Severity vs. Leniency Bias in Performance Appraisal: Experimental evidence

Performance appraisal can be biased in two main ways: lenient supervisors assign pre- dominantly high evaluations (thus rewarding also undeserving agents who have exerted no effort) while severe supervisors assign predominantly low evaluations (thus failing to reward deserving agents who have exerted effort). The principal-agent model with moral hazard predicts that both biases will be equally detrimental to effort provision. We test this prediction with a laboratory experiment and we show that failing to reward deserving agents is significantly more detrimental than rewarding undeserving agents. This finding is compatible with empirical evidence on real world supervisors being preponderantly biased towards lenient appraisals. We discuss our result in the light of alternative economic the- ories of behavior. Our result brings interesting implications for strategic human resource management and personnel economics and contributes to the debate about incentives and organizational performance.