Despite record economic growth for more than a decade, poverty has remained stubbornly high in Afghanistan, especially in the regions that suffered less from conflict. This paper aims to explain this paradox by combining a model of conflict intensity at the province level over period 2007-14 with a model of consumption at the household level in 2011. Provincial data show that higher levels of conflict were positively correlated with both a larger presence of troops (international and Afghan) and larger aid flows. Household data show that the negative impact of conflict on consumption was more than offset by the positive impact of aid and troops. According to the estimates, Afghan troops contributed more to poverty reduction than international troops, possibly because they spent more locally. The paper uses the estimated models to conduct an out-of-sample validation exercise, focusing on the transition initiated in 2014. The results should be interpreted with caution, as the quantitative models cannot account for strategic shifts in the insurgency and watershed political developments. But they suggest that the reduction in the number of international troops and declining foreign aid flows led to an increase in conflict intensity and a decline in consumption per capita, matching current trends.