Previous cross-country regression analyses on the causes of deforestation seem to have misspecified the regression models, in which the causes at different levels are mixed, leading to flawed cause-effect relationship between the rate of deforestation and its explanatory variables. The paper focuses explicitly on underlying/policy-oriented causes of deforestation and then examines their relationship with rate of deforestation across 43 countries of Latin America, Africa and Asia. An environmental Kuznets Curve (EKC) relationship between rate of deforestation and income is also tested. Results across all regions and based on panel data technique show that forest and allied (non-forest) sector policies and population density increase deforestation, while forest products export promotion policies, export prices and technological progress decrease deforestation. The effects of per capita income, economic growth and agricultural production are found to be varying in different region. An inverted U-shaped EKC empirically fits for Latin America and Africa while a U-shaped EKC does the same for Asia.