A study of signalbased choice in absence of information about the other's preference
Version
1
Resource Type
Dataset
Creator
 Sarkar Bose, Arundhati (XLRI  Xavier School of Management)
 Sarkar, Sumit (XLRI  Xavier School of Management)
Publication Date
20181231
Description

Abstract
Subjects: The experiment was conducted using freshmen MBA students. At the time of the experiment the MBA program just commenced, and hence MBA education cannot have any effect on their decisionmaking. 73.66% of the subjects studied engineering or technology, and the rest studied business, economics or science at undergraduate level. The average age of the subjects was 23.73 years, while the minimum and maximum age was 20 years and 28 years respectively. 39.25% of the subjects were female.
Design: A betweensubjects design was used to observe the effect of signals on the subjects’ decisions. The subjects were given a scenario wherein they had to make a choice for a hypothetical individual. According to the scenario, the hypothetical individual (henceforth, ‘other’) was randomly drawn from a population and has a preference over two options – “X” and “Y”. S/he made a choice for herself/himself, and has a belief about whether the subject will make the same choice as her/him. The subjects were required to make a choice decision for the ‘other’ without knowing her/his choice and belief. The subjects were told that “X” and “Y” represent two contrasting preferential attributes, but they were not told what those attributes are. This was done to ensure that their decisions do not get affected by their own preferences. According to the scenario, the population from which the ‘other’ was drawn consists of Xtype and Ytype individuals. Type depends on their preference over “X” and “Y”. The subjects were provided with a pair of frequency distributions of belief among Xtype and Ytype individuals in the population. This paired frequency distribution was the signal for the subjects. The scenario mentioned that the ‘other’ made a prediction about the chances that the subject will make the same choice as her/him.
62 different pairs of frequency distributions of predictions were created and were used in all three treatments. This was done to ensure that the treatments were identical except for their payoff structures. In each treatment 28 subjects were provided with strong signals having highly skewed distributions. Among these 28 subjects in a treatment, 14 got highly leftskewed distribution for Xgroup and rightskewed distribution for Ygroup and the remaining 14 subjects got the opposite. In each treatment 34 subjects were provided with weak signals having less skewed or nearly symmetric distributions. However, even among these 34 subjects in a treatment, 17 got slightly leftskewed distribution for Xgroup and rightskewed distribution for Ygroup and the remaining 17 subjects got the opposite. The ratio of meanP for Xgroup to that for Ygroup indicates the nature of signal as well as provides a measure of the signal strength. For convenience let us denote the signal as Xsignal if the ratio is greater than 1 and as Ysignal if the ratio is less than 1. In each treatment 31 subjects were given Xsignals and the remaining 31 were given Ysignals. Out of the 31 Xsignals, 14 had the ratio of meanP for Xgroup to that for Ygroup in the interval [1.28, 1.59]. These 14 signals were classified as strong Xsignals. The remaining 17 had the ratio in the interval [1.02, 1.24] and were classified as weak Xsignals. Likewise, among the 31 Ysignals there were 14 strong Ysignals having the ratio of meanP for Ygroup to that for Xgroup in the interval [1.28, 1.59] and 17 weak Ysignals having that ratio in the interval [1.02, 1.24].
Payoff determination: The “choice” and the “prediction” of the ‘other’ corresponding to each subject was required for determination of the subjects’ payoffs. For each subject, the “choice” of the ‘other’ and her/his “prediction” was determined by means of handrun random experiments maintaining the same probabilities as given in the scenario.
Availability
Download
Update Metadata: 20190430  Issue Number: 1  Registration Date: 20190430