Essays on the political economy of mass media and representation

Abstract

This thesis consists of three self-contained papers on the political economy of mass media and the economics of representation, all of which deal with the identification of subtle effects, if any. The first paper explores the use of machine learning (ML) and natural language processing (NLP) in detecting political media slant in the mainstream media of Singapore. This paper finds robust evidence of slant towards the ruling party only when using newly defined measures of coverage accuracy—as opposed to the usual measures of coverage intensity. The additional methods from ML and NLP provide measures from the rich textual data, which help deal with identification. The second paper tests the assertions that an increase in women representation on the corporate boards of directors improves firm financial outcomes. The identification strategy exploits the institution setting in Singapore where firms have varying linkages to the government, in the 2000-17 period where there is a substantial increase in women representation in politics. The main finding is that any observed link between female board representation and better financial performance can be attributed to permanent factors, such as corporate culture. The last paper turns to the context of the U.S. elections right after the peak of the women’s MeToo movement in 2018. Using returns from the House elections and county-level measures of the interest in the movement from Twitter data, this paper finds the expected advantage for Democratic women candidates and a disadvantage for the Republican men candidates, but only in those counties with high existing Republican support.