Identifying Reform Priorities: The Role of Non-linearities

IMF Working Papers, December 11, 2020

Source details

Research question: Can countries improve their business climate through reforms in specific policy areas?
Context: Kraay and Tawara (2013) regressed seven different business climate indices on 38 policy indicators and found little agreement across the seven linear models on which policy indicators matter most.
Author’s approach: Revisit the same data but replace linear models with a Random Forest algorithm.
Methodological note: Uses Random Forest and reports consensus across models and heterogeneous marginal effects across countries.

Strong consensus across models on the importance ranking of policy indicators when using Random Forest.
Consistently among the top ten contributors to a better business climate (no matter which business climate index is considered):
- High recovery rates in insolvency proceedings (i.e., cents on the dollar for creditors).
- Shorter border formalities for importers.
- Shorter border formalities for exporters.
- Low costs for starting a business.
The marginal effect of reforms is heterogeneous across countries.
Reform priorities depend on country specific circumstances.

Prioritize reforms that raise recovery rates in insolvency proceedings to improve creditor outcomes.
Reduce border formalities for both imports and exports to enhance business climate indicators.
Lower costs for starting a business as a consistent reform priority across indices.
Tailor reform sequencing and emphasis to country specific circumstances because marginal effects vary across countries.

Data scope referenced: same dataset used by Kraay and Tawara (2013) of seven business climate indices and 38 policy indicators.
Subjects: Economic sectors, Financial crises
Keywords: Business climate, Random Forest, Shapley values, Structural reforms