International migration and informal social protection in rural Mozambique

Accepted Manuscript

International Migration and Informal Social Protection in Rural Mozambique Mariapia Mendola PII: DOI: Reference:

S1090-9443(17)30088-1 10.1016/j.rie.2017.04.002 YREEC 719

To appear in:

Research in Economics

Received date: Revised date: Accepted date:

15 March 2017 6 April 2017 7 April 2017

Please cite this article as: Mariapia Mendola, International Migration and Informal Social Protection in Rural Mozambique, Research in Economics (2017), doi: 10.1016/j.rie.2017.04.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

International Migration and Informal Social

Mariapia Mendola†

AN US

April 6, 2017

CR IP T

Protection in Rural Mozambique∗

Abstract

CE

PT

ED

M

This paper uses a unique household survey from two southern regions in Mozambique to examine the extent to which dispatching family labour abroad is a source of informal social protection for households left behind. We do so by studying the relation between migration, remittances and participation in groups that provide informal social services to migrant-sending communities. We distinguish between a range of community associations related to insurance, credit and production provision and we further correct for the endogenous choice of group membership. Results show that migration associated with remittances increase community group participation. In particular, while there is no significant relationship with group membership related to credit or self-help activities, remittance recipients are more likely to join groups that provide insurance. These findings point to the role of migration as a source of social protection for migrant-sending communities at origin. [JEL codes: O17, O15, O12.]

AC

Keywords: International Migration, Social Protection, Group Participation, Mozambique

We thank Ines Raimundo, Basilio Cubula, Elisio Mazive and the whole fieldwork team from the Eduardo Mondlane University and the National Statistical Institute in Maputo for excellent support during the data collection. Financial support from Centro Studi Luca d’Agliano (LdA) is gratefully acknowledged. The usual disclaimer applies. † Università di Milano–Bicocca, LdA and IZA, ¸ ∗

1

ACCEPTED MANUSCRIPT

1 Introduction Poor households in developing regions such as sub-Saharan Africa (SSA) face substantial risks from multiple sources, but have typically limited access to formal insurance, credit and social protection systems. They therefore have to resort to either intra-

CR IP T

household coping mechanisms or informal mutual arrangements with other households (Fafchamps 2003; Dercon, 2002). Here, we investigate how emigration of family labour abroad may be a source of social protection to both the household and the community left behind, by spreading risk over large distances and providing stable liquidity

AN US

inflows (remittances) home. Research on international migration has highlighted that labour mobility may play a positive role in increasing household consumption and investments at origin (Lucas, 1987; Stark, 1991), but little evidence exists on the extent to which migration affects traditional social protection and insurance mechanisms in village economies at origin.

M

In this paper we use a unique and unusually detailed household survey from two

ED

southern regions in Mozambique to explore the extent to which households dispatching family labour abroad - typically to South Africa - are more able to gain access to

PT

informal social protection arrangements in migrant-sending communities. We do so by examining the relationship between migration, remittances and participation in commu-

CE

nity groups that provide shared economic benefits at the level of the community. There is an extensive literature that shows the key role of local indigenous risk-sharing and

AC

finance institutions as intermediaries to provide ex ante measures of risk-management as well as ex post safety net against uninsured risks. At the same time, the literature on the microeconomics of migration in developing countries has pointed out the importance of cross-border labor mobility for households left behind in terms of better

2

ACCEPTED MANUSCRIPT

risk-managing and higher income diversification (Stark and Levhari, 1982; Yang, 2008; Mendola, 2008). Hence, in the absence of an adequate system of social protection, it is unclear a priori how intra-household coping mechanisms, such as migration, (re-) shape local institutions in the community of origin (Fafchamps and Lund, 2003; De Weerdt and Dercon, 2006; Dubois et al. 2008). On the one hand, households with migrants may

CR IP T

turn better-off and hence withdraw from the network. On the other hand, households with back-up income sources such as remittances may be in a better position to engage in an informal sharing agreement and, at the same time, other houeholds may be more willing to share with them.

AN US

This paper investigates this question and adds to the literature by distinguishing between a range of informal social services provided by associations related to insurance (funeral associations, risk-sharing arrangements, etc.), credit (Roscas, groups with joint liability, etc.) and production (self-help groups with income generating activities). Our

M

empirical analysis provides no evidence of a significant relationship with group membership related to credit or self-help activities, while we find that households receiving

ED

remittances are more likely to join groups that provide insurance. Results are robust to the inclusion of community fixed effects and to endogeneity concerns which we tackle

PT

with an instrumental variable strategy. Overall, as the need to manage risk(s) and secure livelihoods can be a main driver

CE

of migration decisions, a derived demand for inter-household forms of social protection and insurance may also arise from the migration process. This sheds new light on the

AC

degree to which migration may generate not only private benefits, but also social returns in the sending regions. The rest of the paper is organized as follows. Section 2 outlines the background

literature. Section 3 describes the context of the Mozambican economy. In Section 4 3

ACCEPTED MANUSCRIPT

we describe the dataset while Section 5 outlines the empirical strategy. Section 6 reports empirical results and Section 7 concludes.

2 Background and Related Literature

CR IP T

Social protection is generally defined as a set of formal or informal mechanisms which enable households either to reduce vulnerability and risk or to cope with economic shocks. The evidence that households in developing countries are exposed to high risk, with important welfare consequences, is plenty (Morduch 1995, Townsend 1995, Der-

AN US

con 2002, Fafchamps 2003). These risks range from individual-specific ones, such as illness, death or unemployment, to economy-wide risks, such as drought or recession. It has long been acknowledged that these shocks have important implications, not least for the poor, including short-term effects on consumption and income volatility (e.g.

