The Effects of Migration on Collective Action in the Commons: Evidence from Rural China

World Development Vol. 88, pp. 79–93, 2016 0305-750X/Ó 2016 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2016.07.014

The Effects of Migration on Collective Action in the Commons: Evidence from Rural Chinaq YAHUA WANG a,1, CHUNLIANG CHEN b and EDUARDO ARARAL c,* a Tsinghua University, China b Development Research Center of the State Council, China c National University of Singapore, Singapore Summary. — Over the past three decades, scholars have studied the effects of more than three dozen factors on collective action in the commons but little is known about the effects of rural to urban migration. We examine this question with the case of China, which has the world’s most extensive levels of rural to urban migration. Using OLS, Logit and Probit models and data from a survey of 1,780 households from 18 provinces, we find that migration has a statistically significant adverse effect on collective irrigation controlling for a large number of theoretically relevant variables. The effects of migration on collective action in the commons are possibly mediated by a number of factors frequently identified in the literature, including leadership, social capital, sense of community, economic heterogeneity, and dependence on resources. We speculate that massive out migration partly explains the significant drop in the use of collective canal irrigation and exacerbated the significant increase in groundwater irrigation since the start of reforms in 1980s. These findings have important policy implications for commons governance in China given that massive rural to urban migration will continue in the next decade. Because of the increasing rural to urban migration worldwide especially in developing countries, the findings could also partly explain the deteriorating state of rural village infrastructure, natural common pool resources and ecological systems in many developing countries. Ó 2016 Elsevier Ltd. All rights reserved. Key words — commons governance, collective action, labor migration, common pool resources, irrigation, China

1. INTRODUCTION

Irrigation systems in China have played a vital role in feeding the country and reducing vulnerability to uncertain rainfall. Irrigated lands occupy half of China’s farmland but produce three-quarters of its grain and more than 90% of its cash crops. Historically, surface irrigation systems were dominant, but their use declined during the modern era, and they were gradually replaced by the growth of groundwater-based, smallholder irrigation, especially in northern China (Calow, Howarth, & Wang, 2009). In the 1950s, groundwater irrigation was virtually nonexistent in northern China. In the mid-1970s, groundwater likely provided approximately 10–15% of the irrigation supply to the water-short provinces of the north. By the mid-1990s, however, this figure had risen to approximately 40%, and in some downstream provinces, such as Hebei, Shanxi, Henan and Shandong, the share of groundwater-irrigated areas increased to approximately 70% (Wang, Huang, Rozelle, Huang, & Blanke, 2007; Wang, Huang, Rozelle, Huang, & Zhang, 2009). As a result, the prevalence of groundwater water pumps dramatically increased over the past half-century. According to official estimates, the number of wells in China was 138,300

A longstanding academic challenge for scholars is to identify the factors that influence collective action in the commons. Over the past three decades, a significant number of studies have been conducted to explore the effects of key factors on the performance of the commons, among which some early representative studies for example Wade (1987), Berkes (1989), Ostrom (1990), Tang (1992), and some recent works for example Fischer and Qaim (2012), Beitl (2014), Frey and Rusch (2014), Cox (2014). As a result, more than three dozen factors have been identified in the literature as summarized in Agrawal (2001) and Ostrom (2009). These factors can be broadly categorized in terms of the physical characteristics of the goods (e.g., excludability, rivalry in consumption, and scarcity), the attributes of the community (e.g., group size, heterogeneity, and social capital), the institutional context (e.g., communication, rules of use, monitoring and sanctioning), and the broader external environment (e.g., economic development, political stability, and technology), among others. Despite this extensive literature, little is empirically known about the relationship between rural–urban migration and collective action in the commons. Does rural out-migration have an adverse effect on the ability of villagers to solve collective action problems in the commons? This question is important because rural–urban migration is an increasing global phenomenon especially among developing countries. We examine this question using the case of irrigation systems in China. Irrigation systems represent a logical unit of analysis in studying collective action in a common pool resource. This is because an irrigation system gives rise to various potential collective action problems such as appropriation, assignment, provision, and monitoring (Ostrom, Gardner, & Walker, 1994; Suhardiman & Giordano, 2014).

* This work is jointly supported by the National Natural Science Foundation of China (71573151, 71303132), the Major Program of the National Social Sciences Foundation of China (15ZDB164) and Tsinghua University Initiative Scientific Research Program (2014z04083). We are grateful for the comments from three anonymous reviewers and the editor. Also we thank Ms. Ye Tao and Mr. Li He for their research assistance. Any errors are the sole responsibility of the authors. Final revision accepted: July 15, 2016. q An earlier version of this paper has been presented at the Fifth Workshop on the Ostrom Workshop (WOW5) Conference held at Indiana University Bloomington on June 18–21, 2014. 79

80

WORLD DEVELOPMENT

in 1964, 2.7 million in 1980, 4.45 million in 2000, and by 2011, the number had risen to approximately 5.08 million. In contrast, according to the Chinese Rural Survey conducted by the China Institute for Rural Studies (CIRS) in 2012, the proportion of households relying on canal irrigation has declined to one-third. In the past three decades, China has experienced a rapid expansion of urbanization. In 1982, 21.13% of the population lived in urban areas; by 2015 this percentage had increased to 56.1%. This urbanization process was accompanied by massive migration. According to data provided by the Chinese National Bureau of Statistics, in 2014, the number of labor migrants from rural areas exceeded 278 million, which is nearly one-fifth of the total population in China and equivalent to three-quarters of the total population of the United States. Because the majority of rural–urban migrants are younger and more skilled workers, this massive migration has led to a major transformation in rural China, including, we speculate, the rapid decline of collective irrigation and a dramatic increase in the use of groundwater irrigation. We make several contributions to the literature. First, China is an attractive case study because it has the world’s most extensive levels of rural to urban migration, but little is known about how migration affects collective action in the commons. Based on econometric analysis employing data from rural China, we find that out-migration has a statistically significant adverse effect on collective action. Based on the literature, we explain the mechanisms through which out-migration affects collective action in the commons, i.e., out-migration has negative effects on village leadership, social capital and sense of community, economic heterogeneity, and dependency on a resource and how these factors in turn decrease the likelihood of collective action in the commons. Second, our study is based on a survey of 1780 irrigation households in 74 villages from 18 provinces throughout China. In contrast, the empirical literature on China is limited to specific provincial/regional data, i.e., Inner Mongolia (Qiao, Zhao, & Klein, 2009), Yunnan (Ito, 2012), and Northern China (Wang, Otto, & Yu, 2013). Much of the literature on collective action in China examines environmental governance (Yee, Lo, & Tang, 2013) and collective action among migrant workers in cities (Chan & Ngai, 2009). Furthermore, much of the empirical literature on rural–urban migration in China has focused on patterns of migration (Wu, 1994), impediments to migration (Scott, Guo, Shen, Hughart, & Giles, 1999) and the consequences of the Hukou system (Chan & Zhang, 1999; Cheng & Selden, 1994), among other topics. To the best of our knowledge, this is the first study on collective action in the commons in China using large-N survey data. Third, we also control for a variety of theoretically relevant factors that the literature cites as influencing collective action, namely, community attributes (household size, inequality, education of household head), physical attributes (proximity of villages to urban areas), geography (hilly areas, flood plains), and water scarcity, among other factors. We speculate that these factors associated with out-migration jointly explain the significant drop in the use of collective canal irrigation and exacerbated the significant increase in groundwater irrigation since the start of reforms in 1980s. The remainder of the paper is structured as follows. In the next section, we discuss the literature about migration and collective action, and the conceptual framework of the influence of out-migration on collective action in the commons. In Section 3, our survey methods and data, the

