Understanding the effects of an employer-provided health security program on the well-being, productivity, and health seeking behavior of workers in a semi-formal manufacturing setting in Bangladesh

We will partner with Ayesha Abed Foundation, which is associated with BRAC Social Enterprise Division, to implement this project. The Foundation was established in 1983 and employs independent female workers (“artisans”) through its 600 sub-centres located in different parts of the country. Each sub-centre employs about 20-25 artisans who work mostly independently. At the sub-centres, the artisans almost always carry out simple tasks such as stitching (allowing the comparability of productivity between artisans across sub-centres).

The Foundation will roll out a mandatory health insurance policy (called Health Security Scheme) to all its (about 500) sub-centres to cover all the artisans associated with the Foundation with a nominal contribution of BDT 25 from the eligible artisans (working for at least four months over the past six months at the time of the enrolment). The Foundation is in the process of revising the policy and coverage under the health security scheme.

The Foundation will establish a referral system (in partnership with the BRAC’s Health Program). Once the condition is approved, the beneficiary will receive a pre-committed fund to pay for the possible in-patient services purchased from a pre-approved hospital. Atypical of other schemes, there will not be reimbursement and the beneficiary is allowed to keep any surplus of the fund disbursed. The benefit package will cover the husband of the artisan and up to two children. In case the artisan is unmarried, two other adult family members (e.g. parents) will be allowed to be under the program.

2015-10-01

2016-03-31

We will focus on the following sets of outcome variables:

(a) Health Care Seeking Behaviour, Utilization and Expenditure: Using both base line and follow-up surveys, we will measure the health care seeking behaviour (e.g., doctor visits or in-patient stay at hospitals) and expenditure. We will specially pay attention to out-of-pocket payments which are expected to decline because of the program.

(b) Mental Well-Being: The scope of participating in the health security scheme should provide the artisans with a buffer against substantial loss in savings (and possibly other income generating assets) that would be required by the household to meet the health care costs otherwise. This impact should be reflected in the artisan’s mental well-being. The research team will develop and implement standard questions to measure the level of anxiety among the workers.

(c) Productivity: Since the workers are paid at piece-rates, we will be able to measure individual productivity or efficiency. The Foundation uses standard minute values to set the piece-rates and assigns jobs to different workers (as suggested in the initial meetings with them). We will work closely with the Foundation to measure efficiency as rigorously as possible.

(d) Labor Market Participation: We will measure (using both surveys and administrative data), the number of days and hours worked before and after the introduction of the interventions. We will measure the pieces that the artisans turn in and also compensation received as such. Data will also be collected on time spent on other activities to capture the total effect of the interventions.

We will use the sub-centres as units of intervention and the artisans (and their households) as units of observation. We will use a cluster randomized control trial to identify the effects of introducing a health security scheme to the female artisans employed by the Ayesha Abed Foundation. Each sub-centre will constitute a cluster (with 20-25 artisans). The project will use a mixed-method strategy with specific focus on quantitatively assessing the impact of the scheme on health and labour market outcomes (such as salaries absenteeism, and productivity). In partnership with the Foundation, the research team will randomly choose 25 sub-centres from 50 pre-specified sub-centres for interventions. We will assess all relevant outcomes and possibly causally associate them with the intervention by comparing artisans between control and treatment sub-centres. We will measure the intent-to-treat effects for all the outcomes of interest using the following econometric specification:

y(ikt)= αD(k) + βT(t) + γ[D(k)*T(t)] + δX(ikt) + ε(ikt)

where, D(k)=1 if the k-th sub-center received the treatment and 0 otherwise. Similarly, T(t)=1 if the observation is made after the intervention is introduced and 0 otherwise. Hence, the coefficient γ will measure the impact of the program as by design of random assignment, D(k)⊥ε(ikt). Since the intervention will be introduced at the sub-centre level (specified as k), we will adjust the standard errors by clustering at the sub-centre level. In the previous equation, X(ikt) are other observable covariates (measured at the baseline) which would help us enhance the precision of the treatment effect estimates.

Fifty sub-centers (small clusters of production units with 20-30 workers per unit) will be selected by the implementation partner. The research team will select twenty-five sub-centers using a random number generator in a statistical package (e.g. Stata).

The unit of randomization will be sub-center. Sub-centers are small production units headed by "managers" (called sub-center-in-charge). About 20-30 artisans work at each sub-center. The employment relation is not formal and workers are paid in piece-wise fashion depending on the amount of work they have finished over a period of time. The sub-center-in-charge keeps record of the word done by each artisan. The tasks almost always involve embroidery and some other needle work. All the raw materials and the designs are provided to the sub-center by the foundation.

Yes

50 sub-centers.

1,100 workers or artisans

25 sub-centers will receive the health security scheme while 25 sub-centers will receive the scheme six months later that will be used as control arm

Since we do not have exact efficiency measures for the specific artisans included in this project, we just use a baseline efficiency measure of 50 per cent and expect a 5 percentage points increase in productivity with a standard deviation of 15 per cent. With an intra-cluster correlation of 5 per cent and average number of artisans per cluster (sub-centre) of 25, we get the required number of clusters of 50 in total (25 sub-centres in control and 25 in treatment group) to ensure the required statistical power to identify the impact of the program with sufficient precision.