Input towards an empirically calibrated predictive model

This report seeks to provide a proof of concept of Regression Discontinuity (RD) analysis of thresholds in migration-decision-making, with the thresholds being related to the environmental and climate risks faced by populations in migration origin rural areas across study sites in West Africa. The analysis uses well-established analytical frameworks to detect discontinuities, and thus provide the empirical basis for predictive models of altered migration flows. The analysis uses two waves of data
from Ghana and Mali collected directly to enable this RD analysis. The results demonstrate first a divergence in migration flow outcomes between migration propensity in the two countries associated with perceptions of various climate risks, from floods to rainfall variability, from those directly experiencing those risks. Perceived weather shocks, such as the impacts of drought, erratic rainfall, and heat stress, correlate with increased migration rates in Ghana, while the opposite holds in Mali. An increased number of in-situ adaptation measures also lead to higher migration rates in the former, while the opposite holds in the latter. A potential explanation for this divergence could be the diverse levels of development in each of these countries. As Ghanaian households are less resource-constrained than their Malian counterparts, they can draw from a wider range of adaptation strategies, which includes migration. The analysis then highlights that when using measured temperature data, there is a significant dip in migration rates at a statistically significant tipping point in Ghana. Beyond this threshold, there is a continued increase in out-migration rates. This may be interpreted as showing that once temperatures rise above this specific threshold, many individuals initially adjust their migration behaviour but then return to similar migration patterns as before the tipping point. The RD analysis underscores the severe impact of climate stressors on livelihoods but indicates that these stressors may lead to short-term reductions in migration, due to the increasing difficulty or need to increase in-situ adaptation. The results for Ghana, Mali and for the East African countries where these data will soon be available, can be used to generate empirical models with predictive powers, but require careful triangulation with other data sources to ensure policies and actions reflect feasible and equitable adaptation interventions.