【原创】Escaping Low-Level Equilibrium of Urbanization

Escaping Low-Level Equilibrium of Urbanization

-- Institutional Promotion, Social Interaction and Labor Migration

Chen Zhao (陈钊)1,  Jiang Shiqing (蒋仕卿)2, Lu Ming (陆铭)3 and Hiroshi Sato (佐藤宏)4

1 Center for China Economic Studies, Fudan University, Shanghai, China

2 Fudan University; Industrial Securities Co., Ltd, Shanghai

3 Shanghai Jiaotong University; Fudan University, Shanghai

4 Department of Economics, Hitotsubashi University, Tokyo, Japan

Abstract: The policies resulting in urban-rural segmentation have not only directly impeded the transfer of labor from rural to urban areas, but also trapped the process of urbanization in a low-level equilibrium, as the negative effects are amplified by interdependencies between interpersonal decisions. This paper finds from the CHIPS2002 data that there is interdependence of rural residents’ decision-making on labor migration to urban areas, and the interdependence is strengthened by the exchange of information between rural residents. According to the simulation results of the models in this paper, China’s rural-urban labor migration is indeed at a low-level equilibrium. To get rid of the low-level equilibrium, in addition to improving the level of education and promotion of information exchange among rural residents, it is more important to implement “big push” policies to eliminate institutional barriers to labor mobility and accelerate urbanization with the social interaction.

Keywords: urbanization, labor mobility, social interaction, peer effect

JEL Classification: J61, O15, R23 

1.Introduction

Labor transfer from rural to urban areas constitutes the only definition of urbanization according to international standards, and also an important indicator of economic development. Public policies on the promotion of labor mobility provide a structural impetus to the sustainable growth of an economy, developing economies in particular. However, in China, urbanization is fast but lags far behind its industrialization, although the expansion of urban area is much faster than the growth of non-agricultural population (Lu, 2010; 2011). These phenomena prompted us to deeply explore the determinants of labor mobility.

Interdependence of decision-making is an important cause of multiple equilibriums in an economy, and it is difficult for an economy to get rid of a low level equilibrium once it falls into the trap (Zanella, 2004). Based on this principle, we attempt to answer two related questions in this paper: first, how does the peer effect (interdependence of decision-making between neighbours) affect the decision on working in cities? Second, how does the heterogeneous social distance affect the impact of peer effect? Using CHIPS2002 data, or data collected in the Chinese Household Income Project 2002 by the Chinese Academy of Social Sciences, we find that the peer effect can affect the decision of rural residents to seek employment in cities in a significant way. We also construct models to study the heterogeneous peer effect on decision of labor mobility. The study finds that a higher frequency of interaction with other villagers for exchange of information is conducive to enhancing the peer effect, but a higher frequency of time-consuming mutual assistance weakens the positive effects of the peer effect.

The existence of peer effect and the corresponding social multiplier (Glaeser et al., 2003) amplifies the impact of one particular factor on the market outcomes and leads to the multiple equilibrium in the market through behavioral dependence between people. According to the simulation results of the model, China’s urbanization features a low-level equilibrium of labor mobility. Policy simulation of this paper shows that improved level of education can promote labor mobility, but the impact is negligible. Methods like enhancing the exchange of information among the rural residents on migrant working and providing effective public services to the villagers to reduce mutual assistance of rural residents would be more effective. However, these policy measures will not be helpful when the labor mobility is already in a low-level equilibrium trap. Only by formulating institutional “big push” policies, eliminating urban-rural divide and public policies favoring urban areas and accelerating social integration between urban and rural areas can China effectively get rid of low-level equilibrium of labor mobility and promote the urbanization process vigorously.

This paper is structured as follows: Part 2 is the literature review of labor mobility and peer effect. In Part 3, the authors construct models to describe how social interactions affect the peer effect. Part 4 is the data description and variable definitions. Part 5 is the empirical model and results. Part 6 is a robustness test on the empirical results. In Part 7, the authors simulate the equilibrium of labor mobility and the relative effects of different policies based on parameters obtained from the empirical model. The final part is the summary.

