Forest cover, agricultural, and socio-economic development: A weighted beta-logistic approach with ratio response

Abstract

A beta-logistic generalized linear mixed model was used to study the association of national-level forest cover with selected covariates from a sample of 158 nations over the 1992–2013 period. The model avoided the improper assumption of normally distributed forest cover data and used covariates representing land, economic, and social factors. Agricultural land expansion was the most important factor associated with declines in forest cover. Relationships with other covariates are more nuanced. We found no support for an environmental Kuznets curve; however, a Kuznets-like trajectory with population and education variables was discerned. Population density had a negative marginal effect on forest cover up to 220 people/square kilometer (km2) but changed to positive after this threshold. A turning point for proportion of rural population was found at 22% when the association changed from positive to negative. The threshold for education level was 93% when its association switched from negative to positive. Economic variables inclusive of per capita income and 10-year lagged GDP growth rate had a weak but statistically significant association. In the future, high-income nations are expected to continue moderate growth in forest cover; a few fast-urbanizing developing countries are predicted to keep increasing forest area, but most will be likely to see a decline.