Labor Migration and Wage Inequality

Transcription

1 Labor Migration and Wage Inequality ZHONG Xiaohan * Center for China in the World Economy (CCWE) School of Economics and Management, Tsinghua University, Beijing, China Abstract: Building on the model from Kremer & Maskin (1996), the paper gives uniform explanations on two prominent phenomena in China s labor migration during transition: first, immigrant workers have been earning different wages at the same time take different jobs; second, labor flow has been associated with regional divergence rather than convergence. The basic logic is that labor flow provides new opportunities of labor division (or matching), which creates job-taking segmentation as well as increases wages of local workers. The implication is that labor flow not only enhances efficiency as a whole, but may also be a Pareto improvement. The empirical justifications also support our theory indicating positive effects of labor immigration on wage growth. JEL Classifications: J61, J31 Keywords: Immigrant workers, wage differentials * School of Economics and Management, Tsinghua University, Beijing Tel: , Fax: , zhongxh@em.tsinghua.edu.cn.

2 1. Introduction Since China s reform era, especially after late 1980s, the registered permanent residence (Hukou) system and its related government policies, which had been preventing people from migrating and thus resulting segmentations of labor market during several decades, fell into a decline in a gradual yet persistent way. Labors from less-developed areas - rural areas in the mid and west part of China - flocked into developed areas - urban areas along the east coast. The so-called tide of rural labor has been attracting economists attentions by its unusual (but somehow undiscovered) impacts. Specifically, there have been two prominent facts: First, rural labors (Mingong) deploy quite different jobs from those of urban labors (formally called Staff and workers, or Zhigong). Rural labors dominate in almost all kinds of unskillful occupations: hardhats, repairers, waiters and waitresses, dustmen in office buildings, and nurses in households. Related with (but probably not resulting from) these occupational segmentations, rural workers earn much lower salaries than their urban counterparts. Second, labor flow has not been able to induce wage convergence among regions: either urban-rural or east-west gaps have not declined. In contrast, during most periods of recent two decades, there are divergences rather than convergences. The simple theory proposed in this paper turns out to be able to explain these two seemingly irrelevant puzzles at the same time. But before a formal expression, let s turn to a more concrete and convincing observation of these two facts, which helps to draw our basic ideas. 1.1 Labor Migration, Occupation and Wage Differences In developed urban areas, such as cities of Beijing and Shanghai, to tell a worker whether he (she) comes from outside (rural areas) or inside (urban areas) turns out to be quite easy. The answer: looking at what his (her) job is. This answer is true even for workers in same firms. For example, at a renowned university, managers of its logistic departments are all staff and workers granted with urban residence. However, physical workers and servicemen, except for those working at high-tech divisions such as computer networks, are all out-of-town labors. This same university also finds some shanties close to newly built office buildings, accommodating hardhats identified as rural workers. At the same time, their wives and sisters have opportunities to get into these buildings -for cleanings and maintenances. A recent survey (Population and Social Science and Technology project team, thereafter PSST, 004) details such an occupation differences. Table 1 gives occupation distributions of both migrating employment and non-migrating employment. There are some defects: in the non-migrating employment, it does not specify occupation distributions of purely urban workers yet included those residing in rural areas, which prevent us from comparing occupation differences between migrating and local workers in urban areas. However, it distinguishes between intra-provincial migration and inter-provincial migration within the migrating employment. The intra-provincial migration includes short-distance moves within counties or cities, most of which are due to land reforming, study and training (which should be exogenous to job seeking), less (only 19 percent) for job seeking, as workers or businessmen (see table 4, ibid). Thus it is arguable that occupation distributions of the intra-provincial migration tend to be close to that of local urban workers (intra-provincial migration within rural areas seems to be 1

3 small). On the other hand, occupation distributions of inter-provincial (the proportion of workers and businessmen is 64%, as shown in table 4, ibid) can be argued to be similar with those of rural workers in cities. Given these measure imperfections, table 1 still draws a very clear picture. As high as 60 percent of intra-provincial migrating workers engage in operation of manufacture and transportation equipments or the related, compared with only 9 percent in intra-provincial migrating workers, or 40 percent in non-migration population with the exclusion of farmers. The Proportion of workers in Business and Service (most of them are vendors and waiters/waitresses) is 0 percent among intra-provincial migrating workers, which is also comparatively high. Contrarily, there is only 4 percent of inter-provincial immigrants who can work as clerks and the related, compared with 9 percent of intra-provincial immigrants. Still only 4 percent of them become professionally technical workers, compared with as high as 16 percent of intra-provincial immigrants. The evidence is convincing enough that rural workers indeed conduct quite different jobs from their urban counterparts: the former struggles with unskillful (or blue-collar) workloads, while the latter indulges in skillful (or white-collar) circles. Table 1 Occupation Distribution (%) Occupations Migrating Employment Non-migrating Total Total Intra-provincial Inter-provincial Employment Employment I. Principals of Bureaucracies, the Party or Social groups, Enterprises and Institutions II. Professionally Technical Workers III. Clerks and the related IV. Business- or Service-man V. Farmers and Irrigation Workers VI. Operators of manufacture and transportation equipments and the related VII. Non-categorized workers Total Source: PSST, 004. Coexisting with these occupation differences, wage levels between rural and urban workers are also different. Researches on wage differences are abundant (see, e.g., Zhao (004) for a recent literature review). Meng and Zhang (001) assert that urban workers earn up to 50 percent higher than their rural counterparts. Zhao (1999) calculated the monthly wage of immigrants in 1999, which is 533 Yuan and lower than 696 Yuan earned by urban counterparts. One question to be address is: What s the relationship between occupation and wage differences, which exists simultaneously between rural and urban workers? At one extreme, some economists deny (explicitly or implicitly) any systematic relations between these two phenomena. This view believes that wage differences can be explained largely by standard theories such as human capitals. Given workers education (or skill) level and other

