We review theory linking poverty to deforestation and examine this link using multiple observations of Costa Rica after 1960. Country-wide disaggregate data facilitate empirical analysis of poverty’s location and its impact on deforestation. If where the poor live is not controlled for, poverty’s impact is confounded with differences between richer and poorer areas. Without controls for location there is no apparent effect of poverty. Using our data over time to implement controls for location, however, we find that the poor are marginalized, on less profitable land. With our controls for location, the poorer areas appear to be cleared more rapidly. This suggests that poverty reduction aids forest conservation. For the very poorest areas, this result is weaker and another effect is found: deforestation in the poorest areas responds less to productivity, i.e. the poorest people appear to have less ability to expand on productive or to reduce on unproductive land.