but still elevated levels. In Mon, Yangon and Ayeyarwady, exposure is at or below two percent. The geographic concentration of conflict shocks reinforces the analytical point developed below: the rarest shock at the national level is the most concentrated locally, and, as the next subsection shows, the most damaging per exposed household. Figure 5 Household exposure to economic, natural hazard and conflict shocks (by geography) Figure 5. Household exposure to economic, climatic and conflict shocks Share of households reporting exposure, by state/region (%) Kayah records the highest exposure to all three shocks: economic 90.4% · climatic 51.5% · conflict 67.7% Economic shock State/region Climatic shock Kayah 90.4% Kachin 51.5% 85.9% Rakhine 21.4% 74.5% Magway Conflict shock 67.7% 20.8% 18.7% 67.1% 29.8% 51.6% 11.5% Sagaing 63.4% 20.6% 19.8% Chin 61.9% 22.7% 21.1% Nay Pyi Taw 60.4% Yangon 59.4% Mandalay 57.4% 19.6% 54.1% 29.6% 11.7% 31.6% 6.8% Kayin Bago 49.4% Ayeyarwady 48.3% Shan 47.3% Tanintharyi 46.4% Mon 43.0% 24.7% 7.2% 8.2% 1.9% 18.7% 18.1% 4.3% 11.4% 9.2% 1.1% 13.3% 5.9% 2.1% Source: PPS 2025 (UNDP) Differentiating shocks by intensity To isolate the effect of each shock type from underlying household vulnerability, predicted MPI headcounts were compared between households exposed to a given shock and otherwise similar households not exposed, holding structural characteristics constant.13 The resulting amplification effect, that is the percentage-point increase in predicted poverty associated with shock exposure, captures the contribution of the shock itself, separate from the structural risk profile of the household. Table 5 summarizes the three estimates. 13 The amplification effects are estimated from a survey-weighted logistic regression that includes structural household characteristics, namely household size, sex of head, education of head, employment of head, per capita income quintile and migrant status, alongside three binary shock indicators (natural, economic, conflict). Predicted probabilities are computed for each household at observed values of the structural variables, separately for shock-exposed and unexposed scenarios, and averaged across the sample. The amplification effect is the difference between these two averaged predicted probabilities, with confidence intervals derived from the model's standard errors. L AYE R S O F DE P R I VAT I O N : M u l t i di m en si on a l Pov er t y i n M y anmar 25