Series Data, Journal of Econometrics 16:221-236. Bjorn, E. and E.S. Jansen (1983) "Individual Effects in a System of Demand Functions, Scandinavian Journal ofEconomics 85:4 1983. Berninghaus, B. and H G. Seifert-Vogt (1993), The Role of the Target Saving Motive in Guest Worker Migration, Journal ofEconomic Dynamics and Control 17, p 181-205. Eckstein, Z. and K I. Wolpin (1989) "The Specification and Estimation of Dynamic Stochastic Discrete Choice Models, a Survey, The Journal ofHuman Resources, 24:4 Evans, D.S. and B. Jovanovic (1989), "An Estimated Model of Entrepreneurial Choice under Liquidity Constraints, Journal ofPolitical Economy 97:4 pp. 808-826 Evans, D.S. and L. Leighton(1989), "Some Empirical Aspects of Entrepreneurship," American Economic Review 79. Harris, J.R. and M. P. Todaro (1970), "Migration, Unemployment, and Development: A Two Sector Analysis," American Economic Review, 60:1, 126-142. Hsiao, C.(1 986), Analysis of Panel Data, Econometric Society Monograph no. 11, (Cambridge University Press, Cambridge). Jovanovic, B. (1979), "Job Matching and the Theory of Turnover." Journal of Political Economy 87:5 972-90. Nijman, R., M. Verbeek, and A. van Soest (1991), "The Efficiency of Rotating-Panel Designs in an Analysis of Variance Model," Journal ofEconometrics , 49:373-399. Maddala, G.S. (1983) Limited Dependent and Qualitative Variables in Econometrics (Cambridge University Press, Cambridge). 12 Maloney, W.F. (1997) "Labor Market Structure in LDCs: Time Series Evidence on Competing Views," Working Paper University of Illinois. Miller, R.A.(1984) "Job Matching and Occupational Choice" Journal of Political Economy 92 1086-1120. Piore, M. (1979), Birds ofPassage (Cambridge University Press, New York, NY). 13 Appendix I Theorem: Individual j enters the rotating sample of size N in period t only if: (z - 1)m + 15 j - (t - 1)m N (2) Demonstration: The last m entries of the sample of size N=zm can be written as Y(z-I)m+1, t' Y(z-I)m+2, t * * *, YN, t I For individual j to be one of these entries, condition (1) implies condition (2). If (2) holds we can also show that 7,, has an entry and that at time t, the individual has just entered the sample of size N. Observation: Only for t E {1,2,...,T-z+1 } are the last m individuals entering the sample of size N observed for all z periods. There are (T - z + 1)(N - (z - 1)m)= (T - z + I)N _( z+1)m z such individuals of the H total. Corollary: For t E {1,2,...,T-z+1 } condition (2) is necessary and sufficient for individual j to be observed for z periods. Appendix II The sequence of decisions to stay or move for each individual is the vector Y 1-(t-1)m,t? j-Im,t+ls"'9Y}-(I+z-2)m,t+z-1 and the set of variables that determine that choice: Xi X 2 ... ... XP J-(t-1)m,1 J-(t-I)m,t J-(/-I)m,t XJX' xj-tm,t+1 X1 ... ... XP j-(t+z-2)m,i+z-1 j-(t+z-2)m,i+z-1 14 We distinguish two types of marginal effect, across the individuals and across time. Cross Section Marginal Effect The cross section marginal effect measures the change in the probability of a move due to individual differences in the independent variables. For each individual we can write: OpJ, op., (k) Sp1, (k) , p, (k) -E j-(t-)m,t j-tm.+l j-(t+z-2)m,t+z-1 orh= 1,2,. .z a1 op_, (k) sj, (k) 1s, (k) 8XJ-(t+h-2)m,i+h-1 Xaj-(t+h-2)m,i+h-1 s2, (k) axj-(t+h-2)m,t+h-1 e- Kk X,J ' ',x,, s2, (k) axj(+h2)m,t+h 1 p16[;rkxBe- 6[ r,X,, ] z a/'e 1 s2(k) r* axj-(I+h-2)m,t+h-1 r=1 8X-(t+h-2)m,t+h-1 6[,, X,, -k,kX" +efl[[-x,]' * e 6[kXe' ' " jfl kXJ, s2, (k) [ [X _(,+h2)m,t+h-_ , r= axj-(t+h-2)m,t+h-1 we know that: -rX,, 8 ,X)(I+r-2)mt+r-I + 2XJ-(t+r-2)m,+r-I+16PXj-(t+r-2)m,t+r-I 8x'-(ib-2)m.h ax _,t _,th 15 {1,6, Vr=h 0 Vr# h {I V=h where: 3, =l r h therefore apj,(k) 1 [-p@kXp Z fl[,r,X it, ee-P[KkX,I1 [grXI '9X)-(i+h-2)m,t+h-I k - = -p2,(k)Z ,(b,h -Sk k r=1 Marginal Effect Over Time (MEOT) The marginal effect over time measures the change in the probability of a move with diffferent levels in an independent variable. Given that we work in discrete time, for each individual we can write: Marginal Effect Over Time = (p,, (k +1)- p,, (k)IX'(k+1) # Xk), V k = 1,2,--, z - 1 We have that: e fl"k+1Xj es pj, (k + 1)= z ,and p,(k) e Kh 6t[X h=1 h=1 Therefore: P , , )6rXJ-(t+k-2)m,t+k-1 ej,x,-(t+k-I)mt+k e I,X -(t+k-2)m,t+k-1 re +.=) MEOT= z e 16[ ' h=1 16 17 Quitting and Labor Turnover: Microeconomic Evidence and Macroeconomic Consequences Tom Krebs Brown University William F. Maloney LCSPR 1. Introduction Recent work on efficiency wage models has produced an internally consistent macro-theory of involuntary unemployment (underemployment)' whose basic assumptions and implications have largely been corroborated by empirical micro-studies.2 Although the microeconometric work was motivated by macroeconomic theory, it has never been fully integrated into a macroeconomic framework.' In this paper we offer one version of such an integrated approach. More precisely, we first develop an efficiency wage model with labor turnover (Phelps 1968, Stiglitz 1974, Salop 1979, Hoon and Phelps 1992) and show that the worker's decision problem gives rise to a quit-rate function. We then use microeconomic data to estimate this quit-rate function and to test the specification suggested by economic theory. Finally, microeconomic evidence and macroeconomic model are combined to evaluate the quantitative effects of changes in economic policy and other macroeconomic shocks on the wage rate, the turnover rate, and employment in the long-run (steady state analysis). The efficiency wage model with labor turnover we employ is in principle applicable to any type of movement of labor across sectors in any country (Bulow and Summers 1986). However, the original literature on this type of efficiency wage model has mainly focused attention on unemployment in developed countries (Phelps 1968, Salop 1979, Phelps and Hoon 1992). We, on the other hand, test and calibrate our model using panel data on the movement of Mexican workers between the formal salaried and the informal self-employed sector. Our choice of the data set was motivated by the following two considerations. First, the efficiency wage model with labor turnover captures well a number of features of LDC labor markets, and in particular the Mexican case we consider. The literature suggests that the self-employed informal sector comprises both workers rationed out of formal salaried jobs as well as a relatively prosperous "upper tier" that may prefer self-employment.4 In other words, the literature is consistent with one of the central ideas of the efficiency wage model with labor turnover, namely that at each point in time workers are voluntarily leaving their formal-sector job for self-employment and simultaneously self-employed workers are unsuccessfully trying to reenter 'See Katz (1986) and Woodford (1994) for surveys. 2See Katz (1986) and Layard, Nickell, and Jackman (1991) for surveys. 'The quantitative papers by Danthine and Donaldson (1990,1995) and Kimball (1994) on efficiency wage models with shirking (Shapiro and Stiglitz 1984) consider microeconomic evidence when calibrating the macroeconomic model, but do not incorporate a microeconomic estimation equation into the macroeconomic model as we do. In this sense, we feel that we have come closer to a full integration of the two fields of labor economics and macroeconomics. See also Blanchard and Katz (1997) for a statement in favor of such an integrated approach. 4 Harris and Todaro (1970) offer the canonical statement of the dualistic (rationing) view and Fields (1990) discusses the "two-tier" view of the informal sector. 1 the formal sector. Moreover, neither minimum wages nor unions are credible explanations for the observed segmentation.' Finally, Constitutional proscriptions against firing suggest quitting as the dominant mode of job separation. Second, our data set has an important time dimension which allows us to estimate the quitting response of individual workers to changes in macroeconomic conditions. Given our final goal of quantitative macroeconomic analysis, this feature of the data set seems essential. In addition, the data on self-employed workers offer a measure of the benefits (payoffs) to not being employed in the formal sector that displays substantial variations over time. These variations in self- employment benefits are important since they provide us with an additional test of the "quitting theory" which predicts that labor turnover is positively correlated with expected benefits to self- employment and that the quit-rate function is symmetric: the benefits-elasticity of quitting is equal to the negative of the wage-elasticity of quitting. Moreover, if the symmetry property of the quit-rate function is supported by the data on self-employed workers, we may use it as a working hypothesis (until refuting evidence is forthcoming) and apply it to unemployment. This opens the door for an assessment of the quantitative macroeconomic effects of changes in unemployment benefits without directly estimating the elasticity of quitting with respect to unemployment benefits, usually an impossible task given the lack of temporal variations in these benefits. Our empirical estimates of the determinants of labor flows from the salaried to the self- employed sector strongly support the specification suggested by the quitting theory: the individual probability ofjob separation is decreasing in the formal-sector wage (the expected payoff to staying) and increasing in benefits to self-employment and the probability of finding aformal-sectorjob (the expected payoff to leaving). Moreover, the above mentioned symmetry property of the quit-rate function cannot be rejected. When the microeconomic estimates are used to calibrate the macroeconomic model, we find the long-run effects of macroeconomic shocks on wages, labor turnover, and (formal-sector) employment to be substantial. The strong employment response found here stands in stark contrast to the disappointingly small unemployment effects reported by Danthine and Donaldson (1990,1995) and Kimball (1994) who calibrate an efficiency wage model with shirking (Shapiro and Stiglitz 1984) to US unemployment data.' This paper can be interpreted as providing a two-stage "test" of the real world relevance of efficiency wage models with labor turnover. In the first stage, microeconomic data are used to estimate and test what we believe to lie at the heart of this type of efficiency wage model, namely the quit-rate function. If the coefficients are found to be significant and of the correct sign, in a 5Bell(1996) finds no evidence that minimum wages are binding. Maloney and Ribeiro (1998) find evidence of union influence on employment, but none on wage setting. 6Danthine and Donaldson (1990, 1995) explicitly consider aggregate uncertainty by solving a stochastic dynamic general equilibrium model. Kimball (1994) studies the dynamic and steady state effects of macroeconomic shocks in a deterministic model, but his quantitative result on employment variations is obtained by comparing steady state equilibria. 2 second stage the estimated quit-rate function is incorporated into the macroeconomic model and the calibrated model economy is used to assess the quantitative importance of efficiency wages. This second-stage check is important since there seems to be little value in having a macroeconomic theory of unemployment (underemployment) which is supported by microeconomic data but implies an almost constant unemployment (underemployment) rate. In this paper we present one fully worked out example of this two-stage procedure in the hope that it will spur interest in further applications to different countries and different sectors. Such work is likely to add an important dimension to the existing empirical literature which has either completely focused on the micro-level or solely relied on cross-country regressions.' The paper is organized as follows. Section 2 develops the model. Almost all derivations, and in particular the discussion of the worker's decision problem, is relegated to the Appendix. Section 3 presents the empirical analysis of the panel data on Mexican workers and some additional information on the Mexican labor market. The specification for the estimated quit-rate function is dictated by the theory developed in Section 2. In Section 4 the macroeconomic model is calibrated and the simulation results are presented. Section 5 concludes. 2. The Model The model is a discrete-time, neoclassical growth model with a labor market characterized by labor turnover, employment-adjustment costs (hiring and training costs), and wage-setting by firms. The analysis will be confined to equilibria in which a number of economic variables grow at a constant rate equal to the exogenous rate of technological progress (balanced growth path). a) Workers There is a large number of ex-ante identical, infinitely-lived workers. Workers' preferences over random consumption sequences allow for a time-additive expected utility representation. Workers do not participate in financial markets and therefore do not save or dissave. Hence, each worker's consumption level is equal to his current disposable income.' 7See, for example, Phelps (1994), Nickell (1997), and Phelps and Zoega (1998) for empirical work using cross-country regressions. Blanchard and Jimeno (1995) conduct an interesting case study comparing two countries, Spain and Portugal. The method outlined in this paper provides a formal procedure for quantifying the importance of efficiency wages in explaining the different macroeconomic experiences of two countries. 81n a sense, this assumption renders the model classical rather than neoclassical. It is mainly made for tractability reasons since it trivially determines the wealth distribution of workers (no wealth). Without this assumption the wealth distribution is in general non-trivial and has to be computed as part of the equilibrium, except when there is complete consumption insurance. The assumption of restricted capital market participation is also made in Danthine and Donaldson (1990,1995) for the same tractability reason. Kimball (1994) does not treat capital accumulation and therefore does not deal with wealth 3 In each period a worker devotes a fixed amount of time to one of the following two activities: working in the formal sector or working in the informal sector of the economy. Our informal-sector data in the empirical section are taken from the self employed and we will therefore call a person working in the informal sector a self-employed worker. In each period, a worker, regardless of his current employment status, receives an idiosyncratic shock determining the relative attractiveness of employment versus self-employment ("taste-shock", change in expected payoff to self- employment). After observing the shock realization, an employed worker makes a quit/stay decision and a self-employed worker makes a search/no-search decision. Whereas an employed worker automatically becomes self-employed when deciding to quit a job, a self-employed worker who decides to search for formal-sector employment receives ajob offer only with probability less than one. The Appendix Al discusses the Bellman equation associated with the worker's decision problem and analyzes the resulting optimal decision rule. The optimal decision rule gives rise to a quit-rate function, q = (w , w,p; T, b), where q stands for the quit rate experienced by firm i , w for the (growth-adjusted) wage paid by firm i, w the average (growth-adjusted) wage, p for the probability of finding (formal sector) employment when not employed, r for the tax rate on formal- sector labor income, and b for the average (growth-adjusted) pecuniary benefits from self employment. Let q(w,p;T,b) - q(w,w,p ; T,b) be the economy-wide quit-rate function when all firms pay the same wage. Clearly, this function is identical to the quit-rate function in an economy with only one representative firm. The function 4(.), however, is the function entering into the profit maximization problem of an individual firm (see Appendix A2) in a many-firm economy. In Appendix Al we show that the individual quit-rate function, 4(.), satisfies <0 a > > ->0 a >0; (1) aw, aw ap 'ab T and that the economy-wide quit-rate function, q(.), satisfies 8q 8q a.