We estimate the early effects of the pilot project to Reduce Emissions from Deforestation and forest Degradation (REDD+) in the Brazilian Amazon. This project offers a mix of interventions, including conditional payments, to reduce deforestation by smallholders who depend on swidden agriculture and extensive cattle ranching. We collected original data from 181 individual farmers. We use difference-in-difference (DID) and DID-matching approaches and find evidence that supports our identification strategy. We estimate that an average of 4 ha of forest were saved on each participating farm in 2014, and that this conservation came at the expense of pastures rather than croplands. This amounts to a decrease in the deforestation rate of about 50%. We find no evidence of within-community spillovers.