About Open Access
Controlling Risk Experiments in Accounting Experiments
Lookup NU author(s)
Professor Ian Dobbs
Dobbs IM, Miller AD
Conference Proceedings (inc. Abstract)
British Accounting Association Annual Conference 2008
Year of Conference
1-3 April 2008
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
In recent years, laboratory experiments have become increasingly important in generating empirical data useful for testing hypotheses in a wide range of accounting areas, including management accounting, agency theory and auditing. Many of these hypotheses concern the behaviour of individuals, working alone or in teams, in conditions of uncertainty. The unobservable risk preferences of experimental subjects are often not of primary interest to the research and are regarded as nuisance variables affecting behaviour. A common procedure for removing the nuisance effect of risk preferences is to use binary lotteries for inducing in all subjects risk preferences pre-specified by the experimenter. The validity of the binary lottery procedure has already been formally established, but only for the special case of a single subject performing a single task. This special experimental setting is not, however, of great significance to experimenters, because subjects are usually required to perform a series of tasks, both to economise on experimental overheads and also because subject behaviour does not usually converge without some task repetition. Furthermore, experiments involving teams and agencies involve interaction between subjects. This paper formally analyses the validity of the risk-preference-inducing binary lottery procedure when subjects perform a series of tasks. Appropriate necessary and sufficient conditions relating to experimental design are developed for this case. The conditions are particularly relevant to experiments in which information is supplied to subjects for the purpose of revision of their beliefs about the future, one of the main roles claimed for accounting information. The conditions are also suggestive of issues that need to be considered in designing experiments involving subject interaction in teams and agencies. Finally, the paper discusses practical operational issues arising from the formal analysis. The research contributes generally to the literature on experimental design, and specifically to experimental accounting research into the role of information for facilitating and influencing decisions.
British Accounting Association
Newcastle University Library, NE2 4HQ, United Kingdom. Tel: 0044 (191) 222 7657
©2011 Newcastle University Library