construction and measurement of our dependent and independent variables, the descriptive statistics analysis, are discussed. In Section 4, the econometric analysis results are presented and discussed, while conclusions and implications follow in Section 5. 2. MIGRATION AND COLLECTIVE ACTION IN THE COMMONS Rudel (2011) suggests three possible mechanisms through which out-migration affects collective action in the commons and the development of durable common pool institutions (CPIs). First, the prospect of higher wages elsewhere would raise the discount rates of participants in CPIs and reduce the salience of the commons with respect to their livelihoods. Alternatively, individuals considering long distance migration, such as unattached young men or women, might be reluctant to join a CPI because they will feel the opportunity cost of joining a local CPI. Second, accelerated rates of labor migration, by increasing the mobility of capital and labor, present challenges to CPI participants committed to the long-term sustainability of the commons. Third, labor migration creates social pressures that shape organizations, preventing the creation of new CPIs and undermining or destroying existing CPIs. Although Rudel’s explanation of the effect of labor migration on CPIs provides valuable clues, we believe that the actual mechanisms operative in the real world are more complicated. According to the previous studies about migration and collective action, we speculate that out-migration affects collective action in the commons through five mechanisms, namely, leadership, social capital, sense of community, economic heterogeneity and resource dependency. The first mechanism is leadership. It is widely believed that leadership exerts a considerable influence on the results of collective action (Meinzen-Dick, Raju, & Gulati, 2002). The involvement of a charismatic or trusted individual reduces the transaction costs of organizing and provides assurance that makes individuals more willing to participate in collective action (Baland & Platteau, 1999; Kolavalli, 1995). For instance, the presence of college graduates and influential elders had a strong positive effect on the establishment of irrigation organizations in a stratified sample of 48 irrigation systems in India (Meinzen-Dick, 2007). The increased rate of rural–urban migration causes a massive brain drain, leading to decreased rural human capital and a lack of rural elite talents. The loss of leadership resources thus reduces the likelihood of organizing successful collective action. The second is social capital. In situations in which the social capital of formalized groups is high, individuals have the confidence to invest in collective activities, knowing that others will also do so (Pretty, 2003). Similarly, in communities characterized by close social proximity (with low transaction costs and frequent communication), where community members place a greater premium on the importance of social norms, collective action in common property resource management is likely to succeed (Runge, 1986). Studies also show that a larger number of exit options reduces cooperative capacity, as it weakens social cohesion (Bardhan, 1993) and increases the costs of enforcing rules, thereby exerting further negative impacts on collective resource management (Stern, Dietz, & Ostrom, 2002). Labor migration weakens local social connections and attenuates social capital in rural villages, which

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

reduces individuals’ confidence and incentive to invest in collective activities. The third, sense of community, is closely related with the concept of social capital, which fundamentally refers to an individual’s experience of community life (Hyde & Chavis, 2007; Mannarini & Fedi, 2009). In terms of its motivating power, a sense of community is considered a catalyst for social involvement and participation in the community (Chavis & Wandersman, 1990; Davidson & Cotter, 1986), and evidence of the connection between a sense of community and participation is generally consistent across cultures and social groups (Albanesi, Cicognani, & Zani, 2007; Cicognani et al., 2008; Prezza & Costantini, 1998). As Klandermans (2002) demonstrates, a positive relationship with the community could strengthen group identification, which reinforces collective action. In this regard, labor migration exerts a significantly negative impact on rural residents’ sense of community, as the number of persons who remain in the village all year has dramatically declined, leading to decreased time for individuals to participate in community life and interact with other rural residents. This undermines the shared emotional connection with other members and reduces the benefits that individuals derive from their participation in rural community activities or collective action. The fourth factor is economic heterogeneity. Poteete and Ostrom (2008) argue that the relationship between heterogeneity and collective action is nonlinear and contingent upon many factors. Among the various forms of heterogeneity, we focus on economic heterogeneity or inequality. Many studies confirm that farmer groups that are internally differentiated on the basis of income or resources are not as successful at collective action as groups that are more homogeneous (Harris, 2008; Naidu, 2009; Ruttan, 2008). For instance, Easter and Palanisami (1986) find that the smaller the variances in farm size among farmers, the more likely they are to form water user associations. Tang (1992) reports a negative relationship between variance in average family income among irrigators and the degree of rule conformance and maintenance activity. Because rural migrant wages are significantly higher than those of rural residents (Taylor, Rozelle, & de Brauw, 2003), labor migration would increase inequality and thereby produce negative effects on farmer participation. The final mechanism is resource dependency. Runge (1986) and Ostrom (2000) contend that greater dependence on a resource creates incentives for cooperation. Transitioning out of rural areas to urban cities and from the agricultural sector to the industrial and service sectors would reduce migrants’ dependence on rural resources and their willingness to invest in rural infrastructure such as irrigation. An empirical study in Gansu Province of China finds that out-migration is negatively associated with water use per unit planting area (Castro, Heerink, Shi, & Qu, 2010). Another empirical study in Western China confirms that migration has a significant negative influence on physical and natural capital (Li, Li, Feldman, Daily, & Li, 2012). As a result, the weaker dependence on resources associated with labor migration could, in turn, weaken the motivation to engage in collective action such as irrigation system maintenance. Based on the above discussion, we hypothesize that out-migration has an adverse effect on collective action in the commons through its effects on leadership, social capital, sense of community, economic heterogeneity, and resource dependence.

81

3. DATA, VARIABLES, AND DESCRIPTIVE ANALYSIS (a) Survey methods and data The survey data used in the current study constitute a subset of a comprehensive, large-scale survey organized by the China Institute for Rural Studies (CIRS) at Tsinghua University in 2012. The aim of this survey was to collect information on the production and living situations, especially migration and irrigation, of rural households. Approximately 600 university students, the majority of whom were majoring in agriculture-related subjects at various prominent universities in Beijing, including Tsinghua University and China Agricultural University, were recruited and trained at Tsinghua University in basic survey and interview techniques by two survey experts from CIRS. These trained surveyors were divided into 97 groups and sent to 17 provinces and 5 autonomous regions to conduct the survey. For convenience, most surveyors worked in their home province. In general, purposive sampling was required and implemented both at the village and household levels. That is, in each province or autonomous regions, with the assistance of local officers and cadres, we chose 5–15 representative villages according to the economic development level; and in each village, our interviewers asked local cadres to choose 20–30 representative village households from each the upper, middle, and lower income levels. In total, 5165 completed household survey questionnaires and 205 villagelevel questionnaires from 22 provinces were returned. The questionnaire covered seven major topics: land trading and transfers, irrigation, education, health and rural hygiene, rural finance, cooperatives, and other organizations in rural areas, and the well-being of the elderly and children in rural villages. In short, the survey provides a comprehensive picture of rural China. One of the most noteworthy characteristics of the survey design in the current study is that it contained related information at both the village and household levels. To the best of our knowledge, few studies have been designed to analyze collective irrigation at both levels. However, it must also be noted that this design was not implemented without difficulty. For example, in certain regions, the completeness and quality of responses varied significantly at both the household and village levels. Moreover, for those villages with fewer than 10 valid households, it is conceivable that the quality of the survey may be less than satisfactory. With these two considerations in mind and matching village data to household data, the subsample we included in the current study was significantly reduced to 1780 households in 74 villages from 18 provinces 2 of China. This subsample is nevertheless much larger than those typically reported in the empirical literature and is overall regionally representative. Moreover, the test for attrition bias has been done, and fortunately there is no sample selection bias. (b) Framework and variables (i) Conceptual framework Ostrom et al. (1994), Araral (2008) and many others suggests that collective action in the commons can be explained by three general factors, namely the physical attributes of the common pool resource, the attributes of the community and households as well as the institutional context. Following the extensive literature on governance of the commons and what we know about the effects of rural–urban migration, we propose a conceptual framework adapted from the

82

WORLD DEVELOPMENT

Institutional Analysis and Development (IAD) framework (Ostrom, 2005) in Figure 1, to show the relationships among our dependent, independent, and control variables. In this framework, the physical attributes of the resource— for example, scarcity—affects the attributes of the community (e.g. size of the village, likelihood of migration), which in turn affects and is affected by the institutional context. For example, water scarcity can leave to the adoption of water sharing rules in the village, which can affect the farm sizes. Below we discuss in more depth the theoretical relevance and challenges of measuring our variables. In summary, our dependent variable is the likelihood of collective action. Our independent variable is the extent of out migration. Our control variables include the physical attributes of irrigation (topography, village location relative to urban areas, and water scarcity), community and household characteristics (village size, wealth inequality) and microinstitutional context (monitoring and enforcement, village governance). The hypothesized effects of these variables on collective action are summarized in Table 1 below. (ii) Dependent variable A widely acknowledged challenge in empirically studying the determinants of collective action in irrigation is identifying the appropriate measurement of the dependent variable or the degree of collective action. Thus far, two main approaches to measuring collective action have been suggested. The first approach is to assess the outcome of successful collective action using indicators, e.g., maintenance status of irrigation facilities, where the relative absence of conflicts or rule violations in rural or community irrigation systems is defined as good maintenance and active participation (Bardhan, 2001); the magnitude of free riding includes both monetary and labor free riding among users of an irrigation system (Araral, 2009).

