Targeted cash transfers, credit constraints, and ethnic migration in the People's Republic of China

Howell, Anthony Working Paper Targeted cash transfers, credit constraints, and ethnic migration in the People's Republic of China ADB Economics Working Paper Series, No. 575 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Howell, Anthony (2019) : Targeted cash transfers, credit constraints, and ethnic migration in the People's Republic of China, ADB Economics Working Paper Series, No. 575, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS190089-2 This Version is available at: https://hdl.handle.net/10419/203417 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ADB ECONOMICS WORKING PAPER SERIES NO. 575 April 2019 TARGETED CASH TRANSFERS, CREDIT CONSTRAINTS, AND ETHNIC MIGRATION IN THE PEOPLE’S REPUBLIC OF CHINA Anthony Howell ASIAN DEVELOPMENT BANK ADB Economics Working Paper Series Targeted Cash Transfers, Credit Constraints, and Ethnic Migration in the People’s Republic of China  Anthony Howell No. 575 | April 2019 Anthony Howell (ton[email protected]) is an assistant professor at the School of Economics, Peking University. The author would like to thank Björn Gustafsson, Ding Sai, Rachel Connelly, Sylvie Démurger, and the participants at the 18th NBER-CCER Annual Conference for helpful comments. This project received funding support from the School of Economics at Peking University and the Natural Science Foundation of China No. 71603009. Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2019 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 632 4444; Fax +63 2 636 2444 www.adb.org Some rights reserved. Published in 2019. ISSN 2313-6537 (print), 2313-6545 (electronic) Publication Stock No. WPS190089-2 DOI: http://dx.doi.org/10.22617/WPS190089-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” inthis document, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. Notes: In this publication, “$” refers to United States dollars. ADB recognizes “China” as the People’s Republic of China. The ADB Economics Working Paper Series presents data, information, and/or findings from ongoing research and studies to encourage exchange of ideas and to elicit comment and feedback about development issues in Asia and the Pacific. Since papers in this series are intended for quick and easy dissemination, the content may or may not be fully edited and may later be modified for final publication. CONTENTS TABLES AND FIGURES iv ABSTRACT v I. INTRODUCTION 1 II. MINIMUM LIVING STANDARD ASSISTANCE PROGRAM BACKGROUND 3 III. DATA 4 IV. EMPIRICAL FRAMEWORK 7 V. RESULTS 8 VI. COMPARING THE DIRECT AND INDIRECT PROGRAM EFFECTS ON MIGRATION 11 A. Potential Mechanisms 12 B. Heterogeneous Credit Constraints across Ethnic Groups 14 VII. CONCLUSION 16 APPENDIX 1 17 APPENDIX 2 18 REFERENCES 19 TABLES AND FIGURES TABLES 1 Official Statistics for the Rural Minimum Living Standard Assistance Program of the 4 People’s Republic of China 2 Minimum Living Standard Assistance Coverage by Geographical Location 6 3 Minimum Living Standard Assistance Coverage, Migration and Human Capital by Ethnicity 6 4 Total Effect of Minimum Living Standard Assistance Cash Transfer on Migration 9 5 Checks for Confounding Factors and Robustness 11 6 Nonlinear, Indirect, and Direct Minimum Living Standard Assistance Effects on Migration 12 7 Direct and Indirect Effects of Minimum Living Standard Assistance on Private Transfers 14 and Migration 8 Direct and Indirect Effects of Minimum Living Standard Assistance on Migration 15 by Ethnic Groups FIGURES 1 2012 China Household Ethnic Survey Rural Sampling Locations 5 2 Probability of Transferring or Lending Money to Another Household 13 ABSTRACT This paper relies on recent proprietary data from the People’s Republic of China’s (PRC) poor rural minority areas to examine the importance of credit constraints on internal labor migration. Specifically, a liquidity shock via the PRC’s minimum living standard assistance (MLSA) program is decomposed into its direct and indirect parts. The institutional features of the MLSA program permit an identification strategy that relies on a set of verifiable assumptions and an instrument variable framework. The results reveal that the direct effect on migration of MLSA is negative, although the net effect is positive driven by the large indirect effects, which are twice as large for ethnic minorities compared to the Han majority. Subsequent evidence further suggests that the main mechanism behind the indirect effect is informal interpersonal lending fostered by risk-sharing strategies. The findings imply that once liquidity is injected into a village it gets circulated in the community, stimulating migration particularly within credit-constrained minority communities. Keywords: ethnicity, indirect effect, liquidity constraints, migration, risk-sharing mechanisms, targeted cash transfers JEL codes: C21, J18, J61, R23 I.  