77286 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1: 185-209 Labor Earnings in One-Company Towns: Theory and Evidence from Kazakhstan Martin Rama and Kinnon Scott One-company towns, characterized by the presence of a large employer in a local labor market, are a frequent legacy of state-led development strategies. How will downsizing or closing unprofitable state-owned enterprises affect these towns? This article develops a simple model combining monopsony power in the labor market with a Keynesian closure of the product market and uses it to interpret the findings of previous studies. The article evaluates the impact of the company's employment level on the town's labor earnings in Kazakhstan, where one-company towns are still preva- lent. The evaluation is based on data from the 1996 Living Standards Measurement Survey. The results show that labor earnings in the town decrease roughly 1.5 percent when the share of its population working for the company decreases 1 percent. The results are robust to changes in the definition of labor earnings and to the inclusion of a variety of other community characteristics in the analysis. These results and the theo- retical model are combined to evaluate the welfare impact of company downsizing and, consequently, to derive the optimal extent of labor retrenchment. Many developing countries, and especially transition economies, have one- company towns whose distinctive feature is the presence of a large employer in a local labor market. In spite of their name, one-company towns are not neces- sarily urban agglomerations. In some transition economies, the label applies to the rural area surrounding a large agricultural producer, such as a state-owned farm. More often, however, it applies to mining towns and to the communities developed next to large manufacturing plants, such as steel mills or armament factories. In all cases, the company accounts for a substantial share of the jobs in the town, and even those who do not work for the company depend on it to make a living. If the company were to cease its operations, the one-company town could easily become a ghost town. One-company towns also exist, or at least existed, in industrial countries. The coal mining company towns in the U.S. region of Appalachia at the turn of the century were in some respects similar to the one-company towns in tran- sition economies and developing countries nowadays. In coal mining company towns, a vertically integrated enterprise provided a variety of affiliated services Martin Rama and Kinnon Scott are with the Development Economics Research Group at the World Bank. This article is part of the research project Public Sector Retrenchment and Efficient Compensation Schemes, supported by the Research Committee of the World Bank through grant RPO 679-51. © 1999 The International Bank for Reconstruction and Development /THE WORLD BANK 185 186 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 to its workers, from housing to stores to medical services. However, in some of the coal mining company towns no positive externality stemmed from the com- pany because the only jobs in town were company jobs. In this respect, the one- company towns in developing countries and transition economies may resemble more the independent towns considered by Fishback (1992), where housing and stores do not belong to the main employer. Central planning, state ownership, and directed credit have made the one- company town setting more prevalent in developing countries than it was in industrial countries by fostering the settlement of large plants in relatively iso- lated communities. Large plants were supposed to exploit economies of scale, reduce import dependence, or satisfy sentiments of national pride. Remote loca- tions were sometimes justified by the availability of a key natural resource, but they were also chosen to promote regional development, to assert the nation's sovereignty, or to hide plants from the outside world. At present, most of those plants are inefficient and overstaffed. The demise of state-led development will be incomplete, it seems, until these plants are dramatically downsized, truly priva- tized, or closed altogether. Decisions affecting state-owned enterprises in one-company towns should not be based solely on considerations of profitability. Keeping these companies in operation certainly entails a cost for the rest of society because their deficits translate into higher taxes or lower social expenditures. But shutting them down, or significantly downsizing their operations, entails a cost for the populations that surround them and depend on them to sustain their economic activity. The optimal policy decision involves a tradeoff between these two costs. Unfortu- nately, relatively little is known about the magnitude of the cost to the sur- rounding populations. As in other areas of public policy, externalities are more difficult to quantify than deficits. In the absence of information about the im- pact of the company's size on the town's earnings, decisions regarding labor retrenchment could be misguided. This article analyzes the one-company-town problem from both a theoretical and an empirical point of view. The theoretical part develops a simple model of the one-company-town setting, combining aspects of the monopsony model of the labor market and the Keynesian model of the product market. The monop- sony aspect of the model is warranted because the company may not behave as a wage taker in the town's labor market. After all, the one-company-town set- ting provides a textbook example of market power by employers. The Keynesian aspect of the model captures the fact that many of the other, smaller businesses in