<0 ; q > 0 aw ap (2) aq aq 1 aq aT bab waw The empirical analysis conducted in Section 3 tests the sign and symmetry restrictions (2) and finds effects. Phelps (1994) emphasizes wealth effects, but nowhere develops a complete general equilibrium model with endogenize wealth distribution. We hope to dispense with this assumption in future work. 4 strong evidence in favor of them. Our panel data on worker transition provide no information about the quit-rate function of an individual firm in a many-firm economy, 4(.). Appendix Al, however, shows that the two marginal quit-rate functions (approximately) satisfy - (w,w,p;,B) V (w,p; -Lq (w,p;,B) (3) 1;- 1 ,(1 -q(w,p;t,b)) (1 -p) p(w,p;,p,b),-), 1 - Ow (1 -q(wp;r,b)) where O3 is the discount factor of workers. Expression (3) enables us to make quantitative predictions about the impact of economic policy knowing only the economy-wide quit-rate function q(.). The parameter g measures the difference between the reduction in the quit-rate when an individual firm raises its own wage and the reduction in the quit-rate when all firms raise their wages simultaneously. b) Capitalists (Firms) There are i = 1,...,N infinitely-lived capitalist with identical preferences who each own one firm with identical production technology. There is one good which can be used for consumption and investment purposes. Firm i combines capital and labor to produce output. Adjusting the amount of labor employed is costly since there are hiring and training costs. We assume that adjustment costs are fully paid by firms.' Taking the economy-wide wage as given, each firm i chooses a sequence of consumption (of owner i ), capital, investment, employment, hiring, and own-wage which maximize the capitalist' life-time utility subject to the relevant constraints. The decision problem faced by firm (capitalist) i is fully spelled out in Appendix A2. The resulting Euler equations for the growth-adjusted variables are 91n our model with credit rationed workers owning no wealth, only the firm can afford to pay these costs. Even if workers are not credit rationed but human capital created by training is firm specific, the firm is likely to bear the full cost of training (Salop 1979, Hoon and Phelps 1992, Phelps 1994). Of course, we do not consider indirect means by which workers can be made to bear some of the adjustment costs. Wages rising with tenure is one example. 5 it = PC (1+g) Xt 1-8+-aF (k Il,t+d aF , = (1+h + X al ,,(k 1,1.) -w, - T(h ,I) (4) Xi = cI P ; y = k T'(h,) ; ki = - (w ,w9,P) where PC and p are the capitalist' discount factor and coefficient of relative risk aversion, cd her growth-adjusted consumption level, ki the capital stock per efficiency unit of labor, hit the hiring rate, lit the fraction of the labor force employed in the formal sector, and ka' yt, Lagrange multipliers associated with physical, respectively human, capital accumulation constraints. The multiplier y' is the (utility) value of a trained employee to the firm (the analog to Tobin's q). Further, 8is the depreciation rate of physical capital, F(.) a standard neoclassical production function, and T(.) the adjustment cost function. c) The Government The government collects taxes and spends the tax receipts on consumption goods. We assume that tax receipts are equal to outlays in each period and that therefore the fiscal budget is always balanced. For simplicity, we assume that informal-sector workers and capitalists are not taxed so that the tax revenue, which is equal to fiscal spending, is (rwl)N. d) Steady State Equilibrium We are interested in symmetric steady state (balanced growth) equilibria. Such an equilibrium is defined as a list of (growth-adjusted, per-capitalist) values for output, y *, capital k *, capitalists' consumption c *, (formal-sector) employment I *, real wage, w *, and hiring (quitting) rate, h * = q ', such that: i)Given the quit-rate function, 4(.), and expected labor market conditions, (w*,p*), y ', k , , h ' , c * , w ' are the solution to the capitalists' (firms') optimization problem if the initial capital stock and stock of trained employees is (k *, 1). ii)The quit-rate function, 4(.), is the solution to the worker's optimization problem. iii)Expectations are fulfilled, that is, expected values are equal to actual values. 6 Because of the assumption of constant returns to scale, the marginal products only depend 8F 8F on the capital to labor ratio: (k,1) = f' () and a(k,1) =f(F) -f'(C) with ak 81 = k / I and f(k) = F(k, 1). Using the definition of a steady state equilibrium and Equation (4), we immediately derive the following characterization of a steady state equilibrium. The growth- adjusted capital stock per employed worker is *_+g)P+(5 £*=f_ (1+) +8-1 (5) PC For given E',the equilibrium wage rate, w *, and the equilibrium probability of finding a job, p, are the solution to f'(k) -*f(E) - w - T((w *,p )) - 1 - (1+g)P T'(q(w*,p*)) = 0 T'(q(w*,p')) + I = 0 (6) aw Finally, the equilibrium values of the remaining variables are determined by 'o h = q q(w*,p) ; l + = + h ' = 7*) p v(1-q*)) 7 c*= F(k*,1*) - 8k' - wl - T(h*)l' where v (1 - q ) is the fraction of self-employed workers searching for a formal-sector job. The parameter v is a constant which is discussed in more detail in Appendix Al. Evidently, the two equations (6) determining the equilibrium values w * and p are equivalent to Z - w - T(q(w*,p)) - r T'(q(w*,p*)) = 0 T'(q(w*,p*)) + 1= 0 (8) * aq tL(w*',p ) 5;(w',p) 'oThe consumption of a worker if employed is w * (1 -T) and if self-employed is b. The government consumes (tw I*)N. 7 where Z = f(k ') + f' (k )k is the marginal (revenue) product of labor and r = 1 - C(1 + g) the real interest rate. It is often useful to depict the solution to (8) as the intersection of a downward-sloping "labor demand curve" and an upward-sloping efficiency (incentive) wage curve. To the extend that the first equation in (8) expresses the optimal employment choice by firms and the second equation represents the optimal wage setting by firms, this terminology seems justified. The first equation in (6) implicitly defines a function w = w d(p) with dw d (T + rT") w ap (9) dp 1 + (T' +rT") q 8w Since -q < 0 , q > 0, T' > 0, and T" ; 0, the curve w d(.) is downward sloping if and only 8w ap if I (T' + r T") < 1. For small r T" this condition is always satisfied since | | T' = < I (this follows from the second equation in (8)). Note also that if the real interest rate is non-negative, p > 1 is a necessary (and if r T" = 0 a sufficient) condition for the w d schedule to be downward sloping. Thus, in order to have a labor demand function with a negative slope, the reduction in the quit-rate due to the increase of the individual firm's wage must be larger than the reduction in the quit-rate when all firms simultaneously increase their wage, which is generally true in the model considered here. The second equation in (8) implicitly defines a function with T" .1q 1 aq 2 q +f .q dw_ ap A 8w apaw p aw (10) dp T" q 1 aq) -2 2q - 1 0q aw p aw 8w2 aw 8w There is no straightforward way of signing the expression (9)." Our empirical results suggest that the quadratic and cross-derivative terms are zero. If in addition T" = 0, we have dw' _ _ a8p / dp ap aw which turns out to be positive for the range of parameter values considered in this paper. Hence, the efficiency wage curve is upward-sloping. "The sufficiency conditions for the firm's optimization problems do not guarantee an upward- sloping curve. 8 e) An Extension: Vacancies and Matching The efficiency wage model developed so far shares one important feature with traditional search models,12 namely an endogenously determined labor turnover rate. However, it also differs from those models in important ways. First, in the efficiency wage model firms set wages in contrast to the Nash bargaining solution usually deployed by the search literature. In a sense, the efficiency wage model is the limit case of the bargaining model in which firms have all the bargaining power and are therefore in a position to make a take-it-or-leave-it offer to workers. Second, the search literature deploys a matching (hiring) function relating the number of hires, hl, to the number of vacancies, v, and the number of unemployed 1 - 1. This matching function describes the efficiency of the job reallocation process. In this section, we briefly discuss the possibility of incorporating a matching function into the efficiency wage model developed here. For the sake of concreteness, consider the case in which the matching function is of the Cobb-Douglas type, h I = va (1 -1) 1', where we assumed for simplicity that all unemployed (self- employed) workers are searching for a formal-sectorjob." Suppose further that the only adjustment cost is the cost of posting a vacancy and that the unit cost of a vacancy is a constant, c. Hence, the total cost of changing the employment level is T = c v. Eliminating the number of vacancies, v, yields a total adjustment cost T = c (hl)pa (1 -1)(1-1/a) and a cost of adjustment per employee of T = ch 1r (v/-I)Thus, we have an adjustment cost function which is convex in the hiring rate, h, but also depends on the employment level, 1. This dependence of the "training and hiring costs" on the employment level is the only difference to the previous model formulation. It is, however, straightforward to incorporate the more general adjustment cost function into the efficiency wage model resulting into three steady state equations in the three unknowns w *,p *,1*. We plan to conduct quantitative policy analysis for such an extended model with a general matching technology in the future. The efficiency wage model developed in the last sections can be thought of as the special case in which a = 1 since then we have T = c h, that is, an adjustment cost function which only depends on the hiring rate. Incidentally, this function is linear, a property we assume in the quantitative section 4. 3. Microeconomic Data Analysis This section uses micro-economic data from Mexico to estimate the quit-rate function and to test the model's predictions about the determinants of quitting behavior. The quit-rate function 12See Blanchard and Katz (1997) and Rogerson (1997) for recent surveys. 13Recall that in our notation h is the hiring rate and hlis total hires of one firm. For simplicity, we have set the number of firms to one: N = 1. 9 estimated here is used in Section 4 for macroeconomic policy analysis. a) Data Description The National Urban Employment Survey (ENEU) conducts extensive quarterly household interviews in the 16 major metropolitan areas and is available from 1987 to 1993. The sample is selected to be geographically and socio-economically representative. The statistical agency (INEGI) expanded it significantly over the period by adding municipalities, however, we include only those present in every year of the survey to prevent changes in composition. The questionnaire is extensive in its coverage of participation in the labor market, wages, hours worked, etc. that are traditionally found in such employment surveys. INEGI's treatment of sample design, collection, and data cleaning is careful. Surveys and documentation of methodology are available on request. The ENEU is structured so as to track a fifth of each sample across a five quarter period. To construct the panels, workers were matched by position in an identified household, level of education, age and sex to ensure against generating spurious transitions. Using just the first variables to concatenate and following changes in sex across the panel led to mismatching (or misreporting) of under .5 percent. Taken together, we have 24 complete panels of 5 periods spanning a total of 28 quarters where transitions could occur across a seven year period which includes a time of recession (1987-88), recovery (1989-91), and then stagnation (1992-1993). Though the model deals with decisions to be self-employed generally, we further narrow the population by only considering self-employed workers in the "informal" sector. We use the term 'informal' here to refer to those unprotected by labor law, more specifically, owners of firms under 16 employees who do not have social security or medical benefits. In fact, under 1% of these firm have more than 5 employees so the definition corresponds closely to that commonly used in the development literature. Formal salaried workers are defined as those in firms of over 16 workers who enjoy labor protections. To eliminate the "self-employed" in consulting firms or other high end activities, we analyze male workers with a high school education or less between the ages of 16-65. The dependent variable is a 0, 1 index that captures whether the worker moved during a particular quarter. If he should move back to salaried employment and then move again to self- employment, this is counted as a second quit. The macro wage and benefit variables employed in the regression analysis, logw, and log b,, are the median for the entire sample (spanning five panels) for each quarter. The probability of being hired, p,, is the number of the self-employed who transition to the salaried sector, as a fraction of those looking for a salaried job. While we know the total number of self-employed, we do not observe the share searching in each period which theory predicts should vary with macro-shocks. We assume that this fraction is proportional to, or at least highly correlated with, the standard measure of search, the unemployment rate. The probability p, is therefore proxied by 15,, the number of the self-employed moving into salaried work divided by the number of unemployed. Even though we thereby avoid the need for time series data on the search intensity, we still require information about the sample average of this variable in order to rescale the estimated coefficient appropriately when conducting the macroeconomic simulations in 10 Section 4. To this end, we use the survey response from the National Micro-Enterprise Survey (ENAMIN) which in 1992 re-interviewed roughly 11,000 of those in the 1991:4 ENEU who declared themselves self-employed. In particular, it asks the motivation for opening the business and offers eight non-exclusive responses. 13.5 percent responded that they could not find work as a salaried worker and another 3.2 percent responded they were laid off at their previous job. We use the rounded up number of 20% in our baseline model as an estimate for the average fraction of self- employed workers searching for a formal-sector job (see Section 4). The last two columns of table 1 present the summary statistics for the variables used in the regression analysis. On average 2.5% of salaried workers transit into self-employment in one quarter. The standard deviations also suggest substantial variation in the macro variables across the period. Table 1 here b) Econometric Specification and Results We are interested in implementing an empirical procedure allowing us to estimate the quit- rate function q = q (w,p, b). In the Appendix we show that if workers' utility is logarithmic and if the optimal decision rule is approximated by a first-order Taylor expansion, then we can write q = D(ao + a logw + aP p + ab log b + O') , (11) where 0(.) is the distribution function of an unobserved, worker-specific shock variable 0'. The random variables { 0' } are identically distributed and capture all differences in taste and ability across workers influencing their quitting decision. The relevant derivatives of the quit-rate functions are then given by =a q(x), = ab(x), and = a p(x), where p(x)is the probability density evaluated at x = ao + a w + app + ab b. Hence, the inequalities (2) expressing the implications of the theory now read a <0, a > 0, and ab > 0and the symmetry relation says aw = -ab There are two difficulties with implementing (11) econometrically.4 First, the discussion in the Appendix only deals with the case of deterministic changes in the macroeconomic variables w , b , and p whereas the data are better described by a stochastic process 1 There is an extensive literature on structural estimation of Markov decision processes (See Eckstein and Wolpin(1989) for a review and Daula and Moffitt (1995) in the labor literature). In contrast to most of this literature, we do not recover the structural (preference) parameters from the estimated coefficients aq/ax since our macroeconomic policy experiments can be conducted without it. Estimates of preference parameters, however, would be essential for a quantitative welfare analysis. 