INTRODUCTION Internal (rural-to-urban) migration has led to a significant reduction in rural poverty in the People’s Republic of China (PRC) (Ravallion and Chen 2007, Knight 2013) and across the developing world. A number of benefits exist for individuals and households that engage in migration. Most notable are the remittances that get sent or brought back to the rural origins, which have for a long time been the largest contributor to rural household income growth in the PRC (De Brauw et al. 2002). Beyond remittances, the skills that migrants acquire in the destination and bring back to their rural origins are also important, as return migrants are more likely to engage in higher profit-driven entrepreneurial activities (Démurger and Xu 2011). Various barriers to migration exist, however, that prevent would-be migrant households from leaving their rural origins in search of employment in the city. Credit constraints is one important such barrier that affects poorer households, in particular (McKenzie and Rapoport 2010). A number of recent studies test whether credit constraints exist by examining how a liquidity shock from some type of program intervention affects the migration decision (Bryan et al. 2014, Angelucci 2015). In the PRC, a small but growing number of recent studies is understanding how migration is affected by the removal of credit constraints (Eggleston et al. 2016, Cai 2015) and other migration barriers (Pan 2016, De Brauw and Giles 2017). Relying on a simple theoretical model of migration with liquidity constraints, I examine the importance of credit constraints by estimating the impact of a liquidity shock via targeted cash transfers on the decision to migrate in the PRC. The intervention of interest is the minimum living standard assistance (MLSA) or Dibao, the largest antipoverty program in the PRC, and perhaps the world. The MLSA program provides unconditional cash transfers to participants who are subject to eligibility rules and must meet both income and assets criteria set by local governments. The program has absolutely no rule regarding migration or repayment. The main objective of this study is to not only study the direct effect of the MLSA program through the individual decision of program beneficiaries to migrate, but also to study the indirect effects that get promoted throughout the entire community. In this case, the direct effect measures the individual response of households to the program benefit. By contrast, the indirect effect results from the interaction of individual responses and shows the impact that the program has on the whole community. It is important to study the indirect effects because program interventions, especially in poor rural villages, are likely to have an effect on the entire community, not just eligible households. This is because the extra liquidity that gets injected into communities can again be transferred from recipient households to ineligible households in the community. Such interpersonal exchanges between neighbors, friends, and relatives provide an important informal channel for households to obtain credit, especially in the absence of well-functioning financial markets (Besley and Levenson 1996, Fafchamps 2011). Most of the existing studies, however, examine only the direct effect of a liquidity shock but do not consider the potential indirect effects of a liquidity shock on the entire community. Angelucci and Giorgi (2009), for instance, study both direct and indirect effects of a large social program on consumption, yet the authors’ definition of the direct effect is equivalent to the ‘effect on the treated.’ It is important to consider that ‘treated’ households are also subject to spillover effects, since externalities may arise even if all households participate in the social program. 2 | ADB Economics Working Paper Series No. 575 Moreover, it remains not very well understood how the direct and indirect effects of a liquidity shock on migration vary across different types of households. In the PRC, for instance, ethnicity is likely to play an important role that conditions the effects (direct and indirect) of a liquidity shock on migration. Relative to the Han majority, the PRC’s ethnic minorities tend to have lower rates of migration (Howell, Gustafsson, and Ding 2017), receive fewer remittances that get sent back to the rural origins (Howell 2017), and exhibit significantly higher rates of poverty (Gustafsson and Ding 2009). A natural question that arises is whether ethnic minorities’ lower mobility is due, at least in part, to facing comparatively higher credit constraints that preclude them from engaging in