2. Literature Review

Early literature mainly focuses on how the characteristics of individual and family influence the labor mobility. Zhao (1999) finds that a range of personal and family characteristics affects migrant working ratio. For example, the ratio of women participating in migrant working is significantly lower than men; the older the age, the lower the ratio; and the ratio of family with more land per capita is significantly lower. Zhu (2002) finds that the income difference between the agricultural and non-agricultural employment significantly affects the ratio of migrant working, which also fits the Harris-Todaro model. Cai (2003) finds that more workers are migrating from central regions to eastern regions than from the western regions to eastern regions, concluding that the distance affects labor mobility.

Recently, researchers are increasingly focused on the role of social relationships and social networks in the labor mobility. Munshi (2003) finds that social networks play an important role in the migration of Mexico residents to the U. S.. Mckenzie and Rapoport (2010) believe that with the expansion of social networks, more and more low-income families are participating in the out migration, thus help reduce rural poverty. By using data on China, Zhang and Li (2003) find that having friends or family members outside the village will increase the ratio of non-farm employment of rural residents. Bao et al. (2007) finds that the “scale of fellow-villagers” in the labor migration destination will significantly improve the overall labor mobility among provinces. Zhao (2003) finds that the total number of migrant workers of a village will significantly affect an individual’s decision on migrant working. Zhao believes that the presence of a migrant network can reduce the psychological costs and costs of searching information.

Why do we carry out a special study on the impact of peer effect on decision-making on labor mobility? The reason is that it is the only way to explain the low-level economic equilibrium. The network can be divided into bonding and bridging networks. The friends and family members outside the village are the bridging network, while the social network within a village is the bonding network that can reduce the cost of migration by providing information. “Peer effect” occurs between the members of a bonding network, but it is different from the “network effect”, as it only takes place when there is interdependence of people’s behaviour. The peer effect is caused by insufficient information and people’s social psychology of not being different. For a network effect to work, other network members need to act first and disseminate the information. However, the “peer effect” based on the social psychology can take place without much time. Bauer et al. (2002) and Araujo et al. (2004) uncovered separately the peer effect on migration of Mexican farmers to Mexican cities and to the U.S.. In this paper, we control the social networks outside the village and the network effect represented by historical mobility ratios, and find that there are still peer effects in making labor-mobility-related decisions. Peer effect has been widely captured in other researches on social and economic behaviors (Durlauf & Fafchamps, 2004); however, there is little literature on the empirical research on the heterogeneity of the peer effect. This is why this paper carries out a more in-depth study on how the type and frequency of social interactions among the villagers affect labor mobility-related decisions.

3. A Model of the Effect of Social Interaction and Peer Effect

Our model is mainly based on the network model of Ballester et al. (2006). We simplify some of the assumptions in the model, but assume that the social distance is a function of the type and frequency of social interaction.

Questionnaires were collected in February 2003 during the Chinese New Year, so the data covered both migrant workers and local residents comprehensively. The data contain personal information such as gender, age, education, work status, the family demographic and economic conditions, as well as the geographic and demographic characteristics of villages. The data also include social interaction data at the household level, which provides a reliable data source for our research.

The dependent variable of “choosing migrant working or not” is a 0-1 variable. By defining the dependent variable as a discrete variable, we can avoid the reflection problem in identification. The reflection problem is a technical difficulty in peer effect identification proposed by Manski (1993). In simple terms, in linear models, the personal characteristics affect the dependent variables “linearly”. The mean value of personal characteristics and that of the dependent variables (both being the measurements of peer effect) are completely collinear. Therefore, if we both control the personal characteristics and the peer effect, the coefficient of the peer effect would not be identified. Fortunately Brock and Durlauf (2001) proved that the reflection problem can be avoided in nonlinear models, as in nonlinear models such as Probit and Logit, the personal characteristics affect the dependent variables in a non-linear manner.

