ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES Vol. 69, No. 3, March, pp. 221–236, 1997 ARTICLE NO. OB972683 Sequential Decision Making with Relative Ranks: An Experimental Investigation of the “Secretary Problem” DARRYL A. SEALE University of Alabama, Huntsville AND AMNON RAPOPORT University of Arizona To illustrate this class of optimal stopping decision The “secretary problem” models a class of optimal problems and motivate our study, consider two exam- stopping decision tasks in which the binary decision ples from entirely different domains. The first example to either stop or continue the search only depends on (Stewart, 1981) concerns a prospecting company, with relative ranks. We present this problem in a computer- resources to develop a single mine, which commences controlled experiment designed to investigate sequen- exploration in a new region. Mineral deposits are scat- tial observation and selection behavior in the context tered throughout the region and are assumed to be of employer hiring decisions. Our principal objectives discovered in a random order. When a deposit is discov- are to test the descriptive power of the optimal decision policy; characterize and competitively test simple deci- ered, a claim option can be taken out. As is often the sion rules, or heuristics, that individuals might use case, the decision as to whether to work the claim has in optimal stopping decision tasks; and examine the to be made within some prescribed amount of time. If sensitivity of the various decision rules by computer this time is relatively short in comparison with the time simulation. Our simulation results show that the opti- between discoveries, then effectively a decision on one mal policy is insensitive to moderate deviations from option has to be made before the next discovery is made. the optimal cutoff value, and that a simple, non-optimal On the other hand, if the option is allowed to lapse decision rule that counts the number of successive non- there is a high probability that some other company candidates performs remarkably well. Model compari- sons show that simple cutoff decision rules account will take up the option. for the decisions of the majority of the subjects. The The second example—the one actually investigated observed tendency to stop the search too early is ac- in the present study—concerns the hiring of the best counted for by postulating an endogenous cost of time secretary from a fixed pool of potential employees spent in the search.  1997 Academic Press applying for a job vacancy (e.g., Ferguson, 1989). The employer is assumed to interview the applicants se- quentially; following each interview, she must decide Our objective in this paper is to characterize and whether to select the current applicant, thereby termi- study several simple decision rules, or heuristics, that nating the interview process, or reject the current appli- individuals may use to maximize the probability of se- cant and interview another applicant. We assume that lecting the best alternative from a finite set of alterna- rejected applicants cannot be recalled (this assumption tives that are inspected sequentially. After each new can and has been relaxed). This assumption applies in inspection, the decision maker (DM) must either choose situations in which interviews are spaced far apart and one of the available alternatives or exercise an option there is strong competition for secretaries, so that the to inspect another alternative; there is no recall of previ- probability that a rejected applicant is still available is ous alternatives. practically zero. These and related sequential decision problems have Address correspondence and reprint requests to Darryl A. Seale, been modeled by the so-called “secretary problem” in Department of Management and Marketing, College of Administra- probability theory. Although there is some obscurity as tive Science, University of Alabama in Huntsville, Huntsville, AL 35899-0376, E-mail:dseale@ email.uah.edu. to the origins of this problem, it is generally agreed 221 0749-5978/97 $25.00 Copyright  1997 by Academic Press All rights of reproduction in any form reserved.