How Does Part-time Work Affect Labor Productivity? Evidence from a Field Experiment on Female Data Entry Clerks in Ethiopia

1. First randomization

AFF distributed job flyers to women who aged between 18 and 35 with high school diploma during a census of approximately 22,000 households in the catchment area. Part-time job opportunities were given to randomly-selected 41 villages and full-time opportunity to the remaining 40 villages. The total number of applicants was 529. From December 19 to 31, 2016, job applicants will participate in a baseline job survey.
Among those who complete the baseline job survey, about 220 qualified applicants will join two-week long training sessions in January to March 2017, which will entail basic computer training (such as Excel), specifics of data entry, and tests. At the end of the training sessions, AFF aims to hire about 100 full time equivalent data-entry clerks (70 full-time and 60 part-time).

2. Second randomization
Among workers who successfully perform the work during the first 30 days, AFF will give a randomly selected subset of workers (roughly half of those employed during the first 30 days) opportunities to choose between full-time and part-time work according to their own preferences.

2016-07-01

2017-12-31

First, we are interested in multi-dimensional self-selection to the jobs offered. In particular, we will explore whether worker’s self-selection is based on measurable characteristics including 1) demographic factors such as age and family structure (i.e. existence of parental and spousal support), 2) socioeconomic status such as levels of education and income, and 3) ability and motivation and expectations for work.
Second, we will measure selection and causal effects of work incentive under part-time jobs on labor productivity. We will collect various information that captures the quantity and quality of data entry clerks’ labor productivity such as quantity and quality of the data entry during the training and at work. Specifically, we will measure the number of total entries (per hour) and fraction of entries with errors, which allows us to measure quality-adjusted labor productivity. We will also measure other aspects of job performance capturing work ethics and working hours: whether the clerk comes to the work (absenteeism), comes on time (promptness), and whether the clerk works voluntarily over-time. In addition, we will measure other behaviors that may indirectly affect their productivity and firm profits including retention (i.e., turnover) and use of leaves (e.g., sick, market, and funeral leaves).
Lastly, we will measure the impact of part-time jobs, compared to full time jobs, on time use, consumption including investment in children, and intra-household bargaining.

1. Experiment Chronology

AFF, the collaborating NGO, will hire full- (eight hours per day) and part-time (four hours per day) female data entry clerks in their catchment area. AFF distributed job flyers to women who aged between 18 and 35 with high school diploma during a census of approximately 22,000 households in the catchment area from July to September 2016. Part-time job opportunities were given to randomly-selected 41 villages and full-time opportunity to the remaining 40 villages. Flyers provide information on job description, working conditions including part- or full-time work, and expected salaries and benefits. A resume and a copy of high school graduation exam report were required to apply for both types of jobs. The total number of applicants was 529. From December 19 to 31, 2016, job applicants will participate in a baseline job survey which measures demographics, educational background, employment history, household income and assets, job preferences and motivation, personality, cognitive skills (clerical and computation abilities), fine motor ability, and basic computer skills.

Among those who complete the baseline job survey, about 220 qualified applicants will join two-week long training sessions in January to March 2017, which will entail basic computer training (such as Excel), specifics of data entry, and tests. Also, trainees will practice data entry at least for an hour every day. At the end of the training sessions, AFF aims to hire about 100 full time equivalent data-entry clerks (70 full-time and 60 part-time). Among workers who successfully perform the work during the first 30 days, AFF will give a randomly selected subset of workers (roughly half of those employed during the first 30 working days) opportunities to choose between full-time and part-time work according to their own preferences. We expect about half of them to remain in their original assignment. AFF will measure these workers’ job performance in the next 30 business days. Lastly, we will implement surveys of the hired workers and the follow-up surveys for 529 study participants beginning in March 2017.

2. Data

The primary data sources are (1) census data including household demographic and socioeconomic information, (2) administrative data used to select data entry clerks including the baseline job survey and training outcome, (3) data on job performance including labor productivity and attrition rates, and (4) the follow-up survey.

The census data cover approximately 22,000 households in the catchment area, where the job flyers have been distributed. The census has collected a variety of socioeconomic, health, psychological variables. This project will use key household characteristics such as the number of children, marital status, intra-household dynamics, and household income to examine the characteristics of workers who select into part-time (relative to full-time) jobs.

The baseline job survey will collect information on demographics, educational background, employment history, household income and assets, job preferences, attitudes and expectations toward work, and cognitive skills (measuring clerical and computation abilities). Fine motor ability and basic computer skills will be also measured. Also, we will measure time use, consumption, and intra-household bargaining.

Data on workers’ performance at the data-entry firm will be collected by AFF from March to December 2017. Data collection on job performance will begin from training sessions. During the training, AFF will collect training performance such as quiz score to measure understanding of the computer program, and quantity and quality of data entry practice. Also AFF will collect information on labor productivity at work. Data entered by each clerk will be automatically stored on a central server from individual laptop computers. These data from the training sessions will be used as a basis for job offers in the main intervention.

We will implement surveys of the hired workers to measure job satisfaction, job preferences, motivation, and career expectation, etc. Also, the follow-up surveys for 529 study participants will begin in January 2017. It will collect data on household and individual welfare including income, time use, consumption, and intra-household bargaining

1. Randomization for first round RCT
Assigning the treatment to villages in which individuals reside is done randomly in the project office of AFF by a computer random number generator. Job applicants were not aware of this randomization. It remains unknown to them even after they are hired.

2. Randomization for second round RCT
Assigning the hired individuals into the “continuation” and “choice” groups will be done randomly in the project office of AFF by a computer random number generator.

First, villages are randomly assigned to full- vs. part-time groups. Second, individual workers are randomly selected for choosing between full- vs. part-time jobs.

Yes

Total about 22,000 households in 81 villages
Total 529 job applicants

The job flyer was distributed by AFF to women who satisfy the above mentioned requirements (women aged between 18 and 35 with high school diploma) during the census of approximately 22,000 households in the catchment area from July to October 2016. The number of individuals who applied to either full- or part-time jobs is 529.

at least 60 work days of 70 full-time and 60 part-time workers

41 villages for part-time jobs and 40 villages for full-time jobs
70 full-time and 60 part-time workers