M

Townsend, 1994; Udry, 1995). By using household-level panel data and following consumption fluctuations over time, several studies show that households in developing

ED

countries are partly but not fully capable of insuring consumption over fluctuations (e.g. Dercon and Krishnan, 2000; Jalan and Ravallion, 1999; Morduch, 2005). This calls for

PT

the establishment of safety-nets or other social security mechanisms to cope with instability of income and well-being. However, in low-income settings, social services or

CE

formal transfers (for example, cash transfers, health-care, housing assistance, child-care, pensions and other forms of welfare) are rarely in place while people typically organise

AC

their social lives around their kin or their community, which provide mutual assistance through groups, associations and social networks (see Besley 1995; Foster and Rosenzweig 1996; Fafchamps 2005; Udry, 2005). Indeed, when markets are weak or even missing group participation and social arrangements provide households with access to 4

ACCEPTED MANUSCRIPT

informal insurance, credit, employment and production opportunities (La Ferrara, 2002; Fafchamps 2005). The theoretical literature suggests that small groups or networks (for example, Genicot and Ray 2003, Ambrus et al., 2014), with members who care for or trust each other and can punish reneging members, can deal effectively with information and enforce-

CR IP T

ment problems and achieve high levels of social protection (for example, Foster and Rosenzweig, 2001; Karlan et al., 2009; La Ferrara, 2002). The empirical evidence from a disparate set of developing countries is consistent with these predictions (e.g. Ben-Porath, 1980; Platteau, 1991; Fafchamps, 1992). Fafchamps and Lund (2003), for

AN US

example, investigates how rural households deal with income and expenditure shocks in the rural Philippines. By using detailed survey data on gifts, loans, and asset sales they show that gift giving and informal credit allow households to share risk within confined networks of family and friends. These results reject models of risk sharing at

M

village or community level while, instead, support the argument that mutual insurance takes place primarily through networks or within sub-groups. Similarly, De Weerdt and

ED

Dercon (2006) test and reject village level full-insurance in Tanzania. They use detailed panel data from a full census of all risk-sharing networks within a Tanzanian village

PT

to investigate the impact of illness shocks on household outcomes. They find evidence consistent with partial insurance of mainly non-food consumption via networks.

CE

Traditional insurance mechanisms are not without costs, though, and access to them is no guarantee, even despite well-identified ethnic or family links (Dercon 2002). Fric-

AC

tions in group participation may arise because of convex transactions costs, asymmetric information, lack of enforceability or any other process that limits informal exchange (Fafchamps, 1992 and 2005). An interesting stream of the literature on risk-sharing agreements points out that the failure of full insurance can be explained by limited com5

ACCEPTED MANUSCRIPT

mitment. This entails that better-off agents need to realize long-term benefits from sharing with less lucky agents (see Ligon et al. 2002; Dubois et al. 2008). Yet, traditional networks may be under pressure from factors such as population changes, migration and wealth differentiation. The collapse of some of these traditional systems during stress and crises is well-documented and, even for idiosyncratic shocks, evidence suggests that

CR IP T

support is no certainty (Dercon, 2000).

The extensive literaure on migration provides consistent evidence that labor mobility is source of extra income for origin households through remittances that, in contexts of capital market imperfections, allow recipients to relax liquidity constraints and develop

AN US

their productive assets (Yang 2008; Mendola 2008). Moreover, migration has been shown to significantly contribute to the diversification of the family’s income sources such that remittances can play the role of insurance transfers between the migrant and hosueholds members left behind (Stark and Levhari 1982; de la Briere et al. 2002).

M

In a companion paper, Gallego and Mendola (2013) study the interaction between migration and social networks in rural Mozambique. They argue that out-migration may

ED

exert costs on communities because social networks may depend on the future presence of their members. On the other hand, migration and network formation may be com-

PT

plementary if migrants can provide insurance against aggregate level shocks through remittances. They find that households with more current migrants are less likely to

CE

have members in network groups, but households that receive remittances are more likely to join network groups. They interpret these results as evidence that success-

AC

ful migration, mainly embodied by stable remittance-receipt, may both relax participation costs and increase the likelihood that other households will enter co-operative arrangements with migrant-sending households, thereby increasing participation (see also Winters et al., 2001, Davis et al., 2002, Foster and Rosenzweig, 2001). These 6

ACCEPTED MANUSCRIPT

results point to the fact that emigration from low-income settings may have positive but previously undocumented social consequences for the individuals and communities left behind. In particular, if out-migration of a family member is aimed at increasing within-household well-being, then its benefits may be share with the community through networks. While Gallego and Mendola (2013) focus on (formal and informal)

CR IP T

groups that provide economic benefits to their members, Nikolova et al. (2016) find similar results while looking at philanthropic groups in Romania and Bulgaria. They use the recent Gallup Polls for these two countries to show that having close contacts with migrants abroad is consistently positively associated with civic engagement and pro-

AN US

social behavior in the home region. Given the wide definition of social networks and their substantial heterogeneity, though, it is difficult to isolate the different mechanisms through which social arrangements are influenced by emigration. This is due to the fact that emigration of one’s household member - which essentially entails an income effect

M

within the migrant household - may change the economic behavior and incentives of both migrant’s relatives and the rest of the community. By focusing on membership

ED

in groups that provide informal social protection and services at the community level, this paper exploits group heterogeneity as to better separate between different mecha-

PT

nisms through which household labor migration may contribute to the welfare of those left behind. Hence, we distinguish between a specific range of community associations

CE

providing different social services to their members, such as groups related to mutual insurance (funerals, risk sharing arrangements, etc.), credit (Roscas.) and production

AC

provision (income generating activities). From this perspective, if migration increases private returns in terms of higher income, credit and insurance access within the origin hosuehold via remittances, we shall observe a negative effect of having a migrant household member on group participation at origin. On the contrary, finding heteroge7

ACCEPTED MANUSCRIPT

nous effects of migration and remittances on membership across credit, production and insurance groups may shed light on both the nature of the migration process and the broader social returns of such a process in the community left behind.

