In Asia, poverty is highly concentrated in rural areas. Four out of every five poor households are rural, and many of these households rely on risky, weather-dependent agricultural activities. To address the uncertainty of agriculture, agrarian households manage risk using two primary mechanisms: they share risk with other members of their community, and they migrate to diversify income sources. Therefore, to understand the overall welfare effects of encouraging migration, it is necessary to consider the effects of migration on risk-sharing. Hence, the study titled "Migration and Informal Insurance" examines whether new migration opportunities for rural households change the nature and extent of informal risk-sharing.
The study shows the interaction between migration and social safety nets by examining how new migration opportunities affect the nature and extent of informal risk-sharing within those villages. The Bangladesh experiment provided very poor agrarian households money for the round-trip bus fare (worth USD 8.50) conditional on one member migrating temporarily. The subsidy offer led to a 22 percentage point increase in migration in the first year, household consumption increased by 30 per cent for those induced to migrate, and migration rates remained 11 percentage points higher during the next lean season, one year after the one-time subsidy was removed.
The study finds clear experimental evidence that risk-sharing improved in villages. Migrants develop contacts in the city, allowing them to better obtain jobs. This "asset" can lead to persistence in migration episodes. In many settings, migration is a risky lottery: a household gives up some income in the village for a chance at income in the destination.
The experiment randomly offered some households subsidies to temporarily emigrate from villages in a poor region of northern Bangladesh. This region is prone to a period of pre-harvest seasonal deprivation during September to December known as Monga, which was the original motivation for the experimental intervention. The migration subsidy would be an offer of 600 Taka (USD 8.50) conditional on one person from the household migrating, with an additional 200 Taka given if the migrant reported to our enumerators in the destination. This amount is sufficient to cover the cost of a return bus ticket and a few days food at the destination. The subsidy was offered in the form of a grant in 37 villages or a zero-interest loan in 31 additional villages.
Three measures of income in this analysis are used, i.e.: home income, city income, and total income. City income is around 20 per cent of home income, and households in treatment villages have slightly higher home income and city income than households in control villages. Like income, consumption for households in treatment villages is higher than consumption for households in control villages.
Households in the sample typically contain four members, and around 40 per cent of households sent a migrant over the course of the year. 47 per cent of households saved some amount during the year, but the amounts were very small (under 400 Taka). In contrast, many households gave or received transfers from other households in the community: 57 per cent received transfers, averaging 5,600 Taka, and 18 per cent gave transfers, averaging 2,800 Taka.
Based on the end line data, 65 per cent of migrants migrated to a rural area. 54 per cent of migrants worked in the agricultural sector, 24 per cent worked as non-agricultural day labourers, 10 per cent in the transportation sector and 12 per cent in other sectors. 56 per cent of migrants reported earning less than they expected. Survey data from incentivised migrants in the destination show that 98 per cent of migrants found work, with 89 per cent reporting that the wage was higher than their home wage. Migrants predominantly travelled in groups, with only 22 per cent migrating alone. 70 per cent of migrants split food costs and accommodation with their group members, and 73 per cent of migrants exchanged information about jobs with other members in their group.
Households often engage in direct transfers between each other, presumably as a way to protect themselves against bad income realisations. Actual transfers, as well as the household's self-reported willingness to ask for help, were affected by the migration subsidies. The experiment significantly increased the willingness of households to interact financially. Not only are such intentions affected in villages where migration subsidies were offered, but actual amounts of transfers also increased as a result of treatment. Perhaps the act of migration leads to migrants bringing back gifts for friends in the village, which could be why we observe the increase in "transfers given," but this does not signal a broader improvement in risk sharing.
There are three takeaways from these results. First, there is a strong norm that households would provide and receive financial assistance among each other. Second, the migration experiment significantly increased the willingness of households to participate in these arrangements as well as actual transfers between households. Third, this increase is not limited to households that were induced to migrate: non-migrant households in treatment villages also reported an increase in the ability to use these informal arrangements.
Income has a significant impact on consumption, confirming the absence of full insurance: a 10 per cent increase in total household income corresponds to a 1.6 per cent increase in total household consumption. The migration treatment significantly improved risk-sharing opportunities. Specifically, treatment reduced the effect of household income on household consumption by over seven percentage points.
If migration makes it easier to hide income because some of the income is earned away from other villagers' watchful eyes, risk-sharing can break down. But the findings suggest that hidden income may not be a key constraint limiting risk-sharing in this setting.
A change in the correlation between income and consumption does not necessarily imply a change in risk-sharing across households if the migration subsidy offers also increased the household's ability to save. These households could be using savings as an alternative consumption smoothing mechanism.
Net migration income depends on the state of the world and the job contact, less any net financial cost to migrate. Total household income therefore depends on the realisation of the state of the world in the village, the migration decision, the realisation of the state of the world in the destination, the realisation of the migration asset, and the net financial migration cost.
At the end of the period, 44 per cent of migrants who hold an asset lose it and 66 per cent of non-migrants who still had an active asset lose it, although estimated imprecisely. For migrants who leave without an asset, 79 per cent would acquire one. Although the intervention only occurred in one period, its impact on both migration and risk-sharing is felt over future periods.