migration. Extant literature from the United States contends that ethnic minorities tend to face larger binding credit constraints due to a moral hazard problem, precluding them from having the same opportunities as the ethnic majority to interact with formal financial institutions (Bond and Townsend 1996). If a similar situation exists in the PRC, then ethnic minorities may be more strongly affected by a liquidity shock via either direct or indirect effects. To account for questions about ethnicity and migration, I compare the direct and indirect effects of the MLSA program on the migration decision for Han and ethnic minority households. To do this, I rely on proprietary household data obtained from the China Household Ethnic Survey (CHES) project. 1 The CHES data offers the first and only source of detailed socioeconomic, employment, and demographic information for Han and ethnic minorities and is nationally representative of the PRC’s ethnic minority areas. Over 7,000 households are included in the sample, half of which are from ethnic minority groups, including Hui, Tibetan, Uyghur, Miao, Zhuang, Dong, and many others. The obvious drawback of the CHES data is that the data are observational rather than being obtained from a random controlled trial—the gold standard for impact evaluations like the one studied in this paper. Moreover, the CHES project contains only one wave of information from 2011, presenting additional estimation difficulties in identifying program effects. One key identification issue, for instance, is that the assignment of participants into the MLSA program is not random. Nevertheless, the total program effect can be identified by comparing program coverage across villages, but not comparing individuals in the same village. Identification of the total effect is based on the institutional features of the MLSA program. That is, coverage is strongly driven by observables meaning that program coverage at the village level is likely to be independent of unobserved social conditions once differences in economic development across village conditions and location fixed effects are controlled for. As a robustness check, the instrumental variable (IV) approach is used to verify the independence assumption. Next, the total program effect can be broken down into its indirect effect based on a set of verifiable assumptions, and the direct effect can be estimated as the difference between the total and indirect parts. This paper attempts to help advance a rapidly evolving body of literature that studies the role of credit constraints on migration. The main contribution to the literature is the attempt to study the direct and indirect effects of targeted cash transfers on rural-to-urban migration in the PRC and comparing the effects across different ethnic groups. This study also potentially provides important guidance for policy makers in the PRC. Understanding how targeted cash transfer programs help reduce credit constraints and spur migration could prove to be vital in helping to alleviate, at least 1 See the appendix section in Gustafsson et al. (2018) (in Chinese) for detailed discussion about the CHES project and data-collection process. Targeted Cash Transfers, Credit Constraints, and Ethnic Migration in the People’s Republic of China | 9 While adding county-level fixed effects controls any unobserved social conditions that may simultaneously influence program coverage at the county level and the migration decision, they do not eliminate concerns about within and between village targeting. In Column (3), an extensive set of village-level controls are included into the model that take into account differences in development conditions across villages. The coefficient on MLSA in Column (3) increases only slightly in size suggesting that the results are not biased by unobservable social conditions at the village level. A 10 percentage point increase in local coverage raises the probability of migration level by 0.34 percentage points. Table 4: Total Effect of Minimum Living Standard Assistance Cash Transfer on Migration Migrant Household (1 = Yes, 0 = No) OLS IV Dependent Variable (1) (2) (3) (4) (5) (6) MLSA coverage –0.043 0.032 0.034 0.037 0.039 (0.007) (0.009) (0.010) (0.017) (0.018) MLSA participant 0.036 (0.015) Household Characteristics Ethnic minority –0.014 –0.023 –0.020 –0.025 –0.022 –0.026 (0.006) (0.008) (0.009) (0.013) (0.013) (0.014) Prop. of females –0.039 –0.015 –0.007 –0.019 –0.012 –0.022 (0.019) (0.019) (0.025) (0.019) (0.027) (0.040) Number of working adults 0.135 0.118 0.093 0.119 0.093 0.101 (0.010) (0.010) (0.013) (0.010) (0.013) (0.021) Average age –0.126 –0.158 –0.171 –0.157 –0.173 –0.158 (0.017) (0.017) (0.022) (0.017) (0.022) (0.037) Average education 0.002 0.004 0.005 0.007 0.008 0.006 (0.001) (0.001) (0.001) (0.001) (0.003) (0.003) Prop. fluent in mandarin –0.017 –0.014 –0.024 –0.016 –0.026 –0.032 (0.008) (0.008) (0.011) (0.008) (0.011) (0.020) Prop. household members –0.206 –0.193 –0.156 –0.156 0.107 –0.116 with disability (0.138) (0.112) (0.103) (0.109) (0.092) (0.078) Dependency rate –0.201 –0.197 –0.187 –0.194 –0.184 –0.152 (0.010) (0.010) (0.013) (0.011) (0.014) (0.071) Political connection –0.014 –0.005 0.008 –0.002 0.012 0.019 (0.007) (0.007) (0.009) (0.008) (0.011) (0.024) Per capita arable land (mu) –0.002 –0.001 –0.001 –0.001 –0.001 –0.002 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Number of durable goods 0.009 0.012 0.010 0.021 0.017 0.024 (0.004) (0.004) (0.004) (0.004) (0.004) (0.010) Local Village Conditions Per capita income (CNY) 0.048 0.069 0.054 (0.016) (0.028) (0.013) Prop. households with poor 0.096 0.071 0.042 sanitation  (0.029)  (0.023)  (0.017)  Village per capita arable land (mu) –0.006 –0.006 –0.006 (0.002) (0.002) (0.002) continued on next page 10 | ADB Economics Working Paper Series No. 575 Migrant Household (1 = Yes, 0 = No) OLS IV Dependent Variable (1) (2) (3) (4) (5) (6) Prop. farm production 0.001 0.001 0.001 (0.000) (0.000) (0.000) Infrastructure investment 0.014 0.011 0.010 (0.008) (0.010) (0.011) Primary education investment 0.036 0.038 0.030 (0.011) (0.011) (0.017) Village political connection 0.006 0.006 0.009 (0.008) (0.008) (0.010) Natural disaster  0.006  0.010  0.022  (0.010) (0.012) (0.034) Mountainous area 0.055 0.070 0.091 (0.020) (0.032) (0.075) Distance to primary school (km) 0.003 0.001 0.001 (0.004) (0.006) (0.006) Distance to transportation center –0.001 0.001 0.005 (km) (0.004) (0.005) (0.013) County-level fixed effect No Yes Yes No Yes Yes Adjusted R-squared 0.045 0.290 0.328 0.309 0.332 0.349 No. of households 7,767 7,767 7,767 7,767 7,767 7,767 IV = instrumental variable, km = kilometer, MLSA = minimum living standard assistance, OLS = ordinary least squares. Notes: Standard errors are reported between parentheses and are clustered by village. Column (1) presents the regression coefficients obtained by OLS. Column (2) adds county-level fixed effect. Columns (3)–(5) present the coefficients from the IV regression with ‘MLSA coverage’ and ‘MLSA participant’ instrumented by village per capita income in 1992; mu = 0.16474 acres or 666.67 square meters. Source: Author's calculations based on China Household Ethnic Survey 2012 data.  Column (4) relaxes Assumption 1 by instrumenting MLSA coverage with historical county income information. As predicted by Proposition 1 (e.g., selection bias does not affect the estimates for the total effect of program coverage), the coefficient on MLSA coverage is not significantly different from the one in Column (3), albeit slightly larger in magnitude. Column (5) replaces the local MLSA coverage variable with the dummy of individual benefit using the same IV approach as in Column (4). As expected from Proposition 1, the estimated coefficient barely changes because the local-level IV makes observations be compared between villages and not within villages. Therefore, the coefficients of local coverage may be interpreted as the total effect of the program on eligible individuals in case they receive the benefit. Checks for Confounding Factors and Robustness Table 5 takes into account the possibility that confounding factors could be driving the results above. One key concern is that the MLSA program coverage may be strongly correlated to other social programs. 2 In order to test whether the results are driven by other antipoverty efforts, Columns (1) and (2) present the OLS and IV estimates after removing from the sample the households that received cash transfers from the five-guarantees (Wubao) program, the PRC’s second-largest antipoverty program. 2 Note that Columns (4)–(5) in Table 4 above add several village characteristics that help to control, in part, for other social programs, although their coverage is not directly taken into account. Table 4 continued Targeted Cash Transfers, Credit Constraints, and Ethnic Migration in the People’s Republic of China | 11 The size of the coefficients are very similar to above, thus indicating that the estimated results do not seem to be a consequence of other targeted cash transfers like Wubao. Table 5: Checks for Confounding Factors and Robustness Migrant Household (1 = Yes, 0 = No) Confounding Factors  Falsification Wubao Program Emigration Test Only Older and Higher Educated Sample OLS IV OLS IV OLS IV Dependent Variable (1) (2) (3) (4) (5) (6) MLSA coverage 0.034 0.036 0.037 0.040 0.028 0.032 (0.011) (0.016) (0.013) (0.020) (0.021) (0.034) County-level fixed effect Yes Yes Yes Yes Yes Yes R-squared 0.213 0.198 0.211 0.209 0.351 0.344 No. of households 6,412 6,412 6,941 6,941 1,546 