The Direct and Indirect Effects of Employee Performance Improvement Plans

This study evaluates the impact of a structured PIP program during a large-scale downsizing initiative at a garment factory in Southeast Asia.

2025-07-21

2025-12-31

Attendance, retention, performance and violations, job satisfaction, and well-being

We will assess skills, effort, and performance primarily through manager evaluations and administrative data. Additionally, we will survey workers on self-reported assessments of job satisfaction. We are also interested in the types of workers who respond positively or negatively to the PIP. We will conduct heterogeneity analysis along the following dimensions: baseline high/low skills and effort (based on their own and managers’ assessments and administrative data), well-being, differences in assessments between them and their managers, belief about the returns to their effort at work, assessment of the job market, degrees of social connectedness, and demographics (age, gender, province, and tenure), whether they feel effort maps into work outcomes, financial stress, and exposure to a military conflict.

We will also study how the PIP affects workers’ individual components of job satisfaction.
Furthermore, the study will examine indirect effects of PIPs and separation by i) using the randomization of treatment intensity across work lines and 2) surveying peer networks to understand how one worker's performance improvement plan and separation might impact others in the social network.

Individual components of job satisfaction and their assessments about the PIP program. Changes in managers’ behavior. Factory-level performance between PIP factories and non-PIP factories.

We will also examine how the PIP influences individual components of job satisfaction gathered through surveys. In addition, we will compare factories that implement the PIP program with those that do not. To supplement these analyses, we plan to conduct factory-level difference-in-differences regressions on administrative outcomes such as absences, quits, violations, and production, comparing the factory undergoing PIP with other factories owned by the firm.

We partner with a large garment manufacturer in Southeast Asia. Eligible workers are randomized at two levels: (i) production lines are assigned to high or low intensity of PIP coverage, and (ii) within lines, eligible workers are randomized into PIPs. Baseline, midline, and endline surveys, along with administrative records, will be used to measure direct effects on treated workers and indirect effects on their peers.

Not available

Randomization will occur via a computer.

There are two randomizations. First, we randomize intensity at the line level. Lines are randomly assigned to either the high-intensity or low-intensity group, with varying levels of treatment exposure. Line randomization is also stratified by the share of workers on the line who are eligible for PIP. Then, randomization occurs at the individual worker level.

Yes

The 68 lines, where eligible workers are located, are randomized to be either high-intensity or low-intensity lines. Standard errors will be clustered at the line level to account for potential correlations within each line.

680-790 workers for outcomes gathered through the midline survey and administrative data. 680 workers for the primary outcomes analysis, based on the 1-month post-survey and administrative data. For the indirect effects analysis, we intend to survey up to 1,600 workers to capture peer effects across the factory.

Half of the workers in the sample are assigned to the PIP arm.

We calculate power (with power 0.8) based on the baseline performance score (composite of skill and effort) that we have available. The mean is 5.54 with a standard deviation of 2.32. With 680 workers, we expect to detect 8.8% change in the short run from the mean. As for unscheduled absences, with a mean of 1.83 and standard deviation of 3.22, we expect to detect a change of 38% of the mean.