The second approach is the direct measurement of collective action. For instance, Fujiie, Hayami, and Kikuchi (2005) proxy for collective action by counting successfully organized collective activities, such as collective lobbies, and joint maintenance of canals and tanks. Some studies such as MeinzenDick et al. (2002) employ both approaches, i.e., a direct indicator of irrigation organization and an outcome indicator of collective action. This study introduces a somewhat different but also illuminating indicator to denote different levels of collective action. Specifically, we assign an ordered dummy variable for different types of irrigation arrangements, with 3 being the highest score and assigned to open canal irrigation, 2 to well irrigation, 1 to lift irrigation and 0 to relying exclusively on rain-fed irrigation. These rankings were based on the widely recognized facts that the coordination and cooperation required for maintaining open canal irrigation are the highest, that intensive coordination is not particularly essential to the maintenance of well and lift irrigation, and that rain-fed irrigation has the lowest cooperation requirements. Although somewhat subtle and indirect, an obvious advantage of this form of measurement is that only the more objective or observable information regarding the irrigation type in use needs to be taken into consideration. In contrast, ex-ante self-reported maintenance efforts or inputs from households might suffer from exaggeration bias in certain circumstances or from the subjective appraisal of maintenance status as recalled by interviewees. Nevertheless, to maintain comparability with the existing literature, we repeat key regressions for collective action with traditional dependent variables such as the frequency of participation in collective irrigation maintenance or village irrigation meetings at the household level and use them as robustness checks for our baseline regressions that use this novel collective action measurement.

Figure 1. A framework to link migration and collective action in the commons. Source: Adapted from Ostrom (2005), p. 15.

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

83

Table 1. Variables definition and expected impact on collective irrigation Dependent variables

Definition

Expected sign

ODCA CLTMTN CLTMET

Irrigation types: 0 = rain-fed; 1 = lift; 2 = well; 3 = open cannel Frequency of participation in collective maintenance: 2 = often; 1 = normal; 0 = seldom or never Frequency of attending village meetings related to irrigation: 2 = often; 1 = normal; 0 = seldom or never

n n n

Independent variables Community attributes LMIGRATION LMIGRATION2 LMIGRATION3 GINI TOTALHOUSE

The share of households with out-migrants in total village households The share of out-migrants as proportion of village total population The share of out-migrants as proportion of total village labor size Gini index: village level family income inequality measure Village size: the number of total households

    Uncertain

Natural conditions TOPOGRAPHY LOCATION MIRRIGTSCAR SIRRIGTSCAR

Plain = 1 and 0 otherwise Non-suburban village = 1 and 0 otherwise Moderate water scarcity at village level = 1 and 0 otherwise Severe water scarcity at village level = 1 and 0 otherwise

+ + + 

Institutional arrangements VPAFAILURE Village governance failure: 1 for petitions and conflicts, 0 otherwise VSANCTION Monitoring & sanctioning rules: 1 for imposing rules against free riders, 0 otherwise

 +

Household characteristics HLABOR Household labor: percentage of labors in household IGTSHORTAGE Irrigation shortage history: 1 = never; 5 = frequent IMPTSHORTAGE Impact of insufficient irrigation: 1 = bad impact and 0 otherwise MDISTANCE Middle distance to public irrigation LGDISTANCE Long distance of village to public irrigation AGE Age of household head EDU Education of household head

+  + + +/ Uncertain Uncertain

(iii) Independent variables Largely consistent with the empirical commons literature, we include four sets of variables to classify the determinants of collective irrigation: natural conditions, community attributes, institutional arrangements and household characteristics. The variable we emphasize in this study is labor migration in the subset of community attributes, one of the most important facets of the urbanization process in developing economies, which is highly stressed in development models but has surprisingly been less explored in the commons literature Labor migration Development economists have long noted labor migration from rural to urban areas as a key factor for boosting economic development in developing countries. It revitalizes considerable residual laborers in rural areas and raises the marginal productivity of rural land production on the one hand and lowers wage costs for the urban service and manufacturing sectors on the other hand. However, labor migration is also associated with side effects and new challenges for traditional villages. From the discussion in Section 2, we expect that labor migration would have a negative effect on collective irrigation. We use three indicators measuring labor migration: (1) LMIGRATION is the share of households with outmigrants in the total number of village households; (2) LMIGRATION2 is the share of out-migrants in the village total population; and (3) LMIGRATION3 is the share of outmigrants in the total village labor force. We use LMIGRATION to estimate a baseline regression and leave the other two indicators for a robustness check. Other community attributes In addition to labor migration, we employ two other variables to account for socioeconomic differences at the village level, namely, village size and income heterogeneity, denoted

by the Gini coefficient, which may influence the collective irrigation preference at the household level. Village size: The effect of the size of user groups on collective action remains a complex and controversial issue (Poteete & Ostrom, 2004). As Meinzen-Dick, Brown, Feldstein, and Quisumbing (1997) argue, group size represents a tradeoff between potential economies of scale and an increase in transaction costs. On the one hand, the larger the community, the more difficult it will be to maintain the institutions and rules governing local collective resources or commons due to mounting coordination costs and free rider problems associated with increasing numbers of participants. As collective irrigation frequently exhibits characteristics of an economy of scale, in communities and villages with larger populations, it may be increasingly preferred or benefit more individuals while featuring decreasing unit costs. In this sense, the impact of village size on collective irrigation is uncertain and contingent upon the relative forces of mounting coordination costs and benefits from economies of scale (Araral, 2009). We use the total number of households in a village as a measurement of village size. Income heterogeneity: According to Agrawal and Gibson (1999), heterogeneity affects the prospects for developing trust among participants, and thus the likelihood of successful collective action, because of its effect on the divergence of interests. Income heterogeneity among farmers, as discussed in the summary in Section 2, is more likely to be a critical factor that hinders successful collective action. In line with convention, we calculate income inequality measures to capture heterogeneity at the village level based on corresponding household income data. A Gini index measuring village-level family income inequality is used to estimate the baseline regressions.

84

WORLD DEVELOPMENT

Natural conditions We include three variables, namely, topography, location, and water scarcity, to control for the geographic heterogeneity affecting collective irrigation arrangements. Topography: We use a simple 0–1 dummy to denote topographical differences, with 1 indicating plains and 0 indicating hilly or mountainous areas. The cost of the construction and maintenance of collective irrigation systems is obviously considerably lower in areas with plain topography, as there is less need to dig channels through hills and mountains. In additional, economies of scale are more likely to be realized for a large irrigation system located on a plain than in hilly or mountainous areas. Hence, we expect that residents living on plains would prefer collective irrigation systems such as open canals to well irrigation. Location: Studies suggest that greater access to market opportunities encourages participants to engage in collective actions (Meinzen-Dick et al., 2002). The distance to the nearest market is normally used as a proxy for access to market opportunities (Agrawal & Yadama, 1997; Araral, 2009). In this study, however, we do not have an exact distance measurement of this type. Instead, we introduce a related indicator as a substitute, i.e., a dummy denoting the village location type, 1 for a non-suburban village and 0 otherwise. Consistent with the existing literature, we expect that proximity to a town would increase the tendency for collective irrigation, ceteris paribus. Water scarcity: There is general agreement that collective action among resource users is unlikely unless they perceive that the resource is moderately scarce. In the case of irrigation systems, Agrawal (2002) and Bardhan (1993) suggest that resource scarcity and collective action are related in a curvilinear manner, that is, farmers are more willing to manage and maintain systems when water is neither extremely scarce nor extremely abundant but is only moderately scarce. A subsequent study by Araral (2009) also contends that water scarcity has a curvilinear effect on collective action. Thus, we expect that water scarcity will have a curvilinear effect on the adoption of collective irrigation. For this study, we introduce two dummy variables to denote the relative water scarcity of farmland, with MIRRIGTSCARE equaling 1 for moderate water scarcity and 0 otherwise and SIRRIGTSCARE equaling 1 for severe water scarcity and 0 otherwise. Consistent with the curvilinear effect hypothesis, we expect that in the case of moderate water scarcity, the incentive for participating in collective action is greatest compared with severe and minimal water scarcity, meaning that the sign of MIRRIGTSCARE would be positive with minimal or severe water scarcity as the reference point. Institutional arrangements There is rich evidence in the empirical literature that the governance structure of an irrigation system affects the likelihood of collective action (Araral, 2009; Lam, 1998; Ostrom, 1990). In this study, we expect that the governance at the village level would influence the preference for collective irrigation. According to the data accessible in this survey, we construct two variables denoting the institutional arrangements at the village level. The first is a dummy variable VPAFAILURE to denote the village governance level, with 1 denoting villages that had petitions in the last three years over land circulation and 0 otherwise. The need for a petition indicates that there have been conflicts that could be resolved at the village or local level, which reflects a governance failure in the village that may jeopardize collective irrigation.