CR IP T

3 Context Mozambique was characterised by slow economic growth until the beginning of the 1990s. Poor levels of education of economically-active members of households, especially women, high dependency rates in households, low productivity in the family

AN US

agricultural sector, lack of employment opportunities in the agricultural sector and elsewhere were featuring the Mozambican economy over the 1990s. Poor development of basic infrastructures and adequate roads in rural areas, and hence poor integration of rural-urban markets, have been contributing to the isolation of communities in Mozam-

M

bique mainly depending on subsistence agriculture and fishing. Although starting from a low base, Mozambique has realised sustained economic growth over the last decade

ED

and a half, averaging 7.5 per cent per annum between 2000 and 2015. This makes the country one of the highest non-petroleum growth performers in SSA. Mozambique’s

PT

growth until the mid-2000s, though, was largely the result of an economy recovering from a bloody internal war, started after the independence from the Portuguese colony

CE

and ended in 1992. Morover, its economic expansion over the past decade is on account of significant inflows of foreign aid and foreign investment. In 2015, Mozambique was

AC

officially declared to be free of the landmines that were left over from the decades of war. Eliminating the landmines has helped Mozambique’s farmers cultivate crops and graze livestock safely and has given investors access to more natural resources. Despite these improvements, Mozambique is still a country with a poverty rate of over 50 per8

ACCEPTED MANUSCRIPT

cent (PRSP, 2014).1 In the face of such extreme poverty, informal social arrangements between households are an important way of coping with a state of permanent vulnerability and eventually substitute for, or enhance of, existing forms of capital investment. However, the way in which such risk-sharing arrangements are formed and enforced varies with the socio-economic and institutional environment, and Mozambique is pe-

CR IP T

culiar in this respect.

People in Mozambique pre–dominantly organise their social life around their kin, which largely define a vital realm in a person’s life. An individual is "incomplete" if he or she is not linked to an ascent group (including dead ancestors, through spiritism

AN US

or witchcraft) and if he or she does not produce any descendants. However, kinship relations and alliances also reflect the common practice of polygamous marriages and the temporary or impermanent nature of family life in this poor context. Social relations typically extend into non-family forms, like relations with neighbours and xaras (quasi-

M

kin), in addition to church relations, community group participation and friendships of varying degrees of formality. Thus, alliances in southern Mozambique go beyond matri-

ED

monial linkages and beyond the kinship circle. This set of alliances defines a person, and the construction of this network is a subject’s major investment for socio-economic life

PT

in the country. Following this line of argument, social protection is typically provided by social associations and by the members of the community.

CE

Our measure of group participation is standard in the literature (see, for example, La Ferrara, 2002). We further distinguish between the different forms of informal social

AC

services provided by Mozambican community groups. Indeed, functions and purposes 1 On

average, Mozambican growth has been contributing to poverty reduction, even though, the country economy has experienced some major constraints to the actual take off of an inclusive economic development process, in common with other African countries and discussed elsewhere in this issue, such as high inequalities (Fosu, 2017) and inefficiencies in urban manufacturing (Fafchamps and El-Hamine, 2017) as well as in agricultural activity (Bevis et al„ 2017).

9

ACCEPTED MANUSCRIPT

of community-based associations vary a lot in the range of activities carried out in the informal sector. They are typically close to the poor by providing services to their members, who are thought to have no viable alternative. Informal groups are thus seen as being constructed by the poor for themselves, with little or no access to the assets of the richer elements of the population. The range of activities provided broadly includes

CR IP T

finance and credit (Roscas, Ascas, groups with joint liability, etc.), mutual insurance (funeral associations, risk-sharing arrangements, etc.) and production (self-help groups with income generating activities). Burial societies, religious associations and ajuda mutua (mutual help arrangements) are essentially a mean of pooling resources to organ-

AN US

ise and pay for unexpected expenses. There is, in general, only one burial society in a village, there are no fees, but all members are supposed to pay and to provide labour or services when someone is hit by a shock. Self-help groups serve economic functions such as agricultural production and micro-enterprise activities (for example, tree plant-

M

ing, beer brewing, selling food and drinks, etc.). Credit associations or xtique serve the usual functions described in the literature, that is addressing the capital market failures

CE

4 Data

PT

deQuidt, 2017).

