Language, culture and institutions: Evidence from a new linguistic dataset

Abstract

Kashima and Kashima's (1998) linguistic dataset has played a prominent role in the economics of culture, providing the instrumental variables used in two seminal works to identify the causal effect of culture on institutional quality. However, for economists, this dataset has a number of weaknesses, including poor overlap with a key cultural dataset and reliance on sources of linguistic information of uneven quality. We address these issues by constructing a new linguistic dataset based on an authoritative source of linguistic information, the World Atlas of Language Structures. The resulting dataset has greater overlap with key sources of cultural information, is arguably less subject to selection bias, and provides more refined information regarding key dimensions of linguistic variation. We show that the variables in this dataset are significantly correlated with commonly used measures of individualism and egalitarianism. In addition, we reexamine the key results from the literature on culture and institutions, showing the causal relationship between culture and institutions is robust to the use of the new linguistic instruments.