The second dummy variable VSANCTION denotes monitoring and sanctioning rules, with 1 indicating the imposition of sanctions against privately cutting channels or evading irrigation fees and 0 otherwise. We expect VPAFAILURE to take a negative sign in the regressions, while the sign of VSANCTION is expected to be positive. Household characteristics Household-level heterogeneities also influence the preference for collective action, as noted in the literature. In this study, we include five variables to account for heterogeneity at the household level. Household labor: The empirical literature reports that family size has positive effects on a farmer’s participation in irrigation (Khalkheili & Zamani, 2009). Family size is generally associated with shares of land and natural resources in villages, thus with larger family size, collective irrigation is more attractive at the household level because it improves resource management. Moreover, larger family size may also improve the labor supply available for collective irrigation. In this study, we use the percentage of laborers in a household to measure family size and expect that increasing this percentage will have a positive effect on the preference for collective irrigation. Distance to irrigation system: The distance from a household’s land to the irrigation system is a variable commonly employed in the existing literature. There are two typical ways to measure distance, one is relative location to the irrigation system, such as upstream or downstream, and the second is absolute distance. In this survey, we adopt the first approach, using the two separate dummies of MDISTANCE and LGDISTANCE that denote the relative distance to the public irrigation system, namely, middle distance or long distance. According to the empirical literature, we expect that the incentive for collective irrigation is the highest for households at the middle distance, while for households at a long distance or located nearby, the incentive is diminished due to the decreasing benefits or relative water abundance. Irrigation shortage history: Irrigation water scarcity may discourage a household from playing an active role in collective irrigation and encourage it to return to private watering (Kajisa, Palanisami, & Sakurai, 2007). We use a simple evaluation variable IGTSHORTAGE with an integral value ranging from 1 to 5 to denote the relative historical irrigation shortage, with 1 representing never and 5 representing frequent. We expect that this variable will have a negative effect on the collective irrigation preference. Impact of insufficient irrigation: Dependence on resources has commonly been considered a positive factor for collective action in the literature, as per the discussion in Section 2. The more important irrigation is to farmers, the more likely they will be to choose collective irrigation. In this study, we introduce a variable IMPTSHORTAGE for the impact of insufficient irrigation to denote the dependence on resources, with 1 representing a negative impact and 0 representing no impact. It can be argued that the more influential irrigation scarcity is on household living conditions, the more proactive the household will be in acquiring more reliable and effective irrigation. Thus, we expect that this variable for irrigation influence will have a positive sign in our regression. Others: education and age: Finally, we also include the education and age of the head of household in our baseline regression as control variables to control for unobserved heterogeneity at the household level. Unlike the previous sets of controls considering the household level, there is no consensus in the empirical literature on the signs of these two variables (Huang, Rozelle, Wang, & Huang, 2009; Khalkheili & Zamani, 2009; Kiumars, Alibaygi, & Nasrin, 2008).

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

(c) Descriptive statistics The descriptive statistics of the fundamental variables are presented in Table 2. As shown in Table 2, for the independent variables, on average, approximately 45% of the sample households include out-migrants, Gini index on village level is 0.39. In addition, 21% of the villages are located in plains areas, 87% of villages are non-suburban villages and nearly 43% and 23% of households live and produce with modest water scarcity and severe water scarcity, respectively. We also find that the percentage of labors in household is nearly two thirds; most farmers heavily reliant on irrigation; households face middle frequency of irrigation shortage in history; approximately 32% and 15% of households, respectively, have farmland at a moderate or long distance to an irrigation system. Respondents also report village governance failure is measured to 0.49, which means nearly half villages suffered petitions over land circulation; monitoring and sanctioning rules is measured 0.24, which means only one-quarter of villages imposed sanctions for non-compliance. Regarding personal characteristics, our sample interviewees are aged approximately 47, and they generally have fewer than 8 years of schooling, that is, below middle school. These two indicators are quite comparable to the results of most rural surveys in China. To present the relationship between labor migration and irrigation type, we conduct further descriptive statistics of group analysis as shown in Table 3. The village with high out-migration rate (more than 30%) has a much lower proportion of households choosing canal irrigation and well irrigation than the village with low out-migration rate. In contrast, the proportion of households choosing rain-fed irrigation and lift irrigation is obvious higher in high outmigration village than in low out-migration village. These statistics are consistent with the exception that labor migration has a negative effect on collective irrigation. However, the other variables in the categories of natural conditions, institutional arrangements and household characteristics are related with irrigation type choice in group analysis. For example, the households living in plain area, moderate water scarcity condi-

85

tion and villages with better governance, are more likely to choose collective irrigation; and the same for the households with the characteristics of locating in middle distance to irrigation system and seldom suffering irrigation shortage. Although the above descriptive statistics support our hypnosis, multivariate regression analysis is necessary to be developed to explore the mechanism of irrigation type choice, for many variables jointly affect the household behaviors of irrigation participation. 4. ECONOMETRIC MODELS AND RESULTS (a) Regression methodology We use an ordered probit model to empirically test the impacts of labor migration combined with the other factors on farmer preferences for collective irrigation in rural areas. The dependent variable is an ordered dummy denoting different degrees of collective irrigation. Thus, the ordered probit model is an appropriate choice for performing the econometric analysis. The model can reveal the influence of each variable in determining the probability of choosing collective irrigation. In general, typical challenges in obtaining a consistent estimate include unobserved heterogeneity, omitted variable bias, and measurement error in both the dependent and control variables. Unobserved heterogeneity denotes that although some unobserved factors might affect irrigation arrangements in specific places, information on these factors could not be obtained through questionnaires. We trigger or minimize this type of bias in the following two ways. First, we include sets of all common indicators believed to be important factors in explaining collective irrigation to absorb the observed heterogeneity at both the village and household levels. Second, for fixed factors such as the climate, cultural features and water abundance of a certain region, which may also influence irrigation arrangements in specific areas but could not be taken into account in the model, we added province dummies to the model.

Table 2. Variables descriptive statistics Variables

Mean

Std. Dev.

Min.

Max.

Dependent variable ODCA

N = 1780 Different irrigation types (3 = open canal)

Description

1.79

1.21

0

3

Independent variables LMIGRATION GINI TOTALHOUSE TOPOGRAPHY LOCATION MIRRIGTSCAR SIRRIGTSCAR VPAFAILURE VSANCTION HLABOR IGTSHORTAGE IMPTSHORTAGE MDISTANCE LGDISTANCE AGE EDU

Percentage of households with out-migrants Gini index on village level Ln (number of total households) 1 = plain; 0 otherwise 1 = non-suburban village; 0 otherwise 1 = modest water scarcity; 0 otherwise 1 = severe water scarcity; 0 otherwise 1 = petitions and conflicts; 0 otherwise 1 = rules existence; 0 otherwise Percentage of labors in household 1 = never; 5 = frequent 1 = bad impact; 0 otherwise 1 = middle distance; 0 otherwise 1 = long distance; 0 otherwise Age of household head (years) Years of education of household head

0.45 0.39 5.83 0.21 0.87 0.43 0.23 0.49 0.24 0.62 2.24 0.78 0.32 0.15 47.41 7.41

0.89 0.12 1.05 0.41 0.34 0.49 0.42 0.50 0.43 0.25 1.18 0.42 0.47 0.36 12.92 3.42

0 0.11 2.30 0 0 0 0 0 0 0 1 0 0 0.00 15 0

1 0.75 7.56 1 1 1 1 1 1 1 5 1 1 1 88 15

Note: Please refer to Table 1 and Section 3.2 for more detailed description of the variables.

86

WORLD DEVELOPMENT Table 3. Irrigation type distribution of different group Rain-fed

Lift

Well

Canal

Migration LMIGRATION = HIGH LMIGRATION = LOW

40.92 16.39

22.99 17.02

7.37 20.45

28.74 46.14

Natural conditions TOPOGRAPHY = 1 TOPOGRAPHY = 0

6.23 26.85

13.03 19.96

39.38 11.37

41.36 41.82

MIRRIGTSCAR = 1 SIRRIGTSCAR = 1

14.81 31.01

21.74 21.88

13.72 18.75

49.73 28.37

Institution arrangements VPAFAILURE = 1 VPAFAILURE = 0

27.19 18.27

17.96 19.07

17.72 16.57

37.13 46.08

VSANCTION = 1 VSANCTION = 0

14.46 25.15

21.32 17.66

19.85 16.28

44.36 40.9

Household characteristics IGTSHORTAGE = SELDOM IGTSHORTAGE = OFTEN

15.92 32.65

19.42 17.2

15.73 19.24

48.93 30.9

IMPTSHORTAGE = 1 IMPTSHORTAGE = 0

23.37 19.95

17.98 20.47

17.45 16.01

41.2 43.57

MDISTANCE = 0 MDISTANCE = 1 LGDISTANCE = 1

27.73 11.89 15.69

18.26 19.1 25.49

17.23 16.94 9.8

36.78 52.07 49.02

Provided that these factors are constant at the provincial level, our provincial fixed effect framework would significantly reduce the bias caused by this type of unobserved heterogeneity and omitted variable bias. With regard to possible measurement error for both the dependent and control variables, we use different measures for those likely suffering from measurement errors or simply dropped the unsatisfactory measurements and observed whether our main estimates remained comparable. (b) Baseline regression In this subsection, we report the results of the baseline regression with and without provincial fixed effects, taking into account the impacts of both observed and unobserved heterogeneity. The first part of the results concerning baseline regression are presented in Table 4, in which columns (1) and (2) are exact pair-wise estimates and those in column (2) further control for provincial fixed effects to address unobserved heterogeneity or omitted bias at the provincial level. The difference between columns (3) and (4) and column (2) is that we add controls for a long distance to the irrigation system in column (3) and a severe water scarcity indicator in column (4) to more precisely assess the curvilinear effects of distance and water scarcity on collective irrigation. Overall, the estimates from columns (1) to (4) show the goodness of fit of our regression models. Most variables have the expected signs and are significant in most specifications. Labor migration has a significantly negative impact on collective irrigation, consistent with our expectations. Moving from column (1) to columns (2)–(4), which include provincial fixed effect controls, the coefficients of labor migration are uniformly larger, which indicates that the omission of unobserved