4 inherent abilities, and controlled for compensating differentials, wages must be the same for any kinds of jobs. This theory seemingly explains wage differentials pretty well, but leaves the question of occupation differences open. After all, hiring a rural worker (with a lower ability than his/her urban counterpart) as a white-collar can still be profitable for his/her employer, as long as a lower wage level compatible with his/her productivity be paid (and he/she is surely willing to accept it). At the other extreme, some economists believe occupation differences are the major cause of wage differentials, and occupation differences are imputed to governments policy discriminations to rural labors. Rural workers are prohibited from entering high-salary industries even if they are fully competent, or, in the case of lower skills, happy to have a lower wage. This theory predicts that labor reallocation and wage convergence will happen soon after the government removes its labor restrictions. However, looking at China s experiences in the past decades, it seems that governments moved faster than the market itself: we now have much more rural labors roaming in cities, but, ironically, occupation and wage differences seems more persistent. This paper holds another view apart from the above two. I admit that, labors skills are contributable to their wage differentials. Besides, I show that skills differences at the same time cause occupation differences, i.e., local workers to be white-collars and rural incomers blue-collars. Even more, this occupational differences exaggerate wage differentials caused by different skill levels. Such a line of deduction combines wage differentials with occupation differences logically and thus goes well beyond the first view stated above. Our theory also implies that occupation differences (and wage differentials) are results of economic rationalities, rather than governmental interventions. Probably more surprisingly, aggravated wage differentials benefits all in the labor markets, both local workers and rural incomers. This makes the second view above both replaceable and inoffensive. 1. Labor Migration and Regional Gaps Since late 1980s, labor flow from less developed rural areas to developed urban areas, greatly changed the pattern of China s labor market. According to data cited by Sicular and Zhao (00), the scale of rural-to-urban migration had skyrocketed from 8.9 million in 1989 to 70 million in The annual growth rate had been 1% in 9 years. As a result, the stock (or pool) of migrants, who are defined according to disparities of their registered residence (Hukou) from real living places, has become large ever. According to PSST (004), from the end of year 1995 to 000, the pool had become 137 million larger than before. Among them, 37 million are inter-provincial migrants, and in turn, 8 million of them are for employment. By regions, the east-china absorbed most of these migrants. Guangdong Province recruited 41 percent of all, followed by Zhejiang (9%), Shanghai (6%), Jiangsu (6%), Beijing (5%), etc. The mid- and west-china are the major exporters. Among them, Anhui, Henan and Sichuan provinces are top three. A somehow puzzling fact is: although labor migration has been of large scale all through the period, wage gaps among regions turned out to be enlarged at the same time. To briefly illustrate such an enlargement, we divide all 31 provinces in China into three areas, following the common practice: East, Mid and West area, and consider wage gaps between them. The statistic result is included in table, section 1. From the table, we can see a regional reversal of fortune before and after the reform era. East area, which has been long behind the national pace before the reform, 3