The option to migrate leads to a tradeoff for risk-sharing. On the one hand, migration introduces an additional source of income that - in expectation - increases the total resources available to the network and thereby increases the gains from sharing risk. On the other hand, migration may change the private ability to smooth risk and thus change the incentive to share risk. The specific type of subsidy may also matter: a financial subsidy directly affects marginal utility whereas a utility subsidy does not change marginal utility.
The utility subsidy induces migration under autarky. The migration effect under limited commitment also increases as income is riskier, despite the higher baseline rate. The estimated improvement in risk-sharing is largest when income is less risky. Welfare increase is larger under limited commitment than under autarky, reflecting the gain due to the large improvement in risk-sharing. The gap between the welfare gain under autarky and the welfare gain under limited commitment narrows as the gain in risk-sharing falls.
In the case of a mean-preserving spread, there is a clear difference between a financial and a utility subsidy. If income is risky, then providing a financial subsidy provides self-insurance to an individual, increasing the autarkic value to the individual. A financial subsidy also increases the value of migrating under perfect risk-sharing but because income risk is already perfectly pooled the financial subsidy does not provide an additional insurance benefit. Thus, the private return to migrating increases more than the social return, tightening the incentive compatibility constraint and crowding out risk-sharing. A utility subsidy has the opposite effect: because it does not alter marginal utility, it does not provide additional insurance. When income is not risky, independent households are already willing to migrate, and so the utility subsidy simply provides additional utility. When income is very risky, an independent household is not willing to migrate and thus cannot benefit from the utility subsidy. At the same time, the household values having insurance from the network as it allows them to migrate. A utility subsidy does not put additional pressure on the incentive compatibility constraints and so risk-sharing tends to improve.
The probability of migration for someone who did not receive a subsidy was 9.6 per cent higher if the person was in a village where approximately 70 per cent of the residents were offered a subsidy to migrate compared with a control village. This result suggests that migration costs fall when more people migrate, which would be one channel for the utility cost effect we estimate.
The baseline experiment led to an increase in welfare equivalent to a permanent 12.9 per cent increase in consumption, net of the financial subsidy itself, the year the experimental subsidies were disbursed. In comparison, the estimated gain of the experiment without accounting for the risk-sharing response would be a consumption-equivalent gain of 3.4 per cent. In other words, welfare gains are three times higher when accounting for spillover effects of the experiment through risk-sharing.
The model shows that both the financial and utility components of the migration subsidies have important effects on migration, risk-sharing, and welfare and that ignoring the welfare effects stemming from risk-sharing improvements leads to an underestimate of the overall benefits of the subsidies.
This study finds that the permanent subsidy leads to the opposite effect - a crowding-out, rather than a crowding-in - than the temporary subsidy. Making the subsidy permanent would lead to a permanent 2.9 percentage point worsening in risk-sharing. Estimating that a permanent subsidy would generate a 13.5 per cent increase in welfare; this number is just 40 per cent higher than the equivalent value under autarky, reflecting the negative effect on welfare from the loss of insurance. This finding has an important implication for the validity of extrapolating from RCTs to alternative longer-term policies: in this case, the experimental evidence, which is by nature short-lived, has the opposite implications for risk-sharing than the permanent policy.
This study makes several contributions to our understanding of the trade-offs involved with experimentally introducing new income opportunities into a village. It focuses on the interaction between migration and social safety nets and examines how new migration opportunities affect the nature and extent of informal risk-sharing within villages. It is not obvious the direction these effects will take, as they depend on the nature of the intervention and the local context. It shows that the subsidy improved risk-sharing by 40 per cent. The model can produce both improvements and declines in risk sharing. The key forces are the riskiness of the migration option and the expected value of city income. On the other hand, a relatively safe migration option, made easier by the intervention, discourages risk-sharing because it also substantially improves the outside autarky option. A permanent utility subsidy leads to a permanent improvement in risk-sharing while a permanent financial subsidy leads to a permanent decline in risk-sharing. This result serves to illustrate how experimental evidence, which may involve only temporary interventions, may have very different impacts than longer-term policies.
These findings suggest a broader policy implication: the effectiveness of safety net programmes and conditional transfers very much depends on how community networks operate and on the risk/return profile of the activities encouraged. The subsidy led to a positive spillover on risk-sharing, generating larger welfare gains than in the absence of risk-sharing. Nevertheless, it also shows that in slightly different contexts, the opposite result may occur. As a result, it is important to take into account these spillover effects when designing social protection programmes.
Costas Meghir, Yale University, IFS, IZA, CEPR and NBER; [email protected]
Ahmed Mushfiq Mobarak, Yale University and NBER; [email protected]
Corina Mommaerts, University of Wisconsin - Madison and NBER;[email protected]
Melanie Morten, Stanford University and NBER; [email protected]
[This op-ed provides a summary of the key findings of the paper titled "Migration and Informal Insurance" set to be published in Review of Economic Studies.]