ED

and allowing access to financial services to poor borrowers (Ghatak and Guinnane 1999;

The empirical analysis is based upon a tailored household survey of 1,002 households

AC

collected by the author in 2008 from 42 communities (both urban and rural) in 4 districts (Manhica, Magude, Chokweâ, Chibuto) of 2 regions (Maputo and Gaza) in southern Mozambique (see map in Figure 1). Sample households have been selected with a probability proportional to population size estimated from the most recent 2007 Gen10

ACCEPTED MANUSCRIPT

eral Population Census data provided by the Mozambican National Statistical Institute (INE) so that the household survey is representative at regional level. The survey collected detailed information on the demographic characteristics of household members, migration status, household asset endowment, and farm and non-farm occupations of the household head and group membership. With regard to the latter, information on

CR IP T

several dimensions of community group participation was collected, including characteristics of the group, i.e., group related to social insurance or ajuda mutua (risk-sharing arrangements, mutual assistance in daily work, funeral associations), credit or xtique (Roscas, groups with joint liability, etc.) and production, buscato or ganho-ganho (self-

AN US

help groups with income generating activities related to goods and services) (Marsh 2003). Table 1 reports sample characteristics related to international migration experience in the household. The data show that 55 percent of households in the sample have at least one migrant member migrated abroad at least once and the average number of

M

migrants per household is 1.6. Among current migrant-sending households, 60 percent receive remittances from migrants and, most interestingly, 55 percent of them receive

ED

remittance inflows on a regular basis while the rest receive them only occasionally.

[Table 1 about here]

CE

PT

[Figure 1 about here]

Table 2 presents the incidence and characteristics of household participation in groups

AC

and social arrangements. The types of groups considered are insurance related (i.e., burials and religious risk-sharing associations), credit groups (ROSCAs, i.e., rotating saving and credit groups) and self-help groups (farmersâ associations, cooperatives productive and cooperative associations, women groups, civic committees etc.). In our sample, 27 percent of households are member of some groups, with 10% participating 11

ACCEPTED MANUSCRIPT

in either insurance arrangements or in a credit group, and 7% participating in self-help groups. [Table 2 about here] Table 3 presents household characteristics by migration status and group participa-

CR IP T

tion by type. Overall, households belonging to community groups seem to be better off with regard to some indicator of physical and human capital. This is consistent with the idea that group members tend to sort themselves into homogeneous pool of persons with regard to some characteristics such as income, human capital, ethnicity, etc. (La Ferrara,

AN US

2002). Interestingly, there is some heterogeneity across groups in terms of household characteristics, for example, households engaging in credit groups seem to be relatively better off with regard to the other group categories.

M

[Table 3 about here]

ED

5 Empirical Strategy

We estimate a model that relates the household decision to participate in community

PT

groups to household attributes, including migration and remittances, and community-

CE

level characteristics. We specify the decision to join a group for household i in the community j as follows: (1)

AC

P(Yi j = ϑ ) = a0 +a1 Xi j +a2 Hi j +b1 Mi j +b2 Ri j +ei j

in which P(Yi j = ϑ ) is the probability of observing ϑ = [0, 1] outcome of the depen-

dent variabile, conditional on vectors of household and community characteristics (X and H respectively), and on the variables of interest, that is, the number of migrants in the household M and the dummy variable R regarding whether the household receives 12

ACCEPTED MANUSCRIPT

remittances. The dependent variable Y captures household membership in community groups related to (i) insurance (funeral associations, risk-sharing arrangements, etc.), (ii) credit (Roscas, groups with joint liability, etc.) and (iii) production (self-help groups with income generating activities). The vectors of independent controls include indicators of the household and family structure, individual demographics, education, wealth,

CR IP T

and residence characteristics. We estimate the model above with both a linear and a probit model. This is because we distinguish three types of group membership which are not mutually exclusive at household level (i.e., households can belong to more than one at the same time). However, this reduced form model may suffer from potential

AN US

endogeneity bias due to unobservable self-selection of migrants. Therefore, we apply the instrumental variable method to mitigate the potential endogeneity bias of both migration and remittance variables. Similarly to Gallego and Mendola (2013), we use a set of instruments to identify both migration and remittances as follows. In order to

M

identify migration we use (i) a dummy variable regarding whether the household had a migrant before the armed internal war (1984-1992) as a proxy for family migration

ED

networks, which are related to the pre-war period and do not affect the current levels of social capital, unless via migration status. This is because the armed prolonged war

PT

largely destroyed existing social ties and massively displaced refugees and returnees. We further use (ii) the number of male household members in their working/migration

CE

age (i.e., between 20 and 30 years old) in 2005, as, in that year, a free-visa agreement was reached between South Africa and Mozambique. Indeed, as long as we control for

AC

household demographics, the specific gender-age composition at the time of the exogenous migration-policy "shock" occurred in South Africa is randomly assigned and does not affect networks in Mozambique beyond its influence through migration out-flows. In order to identify the remittance equation we use (iii) the short-run deviation in 13

ACCEPTED MANUSCRIPT

rainfall levels, measured as rainfall in 2007 minus the average historical rainfall since 1979; (iv) whether household migrants have a permanent job-contract at destination as another instrument. We argue that variations in rainfall may have an important effect on changes in household income (in a region where most households are either directly or indirectly dependent on agriculture) thereby affecting remittances as well, as a form of

CR IP T

insurance (Yang and Choi, 2007). On the other hand, as long as we control for a large set of household and community socio-economic characteristics, it is unlikely that shortrun covariate shocks such as poor rainfall will affect household membership in social networks directly, unless they make remittance-receipt more likely for migrant-sending

AN US

households. The rationale for including the last instrument lies in the fact that migrant employment conditions at destination are a function of labour market characteristics abroad, and may therefore be related to social networks at home only through remittance

M

behaviour.

