Over the past few decades, improved firewood cooking stoves have been massively distributed around the world, mainly with the purpose of decreasing fuelwood consumption among rural households. Surprisingly, rigorous "on the field" evidence on the causal impact of these devices is very limited. This paper estimates the impact of an improved stove design distributed in the Northern Peruvian Andes on firewood consumption. To identify the causal effect of improved stoves, it exploits a quasi-experiment related to the improved stove intervention. The evidence indicates that a proportion of households that adopted the new device experienced iron frame failures. These failures were not systematically caused by inadequate usage, installation or maintenance, but by faulty iron frame construction. Moreover, faulty iron frames were randomly distributed, and whether an iron frame was faulty or not, was not ex-ante observable to the beneficiaries. Therefore, an iron frame failure indicator is used as an instrumental variable to identify the causal effect of improved stoves. Improved stove usage appears to reduce firewood consumption by approximately 46% in the study area. © 2013 Elsevier B.V.