The CHIPS2002 data registered the number of days when an individual was away from home by means of self-report. Due to data limitations, we assume that cities are the only destination for migration. We define individuals who are away from home for more than six months (180 days) as migrant workers, and removed students who are away for a long time for education and farmers who work at local township enterprises. The biggest deletion of the raw data is that we removed males under 16 years old and over 60 as well as female under 16 years old and over 55 from the sample (a total of 11,404 observations). The reason is that these individuals are not part of the working age population according to China’s labor statistics, and therefore not the object of this study. Finally, our data remove some samples with missing variables or abnormal values. The above processing leaves us with a total of 16,401 valid observations.

In the CHIPS2002 data, we can get the total population of the villagers and the total population of migrant workers of 2002 from the village-level questionnaire. The measurement of the peer effect is: 

as the intensity of social interaction. A family with more social interactions with other villagers will be under bigger influences of the latter. This paper divides social interaction into two types: interaction for information sharing and interaction for labor mutual assistance. In Chinese rural areas, widening income gap between urban and rural areas makes migrant working an effective way to increase income. As there is a split between urban and rural labor markets, a large number of rural laborers have to find work in other provinces, and depend on the exchange of information with other peers to search for better job destinations and higher incomes. The labor services market is still underdeveloped in rural areas, and the villagers usually share labor information among themselves. CHIPS2002 data contain information on the two types of social interaction of rural residents: the labor mutual assistance (“help each other out during the farming season”), and the interaction for information sharing. A series of indicators in the CHIPS2002 data have recorded the intensity of social interaction among a family and the relatives and neighbors. The answers to these questions are (1) very frequently (2) often, (3) just so-so, (4) sometimes, (5) none/few. We divide these indicators into two groups of base indicators to interact with the peer effect to capture the heterogeneous peer effect. We use the continuously measured variables of social interactions as the reference model to get the most concise results, and use the society interaction model measured discretely for robustness test.

All explanatory variables are defined in Table 1, and the statistics description is in Table 2. In Table 2, we can see that of all the 16,401 samples, 2,675 samples, or 16.31% participated in the migrant working in 2002. We can identify the difference between the migrant workers and non-migrant workers even in the most basic descriptive statistics. In the sample of migrant workers, 50.24% are unmarried, compared with the 25.21% of non-migrant workers. The average age of migrant workers is 27.1 years, younger than the average age of 36.1 of non-migrant workers. We will control all the explanatory variables in the regression model. However, in the regression analysis, we focus on degree of peer effect and the effect of the interaction terms.

The regression results are in Table 3. Equation (1) is the benchmark model, and we only control the characteristics of individuals and families. We control the ratio of migrant working in the village (peer effect), and the interaction term between peer effect and social interaction in Equation (2). The goodness of fit increases from 0.1828 to 0.2136, which means that the peer effect plays a significant role in the decision on migrant working. We are concerned that the potential bias of omitted variables may lead to biased estimation of peer effect; therefore, we have included in the equation (3) the data of the ratios of migrant working in 1998, which were provided by villagers based on their memories. It takes time for the network effect to take place; therefore, we assume that the ratio of migrant working in 1998 is based on controlled network effect. As a result, the migration ratio of 2002 can better reflect the peer effect. We control the village-level characteristics in equation (4). Equation (5) adds the dummy variables to avoid the bias of omitted variables in the estimation.

The regression results from the equations (2) to (5) show that the peer effect is positive. Compared to the equation (2), the coefficient of peer effect in equation (3) falls from 1.503 to 0.65. The coefficient of the ratio of migrant working of 1998 is highly significant, and three times the ratio in 2002. This finding is consistent with that of Munshi (2003): the migrating ratio in the past has a greater impact than the current ratio. Intuitively, more information will be accumulated in the network with the passage of time. When we control the dummy variables in the equation (5), the peer effect becomes less significant, but the interaction term between peer effect and social interaction remains highly significant.