ED

6 Results

We summarise the estimation results in Table 4, which reports our reduced form esti-

PT

mates of linear probability models, where the dependent variable is a dummy variable taking the value of 1 if the respondent – generally the head of the household – is a mem-

CE

ber of a group providing (i) insurance, (ii) credit, or (iii) productive activities/income. The column in Group 1 reports regression results using the baseline model with no con-

AC

trols, while columns in Group 2 include standard demographic controls including family structure as well as individual characteristics, educational variables, wealth, ethnicity, religion, and community controls (such as the quality of roads, school, health facilities, formal bank and market availability). The columns in Group 3, eventually, include com14

ACCEPTED MANUSCRIPT

munity fixed effects (in which the community is our primary sampling unit) in order to focus fully on the within-community variation only. [Table 4 about here]

CR IP T

As far as controls are concerned, there are some interesting robust results. Higher education of the household head is positively associated with any group participation, while the household head’s working in agriculture - where risk management is typically hard - is positively related to insurance group membership only. With regard to household characteristics, the higher the number of women in the households, the lower the

AN US

likelihood of joining an insurance group, while a marginal increase in the number of children increases insurance group membership. This is in line with the idea that, while household heads are keen to insure the welfare of their offspring, the supply of female labour (at home or on the market) may perform an insurance function at household level,

M

decreasing risk-pooling arrangements with others. Moreover, while the longer residence of households is associated to insurance group membership (where the ex ante level of

ED

trust is critical in the decision to join), the level of household wealth is a positive and robust determinant of credit group participation, which suggests that the latter may be a

PT

’normal good’.

The results on our migration variables show that, while having a migrant household

CE

member (without remittances) is likely to exert a cost on insurance group participation

AC

(which is less precisely estimated when including community fixed effects), the net effect of migration on mutual insurance is likely to be positive through remittance receipt. The latter positive effect of remittances is robust to the inclusion of controls and community fixed effects. However, the same results do not hold for credit and self-help. In table 5 we report Probit marginal effects that, in the case of most saturated models, 15

ACCEPTED MANUSCRIPT

show that an extra migrant in the household decreases the likelihood to engage in insurance groups by 3 percent whilst households receiving remittances are 14.5 percent more likely to participate in informal insurance arrangements. No significant effects are

[Table 5 about here]

CR IP T

found for credit or self-help groups.

However, as mentioned above, reduced form model estimates may be biased by some unobservable characteristics which affect both migration behaviour and household group membership. In order to correct for this possibility, we use an IV strategy to

AN US

address the causal impact of migration and remittances on participation in social safetynets in the households left behind. Using the set of instruments examined above, Table 6 presents IV results for group participation, where we use a 2sls estimator. The Fstatistics of all the combinations of excluded instruments and the over-identification

M

Sargan test suggest that the instruments are not weak and that they are valid (first stage regressions are reported in the second panel of the table). IV results show the same net-

ED

effect of migration on participation in different group participation. In particular, while labour out-migration has a negative impact on household group membership related

PT

to insurance and self-help activities, there is an off-setting positive effect of receiving

[Table 6 about here]

AC

CE

remittances on the likelihood to join groups that provide social insurance.

We interpret these results as evidence that emigration of family labour abroad may

be a source of social protection and insurance at origin, by spreading risk over large distances and sending stable remittance inflows home. This is so as informal groups of micro-insurance may be more inclined to accept members that have a regular source 16

ACCEPTED MANUSCRIPT

of income uncorrelated with the rest of the economy. Hence, a lower risk of default associated with higher income stability and diversification provided by migration may have a positive impact on social insurance arrangements at the level of the community. The same does not hold with regard to credit and self-help groups, where the level of monitoring and enforcement may be more limited by the distance. These results are

CR IP T

further consistent with the argument that having a migrant in the household may change the social position of the origin family with respect to the rest of the community. In other words, it may be the case that non-migrant households in the village are more willing to share resources – and risk – with migrant households, while the latter are better off in

AN US

terms of within-household well-being (through remittances and diversification) without necessarly sharing this higher welfare with the rest of the community in terms of higher loans or services provision. Eventually, we shall argue that the within-family dimension of the migration impact is only aspect of it: migration may entail several spillover effects

M

on other households, in both social and economic terms, the consequences of which

ED

should deserve further investigation.

PT

7 Conclusions

This paper has examined the relation between migration, remittances and participa-

CE

tion in groups that provide informal social services in migrant-sending communities in two regions in rural Mozambique. We provide evidence along multiple dimensions of

AC

community-based associations by distinguishing between groups related to insurance (funeral associations, risk-sharing arrangements, etc.), credit (Roscas, groups with joint liability, etc.) and production provision (self-help groups with income generating activities). By using an original household survey containing detailed information on family 17

ACCEPTED MANUSCRIPT

migration status, remittances and participation in social protection arrangements, we find that the net effect of migration on inter-household co-operation is likely to be positive through remittance receipt. In particular, the dominant effect is on membership in groups related to mutual insurance (i.e., funerals association, risk-sharing arrangements, etc.), while we find no robust effects on credit and self-help group participation. The

CR IP T

interpretation of our findings, however, is blurred by potential endogeneity concerns. We have tackled this issue by employing an instrumental variable approach and the same results hold. We read these results as evidence that, in a poor developing setting such as rural Mozambique, even though community groups are open, income risk and

AN US

participation constraints may limit both access to informal associations and their effectiveness in facilitating inter-household co-operation and assistance. Thus, by spreading risk over long distances and decreasing aggregate income risk, migration coupled with remittances may lower participation costs and increase household commitment in mu-

M

tual insurance arrangements at the level of the community. This sheds new light on the

nities at origin.

