Although questions surrounding industrial policy are fundamental, we lack both measures and comprehensive data on industrial policy. Consequently, scholars and practitioners lack a sys- tematic picture of industrial policy practice. This paper provides a new, text-based approach to measuring industrial policy. We take the tools of supervised machine learning to a comprehen- sive, English-language database of economic policy to construct measures of industrial policy at the country, industry, and year level. We use this data to establish four fundamental facts about global industrial policy from 2009 to 2020. First, IP is common (25 percent of policies in our database) and trending upward since 2010. Second, industrial policy is technocratic and gran- ular, taking the form of subsidies and export promotion measures targeted at individual firms, instead of tariffs. Third, the countries engaged most in IP tend to be wealthier (top income quintile) liberal democracies, and IP is very rare among the poorest nations (bottom quintile). Fourth, IP tends to be targeted towards a small share of industries, and targeting is highly corre- lated with an industry’s revealed comparative advantage. Thus, we find contemporary practice is a far cry from industrial policy’s past, and tends toward selective, export-oriented policies used by the world’s most developed economies.