1,546 IV = instrumental variable, MLSA = minimum living standard assistance, OLS = ordinary least squares. Notes: Standard errors are reported between parentheses and are clustered by village. Columns (1) and (2) present the OLS and IV estimates after removing from the sample the households that received cash transfers from the five-guarantees (Wubao) program, the People’s Republic of China’s second-largest antipoverty program. Columns (3) and (4) present the OLS and IV estimates after removing all households that had changed their local hukou (household registration system) status within the previous 5 years. Columns (5) and (6) present the OLS and IV estimates after including only the sample of households where the household head has at least a high school degree and is at least 30 years of age. Source: Author's calculations based on China Household Ethnic Survey 2012 data.  Another potential confounding factor is that rather than encouraging households to migrate, the MLSA program might have promoted households to relocate to highly covered areas, who may in turn be more likely to send a household member(s) to work back in the original location or some other place. Columns (3) and (4) present the OLS and IV estimates after removing all households that had changed their local hukou (household registration system) status within the previous 5 years. The results are similar to the ones presented above, indicating that the estimated effects are not likely to be due to changes in the composition of workers in the labor force, but rather to changes in their decisions. Next, a ‘falsification test’ is offered by examining populations that should not be directly affected by the MLSA program coverage. Columns (5) and (6) present the OLS and IV estimates after including only the sample of households where the household head has at least a high school degree and is at least 30 years of age. The results return statistically insignificant coefficients, helping to confirm that the findings are driven by the decisions of households in which the program is intended to affect.  VI. COMPARING THE DIRECT AND INDIRECT PROGRAM EFFECTS ON MIGRATION Table 6 reports the results for the direct and indirect effects of the MLSA program on migration. As outlined in Proposition 2, in order to estimate the indirect effect it is first necessary to verify that it is homogenous. If the total effect of the program is linear, then the indirect effect is homogenous for the chosen sample. The first column shows that the quadratic term for the total effect of local MLSA coverage is nearly zero and not statistically significant. 12 | ADB Economics Working Paper Series No. 575 Table 6: Nonlinear, Indirect, and Direct Minimum Living Standard Assistance Effects on Migration Migrant Household (1 = Yes, 0 = No) All Sample Non-MLSA Participants Only All Sample OLS OLS IV OLS IV Dependent Variable (1) (2) (3) (4) (5) MLSA coverage 0.085 0.066 0.071 0.065 0.069 (0.041) (0.020) (0.031) (0.025) (0.033) Squared MLSA coverage 0.031 (0.088) MLSA participant –0.029 –0.034 (0.009) (0.015) R-squared 0.212 0.216 0.196 0.206 0.199 County-level fixed effect Yes Yes Yes Yes Yes No. of households 7,767 6,269 6,269 7,767 7,767 IV = instrumental variable, MLSA = minimum living standard assistance, OLS = ordinary least squares. Notes: Standard errors are reported between parentheses and are clustered by village. Column (1) presents the OLS model with quadratic effect of MLSA coverage. Columns (2) and (3) present the OLS and IV estimates for the indirect effect on individuals who do not participate in the MLSA program. Columns (4) and (5) present the OLS and IV estimates for the indirect effect (MLSA coverage) and direct effect (MLSA participation). Source: Author's calculations based on China Household Ethnic Survey 2012 data.   As a result of not rejecting the linear total effect assumption, the indirect effect of the program can be estimated using only the sample of individuals who are not in the program. Columns (2) and (3) show the results. The indirect effect is positive and statistically significant. The size of the coefficients indicate that the indirect effect is greater than the total effect discussed above, suggesting that the direct effect should be negative. Negative direct effects are indeed revealed in Columns (4) and (5), where the model estimations include program intervention at both the local (indirect effect) and individual level (direct effect). In Column (5), the coefficient on the direct effect is negative and statistically significant, indicating that cash transfers from MLSA program reduces the probability of beneficiaries to engage in migration by 0.34 percentage points. By contrast, the amount of cash transfers to poor villages seems to stimulate outmigration, with