town cater to the company and its employees. This simple model is used as a framework to interpret the findings of previous studies, to derive the specifica- tions for the empirical analysis, to evaluate the welfare impact of downsizing, and to discuss the corresponding policy implications. The empirical part uses data from Kazakhstan, a country with several fea- tures that make it relevant for our study. First, it is to some extent representative of the transition economies in Central Asia and Eastern Europe. Its output per Rama and Scott 187 capita is close to the median for this group of countries, and other indicators of economic development occupy a roughly intermediate position as well. Second, the one-company-town setting is still prevalent in Kazakhstan, and, because of the vast surface of the country, one-company towns tend to be quite isolated from one another. Third, the data on individual earnings and community char- acteristics that are needed for the empirical analysis have been collected and processed recently. A previous use of these data in the context of a poverty assessment indicates that they are of reasonable quality (World Bank 1998). Section I discusses the one-company-town problem in an intuitive way. Sec- tion II formalizes this discussion under the form of a simple analytical model. In section III, this model serves as the background for a review of the literature. Section IV provides a brief description of Kazakhstan and its labor market. Sec- tion V addresses data issues, including the construction of the variables measur- ing labor earnings and company size. Section VI regresses labor earnings on individual characteristics and the size of the largest employer in town. Section VII shows that the results are not affected when other community characteris- tics, apart from the size of the largest employer, are taken into account. Section VIII combines the empirical results and the theoretical model to assess the wel- fare impact of company downsizing and to derive the policy implications. Sec- tion IX concludes. I. THE PROBLEM The textbook partial equilibrium diagram of the labor market can be used to discuss the one-company-town problem in an intuitive way. This diagram repre- sents the determinants of employment and labor earnings in an integrated, rela- tively frictionless labor market. Consider the labor market of a town, under the assumption that neither local firms nor individuals can move easily in or out of it. Suppose that no local firm is able to affect labor market outcomes on its own. Suppose also that a large number of individuals can either work or perform other activities, such as studying or rearing children. In this simple setting, it is possible to analyze how the settlement of a large firm in the town affects its labor market equilibrium. The results shed light on what would happen if an existing large firm were to cease or downsize its activities. The upward-sloping line in figure 1 represents labor supply. It indicates how many individuals (L on the horizontal axis) would prefer to work rather than perform other activities for any given wage rate (W on the vertical axis). The downward-sloping line marked "labor productivity without company" repre- sents labor demand by local firms. It indicates the marginal productivity of la- bor in these firms or, conversely, the number of workers these firms would be willing to employ at any given wage rate. This line is downward-sloping under the assumption of a decreasing marginal productivity of labor; it would be flat if labor productivity were constant. In the absence of a large company, the labor market equilibrium would lie at the intersection of the supply and demand lines. 188 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 The equilibrium wage in the town would be W°, and the equilibrium employ- ment level would be L°. When a large company settles in the town, at least some workers can be em- ployed more productively, as shown by the upward shift of the marginal pro- ductivity line in figure 1 (the new line would be flat if labor productivity at the company were constant). Assume first that the company behaves in the same way as local firms. Under this assumption, the new labor market equilibrium would obtain for an employment level L 1 and a wage level W1. However, it could be in the interest of the company to pay less than W1. At a lower wage level, more individuals would prefer to perform other activities rather than work for a wage. Because some of those who work for a wage are employed by local firms, employment in the large company would fall. The company's output would fall, but its labor costs would fall, too. It is in the interest of the company to pay less than W1 as long as the resulting fall in output is smaller than the corre- sponding fall in labor costs. Let W2 be the wage level that maximizes the profits of the company. This level is necessarily higher than W°; otherwise, the company would be unable to compete for workers with local firms. It follows that the equilibrium level of employment is higher than it would be in the absence of the company (L2 > L°). Employment by local firms, however, may be either higher or lower than before, depending on the impact of increased employment and wages on local consump- tion. There are many goods and services that cannot be "imported" from other Figure 1. Employment and Wages in a One-Company Town Wage ( W) Labor labor productivity productivity Wlth without company company Labor supply w2 Employment (£) Rama and Scott 189 towns, so their prices would increase with consumption, and local firms would find it profitable to hire more workers after the company settles in the town. The broken line in