ED

role of successful migration as a source of social protection for households and commu-

PT

8 References

CE

Ambrus, A., Mobius, M. and Szeidl, A. 2014. Consumption Risk-sharing in Social Networks, American Economic Review, 104 (1).

AC

Ben-Porath, Y., ââThe F-Connection: Families, Friends, and Firms and the Organi-

zation of Exchange,ââ Population and Development Review, 6 (1): 1-30, March 1980. Besley, T. 1995. Nonmarket Institutions for Credit and Risk Sharing in Low- Income

Countries. Journal of Economic Perspectives 9(3), 115-127. 18

ACCEPTED MANUSCRIPT

Besley, T., S. Coate, and G. Loury. 1993. The Economics of Rotating Savings and Credit Associations. American Economic Review 83(4), 792.810. Bevis L.E.M., Conrad, J., Barret, C. and Gray, C. (2017), "State Conditioned Soil Investment in rural Uganda", Research in Economics, This issue. Cox Edwards, A., and M. Ureta. 2003. International Migration, Remittances and

CR IP T

Schooling: Evidence From El Salvador. Journal of Development Economics 72(2), 429-461

Davis, B., G. Stecklov, and P. Winters. 2002. Domestic and international migration

Population Studies 56(3), 291-309.

AN US

from rural Mexico: Disaggregating the eÂects of network structure and composition.

De Quidt, J. Fetzer, T. and Ghatak, M. (2017), "Microfinance, Market Structure and Borrower Welfare: Regulatory Lessons from the Indian Crisis", mimeo LSE. De Weerdt, J. and Dercon, S. 2006. "Risk-sharing networks and insurance against

M

illness," Journal of Development Economics, Elsevier, vol. 81(2), pages 337-356. Dercon, S. 2002. "Income Risk, Coping Strategies, and Safety Nets," World Bank

ED

Research Observer, Oxford University Press, vol. 17(2), pages 141-166. Dercon, S. and P. Krishnan (2000), "Vulnerability, seasonality and poverty in Ethiopia",

PT

Journal of Development Studies, 36 (6): 25-53 Dubois P., Bruno J. and Magnac T. 2008. Formal and Informal Risk Sharing in

CE

LDCs: Theory and Empirical Evidence. Econometrica Vol. 76(4), 679-725. Fafchamps, M. 1992. Solidarity Networks in Pre-Industrial Societies: Rational

AC

Peasants with a Moral Economy. Economic Development and Cultural Change 41(1),147174.

Fafchamps, M., and S. Lund. 2003. Risk Sharing Networks in Rural Philippines.

Journal of Development Economics 71(2), 261-287. 19

ACCEPTED MANUSCRIPT

Fafchamps, M. 2005. Development and Social Capital. Journal of Development Studies 42 (7), 1180-1198. Fafchamps M. and El-Hamine, S. (2017), "Firm productivity, wages and agglomeration externalities", Research in Economics, This issue. Foster, A., and M. Rosenzweig. 1996. Comparative Advantage, Information and the

CR IP T

Allocation of Workers to Tasks: Evidence from an Agricultural Labor Market. Review of Economic Studies 63(3), 347-374.

Foster, A. and M. Rosenzweig. 2001. Imperfect Commitment, Altruism and the Family: Evidence from Transfer Behavior in Low-Income Rural Areas. Review of

AN US

Economics and Statistics 83(3), 389-407.

Fosu, A.K. (2017), Growth, Inequality and poverty reduction in developing countries: Recent Global Evidence, Research in Economics, This issue. Gallego,J.M. and Mendola, M. (2013), Labor Migration and Social Networks Par-

M

ticipation in Southern Mozambique, Economica, 80 (320): 721-759. Genicot, G., Ray, D. (2003). Endogenous group formation in risk-sharing arrange-

ED

ments. Review of Economic Studies, 70, 87â113 Ghatak, M. and T. Guinnane (1999). "The Economics of Lending with Joint Li-

195-228.

PT

ability: A Review of Theory and Practice." Journal of Development Economics 60:

CE

Jalan, J. and M. Ravallion (1999), "Are the poor less well insured? Evidence on vulnerability to income risk in rural China", Journal of Development Economics 58(1):

AC

61-81.

Karlan, D, Mobius, M., Rosenblat, T. and Szeidl, A. 2009. Trust and Social Collat-

eral. Quarterly Journal of Economics, 124(3): 1307-1361. La Ferrara (2002) Self-Help Groups and Income Generation in the Informal Settle20

ACCEPTED MANUSCRIPT

ments of Nairobi, Journal of African Economies 11(1), pp. 61-89. Ligon, E., J. Thomas, and T. Worrall. 2002. Informal Insurance Arrangements with Limited Commitment: Theory and Evidence from Village Economies. Review of Economic Studies 69(1), 209-244.

Review 77(3), 313.330.

CR IP T

Lucas, R.E.B. 1987. Emigration to South Africa.s Mines. American Economic

Lucas, R.E.B., and O. Stark. 1985. Motivations to Remit: Evidence from Botswana. Journal of Political Economy 93(5), 901.918.

Mendola, M. 2008. Migration and Technological Change in Rural Households:

AN US

Complements or Substitutes? Journal of Development Economics 85(1-2) ,150.175.

Morduch, J. 1995. Income Smoothing and Consumption Smoothing. Journal of Economic Perspectives 9(3):103â14.

Platteau, J., âTraditional Systems of Social Security and Hunger Insurance: Past

M

Achievements and Modern Challenges, in Social Security in Developing Countries, E. Ahmad, J. Dreze, J. Hills, and A. Sen (eds.), Clarendon Press, Oxford, 1991.