a 10 percentage point increase in program coverage increasing the rate of migration of the eligible population by 0.69 percentage points. A. Potential Mechanisms The existing literature contends that the indirect effects of welfare programs are driven by the existence of risk-sharing strategies within communities (Angelucci and Giorgi 2009). Applied to the current context, the MLSA program increases liquidity in the community by stimulating informal credit and private transfers between households. These private transfers, in turn, serve an important liquidity shock to raise migration despite the fact that migrant households do not receive cash transfers directly from the MLSA program. Information on households’ lending and borrowing behavior can be used to help confirm empirically whether or not indirect effects of MLSA result from an increase in interpersonal lending. Targeted Cash Transfers, Credit Constraints, and Ethnic Migration in the People’s Republic of China | 13 Figure 2 shows the probability of MLSA participant households to lend money to another household across the income distribution. The results show that MLSA beneficiaries are indeed more likely to transfer money to another household at each decile except for the top one. Figure 2: Probability of Transferring or Lending Money to Another Household MSLA = minimum living standard assistance. Source: Author's calculations based on China Household Ethnic Survey 2012 data. See the appendix section in Gustafsson, Hasmath, and Sai (2018) (in Chinese) for detailed discussion about the CHES project and data-collection process.  Columns (1) and (2) in Table 7 help confirm the descriptive findings, mainly that MLSA beneficiaries are more likely to become lenders. In Columns (3) and (4), the estimated effect of MLSA coverage on the probability of borrowing for individuals out of the program is positive and significant. A 10 percentage point increase in local coverages raises the probability of borrowing money between 1.42 and 1.55 percentage points. The results suggest that the higher proportion of MLSA beneficiaries in the village, the higher the probability of being financially helped by another household.  14 | ADB Economics Working Paper Series No. 575 Table 7: Direct and Indirect Effects of Minimum Living Standard Assistance on Private Transfers and Migration Migrant Household (1 = Yes, 0 = No)  Indirect Effect (Non-MLSA Participants Only) Lender Borrower Migration OLS IV OLS IV OLS IV Dependent Variable (1) (2) (3) (4) (5) (6) MLSA participant 0.074 0.081 (0.014) (0.032) MLSA coverage 0.142 0.155 0.146 0.165 (0.050) (0.068) (0.051) (0.069) Borrower residual 0.135 0.161 (0.045) (0.064) MLSA coverage × Borrower residual 0.271 (0.091) 0.295 (0.121) County-level fixed effect Yes Yes Yes Yes Yes R-squared 0.212 0.216 0.196 0.206 0.199 No. of households 7,767 6,269 6,269 7,767 7,767 IV = instrumental variable, MLSA = minimum living standard assistance, OLS = ordinary least squares. Notes: Standard errors are reported between parentheses and are clustered by village. In Columns (1) and (2), the outcome is the probability that MLSA participants become a lender. In Columns (3) and (4), the outcome is the probability of borrowing for individuals out of the MLSA program. The columns present the OLS and IV estimates for the indirect effect on individuals who do not participate in the MLSA program. In Columns (5) and (6), the outcome is the probability of migrating. ‘Residual transfers’ are predicted based on the estimated parameters obtained in Columns (3) and (4). The interaction term between ‘residual transfers’ and ‘MLSA coverage’ shows how much the effect of MLSA on migration increases given an unexpected rise in private transfers. Source: Author's calculations based on China Household Ethnic Survey 2012 data.  The parameters estimated in Columns (3) and (4) are used to first predict a residual term for household borrowing. Then, an interaction term is included between the borrowing residual proxy and MLSA coverage in Columns (5) and (6). The results show that the indirect effect on migration is larger where private transfers have increased. While not providing a causal mechanism, the findings support the hypothesis that the indirect effect is stimulated by informal credit through private transfers. B. Heterogeneous Credit Constraints across Ethnic Groups Table 8 compares the size of the direct and indirect effects of the MLSA program on migration for Han and ethnic minorities. Ethnic minorities may face larger credit constraints forcing them to rely more heavily on informal ethnic lending to cover the costs associated with migration. Thus, the parameter coefficient on the indirect effects should be larger for ethnic minority households relative to their Han counterparts. Note