Essays on incentives in bureaucracy: experimental evidence from Pakistan

Public sector performance is a key determinant for a country’s economic growth and development. Yet, poor service delivery remains widespread across the developing world which is often attributed to weak monitoring and incentive systems. This thesis, written in a three-paper format, explores questions around how to improve the performance of frontline workers through the use of financial and non-financial incentives. I design two field experiments in collaboration with the Teacher Training Academy in Punjab Pakistan to explore the impact of employer recognition (a type of non-financial incentive) on the performance of public school teachers in in-service trainings. The first experiment serves as a pilot to the second experiment which is larger in scope and builds upon the findings of the pilot. These experiments are presented as Chapter 1 and Chapter 2 respectively. A third experiment is designed and implemented in collaboration with the Punjab Agriculture Department to evaluate the impact of a province-wide pay-for-performance programme on extension outreach and quality. This study is presented as Chapter 3. In Chapters 1 and 2, I investigate whether non-financial incentives such as employer recognition can improve the performance of public school teachers in in-service trainings as measured through training test scores, and if so, what are the channels through which such tools operate? In chapter 1 (pilot), I randomize 650 primary and secondary school teachers into three employer recognition schemes and a control group to understand the strength of different channels of employer recognition. The employer recognition incentive is a standard tournament-based incentive in which teachers are awarded a prestigious recognition certificate if they qualify. In T1 (Private arm), the certificate is given to the teacher privately; in T2 (Peer arm), the certificate is given in front of peers and colleagues; and in T3 (Career arm), the certificate is given privately but teachers are also told that their name would be added to an ‘excellent teacher list’ which will be shared with departmental leadership that can help them quality for future career opportunities in their districts. Results show that the source of individual motivation of bureaucrats for entering the civil service mediates the effectiveness of employer recognition. We find that teachers who report intrinsic reasons for entering the service such as ‘interested in the profession of teaching’ exhibit a positive treatment response and those who report extrinsic reasons such as ‘salary’ show a negative treatment effect, with the two effects being significantly different from each other. Further analysis shows that the positive treatment effect for intrinsic individuals comes from the “peer" and “career" channels of employer recognition, with no effects on the “private" channel. Treatment effects on endline self-efficacy beliefs show positive effects for intrinsic but negative effects for extrinsic individuals, indicating that self-efficacy beliefs may be an important causal mechanism for these effects. These results directly support the design of the study in Chapter 2. In Chapter 2, I randomize 3,394 head teachers attending a mandatory government training into four different designs of recognition incentives that are tied to training performance and a control group. Treatment 1 (Peer arm) leverages the peer/collegial approval channel of recognition – trainees are told that those with the top score in the training post-test and the most improved score (over the pre-test) will be provided certificates in a district level ceremony which will be attended by their peers and colleagues in their district; Treatment 2 (Career arm) leverages the career-benefits channel of recognition – it is the same as Treatment 1 except trainees are told that those who qualify will receive the certificates privately but at the same time their name will be added to an ‘excellent teacher list’ (which will be shared with the School Education Department leadership) that can help them qualify for future opportunities in their districts. This leverages informal career incentives in the system such as postings to preferred schools, transfers to other lateral postings, or getting selected for promotion earlier once eligible; Treatments 3 and 4 (Public PLUS and Career PLUS) cross the first two treatments with a self-efficacy enhancing frame that aims to bolster teacher perceptions in their ability to do well in the training and their jobs more broadly. I find that employer recognition can improve teacher training performance if it is linked to tangible career benefits in the future but the positive effects can also backfire depending on how these incentives are framed. When a self-efficacy frame is added to either of the first two treatments, it improves teacher self-efficacy but makes teachers overconfident, reducing performance in the training. These findings have implications for how to design more powerful non-financial incentives that are effective for eliciting higher teacher effort, specifically in trainings and more broadly in schools and classrooms. At the same time, they point towards the sensitivity of such incentives to framing effects and suggest caution in how these are framed. In Chapter 3, we partner with the Agriculture Extension Department to evaluate the impact of three alternative pay-for-performance schemes on extension outreach and quality. Extension service delivery is a typical setting where agents carry out tasks that include easily measurable and hard to measure dimensions, and hence the merits and demerits of objective versus subjective incentive systems are unclear. The performance schemes leverage the departmental digital monitoring system called AgriSmart and link incentives to objective metrics available on AgriSmart (Objective arm), supervisors’ own subjective evaluation (Subjective arm), or supervisors’ own subjective evaluation with an additional element of top-down monitoring to align supervisors’ incentives with the objectives of the principal (Subjective Plus arm). Early results show that while all treatments improve performance on AgriSmart metrics, Subjective Plus also shows positive and significant effects on farmer experience of extension outreach and quality, and farmer agricultural practices. In addition, we also find strong and positive effects on AD performance in their managerial tasks (such as assigning farmer training programmes to frontline extension agents) in the Subjective Plus arm, which form an important input into how frontline staff carry out extension outreach. Additional data collection will explore whether the results in Subjective Plus are driven by the effective use of discretion by supervisors in their evaluations, an improvement in AD performance and practices, or both. It will also explore mechanisms that explain why treatment effects observed on AgriSmart outreach indicators in the Objective and Subjective arm do not translate to farmer experience of extension outreach and quality. This thesis generates three key findings. First, it highlights how to bolster teacher engagement, effort, and subsequently the acquisition of knowledge and skills in in-service trainings. It indicates that non-financial incentives that are linked to potential career benefits in the future are a feasible and cost-effective way for improving the impact of in-service trainings, where some of these career benefits such as postings to preferred schools or transfers to other lateral positions may be less formal and easier to exercise in the system as compared to promotions. More broadly, the results in Chapter 1 and Chapter 2 point towards the existence of such informal career incentives in the system and how they could be leveraged to make the design of non-financial incentives sharper. Second, the impact of non-financial incentives is mediated through individual sources of motivation and non-cognitive traits such as beliefs about ability and individual self-efficacy. This makes such incentives highly contextual, where the impact can often be negative depending on the group-type being incentivized. This implies a need for caution in how such incentives are designed and framed. Third, settings where tasks are multi-dimensional and quality of performance is hard to measure, incentive systems which give discretion to supervisors can be effective as long as supervisors incentives can be aligned with the objectives of the principal.