Comparing Inductive and Deductive Modeling of Land Use Decisions: Principles, a Model and an Illustration from the Philippines

Abstract

Understanding the causes of land use change is of great importance for issues of tropical deforestation, agricultural development and biodiversity conservation. Many quantitative studies, therefore, aim to link land use change to its causal ‘driving forces.’ The epistemology of virtually all these studies is inductive, searching for correlations within relatively large, sometimes spatially explicit, datasets. This can be sound science but we here aim to exemplify that there is also scope for more deductive approaches that test a pre-defined explanatory theory. The paper first introduces the principles and merits of inductive and more deductive types of land use modeling. It then presents one integrated causal model that is subsequently specified to predict land use in an area in northeastern Philippines in a deductive manner, and tested against the observed land use in that area. The same set of land use data is also used in an inductive (multinomial regression) approach. With a goodness-of-prediction of 70% of the deductive model and a goodness-of-fit of 77% of the inductive model, both perform equally well, statistically. Because the deductive model explicitly contains not only the causal factors but also the causal mechanisms that explain land use, the deductive model then provides a more truly causal, as well as more theory-connected, understanding of land use. This provides land use scholarship with an invitation to add more deductive (theory-driven and theory-building) daring to its methodological repertoire.