11 {(w,,P,b1)} }. If, however, the process of macro variables, { (w,,p,,b,) }7 , is Markovian, a straightforward extension of the analysis shows that the optimal decision rule of workers still gives rise to a quit-rate function q= (a0o + a log w, + a pt + ab log b, ) + , (12) since the current macro state (w,P, b) is a sufficient statistic for the future evolution of the relevant macro variables (we again assumed a first-order Taylor approximation). With some weak additional assumptions, it can also be shown that the restriction imposed upon the signs of the coefficients a , a b still hold. From a quantitative point of view, the quit-rate function derived in the Appendix corresponds to the quit-rate function when macroeconomic variables follow a random walk (highly persistent macroeconomic shocks). Given our short sample period (28 observations over seven years), we do not attempt to test the random walk assumption. Second, if we only use the macro variables w, ,P,, b, as right-hand-side variables, we certainly loose a large amount of information about economic variables influencing the decision of an individual worker since any observed difference in behavior is automatically attributed to the unobserved idiosyncratic shock, 0'. On the other hand, if we use the observed wage of individual workers on the right-hand-side (plus additional human capital variables), as is done by previous cross-sectional studies estimating quitting equations,"s we mix together (transitbry) micro- and (permanent) macro-shocks since both types of shocks cause the individual wage to change. In this paper, however, we are mainly interested in the quitting response of individual workers to permanent macro shocks. These difficulties with using the currently observed individual wage are also the reason why we do not attempt to estimate the quit-rate function of an individual firm, q, = 4(, w,p, b). Our approach to this problem is to add to the right-hand-side of (12) additional individual-specific human capital variables:"6 qJ = ) (a + a logw + a pt + a log + k Jk + . (13) The particular specification (13) with 0' satisfying standard error-term assumptions could, for example, arise as the linearized solution of the worker's dynamic programming problem under the following conditions. Suppose we decompose the individual wage and the individual self- employment benefits as follows logw/ = logw, + logYW' (14) logb' = logb, + log Tbt 15See, for example, Pencavel (1972), Krueger and Summers (1988), and Campbell (1993). 16Though in the empirical analysis we do not growth adjust either w or b, in the log-formulation such an adjustment would be entirely captured in the intercept term, ao. 12 where the macroeconomic variables, log w, log b,, follow a Markov process, or more specifically a random walk. Moreover, assume that the evolution of the idiosyncratic wage and benefit component is given by log Hkwk w 1ogi.' = Ek -twkHij + ewt log'bt = k 7bkH + Lbt (15) where the y wk bk are constants, the human capital variables H follow a Markov pacess, for each t = 0, 1,... the random variables { e'} and {'bt)j are identically and independently distributed, and for each jeJ the processes }V o- and { ebt)-o are serially uncorrelated. In principle, these assumptions are not too restrictive as long as the observed variables, Hk1, exhaust the list of relevant idiosyncratic variables. In practice, however, there are always unobserved variables relevant for the worker's quitting decision whose existence causes a violation of our error-term assumption. We employ a random effects probit routine with estimates of Huber-While robust standard errors (see below) to ameliorate this problem. The technology for estimating mover-stayer models in a rotating panel context is not presently developed." As Maddala(1987) notes, simply pooling the data and then using standard probit techniques would yield consistent but inefficient estimates because the c6relation across observations is not being exploited. As an intermediate approach, we maintain the integrity of each of the 24 panels, but "pool" them to form one large four-period panel (the first quarter is used to establish the worker in the formal sector) of 100,978 observations. Each observation includes the human capital variables and move-stay index particular to the individual, as well as the two macro- earnings variables and the proxy for the probability of finding a job corresponding to the potential move's location in the span of 28 periods. We then estimate the transition equation using STATA's panel probit routine." This permits estimating Huber-While robust standard errors as a measure to ameliorate violations of the assumptions on e' and e' made above. W1 b The results presented in the first two columns of table 2 strongly support the specification suggested by our theory. The two earnings elasticities enter of predicted sign and significantly at the .1% level. A rise in self-employed earnings relative to formal-sector earnings leads to more 17A substantial literature exists on limited dependent models in a panel context (see Maddala 1983 and Baltagi 1996 for overviews) and continuous dependent variables in an incomplete or rotating panel context (see Nijman, Verbeek and van Soest 1991 for a recent overview), but there is no literature analyzing limited dependent variables in a rotating panel context. 8 As Maddala(1987) notes, random effects probit estimators are consistent while the fixed effect estimators are not, in additional to being computationally difficult. We find it unlikely that there is any correlation between the random individual effects and the macro explanatory variables that would require use of a fixed effects estimator. 13 workers trying their hand at self-employment. Our proxy,f,, for the probability of finding a formal sector job also enters as expected and very significantly offering direct support to the efficiency wage dynamic postulated here. The higher the probability of finding another job in the formal sector, the more likely a worker will risk starting his own business. A more complete model with squares and cross terms of the macro variables was also estimated but none of these additions were significant and the results are not reported. Also consistent with the model, a ) test cannot reject the symmetry of the two income related elasticities at the 5% level. The next two columns present the results when this constraint is imposed. As expected, they are very similar and are used in the simulations. The human capital variables enter significantly in both regressions. Since in the present specification they explain the idiosyncratic component of the wage-benefit differential, the signs of the combined effects (evaluated at the mean) are plausible. Increased schooling may plausibly lead to a larger differential, and hence a lower propensity to move, because of a greater demand for skills in the formal salaried sector. Conditioning on schooling, more experience may raise the probability of success in the self-employed sector, lower the differential and hence raise the probability of moving. The regressions were also run with an alternative dependent variable that dropped observations after the first quit and yielded similar results. We also compared the results to the theoretically consistent standard probit techniques on the pooled sample and found them very similar (available on request). Table 2 here Our estimation results confirm the most basic implications of the "quitting theory". Even though this is not the place for a detailed analysis of the full range of alternative views of job- separation, we point out that the estimation results are inconsistent with the simplest "firing theory", that is, the theory that termination of existing worker-firm matches is mainly a decision made by firms, To see this, suppose that the quit-rate is a constant independent of any economic factors. Suppose further that the firm's decision problem is identical to the one discussed in our efficiency wage model with the only exception that the wage is determined in a competitive labor market, the labor supply function being derived from a standard labor/leisure choice of workers. Consider now the response of the economy, which is initially in steady state, to a negative productivity (demand) shock. The dynamic response is likely to be an increase in firing and a reduction in hiring by firms since the new optimal employment level is lower. In addition, reduced demand for labor can be expected to decrease the wage. Hence, firing (the left-hand-side variable) is negatively correlated with the wage and the probability of finding a job (hiring) and this theory therefore predicts a < 0 and a < 0. If formal-sector and informal-sector productivity shocks are uncorrelated or positively correfated, which is a reasonable assumption, we have in addition the prediction ab 0. Thus, two of the three inequalities are rejected by the data. We should also mention that if we consider an exogenous change in the wage (change in minimum wage) for constant productivity, similar reasoning yields aw > 0 , aP < 0 , ab = 0, which is even more strongly rejected by the,data. 14 4. Macroeconomic Policy Analysis Section 2 developed a macroeconomic model with labor turnover. The previous section estimated an aggregate quit-rate function q(w,p;r,b) using microeconomic data. In this section, we combine the theoretical analysis with the empirical estimates in order to assess the quantitative effects of different macroeconomic shocks on the wage rate, the turnover rate, and formal-sector employment. This is done by implicitly differentiating Equation (8) with respect to the parameters under consideration and thereby calculating the effect on the wage rate and the probability of finding ajob. The effect on the labor turnover rate and formal-sector employment is then calculated using dh = dq - dw + -dp + dr aw ap a'r (16) dl = 1 + dp -h d h p + v(1-q') p 'v(I-q *) p ) for the case of a tax reduction program. Analogous expressions are used for changes in other parameters. a) Calibration In order to make quantitative statements, we have to assign values to several parameters. We first discuss a baseline version of the model and then proceed to explore the sensitivity of our results to variations in the parameter values. To be consistent with the empirical work, we assume that the entire labor force consists of formal-sector workers and self-employed individuals. An essential part of the model is the adjustment cost function T(.). The comparative statics result depend on T'and T". The value of T' is determined by theory through the second equation in (10). Since we have no data allowing us to estimate T", we assume T" = 0, that is, we assume that convexities in adjustment costs are weak (Section 2e provides one scenario in which this is the case). This assumption has the further advantage that the initial level of the equilibrium real interest rate is irrelevant for the quantitative impact of macroeconomic shocks. Another building block of the model is the economy-wide quit-rate function q(.)whose properties determine the (local) effectiveness of economic policy. We take as values for the first- order derivatives the point estimates of Section 3 and assume, in accordance with the empirical 15 results, that all second-order derivatives are zero.19 The function p (.)entering the second equation of (8) is calculated using the expression (3) withthevalues, h * = q *= 0.02448,p * = 0.2148,and p = 0.9. Thequartedyseparationrateof 0.02448 is directly taken from the data and is the averaged, quarterly fraction of formal-sector employees leaving for self-employment. The probability of finding ajob, p = 0.2148, is calculated using the formula p * with a hiring (quitting) rate h = 0.02448, a formal-sector employment lev&l'i *IA-67, and a fraction of self-employed workers searching for a formal- sector job,v(1 - q *), taken to be 0.2 (see the discussion in Section 3 for the value 0.2). The implied value of v is 0.21. Assigning a value to P3 is complicated by the assumption that workers do not participate in capital markets, which implies that we cannot simply follow the real business cycle literature and infer the discount factor from the long-run real interest rate (rate of return to capital). Of course, if workers' discount factor were equal to the discount factor of capitalists, then this procedure would still be valid. But one reason why workers have only a small amount of wealth might be exactly their impatience relative to capitalists. We decide to use a discount factor P = 0.9 for the baseline model, which is considerably lower than the 0.96 used in the real business cycle literature.20 Table 2 summarizes the choice of parameter values for the baseline model. Table 3 here b) Simulation Consider first a tax reduction program which lowers the tax on formal-sector labor income, T, by 1% but leaves the (after-tax) earnings of self-employed workers, b, unchanged. For simplicity, we also assume that any change in government revenues is met by a corresponding change in government spending so that the fiscal budget is balanced before and after the tax reform. As shown in table 4, the quantitative effects of such a tax reform on all three variables of main interest is substantial: the before-tax wage increases by 0.68%, the labor-turnover rate decreases by 2.63% (of its initial value of .02448), and the formal-sector employment level rises by 0.91% (of its original level of 0.637). Table 4 here Figure I shows the effect of the tax reduction program as shifts in the labor demand and efficiency wage curve. The labor demand curve shifts to the right: for given before-tax wage, w, 191nstead of rewriting equation (8) in terms of log-wages, we simply normalize the equilibrium wage before the policy change to one. Thus, we have aqlaw = -aq/aT. The estimates of 8q/8x are constructed by multiplying the estimated coefficients a n by p(x). In the case of aq/ap, the estimate of aq/aI5 is further multiplied by .2148/6.313 yielding an estimate for aq/ap. 200f course, the value of 0.96 is derived by considering US data. 16 a tax reduction increases the after-tax wage reducing quitting which in turn increases the demand for labor and therefore p. The efficiency wage curve shifts to the left. Taken together, we conclude that the wage unambiguously rises, but the effect on p is ambiguous. The decrease in p shows that the change in the efficiency wage curve dominates. Even thoughp decreases, formal-sector employment increases because of the pronounced reduction in equilibrium hiring. Figure 1 here We can also evaluate the welfare consequences of the tax reform. Clearly, workers gain in the ex-ante and in the ex-post sense since the labor income (and therefore consumption) of formal- sector workers increases and the earnings of informal-sector workers has not changed. The consumption of capitalists might increase or decrease depending on the shape of the production and training cost function as can be inferred froin the expression for equilibrium consumption of capitalists (7). Finally, the effect on government revenue, (Twl)N, is in general ambiguous since there is a direct effect (the decrease in r) which tends to reduce revenues but two indirect effects which tend to increase revenues (the increase in w and 1). For the Mexican case considered here the average tax rate is a very small 8% which implies that the direct effect dominates and the tax revenues are reduced. As mentioned above, in our simulation we assumed that this downfall in receipts is matched by an equal reduction in outlays. Finally, it is worth mentioning that output of the formal-sector economy is always increased simply because more workers are employed in the formal sector. If the productivity of the formal sector is higher than the productivity of self- employed workers, then the reallocation of workers across sectors unambiguously raises total output. Hence, in principle both workers and capitalists can always be made better off if the government redistributes (in a non-distortionary fashion) some of the gains of formal-sector employees to capitalists. Consider now a decrease in the earnings of self-employed workers, b, by 1% of its original value.21 One possible reason for such a decrease in b is improved tax collection in the informal sector. Alternatively, one can think a drop in the demand for goods produced in the informal-sector (non-tradable goods). Clearly, if =L1 - T (recall b = 1), a hypothesis suggested by theory and aT b supported by our empirical evidence, then a 1% decrease in b is equivalent to a 1% increase in T except that in the former case the after-tax income of formal-sector employees increases by 1.68% and in the latter case by 0.68%. Notice, however, that the ratio w(1 -r)/b, which measures the pecuniary reward of formal-sector work relative to informal-sector work (earnings gap), rises in both cases by the same amount, namely 1.68%. Let us now turn to the discussion of events shifting the labor demand curve only. First, suppose that the growth-adjusted marginal (revenue) product of formal-sector work, Z, increases by 21For simplicity and for lack of precise data, we assume that initially w = b = 1. 