We are more interested in the interaction between peer effect and social interaction. All interaction terms are highly significant from equations (2) to (5). We can get the following conclusions on the significance and direction of the variables: (1) for the social interaction, the more frequent the interaction, the greater the marginal effect of the peer effect. (2) For the labor mutual assistance, the more frequent the interaction, the weaker the marginal effect of the peer effect. The two different directions of social interaction can be explained as follows: information sharing enhances the peer effect. However, although mutual assistance reduces social distance, it takes up the time of migrant working, and thereby weakens the positive peer effect. If the negative interaction term between the peer effect and social interaction is the result of the time constraints faced by villagers who provide mutual assistance, then we need to consider seriously the causal relationship between migrant working and mutual assistance. Fortunately, when we use discrete social interaction variables, we find that not everyone has a lot of interaction with neighbors. We find in the following robustness test that a certain degree of labor mutual assistance can help enhance the positive effects of the peer effect.

Based on regression (4), the marginal effect of the peer effect is 0.1413 for a representative individual with a moderate level of social interaction and mean value of other characteristics. In other words, a one-percentage-point rise in the ratio of migrant working in the village will increase the ratio of engaging in migrant working of an individual by 0.1413%. The regression results confirm the existence of the peer effect in decision-making on migrant working. However, the positive correlation between the ratio of migrant working in the village and the ratio of engaging in migrant working of an individual may cause a low level equilibrium of labor mobility. The urban-rural division policy makes the villagers less willing to participate in migrant working, and the negative effects of this policy are amplified by the mutual influence of villagers’ decision-making.

Results of other variables are similar to the previous study. Our interpretation of the regression coefficients is based on equation (4) .

(1) Personal characteristics significantly affect the decision on migrant working. The ratio of women engaging in migrant working is 4.45% lower than that of men. The ratio of married individuals significantly dropped by 11.76%. The impact of age on migrant working is inverted U–shaped. When other factors are the same, 33-year-old residents are most likely to participate in migrant working. The ratio of migrant working increases with age for individuals under the age of 33, and decreases with age for individuals over the age of 33. Zhao (2003) finds that all levels of education do not increase the ratio significantly. We find that, compared to illiterate people, the different levels of education will significantly increase the ratio of migrant working. Different levels of education have different impact on the ratio. Junior high school graduates are more likely to work out (7.13%), followed by primary school students (5.25%) and individuals with a higher education (4.75% for technical school or college graduates, and 4.9% of senior high school graduates). According to the study, higher education doesn’t necessarily increase the ratio of migrant working of the rural labor force. Of course, another reason might be that rural residents with junior high school education or better education are more likely to obtain urban household registration and no longer be listed as rural residents engaged in “migrant working”. Another reason is that villagers with higher education are more likely to participate in local non-farm employment (Zhao, 1999), and in our regressions, the local non-farm employment are not counted as the migrant population.

(2) The family characteristics also significantly affect the decisions on migrant working. The ratio of an individual engaged in migrant working will increase by 4.01% if the family has an additional laborer. The land per capita in a household can significantly reduce the ratio of migrant working, as the labor engagement of migrant working and farming is replaceable. A family with more land per capita has higher marginal gains in agricultural production, and is less likely to work out. The demographic structure of the family also has an impact. The ratio of migrant working decreases significantly for families with children aged 6 to 12 (the marginal effect being -0.98 %). The ratio of migrant working doesn’t change much for families with senior citizens aged 65 and above, but the regression coefficient is positive. The network of relationships of a family increases the ratio, and having family members, relatives or friends engaged in migrant working can increase the ratio significantly by 1.2%. Having family members, relatives or friends working as leaders at village level can significantly increase the ratio by 1.03%.

(3) Village characteristics also affect the decisions on migrant working. An increase of RMB 1,000 in the average income of the villagers will reduce the ratio by 0.92 percent. Residents from mountainous and hilly areas are more likely to do migrant working than those in plain areas. The two dummy variables may reflect the harsh living conditions of the mountainous and hilly areas. The distance between the village to the nearest traffic station and to the county seat do not significantly affect the decision on migrant working.

6. Robustness Testing

In the previous section, we make use of the interaction term between continuous social interaction and the peer effect. In order to verify whether the interaction in labor market always weakens the peer effect, we use discrete interaction in the labor market to interact with the peer effect. The hypothesis is that only those villagers engaged in high-frequency interaction in labor market face time constraints, hereby reducing the positive effects of the peer effect. Those engaged in low-frequency interaction are not subject to time constraints, and the positive peer effect would be enhanced because of reduced social distance.

