Asim Ijaz Khwaja
Sumitomo-FASID Professor of International Finance and Development
Making Moves Matter: Experimental Evidence on Incentivizing Bureaucrats through Performance-Based Transfers
Adnan Q. Khan, Asim Ijaz Khwaja, and Benjamin A. Olken. "Making Moves Matter: Experimental Evidence on Incentivizing Bureaucrats through Performance-Based Transfers." HKS Faculty Research Working Paper Series RWP17-006, November 2016.
Postings are often used by bureaucracies, especially in emerging economies, in an attempt to reward or punish their staff. Yet we know little about whether, and how, this type of mechanism can help incentivize performance. Using postings to induce performance is challenging, as heterogeneity in preferences over which postings are desirable non-trivially impacts the effectiveness of such schemes. We propose and examine the properties of a mechanism, which we term a performance-ranked serial dictatorship, in which individuals sequentially choose their
desired location, with their rank in the sequence based on their performance. We then evaluate the effectiveness of this mechanism using a two-year field experiment with over 500 property tax inspectors in Punjab, Pakistan. We first show that the mechanism is effective: being randomized
into the performance-ranked serial dictatorship leads inspectors to increase the growth rate of tax revenue by between 44 and 80 percent. We then use our model, combined with preferences collected at baseline from all tax inspectors, to characterize which inspectors face the highest marginal incentives under the scheme. We find empirically that these inspectors do in fact increase performance more under this mechanism. We estimate the cost from disruption caused by transfers to be small, but show that applying the scheme too frequently can reduce performance. On net the results suggest that bureaucracies have tremendous potential to improve performance by periodically using postings as an incentive, particularly when preferences over locations have a substantial common component.