5 accelerated and surpassed the country s growth by a rate of percent each year, which is in turn higher than the mid and west area, by 1- percent per year. Let s focus on the two periods after the reform, year 1978 to 1991, and year 1991 to 00, to see evolutions of regional wage gaps. It s reasonable to regard the first period as of much less labor flow and the second much more, according to, e.g., Sicular and Zhao (00). Also, in the second period, the east area received immigrants, and the mid- and west-area exports them. The classical theory says that, fixing factors that affect labor demand, such as investments, labor flow would be good for eliminations of regional wage gaps. Labors are deemed to flow from low-wage areas to high-wage areas. Because of the basic rule of diminishing marginal productivities, an increase of labor supply in high-wage areas tends to lower wages there; a decrease of labor supply in low-wage areas tends to raises wages there. Such kinds of labor flow, if keep on, would diminish and finally eliminate regional wage gaps. According to this theory, controlling for others, the wage gaps between the east and west should have decreased significantly during the second post-reform era ( ), at least compared to the first era ( ). The fact can be found from section 1, table. In the first post-reform period, annual growth of wages in the east was higher than the mid and west by 1.95 and 1.73 percent respectively. Since, as we believe, there were little labor flow at that time, this gap can only be explained by other factors, mostly by investments. Section of table shows differences of investment growth in the two post-reform periods. In the first period, the annual growth of investment in the east was higher than the mid and west by.3 and 3.75 percent respectively, which seems enough to explain the wage gaps. Now looking at the second period, wage gaps between the east and the mid/west kept going. Annual growth of the east was still higher than the mid and the west by 1.55 and 0.93 percent respectively, only degenerating slightly 1. However, the investment gap, a significant factor on wage gaps, shrunk and even disappeared: the annual rate of the east was higher than the mid and the west by 0.68 and 1.65 percent respectively, less than half of the first period, and statistically insignificant. More strangely, the mid area decreased its investment gap with the east much more than the west, but its wage gap with the east became much more serious than the west. This gives additional evidence on the weak explanatory power of investment on wage gaps in this period. The comparison between the two post-reform periods can be looked as a natural experiment of effects of labor flow on wage gaps. The result seems to show that, labor flow enlarged rather than reduced wage gaps. Other researchers give the similar conclusions. For example, Zhong (005) clearly illustrated that, even controlled for the investment growth, regional wage gaps increased significantly after 199, with the east increased much faster than its mid and west counterparts. Furthermore, the paper also found that, construction and manufacture, traditionally perceived as major industries receiving rural labors, also saw an enlargement of regional gaps 1 The better performance of the west than the mid might be owing to the great development of the west policy implemented from year 000. Before that, i.e., during year , the wage growth rate of the west is lower than the east by 1.84 percent, which is similar with that of the mid. A similar conclusion can be drawn if provincial level data used. For example, consider the big-five importing provinces Guangdong, Zhejiang, Jiangsu, Shanghai and Beijing) and the big-three exporting provinces Anhui, Henan and Sichuan. Comparing the same two periods, in the importers, only Guangdong s wage growth has significantly slowed down. And in the exporters, only Henan s growth has slightly accelerated. As for investment growth, only Zhejiang, in the importers, has an obvious upward trend (and not surprisingly, its wage increase is the biggest among the five). And in the exporters, only Henan s growth decreased a little. As a whole, the investment growth converged among all the importers and exporters in the second period. 4

6 within their domains. How can we explain such an obvious paradox between classical theory and China s reality? It turns out that, the assumption of diminishing marginal productivities may not be true. It implies homogenous labors, or more precisely, labor all conducting the same jobs. However, if we loosen this assumption and think that different labors can work in different occupation, or at least play different role within one occupation, things will be changed. In fact, essentials of economics believe that division of labor can increase productivity. In our story, if labors holding different endowments (i.e., skills) conduct different occupations, the power of labor division should reveal. Rural labors coming into cities thus provide opportunities for labor division and improved productivities, at least partially offset the quantitative effect of increased labor supply attenuating productivity. On the other hand, however, migrations can also associated with an enlargement of wage gaps among regions, as we show later. But it is a favorable shock, as we emphasized, not only in the sense of efficiency (which is allowed without labor division and increased productivity), but also of equity, since it is a Pareto improvement. 5

7 Table Regional Gaps (Average on Provinces) East Mid West Difference Mid - East West - East 1. Relative Wage Growth (%) (0.8) (0.37) (0.90) (0.35) (0.35) (0.93) (0.83) (0.57) (0.36) (0.38) (1.78) (0.54) (1.38) (0.6) (0.6). Investment Growth (%) (4.0) (.86) (3.66) (1.67) (1.73) (3.36) (3.33) (.69) (1.40) (1.40) 3. University Student Numbers (In every ten thousand persons) (17.81) (3.08) (.68) (5.11) (5.11) (84.36) (.96) (8.94) (5.00) (4.8) Note 1: Numbers in parentheses are standard errors. Note : The east area includes the following 1 provinces: Beijing, Tianjin, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Guangxi and Hainan (after year 1990). The middle area includes the following 9 provinces: Shanxi, Inner Mongolia, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei and Hunan. The west area includes the following 10 provinces: Chongqing (after year 1997), Sichuan, Guizhou, Yunnan, Xizang, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang. Note 3: Relative wage growths are provincial growth rates minus the same year national growth rate. Note 4: Education levels of employment are weighted averages from education composition data. The weights are: zero year for illiteracy, 6 years for elementary school graduates, 9 years for junior high school graduates, 1 years for senior high school graduates, 14.5 years for junior college graduates and above. Source: NBSC (1990), NBSA (1997), NBS (various years), and author s calculations.. The Model and its Interpretations.1 A Modern Parable 6