We report the interaction terms between controlled discrete social interaction and peer effect in Table 4. The results show that when we put the peer effect and the interaction terms in the regression equation, the goodness of fit increases from 0.1828 in equation (1) to 0.2149 in the equation (6). We arrive at the following conclusions from the equations (7) to (9): (1) for the social interaction on information sharing, the higher the frequency of information sharing, the greater the positive effects of the peer effect. The low-frequency information-sharing does not increase the positive effects of the peer effect when the village-level variables are controlled. (2) for the labor mutual assistance, the higher the frequency of interaction, the weaker the positive effects of the peer effect. Interestingly, the low-frequency mutual assistance enhances the positive effects of the peer effect. The intuitive explanation is that people engaging in moderate mutual assistance are not faced with time constraints, and the reduced social distance can help enhance the positive effects of the peer effect. Here, the time constraint is not a tight constraint, so the simultaneous endogeneity is not an important issue.

In Table 3 and 4, when the 1998 ratio of migrant working is controlled, the coefficient and significance of the peer effect decrease greatly. Therefore, we need to test whether the 1998 ratio of migrant working is a better measure of the peer effect. We interact the variable of the ratio of migrant working with the variable of social interaction of 1998, and repeat the above regression. Results are shown in Table 5. Based on our theoretical analysis, if the 1998 ratio of migrant working is a better measure of the peer effect, its interaction term with the social interaction should be significant. By comparing with the results in Table 3, we can see that most of the regression results in Table 5 are robust. The only difference is that the interaction terms in Table 5 become less significant. In equation (11), we control the interaction between discrete social interaction and the peer effect as well as the dummy variables of the village, and most of the interaction terms become less significant, but the direction of the coefficients is consistent with that of table 3. In conclusion, the heterogeneous peer effect is more pronounced when controlling the current ratio. This is understandable, as theoretically, the peer effect occurs in the current period, and the impact depends on the social distance influenced by social interaction.

Note: *, **, and *** means that the coefficient is significant under the significance of 10%, 5% and 1%. Standard errors are reported in parentheses. 

7. Peer Effect and Public Policy: Why Do We Need the “Big Push” at the Institutional Level?

The empirical results prove the existence of the peer effect in Chinese rural labor mobility. Meanwhile, the effect has a non-linear nature, as it is weakened by more interaction in the labor market and enhanced by interaction of information exchange. The existence of heterogeneity of the peer effect has rich policy implications. Theoretically, the existence of the peer effect means that there might be multiple equilibrium in the economic processes: when the average behavior of a community (group) is at a low level, the economic process might converge to a low level of equilibrium, and when the average behavior exceeds a certain level, the economic process might converge to a high level of equilibrium due to social interactions (Zanella, 2004). Under the theme of this paper, if there is a low level of equilibrium in labor mobility, the equilibrium will be very unfavorable to the process of urbanization in China. We use the parameters of the equation (8) in Table 4 to simulate the equilibrium in decision-making on migrant working. In Figure 1, we establish the numerical simulation chart of the average ratio of migrant working in a village and of an individual. X-axis represents the ratio of migrant working in a village, and the Y-axis represents the ratio of an individual. The solid line is the 45-degree line, representing the equilibrium when the two ratios are the same. The dotted line is the response curve of an individual, representing the relationship between the two ratios when other explanatory variables are of mean value. We can see that there is only one intersection between the curve of the ratio of migrant working of an individual and the 45-degree line, and the corresponding ratio of a village is 8.56%. The slope of the response curve at the intersection is less than 1, indicating a steady-state of equilibrium. The ratio density function of Probit model is the standard normal distribution, and the cumulative distribution function is S-shaped, therefore, the level of equilibrium is decided by the intersection of the 45-degree line and the response curve. If the intersection is below the ratio of 50% of the migrant working of an individual, then there is a low-level equilibrium. In this case, the deviation within a certain range around the equilibrium point will converge to the low-level equilibrium by dynamic adjustment. If the intersection is above the ratio of 50% of the migrant working of an individual, then there is a high-level equilibrium. It is also a steady-state of equilibrium, and deviation within a certain range around the equilibrium point will converge by dynamic adjustment to it as well. It can be seen from the figure that the intersection of the obtained numerical simulation and the 45-degree line is below the S-curve. It means that if the values of the model parameters keep the same, even if the ratio of migrant working in a village rises (moving along the curve), the ratio of labor mobility will also eventually converge to the “low-level equilibrium trap” because of the peer effect (Moffitt, 2001). 

