Project 1: Cheating We use an online real-effort experiment to investigate how bonus-based pay and worker productivity interact with workplace cheating. Firms often use bonus-based compensation plans, such as group bonuses and firm-wide profit sharing, that induce considerable uncertainty in how much workers are paid. Exposing workers to a compensation scheme based on random bonuses makes them cheat more but has no effect on their productivity. We also find that more productive workers behave more dishonestly. We explain how these results suggest that workers’ cheating behavior responds to the perceived fairness of their employer’s compensation scheme. Project 2: Observation We propose a behavioural model of effort provision under a fixed wage payment scheme that accounts for competitive social preferences such as non-competitive self-esteem, competitive self-esteem and status-seeking. We designed and ran a sequence of controlled laboratory experiments that allow for the separation of these effects using four distinct combinations of relative performance feedback (explicit rank in a peer group) of three forms - none, computerised and personal, and either a private or a public information condition. All the data is collected using a real-effort task. Preliminary analysis reveals an important role for relative performance feedback in driving performance, despite the lack of tangible benefits. Moreover, we find evidence of reference-dependence over rank information
Laboratory Experiment with Human Subjects. Project 1: Cheating We ran an online experiment during a two-week period in May 2011. To recruit subjects we sent email announcements to students at the University of Southampton. During the two-week period, 641 subjects completed the experiment at a time and location of their choosing. The experiment lasted 20 to 25 minutes and had to be completed in one go. The subjects were informed that they would be paid £3 for completing the experiment, on top of any money that they earned during the experiment (all payments were in pounds Sterling). The average total payment to the subjects who completed the experiment was £19.32. At the end of the experiment, we asked the subjects to report their gender. We paid the subjects privately in cash on two pre-advertised dates. The experiment was implemented using the Django open source web application framework, which is written in Python. We used the computerized real-effort slider task', which was first developed by Gill and Prowse (American Economic Review, February 2012). The slider task consists of a screen displaying a large number of sliders. As the task proceeds, the screen displays the subject's current points score and the amount of time remaining. In the first of the three stages, subjects were given two minutes to practice the slider task (with 102 sliders) and become familiar with it. This stage was not compensated. In the second stage, subjects were given ten minutes to achieve a points score of 40 in the slider task (again with 102 sliders). We chose a long time limit to give the subjects more than ample time to achieve the required score, but subjects were aware that as soon as they reached a points score of 40 they would move on to the third and final stage and that if they did not achieve the score they would exit the experiment. Our experimental conditions varied how the subjects were compensated for completing the stage. In theControl' condition (C), the subjects were paid a fixed pre-announced amount for completing the stage. The control subjects were split into two subgroups: in Control 1' (C1) subjects were paid £2 for completing the stage; inControl 2' (C2) subjects were paid £8 for completing the stage. In the Treatment' condition (T), subjects were told that they would be paid £2 for completing the stage, and that there was a 50% chance they would earn a bonus of an additional £6 for completing stage 2, in which case they would be paid a total of £8 for completing the stage. The treated subjects were informed that a random number generator would decide randomly whether or not they had earned the bonus, and that they would find out the outcome after completing the stage.Treatment 1' (T1) refers to the subgroup of treated subjects who did not receive the bonus, and who were thus paid £2 for completing the stage. `Treatment 2' (T2) refers to the subgroup of treated subjects who did receive the bonus, and who were thus paid a total of £8 for completing the stage. Subjects were randomly allocated to C1, C2, T1 or T2 with equal probability. The third stage is designed to measure cheating and productivity. Subjects were given five minutes to work on the slider task (now with 201 sliders), and were paid £0.05 for every point they scored. In addition, we asked the subjects to report the last digit of their best friend's phone number and told them that they would be paid an amount in pounds equal to the number that they reported (we also gave them the option of reporting ``don't know", which paid nothing). The subjects were informed that they could report the number at any time during the five minutes, that they could change their report if they made a mistake, and that the report at the end of the five minutes would determine payoffs. Project 2: Observation A total of 18 sessions were conducted in the experimental laboratory of the Nuffield Centre for Experimental Social Sciences (CESS), University of Oxford, in February and March 2011 using Java software. In these, 306 students of any degree other than Psychology performed identical computerised real-effort task in groups of 17 per session during seven (one practice and six paying) 6-minute rounds. Participants were randomly allocated to the computers and worked in completely isolated cubicles. Similarly, the experimenter was located in a separate cubicle, away from the sights of the subjects, and the laboratory assistant would not be observing the work of the participants. The specific inter-subject and subject-experimenter anonymity condition then varied across the treatment groups. The experimental design aimed to separate the causal relationship between relative performance feedback of two forms, computerised and personal rank information, under a private or public information condition and the dependent variables - subjects’ performance and change in performance. It is hypothesised that the isolated tendencies are driven by subjects’ non-competitive and competitive self-esteem as well as their status-seeking with the experimenter or their peer group. Each participant received their own copy of the instructions (see below) in an envelope when entering the laboratory room. The instructions used abstract framing and neutral language that would not give away the purpose of the study and hence the expected behaviour by the subjects. They were read out by the experimenter. Participants could communicate their questions to the experimenter and receive answers solely in a written form through the laboratory assistant. Before the start of the scored task, the instructions reminded the participants that their payment would not depend on their performance on the task. The experiment lasted for about 90 minutes and all participants were paid a fixed wage of £20 (including the £5 show-up fee) after completing a demographics questionnaire that followed the experimental task. Participants of the experiment performed either on the Control Group (C) with no relative performance feedback or on a Treatment Group with computerised private (T1), personal private (T2) or personal public (T3) relative performance feedback after each paying round. The relative performance feedback would take the form of the relative rank of the individual within the group of 17 subjects participating in the session. Additionally, all participants would learn about their individual total point score (the sum of their points score from the verbal and the numerical part in a particular round) on the task in each of the 7 rounds. The treatment groups would hence differ by the form of the relative performance feedback (none, computerised or personal) and the information condition (private or public) used.