ED

Stark, O. 1991. The Migration of Labour. Basil Blackwell, Cambridge. Towsend, R. 1994. Financial Systems in Northern Thai Villages. Quarterly Journal

PT

of Economics 110(4),1011-1046.

Udry, C. and Conley, T. (2004) Social Networks in Ghana Economic Growth Center

CE

Working Paper no. 888.

Udry, C. 1994. Risk and Insurance in a Rural Credit Market: An Empirical Investi-

AC

gation in Northern Nigeria. Review of Economic Studies 61. Winters, P., A. de Janvry, and E. Sadoulet. 2001. Family and Community Networks

in Mexico-U.S. Migration. Journal of Human Resources 36(1), 159-184. Yang, D. 2008. International Migration, Remittances, and Household Investment: 21

ACCEPTED MANUSCRIPT

Evidence from Philippine Migrants.Exchange Rate Shocks. Economic Journal 118, 591-630. Yang, D., Choi, H., 2007. Are remittances insurance? Evidence from rainfall shocks

AC

CE

PT

ED

M

AN US

CR IP T

in the Philippines. World Bank Economic Review 21(2), 219-248.

22

AC

CE

PT

ED

M

AN US

CR IP T

ACCEPTED MANUSCRIPT

23

ACCEPTED MANUSCRIPT

9 Figures and Tables

AC

CE

PT

ED

M

AN US

CR IP T

Figure 1: Household survey in Mozambique (Gaza and Maputo Regions)

24

ACCEPTED MANUSCRIPT

Table 1: Incidence of migration and remittances (household level) Mean s.d. 1.59 (1.18) 55% (0.5)

N. of current migrants in the HH HH with migration experience

CR IP T

HH receiving remittances (total) 24% (0.43) HH receiving remittances (out of migrant HHs) 60% (0.49) HH receiving remittances on a regular basis (weekly, 55% (0.48) monthly, trimestrally, yearly) Note. Standard deviations in paretheses.

AN US

Table 2: Social network participation (household level)

M

Participation in any group (dummy) By type Insurance group (total) ROSCAs Self-help groups Other

AC

CE

PT

ED

Note. Standard deviations in paretheses.

25

Mean s.d. 27% (0.45) 10% 10% 7% 1%

(0.35) (0.29) (0.27) (0.13)

ACCEPTED MANUSCRIPT

Table 3: Household characteristics by group membership and migration status Insurance groups Credit groups Self-help groups No Yes No Yes No Yes 55% 49% 54% 24% 23% 24% 40% 35% 39% 46.64 42.73 46.26 45% 14% 43% 46% 52% 47% 9% 34% 10%

69% 22% 47% 46.38 35% 36% 29%

54% 5.24 2.84 1

44% 4.94 2.88 0.71

30% 5.88 3.23 0.72

42% 4.96 2.87 0.69

50% 6.01 3.38 0.91

89%

85%

82%

85%

89%

84% 12% 1% 3% 0.14 28%

88% 9% 1% 1% -0.33 19%

57% 22% 2% 18% 1.64 55%

87% 9% 1% 3% -0.2 21%

64% 36% 0% 0% 0.58 44%

0.28 0.58

0.2 0.65

0.27 0.62

0.2 0.65

0.28 0.61

CR IP T

63% 33% 41% 43.89 40% 42% 18%

M

AN US

HH with int.migration experience 53% HH receives remittances 23% Female HH head 39% Age of HH head 46.67 Head no education 43% HH head education- primary 47% HH head education- secondary or 10% more HH head occupation- farmer 41% Household size 5 N. of females in the HH 2.92 N. of children in the HH 0.66 (<5years-old) Residence 5 or more years 85% (dummy) Ethnicity-changana 85% Ethnicity-Ronga 10% Ethnicity-Chope 1% Ethnicity-Other minorities 3% Wealth index -0.19 Urban area 21% Community characteristics Ethnic fractionalization index* 0.19 Religion fractionalizaiton index* 0.66

AC

CE

PT

ED

Note: *Ethnic and religion fractionalization indexes measure community diversity (when equal to 1 the community is fully heterogeneous). Hence, equal communities are represented by the two first quantiles of each index distribution, and unequal communities are in the top two quantiles.

26

ACCEPTED MANUSCRIPT

Table 4: Impact of migration and remittances on heterogeous group participation-LPM

N. of current migrants in the HH HH receives remittances Female HH head

(2)

Insurance

Roscas

Self-help

Insurance

-0.034** (0.014) 0.116*** (0.041)

-0.010 (0.014) 0.008 (0.035)

-0.003 (0.009) -0.008 (0.031)

-0.020 (0.013) 0.114*** (0.033) 0.012 (0.025) -0.003 (0.003) 0.000 (0.000) -0.027 (0.028) 0.102**

Age of HH head Age of HH head squared HH head educ- primary

HH head occupation- farmer HH operating land HH size N. of females in the HH

M

N. of children in the HH (<5years-old)

Urban area (dummy) Community with paved-road

ED

Residence 5 or more years (dummy) Wealth index

Self-help

-0.008 -0.010 (0.011) (0.010) 0.045* -0.005 (0.027) (0.024) 0.015 0.051*** (0.021) (0.019) 0.003 0.003 (0.003) (0.002) -0.000 -0.000 (0.000) (0.000) 0.048** -0.009 (0.023) (0.020) 0.130*** 0.116***

