The Effect of Certain Research Design Choices on the Assessment of the Market's Reaction to LIFO Changes: A Methodological Study
The Effect of Certain Research Design Choices on the Assessment of the Market's Reaction to LIFO Changes: A Methodological Study
复制标题
某些研究设计选择对评估市场对后进先出变化反应的影响：方法论研究
DOI：
10.2307/2490940
复制
发表时间：
1983
影响因子：
4.4
通讯作者：
D. Murray
中科院分区：
文献类型：
--
作者：
D. Murray
实证研究的研究设计的一个主要目标是消除尽可能多的竞争假设。为此目的经常使用的一种策略是使用对照组作为比较标准（例如Dyckman和Smith [1979]以及柯林斯和登特[1979]在评估FASB标准19的征求意见稿时使用的方法）。理想情况下，实验组和对照组之间观察到的任何差异都应该是由于实验条件造成的。在真正的实验中，这更有可能通过随机选择受试者，然后将他们随机分配到实验和控制条件来实现。然而，基于市场的研究是事后的;研究人员不行使对自变量的控制，也不对实验组或控制组的企业分配。公司的自我选择进入各自的群体引入了自我选择偏见可能混淆结果的可能性。除了自变量之外，企业可能在其他特征上存在差异，这些差异可能导致因变量的变化。一种用于减少自我选择偏差可能性的方法是将实验组和控制组的变量进行匹配，这些变量在先验基础上被认为会影响因变量。
A major objective of the research design of an empirical study is the elimination of as many rival hypotheses as possible. One strategy frequently used for this purpose is to employ control groups as standards of comparison (such as used by Dyckman and Smith [1979] and Collins and Dent [1979] in their assessments of the exposure draft of FASB Standard 19). Ideally, any observed difference between an experimental group and a control group should be due to the experimental condition. In true experiments, this is more likely to be achieved through a random selection of subjects, followed by their random assignment to the experimental and control conditions. However, market-based research is ex post; the researcher does not exercise control over the independent variable(s) nor over the assignment of firms to the experimental or control group. The self-selection of firms into their respective groups introduces the possibility that self-selection biases may confound the results. Firms may differ on characteristics other than the independent variable(s), and these differences may induce variations in the dependent variable. One approach used to reduce the likelihood of self-selection bias is to match the experimental and control group on variables that, on an a priori basis, are believed to influence the dependent variable.1 Harrison [1977], for