Evaluating the Disaster Risk Model of Myanmar: From Index to Negative Binomial Regression

This study evaluates the predictive efficacy of conventional disaster risk index (DRI) modelling against empirical generalised linear model (GLM) analysis using data from 71 districts in Myanmar. While the standard DRI framework (Hazard $$\:\times\:$$ Exposure $$\:\times\:$$ Vulnerability) is widely used for risk ranking, its reliance on a composite vulnerability score can mask the underlying factors that shape disaster impacts. This study employed negative binomial regression to model the affected population count, comparing an initial model using a categorical vulnerability index with a subsequent model using disaggregated socioeconomic components. The analysis revealed that the composite index lacked statistical significance, while disaggregation was essential for policy relevance. The model identified four highly significant predictors, with flood risk emerging as the sole dominant hazard (IRR $$\:\approx\:$$ 1.44). Three socioeconomic factors were highly significant: districts with the highest secondary education (IRR $$\:\approx\:$$ −1.74) and the highest conflict index (IRR $$\:\approx\:$$ −0.71) were major contributing factors, while economic precarity, such as unpaid family workers, was an extreme exacerbating factor (IRR $$\:\approx\:$$ 1.52). These findings demonstrate that, compared to vulnerability indices, GLM offers a superior, evidence-based framework for identifying high-impact intervention areas. Shifting resource allocation towards flood mitigation and targeted resilience investments in education and economic stability is recommended.