heterogeneity may cause an underestimate of the impact of labor migration on collective irrigation. The estimates of labor migration from columns (2) to (4) are relatively comparable in magnitude, thereby improving our confidence that the estimates are to some extent robust and less likely to suffer from specification bias. For other community attributes, increasing heterogeneity as captured by the village Gini coefficient has a consistently negative effect, which is consistent with studies by Chambers and Conway (1992) and Stern et al. (2002) while contradicting other studies (Bardhan, 1999). The result here shows that increasing the village size has a positive impact, indicating that the benefits of economies of scale in an irrigation system override the effect of increased coordination costs. In natural conditions, it is less costly both to build and maintain collective irrigation facilities such as channels in plains than in hilly mountains, and thus residents of plains areas had a significantly greater preference for collective irrigation, as shown in Table 4, which is highly consistent with the empirical literature (Agrawal, 2002; Araral, 2009; Bardhan, 1993; Ito, 2012). Consistent with the hypothesis regarding the curvilinear effect of water scarcity, both in the first three columns using severe and minimal water scarcity as reference points and in the fourth column using solely the minimal water scarcity indicator as a baseline, the estimates for the indicator of moderate scarcity are unanimously positive and significant. This indicates that the incentive for participating in collective irrigation is greatest in this case compared with those of severe or minimal water scarcity. The significantly positive estimates of the proximity to town from columns (1) to (4) show that non-suburban villages prefer collective irrigation more than suburban ones, which means that for villagers located far from a town or county seat, the incentive for collective irrigation is

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

87

Table 4. The determinants of participation in collective irrigation Variables

Dep. = ordered collective irrigation (odca) (1)

LMIGRATION

Other community attributes GINI TOTALHOUSE

Natural conditions TOPOLOGY LOCATION MIRRIGTSCAR

(2) ***

(3) ***

(4) ***

0.133 (0.041)

0.191 (0.049)

0.187 (0.049)

0.188*** (0.050)

1.493*** (0.233) 0.163*** (0.029)

1.516*** (0.269) 0.182*** (0.033)

1.489*** (0.270) 0.180*** (0.033)

1.566*** (0.275) 0.171*** (0.034)

0.420*** (0.071) 0.067 (0.087) 0.352*** (0.058)

0.601*** (0.101) 0.160 (0.100) 0.236*** (0.070)

0.599*** (0.101) 0.167* (0.100) 0.221*** (0.071)

0.613*** (0.102) 0.196* (0.102) 0.146* (0.087) 0.137 (0.092)

0.256*** (0.056) 0.167** (0.065)

0.157** (0.063) 0.171** (0.077)

0.152** (0.063) 0.160** (0.077)

0.140** (0.064) 0.170** (0.077)

0.410*** (0.108) 0.190*** (0.024) 0.068 (0.067) 0.447*** (0.060) 0.002 (0.002) 0.010 (0.009)

0.196* (0.113) 0.140*** (0.026) 0.094 (0.074) 0.386*** (0.063) 0.005** (0.002) 0.017* (0.009)

0.206* (0.114) 0.143*** (0.026) 0.096 (0.074) 0.448*** (0.066) 0.005** (0.002) 0.017* (0.009) 0.244*** (0.084)

0.207* (0.114) 0.138*** (0.027) 0.092 (0.074) 0.451*** (0.066) 0.006** (0.002) 0.018* (0.009) 0.246*** (0.084)

SIRRIGTSCAR

Institution arrangements VPAFAILURE VSANCTION

Household characteristics HLABOR IGTSHORTAGE IMPTSHORTAGE MDISTANCE AGE EDU LGDISTANCE Province fix

No

Yes

Yes

Yes

Observations chi2 r2_p

1,780 357.4 0.0763

1,780 636.9 0.136

1,780 645.5 0.138

1,780 647.7 0.138

Standard errors in parentheses. *** p < 0.01. ** p < 0.05. * p < 0.1.

greater, consistent with the prevailing hypothesis that market opportunity may encourage villagers to opt out of nonagricultural opportunities, thereby discouraging collective action (Agrawal & Yadama, 1997; Bardhan, 1999; Dillon, 2011; Meinzen-Dick et al., 1997). The results regarding the impact of institutional arrangements at the village level generally meet our expectations. Less satisfactory governance, represented by more petitions and conflicts in villages, tends to reduce the incentives for collective irrigation, while when there are monitoring and sanctioning rules against free riders and rule violators, collective irrigation

is preferable, which is consistent with the broader literature (Araral, 2011; Gibson, Williams, & Ostrom, 2005). With respect to the impacts of household characteristics on collective irrigation, the estimates in Table 4 show that a larger share of laborers in a given household significantly increases the incentives for collective irrigation, a history of water shortages has a significant adverse effect on the preference for collective irrigation, and the impact of insufficient irrigation on family living has a positive effect but one that is insignificant in columns (1)–(4). Concerning the impact of the distance from farmland to the irrigation system, the estimates in

88

WORLD DEVELOPMENT Table 5. Accounting for measurement error of control variables Variables

Dep. = ordered collective irrigation (odca) (1)

LMIGRATION2

0.468* (0.272)

LMIGRATION3

(2)

0.329*** (0.090)

LIMIGRATION MLD

(3)

(4)

(5)

0.239*** (0.050) 0.854*** (0.150)

0.267*** (0.049)

0.244*** (0.047)

LINCMSD

0.210*** (0.044)

1.549*** (0.276)

1.697*** (0.281) (other controls omitted)

Province fix

Yes

Yes

Yes

Yes

Yes

Observations chi2 r2_p

1,769 586.4 0.126

1,735 622.6 0.137

1,716 639.0 0.142

1,716 629.1 0.139

1,770 647.8 0.139

GINI

1.336*** (0.279)

Standard errors in parentheses. Notes: (1) MLD = mean log deviation family income; (2) LINCMSD = the standard deviation of logs family income. *** p < 0.01. * p < 0.1.

columns (1) to (4) show that farmland at a middle distance is associated with the largest incentive to join collective irrigation, while for those at a long distance, the incentive to join is also positive and significant but slightly smaller, as shown in columns (3) and (4). Note that in columns (1) and (2), the reference point for the distance dummy is the non-middle distance location including both the upper stream and the lower stream, while in columns (3) and (4), the upper stream serves as a reference point. Thus, the coefficient of middle distance increased slightly from column (2) (0.386) to column (3) (0.448), providing additional indirect evidence consistent with the prevailing observation that the middle location has the largest incentive for collective irrigation. Age and education present negative signs and are significant in the last three columns. This implies that the elder group may have less interest in participating in collective irrigation, which is consistent with the empirical result of Li et al. (2012) in China. Moreover, a farmer with higher education may also exhibit a lower level of participation, which is consistent with the empirical result of Agrawal and Gupta (2005) in Nepal. However, the effects of age and education do not always hold in the following robustness check. Therefore we speculate that the effects of age and education on choosing an irrigation organization method are, to a large extent, context-dependent. (c) Robustness check To test the possible impacts of endogeneity and selfselection bias, the robustness check of the baseline regression has been reported in this subsection. We address the issues of unobserved heterogeneity, omitted variables and measurement error as follows. But there still remained a possible endogeneity issue, i.e., the reverse causality between dependent variable and independent variables, which cannot be examined completely for this study employs a single period dataset. However, we think the endogeneity issue caused by reverse

causality is not a major problem, for migration is more influenced by the rural–urban wage differentials, which is discussed briefly as follows. The above econometric analysis has reported a statistically significant adverse effect of out-migration on collective irrigation. In theory, decline of collective irrigation may influence the households’ decision to migrate. The possible mechanism is that lack of irrigation may reduce households’ income from agricultural production and thus increase the incentive to migrate for higher income. However, such a possibility is very small in real China. According to the statistics of China in 2015, the income from agricultural production in annual income of farmers has dropped to one-quarter, less than average monthly wages of out-migrant labors. Therefore, for the possible issue of reverse causality in this case, the overwhelming mechanism is the out-migration explains the decline of collective irrigation, not vice versa. (i) Accounting for measurement errors on the control side Table 5 reports the impact of possible measurement error in the labor migration indicator and income inequality measure on our baseline regression, with columns (1)–(2) using slightly different measurements of labor migration and columns (3)– (5) replacing the Gini coefficient with two other income inequality measures, MLD & LINCMSD. Due to the larger denominator used in constructing LMIGRATION2 and LMIGRATION3, the coefficients rose slightly compared with those used in Table 4 for LMIGRATION. However, overall, the coefficients of labor migration remain negative and significant in columns (1)–(2) of Table 5, and the estimates for the other control variables are generally comparable to or the same as their counterparts in Table 4. These comparisons assure us that the baseline regressions are robust to possible measurement errors in the labor migration indicators. Roughly the same applies to the assessment of the measurement error of the income inequality measure, in that although there are differences in the magnitudes of the estimates, their