17 I%.22 Such an increase in Z could arise, for example, if along the development path the ratio of formal-sector productivity to informal-sector productivity increases because secular technological progress disproportionally favors the formal sector. As a second example, one can think of product and/or labor market deregulation as well as a reduction in labor taxes paid by firms. It also follows from equation (8) that a reduction in the average training cost (for constant T' and T") is equivalent to an increase in Z by the same amount.23 The qualitative impact of any of these events is illustrated in Figure 2. For given wage w, a higher marginal product of labor increases the demand for labor which in turn increases hiring and therefore the probability of finding ajob -- the labor demand curve shifts to the right. Since the efficiency wage curve is unchanged but downward sloping, the wage rises but the probability of finding a job decreases in the new equilibrium. Figure 2 here The second line in table 4 summarizes the quantitative effects of a 1% increase in Z which are very similar to the previous case: the before-tax wage increases by 1.56%, the turnover rate declines by 2.19%, and the employment level rises by 0.93%. Finally, we turn to changes in the real interest rate, r. It follows immediately from Eq. (8) that a 1% (100 basis points) drop in r is equivalent to an increase in Z by T'%. Since T'= ( q )- = 10.53 in the baseline model, the impact on all endogenous variables is very large. The linear approximation used here, however, is certainly not appropriate for changes of such a magnitude and the implied changes of the endogenous variables therefore appear unrealistically large. On the other hand, our analysis indicates that the real interest rate is one of the most important variables, a result consistent with the finding by Phelps and Zoega (1998) using unemployment cross-country regressions. Notice also that even though fiscal policy has no effect on the real interest rate in the basic model developed in this paper, leaving changes in the discount factor of capitalists and/or changes in productivity growth as the only sources of interest rate changes, an OLG version of the model along the lines developed by Blanchard (1985) and used by Hoon and Phelps (1992) would deliver a link between fiscal policy and the real interest rate. Moreover, the equation system (8) can also be viewed as representing the equilibrium conditions of a small open economy version of the model with exogenous world interest rate r. c) Sensitivity analysis This section discusses how parameter uncertainty affects the simulation results. Our approach is to change one parameter value at a time and to report the effects of an increase in Z for the new parameter constellation (results for changes in Tare similar). First we decrease the discount factor 22For simplicity, we set Z = w = 1, that is, we assume that training cost are "small" compared to wage cost. 23Observe, however, that an increase in Z by 1% of its initial value is very different from a decrease in training cost by 1% of total training cost as long as T(average training cost) is considerably smaller than Z 18 of workers, P3I from its original value of 0.90 to 0.80. In another experiment, we change our estimate of the fraction of self-employed workers searching for a formal-sector job, v(1 -q *), from its original value of 0.20 to 0.30. Finally, we also consider decreases in the absolute value of the quit-rate elasticities by one standard error. Table 5 summarizes the results and reveals that the parameter variations do not change our main conclusions. Indeed, the response of the economy is surprisingly robust to changes in parameter values. Table 5 here 5. Conclusion In this paper we incorporated microeconomic evidence about Mexican workers moving into self-employment into a macroeconomic efficiency model with labor turnover and used the calibrated model economy to evaluate the quantitative effects of changes in economic policy and other macroeconomic shocks. As we have already mentioned in the Introduction, an application of this method to different countries and unemployment is high on the agenda for future research. In this respect, an interesting question is whether the calibrated model economy generates a wage curve (Blanchflower and Oswald 1994), that is, a negative relationship between wages and unemployment. A glance at table 3 confirms that at least for the case considered here, a wage curve does indeed arise, that is, regardless of the type of macroeconomic shock there is always a negative association between the formal-sector wage and self-employment. There are several extensions of the model analyzed in this paper which seem promising. First, allowing for a more realistic matching technology (see Section 2e) establishes an interesting link between search intensity and efficiency wages. Second, a comparison of the market outcome with the (constrained) efficient allocation could offer valuable theoretical insights. Finally, a stochastic version of the model with aggregate uncertainty would permit the study of business cycles. We plan to study these extensions in future work. 19 References Baltagi, B.H. (1996) Econometric Analysis ofPanel Data, John Wiley and Sons (New York). Bell, L (1997) "The Impact of Minimum Wages in Mexico and Colombia"Journal ofLabor Economics, 15:3 s102-35. Blanchard, 0. (1985) "Debt, Deficits, and Finite Horizons", Journal ofPolitical Economy 93: 223-47. Blanchard, 0. and J. Jimeno (1995) "Structural Unemployment: Spain versus Portugal", AEA Papers and Proceedings, 85: 212-218. Blanchard, 0. and L. Katz (1997) " What We Know and Do Not Know About the Natural Rate of Unemployment", Journal ofEconomic Perspectives, 11: 51-72. Blanch flower, D. and A. Oswald, The Wage Curve. Cambridge, Mass. and London: MIT Press, 1994 Bulow, J. and L. Summers (1986) "A Theory of Dual Labor Markets with Application to Industrial Policy, Discrimination, and Keynesian Unemployment", Journal ofLabor Economics 4, 376-414. Campbell, C. M. (1993) "Do Firms Pay Efficiency Wages? Evidence with Data at the Firm Level",Journal ofLabor Economics, 11: 442-70. Danthine, JP. and Donaldson, J. (1990), "Efficiency Wages and the Business Cycle Puzzle", European Economic Review, 34: 1275-1301. Danthine, JP., and J. Donaldson (1995), "Non-Walrasian Economies", Ch. 8 inFrontiers ofBusiness Cycle Research, T. Cooley (ed.). Princeton University Press, Princeton, New Jersey: 1995. Daula, T. and R. Moffitt (1995), "Estimating Dynamic Models of Quit Behavior: The Case of Military Reenlistment," Journal ofLabor Economics, 13:499-523. Eckstein, Z. and K I. Wolpin (1989) "The Specification and Estimation of Dynamic Stochastic Discrete Choice Models, a Survey", The Journal of Human Resources, 24:4 Fields G. S. (1990), "Labor Market Modelling and the Urban Informal Sector: Theory and Evidence," in OECD, The Informal Sector Revisited, Paris. Harris, J. and M. Todaro (1970) "Migration, Unemployment, and Development: A Two Sector Analysis", American Economic Review, 60: 126-142. Hayashi (1982) "Tobin's Marginal q and Average q: A Neoclassical Interpretation" Econometrica, 50(1) 213-24. Hoon, H. and E. Phelps (1992), "Macroeconomic Shocks in a Dynamized Model of the Natural Rate of Unemployment", American Economic Review, 82: 889-900. 20 Katz, L. (1986) "Efficiency Wage Theories: A Partial Evaluation",NBER Macroeconomics Annual, 1:235- 276. Kimball, M. (1994), "Labor-Market Dynamics When Unemployment is a Worker Discipline Device", American Economic Review, 84: 1045-58. Krueger, A. and L. Summers, (1988), "Efficiency Wages and the Inter-Industry Wage Structure", Econometrica, 56: 259-93. Layard, R,Nickell, S., and R. Jackman, Unemployment: Macroeconomic Performance and the Labor Market, Oxford University Press, 1991. Maddala, G.S. (1983) LimitedDependent and Qualitative Variables in Econometrics (Cambridge University Press, Cambridge). Maddala, G.S. (1987) "Limited Dependent Variable Models Using Panel Data," The Journal of Human Resources, XXII(3). Maloney, W.F. and E. Ribeiro (1998), "Efficiency Wage and Union Effects in labor Demand and Wage Structure in Mexico" Mimeo, The World Bank and Universidade Federal do Rio Grande do Sul, Brazil. Nickell, S. (1997) , "Unemployment and Market Rigidities: Europe versus North America", Journal of Economic Perspectives, 11:3 pp. 5-74. Nijman, R., M. Verbeek, and A. van Soest (1991), "The Efficiency of Rotating-Panel Designs in an Analysis of Variance Model," Journal ofEconometrics , 49:373-399. Pencavel, J.H. (1972) "Wages, Specific Training, and Labor Turnover in US Manufacturing Industries", International Economic Review. Phelps, E. (1968), "Money-Wage Dynamics and Labor Market Equilibrium",Journal ofPolitical Economy, 76: 678-711. Phelps, E., Structural Slumps: The Modern Theory of Unemployment, Interest, andAssets, Cambridge, Mass: Harvard University Press, 1994. Phelps, E. and G. Zoega (1998), "Natural-Rate Theory and OECD Unemployment",Economic Journal, 108: 782-801. Rogerson, R. (1997) "Theory Ahead ofLanguage in the Economics of Unemployment",Journal ofEconomic Perspectives 11: 73-92. Salop, S. (1979), "A Model of the Natural Rate of Unemployment",American Economic Review, 69: 117-25. Shapiro, C. and J. Stiglitz (1984), "Equilibrium Unemployment as a Discipline Device",American Economic Review, 74: 433-44. 21 Stiglitz, J. (1974) "Wage Determination and Unemployment in LDCs: The Labor Turnover Model", Quarterly Journal of Economics 88: 194- 227. Woodford, M. (1994) "Structural Slumps", Journal ofEconomic Literature 32:1784-1815. 22 Appendix. Al. Worker's decision problem a) Representative firm economy There is a continuum of ex-ante identical, infinitely-lived workers whose total probability mass is normalized to N (the number of firms). Each worker has preferences over random consumption sequences, { Cw o' that are time- and state-additive with logarithmic one-period utility function24 T U }%) = limE L O' log(C ) , (Al) w Wt t=0 Tw wt where Ow stands for the worker's pure discount factor. Workers are either employed in the formal sector (employed) or work in the informal sector (self-employed). Workers do not participate in capital markets and their consumption level is therefore equal to their current disposable income. Hence, if we denote the formal sector wage by W, the tax rate on labor income from formal-sector work by T, and the net pecuniary benefit to self-employment by B,Ot, we have W1(1 -T) if employed in formal sector in period t { W ( BI Of if self-employed in period t (A2) Here B, is the average benefit from self-employment and 0 a worker-specific (random) component. For given wage rate and benefits, a worker chooses a quit/search rule which maximizes (Al) subject to the budget constraint (A2). We are interested in an equilibrium growth path along which consumption,C , and wage, W, grow at a constant rate equal to the growth rate of technological progress,g . Thus, we introduce the following growth-adjusted variables: cw = CI/A, w =W/A, with A = A0 (1 +g)' being the exogenous parameter of labor efficiency. We also assume that average self-employment benefits, B,, grow at the rate g and define: b = B /A . We have dropped the time-subscript on growth-adjusted wage and benefit payments to indicate that these variables are expected to be constant, an expectations turning out to be correct in equilibrium. To ensure that the worker's optimization problem is well-defined, we assume PWI (1 +g) < 1. In this case, maximization of (Al) subject to (A2) is equivalent to solving: T max limE [ T log(c W,) (A3) T-_ t=0 24The log-utility assumption is not essential (CRRA-utility would suffice), but provides a direct link to the empirical section in which log-wages are used in the regressions. 23 w (1-r) if employed in formal sector in period t subject to: c. w b O, if self-employed in period t There are two sources of idiosyncratic uncertainty: uncertainty about the idiosyncratic component of self-employment benefits and uncertainty about the individual employment status. We assume that the sequence of shocks , { 0 }- with 0 = logO , is a sequence of identically and independently distributed random variables with dis1rbution function Ct(.)and density function (p(.)." Further, the probability of finding a formal sector job when searching is a constant, p. The decision problem (A3) therefore displays a recursive structure and the corresponding Bellman equation reads Ve(9,x) = max H log(w(1-r)) + , fVe(',x)d(f) ;A4 { log b + 0 + f Vs(0',x) d(O') V'(O,x) = max I{logb + 0 + f V3(6',x) d0(f') ; p+log (w(-f)) Ve(0',x)d0(0')) + (-p) (logb +0+ f VP(0',x)de('0)) , where x - (w,p,t,b). Further, Ve(O,x), respectively V'(O,x), denotes the (utility) value of pursuing the optimal policy when employed, respectively self-employed, when the macroeconomic state is x. It follows from (A4) that the optimal strategy is to define acut-off value, 0 = 0 (x), and to quit CC a formal-sector job if 0 ; 0 and to search for a formal-sector job if 0 0 . Moreover, the optimal cut-off C C value is implicitly defined by the following equation 0 - log(w(1-T)/b) - W EAV(6c,x) = 0 (A5) EA V(6 ,x) -P 1(D (0 )log(w(1-t)/b) -f O'd1(0') c 1 ((1 -p) (D(O) c, zt e [ C The economy-wide quit-rate function is then simply given by q(x) 1 - D(0c(x)), which establishes the link between the solution to the worker's decision problem and the quit-rate function entering into the firm's decision problem. Using a probit model in Section 3 amounts to assuming a normally distributed0 and a linear approximation for the function 0 (x). Finally, the fraction of self-employed workers searching (applying) for a formal sector job is 4(0 (x)). 25Although most of the properties of the quit-rate function also hold for general Markov processes, the particular expression for p (see A9) would of course change. 24 Since = -cp (0) ae for x = w,p,T, b, the properties about the partial derivatives of the quit- axn c axh n rate function stated in Eq. (2) follow from implicitly differentiating (A5) and signing the result. In order to calculate the function p(.), we use the following expression: 1 + EAV c _ ww aw O EAV 1-I3 a C (A6) 1+ P.0(l-P)ONOd 1+ 1 I - k,,(1-P)A(ed w 1 + (P() g(cx) where g,(0,,x)is a function independent of the probability density 9(0c). Suppose now that there is a (growth-adjusted) fixed cost of searching, f = logF. In this case, the optimal policy is to define to cut-off values, Oc and Oc2, and to quit a formal-sector job if 0 ; 0c2 and to start searching if 0 0c. Moreover, we have 02 = 02(x) implicitly defined by an equation analogous to (A5) and0, = ci (x) = 6c2(x) - flp. Forthe quantitative analysis both the derivatives ofthe quit-rate function, a -P(6 2 c, and the derivatives of the search-rate function,- as = aC, play an ax c2 axT ax cd ax n nn important role. The empirical analysis provides us with detailed information about-L-, but good data on as are lacking. However, since we have Oci(x) = c2(x) - f/p, the following exprextions relate the searcIf- rate function to the (observed) quit-rate function: a (C2flP) aq and ax. (P( (c2) ax n as _ c(e2-flp) [ aq /1 Tp = - [_+ <P (9c2) f2. ap <p(Oc2) ap c 2 In principle, the above relationship in conjunction with information on search cost, c, and the distributional characteristics of the idiosyncratic shocks, 0, could be used to calculate the response of the search intensity to aggregate shocks. However, the expression also makes clear that the quantitative answer crucially depends on distributional assumptions determining the ratio of probability densities evaluated at the COC62 -flp) cut-off value, . In this paper, we choose not to make assumptions about the distributional characteristics of 0. ? Instead, we impose simplifying assumptions rendering all results independent of the probability density evaluated at the respective cut-off value. More specifically, we deal with the problem of 26Especially when 0 is interpreted as "taste-shock", such a procedure seems very questionable. 25 unobserved search intensity by assuming that search costs are small, that is, f - 0. In this case, the derivatives of the search intensity function are simply given by the negative of the derivatives of the quit-rate function: as 8q ax" ax" xn nx The assumption of small search cost seems to create a problem by itself since in this case the quitting/search model developed so far implies that s = 1 - q, which is not the case for the data set we consider. However, even if there are no search costs, the above relationship need not hold if the idiosyncratic shock process exhibits serial correlation. In the case of serial correlation, there is still one common cut-off value for employed and self-employed, but the (stationary) distributions are different because of self- selection, that is, the quit rate is 1 - (O c)but the search intensity is 0 (O ) and therefore s * 1 - q. Since we only have a very limited knowledge of the serial correlation properties of this idiosyncratic process, we only deal with the extreme case in which there are two types of workers. The behavior of type I workers is correctly described by the Bellman equation (A4) (no serial correlation). Type two workers, on the other hand, never move (no stochastic shock) and therefore stay either in the formal or the informal sector forever (for the time span covered by our data). If a fraction v of the total labor force is correctly characterized by the Bellman equation (A4) (type I), then the search rate is given by s(x) = v(1 -q(x)), which is the expression used in the text.2 Notice that v is a number independent of x. b) Many Firms We consider now an economy in which i = 1,...,N identical firms set wages in an uncoordinated fashion (they play a Nash game). Except for this modification, the economy is identical to the one considered in part a). For simplicity, we again consider the case of no search cost. The Bellman equation of a worker who is currently employed by firm i paying wage w, reads pe(O,W x) = max log (w(1-r)) + 0 f pe(6',w x) def) (A7) Ilog b + 0 + Ow V'(9',x) d((') , where VS(.) is the value function of the Bellman equation (A4) (assuming all other firms are paying wage w). The optimal strategy is again to define a cut-off value, Oct = c(w ,x), and to quit a formal-sector job if 0 0 and to search for a formal-sector job if 0 s Oc . Moreover, tie optimal cut-off value is implicitly defined gy the following equation cl - log (w,(1-)/b) - 0WEAP(w,0c,,x) = 0 EA V(w, , 0,x) - 1 - D(O c) Ilog(w (1 -T)) - p log (w(1 -C)) - (1 -p) log b - p EA Vj (A8) -_4 (1-p) d w cr 9OcI 27This formula changes once we allow for the possibility that the fraction of formal-sector-only workers is different from the fraction of informal-sector-only workers. 