Establishing legal protection for forest areas is the most common policy used to limit forest loss. This article evaluates the effectiveness of Indonesian forest protected areas introduced between 2000 and 2012. Specifically, we explore how the effectiveness of these parks varies over space. Protected areas have mixed success in preserving forest, and it is important for conservationists to understand where they work and where they do not. Observed differences in the estimated treatment effect of protection may be driven by several factors. Indonesia is particularly diverse, with the landscape, forest and forest threats varying greatly from region to region, and this diversity may drive differences in the effectiveness of protected areas in conserving forest. However, the observed variation may also be spurious and arise from altering degrees of bias in the estimated treatment effect over space. In this paper, we use a difference-in-differences approach on treated observations and matched controls to estimate the effect of each protected area and then differentiate between the true variation in protected area effectiveness and spurious variation driven by several sources of estimate bias. We find that the change in primary forest cover ranges from a decrease of 3.4% to an increase of 5.3% and the effect of several parks significantly differ from the national average effect of protection of 1.1%. Potential biases may affect estimates in two parks, but results consistently show Sebangau National Park has significantly greater effectiveness while two parks are significantly less able to protect forest cover than the national average.