Figure 1 shows the relationship between village migration ratio and mean individual out migration probability (simulation parameters are from table 3). When the two values equal (cut the 45 degree line), it is the equilibrium migration ratio. As shown in the graph, the equilibrium migration ratio is 8.56%.

Figure 2 shows the overall policy effect of increasing both education level and pro-peer effect social interaction. The combining policy will lift up equilibrium migration ratio to 11.89%.

Figure 3 shows the effect of rural-urban labor market integration on out migration decision (long dash line). Though in our framework we do not have explicit parameters to measure the extent of labor market discrimination against rural migrants, we increase the intercept term, which is exogenous and homogenous to every sample individual and thus can represent the “institutional change”, to demonstrate the effect of market integration. We increase intercept from -4.6672 to -4.1255 and the equilibrium migration ratio reaches 50%.

Enhanced labor mobility will not only increase villagers’ income, but also reduce the income gap between urban and rural areas. At the same time, the migration of rural labor to urban areas is conducive to China’s economic growth. Therefore, the policy objectives should be to promote the transfer of labor from rural to urban areas. To this end, we distinguish three levels of policy measures and simulate their effect.

The first policy is to shift the policy reaction function, that is, issue policies to increase the ratio of migrant working without affecting the social interaction among the villagers, and thereby not changing the slope of the S curve. Using the model language, what we change is other explanatory variables with the exception of the intensity of social interaction. Of all the variables we control, only the educational level can be significantly altered by economic policy. We assume that the public policy aims to improve the level of education of the villagers so that all villagers who are illiterate or primary school graduates can receive a junior middle school education. We know from the regression results that improved level of education can increase the ratio of migrant working. In fact, the results of the policy simulations show that the ratio of migrant working at equilibrium is 9.47% (Figure 2). Therefore, improving the education level of the villagers can indeed promote labor mobility. However, the impact of this policy is small, and the intersection is still at a low-level equilibrium.

The second policy is to improve the social interaction in favor of migrant working among villagers to enhance the peer effect. In Figure 2, this means that the intercept of the response function remains the same, but the curve rotates counterclockwise. Our study has found the heterogeneity in the peer effect, which is that the peer effect depends on the intensity of the interaction among the villagers. Therefore, if we increase the social interactions that can enhance the peer effect, and reduce social interactions that weaken the effect, we will be able to increase labor mobility. If the policy measures can facilitate better information exchange among villagers (we define the state of “exchange of migrant working information” as “a lot”), and establish labor service market in rural areas to reduce the interaction among the villagers in the labor market (we define the state of three variables on the intensity of labor force interaction as “little or no”), we can see that the slope of the curve of migrant working will be significantly increased to intersect with the 45-degree line at a higher position. The corresponding average ratio of migrant working of the village would be 10.51% (Figure 2). However, the intersection is still at a low-level equilibrium.

The combination of the two policies yields a ratio of 11.89%, still a low-level equilibrium (Figure 2). In other words, we must find new policies and measures to get rid of the low-level equilibrium of labor migration and urbanization.