26 The quit-rate function of firm i is then given by 4(ws,x) a1 - (D(,(w,,x)). Since aq= -q(6 ) c for x = w,p,r,b and a _q_ c , the properties Snex c x n aw cl n I about the partial derivatives of the quit-rate function stated in Eq. (1) follow from implicitly differentiating (A5) and signing the result. In order to calculate the function .(.), we need an explicit expression of the following derivative: 1 + aEA aa 8w '"J3* aEAv1 cl (A9) PW O(O) 1+ 1 1- PW(O) w 1 + g2(Ec,x)p(O where g2(.) is a function independent of the density 9(0). We are interested in the ratio aw = ' . (A10) aq ao aw aw Expression (A 10) depends on the probability density evaluated at the cut-off value, (p(Oc), which is difficult to pin down without a very detailed knowledge of the distributional characteristics of the random variable 0. Evoking again the principle of "independence of results from irrelevant distributional details", we ask for an approximation yielding an expression for p which is independent of ( (O). One such approximation is to take the limit 9(Oc) - 0, that is, to evaluate p at small values of 9(O). In this case, (A 10) becomes Equation (5) used in the text (observing that 4(D) = 1 - q). A2. Capitalist' Decision Problem There are i = 1,...,N identical, infinitely-lived capitalist with time-additive preferences over consumption sequences, { C1 )=o U({Cit)70) c1 i (All) 27 In (All) 03cstands for the pure discount factor of capitalists and pfor their coefficient of relative risk aversion. Each capitalist owns one firm. Each firm i (owned by capitalist i )combines capital K,, and labor L to produce output Y . Changes in the employment level create adjustment cost. Hiring new workers is associated with recruitment and training cost. In the case of training cost, we assume that training takes one period and that the human capital created through training is destroyed once the worker quits (match specific human capital), that is, newly hired workers always have to be trained regardless of their previous work experience. Following Phelps and Hoon (1992), whose specification of employment adjustment cost derives from Hayashi's treatment of capital adjustment cost (Hayashi 1982), we assume that the total cost of training and hiring is independent of the capital stock and that net output produced is linear homogenous in (K ,A L ,A H,), where A, stands for the (common) efficiency of labor and H for the number of I itI it worers hired by firm i. More precisely, we assume that output net of adjustment cost is equal to F(K, A L ) - T(A H A Lt), where F(.,.) is a standard neoclassical production function and T(.,.)is a linear homogenous function. Moreover, the adjustment cost function satisfies: aT(x,y) > 0, ax at(x,y) < 0, t(x.y) ; 0. Notice that we permit the case of linear adjustment costs, a(x'y) = 0, for ay ax2 ax2 which adjustment to the optimal employment level is instantaneous. Each firm employs a large number of workers. Firm i chooses sequences of consumption (of owner i ), capital, investment, employment, and hiring as well as a wage rate which maximize (Al 1) subject to the constraints29 K 1+= (1-8)K + Ii L = (1 -q(w ,x))L + H I,t+1 I it it Yt = F(K ,A L,,) (A12) Y = C +I +wA L +T(AL ,AH it it It I t it t it t it Kio , L,o given . Introduce the following growth-adjusted: k = K / A , i = I /A,,c = C /A Y Y / A If It 1 It I t I II t t It t (recall that the size of the labor force is normalizedto one). Introduce Airther the hiring rate = H /L It 1t 11 28We only allow the firm to choose one wage rate, not an entire sequence of wage rates. Deriving the quit-rate function from the worker's Bellman equation when the firm can choose arbitrary wage sequences { W }=o is a nightmare. 29We assume that each self-employed worker produces what he consumes, namely b. If b is interpreted as unemployment benefits, then it should be added to the right-hand-side of the last last constraint . 28 and let l1' = L . Clearly, maximizing (Al 1) subject to (A12) is equivalent to solving the following optimization problem: max t t 1=0 C ~ subject to : k, _ 1 [(1-8)kl + ill] I,t+1 l+g 1t at3 1 = [1 + h - 4(w.,x) 1i F(ki,,l,) = ci + i i + +T(h,)1, klo , lI given , with T(h1) = T(h ,1) and the growth-adjusted discount factor Oc = (1 +g) i. To render the capitalist' maximization problem well-defined, we require i < 1. Writing down the Euler equations associated with the optimization problem (A13) leads to Equation (4).30 3oThe sufficiency conditions require certain assumptions about the second-order derivatives of the quit-rate function. We do not attempt to derive these properties from the Bellman equation (A4) and (A7). 29 Table 1. Summary Statistics Var Nobs Mean Std. Dev. Min Max move 100978 .02448 .1545 0 1 hire 100978 .6313 .1866 .34 1.13 mwl 100978 1.674 .05319 1.51 1.76 mw3 100978 1.856 .1060 1.54 1.94 exp 100978 21.61 13.01 0 62 exp2 100978 636.3 684.9 0 3844 sch 100978 7.302 2.808 0 11 sch2 100978 61.20 37.02 0 121 Table 2. L_ _ Probit Regression Unconstrained Symmetry Imposed Summary Statistics Coeff. Std. Err. Coeff. Std. Err. Mean Std. Err. move .0245 .1545 log w -1.3153 .3895 1.674 .05319 log b .8879 .1965 1.857 .1060 log w - log b -.6436 .1522 -.183 .06625 .3122 .05647 .3650 .04801 .6310 .1866 school -.08249 0.01231 -.08253 0.01231 7.302 2.808 (school)' .004241 0.000968 .004232 0.000978 61.204 37.02 experience .03200 .002887 .03200 .002886 21.612 13.00 (experience)2 -.000535 .0000556 -.000536 .0000556 636.327 683.9 constant -1.6167 .3860 -2.3195 .06131 Nobs 100978 100978 Significance X2(6)=440.8 p=0.00 X2(5)=440.7 p=0.00 Test of aw--ab X2(l)=3.41 p=.065 30 Table 3. Parameter Values for Baseline Model Quit-rate elasticity aq/8w = - aq/ab = - aq/aT -.03737 Quit-rate elasticity aqlap .06231 Formal-sector employment 1' .637 Average turnover rate q * = h .02448 Fraction of self-employed searching for a job v(1 -q ) .20 Workers' discount factor O .90 Implied average probability of finding a job p .2148 Implied parameter p 2.54 Implied parameter v 0.21 Table 4. Policy Effects: Baseline Model Tax reduction AT = -.01 MPL increase AZ = .01 Wage change Aw .0068 [0.68%] .0156 [1.56%] Probability change Ap -.000281 [-0.13%] .000776 [.36%] Turnover rate change A q -.000643 [-2.63%] - .000535 [- 2.19%] Employment change Al .00577 [0.91%] .00589 [.93%] Table 5. Sensitivity Analysis: Effects of AZ = .01 O= .80 v(1 -q) = .3 dqlaw = -.0285 aqlap = .0541 Aw .0222 .0183 .0156 .0157 Ap .002407 .000579 .000593 .000789 Aq -.000803 -.000628 -.000409 -.000544 Al .00802 .00687 .00449 .00599 31 Figure 1. A reduction in labor-income tax. A W' EW / D' P Figure 2. An increase in the marginal revenue product of labor EW w1* D P 32 Self-Employment and Labor Turnover in LDCs: Cross Country Evidence William F. Maloney LCSPR 1.1 Introduction This paper develops and tests an integrated approach to understanding two outstanding questions central to understanding the functioning of LDC labor markets and the impact of labor legislation. Though the paper focuses primarily on Latin America, the issues and analysis are germane both to other LDCs and the industrialized countries. The first is the role of the large informal sector in the region (see table 1). A traditional view argues that the sector testifies to government or union induced rigidities that force formal remuneration above market clearing and ration workers into informality.' We argue that this view is probably incorrect and that it is difficult to draw any conclusions about efficiency from sector size alone. The second question centers on what recent findings of high turnover, a common measure of rigidities (see Nickell 1997), imply about the flexibility of labor markets in the region.2 It is often asserted that high firing costs and excessive benefits in the formal sector prevent the efficient allocation of workers among jobs.' However, as table 1 suggests, average tenure is shorter, and a larger fraction of the work force has been employed in their current position for less than two years in Latin America than in the OECD. We argue that this probably cannot be interpreted a priori as evidence of greater flexibility. The structure of the paper is as follows: Section 2 heuristically develops a model that moves beyond the standard segmentation-based view of the relationship of formal and informal sectors, and incorporates the increasing evidence that a large fraction of the employment in the informal sector is voluntary. It is developed in an efficiency wage context both because recent evidence suggests that much observed segmentation may arise endogenously rather than being imposed by labor unions or minimum wages, and because it permits explicit modeling of the determinants of turnover. Predictions can be made about how the size of the self-employed sector, the degree of segmentation in the market, and turnover should move with the development process and policy innovations. Section 3 examines cross country data from Latin America, Europe and Asia with three objectives. The first two are straightforward: to test the predictions of the model about the size of the informal sector and rates of turnover with respect to several key labor market, productivity, and demographic variables suggested by the theoretical framework and second, to suggest the direction of possible influence of variables that are theoretically ambiguous. But somewhat speculatively, we also attempt to provide more informed estimates of the incidence of unmeasurable distortions that ration more workers into the informal sector or rigidities that decrease turnover. Since our theoretical 1 See Harris and Todaro (1970) for an early presentation of this view. 2 See Maloney (1995) for Mexico, Gonzaga (1996) for Brazil, Anderson Shaffner (1997) for Colombia, Mirquez and Pag6s (1998) more generally. See Hopenhayn and Rogerson (1993) for a recent theoretical discussion. 3 See for example Burki and Perry (1997) The Long March. framework abstracts from such exogenous phenomena, we tentatively measure their impact by the deviations from the model's predicted values. Though it is trivial to raise objections to this approach on either theoretical or empirical grounds, the results at once strongly coincide with the stylized facts about industrialized countries and challenge what is commonly thought about Latin America: with some predictable exceptions, regional labor markets do not appear unusually distorted or inflexible. 2. Motivation and Theoretical Overview The empirical work here is motivated by a macroeconomic model based on micro behavior of workers describe in detail in Krebs and Maloney 1998. It is built as a growth model so that secular movements in labor productivity can be incorporated and makes predictions about movements in formal and informal sector employment, the degree market segmentation, and labor turnover rates across the course of the development process. It also attempts to incorporate two emerging stylized facts about LDC labor markets. 1. The informal sector is extremely heterogeneous containing both voluntary and involuntary members. The informal sector is frequently considered the disadvantaged segment of a labor marketed segmented by government or union intervention in the wage setting process in the formal sector.' During downturns, the sector is thought to expand as it absorbs displaced workers, then contracting again with recovery. While some fraction of the sector corresponds to this view, recent studies find that many of the informal employed are voluntarily so and should probably be viewed as unregulated entrepreneurs. Comparisons of formal/informal wage differentials traditionally used to show segmentation have been shown to be meaningless, and there appears to be high degrees of mobility among sectors.' The Mexican micro-enterprise survey suggests that 70% of workers enter the sector voluntarily for reasons of independence or higher income and recent time series data from Mexico and Chile suggest that both the size of and transitions into the self-employed sector behave procyclically.6 There is increasing evidence both in the sociology and economics literature that suggests a life cycle view of the trajectory between formal and informal self-employment: in the absence of well-functioning credit markets and effective educational systems, workers may take 4 See the classic statement of this view in Harris and Todaro(1974). 5 In the absence of any distortions, we should find a wedge between formal and informal incomes that incorporates the value of benefits forgone, the value of taxes evaded, the value of lifestyle differences between wage and self-employment, capital costs, implicit training costs and payments in kind. Without this information, wage comparisons tell us nothing about segmentation or relative welfare between sectors. See MacIsaac and Rama (1997) and Maloney (1995, 1997a.) 6 See Maloney (1997b) and Pages(1998). Pais de Barros finds no cyclical movement in Brazil. Saavedra in Peru finds a broadly countercyclical movement but this may be largely driven by secular trends. 2 formal sector jobs to accumulate human and financial capital and then quit to open their own business.' In sum, there is ample evidence suggesting that self-employment is a desirable destination for many workers who voluntarily leave formal employment. 2. In the absence ofgovernment or union induced rigidities, there is still strong evidence of "segmentation. " Recent work on Mexico challenges the customary view of the sources of labor market segmentation. Minimum wages are not binding (See Bell 1998) and the evidence suggests that union power is directed largely to the maintenance of employment and find no significant effect on wages.' As Marquez and Ros (1990), noted, and has been confirmed by later studies for Peru (Shaffier,1998)and Guatemala (Funkhauserl998), wages of similar workers rise with firm size, much as they do in industrialized countries. Further, Mirquez (1990), and Abuhadba and Romaguera (1993) find evidence consistent with efficiency wage effects in the high correlation of wage differentials among Chile, Venezuela, and Brazil and the US. This evidence suggests that the conditional wage dispersion (wages adjusted for human capital) may be emerging endogenously and is not due to either government or union intervention. Both stylized facts suggest an interpretation of the interaction of formal and informal markets rooted in the extensive literature on efficiency wages where firms voluntarily pay wages above the market clearing level.' One common variant of these models arises from the difficulty of monitoring individual workers and the lack of any penalty from being caught "shirking" - any activity, or lack thereof, that might be detrimental to the firm. If wages are market clearing, a worker fired for shirking can simply get anotherjob at the same wage. However, if all firms pay higher than market clearing wages, unemployment will be created in the economy that creates a disincentive to being laid off and hence to shirking. Since in many Latin American countries, workers can be fired only with difficulty, the "turnover" variant of efficiency wage models seems more appropriate: Firms must invest resources in workers when they are hired, perhaps through training or through the process of recruitment, that will be lost if the worker leaves. Hence, it is worthwhile for firms to pay higher wages and raise the opportunity cost of leaving to other firms or jobs. This argument may be particularly compelling in LDCs given the life cycle model of self-employment developed above. In an inversion of the commonly held view that higher than market clearing wages create informality, it may be that the attractiveness of self-employment causes firms to pay above market clearing wages. This, in fact, does create a subset of the informal sector that is involuntarily self-employed and who are 7Aroca and Maloney (1998) model the transitions into informality as a destination of entrepreneurs and find evidence using logit techniques adapted to panel logit context. 