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

89

Table 6. Duplicating baseline regressions with traditional measure for collective irrigation Variables

CLTMTN (1)

CLTMET (2)

CLTMTN (3)

CLTMET (4)

LMIGRATION

0.588** (0.292) 1.142*** (0.298) 0.173*** (0.036) 0.408*** (0.116) 0.299*** (0.104) 0.326*** (0.073) 0.039 (0.067) 0.013 (0.077) 0.179 (0.124) 0.003 (0.030) 0.009 (0.080) 0.194*** (0.066) 0.006** (0.003) 0.031*** (0.011)

0.946*** (0.284) 0.125 (0.303) 0.172*** (0.037) 0.252** (0.121) 0.226** (0.106) 0.287*** (0.075) 0.054 (0.068) 0.109 (0.078) 0.329*** (0.126) 0.018 (0.031) 0.138* (0.081) 0.178*** (0.068) 0.007** (0.003) 0.010 (0.011)

1.106** (0.449) 0.518 (0.399) 0.185*** (0.051) 0.177 (0.139) 0.144 (0.140) 0.848*** (0.105) 0.483*** (0.092) 0.193* (0.100) 0.332** (0.148) 0.066* (0.037) 0.063 (0.098) 0.101 (0.079) 0.006* (0.003) 0.028** (0.013) 0.251*** (0.051)

1.417*** (0.453) 0.199 (0.395) 0.204*** (0.052) 0.001 (0.135) 0.048 (0.143) 0.321*** (0.100) 0.224** (0.091) 0.079 (0.101) 0.282* (0.150) 0.004 (0.038) 0.067 (0.099) 0.240*** (0.081) 0.010*** (0.004) 0.005 (0.013) 0.062 (0.049)

1,863 415.6 0.131

1,891 323.4 0.114

1,153 292.0 0.132

1,167 162.4 0.0837

GINI TOTALHOUSE TOPOGRAPHY LOCATION MWATSCAR VPAFAILURE VSANCTION HLABOR IGTSHORTAGE IMPTSHORTAGE MDISTANCE AGE EDU CANALMATAIN Observations chi2 r2_p Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

signs and significance in determining the preference for collective irrigation are the same. Finally, in column (5), we eliminated water scarcity variables out of a concern that some variables at the village level may be closely correlated with similar controls at the household level. Similar coefficients for the labor migration indicators in columns (3)–(5) indicate that our results in the baseline regression are less likely to be challenged by measurement errors in the other control variables or specifications. (ii) Repeating baseline regressions using traditional measures of collective irrigation To maintain comparability with the existing literature, we use two other two traditional indicators, namely, the selfreported frequency of participation in maintaining a collective irrigation system (CLTMTN) and the self-reported frequency of attending village irrigation-related meetings (CLTMET). Both of these are constructed as ordered dummies with 2 denoting frequent participation, 1 normal participation and 0 for the lowest frequency and serve as substitutes for our ordered collective irrigation measure to determine whether our key hypothesis that labor migration has a negative effect on participation in collective irrigation

still holds. The results of this robustness check are presented in Table 6. In columns (1) and (2), we repeat the baseline specification in Table 4 with provincial fixed effects using an ordered frequency indicator of participation in collective irrigation maintenance and irrigation-related meetings as the dependent variables, respectively. In column (3) and column (4), we add a further control denoting who is responsible for irrigation maintenance, constructed as 1 for an irrigation unit office, 2 for a village committee and 3 for a water user association, duplicating the estimates in columns (1) and (2). Because this information is available only for a limited subsample, the sample sizes are further reduced to 1153 and 1167, respectively, as shown in Table 6. Consistent with the findings in the baseline regressions, columns (1)–(4) show that labor migration uniformly and significantly discourages participation in collective irrigation, whether it be the maintenance of irrigation systems or attending irrigation-related meetings. In line with the literature, the estimate for CANALMATAIN is positive, indicating that more formal and bureaucratic control may hinder or substitute for the need and urge for self-coordination and cooperation among villagers in common pools. Finally, in Table 6, the other control variables are generally of similar

90

WORLD DEVELOPMENT Table 7. Repeating key regressions with village level data Variables

Dep. = ordered collective irrigation maintenance (1)

LMIGRATION3

Other community attributes TOTALHOUSE ECONOMYPOWER

Natural conditions TOPOGRAPHY LOCATION DISTOTOWN

Institution arrangements VPAFAILURE VSANCTION

(2) **

(3)

(4)

0.218 (0.100)

0.264 (0.117)

0.261 (0.139)

0.055* (0.028)

0.123 (0.103) 0.395** (0.174)

0.157 (0.106) 0.415** (0.177)

0.187 (0.118) 0.388** (0.187)

0.253* (0.144) 0.397* (0.221)

0.053* (0.030) 0.083* (0.046)

0.154 (0.244) 1.231*** (0.388) 0.275 (0.299)

0.352 (0.277) 1.277*** (0.390) 0.332 (0.302)

0.198 (0.319) 1.368*** (0.398) 0.390 (0.313)

0.143 (0.368) 1.183*** (0.431) 0.189 (0.371)

0.030 (0.077) 0.249*** (0.091) 0.040 (0.078)

0.506** (0.213) 0.353 (0.244)

0.520** (0.214) 0.347 (0.244)

0.578** (0.226) 0.335 (0.254)

0.483* (0.256) 0.051 (0.306)

0.102* (0.053) 0.011 (0.064)

0.165 (0.107)

0.352* (0.180) 0.606 (0.552)

14.594 (9.527) 20.175* (10.295)

3.068 (1.977) 4.241** (2.095)

RIVEREWAT

**

Margins

0.228 (0.104)

Provincial water endowment WATERENDOW

**

*

Province fix

No

No

No

Yes

Yes

Observations chi2 r2_p

148 37.92 0.123

148 40.30 0.131

140 43.98 0.152

140 79.54 0.275

140

Standard errors in parentheses. Notes: ECONOMYPOWER is village collective income, with 1 representing high income village, 2 representing middle and 3 representing low. DISTOTOWN is log (distance between village and the closet township). WATERENDOW is log (water endowment per capita for each province in 2010). RIVEREWAT is the ratio of III class and above river length, which indicates river water quality for each province in 2010. *** p < 0.01. ** p < 0.05. * p < 0.1.

sign, although in some circumstances of less significance. Therefore, to save space, we do not proceed to discuss these findings in further detail. In summary, these findings justify the rationale underlying our ordered collective irrigation measure. (iii) Repeating regressions with a village level sample Besides the biases related to unobserved heterogeneity, omitted variables and measurement error discussed above, there is still the risk that the subsample we used is not representative of the whole survey. We address this concern by repeating key regressions with a subsample covering more villages, and restrict our empirical analysis of collective action in irrigation at the village level rather than at the household level. Notably, the collective irrigation variable we used for the village level regression is another ordered dummy constructed using evaluation points ranging from 0 to 2 such that 2 points denoted good maintenance and 0 denoted poor maintenance of irrigation facility, as maintenance outcome is commonly used as a measurement of level of participation, which could then also be viewed as providing a second opportunity to

check whether our findings remain robust for different collective action measurements. In addition to the control variables at the household level, we include similar control variables for the village regressions, and the results are presented in Table 7. From columns (1) to (3), we used water endowment at the provincial level to account for provincial level heterogeneity. And in column (4) we control provincial fixed effects. In column (5) we calculated marginal effect based on the setup of column (4). It is noteworthy that the coefficients before the labor migration are all negative and significant in columns (1)–(4), and are comparable in terms of magnitude, which means that our key hypothesis which argues that labor migration would discourage participation in collective irrigation can also be justified in a sample covering many more villages, and is less likely to be suffering from sample selection bias. (iv) Using different models and marginal effects To assess the impact of using different estimation models on our results, the baseline regression in Table 4 is further

