An Examination of Event Dependency and Structural Change in Security Pricing Models

This paper considers two aspects of the tendency for systematic risk to change during the period surrounding a firm-specific event. First, a statistic allowing for heteroskedasticity is presented as a means of more precisely testing for the incidence of structural change in the market model. Secondly, the bias resulting from the imposition of a single, arbitrary event period on every firm in a market efficiency study is formally demonstrated. Using a sample based upon stock splits, the switching regression technique of Quandt is then adapted to show that event intervals are more appropriately considered on a case-by-case basis. A comparison of alternative residual measures illustrates these procedures.