8 Before a formal statement of our model, let s begin from a fictitious but meaningful story. At a small village in west part of China in the 1980s, Ms Li is a typical young woman farming with her husband in their homeland. Although they had been working very hard, the standard of livings were not satisfying. As most of their neighbors, they were lack of scientific as well as marketing skills which are largely due to their poor education background - both being graduates from elementary school. However, since the 1990s, their situation has changed. Li, through recommendations of a fellow villager, got into a property management company located in a big city, working as a dustman. Her wage has been raised yet the job is quite easy to do. A middle-age woman, Ms Zhang is their forewoman at this company. She originally worked in a textile factory, which went bankruptcy when competitions from Towns and Villages Enterprises (which hired large amount of rural labors) became fierce. She soon got this offer as a forewoman. Thanks to her comparatively higher education (a graduate from high school), she works pretty well and earns higher than before, and also much higher than her underlings as Ms Li. We will formalize this story into a model, and show its implications. The model is based on Kremer and Maskin (1996). There they use a similar model to explain job segregations and wage inequality. Our model, in the mathematical form, is an application and extension of theirs; but the backgrounds and implications are totally different.. The Production Function The key of the model is a specific production function. Consider a typical production process. Generally speaking, it can be accomplished only through cooperation of workers from different positions, or occupations 3. For simplicity, let s assume a production much be accomplished by a combination of two jobs. Let s call them design and implementation. The worker conducting the design is called a white-collar and implementation a blue-collar. Suppose each worker has a given skill, which determines output. Specifically, the production function is, (, ) = f q q q q ( 1 ) w b w b Where q w, q b > 0 represent skill levels of workers as white-collar and blue-collar respectively (equation (6), page 6, in Kremer and Maskin (1996)). Obviously, the production function is an increase function of skills. A production of this kind has two distinguished properties: First, there are imperfect substitutions between skills devoting to two positions. Recall that a production function of perfect substitution should have the form as f (q w, q b ) = aq w +bq b, where division of labors does not matter two position are completely replaceable and one of them can even be left vacant (e.g. blue-collar can be vacant if a>b). However, if the production function has no substitution at all, i.e., has perfect complements, formed as f (q w, q b ) = Min{aq w, bq b }. Two 3 There are different taxonomies of occupation. In many cases, we distinguish them by firms they are embraced in. A practitioner in a school is a teacher; in a factory is a manufacture worker, and so on. However, this paper does not follow such a definition. I regard occupation as different positions (or divisions of labor). There can be different kinds of occupations, or positions within a firm. In fact, the diversity of occupations is a key feature of modern firms. 7

9 positions are both so important that any allocation of workers (or skills) between them might be useless. For example, if a=b=1, any allocation leads to the same output. The production function of imperfect substitution, given in equation (1), has a property of diminish but also finite marginal rate of substitution. In this sense, the two positions are both important but not too much. But, notice that imperfect substitution itself cannot reveal the relative importance of different position, or allocations of skills on different positions yet. Consider such a production function: f (q w, q b ) = q w q b. It has a symmetric form and thus the allocation of skills is also unimportant. To solve this, the following property is required. Second, the relative importance of two positions is different, in the sense that elasticities of skills on productivity are different. Our production function satisfies this condition: one percent of increase of skills on the white-collar position will increase the output of each pair of workers by percent, while that of the blue-collar can only increase output by 1 percent. In fact, for two workers of given skills, one is higher and the other is low, we should allocate the higher to the white-collar and lower blue-collar. The allocation of workers on positions does matter..3 An Important Assumption Before providing our formal analysis using the production function above, an important assumption should be raised and explained. We assume regional gaps of skills, including gaps between urban and rural areas and within urban areas. As all know, labors from urban areas have higher skill levels than those from rural areas. It can also be shown that even rural labors that have access to urban labor markets are still poorer in skills than their urban counterparts. As evidence, search table 7 of PSST (004). Schooling years (which is a natural measure of skills) of intra-provincial migrating employment are 8.95 on average, compared with the number of intra-provincial ones, which is 10.3 years. For non-migrants, most of them being farmers, this number is 7.7. Another yet indirect evidence comes from NBSPE, et al. (various years). From it we can calculate that, in year 00, the total average of schooling years of all employment including those in urban and rural areas is 8.1. This number is slightly lower than that of inter-provincial migrants mentioned above, indicating the even lower education level of labors trapped in rural areas. The number for workers from urban areas is 10.0, which is close to that of the intra-provincial migrants mentioned above. This might prove our previous assertions that intra-provincial migrants are very similar to local labors. From these data, we can derive that education levels of rural labors as a whole cannot be higher than 6, with a gap of 4 years. Of course, labors working in rural areas should have poorer education than those go out. But it may not be too much: labors still in rural areas are largely the old; the young should be uniformly educated due to the compulsory 9-year education system. Besides the education (or skill) gaps between rural and urban areas, differences within urban areas, especially between the developed and undeveloped urban areas, also exist. Section 3 of table lists numbers of university students in the east, mid and west areas respectively, as a measure of average education and skill levels of urban workers. The difference is obvious 4. 4 We cannot find a better measure for skill differences of urban labors among regions. NBSPE, et. al (various years) contains data of education composition of employment of different regions, but it does not distinguish between rural and labor employment. Since a large proportion of labors are rural ones, who have in general lower education level, this variable can barely be regarded as a good measure for solely urban labors. However, college education, as a higher level of education, has its students almost from urban areas and thus can reflect educational 8