8 See Maloney and Ribeiro (1998), and Hernandez -Laos (1998). Panagides and Patrinos (1994) find some wage effects, but these are likely to disappear when relevant firm characteristics, such as size, are included. 9 For discussions of the theory of efficiency wages see Stiglitz (1974), Krueger and Summers (1988), Phelps (1994). 3 unable to easily move back into the formal sector.'o Thus, potential self-employed, aware of the high rates of failure of small businesses, will think twice about leaving formal employment if the probability of being rehired is reduced." The efficiency wage approach has the advantage of dealing explicitly with the issue of turnover, the second of our central issues to be examined. However, it also complicates our view of the informal sector and what its existence reveals about inequality, poverty or labor market distortions. A large fraction of workers may treat the sector as a very desirable destination either to attempt to run a business, or as a place where workers quitting an undesirable formal sector job may search for another and for them the traditional conflation of informality with disadvantage or relative poverty are inappropriate. However, it is also clear that some fraction is trapped there involuntarily- that is the expected byproduct of efficiency wages. Despite abstracting from wage rigidities introduced by minimum wage or unions, this approach is useful for understanding labor markets in the region. First, in Brazil, Chile, as in Mexico, government mandated minimum wage and the curbed power of labor unions are unconvincing as the principal sources of segmentation. Second, efficiency wage phenomena are likely to exist as an important underlying determinant of wage structure, however overlaid by other local institutions, and they have long term implications for labor, education, and poverty alleviation policies. Third, careful modeling these effects aids in identifying abnormalities in informal sector size or turnover that may be interpretable as more reliable evidence of distortrons that are not explicitly introduced through the model. Though the model is general equilibrium in design, the intuition can be distilled to two equations. These can be broadly represented as an upward sloping "incentive curve" and as a downward sloping labor demand curve shown in figure 1. The curves are plotted with the probability of being hired in the formal sector on the X axis and formal wages relative to the average in the self-employed sector on the Y axis. The incentive curve II captures the essence of the efficiency wage story. It represents the constraint that firms face in trying to prevent workers from leaving with their training and opening a business in the informal sector. The higher the probability that a worker will be hired in the formal sector if the business turns out to be less successful than expected, the greater the likelihood of quitting his current formal sector job, and hence the higher 10 Some fraction of the informal sector serves the role of unemployment benefits in industrialized countries. This raises the possibility of an analogy between the size of the sector and the natural rate of unemployment (NAIRU). The large movements in formallinformal remuneration in LDCs can generating insights on the elasticity of the NAIRU to unemployment benefits in the industrialized countries where the lack of variation in wages/unemployment benefits have prohibited rigorous testing. (Katz and Blanchard 1997). " For the worker's decision to enter self-employment, we have in mind a model something like the "noisy selection" model of Jovanovic (1982). Here, workers have only a very diffuse idea of their ability as entrepreneurs and whether they will be able to stay in business. Only by actually opening a business can they learn about their true underlying abilities. The ability to be rehired is therefore and important consideration in risking self-employment. 4 the formal sector wage must be to persuade him from trying his luck. Though it is not a traditional labor supply curve, it incorporates the usual depressing effects on labor supply to the formal sector of increased attractiveness of the informal sector, or a rise in taxation of formal sector wages. The second curve DD is similar to the traditional labor demand curve. It can be argued that as wages rise, formal firms hire fewer workers and the probability of being hired falls. 2.2 Comparative statics These two curves allow analysis of the impact of several important variables on the size of the informal sector, what share of it is likely to be voluntary, and on rates of turnover. We analyze the impact of three broad classes of policy interventions or economic innovations: Increases in labor productivity or firm profitability, a rise in the benefit to being self employed and changes in hiring costs. 2.2. a Increases in labor productivity or firm profitability. (7) This includes technological progress, and of particular interest, a fall in labor taxes, or a reduction in any regulation that adversely affects productivity. Any of these changes has the effect of shifting the DD curve to the right along the II curve (figure 2). As productivity increases, firms are willing to hire more workers, and hence to increase the probability of being hired. The movement along the incentive curve implies that a higher wage relative to that in the informal sector must be paid to retain workers. But this also necessarily implies that a larger fraction of informal workers is involuntary and would experience welfare enhancements upon finding ajob in the formal sector. For each of the shocks discussed below, the rise in formal sector employment and wage therefore may have negative distributional effects. The impact on turnover is ambiguous. Both wages and the probability of being hired rise over time with opposing effects on turnover and it is not clear, ex ante, what the net effect should be. Though the implied shifts of curves are the same for the following cases, it is worth highlighting certain aspects. Technological progress: A secular rise in formal sector productivity due to technological progress has the effect of raising both the level of employment in the formal sector, and the wage paid there. The model has the prediction, then, that as countries grow, a larger and larger fraction of those self-employed are involuntary and segmentation increases among the sectors. For very poor countries, salaried vs. self-employment maybe very close substitutes, but in richer countries, perhaps Argentina, self-employment is, on average less desirable. This dynamic may offer some insight into the elusive Kuznets relation of worsening and then improving distribution with development. A poor country has a very large self-employed sector. As productivity rises, segmentation between the formal and informal sector increases and increases the wage differential between formal and informal sector workers, worsening distribution. However, eventually, the self-employed sector shrinks to so small a size that, though the differentials are great, the number of workers affected is small, leading to a 5 relative improvement of the Gini. Using estimates of the important elasticities, section 4 shows this to be a plausible dynamic. The indeterminacy of turnover suggests that it is not obvious that LDC's should have higher or lower rates of turnover. Regulations and Taxes: Any regulation that can be reinterpreted as a tax on firms-non-wage benefits, firing costs- or any economy wide regulation that leads to lowering the marginal product of labor reduces the size of the formal sector and lowers the formal sector wage. It is important, however, to bear in mind that this effect is most compelling if workers do not value these benefits. To the degree that they do, this is simply payment in a different form. This even applies in some measure to restrictions on firings which can be seen as a tax equal to the option value of the ability to divest of an underperforming asset. To the degree that the worker sees these costs as an insurance premium against termination, they are passed along to workers as lower wages with no impact. However, it is easy to generate scenarios where this might not be the case, and the net result is to reduce employment, and turnover.12 It is important to highlight that the distributional impacts of a reduction in labor taxes are the opposite of those generally postulated in the traditional view with a minimum wage or union induced wage rigidities. There, a reduction in taxes reduces total remuneration to the formal sector relative to the formal, and at the same time increases the size of the formal sector, with likely positive distributional effects. Here the result is less clear. 2.2. b Rise in the benefit to being self-employed, or a reduction in worker taxes (r) Anything that raises the benefit to being self-employed relative to being formally employed increases the rate of turnover and causes a shift in both the incentive curve and the demand curve (figure 3). In the former, for any probability of being hired, the formal wage must rise to offset the increased desirability of the informal sector. In the latter, the increased cost of retaining workers also shifts the labor demand curve left. What is clear is that employment in the formal sector falls. This coincides with existing literature on unemployment in the OECD countries that increasingly focuses on the level and duration of benefits as the key determinant of unemployment. Nickell finds the duration to be the key determinant of long-run unemployment levels while Blanchard and Jimeno(1995) attribute the relatively high Spanish unemployment to the fact that Spaniards get access to benefits of indefinite duration if employed only 6 months of the last 4 years while Portugese workers must have been working 1.5 of the last two years. Benefits of indefinite duration are similar in principal to self-employment as an alternative to formal work. The absence of unemployment benefits in LDCs has the effect of collapsing both the self-employed and the 12 See Bentolila and Bertola(1990) for a discussion of the impact of firing costs and labor demand. 6 "unemployed" into one sector. The impact on relative wages, however, is ambiguous since both curves shift left, hence it is difficult to say anything definite about distribution. This indeterminacy also prevents any ex ante statement about turnover despite the increased difficulty of finding a formal job. Income or other taxes: Any tax that finances a public good or whose benefits are perceived as below its cost in terms of taxation renders the informal sector more attractive. Internal mobility restrictions: Where internal reallocation in the formal firm is highly regulated, talented workers may choose to work on their own. 2.2. c Changes in Hiring Costs Any policy that serves to lower the fixed costs of hiring (recruitment, training, etc.) reduces the loss involved with a quit and hence the magnitude of efficiency wage effects. Showing the effect graphically is difficult since it involves both shifting and changing the slopes of both curves. However, what is clear, is that in the limit where training costs fall to zero, there is no longer any need to pay efficiency wages, no segmentation, and there is an increase in formal sector employment. The impact on turnover is positive since there is no reason for firms to prevent identical workers from leaving and replacing them with new ones. Public education: Public education has long been justified on the grounds that it addresses the externality implicit in the efficiency wage story: the private sector will under-invest since the basic skills they pay to impart can be easily transferred elsewhere. To the degree that poor LDC education systems force both training and socialization costs on individual firms, the wage gap between self-employed and formal salaried workers will be larger, segmentation greater and distribution worse, and a larger fraction of the self- employed involuntarily employed. This offers another channel through which improving education may equalize the distribution of income in the economy. Reduced Interest Rates: Reduced interest rates lower the cost of investment in human capital and thus lower hiring costs. Better job matching and signaling: If the recruitment and selection process constitutes a sizeable fixed cost, any improvement in mechanisms to promote good matching, or that reliably signals workers' skills, such as the education certification schemes in Mexico, cause all the same desirable outcomes. Trade Reforms: To the degree to which other reforms, such as that of the current account, increase the demand for skilled labor and raise implicit training costs, segmentation and wage dispersion may increase. This may offer one explanation for the increasing wage dispersion observed with trade liberalization in Mexico and Chile. 7 3. Cross Sectional Regressions We next test these hypotheses on a cross section of countries for which a consistent set of productivity and distortion variables are available. -In the first set of regressions, we examine the determinants of self-employment as a share of the total work force. In the second set, we focus on two measures of turnover. Using deviations from the predicted values of these regressions, we construct somewhat speculative measures of the magnitude of distortions and of rigidities. 3.1 Variables The data sources and more detailed descriptions are listed in Appendix I. 3.1 Dependent Variables Share of Workforce in Self-employment: The OECD tabulates the share of non-agriculural workers in self-employment or as owners of firms. To the degree possible, the same variable was created from the employment and household surveys from Latin America. We focus on self-employment rather than informality more generally for two reasons. First, we believe that it is the act of opening a business that is the central issue and that informality while important and often highly correlated is secondary." Second, data on informality in the OECD is largely unavailable and the LAC data sets differ in the variables available to use as proxies. Third, data on those employed in micro-enterprises is not available for the OECD. We assume that the total population employed in the self-employed sector, both as owners and workers is proportional to the share declaring themselves self-employed. Mean Tenure in the Manufacturina Sector and Share of the Manufacturing Work Force with under Two Years of Tenure: These are two alternative variables available from the OECD and then generated from the LAC household and labor market surveys. We focus on manufacturing turnover because this is the best proxy for formal sector turnover that is available. 3.1.a Formal Sector Labor Productivity/Profitability (Z) Industrial Value Added (Indust. V.A.): The log of industrial value added per industrial worker is the proxy for formal sector labor productivity. Social Security Tax-Employers (SSEmp): Social security (broadly defined) taxes as a share of the wage by worker. The model suggests that there may be different effects. Employment Protection(Protection): An index of employment protection constructed by Gustavo 1 See Levenson and Maloney(1998) for a development of this view. 8 Marquez that captures both the difficulty of laying off workers and the cost in terms of severance pay." Unfortunately, this leads to roughly a halving of the available observations and hence a separate set of regressions are run using this reduced sample. 3. 1.b Rise in the benefits to being self employed, or a reduction in worker taxes. (T) Social Security Tax-Workers (SS Worker): Social security (broadly defined) taxes as a share of the wage by worker. Ideally, we would have a measure of labor productivity in the self-employed sector. Unfortunately, this is not feasible as even in the OECD possible proxies, such as wages in the commerce or other services are not consistent across countries. However, social security tax incidence on workers does capture an important element of the relative attractiveness of each sector. 3.1.c Hiring Costs. Education (Education): The share of the appropriate age group with secondary education. No cross country direct measures of hiring costs are available. However, public education is a public good that addresses exactly the externality identified in the efficiency wage model. The more firms have to train, the more they have to lose by workers moving to another firm or to self-employment. We interpret any effect of this variable as working through training costs. Real Interest Rate (Real Interest): The real interest rate affects the cost of training workers to raise future productivity as it would in the case of any other investment. We use the average of the 30-90 day borrowing rate deflated by changes in the CPI in most cases. While this is not generally the rate at which larger corporations borrow, it is none the less a rough indicator of the cost of investing in workers in the economy. 