figure 1 represents the labor demand of local firms in the new equilibrium (as the model in the next section shows, this line is not necessarily downward-sloping). For a wage level W2 employment by local firms would be Ly. This new employment level is larger than L° when local goods and services account for a large share of the town's consumption. It is also larger than L° when the productivity of labor is constant. Whether the company behaves as a wage taker or as a wage maker depends, among other things, on its objectives. In state-owned enterprises, those objec- tives are not always clearly specified. And even when they are, managers do not necessarily face the appropriate incentives to fulfill them. However, figure 1 shows that the presence of a large company raises labor earnings in town as long as labor market equilibrium lies on the town's labor supply line. This result holds regardless of whether the company exploits its ability to push wages down. n. A SIMPLE MODEL To interpret the findings of previous studies as well as to implement the em- pirical study of the case of Kazakhstan, we organize the preceding discussion under the form of an analytical model. This model does not thoroughly describe the one-company-town setting. Instead, it highlights how the size of the com- pany can affect employment and earnings in the town. With this objective in mind, we keep our hypotheses regarding technology and preferences simple. The introduction of a Keynesian closure in an otherwise standard monopsony model of the labor market distinguishes this model from the previous literature (see Boal and Ransom 1997). Assume that local firms use labor only and that their productivity is constant regardless of their volume of output. An example of a local firm with these characteristics would be a barbershop. Haircuts involve little more than the barber's time, and the average time per haircut does not vary much with the size of the shop. If units are chosen appropriately, each employee produces one unit of output over the relevant time period. Assume that labor productivity is con- stant in the company as well, at level K. Parameter K can be seen as a measure of the capital stock or the size of the company. These assumptions on technology can be written as follows: (1) YT = LT (2) YC = KLC where Y is output, T refers to local firms, and C refers to the company. Under these assumptions, the demand lines in figure 1 would be flat. Total employment in the town is the sum of employment in local firms and employment in the company. But for the town's labor market to be in equilib- 190 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 rium, total employment must equal the number of individuals willing to work at the prevailing wage rate. Those individuals who have a better alternative prefer not to work, unless the wage rate increases enough to match the value of that alternative. This value represents their reservation wage, R. As the wage rate goes up, individuals with increasingly higher reservation wages are drawn into the labor force, and total employment expands: (3) L = Lr+Lc (4) W=R Jn T -I (5) R = R(L) with 5f~ = -. dL R e Parameter e can be interpreted as a measure of the elasticity of the town's labor supply. The larger is e, the more responsive is labor supply to changes in the wage level. Increased labor mobility between towns should therefore raise the value of e. Finally, it is assumed that the output of local firms is consumed in town, whereas the output of the company is sold out of town. The unit price of the company's output is supposed to be given. The unit price of local output, in turn, is determined by supply and demand. On the supply side, competition between local firms implies that this price has to be equal to the wage level W (in the example above, the price of the haircut cannot be distinguished from the earnings of the barber). On the demand side, it is assumed that a fraction a of total income in the town is spent on locally produced goods. (This assumption corresponds to the case where the households' utility function is of the Cobb- Douglas type, exhibits constant returns to scale, and includes the consumption of local goods among its arguments.) Total income is the sum of the value added by local firms and the wage bill of the company. Equilibrium in the market for local goods is attained when the supply of local goods measured in money terms WYT equals demand DT: (6) WYT = DT (7) D T = a(WYT+ WLC). Parameter a can be interpreted, in Keynesian terms, as a marginal propensity to consume. Employment by local firms depends on the employment and pay decisions of the company because those decisions affect income and, therefore, the demand for local output.1 It follows from equations 1, 3, 6, and 7 that 1. The simple specification chosen for DT implies that there is no equilibrium in the absence of a company, except as a limit case when K — » 0. A full analysis of the case where no large company operates in town would require the introduction of a second group of local firms, producing goods that can be sold out of town. This would complicate matters without shedding additional light on the consequences of downsizing. Rama and Scott 191 (8) L= J—LC. 1-a This relationship between L and L c is reminiscent of the Keynesian multiplier. For every job created by the company, more than one job is created overall. The size of this multiplier depends on the marginal propensity to consume local goods oc The number of jobs created by the company, in turn, depends on whether it behaves as a wage taker or as a wage maker. The company's profits 7t are a function of its labor productivity, its labor costs, and its employment level: (9) n = (K-W)Lc. If