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA

91

Table 8. Using different model and marginal effect Variables LMIGRATION GINI TOTALHOUSE TOPOGRAPHY LOCATION MIRRIGTSCAR VPAFAILURE VSANCTION HLABOR IGTSHORTAGE IMPTSHORTAGE MDISTANCE LGDISTANCE AGE EDU

(1) OLS

(2) Ologit

(3) Oprobit

(4) Margins

0.100*** (0.023) 1.175*** (0.216) 0.125*** (0.026) 0.461*** (0.058) 0.010 (0.081) 0.298*** (0.056) 0.240*** (0.055) 0.132** (0.060) 0.384*** (0.103) 0.181*** (0.023) 0.061 (0.066) 0.505*** (0.060) 0.325*** (0.077) 0.002 (0.002) 0.007 (0.008)

0.373*** (0.096) 2.262*** (0.446) 0.322*** (0.057) 0.951*** (0.168) 0.309* (0.168) 0.366*** (0.120) 0.259** (0.108) 0.214* (0.130) 0.350* (0.191) 0.229*** (0.044) 0.161 (0.125) 0.796*** (0.113) 0.391*** (0.141) 0.009** (0.004) 0.031* (0.016)

0.189*** (0.049) 1.464*** (0.269) 0.180*** (0.033) 0.595*** (0.101) 0.156 (0.100) 0.221*** (0.071) 0.153** (0.063) 0.141* (0.076) 0.207* (0.114) 0.143*** (0.026) 0.095 (0.074) 0.452*** (0.066) 0.249*** (0.084) 0.006** (0.002) 0.017* (0.009)

0.045*** (0.011) 0.345*** (0.064) 0.042*** (0.008) 0.140*** (0.024) 0.037 (0.024) 0.052*** (0.017) 0.036** (0.015) 0.033* (0.018) 0.049* (0.027) 0.034*** (0.006) 0.022 (0.017) 0.107*** (0.016) 0.059*** (0.020) 0.001** (0.001) 0.004* (0.002)

Province fix

No

Yes

Yes

n

Observations R-squared chi2

1,780 0.182

1,780 0.136 638.7

1,780 0.137 642.3

1,780

Robust standard errors in parentheses. *** p < 0.01. ** p < 0.05. * p < 0.1.

adjusted and estimated using naı¨ve OLS and the ordered logit model, and the results are presented in columns (1) and (2), respectively, of Table 8. Compared with their counterparts in columns (2) and (3) of Table 4, the estimates from the ordered logit model are similar or comparable, while the OLS results are slightly smaller for our key variable, labor migration, confirming that the baseline regressions are better estimated using the ordered probit model rather than the OLS model. For column (4) in Table 8, we calculate the marginal effect of each variable based on estimates in column (3), duplicating the baseline specification in column (3) in Table 4. Consistent with previous regression findings, the impact of labor migration on choosing collective irrigation is underestimated in specifications lacking controls for provincial fixed effects. The calculation in column (4) tentatively suggests that a one-percent rise in the labor migration rate will lead to a 4.5% increase in the probability of choosing private irrigation. Being located at a middle distance from an irrigation system tends to incentivize collective irrigation the most, increasing the probability of collective irrigation by 10.7% compared with the upper or lower stream. Because the rural–urban

migration rate has been very high in the past three decades, we can expect that the effect of labor migration on collective action is generally destructive in China. 5. CONCLUSIONS The main contribution of this paper is to highlight the effects of labor migration on collective action in the commons literature. Based on a survey of 1780 households from 18 provinces throughout China, we find that labor migration has a statistically significant adverse effect on participation in collective irrigation, controlling for the type of irrigation and a host of theoretically relevant variables. We speculate that the problems of collective irrigation resulting from out-migration can partly explain, perhaps indirectly, the significant decline in surface irrigation in China, and conversely, the significant increase of groundwater irrigation since the 1980s. However, we also speculate that labor migration is a critical but an intermediate variable affecting collective action through the other ‘‘direct” factors frequently identified in the literature,

92

WORLD DEVELOPMENT

including leadership, social capital, sense of community, economic heterogeneity, and dependence on resources. This suggests that the effects of migration on collective action in the commons are possibly mediated by a number of mechanisms. Similarly with the effect of labor migration, and consistent with the literature, we also find that collective action in irrigation is affected by proximity to urban centers or towns, increased inequality, lower levels of household labor, and in hilly areas, non-suburban villages and areas with a history of water shortages. Both water scarcity and distance to the irrigation system have a curvilinear effect on collective action, i.e., farmers are more likely to participate in collective irrigation when water is moderately scarce and their farmland is located at a middle distance to water sources. Meanwhile, better village governance, and more monitoring and sanctioning rules would increase the likelihood to participate in collective irrigation as our expectations. Based on these findings, the policy alternative to save the decline of collective irrigation is to improve the rural governance and institutional arrangements, for the certain biophysical and socioeconomic conditions are difficult to change in the short term. Although our econometric results are robust, there are nevertheless some limitations of this study. First, the sample size

could be further enlarged to test the robustness of the findings, and the accumulation of time series dataset will provide more convincing evidence for the effects of labor migration and completely address the endogeneity concern. Second, some positive effects of migration on collective action have been omitted in the survey, e.g. remittances could be used to support local schools, culture, infrastructure, or in general human capital, which could be explored in another survey. Third, more types of commons in China remain to be examined to enrich the theory of collective action in the commons. Although there is growing consensus regarding the critical factors that influence collective action in the commons, the complex interactions among the factors remain to be studied to deepen the understanding of complex institutional arrangements in diverse socio-ecological settings. Nevertheless, if confirmed by further studies, our findings have important implications for the governance of the commons in the face of rising pressures for rural–urban migration. Because labor migration is a widespread phenomenon worldwide, especially in developing economies, the findings could perhaps partly explain the deteriorating state of rural village infrastructure, natural common pool resources and ecological systems in many developing countries.

NOTES 1. A visiting scholar 2009–10 at the Workshop in Political Theory and Policy Analysis of Indiana University, Bloomington, USA.

2. There are 22 provinces, 5 autonomous regions and 4 direct municipalities in China. Our sampling approach includes the provinces of Anhui, Chongqing, Fujian, Gansu, Guangxi, Guizhou, Hebei, Henan, Hubei, Hunan, Jiangsu, Jiangxi, Inner Mongolia, Ningxia, Shandong, Sichuan, Yunnan, and Zhejiang.

REFERENCES Agrawal, A. (2001). Common property institutions and sustainable governance of resources. World Development, 29(10), 1649–1672. Agrawal, A. (2002). Common resources and institutional sustainability. In E. Ostrom et al. (Eds.), The drama of the commons. Washington, DC: National Academy Press. Agrawal, A., & Gibson, C. C. (1999). Enchantment and disenchantment: The role of community in natural resource conservation. World Development, 27(4), 629–649. Agrawal, A., & Gupta, K. (2005). Decentralization and participation: The governance of common pool resources in Nepal’s Terai. World Development, 33(7), 1101–1114. Agrawal, A., & Yadama, G. (1997). How do local institutions mediate market and population pressures on resources? Forest panchayats in Kumaon, India. Development and Change, 28(3), 435–465. Albanesi, C., Cicognani, E., & Zani, B. (2007). Sense of community, civic engagement and social well-being in Italian adolescents. Journal of Community & Applied Social Psychology, 17(5), 387–406. Araral, E. (2008). The strategic games that donors and bureaucrats play: An institutional rational choice analysis. Journal of Public Administration Research and Theory, 19(4), 853–871. Araral, E. (2009). What explains collective action in the commons? Theory and evidence from the Philippines. World Development, 37(3), 687–697. Araral, E. (2011). The impact of decentralization on large scale irrigation: Evidence from the Philippines. Water Alternatives, 4(2), 110–123. Baland, J. M., & Platteau, J. P. (1999). The ambiguous impact of inequality on local resource management. World Development, 27(5), 773–788. Bardhan, P. (1993). Symposium on management of local commons. The Journal of Economic Perspectives, 87–92. Bardhan, P. (1999). Water community: An empirical analysis of cooperation on irrigation in South India. Berkeley: University of California, Working paper.

