Building on the contributions of Mather and others, this paper offers an approach for developing a more comprehensive theory of the forest transition. We argue that long-run changes in forest cover in a country or region cannot be separated from the overall pattern of land use changes. Moreover, this pattern is determined by relative land values; forest cover changes over time as the value of one land use relative to the value of its competing use changes over time. However, the actual values that are used to allocate land may be far from optimal; that is, the presence of market, policy and institutional failures can distort economic and political incentives that can lead to bias in favour of one type of land use over the other, and may ultimately explain why a forest transition may be delayed unnecessarily in some countries and regions.