that separate regressions are first carried out for each regression to verify that the total effect is linear for each ethnic subgrouping. The results in Columns (1) and (2) confirm expectations. The size of the coefficient on ethnic minorities is twice as large as the one for Han households irrespective of whether OLS or IV is used. The indirect effect is largest for Miao, yet fails to have any statistically significant effect for Tibetans and Uyghurs. Thus, while ethnic minorities appear to generally rely more on interpersonal lending to Targeted Cash Transfers, Credit Constraints, and Ethnic Migration in the People’s Republic of China | 15 finance their migration trips, the strength of those informal networks depends significantly on the particular minority group. Table 8: Direct and Indirect Effects of Minimum Living Standard Assistance on Migration by Ethnic Groups Migrant Household (1 = Yes, 0 = No) Han All Minority Hui Tibetan Uyghur Miao Dong Zhuang Other Minority Dependent Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) Panel A: OLS MLSA coverage 0.044 0.087 0.113 0.040 0.077 0.162 0.079 0.104 0.125 (0.012) (0.022) (0.027) (0.029) (0.051) (0.019) (0.023) (0.039) (0.054) MLSA participant 0.011 –0.034 –0.024 –0.014 –0.016 –0.104 0.037 0.021 –0.178 (0.012) (0.010) (0.018) (0.006) (0.007) (0.055) (0.023) (0.019) (0.019) County-level fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.089 0.062 0.067 0.044 0.040 0.034 0.104 0.149 0.078 No. of households 3,145 4,622 616 329 513 1,070 589 421 1,046 Panel B: IV MLSA coverage 0.048 0.094 0.117 0.052 0.084 0.171 0.084 0.110 0.131 (0.020) (0.034) (0.047) (0.062) (0.072) (0.056) (0.041) (0.045) (0.061) MLSA participant 0.015 –0.032 –0.022 –0.011 –0.010 –0.980 0.042 0.028 –0.067 (0.021) (0.016) (0.021) (0.006) (0.006) (0.067) (0.064) (0.027) (0.024) County-level fixed effect Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.086 0.029 0.067 0.042 0.040 0.034 0.105 0.147 0.042 No. of households 3,145 4,622 616 329 513 1,070 589 421 1,046 IV = instrumental variable, MLSA = minimum living standard assistance, OLS = ordinary least squares. Notes: Standard errors are reported between parentheses and are clustered by village. Column (1) includes only the Han subsample, while Column (2) includes only ethnic minorities. Columns (3)–(9) include the specific ethnic minority groups. Panel A reports the estimated coefficients from OLS, while Panel B reports the IV coefficients. Source: Author's calculations based on China Household Ethnic Survey 2012 data.  It appears that the negative direct effect on migration is driven by the incorporation of a large amount of ethnic minorities into the sample. With the exception of Dong, receiving MLSA benefits for the other ethnic minority groups reduces their likelihood of migrating. It is unlikely that these results are due to differences in credit constraints, since all households that receive MLSA are assumed to be unable to borrow due to insufficient collateral. Why does directly participating in MLSA reduce migration for most ethnic minority groups, but not the Han majority? An alternative explanation is that poorer ethnic minorities face additional obstacles to migration, such as lacking access to information and insufficient language skills. Another reason is cultural or religious preferences that make poorer ethnic minorities less likely to migrate even after credit constraints are reduced. A third plausible explanation, and supported empirically, is that poorer ethnic minority households that receive MLSA assistance prefer to lend through informal ethnic channels to more capable or willing members of their community. 16 | ADB Economics Working Paper Series No. 575 VII. CONCLUSION Low mobility among poor households and increasing ethnic inequality in rural areas, combined with labor shortages in urban areas, present a complex set of challenges that the PRC faces. Boosting migration via reducing credit constraints highlight the potential role that targeted cash transfer programs can have on helping to mitigate, at least partially, some of these social challenges. This paper investigated the impact of MLSA, the largest antipoverty program in the PRC, on the migration decision. Focus is placed on comparing both the direct and indirect program effects across Han and ethnic minority households. The main results are as follows. Despite a negative direct effect, the total net effect of the MLSA program is positive. As the MLSA program grows 10 percentage points, the migration rate increases between 0.28 and 0.35 percentage points on average. The positive net effect is driven almost entirely by the indirect effects, where the liquidity injected into poor rural villages via the MLSA program helps to spur migration. As suggested by