The forest area of the Philippines declined in twenty years, during 1970 - 1990, from about one third to about one fifth of the total land area, i.e. from 10 to 6 million hectares. The relative significance of the various direct and indirect causes of deforestation obviously have changed over the course of time. It has been suggested that during the last decades, the expansion of subsistence or small-scale cultivators into the previously forested upland areas has been the major human activity leading to deforestation. The indirect causes of deforestation include economic, political, demographic, and environmental factors. In this paper, it is hypothesised that the indirect causes increasing the expansion of agriculture into the uplands include factors like population density, conditions on farms in the lowlands, as well as poverty and lack of non-farm employment opportunities. The aim is by no means to present a comprehensive causal model but rather to analyse and understand one part of the complexity related to deforestation.
Deforestation or forest cover changes in the Philippines are analysed using multiple regression with pooled data from 55-64 provinces and from two years, 1969 and 1990. In the empirical models, the dependent variable is the logit-transformation of the forest cover of each province, and the independent variables include population density, the share of small farms, and the tenancy rate of each province. First, a model with pooled data and a common intercept is analysed using the ordinary least squares (OLS) method. In addition, the data are analysed using the so-called fixed effects (FEM) and random effects models (REM). According to these estimated models, forest cover is negatively related to all of the three independent variables used, i.e. the bigger the population density and the larger the shares of small and tenant farms, the smaller the forest cover in each province. Unfortunately, omitting some theoretically relevant variables due to the lack of data may have caused bias in the models, and bivariate correlations make the interpretation of the results difficult.