10 .4 Explaining Occupation and Wage Difference Based on the production function and the basic assumptions, we now ready for analyzing the effects of labor migration. Let s focus on the occupation and wage differences first. The skill level of a typical rural labor (or, potential migrant) is denoted as q = L. An urban labor s skill level is denoted as q = H. Urban labors have higher skill levels than rural ones, i.e., H > L. Consider the situation when the urban-rural separation policy has not been unlocked. Thus rural labors are not allowed to enter urban labor markets. In a typical production process in rural areas, two rural labors, both having low skills, combined with each other and produce an output of L 3. The real wage for each is L 3 /. At urban areas, two urban labors, both having high skills, combined and produce an output of H 3. The real wage for each is H 3 /. Let s assume numbers of potential immigrants are much larger than urban labors 5. Now consider the situation after labor migration. We have the following proposition. Proposition 1. (1) Rural labors will match only with urban labors after entering urban labor markets. () Rural labors will do blue-collar work only, and urban labors will do white-collar work only. (3) Real wage of rural labors will be unchanged, and that of urban labors will be raised. These three conclusions hold if and only if: 3 3 L + H <LH, or 1+ 5 H < L () Let s interpret this proposition. Interpretations of conclusion 1. L 3 + H 3 < LH says that, when a single rural labor and a single urban worker match, their possibly highest aggregated productivity (LH ) should be larger than the sum of individual productivities when they do not match mutually but internally ((L 3 + H 3 )/). Thus mutual match (or, inter-match) is better for both, and should be realized, according to the Coase s theorem. More concretely, imagine that rural labors still choose to match within themselves (called rural-labor self-match ), and urban labors also choose self-match (called urban-labor self-match ). Suggest the following change: exchange the urban worker who works in the urban-labor self-match as a blue-collar, by a rural worker who works in the rural-labors self-match. The (original) urban-labor self-match will then get a lower aggregated productivity by H (H L), but the rural-labor self-match will get a higher aggregated productivity by (H L )L. Such an exchange can find a way to benefit all if and only if the increase more than offsets the decrease. Notice that productivity improvements as a whole do not hold unconditionally. It requires skill gaps between two kinds of labors are not too huge. Otherwise self-matches will still be maintained. Intuitively, if the rural worker has too poor skills, the urban labor as white-collars can be so cumbered that, even if the rural worker asks for an amount as low as L 3 /, which is his/her difference within urban areas. Such a variable may even underestimate the actual education difference, since college graduates usually flow from less-developed areas to developed ones. 5 According to China Statistic Yearbook (004), by the end of 003, there are 60 million urban employments and 490 million rural ones. Thus this assumption is reasonable. 9

11 reservation wage and lower than a high-skill worker by (H 3 - L 3 )/, he/she cannot compensate the high-skill partner if the aggregated productivity decrease, H 3 H L, is even larger, or equivalently, 1+ 5 H > L. Take babysitting as an example. Babysitting is a job requiring both a white-collar and blue-collar role. Baby education is a white-collar job, while holding baby in the arm and transferring are examples of blue-collar jobs. If a rural woman has a pretty good skill, hiring her as a nurse to take the blue-collar role will relieve from slaving one of the parents completely, or both partially, and conduct other white-collar jobs. However, if the nurse is too clumsy, always annoying or even hurting the baby, the quality of babysitting can be seriously downgraded. In such cases, the parents would rather take the full responsibility of blue- and white-collar both, although it seems wasteful of their time. Interpretations of conclusion. As we just said, the self-match within rural or urban labors must be inefficient, thus will be defeated in competitions. Besides, notice that H L > L H always holds. It is impossible, in inter-matches, which rural workers do white-collar jobs and urban workers do blue-collar jobs. Interpretations of conclusion 3. Since we assume that the potential amount of rural labors is much larger than that of urban labors, there must have been left some rural labors that can only self-match within themselves in rural areas. As workers with the same skills, competitions will drive their wages to be the same, i.e., L 3 /, no matter what kinds of match they are in. Thus, urban workers will have a wage level of H L - L 3 /. According to expression (), it is higher than that before labor migrations. Proposition 1 explains why rural labors, when enter into urban labor markets, will work in different occupations from their urban counterparts, and why wage gaps will be enlarged wage levels of rural worker remain unchanged but those of the urban increase. Although equality may deteriorate, efficiency definitely improved. Even more, it is a Pareto improvement since no one is worse off..5 Explaining Regional Gaps Proposition 1 also implies that regional gaps between rural and urban areas are increased with labor flows. We now go one step further to explain not only rural-urban gaps, but gaps between the developed and less-developed urban areas. The fact can be found in section 1 of table. Wage growth of the developed east areas has been significantly higher than that of the mid and west areas. We extend our model to explain this. Like skill gaps between rural and urban labors are the roots of corresponding wage gaps, it s easy to imagine wage gaps between different urban areas be contributed to skill gaps between these areas. Consider labors with three different skill levels: rural labors (with a skill level of L), urban labors in less-developed urban areas (with a skill level of M), and urban labors in developed urban areas (with a skill level of H), H>M>L. We assume the amount of rural labors is larger than the sum of the two types of the urban, which fits China s reality, as we mentioned above. Consider the situation where rural labors flow into both the developed and less-developed urban areas and match with labors there. Rural labors conduct blue-collar jobs and urban labors 10

