Source: Dissertation Abstracts International, Volume: 74-12(E), Section: A.

Adviser: Andrew Metrick.

A small but growing literature in financial economics has attempted to understand the role that industries collectively play in shaping both governance decisions and equity prices. However, little work has been done on understanding how the process of defining industries can affect empirical tests of governance or markets.

My dissertation attempts to fill in this important gap in our knowledge. The crux of the "industry definition" question is simple: given some set of firms, how can we logically divide this set into subsets based on one or more key economic characteristics? There is no single answer to this question. However, my dissertation identifies many of the key benefits and tradeoffs associated with different industry classification approaches. In Chapter 1, I explore the tradeoff between industry size and statistical power in the context of the debate over whether corporate governance decisions impacted stock returns during the 1990s. In Chapter 2, I introduce a new set of firm-specific industry definitions and show that these new industry definitions relax many of the constraints associated with traditional industry classification standards. Finally, in Chapter 3 I apply my new set of industry classifications to a series of important open questions in the executive compensation literature: namely, how does CEO compensation depend on industry-adjusted firm performance, and how does compensation depend upon the pay of other CEOs within a firm's industry peer group?

My dissertation contributes to our knowledge of the relationship between industry peer groups, corporate governance, and equity markets in at least five distinct ways. First, it formally documents the tradeoffs between narrowly-defined and broadly-defined industry classifications in the context of empirical asset pricing tests. Second, it provides evidence that the previously-documented link between corporate governance and equity prices during the 1990s is not a function of unexpected industry performance. Third, it formally documents the transitive and disjoint nature of traditional industry classifications such as SIC codes. Fourth, it introduces a new set of industry definitions that allow industries to be defined at the firm level based on sound, consistent economic principles, thereby breaking the transitive and disjoint structure of traditional industry classifications. Finally, my dissertation provides a host of evidence suggesting that industry-adjusted stock returns play an important role in CEO compensation decisions, in contrast to much of the previous literature on executive compensation. All in all, my dissertation suggests that industry-related effects are extremely important---both in the context of corporate governance decisions and in the context of stock returns---and as a result, industry peer groups should play an important role in future research aimed at better understanding the complex relationships between corporate decision-making and financial markets.

Chapter 1 documents that an important tradeoff exists between narrow and broad industry definitions: in particular, narrow industry definitions should minimize statistical bias at the expense of reducing statistical power, while broad industry definitions should have the opposite configuration. When I examine this tradeoff within a specific context (the relationship between governance decisions and stock returns during the 1990s), I find that industries in the middle of the granularity spectrum appear to minimize the total amount of type I and type II error within my specific empirical application. Despite the narrow focus of Chapter 1, this result is broadly applicable and has a number of potential implications for future research.

I begin Chapter 2 by noting that all of the industry classification standards favored by the academic community are defined in a transitive and pairwise disjoint fashion. In contrast, analysts and investment bankers typically customize industry peer groups at the firm level. In Chapter 2, I use this practical insight to create a set of firm-specific industries, or FSIs. In particular, I exploit a federal law requiring publicly-traded firms to provide an unbiased list of their product market competitors in their 10-K filings. I hand-collect these competitor names from more than 36,000 10-Ks and show that these self-reported competitor lists perform favorably relative to traditional industry classifications in a number of empirical contexts.

In Chapter 3, I apply FSIs to an important open question in the executive compensation literature. In particular, while theory suggests that CEO compensation should depend upon industry-adjusted firm performance, thirty years of empirical research has largely failed to produce evidence consistent with this hypothesis. However, when traditional industry classifications are replaced with FSIs, I find strong support for this hypothesis. Together, these chapters suggest that industry definitions play a crucial role in empirical tests of governance and financial markets.