Insurer Ownership Structure and Executive Compensation as Complements
Marx, Leslie M., Mayers, David, Smith, Clifford W., Jr., Journal of Risk and Insurance
The authors apply results on complementarities to theories of insurance companies' choices of ownership structure and executive compensation. They identify minimal restrictions on the interaction between firm policies and exogenous characteristics for theories to have testable implications for reduced-form regression coefficients. Obtaining testable implications for structural-equation regression coefficients requires additional identifying restrictions. The authors' analysis highlights a basic tradeoff between theory and statistical methods.
An insurance company faces a broad array of strategic choices. It must choose its lines of business as well as its geographic reach, its organizational form, the composition of its board of directors, and the structure of its executive compensation plan, distribution system, risk management policy, and so on. While progress has been made in examining facets of these strategic policy choices, virtually all of the analyses fail to incorporate the facts that the various policies are endogenous and that many are jointly determined. This article explicitly recognizes this simultaneity and examines the requirements for a theory to have empirically testable implications for correlation and regression coefficients.
To provide unambiguous predictions for the signs of correlation and reduced-form regression coefficients, a theory of insurance companies' choices must imply monotone comparative statics. When the relation between underlying firm characteristics and choice variables is not monotonic, the theory produces ambiguous sign predictions in terms of these coefficients. In this case, although a reduced-form regression will estimate the average relation in the data, the theory is not falsifiable in that either a positive or a negative observed relation would be consistent with the theory. If the theory does not imply monotone comparative statics, then testable implications require structural-equation estimation methods--but these methods require identifying restrictions.
While the requirements for monotone comparative statics are readily satisfied in models with a single choice variable, insurance companies are complex organizations that make a number of choices. When firms choose several policy variables, additional care is required in developing theory and in undertaking empirical work to analyze these choices.
Research on complementarities among policy choices has led to a more complete understanding of the requirements for monotone comparative statics, (1) but this literature has only recently begun to focus on the implications for empirical work. While Arora and Gambardella (1990), Arora (1996), and Athey and Stern (1998) discuss how theories about complementarities can be tested empirically, (2) these ideas have not been applied in studies of the insurance industry, even though they have important implications for research on insurance companies' strategic choices.
To keep the analysis simple, the authors model the firm's problem as focusing on two choices: ownership structure and executive compensation. The authors employ existing results on complementarities, applying these results to questions of strategic policy choices within the insurance industry. The analysis provides better interpretations of existing evidence, identifies weaknesses in existing theory, and suggests additional tests.
The following sections of this article discuss how results on monotone comparative statics can be applied to issues in the insurance industry by examining the impact of a firm's lines of business on its choices of ownership structure and executive compensation as well as implications for empirical work. The article also proposes tests of the theory.
COMPLEMENTARITIES IN INSURANCE
Lines of insurance vary in the level of required managerial discretion. For example, managerial discretion should be less important within lines for which more extensive loss data are available, the variance of losses is lower (Lamm-Tennant and Starks, 1993; Doherty and Dionne, 1993), screening is less valuable (Hansmann, 1985; Smith and Stutzer, 1990), and the legal environment for claims administration is expected to be more stable. …