Working Paper: NBER ID: w17912
Authors: Abhijit Banerjee; Raghabendra Chattopadhyay; Esther Duflo; Daniel Keniston; Nina Singh
Abstract: The role of good management practices in organizations has recently been emphasized. Do the same principles also apply in government organizations, even the most bureaucratic and hierarchical of them? And can skilled, motivated managers identify how to improve these practices, or is there a role for outsiders to help them in this task? Two unique large-scale randomized trials conducted in collaboration with the state police of Rajasthan, India sought to increase police efficiency and improve interactions with the public. In a sample of 162 police stations serving almost 8 million people, the first experiment tested four interventions recommended by police reform panels: limitations of arbitrary transfers, rotation of duty assignments and days off, increased community involvement, and on-duty training. Field experience motivated a novel fifth intervention: “decoy” visits by field officers posing as citizens attempting to register cases, which gave constables incentives to behave more professionally. Only two of these, training and decoy visits, had robust impacts. The other three, which would have reduced middle managers’ autonomy, were poorly implemented and ineffective. Building upon these findings, we designed a second experiment that provided explicit incentives to police officers to carry out sobriety traffic checkpoints and did not rely on middle managers. Linking good performance with the promise of a transfer from the reserve barracks to a desirable police station posting, these incentives worked within existing organizational constraints and had very large effects on performance.
Keywords: police performance; management practices; incentives; randomized trials; public sector reform
JEL Codes: H11; H76; K42; O22
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| decoy visits (Y50) | increased probability of case registration (C24) |
| decoy visits (Y50) | improved police behavior (K40) |
| sobriety checkpoints (R48) | increased number of drunk drivers brought to court (K40) |
| incentives (M52) | enhanced productivity in police performance (J24) |
| training interventions (M53) | increased victim satisfaction with police interactions (K42) |
| reduced middle managerial autonomy (M54) | less effective interventions (I24) |