发展中国家的劳动力市场是二元的吗？

WPS jq4lp POLICY RESEARCH WORKING PAPER 1941 Are Labor Markets There is little evidence to support the traditional in Developing Countries dualistic view of a labor Dualistic? market segmented between Duallstlc H ~~~~~~~~~~~~~~~formal and informal sectors as the principal paradigm William F. Maloney through which to view the informal sector. The division between good jobs and bad jobs seems to cut across issues of formality - and for many workers, inefficient labor codes and low levels of human capital may make employment in the informal sector more desirable. The World Bank Latin America and the Caribbean Region Poverty Reduction and Economic Management Unit June 1998  | POLICY RESEARCH WORKING PAPER 1941 Summary findings There is a long tradition of viewing as disadvantaged the finds little evidence to support the traditional dualistic roughly 40 percent of workers in developing countries view. who are unprotected by labor legislation and work in He shows that traditional earning differentials cannot small "informal" firms. prove or disprove segmentation in developing countries, Maloney offers an alternative to traditional views of and patterns of worker mobility do not suggest a rigid the relationship between formal and informal labor labor market - or one segmented into formal and markets: For many workers, inefficiencies in present informal divisions. labor codes and relatively low levels of human capital It is possible that the market is dualistic in the sense (labor productivity) may make employment in the used in the industrial world, but the division between informal sector more desirable. good jobs and bad jobs seems to cut across issues of He offers the first study of worker transitions among formality. sectors, using detailed panel data from Mexico, and This paper - a product of the Poverty and Economic Management Unit, Latin Arnerica and the Caribbean Region- is part of a larger effort in the region to reexamine the role of the informal sector. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Marta Cervantes, room 18-095, telephone 202- 473-7794, fax 202-522-0054, Internet address mcervantes@worldbank.org. The author may be contacted at wmaloney@worldbank.org. June 1998. (34 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center  Are Labor Markets in Developing Countries Dualistic? William F. Maloney* Keywords: Dualism, informal sector, Mexico, segmentation, transitions, wage differentials. * This work was undertaken while I was on leave from the World Bank, 1994-95. I am grateful to Enrique Davila Capalleja, Aslan Cohen, Rene Cortazar, Bill Dickens, Hadi Esfahani, Ariel Fiszbein, Alec Levenson, Frank Lysy, Doug Marcouillier, Gustavo Marquez, David Nielson, Ana Revenga, Eric Rice, Bill Savadoff, and Guilherme Sedlacek for helpful discussions. I also thank Roberto Flores Lima and Agustin Ibarra Almada of the Mexican Secretariat of Labor and Social Welfare for their advice and help with the project. Finally, I thank the Mexican National Institute of Statistics, Geography, and Information (INEGI) for the use of the data. INEGI is not responsible for any incorrect manipulation of the data or erroneous conclusions drawn from it. This paper was formerly titled "Dualism and the Unprotected or Informal Labor Market in Mexico: A Dynamic Approach."  L Introduction A long tradition views the roughly 40% of LDC labor forces that are unprotected by labor legislation and working in small "informal" firms to be the disadvantaged segment of a dual labor market.' The origins and dynamics of this sector have attracted renewed interest for at least two reasons. First, increasing labor market efficiency and flexibility is considered an essential complement to the market-based reforms underway throughout the developing world.2 To the degree that segmentation is driven by government or union imposed regulations that induce rigidities and push labor costs above market clearing, the large size of the informal sector stands as a measure of the magnitude of required reforms. Second, a related literature with a very different emphasis sees informality as the result of an ongoing effort by large modern enterprises to evade mandated protections through subcontracting to unprotected workers, a process accelerated by heightened global competition in labor intensive manufactures. The existence and-behavior of the sector are thus directly relevant to the debate over establishing common labor standards throughout free trade agreements: whatever safeguards may be enshrined in labor codes, LDC-based firms could employ an effectively unprotected work force to 'See, for example, Harris and Todaro (1970), Sabot (1977), Mazumdar(1983) or Fields (1990). A recent World Bank document argues that "Protected workcers in the 'modem' or 'formal' sector ...enjoy high wages, social security, vacation, pension and employment security as mandated by legislation. By contrast, those unable to find work in such firms resort to the next best alternative, the so-called 'informal sector' in small firms or self-employment, engaged in labor-intensive activities, withoutjob security or benefits."(Ozorio de Almeida et al. 1995, p. 1). See Fields (1990), Tokman (1992), Portes (1994), Rosenzweig (1988), Thomas (1992) for excellent overviews of the informality literature. See Stiglitz (1974), Esfahani and Salehi-Isfahani (1989) for efficiency wage models of LDC dualism. 