Introduction
Agency in the workplace—the influence over organisational decisions—is increasingly viewed as critical for organisational performance. Firms have begun to increase the level of agency given to workers through mechanisms such as peer reviews, nominations for employee recognition, or voting rights over specific decisions. This paper presents results from a firm-level randomised controlled trial (RCT) in India that studied the impacts of increased worker agency relative to increased managerial discretion in employee recognition programmes. The researchers partnered with Obeetee Ltd, India’s largest carpet manufacturer, to evaluate how different reward allocation mechanisms affect worker attendance, productivity, and firm culture.

Background
The study was conducted in eastern Uttar Pradesh, often referred to as the “carpet capital of India.” Obeetee Ltd outsources the production of hand-knotted carpets to hundreds of smaller supplier firms. These firms generally employ 10 to 20 workers, and worker attendance is highly variable. In control-group firms, an average of 48.3% of workers were present during daily spot checks. Productivity is individualised and measured as knots woven per day, adjusted for carpet complexity. Absenteeism is a problematic issue for firms attempting to meet strict production deadlines.

Experimental design
The researchers introduced an employee recognition programme in 125 firms, providing monetary rewards every two weeks for 12 weeks. The reward amount was fixed at 10% of the average baseline monthly earnings in the firm, ranging between ₹300 and ₹1200. To be eligible, workers needed to be present for at least one-third of the working days in the two-week cycle. Firms were randomised into three arms: worker vote (agency treatment), managerial discretion, and a public lottery (control). This design allows for a direct comparison between worker and manager control while controlling for income effects via the lottery arm.

Effects on main outcomes
The study find that different allocation mechanisms impact different margins of worker behaviour. The worker agency treatment increased attendance, raising the fraction of workers present on a given day by 4.1 percentage points (p.p.) relative to the control arm and 5.5 p.p. relative to the manager arm—an 11% increase over the control mean. Conversely, managerial discretion increased productivity by 0.30 standard deviations (SDs), or 6% relative to the control. However, the manager arm significantly reduced work-related interactions between workers by 9.3 p.p., representing a 39% reduction. This suggests that while managerial discretion can improve output, it may come at the cost of lower knowledge spillovers and reduced workplace cohesion.

How were rewards allocated?
Winners in the manager arm were positively selected on both attendance and productivity. In contrast, winners in the worker vote arm were positively selected on attendance and social interactions. Managers tended to rotate winners more frequently; only 7% of winners in the manager arm were repeat winners, compared to 23% in the worker arm, likely to avoid perceptions of favouritism. Reward sharing was common across all arms, with 72% of control workers reporting that winners shared bonuses. However, the likelihood of sharing was 14.2 p.p. higher in the voting arm. Specifically, workers in the voting arm were 37.7 p.p. more likely to report receiving a share because they voted for the winner, indicating that workplace democracy can incentivise non-productive behaviours like reward-trading or collusion.

Mechanisms: Incentives to win versus impacts of winning
Using a shortlist design, the researchers found that the impacts on attendance and productivity were driven by workers responding to incentives to win rather than the act of winning itself. Winning a reward generally had no impact on subsequent performance, with the notable exception that winning a reward allocated by a manager reduced work-related interactions. This reduction could be attributed to peer resentment or a reduced perceived need for winners to collaborate with others.

Other outcomes
The manager arm significantly increased perceived recognition and the perception that effort or skill determined rewards (0.25 SD). However, it also increased perceived favouritism, with perceptions that personal relationships with the manager were rewarded rising by 0.27 SD. Daily earnings in the manager arm increased by ₹28.46 relative to the control group, a 12% increase consistent with higher productivity gains. In the worker vote arm, earnings increased by ₹19.36, which suggests that attendance also influences compensation in this setting.

Choice of allocation mechanism
When asked to choose a mechanism for future rounds, 80% of managers and 59% of workers in the control group initially preferred worker voting. However, experience with the treatments shifted preferences; experience with managerial discretion reduced managers’ demand for worker democracy by 40.5 p.p. as they learned about productivity returns. Workers generally showed increased demand for whichever mechanism they had experienced during the programme.

Conclusion
The findings underscore that what is valued in the workplace significantly impacts worker behaviour and firm culture. Managerial discretion improves productivity but may reduce workplace interactions and increase perceptions of favouritism. Worker democracy improves attendance but can foster collusive agreements where rewards are shared in exchange for votes. The effectiveness of workplace policies depends not only on their formal structures but also on how workers collectively respond to and potentially circumvent them.