Multi-level poverty traps

At the top of the Sustainable Development Goals sit two important and unresolved global challenges of our modern era: eradicating extreme poverty and reducing inequality. Although extreme poverty has decreased in the last three decades, the global poverty rate has recently increased from 7.8 to 9.1 percent due to the pandemic and climate shocks. Therefore, identifying the mechanisms at the micro (individual) and meso (community) levels, their relationship with social mobility, and implementing effective intervention strategies are critical research objectives to achieve sustainable development targets. Numerous models have been developed to examine the complex social interactions giving rise to inequality and persistent poverty, yet few approaches include multilevel dynamics. Here, we introduce a heterogeneous agent-based model to identify conditions underlying poverty traps at different levels. Three distinct regimes emerge in our model. The first regime is a single equilibrium poverty trap characterized by low wealth and an ineffective response to poverty alleviation interventions. The second regime is a double equilibrium trap that displays high wealth inequality and responds relatively well to interventions. In the third regime, all agents prosper economically. We identify key individual (behavioral) and community (institutional) mechanisms that are important for achieving sustainable reductions in poverty and inequality. At the individual level, behavioral characteristics like risk aversion, attention, and saving propensity can lead to sub-optimal diversification and low capital accumulation. At the community level, institutional drivers such as lack of financial inclusion, access to technology, and economic segregation are key drivers of inequality and poverty traps. Our results show that addressing the above factors can yield 'double dividend' -reducing poverty and inequality within-and-between communities and create a positive feedback loops that can withstand shocks. Finally, we demonstrate that our theoretical model can be used as a sandbox for cost-benefit analysis of intervention strategies.