12 conduct white-collar jobs. We can summarize wage levels of different groups before and after labor flow as table 3. Table 3 Wage levels before and after migrations (1) Rural () Urban labors in (3) Urban labors (3)/() labors less-developed in developed areas areas Before L 3 / M 3 / H 3 / (H/M) 3 After L 3 / M L - L 3 / H L - L 3 / (H L - L 3 /)/(M L - L 3 /) We have the following proposition: Proposition. (1) Rural labors enter into labor markets of both the developed and less-developed urban areas and match only with local urban labors; () Rural labors will do blue-collar work only, and urban labors will do white-collar work only. (3) Wage level of rural labors will not change, and wage level of urban labors in both the developed and less-developed areas will increase. (4) Urban labors in the less-developed areas will not enter labor markets in the developed area. (5) Wage gaps between the urban workers in the less-developed and developed urban areas will be enlarged. These five conclusions hold if and only if 5+ 1 x + x+ 1 xy < min,, y where x H/M, y M/L (thus xy=h/l), satisfying x, y>1. Proof. Conclusions 1 to 3 repeat what we say in proposition 1. By our new denotations, this 5+ 1 requires that xy < and y < Since y > 1, the tight one is xy < 5+ 1 Conclusion 4 requires: H M (H L - L 3 /)+ (M L - L 3 /), solved as xy< y + 1. Conclusion 5 requires (ref. Table 3): (H L - L 3 /)/(M L - L 3 /) > (H/M) 3, solved as xy < x x. Summarizing all these inequalities will lead to our conditions as expression (3). For example, x = 1., y = 1.1 is a pair of solution which satisfies expression (3), but x = 1.1, y = 1. is not. Notice that when x approaches 1, the upper limit (the tight constraint) of y will be the second item of right-hand side of the inequality, i.e., This is exactly the condition for conclusion 5 to hold. When levels of skill are roughly the same between the developed and the less-developed urban areas, if the skill level of rural labors is not much lower than the urban ones (the ratio is less than 1.), wage gaps of urban labors between the developed and the 11

13 less-developed areas will be increased; otherwise (the ratio is bigger than 1.) these gaps will be decreased. When y approaches 1, the upper limit of x would be the third item of right-hand side of the inequality. This is the condition for conclusion 4 to hold. It says that, when the skill levels between rural labors and urban labors in the less-developed areas are very close to each other, if the skill level of urban labors in the developed areas is not much higher than theirs (the ratio is less than 1.41), urban labors in the less-developed areas will not leave for the developed urban areas; otherwise (the ratio is bigger than 1.41) they will 6. As a whole, proposition, based on proposition 1, points out the possibility that labor flow from rural to urban areas can lead to enlargements of income inequality within urban areas. The conditions become more restricted, skill differences (represented by xy = H/L) must be smaller than before..6 Limitations of the Analysis The major limitation of our analysis is that, we consider only the allocation effects of labor flow. We did not consider the amount effects. I.e., when the rural labors get into the urban labor market and match the urban workers, urban workers who had been doing blue-collar jobs before are crowded out, increasing the supply of urban labors and putting pressures to lower wage levels. Following our model, if there are N urban workers originally, half in blue-collar and half in white-collar. When the same amount of rural labors come in, all these urban worker have opportunities to become white-collars, which increase the supply of white-collars by one time and probably decrease wage levels for white-collars. The key is to see whether the productivity (thus the labor demand) increase due to the matching effects can offset the productivity decrease due to the amount effects of labor supply increase. In areas with rapid growth, such as cities of Beijing and Shanghai, or in highly growing industries of all cities, such as real states, or nurse, labor demand has already been rising large enough to offset the labor supply increase caused by amount effects. Thus wage increase of urban worker predicted by the matching effects can be revealed. 3. An Empirical Justification In this part, we try to test the causal effects of labor flow on regional wage. We expect that a region receiving more labors from outside will have a higher wage level. Or, equivalently, a region with a high rate of migrating labor flow will have a higher wage growth. In fact, even if wage does not increase in some regions where labor flow in, our theory cannot be rejected certainly, since it 6 Notice also that expression (3) cannot be used for two extremes, i.e., x = 1 or y = 1. (This is because the item (x - 1) or (y - 1) has been crossed out from both sides of the inequality during our deductions.) In fact, when x = 1 and y >1, there are no difference between urban labors in the developed and less-developed areas. Thus urban labors in the less-developed areas will definitely not enter labor markets in the developed urban areas. Thus the problem degenerates to one specified in proposition 1. All the conclusions in proposition hold if and only if 5+ 1 y <, except that in conclusion 5, the wage gap of urban labors between the developed and less-developed areas does not strictly increase but remains the same. On the other hand, when x > 1and y = 1, proposition holds if and only if 5+ 1 x <, except that in conclusion 3, the wage level of urban labors in the less-developed areas does not strictly increase but remains the same. 1

