Leader Selection and Performance in Public Sector Organizations: Experimental Evidence from Ghana

In collaboration with Ghana Health Services (GHS)—Ministry of Health (MoH) and Kybele NGO, this project investigates how to select leaders among midwives to lead the implementation of a new protocol called obstetric triage package (OTIP). OTIP consists of a one-week training on clinical knowledge to accurately assess and prioritise the care of pregnant women with the goal of quickly identifying the severity of a patient’s condition, determine the appropriate level of care, and ensure that those with the most critical needs receive immediate attention. OTIP aims to tackle the inefficient practice of seeing obstetric patients on a 'first-come, first-served' basis, which puts patients at risk.

A key innovation of OTIP is the designation of up to ten midwives per hospital as 'Champions'—those selected to attend training, train their peers, as well as monitor and motivate the adoption of the protocol among colleagues.

Although OTIP was designed with a collaborative leadership approach in mind, the current method of selecting Champions remains a top-down process led by managers, reflecting a hierarchical model and potentially missing the advantages of tapping into peer networks. To address this, we introduced a bottom-up selection process, which fosters shared responsibility, teamwork, and collective decision-making, as commonly used in community-based service delivery.

The selection criteria are the same for the two treatment arms, asking to identify midwives with demonstrated ability to be a good leader, work well in teams, flexibility and ability to handle stressful situations, good communication skills, understanding of compassionate care, and ability to take initiative in difficult situations. In the `bottom-up’ treatment arm, we ask midwives in each hospital to nominate leaders through an online platform. The top nominees are then selected as OTIP champions in each bottom-up hospital.

2024-10-24

2025-01-31

1. Service quality
2. Knowledge
3. Collaboration

1. Service quality: including time until assessment, assessment checks and satisfaction, as reported by patients.
2. Knowledge: standardized test, developed by Kybele’s nursing researchers and GHS healthcare staff, measuring midwives' clinical knowledge.
3. Collaboration: behavioural games that midwives play with their teams, incentivized to achieve the highest output.

4. Leadership effectiveness
5. Social preferences
6. Attitudes

4. Leadership effectiveness: adapted methodology of Weidmann et al. (2024) to be used in the field. Teams of three midwives are formed, with the 'Champion' acting as the leader and two non-champions as the 'workers.' Each team is tasked with completing three jigsaw puzzles, and the team’s score is determined by the least-completed puzzle. The method involves assessing individual skills and team performance by randomly reassigning leaders across different teams.
5. Social preferences: modified dictator game following Rafkin and Soltas (2024), where midwives have the option to donate or avoid donating money to a champion.
6. Attitudes: midwives' self-reported perceptions of autonomy, empowerment and motivation on the job.

This study is embedded within our ongoing impact evaluation of the OTIP Champions Programme. The impact evaluation randomizes the roll-out of the programme across 25 hospitals located in the Greater Accra, Central and Western regions, corresponding to the final phase of the national scale-up among all high-volume hospitals (above 1,200 annual deliveries in 2022). Hospitals are randomly assigned to either the early-treatment arm, where OTIP is implemented in September 2024, or the later-treatment arm, where OTIP is scheduled for introduction in January 2025.

We cross-randomize across both early- and later-treated hospitals whether leader selection remains the status quo, e.g. appointed by hospital management (top-down in 13 hospitals), or whether midwives select the ‘Champions’ among their peers (bottom-up in 12 hospitals).

Not available

The statistical software Stata, and specifically the random number generator setting a seed, were used to generate the randomization.

Hospitals

Yes

25 hospitals

- Hospitals: 25 hospitals, same as cluster size. - Midwives: panel data for roughly 30 midwives per hospital (randomly selected midwives plus all champions), totalling 750 midwives per hospital. - Patients: cross-section of 50 mother-newborn pairs per hospital, on average, in each round, totalling 1,250 mother-newborn pairs per round and 3,750 across 3 data collection rounds.

1. Top-down: 13 hospitals, roughly 390 midwives and 650 mother-newborn pairs per round.
2. Bottom-up: 12 hospitals, roughly 360 midwives and 600 mother-newborn pairs per round.

- Patient-level outcomes: MDE of 0.20-0.64 standard deviations - Midwife-level outcomes: MED of 0.24-0.41 standard deviations