2 See, for example, World Bank (1995) World Development Report.  compete with U.S. firms. 3 This paper argues that the traditional conflation of issues of formality and dualism is probably inappropriate and offers an alternate view of the role of the unregulated small firm sector. As in the industrialized countries, many workers may choose to start or work in small firms and in the developing country context the incentives to do so may be greater. First, the inefficiencies and rigidities that often accompany labor protections, and the implicit taxation they imply in the absence of a binding minimum wage, may serve to reduce the attractiveness of formal sector employment. More generally, the low level of formal sector productivity for the mass of poorly educated workers in developing countries reduces the opportunity cost of being independently employed. Much of the informal sector may therefore reflect an efficient allocation of labor. Despite the long tradition and a voluminous literature, support for the dualistic view is not strong and has rested largely on case studies and on comparisons of earnings across sectors adjusted for observable worker characteristics.4 This paper employs detailed panel data from Mexico to take a more comprehensive approach than has been previously feasible. It examines the earnings differentials and mobility patterns of individual workers transitioning among formal salaried employment and three modalities of informal work: The self-employed, owners of informal firms with or without additional employees; the infonnal salaried, those working for these informal firms and who are usually considered the least advantaged of the work force; and contract workers, those who do not receive a regular wage or salary, but who are paid as a percentage, by piece, on commission, 3 See Portes, Castells and Benton (1989), Tokman (1992).The U.S. Department of Labor has sponsored two studies on the informal sector. See Workers Without Protections: Case Studies of the Informal Sector in Developing Countries (1993) and The Informnl Sector in Mexico (1992). 4 See Rosenzweig (1988). 2  or fixed contract and are often connected to larger firms. The earnings differentials generated are more precise and reliable than those previously generated, but are shown.*t6 be fundamentally unable to prove or disprove segmentation in this context. As an alternative, the paper examines the patterns of worker mobility predicted by the dualism hypothesis, generating both a dynamic overview of movement through the labor market, and characterizations of the interactions among the four classes of work. While unable to provide conclusive evidence, this approach offers substantial reason to question the dualistic view as the primary explanation for the existence and dynamics ofthe sector. The labor market appears relatively fluid with large and symmetric flows of work:ers among all sectors. The data is consistent with self- employment being a desirable destination for many workers and with contract work being closely related. It also suggests that informal salaried work is the entry point and perhaps training area for young workers who, even if queuing, very quickly leave to take both formal and informal jobs. rr. Data: The study employs two sources of data on men aged 16-65 with a high school education or less in 16 major metropolitan areas. The National Urban Employment Survey (NUES) conducts extensive quarterly interviews and is structureid so as to generate panels that allow tracking a fifth of the samnple across five quarters. Workers are matched by household, role in household, sex, level of education, and age, to ensure against generating spurious transitions. Though five quarters does not permit a full description of the life cycle of an individual, it is nonetheless possible to sketch patterns of mobility among sectors and to identify worker characteristics that correspond to them. To generate a sufficiently large sample of roughly 15,000 observations, three contiguous NUES cohorts were 3  combined: 1990:3-1991:3, 1991:1-1992:1, 1991:2-1992:2.5 Another panel was created spanning 1991:1-1992:1 that terminates in the more detailed Micro-Enterprises Survey (MES). This survey was constructed by identifying 11,000 owners of micro-enterprises, defined as firms of fewer than six individuals, from the 1990:4 NUES and re- interviewing them in 1991: 1 in more depth about issues of capital structures, costs, and employment patterns. It also specifically asks why they left their previous job and why they started the present business which offer a partial alternative to naive corrections for selectivity bias.6 Thus, for a reduced sample, we have far more information on both earnings differentials, and motivations for moving. The panels in the combined NUES sample above were chosen to include and span either side of the MES panel. Two popular definitions of informality are employed. The first focuses specifically on the issue of "protectedness" and comprises owners or workers in firms of fewer than 16 employees who ; The cohort beginning in the fourth quarter of 1990 was not used because it would incorporate the end of the year bonuses, the aguinaldo. This normally amounts to a 13th monthly payment but may vary by year and firm imparting an undetermined upward bias to the monthly wage reports. Since we are concerned with expected differentials in income, the wages of workers who reported nonnally receiving the aguinaldo were increased by 13/12. 