Mobility or traps? Unpacking the heterogeneity of climate-induced migration dynamics

A report by the World Bank estimates that more than 200 million people could be displaced due to climate change impacts by 2050. Despite the significance of this figure, making accurate projections remains problematic, and scientific evidence on climate-induced displacement is still inconclusive. This lack of clarity hampers conceptual and regulatory frameworks, preventing effective policy design. 

For instance, are there “climate migrants”? And if there are, who are they? These seemingly simple questions still lack a definitive answer. Climate change acts as a ‘threat multiplier’, amplifying existing constrains and worsening socioeconomic and environmental conditions. Additionally, there is uncertainty about its effects on local development, and consequently, on coping and mitigating strategies, such as the temporary or permanent decision to migrate. 

Answering such questions requires innovative approaches that enable a deeper understanding of climate migration dynamics and the socioeconomic factors that shape them.

In a recent review, we found that households’ mobility responses to extreme weather events are heterogeneous and context-dependent. These diverse responses stem from the role played by key geographic and socioeconomic factors — such as wealth and human capital — that vary across households. They define the impact of extreme climate events in a surprising way, as the same event can result either in an increase or a decrease in mobility, depending on whether it pushes people to leave their homes or impoverishes them by forcing them to stay.

This implies that, without unpacking the heterogeneity puzzle, it is not possible to make long-term projections to inform the targeting and the design of climate-resilience policies.
 

Chasing heterogeneity

In a new working paper, we propose novel ways to uncover the relationship between the shock caused by extreme climate events, (im)mobility and adaptation, while also tackling the heterogeneity issue. Our contribution is at the intersection between the New Economics of Labor Migration and the poverty traps literature. We set a data-driven approach based on causal machine learning that allows the estimation of heterogenous climate impacts for different groups based on their characteristics.

We leverage the Nigeria General Household Survey (GHS) — Panel, implemented in four waves between 2010 and 2019, with the support of the Living Standards Measurement Study (LSMS) — the World Bank’s flagship household survey program. Our work integrates this data with information on drought conditions from the Global Standardised Precipitation-Evapotranspiration Index (SPEI) database. 

We identify ‘migrant-sending households’ as those with at least one member migrating between two survey waves, representing 27 percent of the families in our sample. Our climate variable is a well-known proxy for agriculture-relevant rainfall shocks: the average SPEI during the growing season months in the period between the two waves.

The impacts of these shocks are first estimated for each household and then aggregated into group-average effects, which depend on pre-shock household-specific characteristics chosen by the existing literature.
 

A babel of (im)mobility outcomes

We find that, on average, households exposed to rainfall shocks are about 6 percentage points more likely to have a family member who migrates.  This effect is, however, not statistically significant, and it is misleading because it conceals substantial heterogeneity of mobility responses.

As Figure 1 shows, migration responses are very different in size, but also, and more critically, in the direction of the effect — which means that the relationship between rainfall and migration can have two opposite outcomes — suggesting that rainfall shocks can also make migration implausible for certain groups.
 

Figure 1. Heterogeneity of rainfall shock impacts on the probability of sending migrants.

^{Note: Values below zero indicate a reduction in the probability of households sending migrants when experiencing a rainfall shock. Conversely, values above zero indicate an increase in the probability of households sending migrants when experiencing a rainfall shock.}

But where does such heterogeneity come from? We identify pre-shock asset levels, local adaptive capacity, and repeated climate shocks exposure as its main drivers: while adaptive capacity acts as a substitute for migration, low assets and repeated exposure exert a trapping effect on vulnerable households, tightening their liquidity constraints and decreasing their probability of migrating in response to shocks.

Our findings are consistent with recent research suggesting that climate-induced poverty is likely to be the real threat at the global level and that policies that facilitate local adaptation, as well as safe, orderly, and regular mobility to lift vulnerable people out of poverty traps are needed.
 

Key takeaways

Our study identifies three key elements to consider improving climate migration research and policy:

1. Data Scarcity: Knowledge gaps in climate migration are mostly due to data scarcity. Yet, much can be done to inform policymaking with data currently available, our methodology is based on public LSMS microdata that can be easily tailored to different contexts.
2. Longitudinal Surveys: From an operational perspective, and to strengthen the climate migration knowledge base, more and better data is essential. Thus, it is critical to extend existing longitudinal survey systems, such as the LSMS, across longer-term panels, and set them up in climate-vulnerable countries that currently lack them.
   
3. Group-Targeted Policies: From a policy perspective, our empirical evidence suggests that any ‘one-size-fits-all’ approach for managing climate migration is doomed to fail. Instead, it calls for group-targeted programs and adaptation policies to tackle this complex challenge at multiple levels.
   

These recommendations shed new and decisive light on the climate-mobility nexus, enabling more reliable long-term projections and more targeted and effective policy interventions.