Bardhan, P. (2001). Distributive conflicts, collective action, and institutional economics. Frontiers of Development Economics – The Future in Perspective, 269–290. Beitl, C. (2014). Adding environment to the collective action problem: Individuals, civil society, and the mangrove-fishery commons in Ecuador. World Development, 56(04), 93–107. Berkes, F. (1989). Common property resources: Ecology and communitybased sustainable development. Belhaven Press. Calow, R. C., Howarth, S. E., & Wang, J. (2009). Irrigation development and water rights reform in China. Water Resources Development, 25(2), 227–248. Castro, V., Heerink, N., Shi, X., & Qu, W. (2010). Water savings through off-farm employment? China Agricultural Economic Review, 2(2), 167–184. Chambers, R., & Conway, G. (1992). Sustainable rural livelihoods: Practical concepts for the 21st century. Institute of Development Studies. Chan, K. W., & Zhang, L. (1999). The Hukou system and rural-urban migration in China: Processes and changes. The China Quarterly, 160 (1), 818–855. Chan, C. K., & Ngai, P. (2009). The making of a new working class? A study of collective actions of migrant workers in South China. The China Quarterly, 198(3), 287–303. Chavis, D. M., & Wandersman, A. (1990). Sense of community in the urban environment: A catalyst for participation and community development. American Journal of Community Psychology, 18(1), 55–81. Cheng, T., & Selden, M. (1994). The origins and social consequences of China’s Hukou system. The China Quarterly, 139(1), 644–668. Cicognani, E., Pirini, C., Keyes, C., Joshanloo, M., Rostami, R., & Nosratabadi, M. (2008). Social participation, sense of community and social well-being: A study on American, Italian and Iranian University students. Social Indicators Research, 89(1), 97–112.

THE EFFECTS OF MIGRATION ON COLLECTIVE ACTION IN THE COMMONS: EVIDENCE FROM RURAL CHINA Cox, M. (2014). Understanding large social-ecological systems: Introducing the SESMAD project. International Journal of the Commons, 8(2), 265–276. Davidson, W. B., & Cotter, P. R. (1986). Measurement of sense of community within the sphere of city. Journal of Applied Social Psychology, 16(7), 608–619. Dillon, A. (2011). The effect of irrigation on poverty reduction, asset accumulation, and informal insurance: Evidence from Northern Mali. World Development, 39(12), 2165–2175. Easter, K. W., & Palanisami, K. (1986). Tank irrigation in India: An example of common property resource management. In Proceedings of the conference on common property resource management. National Academy Press. Fischer, E., & Qaim, M. (2012). Linking smallholders to markets: Determinants and impacts of farmer collective action in Kenya. World Development, 40(6), 1255–1268. Frey, U., & Rusch, H. (2014). Modeling ecological success of common pool resource systems using large datasets. World Development, 59(7), 93–103. Fujiie, M., Hayami, Y., & Kikuchi, M. (2005). The conditions of collective action for local commons management: The case of irrigation in the Philippines. Agricultural Economics, 33(2), 179–189. Gibson, C. C., Williams, J. T., & Ostrom, E. (2005). Local enforcement and better forests. World Development, 33(2), 273–284. Harris, L. (2008). Water rich, resource poor: Intersections of gender, poverty, and vulnerability in newly irrigated areas of Southeastern Turkey. World Development, 36(12), 2643–2662. Huang, Q., Rozelle, S., Wang, J., & Huang, J. (2009). Water management institutional reform: A representative look at Northern China. Agricultural Water Management, 96(2), 215–225. Hyde, M., & Chavis, D. (2007). Sense of community and community building. In Handbook of community movements and local organizations. US: Springer. Ito, J. (2012). Collective action for local commons management in rural Yunnan, China: Empirical evidence and hypotheses using evolutionary game theory. Land Economics, 88(1), 181–200. Kajisa, K., Palanisami, K., & Sakurai, T. (2007). Effects on poverty and equity of the decline in collective tank irrigation management in Tamil Nadu, India. Agricultural Economics, 36(1), 347–363. Khalkheili, T. A., & Zamani, G. H. (2009). Farmer participation in irrigation management: The case of Doroodzan dam irrigation network, Iran. Agricultural Water Management, 96(5), 859–865. Kiumars, Z., Alibaygi, A. H., & Nasrin, A. (2008). The utility of discriminant analysis for predicting farmers’ intentions to participate in farmer-managed irrigation systems in Iran. Journal of Applied Sciences, 8, 697–701. Klandermans, B. (2002). How group identification helps to overcome the dilemma of collective action. American Behavioral Scientist, 45(5), 887–900. Kolavalli, S. (1995). Joint forest management: Superior property rights? Economic and Political Weekly, 1933–1938. Lam, W. F. (1998). Governing irrigation systems in Nepal: Institutions, infrastructure, and collective action. ICS Press Institute for Contemporary Studies. Li, C., Li, S., Feldman, M. W., Daily, G. C., & Li, J. (2012). Does outmigration reshape rural households’ livelihood capitals in the source communities? Recent evidence from Western China. Asian and Pacific Migration Journal, 21(1), 1–30. Mannarini, T., & Fedi, A. (2009). Multiple senses of community: The experience and meaning of community. Journal of Community Psychology, 37(2), 211–227. Meinzen-Dick, R. (2007). Beyond panaceas in water institutions. Proceedings of the National Academy of Sciences of the United States of America, 104(39), 15200–15205. Meinzen-Dick, R. S., Brown, L. R., Feldstein, H. S., & Quisumbing, A. R. (1997). Gender, property rights, and natural resources. World Development, 25(8), 1303–1315. Meinzen-Dick, R., Raju, K. V., & Gulati, A. (2002). What affects organization and collective action for managing resources? Evidence from canal irrigation systems in India. World Development, 30(4), 649–666.

93

Naidu, S. (2009). Heterogeneity and collective management: Evidence from common forests in Himachal Pradesh, India. World Development, 37(3), 676–686. Ostrom, E. (1990). Governing the commons: The evolution of institutions for collective action. Cambridge University Press. Ostrom, E. (2000). Collective action and the evolution of social norms. The Journal of Economic Perspectives, 137–158. Ostrom, E. (2005). Understanding institutional diversity. Princeton, NJ: Princeton University Press. Ostrom, E. (2009). A general framework for analyzing sustainability of social-ecological systems. Science, 24(7), 419–422. Ostrom, E., Gardner, R., & Walker, J. (1994). Rules, games, and commonpool resources. University of Michigan Press. Poteete, A. R., & Ostrom, E. (2004). Heterogeneity, group size and collective action: The role of institutions in forest management. Development and Change, 35(3), 435–461. Poteete, A. R., & Ostrom, E. (2008). Fifteen years of empirical research on collective action in natural resource management: Struggling to build large-N databases based on qualitative research. World Development, 36(1), 176–195. Pretty, J. (2003). Social capital and the collective management of resources. Science, 302(5652), 1912–1914. Prezza, M., & Costantini, S. (1998). Sense of community and life satisfaction: Investigation in three different territorial contexts. Journal of Community and Applied Social Psychology, 8(3), 181–194. Qiao, G., Zhao, L., & Klein, K. K. (2009). Water user associations in inner Mongolia: Factors that influence farmers to join. Agricultural Water Management, 96(5), 822–830. Rudel, T. K. (2011). The commons and development: Unanswered sociological questions. International Journal of the Commons, 5(2), 303–318. Runge, C. F. (1986). Common property and collective action in economic development. World Development, 14(5), 623–635. Ruttan, L. (2008). Economic heterogeneity and the commons: Effects on collective action and collective goods provisioning. World Development, 36(5), 969–985. Scott, R., Guo, L., Shen, M., Hughart, A., & Giles, J. (1999). Leaving China’s farms: Survey results of new paths and remaining hurdles to rural migration. The China Quarterly, 158(2), 367–393. Stern, P. C., Dietz, T., & Ostrom, E. (2002). Research on the commons: Lessons for environmental resource managers. Environmental Practice, 4(2), 61–64. Suhardiman, D., & Giordano, M. (2014). Is there an alternative for irrigation reform? World Development, 57(5), 91–100. Tang, S. Y. (1992). Institutions and collective action: Self-governance in irrigation. San Francisco, CA: ICS Press. Taylor, J., Rozelle, S., & de Brauw, A. (2003). Migration and incomes in source communities: A new economics of migration perspective from China. Economic Development and Cultural Change, 52(1), 75–101. Wade, R. (1987). The Management of common property resources: Collective action as an alternative to privatization or state regulation. Cambridge Journal of Economics, 11(2), 95–106. Wang, J., Huang, J., Rozelle, S., Huang, Q., & Blanke, A. (2007). Agriculture and groundwater development in northern China: Trends, institutional responses, and policy options. Water Policy, 9(S1), 61–74. Wang, J., Huang, J., Rozelle, S., Huang, Q., & Zhang, L. (2009). Understanding the water crisis in northern China: What the government and farmers are doing. International Journal of Water Resources Development, 25(1), 141–158. Wang, X., Otto, I. M., & Yu, L. (2013). How physical and social factors affect village-level irrigation: An institutional analysis of water governance in northern China. Agricultural Water Management, 119, 10–18. Wu, H. (1994). Rural to urban migration in the People’s Republic of China. The China Quarterly, 139(2), 669–698. Yee, W. H., Lo, C. W., & Tang, S. Y. (2013). Assessing ecological modernization in China: Stakeholder demands and corporate environmental management practices in Guangdong province. The China Quarterly, 213(2), 101–129.

Available online at www.sciencedirect.com

ScienceDirect