-0.008 (0.009) 0.028 (0.023) 0.036 (0.031) 0.004** (0.002) -0.000* (0.000) 0.002 (0.025) 0.111* (0.057) 0.033 (0.029) 0.042 (0.025) 0.005 (0.006) -0.002 (0.008) 0.013

(0.016) 0.081**

(0.013) -0.005

(0.012) 0.017

(0.021) 0.076**

(0.009) 0.001

(0.019) 0.011

(0.033) 0.010 (0.007) 0.238 (0.275)

(0.028) 0.015** (0.006) 0.083 (0.228)

(0.025) 0.005 (0.005) 0.102 (0.204)

905 0.127

905 0.093

905 0.114

(0.030) 0.015** (0.007) 0.078* (0.042) -0.064 (0.074) 0.082 (0.052) -0.135** (0.065) -0.065 (0.073) 0.048 (0.083) 905 0.148 42

(0.031) 0.014*** (0.005) 0.048 (0.032) 0.014 (0.031) 0.012 (0.026) 0.211*** (0.054) 0.016 (0.026) -0.014 (0.023) 905 0.238 42

(0.027) 0.010 (0.006) 0.081** (0.036) -0.022 (0.053) -0.005 (0.029) -0.120** (0.046) -0.011 (0.024) 0.082 (0.053) 905 0.157 42

PT

CE

AC

-0.006 (0.013) 0.119** (0.034) 0.027 (0.025) 0.001 (0.002) -0.000 (0.000) 0.052** (0.020) 0.153** (0.058) -0.003 (0.020) -0.008 (0.040) 0.006 (0.006) -0.006 (0.008) -0.006

Community with health facility 905 0.000

-0.027* (0.014) -0.004 (0.045) 0.007 (0.030) -0.002 (0.004) 0.000 (0.000) -0.028 (0.030) 0.109*

(0.056) 0.085** (0.033) -0.014 (0.038) -0.003 (0.009) -0.024** (0.011) 0.057***

Community with a market

905 0.001

Self-help

(0.033) 0.018 (0.019) 0.041* (0.024) 0.007 (0.005) -0.004 (0.008) 0.011

Community with a Bank

905 0.015

Roscas

(0.037) 0.007 (0.021) 0.009 (0.027) 0.002 (0.006) -0.006 (0.009) -0.004

Community with primary school

Obs. R-squared N. of community fe

Insurance

(0.045) 0.107*** (0.025) -0.025 (0.032) -0.006 (0.007) -0.023** (0.010) 0.050***

AN US

HH head educ- secondary or more

Roscas

(3)

CR IP T

(1)

Note. The dependent variable is a dichotomous indicator of household group participation. The model is estimated using a linear probability model (LPM). Further controls (not reported) include religion and ethnicity dummies. Standard errors clustered at the community level are reported in parentethesis in parentheses. *,** and *** denote statistical significance at 10, 5 and 1 percent level, respectively.

27

ACCEPTED MANUSCRIPT

Table 5: Impact of migration and remittances on heterogeous group participation– Probit marginal effects (1)

HH receives remittances Controls Community FE Obs.

(3)

Insurance

Roscas

Self-help

Insurance

Roscas

Self-help

Insurance

Roscas

Self-help

-0.042* (0.022) 0.133*** (0.050) no no 905

-0.011 (0.017) 0.009 (0.038) no no 905

-0.004 (0.010) -0.004 (0.032) no no 905

-0.027* (0.016) 0.125*** (0.048) yes no 905

-0.001 (0.010) 0.021 (0.027) yes no 905

-0.004 (0.005) -0.014 (0.016) yes no 905

-0.032* (0.017) 0.145*** (0.052) yes yes 905

-0.005 (0.012) 0.063 (0.040) yes yes 905

-0.001 (0.009) -0.021 (0.029) yes yes 905

CR IP T

N. of current migrants in the HH

(2)

AC

CE

PT

ED

M

AN US

Note. The dependent variable is a dichotomous indicator of household group participation. The model is estimated using a Probit model. Further controls (not reported) include religion and ethnicity dummies. Robust standard errors are reported in parentethesis in parentheses. *,** and *** denote statistical significance at 10, 5 and 1 percent level, respectively.

28

ACCEPTED MANUSCRIPT

Table 6: 2SLS results of the impact of migration and remittances on heterogenous group participation

Second stage results N. of current migrants in the HH HH receives remittances Overid. Sargan test (Chi−sq. statistic) P-value

Insurance

Roscas

Self-help

0.330* (0.135) 1.135** (0.517)

-0.085 (0.072) -0.344 (0.254)

-0.121* (0.072) 0.393 (0.259)

1.915 0.384

0.296 0.262

1.064 0.587

First stage results

CR IP T

Variables

(Migration) (Remittances) 0.165* 0.078*** (0.064) (0.029 N. of male hh members aged 20-30 in 2005 0.376* 0.073*** (0.050) (0.022) Short-run rainfall deviation 2.146 3.217* (3.760) (1.837) Permanent job contract of migrants 0.837*** 0.250*** (0.095) (0.039) Joint F−statistic all instrument(s) 43.5 25.59 p-value 0.00 0.00

ED

HH controls Community indicators Observations

M

AN US

Relatives migrated before the war

YES YES 905

YES YES 905

YES YES 905

AC

CE

PT

Note. The dependent variable is a dichotomous indicator of household group participation. The model is estimated using a stage least squares estimator (2SLS). Further controls (not reported) are the same as in previous tables. Robust standard errors clustered at the community level are reported in parentethesis in parentheses. *,** and *** denote statistical significance at 10, 5 and 1 percent level, respectively.

29