The Sisyphean Pursuit of Evidence for Poverty Traps (WP-26-24)
Dean Karlan, Amol Singh Raswan, Christopher Udry
Much of development economics, both micro and macro, posits theories and explores the empirics of poverty traps. A common theory centers around asset-threshold poverty traps in which marginal returns to capital increase sharply over a certain threshold and financial markets are incomplete. This combination leads households above this threshold to prosper while those below continuously fall back, trapped in poverty. Yet in economics, as best as the authors can tell, far more papers assume such thresholds than empirically demonstrate them. In a key recent exception, Balboni et al. (2022) (hereinafter “BBBGH”) argues that data from Bangladesh identifies clear evidence of such a threshold at slightly above the cost of a cow for recent recipients of a big push transfer program. The authors apply this analysis to data from seven other similar programs in low-income countries and find no evidence of an asset-threshold poverty trap. They then return to BBBGH and argue that their data do not support the identification of a poverty trap, with the point of disagreement being a log transformation challenge and a potential geographic confound. In long-term representative rural panel data from Bangladesh and Ghana the authors find no evidence of divergence away from any particular threshold. While the threshold theory is undoubtedly true for some households, they argue that the pursuit of clear evidence of an asset-threshold poverty trap for a broad population is likely an aspirational goal rendered unachievable by the intersection of a multitude of possible mechanisms as well as heterogeneous agents.