6Heckman (1979) is insistent that the standard two stage methods of correctingfor selectivity bias depend on having confidence in the underlying model of how workers choose among sectors. A bad first stage selection specification may induce bias rather than correct for that existing. This is likely to be the case as the premise of this work is that we have little knowledge of the role each sector serves. Further, since a principle argument of this work is that, in the absence of knowledge of the value of unobservable components ofthe differential, we cannot use the differentials as evidence of segmentation, the issue becomes of somewhat less importance. The standard Heckman procedure was employed, however, for transitions in and out of formal salaried work using the variables employed in the logit analysis in section IV for the probit first stage. The model for Formal Salared to Self-Employment did not converge and yielded improbable results. For Self-Employment to Formal Salaried, Formal Salaried -Informal Salaried , Formal Salaried to Contract, Contract-Formal Salaried, it could not be rejected that the transformed correlation coefficient between the errors in the selection and wage equation was zero and hence the correction for selectivity bias was inappropriate. Only for the Informal Salaried to Formal Salaried regression was the transformed correlation coefficient significant. The correction changed the sign ofthe differential and greatly changed the magnitude. Again, however, in the absence of a model of what is being selected for, it is unclear how to interpret the restat. 4  do not have social security or medical benefits. The second addresses the issue of the role of the small firm or "micro-enterprise" using the Mexican government definition of firms of under six workers. Because only the latter definition is consistent with the sampling of the MES, it is the one used for the joint NITES/MES panel. However, the similarities between the two sets of descriptive statistics (table 1) on age, work experience, level of schooling and an index of the real wage at the beginning of the period for each transition, confirm that there is great overlap in the composition of the two populations and that the results are unlikely to be driven by the particular definition chosen. Though the differences in sample means are sometimes statistically significant, they are never large. This is due partly to the fact that among entrepreneurs of firms with less than six workers, only 5% are covered by benefits and that the vast majority of micro-firms are concentrated at sizes of below three workers. IL Wage Differentials as Evidence of Segmentation Traditional efforts to identify segmenlation by comparing conditional means between sectors are unsatisfying for two reasons. First, as is well-documented in the literature, unobserved worker characteristics that affect productivity-- ability to tolerate authority, punctuality, entrepreneurial ability- may also influence in which sector an individual chooses to work and bias estimates of the sector differential.' But a greater concern is that the specific characteristics of work that pertain to or even define 7 Recognizing this, Marcouiller et al. atlempt to mitigate the selection bias but are clear about the limitations of the standard techniques when there is no clear consensus about how workers sort themselves between the informal and formal sector. They, in the end, report higher earnings in the informal sector. 5  the formal and informal sectors affect the earnings paid in each sector and make it unclear what the magnitude or sign of the differential should be even in an unsegmented market. Informal earnings should exceed formal sector wages by the expected value of unreceived benefits, and fall below them by the taxation that is often evaded. Earnings both in contract and self-employment may reflect a premium for risk, lifestyle, and in the latter case, the implicit costs of capital invested, and the value of unpaid work by family members that the MES suggests comprise 34% of micro-firm employees. Informal salaried workers are among the youngest (table 1) and the MES reports that roughly 30% are related to their employer. Their reported earnings may therefore incorporate training costs (see Hemmer and Mannel, 1989 and Roberts, 1989) or unobserved payments in kind. i. Estimates of earnings differentials Using the vast but undetailed NTJES, the first columns of tables 2a and 2b show the percentage change in hourly real earnings generated by movements of individuals among sectors. This holds worker characteristics constant and leaves the variations in the characteristics ofthe work itself as the residual explanatory factor. The next columns account for these characteristics more than has been possible previously and present changes in real wages net of taxes,' and real wages net of taxes per hour among the four sectors.9 For all three- transitions into self-employment, the s Payments were calculated based on the Mexican tax tables. It was assumed that all informal workers avoided paying taxes and all formal workers paid. 9The non-normality of the distributions makes the sample mean an inadequate measure of the central tendency of the data and its significance. Two alternate measures are presented in each column: the mean as determined by a robust estimation technique, and the median. The median is largely unaffected by the non- normality of the distribution and is robust to outliers. The robust mean attempts to recover the information in the tails while compensating to some extent for outliers and non-normality. All calculations are done in STATA. 