Welfare Effects of Subsidizing Seasonal Migration

Main article

We build and estimate a structural model of migration to help interpret a series of experiments that encouraged rural families in Bangladesh to seek jobs outside their villages during the “lean season.” Our model helps rule out several enticing explanations as to why so many villages pass up opportunities to bolster their incomes through temporary migration, including risk of not finding a city job, credit constraints, and high migration costs. A key innovation of our approach—which was missing from earlier literature—is to carefully consider the non-monetary hardships associated with migration, which include family separation and poor urban living conditions. Once this “unobserved disutility” is considered, our model suggests that temporary migration functions as a form of insurance, with most villagers choosing to migrate only when they have no other options.

By the year 2050, around 2.5 billion more people will be living in urban areas than there are now (United Nations, 2018). Around 90% of these new city dwellers will be in Africa and South Asia, expanding already populous cities like Dar es Salaam and Dhaka. Policymakers in developing nations therefore often discuss whether to take steps to control the rate of new population inflows into their capitals and second-tier cities (Lucas, 2021; McKenzie, 2024).

Yet, rural-urban migration may be a key pathway out of poverty. Urban centers typically offer better earning opportunities for adults and better education and training for young people. Developing countries are rapidly urbanizing (see, for example, Young, 2014; Brueckner and Lall, 2015; Lagakos, 2020), and with it, poverty is urbanizing as well (Duflo et al, 2012).

Experiments related to rural to urban migration in Bangladesh

A series of migration experiments in rural Bangladesh (Bryan et al, 2014; Akram et al, 2018) offer a unique perspective on the potential benefits and costs of out-migration for rural individuals. These studies look at poor villages in the Rangpur region of northern Bangladesh. As part of the studies, migration subsidies were offered to any household with low asset levels that contained individuals willing to leave during the lean season, when farming work in rural areas is scarce. The migration subsidies were relatively small, covering little more than the roundtrip bus ticket needed to get to places like Dhaka, where job prospects were better. Many households eagerly took the subsidies, however, and seasonal migration rates rose by 22 percentage points in villages that were randomly offered the subsidies (compared to control villages).

Households that sent a migrant saw big increases in consumption, of around 30% on average per household member. High-frequency phone interviews with the migrants confirmed that the higher family consumption came from the higher incomes earned in urban jobs, such as construction work or the pulling of rickshaws, which are basically human-powered taxis. 

The migration subsidies were offered only temporarily but migrants chose to travel back to the city in subsequent lean seasons, a year later and three years later. This effect dissipated over time, and migration rates in treatment villages were statistically indistinguishable from those in the control villages five years after the original experiments. If the experiments had any long-term impact on the villagers, it was not apparent in any of the (extensive) data collected.

Modelling the migration decision

The experiments underscore the need to understand why these rural families regularly pass up opportunities to raise their (low) consumption levels by sending a migrant out in the lean season. To help make progress on this question, we built a dynamic migration model with a rich set of migration motives and confronted it directly with the original data from the experiments (Lagakos, Mobarak, and Waugh, 2023). Within this dynamic migration model, rural households are modelled as risk-averse but unable to borrow and only partly able to save to self-insure against fluctuations in village incomes. Temporary or permanent migration is allowed, but it carries a financial cost (the bus ticket and enough pocket money to get started) and, importantly, a non-monetary cost (e.g., being away from home and living in crowded quarters). Cities are also assumed to be risky places, meaning that migration might not lead to smoother incomes. 

We estimate the model to match the experimental data as closely as possible. Starting with the stationary distribution of the model, we offer the same migration subsidies to model households  and compute their take-up, the resulting consumption increases, and other statistics. We make sure the model responses are as close as possible to people’s actual responses to the field experiment with migration subsidies described above.

The estimated model provides little support for the hypothesis that workers cannot leave the village to take urban jobs because they are credit or savings constrained. It only takes about two weeks of savings to migrate and get started in the city. It is unlikely that many people would struggle to come up with sufficient savings to pay the transport costs.

One might imagine that households follow a cutoff migration rule based on assets, where they only move out when their assets are high enough to cover the migration cost plus a bit more. But this is not the case at all. Both in the model and data, we see that migration is more common among households with the lower asset levels, not higher ones (Table 1). The asset threshold is taken as 800 Taka since this is around the size of the original migration transfer. 

Table 1: Migration Rate by Asset and Consumption Level

Panel A: Data
		Assets
		≤ 800 Taka	> 800 Taka
Consumption	Below Median	40	29
	Above Median	36	31
Panel B: Model
		Assets
		≤ 800 Taka	> 800 Taka
Consumption	Below Median	41	28
	Above Median	41	38

Another theory that seems unlikely here is that migration risk is holding back worker migration (as in the famous theory by Harris and Todaro, 1970). Villages are risky too, and, at least in this setting, workers managed to get jobs in cities without too much of a struggle. 

Looking at the variance of log household consumption growth from before to after the lean season—which serves as a simple proxy for consumption risk—there is little evidence that villages are places where consumption stays steady over time (Table 2). In fact, households face significant volatility in their consumption. Nor does migration appear to greatly increase this risk: the distributions of log consumption growth are only modestly higher in treatment villages, though migration rates differ significantly.