3. 1.d Other Variables: Duration of Unemployment Benefits: (U Benefits): Some share of the self-employed in Latin America would be found unemployed in the industrialized countries where unemployment benefits exist and are often generous. Its exclusion as an alternative to self-employment may bias results. For OECD countries with traditional unemployment benefits, the variable takes the value of the duration of unemployment benefits which Nickell (1997) found this to be the most important variable for explaining levels of unemployment. For Latin America and other countries, we calculate the standard severance pay package given the mean tenure (or predicted if unavailable). 14 Chief Economist Office of the IDB. 9 Youth (Youth): We include one demographic measure as well, the share of the working population found between the ages of 16 and 20. The model implicitly assumes homogeneous work forces across countries. This is clearly not true as the share of young workers is much higher in Latin America than in the OECD countries. This variable is most relevant to the turnover regressions where traditionally young workers have higher rates of turnover as they shop around for careers. But it may also have a similar interpretation in the self-employment regressions. Latin Dummy (LA): We include a dummy for being from the Latin American region. In theory, this may capture any difference between OECD economies and the region, including labor legislation." Ideally, we would like to eliminate the significance of this variable by including the labor market variables that it may be proxying for. 3.2 Results The results must be interpreted with caution. First, we have at best 40 observations and in our most courageous moments, only 17. This in some cases can make the results sensitive to the countries included. Second, the data are not uniform. Most of the OECD variables were gleaned from presumably consistent publications of that organization while the Latin variables were individually extracted from not necessarily consistent survey data. The LAC dummy may pick up these data discrepancies. Despite these potential pit-falls, the regressions prove surprisingly robust and consistent with the model. The three Formerly Socialist Countries in the sample, Czech Republic, Hungary and Poland have extraordinarily low self employed sectors given their income level and affected the results. Since it seems likely that the repression of entrepreneurial freedom under communism is related to this result, we did drop these. However, the addition or subtraction of most of the other countries might change the parameter values some, but the overall story remains the same. 3.2 Results: Self Employment The results are broadly consistent across the regressions. Column la in table 2 presents the complete regression with all variables included with the exception of the employment protection measure. As is clear, the employers' social security tax, the real interest rate, the level of education, the share of young people enter significantly and of predicted sign. Progressive parsing out of the less significant variables in column lb makes the labor productivity and educational variables significant. In no case was the worker's share of social security remotely significant. Part of this poor performance may be due to measurement error. In roughly 25% of the cases, the social security law dictates a progressive tax that varies greatly across the range of incomes. We chose the midpoint 1 We included a squared income term as well but it was never significant. This does not, however preclude more exotic non-linear functions. 10 of this range, but we can have no way of knowing if, in fact, this represents that average tax on labor. Similarly, the unemployment benefits variable is never significant, nor, in this case, is the LA dummy. The significant variables enter of the signs predicted by theory. Most important in terms of magnitude is formal sector labor productivity. Taking the extreme values of this variable would account for 9.6 points of the variance in the share of self-employments detailed in column I of table 3. Figure 4 illustrates the same, but important conclusion: a large self-employed sector is not ipso facto evidence of distortions, but that the opportunity cost of self-employment is lower in poorer countries. The relative youth of Latin America's population also explains much of the variance with the spread across the sample accounting for 8.8% of the higher share in self-employment. The education variable that measures the degree to which firms must bear the burden of financing overall education would account for roughly 7.5 percentage point difference. Real interest rates are also surprisingly important. Peru's very high self-employed sector (46%) and high real interest rates (67%) are clearly dominating the relationship although even dropping this outlier yields a significant coefficient. The difference between real interest rates of under 5% as in the OECD vs often above 30% in Latin America is worth over 6%. Once again, the importance of ensuring macro-economic stability, and reducing risk to bring down interest rates seems clear. In general, these effects dwarf the impact of any of the three labor market variables. Across the range observed, social security taxes on employers explain relatively little of the size of-the sector, 3.2 percentage points in the share of self-employment. The same exercise was repeated with the smaller sample arising from using the Marquez protection index, Ic-le. With fewer observations, the apparent collinearity of the productivity variable and education variables makes identifying the "correct" parsimonious form difficult (Id-i e). However, in all cases, the employment protection index has the impact of increasing the size of the self-employed sector. But, again, the difference between the highly protective Bolivia, Honduras, Mexico, Peru or Spain compared to the unprotective US or UK is worth only about 1.35 percentage points in the share of self-employment, a fairly small impact. The results are similar to those of Marquez (1998) who also found a positive impact of his protection index, as well as a negative sign on GDP per capita which may be seen as a proxy for formal sector productivity. In sum, the three labor distortion variables, the tax on salaries, on payroll, and restrictions on hiring and firing have relatively small impacts compared to those of the productivity, real interest rate, education variables and, in the larger sample, the relative youth measure. Thus, given the level of productivity in a country which again, may be affected by labor legislation, it is hard to argue that these distortions are responsible for the size of the sector. II The Adjusted Size of the Informal Sector: A Measure of Unobserved Distortions? Can we say anything about distortions on which we have little reliable information, such as union or government induced wage rigidities from these regressions? Perhaps. In theory, the residuals of the regression capture the impact of all variables not explicitly included in the regression, including other labor market distortions. It is absolutely correct to argue that they also include any country specific variables, and any error in measuring sector size which cast doubt on using the residuals for this purpose. However, this critique applies to the use of the raw sector sizes as well and if sector size is thought to contain information, then the adjusted values obtained from the residuals are probably more appropriate. Table 3 tabulates three sets of residuals from regressions on subsets of the significant explanatory variables. The results tell a reasonable story with a few exceptions. The first set of residuals are those from regressing only on formal sector labor productivity only. The complete re- ranking of size shows immediately the importance of compensating for productivity, or more generally level of development when drawing inferences from the size of the sector. Among the OECD countries, they tell a story that is broadly consistent with the literature and suggest that the approach is not entirely misguided. The U.S., Canada show up as relatively clean, while Spain, Greece and Italy appear heavily burdened. Among the Latin countries, Costa Rica, Honduras, Guatemala, Chile, Paraguay, Brazil are all below trend in their share of the work force in self- employment and hence, arguably, with less onerous legislation than the mean, 'while Colombia, Argentina, Uruguay, Venezuela and Peru far above. The second set of residuals add the structural variables measuring share of young people in the work force and the level of education, as well as the level of interest rates. The extremeness of this last variable and the difficulty of measuring it may distort the results some. Mexico is now firmly below trend and Chile, with relatively moderate interest rates, is actually above, although not by a significant amount. Peru now emerges as far less pathological that previously, but it still is joined by Argentina, and Venezuela as appearing to have a high level of unobserved distortions. The third set of residuals add to the regressors the burden of taxation for social security. Consistent with the previous findings, this does little to change the overall ranking although Mexico's sector share is now even more below trend and Colombia approaches trend. The final set of residuals attempts to incorporate the Marquez index. To take advantage of the more precise parameter estimates from the large sample, they are constructed by taking Resid 3 and regressing it on the measure of employment protection for the countries for which it was available. In general, Latin America shifts up in the rankings due to their overall higher firing costs. Now, Costa Rica, Bolivia, Mexico, Honduras, and El Salvador, and Paraguay appear to have fewer residual distortions than the U.S. or most OECD countries. The Latin Countries that still appear with residual distortions are Venezuela, Argentina, Peru and to a lesser degree Colombia and Chile. Again, this entire exercise is distorted to the degree that sector size is poorly measured, or that included variables are correlated with those excluded. Still, as a measure of segmentation, it is 12 probably better than conventional comparisons of wages among sectors, and suggests a fairly robust story. Numerous countries in the region do not appear to be unduly saddled by labor legislation relative to that elsewhere. These results are somewhat at odds with some previous work. Loayza (1996) built a model of informality that focuses largely on issues of taxes and regulation of business. He also generates standardize estimates of the size of the informal sector as a share of production, based on the VAT evasion rate. Chile, Argentina, and Costa Rica to have the lowest adjusted size and Peru, Panama, and Bolivia to have the largest. Given the possible divergence of VAT evasion rates from the relative share of employment in self-employment, these differences are, perhaps, not surprising. The results are also not entirely consistent with the indexes of distortion calculated in The Long March, a regional reform perspective published by the World Bank, which showed Peru, Chile and Colombia among the most liberalized and Mexico, Bolivia and perhaps Brazil among the least. This divergence, again, may be due to data problems-Chile may count its self -employed more conscientiously. But it also may be the case that enforcement varies greatly, and that a formally rigid market may, in practice, be quite fluid. 3.3 Results: Turnover The turnover results are more difficult to interpret first, because of the fewer degrees of freedom available, and second because the theory is far less clear about what the signs should be. Nonetheless, the results are provocative.'" Mean Tenure Columns 1 a to 1 c in table 4 present various specification of average tenure in manufacturing employment. Column a includes all variables in the specification, again with the exclusion of the protection variable. Neither the LAC, Youth, Unemployment Benefits enter significantly. Labor productivity enters both in levels and with its square suggesting a non-linear relationship. Taken at the mean, labor productivity appears to have a negative impact on tenure. This can be reversed with the exclusion of all other variables, but the inclusion of the share of the population with secondary schooling reverses its sign. Thus, although the OECD countries have more stable work forces, it appears that it is the fact that they are educated, rather than rich that drives the result. Both taxes on social security appear to increase tenure and real interest rates decrease it. Again, these results are consistent with the theoretical framework. The reduction in sample to 17 observations with the inclusion of the employment protection variable leaves only the productivity, social security tax on employers and protection variables 16 See Mdrquez and Pag6s (1998) for a graphical treatment of these issues. The econometric results here are broadly consistent with their findings. 13 significant. The latter enters with predicted sign, suggesting that it does negatively affect turnover and importantly. The difference across the range from the U.S.(1) to Mexico to Venezuela (37) accounts for 3.24 years on a mean of about 9. Table 5 presents a set of residuals for the average job tenure regression analogous to those previous. With the usual caveats about small sample size and the possible correlation of ommitted variables with those included, Venezuela, Bolivia, Brazil, Bolivia appear with greater than average turnover when adjusted for productivity while Paraguay, Argentina, Honduras and Panama are below suggesting less. When adjusted for education levels and interest rates, Brazil shows closer to mean turnover and Panama now is above. Adding employers' social security contributions has the effect of bringing Honduras up above mean turnover and moving Panama below. The share of the manufacturing workforce with under two years of seniority. The results for the second measure of turnover, the share of the manufacturing work force with under 2 years of tenure are broadly consistent but suggest the sensitivity of the results when sample sizes are so small. Labor productivity, the real interest rate, the employers contribution to social security, and education variables enter significantly and with signs consistent with the previous results. As column 2a suggests, the education variable proved very unstable with the youth variable included. However, because the youth variable entered with the sign opposite to that expected, the preferred regression was that presented in 2b. 2c suggests that in these regressions, the Marquez protection variable does not enter significantly. The residuals are not presented in tabular form for this regression. 3. The Overall Picture Figures 5a and 5b plot the residuals from the second set of residuals (productivity, education, real interest rate, youth) from the self-employment regression and from the turnover regressions to see if the combination of the two can reveal anything about the functioning of a given labor market. Though speculative, we will interpret these residuals as measuring labor market distortion (self - employment above the conditional mean) and rigidity (average tenure above the conditional mean). In the North East quadrant of figure 5a, rigid and distorted economies, we find Greece, Italy and Spain, all European countries renowned for repressive labor codes. It is, however, interesting to note that only one Latin country appears in this quadrant, perhaps unsurprisingly Argentina, but in figure 5b it will show higher than average turnover. In both graphs, Honduras, Paraguay, also appear somewhat rigid although less distorted than average. Among most flexible and undistorted in the South West quadrant we find, unsurprisingly, the U.S. and Canada, accompanied by Panama, Bolivia, Brazil. Venezuela, Peru. There is some sensitivity of the results to which measure of turnover is used. Using the share of workers with under 2 years of tenure, Figure 5b suggests that Argentina has above average flexiblity and Brazil, below average. The graphs were redone replacing industrial value added per worker with purchasing power parity adjusted per capita GNP. The placement of countries changed little suggesting that the results are not very sensitive to the exact measure of labor productivity used. In sum, with some exceptions and with strong caveats about the 14 reliability of the results, Latin labor markets do not appear exceptionally distorted, or inflexible. 