6  NUES/MES panel allows further adjustment for imputed return on the value of capital (tools, inventories, and location if owned), and hours worked by unpaid workers (table 3).10 The results would appear to invert the conventional view of the relation between formal and informal sectors, regardless of the definition chosen: movement into formal salaried employment is associated with a significant decline in remuneration, except from informal salaried employment; movement from formal salaried employment always leads to a significant increase. However, in the absence of information on the value of benefits, compensation for risk, the value of independence, or in-kind payments or implicit training costs, the magnitude of the distortion-free differential cannot be known ex-ante, and neither these, nor any previous reported sectoral differentials are reliable measures of segmentation. Further, though table 3 suggests that those reporting voluntary moves into self-employment do far better than those moving involuntarily, we cannot necessarily interpret asymmetries in differentials in tables 2a and b as evidence that there is a larger component of voluntary movement into the formal salaried sector: Involuntary separations due to formal sector firm closures imply the loss of well paying jobs and a large differential, while the analogous failure of a micro-enterprise may imply low earnings prior to transition and a smaller differential. However, the differentials among informal sectors are less affected by unobservables, such as the loss of formal sector benefits, and are suggestive of similarities and differences among them. 10 Despite choosing the NUES sample to straddle the MES sample, predictably, the raw differentials differ somewhat. For the critical salaried formal salaried/self-employed transition they are extremely close. For the other cases, although we are dealing with the same relative orders of magnitude, it is less easy to consider these simply as adjustments to the NiJES results reported above. The formal/self-employed differential is slightly, although not significantly higher (11,5 vs 9,4) and the contract/self-employed differential is also broadly similar (10,-i vs 7, 8) with a substtial but insignificant difference in the median. The differential from informal/ salaried employment is substantially higher (33,25 vs 19,16) and significant different. 7  Tables 2a and b show that movement into self-employment is always associated with a substantial and significant rise in per hour after-tax remuneration from every other sector, reaching over 25% from the fornal and informal salaried sectors. However, the joint NUES/MES panel (table 3) shows that after modest adjustments for capital costs, and much larger ones for unpaid labor,"' both contract work and self-employment share simrilar premia with respect to formal salaried and informal salaried work, and the differential between them is insignificant. Contract work appears similar to self- employment and may therefore share the same composition of voluntary and involuntary entrants. Salaried informal employment, however, suffers a discount relative to all other sectors. After adjusting the differentials for standard firm size effects observed in the U.S.,"2 moving to an informal sector firm from a comparably-sized formal sector firm now yields a 12-15% rise in the "micro- enterprise" sample, comparable to those above, but only a slight rise in the "unprotected" sample. Further, the large asymmetries in both cases suggest that informal salaried workers gain far more entering salaried formal work than they lose leaving, as would be predicted if they were queuing for 11 A return of 10% was imputed. Most micro-enterprises who save in commercial accounts or cajas de ahorro received 3% real. Implicitly we are assuming 7% for depreciation. The low level of capital employed results in the overall differential being relatively insensitive to the cost of capital value chosen. The more detailed treatment of taxation in the MS allows dropping the previous assumption of complete avoidance by the self-employed and induces a slight moderation in the after-tax differential. The adjustment for unpaid labor may be overstated if Balan et al are correct that "Since in most cases these family members would not otherwise be employed outside the household, their contribution to family finances is a 'net' one..."(p.218) or if the unpaid labor was in training. 