Angelucci and Giorgi (2009), a plausible explanation for the positive indirect effect is due to the existence of risk-sharing strategies between households within the same community that help to promote an informal credit market once the program arrives. In line with this explanation, the supporting evidence presented in this paper shows that beneficiary households are more likely to become informal lenders, and nonbeneficiaries are more likely to receive private transfers in areas with higher program coverage. In addition, the indirect effect on migration is larger in villages with a higher increase in private transfers. Importantly, credit constraints are heterogeneous across different households. The indirect effects, for instance, of the MLSA program on labor migration are more than twice as large for ethnic minority versus Han households. In line with the extant literature in the United States (Bond and Townsend 1996), this finding helps to confirm that ethnic minority households face larger credit constraints, which limits the frequency and quantity of credit transactions between potential (Han-dominated) lenders and ethnic minorities. Ethnic minority communities must depend more heavily on interpersonal lending within their own ethnic community as a way to finance their migration trip. While this paper reveals a number of important insights into the impacts of targeted cash transfer programs on migration and economic development more broadly, there are several ways to extend this research. Subsequent studies may wish to compare the direct and indirect effects of other types of programs, including conditional cash transfers and microfinance programs, on migration as well as other household decisions and whether those effects vary across different subpopulations. It is also important to develop further the ethnicity angle and better understand the different types of risksharing strategies for ethnic minorities in the PRC, as well as the reason why ethnic minorities tend to be less likely to migrate compared to Han even after credit constraints are relaxed. APPENDIX 1: SUMMARY INFORMATION Table A1: Summary Statistics Statistic Mean SD Min Max Household Characteristics Migrant household (1 = yes, 0 = no) 0.365 0.481 01 MLSA recipient 0.147 0.354 01 Ethnic minority 0.629 0.483 01 Prop. of females 0.485 0.144 0 1 Number of working adults 4.888 1.684 1 27 Average age of working adults 37.461 6.834 17 64 Average education 6.920 3.405 0 16 Prop. fluent in Mandarin 0.250 0.377 0 1 Prop. household members with disability 0.039 0.048 0 0.518 Dependency rate 0.478 0.510 0 5 Political connection (CPC member = 1, Non-CPC member = 0) 0.173 0.378 01 Per capita arable land 14.055 19.719 0 130 Number of durable goods 12.175 15.264 0 70 Local Village Conditions Per capita income (CNY) 2,693.765 1,407.619 190 9,800 Prop. households with poor sanitation 0.872 0.231 0 1 Per capita arable land (mu) 2.912 4.169 0 42.326 Prop. farm production 63.881 23.079 0 100 Infrastructure investment (1 = yes, 0 = no) 0.559 0.496 01 Primary education investment (1 = yes, 0 = no) 0.814 0.389 01 Political connection (Higher level official from village = 1, otherwise 0) 0.565 0.496 01 Natural disaster (1 = yes, 0 = no) 0.639 0.480 01 Mountainous area (1 = yes, 0 = no) 0.604 0.489 01 Distance to primary school (km) 0.816 0.991 0 7.601 Distance to transportation center (km) 1.787 1.089 0 4.875 CPC = Communist Party of China, km = kilometer, MLSA = minimum living standard assistance, SD = standard deviation. Note: 1 mu is equal to 666.67 square meters. Source: Author’s calculations. 18 | ADB Economics Working Paper Series No. 575 APPENDIX 2. INSTRUMENTAL VARIABLE STRATEGY In the instrumental variable (IV) model, the village 1992 income is used to instrument for program coverage, dĀ v:  d Ā v  Ȗ  Ȗ  Inc  ș v  e iv (2)  Table A2 reports the results from the first-stage regression. The instrument is useful for identification since it is highly correlated to program coverage. The partial R-squared of a regression of program coverage on historical village income, controlling for the other covariates, is 0.609, and the first-stage F-statistic is 66. This strong correlation suggests that weak IV should not be a problem in this case.  Table A2: First-Stage Regression Results Dependent Variable MLSA Coverage (1) (2) (3) (4) 1992 County-level per capita income –0.102*** –0.077*** –0.088*** –0.058*** (0.002) (0.003) (0.005) (0.006) Village characteristics No Yes Yes Yes County-level fixed effect No No Yes Yes Household characteristics No No No Yes Adjusted R-squared 0.386 0.424 0.596 0.609 F-statistic 294.229 150.069 117.861 66.099 MLSA = minimum living standard assistance. Notes: *** denotes 1% level of significance. Standard errors in parentheses. Source: Author’s calculations.