the company treats the wage level as given, it expands its employment level as long as K is higher than W. But for more individuals to be drawn into the labor force, the wage level W has to increase. Eventually, the marginal profit from hiring an extra worker falls to zero, and the labor market reaches a new equilib- rium. In this equilibrium the wage level W1 verifies (10) W1 = K. The company could take advantage of its size to influence labor costs, how- ever. This is because the number of workers Lc it attracts varies with the wage level W it pays. The relationship between L c and W depends on the elasticity of labor supply by individuals, as well as on the endogenous response of employ- ment by local firms to changes in employment and pay in the company. In ana- lytical terms, equations 4, 5, and 8 imply (11) Equation 11 can be interpreted as the labor supply curve faced by the company. The company's labor supply curve is flatter than the town's labor supply curve because a higher wage level leads to a higher labor demand by local firms, so that L c increases by less than L. However, under the chosen assumptions, the elasticity of the company's labor supply curve is the same as that of the town's labor supply curve, namely e. The optimal wage level from the point of view of the company, W2, can be obtained by replacing the labor supply curve faced by the company in the profit function represented by equation 9. The first-order condition for this problem is 192 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 As long as the local labor supply curve is not infinitely elastic, this wage level is lower than W1. Because fewer individuals are drawn into the labor force, total employment is lower when the company behaves as a wage maker. m. PREVIOUS STUDIES Previous studies have evaluated the empirical relevance of some of the key assumptions in the model. For instance, in the model the town's labor supply is not infinitely elastic, implying that workers do not move in and out of town in large numbers in response to regional differences in labor earnings. The litera- ture on local labor markets has implicitly dealt with this assumption, by assess- ing whether the labor earnings of otherwise similar workers vary across regions in a country. The model also assumes that the company is large enough to enjoy a significant monopsony power and possibly take advantage of it. Several stud- ies have focused on labor markets with only a few employers and tried to deter- mine whether wages in those markets are below their competitive levels. An- other key assumption is that employment and pay in the company have an impact on labor earnings in the town. Some research has implicitly dealt with this as- sumption by documenting everyday life in one-company-town settings.2 Even in a well-integrated economy, with good infrastructure and transporta- tion, efficient housing markets, and a highly mobile labor force, regional dis- parities in labor earnings can be significant. In the United States, wage inequal- ity has increased considerably over the past quarter of a century, but it has done so at markedly different paces depending on the region. Using data from popu- lation surveys covering almost two decades, Topel (1994) estimates the determi- nants of the relative wages of low-skilled and high-skilled males at the regional level. His results suggest that distinctly local factors affect relative wages. The extent of labor markets thus appears to be limited by geography. In a previous study in the same spirit, Topel (1986) also shows that wages are responsive to employment shocks at the state level. The specification he uses can be inter- preted as a version of the labor supply function in equations 4 and 5, with con- trols added for education, experience, marital status, and other characteristics of workers, all measured at the state level. Topel finds that wages are indeed responsive to local employment shocks, particularly in the case of older and less-educated workers. This finding suggests that local labor supply is upward- sloping, as is assumed in the theoretical model. A related issue is whether large employers enjoy some monopsony power in local labor markets. This issue has been addressed for workers with specific labor skills in well-delimited geographic areas. The market for hospital nurses is 2. There is also a vast literature on the consequences of company downsizing on the earnings of the workers dismissed by the company. Whereas these consequences tend to be sizable, workers in other firms are not necessarily affected by the downsizing process. Actually, most analyses in this literature are carried out under the assumption that downsizing entails no externalities. As a result, this literature does not shed much light on the one-company-town problem. Rama and Scott 193 a case in point. Because there are only a few hospitals in each city, monopsony power by hospitals could be considerable. Frequent complaints about the "chronic shortage of nurses" indicate that hospitals would be willing to hire more nurses at the prevailing wage rate, but not to raise their wage rate to attract them. Equilibrium vacancies of this sort can be expected when firms pay wages below the relevant marginal product. Sullivan (1989) assesses the monopsony power hypothesis by estimating the elasticity of the supply curve of nurses faced by individual hospitals. The speci- fication used by Sullivan is similar to that of the labor supply function faced by the company, as described by equation 11, with wages and employment mea- sured at the hospital level. Sullivan's model includes controls for employment, wages, number of patient-days, and average length of stay in other