14 might be the case where the negative amount effects dominate the positive matching effects. But if empirical results really show that labor flow inspires wage increase, our theory must be reasonable. The key issue in the empirical test is the problem of endogeneity (or, simultaneity), i.e., observed positive correlations between labor flow and wage level might reflect the fact that a region with a higher growth of wages attracts more labor flow 7. There are several ways to rule out endogeneities. One is introducing instrumental variables, which are only correlated to labor flow but not wage increase. However, it is very difficult to find a proper instrument immediately so we leave it for future researches. The other way is to adopt a time-series method, such as Grange Test, to identify causalities through time lags. This is also difficult due to lack of continuous time series on labor flows. Yet another way is to include control variables as many as possible, making the explanatory variable, labor flow, not to able to correlate with the residuals. As a first step, this paper uses the third method. There also has been no perfect measure on labor flow until now. Our paper adopts the commonly used way, i.e., to use migrating population scale to reflect that of labor flow. The data comes from the two population censuses on 1990 and 000 (PCO, et. al, 1993, 00). From 1990 data we can find the amount of migration between years 1985 to The method used in the census is to count the people whose current residence is different from that of 5 year ago. The 000 census uses roughly the same way to get the amount of migration between years 1999 to 000. Besides, the 000 census provides the amount of population whose current residence is different from its permanent residence (Hukou). Since the permanent residence is still very permanent until now, we regard these data as a measure of all the migration since the reform era. Table 4 gives the comparative scale of migrations (net immigration population divided by total amount of end-year population). From this, we can see that the scale has been steadily increasing since the reform era. At the end of 1980s, migrating population (either in or out) cannot exceed 5 percent of total population in any province. But at the end of 1990s, many provinces have a proportion higher than 10 percent. The patterns of flow are also quite stable: Guangdong, Shanghai and Beijing are always the major receivers, and Sichuan, Jiangxi and Anhui provinces are always the large senders. Table 5 gives out the basic empirical results. All variables are averaged among years. Thus the regression is based on a cross-sectional data. Regression (1) shows that throughout all years when data is available, the scale of labor flow is highly correlated with wage growth. In regression () several control variables are added, such as comparative scales of investment (the base year is 1978) 8, employment (or, social workers amount), the ratio of foreign investment on total investment (to reflect the quality of investment). All control variables have signs consistent with economic theories and the investment variable is significant. Our labor flow variable keeps being significant: one percentage increase of immigrating proportion will increase annual wage growth by 0.08% percent. Regression (3) includes years from 1985 to The labor flow is still positive, with a larger economic value, but ceases to be significant. The reason might be that before 1990s, the scale of flow is small thus the influence is buried in noises. Regression (4) 7 However, in theory, what determines labor flow is the level rather than growth of wages. However, given wage levels, higher growth rate gives labors higher expectations thus attract more of them. 8 Using comparative rather than absolute scales of investments is because levels of investments themselves do reflect the increase of capital stocks. Although the investment growth in the developed areas has slowed down since the 1990s, we believe their investment levels has by no means been diminished, not to say lower than the less-developed area, due to their accumulations during all the reform years. 13

15 estimates the relationship in more recent year of 1995 to 000; the result is similar with regression (). The effect is even stronger: one percent increase of migration-population ratio leads to 0.15 percent increase of wage growth rate. Regression (5) included an additional control, education levels of labors. In years when data is available, i.e., 1997 to 00, the regression still shows a positive sign of labor flow. As a whole, the empirical results justify that labor flow is favor to regional wage growth 9. 9 Here we believe that wage indexes reflect those of only local workers, excluding immigrants. If not, i.e., these indexes include those of both types; our conclusion can only be underestimated, because immigrants earn lower wages thus work opposite to our interested effects. 14

16 Table 4 Inter-Regional Migration after Reform Province Net Immigration on Total population (%) All yeas Beijing Tianjin Hebei Shanxi Inner Mongolia Liaoning Jilin Heilongjiang Shanghai Jiangsu Zhejiang Anhui Fujian Jiangxi Shandong Henan Hebei Hunan Guangdong Guangxi Hainan Chongqing NA Sichuan Guizhou Yunnan Xizang Shaanxi Gansu Qinghai Ningxia Xinjiang Source: PCO, et. al, (1993, 00); Author s calculations. 15