4. Conclusion This paper has presented the results of cross country regressions motivated by an efficiency wage model of the LDC economy. The model departs from the assumption that for many workers, informal self-employment is a desirable destination for salaried workers rather than the disadvantage sector of a market made dual by union or government induced wage rigidities. Nonetheless, segmentation will always be present in all labor markets, even in the absence of unions or minimum wages so long as firms seek to retain workers in whom they have invested. Firms will pay above market clearing "efficiency wages" to lower turnover and in the process, create unemployment or segmentation that may cut across lines of formality. The predictions about the size of the self employed sector and turnover were tested using cross sectional OECD and Latin American data and generally supported. The size of the informal sector can tell something about the impact of labor legislation on efficiency or distribution, but only if adjusted for demographic and other variables that theory suggests are important. Raw measures of sector size are not a reliable measure of distortions or inequality. The adjusted size of the informal sector suggests that, Venezuela, Peru, Argentina, and perhaps Colombia emerge as having relatively high levels of distortion by global standards, while Brazil, Costa Rica, Panama, Bolivia, Mexico, Guatemala, Honduras, and Paraguay have lower than average distortions. Such crude comparisons of adjusted sector sizes may offer an alternative to meaningless comparisons of earnings across formal and informal sectors. The divergence of these measures from those created from formal labor legislation in The Long March may suggest differences in enforcement, or the ability of the market to work around oppressive regulations. Comparisons of raw turnover rates across countries tell us little about true labor market flexibility. Theory and preliminary empirical work suggests that many variables affect turnover in significant ways. Once these are considered, Latin American labor markets appear of average flexibility: Paraguay, Honduras, and perhaps Argentina and Brazil appear more rigid than average while Bolivia, Panama, and perhaps Peru are less. Barriers to firing workers, and social security taxes on firms appear to reduce the size of the formal sector. Anticipation of costly firing may lead to a reluctance to employ new workers while high non-wage benefits raise labor costs. That said, empirically, the level of formal sector productivity, real interest rates, and education levels in general have a larger impact on the size of the informal sector than labor market taxes or barriers to firing. 15 References Abuhadba, Mario, Romaguera Pilar (1993), "Inter-industrial Wage Differentials: Evidence from Latin American Countries" Journal-of-Development-Studies. Andersen Schaffner, Julie (1997), "Job Stability in Developing and Developed Countries: Evidence from Colombia and the United States." Mimeo, Stanford University. Aroca, P. and W. F. Maloney (1998), "Logit Analysis in a Rotating Panel Context and an Application to Self-Employment Decisions," mimeo The World Bank and Universidad Catolica del Norte, Chile. Begin, James P. (1995) Singapore's Industrial Relations System in Stephen Frenkel and Jeffrey Harrod Eds. Industrialization and Labor Relations Contemporary Research in Seven countries. ILR Press, Ithaca. Bell, L (1997) "The Impact of Minimum Wages in Mexico and Colombia" Journal of labor Economics. Bentolila, Samuel and Giuseppe Bertola (1990) Firing Costs and Labour Demand: How Bad is Eurosclerosis?" Review of Economic Studies, 57:381-402. Birdsall, Nancy and Frederick Jaspersen (1997) EDS. Pathways to Growth, Interamerican Development Bank, Johns Hopkins. Blanchard, 0. and J. Jimeno "Structural Unemployment: Spain vs. Portugal" American Economic Review 85:2 p. 212-218. Blanchard, Olivier and Lawrence, F. Katz, "What we Know and do Not Know about the Natural Rate of Unemployment," Journal ofEconomic Perspectives, 11:1. Blanchard, Olivier and Juan F. Jimeno, "Structural Unemployment: Spain vs. Portugal," American Economic Review, 85:212-217. Boyer (1994) "Do Labour Institutions Matter for Economic Development" in Workers, Institutions, and Economic Growth in Asia. Gerry Rodgers Ed. ILO Geneva. Cox- Edwards, Alejandra (1993), Labor Market Legislation in Latin American and the Caribbean, Report 31, Latin America and the Caribbean Technical Department, the World Bank. Fields, Gary (1994), "Changing labor Market Conditions and Economic Development in Hong Kong, the Republic of Korea, Singapore, and Taiwan, China, World Bank Economic Review 8:3. 16 Funkhouser, E. (1998) "The Importance of Firm Wage Differentials in Explaining Hourly Earnings Variation in the Large Scale Sector of Guatemala" Journal ofDevelopment Economics 55(1), 115- 131. Gonzaga, Gustavo (1996), "The Effects of Openness on Industrial Employment in Brazil" mimeo, Departmento de Economia, PUC-RIO, Brazil. Gonzalez, Jose Antonio Labor Market Flexibility in 13 Latin American Countries and the United States: Stylized Facts about Structural Relationships Between Output and Employment- Unemployment-Wages" Mimeo, The World Bank. Harris, J.R. and M.P. Todaro (1970) "Migration, Unemployment and Development: A Two Sector Analysis,"American Economic Review, 60:1, 126-142. Hernandez Laos, (1995). "Costo Laboral y Competitividad Manufacturera en Mexico (1984-1993)" mimeo. Hopenhayn, Hugo and Richard Rogerson (1993), "Job Turnover and Policy Evaluation: A General Equilibrium Analysis." Journal ofPolitical Economy, 101:5:915-938. Jovanovic, Boyan (1982), "Selection and Evolution of Industry," Econometrica, 1649-670. Krueger, A.B and L. H. Summers (1988), " Efficiency Wages and the Inter-Industry Wage Structure," Econometrica 56:2 p 259-293. Krebs, T. and W. F. Maloney (1998), "Quitting and Labor Turnover: Microeconomic Evidence and Macroeconomic Consequences" mimeo, World Bank, Washington, D.C. Levenson, Alec and William F. Maloney (1997), "The Informal Sector, Firm Dynamics and Institutional Participation," Mimeo, University of Illinois IBRD and Policy Research Working Paper No.1988. Levine, Marvin (1997) Worker Rights and Labor Standards in Asia's Four New Tigers, Plenum Press, New York. Loayza, N.(1996), The Economics of the Informal Sector: A Simple Model and Some Empirical Evidence from Latin America." Carnegie-Rochester Conference Series on Public Policy 45:129-162, North Holland. MacIsaac, Donna and Martin Rama(1 997), "Determinants of Hourly Earnings in Ecuador: The Role of Labor Market Regulations" Journal ofLabor Economics; 15(3), Part 2 July 1997, pages S 136-65. 17 Mdrquez, Gustavo(1 990), "Wage Differentials and Labor market Equilibrium in Venezuela, Unpublished Ph.D. Dissertation, Boston University. Mdrquez, Gustavo(1998), Protecci6n al Empleo y Funcionamiento del Mercado de Trabajo: una Aproximaci6n Comparativa. Uncitable mimeo, Chief Economist Office, IDB. Mdrquez, Gustavo and Carmen Pagds (1998), "Ties that Bind: Employment Protection and Labor Market Outcomes in Latin America." IDB mimeo. Maloney, William F. (1997a) "Are LDC Labor Markets Dualistic?" mimeo, the World Bank. Maloney, W.F. and E. Ribeiro (1998), "Efficiency Wage and Union Effects in labor Demand and Wage Structure in Mexico" Mimeo, The World Bank and Universidade Federal do Rio Grande do Sul, Brazil. Maloney, William F. (1997b) "Labor Market Structure in LDC's: Time Series Evidence on Competing Views," mimeo, the World Bank. Maloney, William F. (1995), The Informal Sector in Mexico: A Dynamic Approach. mimeo, The World Bank. Marquez, Carlos and Jaime Ros (1990), Segmentacion del Mercado de Trabajo y Desarrollo Economico en Mexico, El Trimestre Economico, Fondo de Cultural Economica, Mexico, LVII:2 Nickell, S. (1997), "Unemployment and Market Rigidities: Europe versus North America," Journal ofEconomic Perspectives, 11:3 pp. 5-74. Revenga, Ana and William F. Maloney (1994) "Mexico Labor Market Strategy Paper" The World Bank. Schaffner, J.A.(1998) "Premiums to Employment in Larger Establishments: evidence from Peru. Journal ofDevelopment Economics 55(1), 81-113. Shapiro, Carl and Joseph Stiglitz (1984), Equilibrium Unemployment as a Worker Discipline Device" American Economic Review, 74:3 431-444. Stiglitz, J.E.(1974) "Alternative Theories of Wage Determination and Unemployment in LDC's: The Labor Turnover Model, " Quarterly Journal ofEconomics, 88:194-227. 18 World Bank (1990,95), World Development Report. You, Jong Il, (1994) "Labour institution and Economic Development in the Republic of Korea" in Workers, Institutions, Economic Growth in Asia. Gerry Rodgers Ed. ILO Geneva. Appendix I: Data Sources 1. Tenure variables of the OECD countries are from the following sources: Table 2: Measures of the Sluggishness of Employment and of Adjustment Costs, Page 11, in Stephen Nickell, "Labour Market Dynamics in OECD Countries", Centre for Economic Performance, Discussion Paper #255, August 1995 Table 5.5: Distribution of Employment by Employer Tenure, 1995, Page 138, Table 5.6: Average Employer Tenure by gender, Age, Industry, Occupation, 1995, Page 139, in "OECD Employment Outlook, July 1997", 2. Self-employment rate of the OECD countries is from "OECD Labour Force Statistics 1976-1996." 3. Self-employment rate and tenure variables of the Latin America and the Caribbean countries are from various CEPAL surveys of the following years: Argentina 1992,Bolivia 1995,Chile 1995, Colombia 1995, Costa Rica 1995, El Salvador 1995, Guatemala 1989, Honduras 1995, Mexico 1994, Panama 1995, Paraguay 1995, Peru 1996, Uruguay 1995, Venezuela 1995. 4. Per capita GNP and wages of Industrial workers are from Table 1: Basic Indicators, Page 214,215 and Table 12: Structure of the Economy: Production in "World Development Report 1997", 5. Employment Protection variables provided by Gustavo Marquez, Chief Economist Office, IDB. 6. Social Security variables are from Table3: Contribution rates for social security programs -OECD countries(1997), and Table 5.8: Social security and non-wage labour costs, in "Social Security Programs Throughout the World - 1997". 19 Figure 1: Equilibrium in Formal Sector Formal Wage D I w > II Figure 2: Formal Sector Probability of Being Hired Productivity Gain (Z) \D' Formal Wage D W w/ I D Figure 3: Rise in Informal P P' Attractiveness (T) Probability of Being Hired Formal I Wage D . D P' P Probability of Being Hired 20 1 I I .461 Peru E o Bol EEI wEI COl Van Hon Gua Par Uru U, Arg Gre Pan Chi Meu a) ~CR Ch Bra Kor Ita 0 LL L- Por O Spa New Ice O- Ire ael o Ad.~ Ca Pm t A Den .0576 Nx II I 6.63332 10.5322 Log of industrial V.A./ Worker Figure 4: Self-Employment vs. Industrial Productivity 21 - .1Von 3r1- E Ita 1 - 0 Arg a .05 - Spa EBe 0) UK Swit Lux c 0 Ger u u Bol Pan AustUS A4qDINe th SweFr mAusi us A0 hFrm . Den Q&n Jap U) -.05- .1 Br -4 -2 0 2 4 Adjusted Mean Tenure Figure 5a: Distortion and Rigidity? I I I I .1 Von c a) E t . 05 Arg 2L Spa E P*ru- wL Fin Bel Swit UK 0) 0- Par Ger I0th Jap Hon .cPon n Bo Fra US Can Nor Ire -.05 Bra -20 -10 0 10 20 Adjusted Share < 2 Years Job Tenure [negative] Figure 5b: Distortion and Rigidity? 22 Table 1: Size of Informal Self-Employment and Turnover Rates LAC OECD % Workforce in Informal Self-Employment 31.5 12.9 % <2 Years Seniority (Manufactures) 38.1 24.5 Average Tenure (Manufactures) 7.61 10.5 23 Table 2: Determinants of Self -Employment Self- Employed as Share of Work Force 1-a I-b 1-c 1-d 1-e C 0.26 0.36 0.11 0.25 0.50 (1.33) (2.46) (.36) (1.46) (3.29) Indust. V.A. -0.02 -0.03 0.00 -0.01 -0.05 (1.26) (2.24) (.02) (.25) (3.19) SS Worker -0.09 0.02 (.89) (.14) SS Emp 0.16 0.16 0.09 0.00 0.15 (2.81) (2.81) (.95) (3.69) (1.63) Protection 3.90E-03 4.50E-03 4.OOE-03 (2.58) (3.69) (2.93) U Benefits 4.OOE-04 -1.20E-04 (.72) (.01) Real Interest 0.22 0.23 0.23 0.24 0.21 (3.91) (4.71) (3.29) (4.13) (3.21) Secondary -1.OOE-03 -1.OOE-03 -1.50E-03 -1.60E-03 (1.91) (2.37) (1.53) (2.4) Youth 3.86 4.14 1.79 (2.33) (2.8) (.63) LAC 0.04 0.02 (1.14) (32) NOBS 36 40 19 20 201 R2 0.92 0.90 0.94 0.93 0.90 Note: t-statistics below coefficients. 24 Table 3: Residuals of Self- Employment Regressions (Ranked by Deviation from Predicted Share) Country % Self-Emp Resid 1 Country Resid 2 Country Resid 3 Country Resid 4 Costa Rica 24.4 -7.6 Costa Rica -7.9 Brazil -8.9 Brazil -7.2 Honduras 31.8 -6.8 France -7.5 France -5.5 Costa Rica -6.4 Austria 6.6 -5.9 Austria -7.2 Costa Rica -5.5 Bolivia -5.2 Panama 26.2 -5.1 Honduras -4.7 Austria -3.9 Mexico -4.8 US 7.3 -4.8 Sweden -4.5 Sweden -3.5 France -3.3 Canada 8.9 -4.6 US -3.6 Mexico -3.3 Honduras -3.1 Guatemala 32.3 -4.1 Brazil -3.0 Norway -2.9 Portugal -2.3 France 8.6 -4.1 Guatemala -2.7 US -2.9 Guatemala -1.1 Denmark 6.9 -3.6 Germany -2.5 Ireland -2.8 El Salvador -0.4 Ireland 13.5 -3.3 Portugal -2.5 Bolivia -2.6 Ireland -0.4 Chile 25.2 -3.2 Finland -2.5 Canada -2.2 Paraguay -0.2 Paraguay 31.7 -3.2 Canada -2.4 Portugal -2.1 Denmark 0 7 Germany 8.5 -2.8 Paraguay -2.3 Netherlands -1.2 Spain 0.9 Netherlands 9.6 -2.7 Panama -1.9 Japan -1.1 US 1.0 Norway 5.9 -2.7 Ireland -1.9 Guatemala -1.0 Germany 1.0 Luxembourg 5.8 -2.3 Norway -1.6 Honduras -1.0 Chile 1.1 Sweden 9.3 -1.9 Bolivia -1.4 Panama -1.0 Netherlands 1.5 Brazil 23.2 -1.8 Mexico -1.3 Germany -0.5 Colombia 1.7 Finland 9.7 -1.8 Belgium -1.1 Paraguay -0.3 Peru 2.4 Australia 12.3 -1.4 Chile -1.1 Denmark 0.0 Italy 2.6 Bolivia 39.2 -1.2 Netherlands -1.0 Korea 0.3 Argentina 2.7 UK 12.6 -0.7 Luxembourg -0.9 Australia 0.3 UK 2.7 Portugal 19.5 -0.7 Spain -0.5 Uruguay 0.4 Belgium 3.5 New Zealand 16.2 -0.5 Korea -0.2 Luxembourg 0.7 Venezuela 6.1 Mexico 26.5 0.2 Turkey 0.4 El Salvador 0.8 Greece 6.6 El Salvador 35.2 0.2 Colombia 0.4 Belgium 1.5 Turkey 26.0 0.7 Denmark 0.5 Turkey 1.5 Belgium 13.3 0.8 UK 1.1 Chile 1.8 Japan 9.9 1.7 El Salvador 1.9 Spain 2.2 Colombia 34.2 2.3 Australia 2.7 Italy 2.2 Spain 18.6 2.4 Japan 2.8 Finland 2.2 Switzerland 10.6 2.5 Argentina 2.9 Iceland 2.3 Iceland 15.5 4.0 New Zealand 3.1 UK 2.3 Korea 23.6 4.3 Italy 3.2 New Zealand 2.9 Argentina 28.5 5.9 Iceland 4.7 Argentina 3.1 Uruguay 32.8 6.8 Switzerland 5.0 Switzerland 3.1 Venezuela 34.4 8.4 Uruguay 5.3 Colombia 3.9 Greece 27.7 8.4 Greece 7.2 Peru 4.3 Italy 22.8 9.4 Venezuela 8.2 Greece 7.3 Peru 46.1 18.6 Peru 17.2 Venezuela 8.9, Notes: Res I from regression on formal productivity. Res 2 from regression that also includes employers social security contribution. Res 3 includes also education, youth, and real interest rates. Res 4 are residuals of Res 3 on Job Protection. 25 Table 4: Determinants of Turnover Manufactures Mean Tenure %< 2 Years 1-a 1-b 1-c 2-a 2-b 2-c C 60.58 74.35 43.30 -288.38 -288.67 -272.71 (3.09) .(4.23) (6.03) (3.82) (4.78) (2.23) Indust. V.A. -11.95 -16.21 -3.19 84.40 77.61 73.37 (2.35) (3.99) (4.88) (4.17) (4.02) (2.47) I.V.A. sq 0.64 0.92 -4.84 -4.31 -3.98 (2.07) (3.99) (3.84) (3.89) (2.12) SS Worker 8.13 8.68 -2.51 (3.29) (3.62) (.16) SS Emp 7.00 6.23 11.65 -46.92 -44.45 -41.84 (1.79) (1.79) (5.86) -(4.63) (3.92) (2.54) Protection 0.09 -0.05 (2.99) (.2 1) U Benefits -0.03 -0.13 -(1.13) -(1.42) Real Interest -7.41 -7.10 14.45 21.44 24.14 (1.92) (1.89) (1.65) (2.22) (1.98) Secondary 0.05 0.04 -0.07 -0.24 -0.31 (2.15) (2.23) (.69) (2.57) (2.11) Youth -39.23 -754.73 (.48) (2.38) LAC -0.70 10.13 (.41) (1.33) NOBS 25 26 171 23 24 16 R2 0.63 0.68 0.811 0.50 0.65 0.54 Note: t-statistics below coefficient. 26 Table 5: Residuals of Average Job Tenure in Manufacturing Regressions (Ranked by Deviations from Predicted Tenure) Country vg. Tenure Resid 1 Country Resid 2 Country Resid 3 Denmark 7.80 -3.35 Denmark -3.76 Bolivia -1.84 Australia 7.00 -3.08 Australia -2.69 Venezuela -1.56 Venezuela 5.77 -1.87 Ireland -1.78 Austria -1.49 Bolivia 6.22 -1.84 Bolivia -1.69 Denmark -1.47 Brazil 6.25 -1.50 US -1.38 Netherlands -0.95 Switzerland 10.60 -1.43 UK -1.11 US -0.81 US 9.20 -1.40 Venezuela -1.10 Ireland -0.79 Canada 8.90 -1.25 Canada -0.98 Italy -0.54 UK 9.00 -1.20 Switzerland -0.65 Australia -0.46 Ireland 8.30 -0.93 Austria -0.36 UK -0.45 Netherlands 10.30 -0.21 Panama -0.29 France -0.43 Germany 10.80 -0.09 Brazil -0.24 Germany -0.27 Austria 10.60 0.14 Germany -0.03 Spain -0.26 Greece 9.00 0.34 Netherlands 0.03 Honduras -0.16 Panama 7.84 0.45 Honduras 0.21 Switzerland 0.16 Honduras 8.39 0.56 Greece 0.39 Sweden 0.17 Sweden 11.50 0.60 Spain 0.40 Canada 0.29 Argentina 8.89 0.83 Finland 0.75 Argentina 0.30 Italy 11.20 1.03 Sweden 0.75 Brazil 0.36 Japan 13.10 1.09 Argentina 0.97 Belgium 0.39 Belgium 11.80 1.34 Belgium 1.15 Greece 0.47 Finland 12.30 1.53 France 1.32 Finland 0.69 Spain 10.90 1.53 Japan 1.63 Panama 1.00 France 12.10 1.71 Italy 1.68 Japan 1.45 Portugal 10.40 1.92 Portugal 1.80 Portugal 1.55 Paraguay 9.93 2.44 Luxembourg 2.43 Luxembourg 2.22 Luxem bourg 14.70 2.63 Paraguay 2.54 Paraguay 2.44 Notes: Resid I from regression on formal sector productivity. Resid 2 from regression that also includes real interest rate, and education. Resid 3 includes also employers' social security contribution 27 CONFIDENTIAL CONFIDENTIAL Repovt No.: 19438 ME Repor No.: 19438 ME Type: SR Type: SR