Extraction, Contestation, and Conservation: Natural Resource Dependence and Protected Area Designation

Amanda Kohn

If the border dissects a mountain range, as with China and Nepal, the land immediately proximate to the border (say, within 25 kilometers, which is the smallest bandwidth used in the regression discontinuity) is still similar — much of it is in the mountains.12 Thinking of the border as an abrupt, fine line rather than the mountains themselves helps reduce concerns about differences across borders. [...] In the main model, I allow for the functional form of the regression line to be a quadratic in order to not overweight observations very close to the border.15 I use the Imbens-Kalyanaraman (I-K) optimal bandwidth to measure how far from the border grid cells should be included, with an upper limit at 300 kilometers (Imbens and Kalyanaraman 2012).16 Repeating the RD exercise on each of the 314 ter. [...] While these differences are estimates, the nature of the RD exercise means that the error is classical, and because the variable is used as the dependent variable, the resulting coefficient estimates on the independent variables in the second step are not attenuated by the dependent variable’s error term. [...] Robustness Checks To test the robustness of my results, I make changes to the models in both the first and second steps of the research design.19 In the first step, I specify the bandwidth of distance away from the border explicitly, rather than using the I-K method. [...] I provide evidence for this mechanism because it is the “weaker” of the two, given the large body of evidence in the literature for the mobilization and democracy mechanism.