12 The wage/size elasticities were taken from Barron et al's (1987) U.S. study which estimates the impact net of unionization effects and capital-labor ratios in a society where labor laws are enforced across size. A simple regression of the log of the wage on the log of firm size yields coefficients on the order of .042 to .069 and very significant, roughly double the 2.8% found by Barron et al.. Behar (1988) and Ros and Marquez (1990) find that in Mexico, as others have found elsewhere, wages increase with firm size and urther, large firms often pay benefits substantially above those established by law. Because we cannot control for capital-labor ratios and unionization, and because this size effect may reflect the very fact that the infornal in small firms are rationed out of larger firms by excessive wage legislation, including a change in firm size variable in the differential regressions might obscure the very effect to be measured. 8  formal salaried jobs. On the other hand, this asymmetry also exists with the contract sector so it is not limited to entering formal employment. Further, should deductions for training costs be a substantial fraction of the wage, such an asym;metry would be expected since returning workers may not work as apprentices. More fundamentally, the discount with respect to the other informal sectors suggests that the composition of the premium in the absence of distortions is distinct for this sector. This might be the case, for example, if the premium for risk is lower, if the 30% of the informal salaried who are related to the owner receive a large fraction of their income in kind, or if training costs are substantial. ii. Are the differentials consistent with an integrated labor market? Though we cannot credibly prove or disprove segmentation based on these differentials, we can ask whether they seem plausibly consistent with an absence of segrnentation: If the responses of those reporting voluntary moves are reliable, is the 15-20% differential plausibly large enough to cover the value of benefits (which, on paper constitute between 31-60% of the wage), some return to risk, and whatever value is placed on one styl[e of work vs another? The breakdown by motivation in table 3 could, in theory, offer some measure of the value of the latter but the differential for "independence" does not vary significantly from that for "higher pay" and, depending on the question, implies contradictory signs on the premium."3 If'we interpret this ambiguity as implying a small effect, and the risk premium is positive, these leave the value of benefits even smaller-- a half or even a quarter of those on paper. Three factors make this plausible and suggest a reason why the attraction of formal sector 13 World Bank estimates. See also Davila-Capalleja(1996) 9  employment may have been overstated. First, since the medical benefits program in Mexico, as elsewhere, covers a worker's entire family, the marginal value of benefits to the second formal sector worker in a family is zero. Second, administrative overhead costs are high and the benefits may be of low value given their cost. In his interviews with Guadalajaran workers, Roberts (1989, p. 50) found that "many informants cited the deduction made for welfare as a disadvantage of formal employment, particularly since the services they received were poor." Third, rapid rates of turnover mean that leaving does not necessarily imply the loss of nominally very generous separation benefits and pensions since as Balan et. al (1973, p. 212) found in their extensive surveys of worker career trajectories in Monterrey, "many change enterprises quite often and thus they cannot benefit from the seniority accumulated in each of them." In each case, the value to workers of formal sector benefits is below their value on paper, and what, in the absence of a binding minimum wage, workers implicitly pay.'4 In sum, earning differentials do not offer compelling evidence in favor of the segmentation hypothesis in Mexico and, given the difficulty of quantifying the unobservables, are unlikely ever to be convincing tests. IV. Patterns of Mobility In contrast, the patterns of worker transitions can offer additional information on the validity of the dualistic view. Ideally, a model of the behavior for each of the four sectors and workers' choices to enter them would be postulated and held up against the evidence. However, this is a vast "Bell (1994) argues that the minimum wage has virtually no impact on the distribution of average wages reported by firms in Mexico and little evidence of non-compliance since few firms paid wages below the 10  research agenda in itself and the present paper seeks only to develop a few provocative stylized facts about the dynamics ofthe market and sketch some characterizations ofthe nature and the interaction of the sectors. To begin, the dualistic view predicts some: general patterns should be evident. If formal sector work is preferred to informal work, we would expect that workers would queue up for formal sector jobs and relinquish them only under the limited conditiontpermitted by the Constitution-- egregious conduct or "acts of god" that induce firm downsizing." This should imply: i. Very low rates of formal sector turnover. ii. A largely unidirectional flow of workers who graduate from the informal sector into the formal sector where they stay until retirement. Flows in the other direction