The Use of Control Groups in Capital Market Research
以会计估计变更事件为例，扩展了Brown和Warner的分析，比较四种对照组在评估股票市场反应时的效果，发现更复杂的对照组可能带来额外收益。
Capital market researchers have used several different types of control groups in attempting to assess stock market reactions to accounting events. In recent years, the properties of various control groups have been examined to determine the sensitivity of experimental results to a particular control group (Gonedes [1975; 1978], Foster [1980], Brown and Warner [1980], Beaver [1981]). Except for the Brown and Warner study, these studies have been of an analytical nature. The purpose of this study was to extend the Brown and Warner analysis to four alternative control groups using a particular accounting event study as an example. The example involved an assessment of whether the stock market reacted to actual changes in accounting estimates; the four types of control groups examined were those which have been used in previous capital market studies (see section 3). One of Brown and Warner's main conclusions was that . . . beyond a simple, one-factor market model, there is no evidence that more complicated methodologies [such as risk-equivalent control portfolios] convey any benefit. Our results, which are based on a market setting typical of many information content studies, lead to a potentially different conclusion. Brown and Warner injected simulated market effects into randomized data, and by replicating their analyses 250 times, they were able to