To get rid of the low-level equilibrium of labor mobility, one way is to expand the income gap between urban and rural areas. The reaction curve will move up if the average income of the villager is lower. The intersection will be a high-level equilibrium when the income gap between urban and rural areas is wide enough. At this time, due to peer effect, the positive impact of the slight increase in labor mobility will naturally converge at a high level of equilibrium. However, there is a price to pay for the widening income gap between urban and rural areas. The gap has been already very large, and its growth will directly harm the economic growth (Lu Ming, etc., 2005; Wan et al, 2006). To switch the low-level equilibrium of labor mobility to a high-level one, a most important priority is to promote the integration of urban and rural labor markets through reform of system, which can also improve the intercept of the response curve. In fact, the labor mobility between rural and urban is essentially based on free decision-making, but the governments’ economic policy that favors cities have led to widespread discrimination against rural migrant workers in urban areas. As a result, there is still a division between the urban and rural labor markets. (Zhao and Lu, 2008). If we can eliminate the city-favoring economic policy to promote the convergence of rural and urban residents, we can raise the expected income of rural migrant workers and ratio of migrant working. We carry out a simulation of the effect of the policy. In Figure 3, on the basis of improving the education levels of the rural labor force and favorable social interaction, we increase the intercept from -4.6672 to -4.1255, that is, an increase of 0.5417, and the balanced labor mobility ratio can be increased to 50%, which is the threshold point of a high-level equilibrium of labor mobility. If the high-level equilibrium appears in the figure, then with the peer effect, a small positive impact favorable to increasing the labor mobility can make the ratio converge at a high-level equilibrium. Eliminating economic policies resulting in urban-rural division, promoting social inclusion of China’s urban and rural labor force to enhance labor mobility and urbanization are the necessary means to build a harmonious society and promote sustained economic growth. This paper emphasizes the jump from a low-level to a high-level equilibrium, which aims to call for an institutional change to promote urban-rural integration and the “big push” policies on urbanization. The education-oriented policies are economic policies, policies on the changing intensity of social interactions are social policies, and the system-level reform is a political policy.

8. Conclusions

The paper examines the existence and the impacts of peer effect in decision on labor mobility. The empirical results can be summarized as the following: (1) there is peer effect in the decision-making related to labor mobility, and its effects are still significant after we control the historical labor mobility ratio (network effect) as well as other social capital. (2) There is heterogeneity in the peer effect. The peer effect is more significant in families participating in more information-sharing and less labor mutual assistance. More labor force mutual assistance weakens the positive effect of the peer effect.

China’s low level of urbanization is very disproportionate to the high level of industrialization. Urbanization and rural-urban migration are not only the key to narrowing the income gap, but also conducive to promoting China’s sustained economic growth. With information asymmetry in the institutional environment of labor mobility, there exists peer effect in the decision related to labor mobility. Urban-rural division policies reduce labor mobility, and the negative effects are enlarged through social interaction, “trapping” the long-term labor mobility at a low-level equilibrium. This will increase the income gap between urban and rural areas and harm the development of a social harmony.

The findings of this paper have important policy implications of eliminating the low-level equilibrium of China’s urbanization and labor mobility. First of all, the effect of policies increasing ratio of migrant working will be amplified by the social interaction. Secondly, the promotion of information sharing and replacing labor mutual assistance with more effective social services will help social multiplier play a greater role in promoting urban-rural labor mobility. Thirdly, when there is peer effect, the labor mobility of a society might be at a low-level equilibrium and the equilibrium can only be eliminated through the institutional “big push”. Based on China’s reality, the government should reform the systems of land, household registration, social security and public services to eliminate the segmentation of urban and rural labor markets and promote social integration. These measures will be more conducive to urbanization than investing in human capital and increasing social interaction.

References:

(1) Araujo, Caridad, Alain de Janvry, and Elisabeth Sadoulet. 2004. “Peer Effects in Employment: Results from Mexico’s Poor Rural Communities.” University of California at Berkeley, Working Paper.

(2) Ballester, Coralio, Antoni Calvó-Armengol, and Yves Zenou. 2006. “Who’s Who in Networks, Wanted: The Key player.” Econometrica, 74(5):1403–1417.

(3) Bao, Shuming, Örn B. Bodvarsson, Jack W. Hou, and Yaohui Zhao. 2007. “Interprovincial Migration in China: The Effects of Investment and Migrant Networks.” IZA, Discussion Paper No. 2924.