hospitals. The regression analysis yields values of the elasticity e ranging from 1.3 in the same year to 3.8 over a three-year period. From these results, Sullivan concludes that hospitals enjoy a significant monopsony power in the market for nurses. Surprisingly, however, the estimated values of £ are similar for metropolitan and nonmetropolitan areas. Because there are many more hospitals in metro- politan areas, the labor supply curve they face should be much flatter (that is, e should be higher). Therefore, Sullivan's results may capture something different from monopsony power. They would be consistent, for instance, with efficiency wages. If the probability of being caught shirking fell with the size of the hospi- tal, large hospitals would have to pay higher wages than small hospitals in order to elicit the same level of effort from their nurses. But in that case, the equation estimated by Sullivan could not be interpreted as a quasi labor supply curve. Monopsony power by employers was also likely in the coal mining company towns of Appalachia at the turn of the century. Miners in some of these towns lived in company housing and were paid in scrip that could be used only in the company store. Some of the companies controlled even the police. If these com- panies could not set wages below the corresponding marginal product of labor, then it is difficult to claim that companies elsewhere can do it. Boal (1995) con- tends that they could not. His regression analysis, based on data from 30-odd coal mining counties in West Virginia over the period 1897-1932, explains county-level wages as a function of county-level employment in the same year and in the previous year. His specification is similar to that of the local labor supply curve in equations 4 and 5. The population of the county and county dummies are included among the independent variables in the regression. Con- trols for average wages and employment in the other mining counties are also added, depending on the monopsony hypothesis used. Boal finds a large coeffi- cient for employment in the same year, but the coefficient is almost as large, and has the opposite sign, for employment in the previous year. He concludes that employer monopsony power was limited in turn-of-the-century coal mining, except in the short run. These findings have led to a reinterpretation of the Appalachian coal mining company town. Fishback (1997) argues that company housing and company 194 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 stores were mutually advantageous arrangements for workers and employers, rather than mechanisms set up by employers to exploit their workers. But it is worth mentioning that Appalachian company towns were not far apart and were linked by rail lines built to haul the coal. Findings concerning Appalachian com- pany towns may therefore be of limited relevance for evaluating the impact of downsizing on more isolated one-company towns. Detailed case studies are another source of information about one-company towns. The history of the town of Pullman (currently a suburb of Chicago) by Buder (1967) confirms how difficult it is to disentangle paternalism from exploi- tation, or mutually advantageous arrangements from monopsony power. George Pullman, who revolutionized railroad travel with his development of the sleep- ing car, also had a serious interest in the labor problems of the time. In 1880, he launched the construction of a model town to house his workers next to his new factory. His intention was to apply principles of business efficiency to meet the needs of his workers and establish a more peaceful system of labor relations. The bitter strike of 1894 suggests that workers thought otherwise. Another case study implicitly dealing with monopsony power concerns gold mining in southern Africa. The account of this history by James (1992) makes it clear that the Southern African Chamber of Mines was not as enlightened as George Pullman. This chamber was formed in 1889 to organize labor supply and mitigate competition for workers between mines. Shortly after it was formed, it established centralized labor-recruiting agencies that fixed the African work- ers' food supply, length of shift, working day, and wage rate. Individual mines could set their own wage scales, but their average wages could not be higher than those of other mines. This system ensured that wages for miners remained constant, in real terms, between 1911 and 1969. Although James's account does not prove that gold mines paid wages below the corresponding marginal pro- ductivity of labor, it suggests that they did. IV. THE CASE OF KAZAKHSTAN From an empirical point of view, it would be ideal to have detailed data not just on one but on several one-company towns. Variation in the size of the com- pany relative to the size of the town could then be used as a natural experiment. Conceptually, the experiment would involve a comparison of the labor earnings of workers who are alike in all respects, except that some live in towns where large companies operate whereas others live in towns with smaller companies. If the labor earnings of these workers differed in a systematic way across towns, the difference could be attributed to the externality created by the company. The regression analysis in section VI emulates this natural experiment. Kazakhstan has dozens of one-company towns. Labor mobility in and out of them is quite low, both for geographic and for institutional reasons. Kazakhstan is the ninth largest country on earth, with a population of less than 17 million people scattered across 2.7 million square kilometers. Labor mobility is con- Rama and Scott 19S strained, among other reasons, by the system of unemployment