The Problem of Institutional Fit: Uncovering Patterns with Boosted Decision Trees

Abstract

Complex social-ecological contexts play an important role in shaping the types of institutions that groups use to manage resources, and the effectiveness of those institutions in achieving social and environmental objectives. However, despite widespread acknowledgment that “context matters”, progress in generalising how complex contexts shape institutions and outcomes has been slow. This is partly because large numbers of potentially influential variables and non-linearities confound traditional statistical methods. Here we use boosted decision trees – one of a growing portfolio of machine learning tools – to examine relationships between contexts, institutions, and their performance. More specifically we draw upon data from the International Forest Resources and Institutions (IFRI) program to analyze (i) the contexts in which groups successfully self-organize to develop rules for the use of forest resources (local rulemaking), and (ii) the contexts in which local rulemaking is associated with successful ecological outcomes. The results reveal an unfortunate divergence between the contexts in which local rulemaking tends to be found and the contexts in which it contributes to successful outcomes. These findings and our overall approach present a potentially fruitful opportunity to further advance theories of institutional fit and inform the development of policies and practices tailored to different contexts and desired outcomes.

Keywords

• institutional fit
• collective action
• context
• community-based management
• environmental governance
• machine learning