Much of the world is heavily dependent on rain-fed agriculture. As a result, the rural economy is highly seasonal. In some areas during the lean season, work opportunities become scant, wages decline, and food prices rise, leading to seasonal deprivation. Seasonal labor migration is one of the coping strategies households use to smooth against this predictable, recurring event. In Myanmar’s Dry Zone, encompassing Magway, Mandalay and lower Sagaing regions, migration occurs within regions, as well as outmigration to other parts of the country during the lean season to work in brick factories, large rubber plantations (weeding, clearing bushes), construction, tea plantations, and jade mines, among others.
While some households already engage in seasonal migration, others do not, despite significant spatial wage differentials. While this may partly be explained by subjective costs associated with migration (being away from family and friends, loneliness), liquidity constraints may also play a role, especially when a household lives close to subsistence level and the risk of migrating and not finding a job is present (Bryan, Chowdhury et al. 2014). Under such circumstances, small subsidies may act as insurance against not finding a job and thereby help poor households overcome the constraint. A primary aim of the experiment proposed herein is to evaluate if such facilitation of seasonal migration help poor farm households smooth their consumption during the lean season in Myanmar.
While the Bangladesh trial used a pure control group, a pilot trial, the research team conducted in West-Timor, Indonesia, compared cash transfers conditional on migration (CCT) against an unconditional cash transfer (UCT). Results showed that encouragement is needed to induce the experimentation needed for households to exploit this profitable investment opportunity. However, previous research into seasonal migration has not investigated rigorously how migration returns and impacts differ by gender and whether and how those returns and impacts shape households’ decision of whom to send as outmigrant.
They compare a CCT on seasonal migration to a UCT, and a pure control group for 138 villages affected by seasonality in the Dry Zone. For the CCT group, the second part of the subsidy is disbursed when the migrant has been verified to be at the destination (through mobile money transactions), to impose the conditionality. There are three types of migration villages: (i) those where the male household head or another adult male household member is eligible to receive the subsidy to migrate, (ii) those where his wife or another adult female household member is eligible to receive the subsidy to migrate, (iii) those where the household has the free choice of which adult household member to send to migrate. Splitting the migration villages into three types helps us to address the question of gender-specific returns to seasonal labor migration, as well as (through comparison with the “CCT any” group) intra-household decisions of whom to send (a male or female household member).
The inclusion of a UCT group is an improvement over earlier experimental work on seasonal migration (Bryan, Chowdhury et al. 2014), as it provides a natural benchmark (Baird, McIntosh et al. 2011). First, economic theory suggests that if markets function perfectly, then the UCT should have the most favorable welfare effects. On the other hand, when market imperfections lead people to under-experiment with certain profitable investment opportunities (such as seasonal migration), nudges may be needed to encourage those behaviors through tying conditions to cash transfers. Targeting may also be improved through tying migration condition to a cash transfer, as only those who are poor are willing to go through the “ordeal” of migrating (Alatas, Purnamasari et al. 2016). Given lower take-up and better targeting, the welfare impact per dollar spent may well be better for the CCT.
The study further cross-randomizes the gender of the household member receiving the transfer with the amount of the second part of the subsidy the destination to evaluate if the small CCT is enough to overcome liquidity constraints under risk and to evaluate the extent of sorting/selection into migration. The first part of the subsidy, to be received at the offer stage, is the same for all CCT-assigned households (MMK 20,000). The second part of the subsidy is MMK 20,000 for the CCT low group and MMK 40,000 for the CCT high group. This will allow us to address what explains (seasonally exacerbated) rural-urban wage gaps: frictions in movement vis-à-vis selectivity of types (Miguel and Hamory 2009, McKenzie and Rapoport 2010, Hicks, Kleemans et al. 2017). To control for direct income effects and increases in budget sets on end line outcomes, we randomize the CCT low group so that half of this group receives MMK 20,000 at end line (as they were told at baseline and hence expected), and the other group receives a “surprise” MMK 40,000. Findings of the study will thus fit in the literature on the misallocation of production factors (labor, in this case) (Hsieh and Klenow 2009, Restuccia and Rogerson 2013, Restuccia and Rogerson 2017).
The different sizes of transfers will also provide evidence on the size of cash transfers when conditions are tied to the transfer – these may have an optimal size beyond which distortion effects (in our case, selection of types who stand little to gain from migrating) start taking up the offer and migrate. In migration villages, the lottery that determines the amount of the transfer in some cases assigns the household member to control. The contrast between these “unlucky eligible” households and the households in the pure control villages will serve as an estimate of possible spillover effects. Such spillover effects have been documented in the scaled-up seasonal migration interventions in Northwest Bangladesh (Akram, Chowdhury et al. 2017). For instance, wages may increase in villages in response to outmigration