Using Laboratory Experiments to Evaluate Accounting Policy Issues

Abstract

INTRODUCTION Accounting scholars have long been challenged to demonstrate the relevance and timeliness of their research to the issues faced by accounting policymakers. Among others, Schipper (1994), Swieringa (1998), and Barth et al. (2001) provide examples of the progress made by empirical-archival studies in addressing this challenge. Notwithstanding this progress, the ability of empirical-archival accounting research to offer policy-directed insights is inherently limited, for two reasons. First, real-world data can only inform us of reactions to policies that already exist, whereas the policymaker's perspective demands ex ante insights of likely reactions to policies that could exist. Second, the variation among firms in accounting practices coincides with a multitude of other factors that vary at the same time, making it difficult to isolate the incremental effect of accounting alternatives. Given these limitations, several observers, such as Maines (1994), Beresford and Johnson (1995), and Swieringa (1996), sugges t a role for true experiments--designed studies in which participants make decisions under experimenter-controlled conditions--in addressing accounting policy questions. Only in a true experiment can we vary an existing or proposed policy alternative of interest and measure its incremental influence on decision makers, holding all other influences constant. This commentary explores this idea in greater depth, especially with respect to the different strengths and weaknesses of individual judgment and decision-making studies, often termed behavioral research, and multiperson interactive studies with competitive incentives, often termed experimental economics, for producing useful policy insights. (1) Our premise is that, viewed as a whole, these two styles of experimentation help to mitigate the concern that experiments are of limited value to policy debates because they lack realism (Abdel-khalik 1994). Behavioral studies of individual judgment and decision making often lack the realism of explicit performance incentives and the multiperiod, multiperson nature of economic competition. For different reasons, experimental economic studies also lack realism, because they often involve student subjects performing stylized tasks that are abstracted from the real-world environment. When viewed as a whole, however, these two approaches complement each other, addressin g the other's weaknesses and offering unique strengths. We provide some historical perspective below, followed by examples from each approach that illustrate our premise. For the sake of brevity, we focus our examples on financial-accounting policy questions, although we could offer similar observations for policy questions involving auditing, management accounting, and taxation. We then comment on the strategic implications of accounting policy questions, distinguishing between experiments that manipulate policy alternatives and those that treat accounting choices as strategic decisions by opportunistic managers. Our concluding comments return to our overall premise and suggest ways to integrate the benefits of different experimental styles. SOME HISTORY Although marked by recent interest, the use of experiments to address accounting policy questions dates back several decades. Swieringa and Weick (1982) review several experimental studies of accounting policy questions conducted in the 1960s and early 1970s. The popularity of such experiments ebbed significantly by the mid-1970s, concurrent with the critical view of this literature that emerged by that time (Gonedes 1972; Gonedes and Dopuch 1974). Criticism reflected a two-pronged attack, noting both the lack of theory in the experimental literature of the day and the inconsistency of this literature with the prevailing belief in informational market efficiency. The following passage from Gonedes and Dopuch (1974, 106) is telling: Since laboratory studies concentrate on individual behavior rather than competitive market phenomena, their relevance to the issues at hand seems nonexistent. …