Table 2: Variance of Log Consumption Growth

	Control Group				Treatment Group
	Stay	Migrate			Stay	Migrate
Data	0.15	0.18		Data	0.16	0.19
Model	0.18	0.19		Model	0.17	0.19

So why do rural families pass up potentially profitable opportunities to migrate seasonally? The most likely answer suggested by our analysis is that families send a migrant out only as a last resort when work opportunities are particularly scarce and they have used up their savings. At that point, workers are willing to endure the hardship of temporary migration and the cramped living conditions in an unfamiliar place that go along with it. In other cases—when village activities are productive, or they still have savings to draw on—villagers mostly prefer to stay in their home environment.

We conducted an additional “out of sample” test of our theory using a second experiment on the same villages that gave out cash with no conditions whatsoever. A modest number of people in villages offered this unconditional transfer did respond by migrating, but the change in seasonal migration rates between treated and control villages was small and not statistically significantly different from zero. 

We simulated the effects of the same type of unconditional transfers in our model and found a negligible migration response, just as in the data. Our model households—particularly the poorest ones—are better off with the transfer because it helped finance more consumption. But this unconditional transfer didn’t make them more likely to migrate. 

Policy implications

One value of our model is that we can simulate the effects of realistic, large-scale counterfactual policies that would be prohibitively expensive to run as a research experiment. We simulated the welfare effects of a government policy that permanently offers migration subsidies to any villager willing to migrate out in the lean season. These subsidies take the same form as those in the experiments, meaning they are only available for rural people with low assets. However, we finance them by taxing urban households and offer them every period. We assume that households know about this option and plan accordingly.

Table 3 reports the results. Rural households with low assets value these transfers as much as a permanent increase in consumption of 2.3%. Urban households dislike the subsidies because they pay for them through higher taxes. Across all households, the permanent migration subsidies raise welfare by 0.8% and greatly increase the percentage of rural households that send a migrant out during the lean season each year. 

Table 3: Welfare Effects of Permanent Migration Subsidies

	Migr. Endogenous Tax Financed (G.E.)
Rural & Low Assets	2.3
All Rural	1.9
All Urban	-1.3
All Households	0.8
% in Rural Area	66
% of Rural Seasonally Migrating	56
% of Rural with Low Assets	74
Tax Rate (% of Labor Income)	1.3

This is not a large policy—it would require about 1.3% of labor income in tax revenue. While it might be infeasible to implement in reality, the point is that permanent migration subsidies effectively provide insurance, not unlike unemployment insurance programs.

Another value of our model is to ask what the efficient migration decisions would look like, which is the migration rates that would be chosen by a (hypothetical) benevolent social planner looking to maximize welfare. This exercise isn’t meant to guide real-life policies but instead to shed insights, at a basic level, on who should move, and who should receive or give money, in an efficient system.

Interestingly, the planner’s solution calls for a lower rate of seasonal movement than we see in real life. Instead, the planner gives money to those with few assets whenever they face low productivity in the rural area. Relative to the competitive outcome, the planner’s solution economizes on short-term trips taken by those facing hardships in the village. The planner finds it more efficient to simply transfer money to these individuals when they experience a bad shock, sparing them the ordeal of urban migration.

The key insight from this exercise is that the planner provides insurance to poor rural residents who would otherwise be forced to leave villages temporarily and return once they have enough money to make ends meet. This highlights how the possibility of temporary migration serves as an insurance mechanism against the risk of being faced with limited work opportunities in the village and having no savings to fall back on.

This article summarizes ‘The Welfare Effects of Encouraging Rural-Urban Migration’ by David Lagakos, Mushfiq Mobarak, and Michael Waugh, published in Econometrica in May 2023.

David Lagakos is at Boston University and NBER. Mushfiq Mobarak is at Yale University and NBER. Michael Waugh is at the Federal Bank of Minneapolis and NBER.

References

Akram, Agha Ali, Shyamal Chowdhury, and Ahmed Mushfiq Mobarak (2014) “General Equilibrium Effects of Emigration on Rural Labor Markets,” Unpublished Manuscript, Yale University

Brueckner, Jan K. and S. Lall (2015) “Cities in Developing Countries: Fueled by Rural-Urban Migration, Lacking in Tenure Security, and Short of Affordable Housing” in G. Duranton, J.V. Henderson, W.C. Strange (Eds.), Handbook of Regional and Urban Economics, 5B, Elsevier, pp. 1399-1455

Bryan, Gharad, Shyamal Chowdhury, and Ahmed Mushfiq Mobarak (2014) “Underinvestment in a Profitable Technology: The Case of Seasonal Migration in Bangladesh,” Econometrica, 82, 1671–1748

Duflo, E., Galiani, S., & Mobarak, M. (2012). Improving access to urban services for the poor. Abdul Latif Jameel Poverty Action Lab Report.

Harris, John R. and Michael P. Todaro (1970), “Migration, Unemployment and Development: A Two-Sector Analysis,” American Economic Review, 60 (1): 126–142

Lagakos, David, Ahmed Mushfiq, and Michael E. Waugh (2023) “The Welfare Effects of Encouraging Rural-Urban Migration,” Econometrica 91 (3), 803-837

Lucas, Robert E. B. (2021) Crossing the Divide: Rural to Urban Migration in Developing Countries. New York: Oxford University Press