(4) Bauer, Thomas, Gil Epstein, and Ira N. Gang. 2002. “Herd Effects or Migration Networks? The Location Choice of Mexican Immigrants in the U.S.” IZA, Discussion Paper No. 551. 

(5) Brock, William A., and Steven N. Durlauf. 2001. “Discrete Choice with Social Interactions.” Review of Economic Studies, 68(2): 235-260.

(6) Cai, Fang, Meiyan Wang, and Yang Du. 2003. Political Economics of Labor Flow. Shanghai:  Shanghai Joint Publishing Press.

(7) Calvó-Armengol A., and Yves Zenou. 2004. “Social Networks and Crime Decisions: The Role of Social Structure in Facilitating Delinquent Behavior.” International Economic Review, 45(3): 939-958.

(8) Chen, Zhao, and Ming Lu. 2008. “From Segmentation to Integration: Political Economics of Urban-rural Economic Growth and Social Harmony.” The Journal of Economic Research, 1: 21-32.

(9) Durlauf, Moffitt S., and P. Young, ed. 2001. Policy Interventions, Low-Level Equilibria, and Social Interactions. Cambridge: MIT Press.

(10) Durlauf, Steven N., and Marcel Fafchamps. 2004. “Social Capital.” NBER, Working Papers 10485.

(11) Glaeser, Edward L., Bruce I. Sacerdote, and Jose A. Scheinkman. 2003. “The Social Multiplier.” Journal of the European Economic Association, 1(2-3): 345-53.

(12) Gustafsson, Björn, Shi Li, and Terry Sicular. 2008. Inequality and Public Policy in China. New York: Cambridge University Press.

(13) Lu, Ming, Zhao Chen, and Guanghua Wan. 2005. “Inequality out of Scarcity- Interactions of China’s Income Gap, Investment, Education and Growth.” The Journal of Economic Research, 12: 4-14.

(14) Lu, Ming. 2010. “Cross-regional Redistribution of Construction Land Rights- New Driver for China’s Economic Growth.” The Journal of World Economy, 1: 107-125.

(15) ——. 2010. “Restructuring Urban System- On the Sustainable Development Strategy of China’s Region and Cities.” Journal of Nanjing University, 5:15-26.

(16) Manski, Charles F. 1993. “Identification of Endogenous Social Effects: The Reflection Problem.” Review of Economic Studies, 60(3): 531-542.

(17) McKenzie, David, and Hillel Rapoport. 2010. “Self-Selection Patterns in Mexico-U.S. Migration: The Role of Migration Networks.” The Review of Economics and Statistics, 92(4): 811-21.

(18) Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market.” Quarterly Journal of Economics, 118(2): 549-599.

(19) Patacchini, Eleonora, and Yves Zenou. 2008. “Juvenile Delinquency and Conformism.” Stockholm University, Working paper.

(20) Wan, Guanghua, Ming Lu, and Zhao Chen. 2006. “The Inequality-Growth Nexus in the Short and Long Runs: Empirical Evidence from China.” Journal of Comparative Economics, 34(4): 654-667.

(21) Zanella, Giulio. 2004. “Social Interactions and Economic Behavior.” University of Siena, Working Paper.

(22) Zhang, Xiaobo, and Guo Li. 2003. “Does Guanxi Matter to Nonfarm Employment?” Journal of Comparative Economics, 31(2): 315–331.

(23) Zhao, Yaohui. 1999. “Labor Migration and Earnings Differences: The Case of Rural China.” Economic Development and Cultural Change, 47(4): 767-782.

(24) ——. 1999. “Leaving the Countryside: Rural-To-Urban Migration Decisions in China.” American Economic Association Papers and Proceedings, 89(2): 281-286.

(25) ——. 2003. “The Role of Migrant Networks in Labor Migration: The Case of China.” Contemporary Economic Policy, 21(4): 500-511.

(26) Zhu, Nong. 2002. “The Impact of Income Gaps on Migration Decisions in China.” China Economic Review, 13(2-3): 213-230.