benefits. Unem- ployed workers need to hold a propiska (a local passport) in order to draw these benefits, which may dissuade them from searching for jobs outside their region of residence. Mobility is also constrained by a poorly developed housing market outside Almaty. The provision of social services by firms may further discourage mobility, although substantial progress has been made in transferring these ser- vices to local authorities. Kazakhstan has some features that make it rather atypical among transition economies. For example, the sharp ethnic divide between the Russian north and the Kazakh south has not so far led to conflicts such as those experienced by some of the other former Soviet republics. We take into account this divide when we assess the robustness of the empirical results. Another atypical feature is that during the Soviet period Kazakhstan received the highest amount of sub- sidies under the old system. This massive transfer of resources is probably one of the reasons underlying the large number of one-company towns. In other respects Kazakhstan is not an exceptional country. Its level of devel- opment is close to the median for transition economies. With an output per capita of roughly $3,000 measured at purchasing power parity, it ranks number 14 among 23 former socialist countries in Eastern Europe and Central Asia. The private sector's share in the economy is not particularly low. A gradual process of privatization and enterprise restructuring has occurred since 1991, and the assets of many nonagricultural firms were transferred to managers' and work- ers' collectives in 1993. Since then, privatization has been extended to more than 90 percent of farms and 80 percent of farmland. Finally, few institutional constraints exist on the employment and pay deci- sions of firms in Kazakhstan (Klugman and Scott 1997). Unionization is high, but it is widely believed that trade unions are not a significant source of labor market rigidity. The minimum wage has been drastically eroded by price liberal- ization and high inflation, to the point where it lacks any significance, except as a scalar for wages in the public sphere (but even the lowest-paid public sector worker earns three times the minimum wage). Also, employers obtained the right to set wages in 1987 and to lay off workers in 1990. Workers are entitled to "adequate" notice before dismissal and to severance pay amounting to three months of wages. However, newly privatized firms are under no direct or indi- rect obligation to maintain employment levels. In light of these weak institu- tional constraints, the labor market equilibrium concepts used in the theoretical part of the article should be appropriate for Kazakhstan. V. THE DATA Another factor that makes Kazakhstan an interesting case is the availability of good data. An integrated multipurpose household survey was conducted in 1996. This survey had the same features as the Living Standard Measurement Surveys (LSMS) implemented with the assistance of the World Bank in many de- 196 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 veloping countries and transition economies since 1985 (see Grosh and Glewwe 1995 for a description of these surveys). The LSMS for Kazakhstan covered a nationally representative sample of more than 7,200 individuals in about 2,000 households. A probability sample was used, instead of the old quota sample of the existing Family Budget Survey, which was typical of the former Soviet Union. Because the data were collected during a single month, the survey can be seen as a pure cross-section of households, with no time dimension involved. The sur- vey gathered information on individual characteristics and earnings, as well as on household consumption. In addition, its community questionnaire reported the distance between each community and the nearest large farm or industrial enterprise, as well as the number of individuals from the community who worked for that farm or enterprise. Measuring labor earnings in a transition economy like Kazakhstan is not an easy task. Official wages do not provide a complete picture of official labor compensation. Official wages may underestimate total earnings because they do not include nonmonetary benefits. During the Soviet period, these benefits tended to be an important component of total labor earnings because they served to attract and retain workers when the firms' autonomy to set wages was limited. Typically, nonmonetary benefits took the form of social services, such as health and child care, or access to sanatoriums and recreational places (Klugman and Scott 1997). A portion of these benefits has been retained. Alternatively, official wages may overestimate total earnings, due to substantial arrears in their pay- ment. Many firms facing a decline in demand for their output have refrained from adjusting employment downward, relying instead on reduced hours and "administrative" leave. To compensate for the shortfall in earnings, workers increasingly combine attachment to their formal employers with growing in- volvement in informal activities (Murthi 1998). In light of these measurement problems, in the empirical analysis we use three different measures of labor earnings: wages officially earned, wages actually re- ceived, and per capita consumption. The first two measures are defined at the individual level, whereas the third one is calculated at the household level. Because of nonmonetary benefits, wage arrears, and informal activities, consumption per capita could in principle provide a better picture of actual earnings from labor than the other two measures. However, consumption per capita is also affected by the size, composition, and asset