Three Essays in Empirical Economics: Data-intensive Methods Applied to Media Economics and Executive Compensation

Justin Briggs

Major Professor: Alex T Tabarrok, PhD, Department of Economics

Committee Members: Timothy J Groseclose, Bryan D Caplan

Carow Hall, Conference Room
May 09, 2019, 10:30 AM to 12:00 PM

Abstract:

This dissertation consists of three studies in empirical economics. The first two examine the content of television and radio news coverage using computer methods. The third uses data on public companies to examine executive compensation as a function of the companies' ownership structure and concentration. 
 
The first paper studies the political slant of news content from all nationwide American television and radio news networks in a systematic way, including how the slant changes over time. Analyzing 270,000 news programs and nearly 1 billion phrases over the six years from 2010 through 2015, the study performs an objective computational analysis which uses the Congressional Record to learn how language usage varies with ideology. This language usage pattern is applied to each network's complete coverage (as represented by transcripts) in order to obtain a ``slant index'' of the news content. In terms of ordering, the slant index largely accords with public perceptions. However despite perceptions, all national television and radio networks are much more centrist (or balanced) than congressmen. Although there is evidence they may be becoming more ideologically diverse over time, it finds no evidence that any major networks report news with a strictly partisan slant. Moreover, all stations respond similarly to events, indicating that it is more important for politicians to get their story in the news than it is to get it on any particular station. 
 
The second paper examines other qualities of U.S. television and radio news, over the period 2010-2016. These get to the heart of whether the media adequately fulfill the role of informing voters in a democracy, including the topics being discussed, whether coverage is fact- or opinion- based, foreign or domestic, negative or positive, and/or emotionally charged. In an unprecedented effort, it uses cutting edge deep learning techniques to evaluate these aspects of quality at a minute-by-minute level, providing in aggregate a fine-grained summary of what the networks cover and how they cover it, plus how this changes over time. It finds that there is over twice as much negative coverage as positive coverage, and scary, shocking or outrageous coverage as pleasant coverage. The majority of cable networks' coverage is opinion, while the broadcast networks present more fact-based content. Importantly for democratic elections, only 20 percent of electoral coverage, accounting for 2.2 percent of total coverage, is on candidates' backgrounds, platforms, and speeches. 
 
The third paper explores the relationship between company ownership concentration and executive compensation packages. It investigates the idea that, other things being equal, we should see companies with more concentrated ownership structures choose more closely optimal compensation contracts for their executives. It finds that compensation depends upon the identity of the owners as well as their concentration. Institutional ownership, by professional investment managers, is associated with much higher executive incentive pay, whereas outside ownership by others is associated with much lower pay. It points to a new hypothesis in executive compensation, whereby the principal-agent problem between investors and investment managers may be allowing higher pay packages among CEOs as an external effect.