should be largely involuntary and in relatively prosperous times, far less. At the time of this sample, the Mexican economy was growing and since 1989, unemployment had been at decadal lows of around 2.6%. iii. Given a probability of being selected from the queue in each time period, the probability of entry into formal salaried work should be an increasing function of experience. This section tests for these patterns in two ways: First, for each definition of informality, Matrices 1 and 2 provide summary data on transitions among sectors by tabulating the conditional probability of finding a worker in sector j at the end of the period (columns) given that the worker began in sector i at the beginning (rows), P1 .16 The row percentages sum to 100% and the totals at /~~~~ 15 The Constitution conceives of the employment relationship as a lifetime contract and workers may only be fired under extreme circumstances and at great cost 16 In a spirit similar to the work at hand, Sedlacek et. al.(1995) studied the mobility of Brazilian workers with and without signed working cards, and hence worker projections, and found little evidence of strong barriers to mobility. 11  the bottom represent the share of the workers to be found in each category at the end of the period P j. The first three columns and rows represent individuals who are not working: those out of the labor force (OLF), not currently working and not searching; those studying; and those looking for work, the unemployed. The bold rectangle borders five categories of work beginning with unpaid labor and the shaded area comprises the paid jobs described earlier which are our chief focus. Since in a random shuffling of workers, P1u would clearly increase with Pj the second panel of the matrix standardizes the transition probabilities by terminal sector size, Pi/Pj. Though this is a better measure of fluidity among sectors, it is an imperfect measure of ease of entry since a low desire to leave the initial sector will yield a low value as well as a distortion induced low level of turnover in the terminal sector: PY/PJ = (1 - Pd )Vr(1 - P.) Vc, tabulated in the third panel, captures the disposition or economic or institutional logic that compels a worker leaving the initial sector to enter an open position in j: e.g. though both third and fourth grade elementary school classes may fully turnover every year, we would expect V to be large for transitions in the ascending direction and zero in the reverse. In the present case, the disposition to enter paid employment from school is two to three times that of the reverse transitions as would be expected if workers generally graduate from school to employment. Ifthe dualism hypothesis that workers graduate from informal to formal employment is correct, we should expect similar asymmetric Vs between the sectors. Further characterization of the patterns of mobility is offered by a multinomial logit model of 12  sector choice that includes experience, experience squared, and schooling. Since these are often the factors included in Mincerian earnings equations, the initial period real wage is included to ensure that it is not simply the wage effect that is reflected, The coefficients in tables 4a and b are those from the standard exponential form where the vector fj measures the degree to which an increase in worker characteristic X increases the probability of a worker going to sector j relative to the probability of staying in sector i. In the second half of each panel are presented the dummy and interactive effects of involuntary separation from the previous job, tabulated in the relatively infrequent case of a spell of unemployment between jobs. In only four cases does the likelihood ratio confirm the significance of these terms as a block (at the 10% level), but in no case does their inclusion cause a substantial alteration of the initial parameters. Again, with few exceptions, the results for the two samples are very similar. Together, these first offer a view of overall labor force dynamics and then of how the four sectors interact. I. Overview of the Labor Market Three important general findings are immediately apparent. First, the matrices reveal high levels of mobility with turnover rates (and implicitly the length of tenure at 5.21-5.7 years) in the formal sector similar to those in the U.S..'7 There appears little evidence of the rigidities that the 17 MediantenureforaU workers over 16 years old in 1991 was 5.1 years. BLS NEWS, USDL 92-386. The implicit tenure based on turnover rates on mean Juily/August 1994 median separation rate of 1. 1%/month was 7.6 years. Bureau of National Affairs, Bulletin to Management, Dec. 8, 1994. Mean tenure calculated as 13  incentives in the labor code would lead us to expect. Second, the symmetry of Vs across directions of movement in all sectors of paid work seems more consistent with a well-integrated market where workers search across sectors forjob opportunities, than one where informal workers seekpermanent status in the formal sector and stay until they retire." Finally, the logit results show that in no sector does the probability of moving into the formal sector relative to staying increase in overall experience as would be expected if there were queuing to enter the sector.19