The Who, What, When, and How of Industrial Policy: A Text-Based Approach

Abstract

Since the 18th century, policymakers have debated the merits of industrial policy (IP). Yet, economists lack basic facts about its use. This study sheds light on industrial policy by measuring and studying global policy practice for the first time. We first create an automated classification algorithm for categorizing industrial policy practice from text. We then apply it to a global database of commercial policy descriptions and quantify policy use at the country, industry, and year levels (2009-2020). These data allow us to study fundamental policy patterns across the world. We highlight four findings. First, IP is common (25% of policies in our database) and has expanded since 2010. Second, instead of blunt tariffs, IP is granular and technocratic. Countries tend to use subsidies and export promotion measures, often targeted at individual firms. Third, the countries engaged most in IP tend to be wealthier (top income quintile) liberal democracies. In our data, IP is rarer among the poorest nations (bottom quintile). Fourth, IP is targeted toward a subset of industries and is highly correlated with an industry’s revealed comparative advantage. We show that industrial policy is a prominent feature of the global economy and a far cry from industrial policies of the past.

Keywords: industrial policy, industrial development, supervised machine learning, targeting, text-as-data, natural language processing

JEL Classification: D72,O14,P16