ownership of households, so that it is not strictly comparable across individuals. These shortcomings imply that none of the three earnings measures is entirely reliable on its own. But econometric results that are consistent across the three measures should carry some credibility. Table 1 reports the means and standard deviations of the three earnings mea- sures and the individual characteristics that are usually considered among the determinants of labor earnings. Figures for consumption per capita are con- structed using the same methodology as in the poverty assessment of Kazakhstan in World Bank (1998). Table 1 shows that the average per capita consumption is similar to the average wage, although some household members do not earn Rama and Scott 197 Table 1. Sample Statistics from the Kazakhstan Living Standards Survey, 1996 Standard Number of Variable Mean deviation observations Age (years) 37.162 9.707 2,278 Gender (male = 1) 0.522 0.497 2,278 Schooling (years) 12.122 2.332 2,278 Wages officially earned 5,077.3 4,593.8 2,278 Wages actually received 5,791.0 6,492.4 1,311 Per capita consumption 5,063.9 3,595.5 2,354 Company's population share 0.055 0.096 98 Note: Monetary variables are measured in tenges per month, at 1996 prices. Source: Authors' calculations based on survey data. any income. This similarity confirms that households have other earnings that are not necessarily legal or reported in addition to their wage earnings. How- ever, the importance of these other earnings should not be overstated because labor force participation rates are high in Kazakhstan. Moreover, ethnic Rus- sian families, who account for roughly 40 percent of the population, typically have one child. As a result, the number of household members not earning any income is lower than elsewhere. Other measurement issues concern the size of the company. Unfortunately, the number of individuals who work for the nearest large employer is not avail- able for 34 of the 132 communities. Table 2 displays the medians and standard deviations of the available data, expressed as percentages of the community's population. When there is no large employer in or close to the community, a zero is reported. Unfortunately, the questionnaire did not foresee the possibility Table 2. The Company's Population Share, Kazakhstan, 1996 (distance at which the nearest large employer is located) Nature of In the Within a 10- Between 10 and More than SO the community community kilometer range SO kilometers kilometers Town Median 0.013 0.108 0.009 0.015 Standard deviation 0.037 0.147 0.016 0.008 Number of observations 25 11 4 2 Other urban Median 0.035 0.043 0.013 0.000 Standard deviation 0.093 0.042 0.018 0.000 Number of observations 8 4 2 1 Rural Median 0.178 0.005 0.000 0.000 Standard deviation 0.145 0.029 0.031 0.048 Number of observations 12 7 10 12 Note: The company's share of population is defined as the fraction of total population in the enumerating district working for the largest local employer. Source: Authors' calculations based on survey data. 198 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1 of more than one large employer per community. The data are disaggregated according to the three types of community considered by the L SM S and the dis- tance from the community to the company. We use this disaggregation to im- pute the company's share of population for the communities with missing data. Consider, for instance, the case of a town where the nearest large company is within a 10-kilometer range but where data on its employment level are missing. Our imputation procedure assumes that the share of the town's population work- ing for the company is equal to 0.108. Econometric analyses using the imputed data are likely to yield more significant estimates than those obtained using the reported data only. If the results turn out to be consistent regardless of whether imputed data are used, these estimates should carry some credibility. VI. COMPANY SIZE AND EARNINGS The impact of the size of the company on the town's labor earnings is as- sessed by estimating a quadratic version of the quasi labor supply curve in equa- tion 11, as follows: The X variables capture earnings determinants such as education and experi- ence. These variables and the earnings variable Ware measured at the individual level. The share L c of the population employed by the company, however, is measured at the community level. Consequently, equation 11' can be interpreted as an earnings function augmented so as to include one community characteris- tic among its explanatory variables. The idea of introducing community characteristics among the determinants of individual earnings or individual consumption is not new. For instance, Borjas (1995) shows that individual wages in the United States are affected not only by individual educational attainment but also by the average educational attain- ment of the neighborhoods in which the individuals grew up. Community char- acteristics are also included among the determinants of consumption per capita in developing countries. Thus, Narayan and Pritchett (1997) show that village social capital (measured as the membership of a variety of groups, such as burial societies) has an impact on individual consumption in Tanzania. Similarly, Ravallion and Wodon (1997) find that location dummy variables are important in explaining per capita consumption in Bangladesh. To our knowledge, the size of the largest employer in town has not been considered among the relevant community characteristics. The quadratic specification chosen for equation 11' allows the elasticity of the labor supply curve faced by the company to vary with its size. More specifically, n2) dlogLc _ dLc 1 _ 1