17 Table 5 The Effect of Labor Flow on Wage Growth Explained Variable: Wage Growth Rate Years (1) () (3) (4) (5) Explanatory Variables Net Immigration Scale *** 1.0 *** ** (0.36) (0.376) (0.903) (0.35) (0.389) Control Variables Investment (year 1978=1) ** (0.000) (0.00) (0.001) (0.001) Foreign Inv./ Total Investment *** (0.0) (0.036) (0.037) (0.047) Employment Growth (0.169) (0.37) (0.141) (0.55) Growth of Labors Education Level (0.316) Adjusted R Samples Note: Standard Errors are in parentheses. *** Significant on 1% level; ** significant on 5% level. Sources: For wage growth see table ; for immigration scale see table 4. Others are from: NBS (various years); NBSA (1997), NBSC (1990). 4. Implications This paper has several important implications. 4.1 Skill, Occupation and Wage Difference Classical wage theory notice that skill differences can lead to wage differences, but often ignore that skill differences can also lead to occupational differences high-skill workers do white-collar jobs, while low-skill workers do blue-collar jobs. These occupation differences have always been contributed to irrational economic behaviors such as cultural impediments, or purely policy discriminations. Our analysis shows that, these differences indeed have economic rationalities. The consequences of occupation differences are various. For the positive side, although occupation differences exist, they do not segregate urban workers from their rural counterparts: they are even willing to work with each other. In fact, divergence of occupations as well as commonality through working in the same firm is the two sides of the same coin, showing the essence of goodness of labor division. Combined production results in extra benefits. It is better for learn from and interact with each other. If on the contrary such a combination cannot be realized, results might be bad for both. For example, when the Chinese New Year comes and rural labors fly back to their homeland, many firms will be forced to shut down. If you believe such 16

18 shutdowns are only for vacations, consider families raising children. They cannot shutdown at any time so are definitely hurt. For the negative side, although people work together, they belong to different labor market due to different occupations. Such a phenomenon has been long recognized by economists and coined as dual labor markets. Although dual labor markets themselves are not discriminations, they pave ways for discrimination. Imagine that, if physical workers in construction industries are mixture of urban and rural workers, wage defaults may not be able to prevail. Since urban labors have their locally social capital, it can be hard for the employer to risk defaults a discriminative default sounds difficult for implementations. These two sides show a complex an arm s length relationship between rural and urban workers, a basic picture of labor markets in transitional China. For Government policies, we need not worry about occupation differences themselves, because they are not discriminations at all. However, we still need to watch discriminations like wage defaults, resulting from labor market separations. Besides, this paper implies that, regional gaps roots in skill gaps among regions thus can only be solved through eliminations of education gaps. This comes back to classical economic theories, however, with additional power: Don t expect labor flow itself can resolve regional gaps; in fact, if education and skill gaps persist, labor flow might enlarge rather than diminish regional gaps. 4. Reform without Losers Rural workers allowed to work in towns is a big policy reform in transitional China, featured by a bundle of reforms concerning residence and employment. Generally, a key issue in reforms is whether the reforms can make all people better off thus is Pareto Improvements. Because such reforms, compared with those only increasing total welfare, have more favorable political feasibilities, since they have well balances between efficiency and equity. In Lau, Qian & Roland (1998), they proposed that price dual track (including employment dual track) is close to a reform without losers. In our analysis, free policy on urban-rural labor markets can also be regarded as one it raises wage levels of urban workers, yet not decreases those of rural immigrants and non-immigrants. More interestingly, the pattern looks like a dual track arrangement a dual labor market. But such a dual track is more like a natural choice of economic selfishness than deliberate policy innovation - governments expect at most efficiency improvement but not a Pareto one Economic Development and Income Inequality Our analysis also adds new elements for discussions on economic development and income inequality. We show that, economic developments, especially those driven by resource re-allocations, can lead to income inequality instead of equality. However, we believe such an inequality has no reason to be regarded ill, since it is a natural choice of economic rationalities, not a compulsive redistribution policy (such as a distorted taxation policy). 5. Conclusions 17

19 The model and theory of this paper explains how rural immigrants, through division of labors, bring in occupation differences: urban workers doing white-collar jobs, rural workers doing blue-collar jobs. And also predicts that, labor flow can be favorable shocks for all the people in labor markets, either coming from rural areas or residing in urban areas, either in developed urban areas or less-developed ones. This paper adds new elements to classical theory advocating free-market ways of resource allocations: The free flow of labors has efficiency gains other than allocations themselves; it can improve divisions of labors thus productivities. Even more, if such an effect is strong enough, labor flow can even result in Pareto improvement and avoid political conflicts prevailing in free trades. 18