The employment and welfare effects of job displacement insurance

How to cite

Hensel, Lukas; Caria, Stefano; Abebe, Girum; Gerard, Francois. 2022. "The employment and welfare effects of job displacement insurance." AEA RCT Registry. AEARCTR-0010551. https://doi.org/10.1257/rct.10551-2.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Lukas Hensel* , Stefano Caria , Girum Abebe , Francois Gerard

* Primary investigator

Registry ID
AEARCTR-0010551
DOI
https://doi.org/10.1257/rct.10551-2.0
First registered
Last updated
Status
Completed
Trial dates
2022-08-01 – 2023-12-31
Countries
Ethiopia (Hawassa)
Registry record
http://www.socialscienceregistry.org/trials/10551

Abstract

We have designed an experiment to study the welfare and labor market impacts of expanding job-displacement insurance in Ethiopia. A large ready-made garment factory in the Hawassa Industrial Park were laid off a large number of female workers. As is common in low-income countries, these workers will be eligible for a modest lump-sum severance pay and will not receive unemployment insurance. Recent research suggests that many of them will be unable to secure another job quickly, food security will deteriorate, and they will return to their homes in the countryside (Hardy et al., 2022). In this project, we will evaluate the impacts of offering (i) a scheme which pays 60 percent of the worker’s wage for 5 months, irrespective of employment status and (ii) a single lump-sum payment of the same value as the income support scheme. Through phone-based high-frequency surveys, we will study impacts on consumption, job-search, and employment outcomes. We will also look at broader impacts on migration decisions, mental health, and women empowerment. Finally, we will study how exposure to additional job-displacement insurance payments affects the demand for future insurance.

Intervention

We evaluate two interventions designed to boost the job-displacement insurance payments received by a sample of young female industrial workers in Southern Ethiopia who lost their job due to a fall in international orders suffered by their employer.

The first intervention – “monthly payment” - provides displaced workers with a payment equivalent to 60 percent of their past monthly wage for a total of 5 months, irrespective of the workers’ (re-)employment status after displacement. Monthly payments start about 30 days after job loss.
The second intervention - “lump-sum payment” - pays a single lump-sum payment of the same value as the (discounted) sum of the monthly payments in the other intervention. The lump-sum payment is offered at the same time as the first monthly payment in the other intervention.

We also collect data from a “working sample”— workers in a similar firm who were not laid off. This sample will provide a benchmark for the expenditure patterns during the study period.

Experimental design

We randomize individuals into one of three conditions:
A control group that does not receive any additional job-displacement insurance payments
A monthly payment group that is offered the monthly payment intervention
A lump-sum payment group that is offered the lump-sum payment intervention

Design details

Randomization method
Randomization was stratified on tenure and job type.
Randomization unit
Individual
Clusters
1500 workers
Observations
1500 workers
Arms
500 per treatment arm.
Primary outcomes
We will test the following 10 core hypotheses: 1. The interventions boost expenditure and savings compared to the control group. 2. The lump-sum intervention boosts expenditure the most in the short run, the monthly intervention boosts expenditure the most in the medium and long run. 3. The interventions increase the probability of living in a large city. 4. Null: The interventions do not change job search. The interventions could decrease job search (through an income effect) or increase job search (through a liquidity effect). 5. Null: The interventions do not change self-employment rates. 6. Null: The interventions do not change wage-employment rates. 7. The interventions improve match quality (earnings, spell and expected future duration, match with skills, job satisfaction). The monthly payments will be more effective than the lump-sum treatment. 8. The interventions boost mental health. The monthly payments more so than the lump-sum payment. The effects would be larger for workers assigned to the monthly payments than the lump-sum payment. 9. The interventions boost female autonomy and friendship networks. 10. People are able to select into the scheme that benefits them the most financially (expenditure at endline) and in terms of their overall well-being (life satisfaction + mental health). These core hypotheses will be tested using the data collected with the bi-monthly high frequency surveys. We will submit a separate PAP for the endline data collection. 14 months after treatment, we will run an in-person endline survey. We edited this registration to include the endline outcomes, without changing the pre-specification for high-frequency outcomes. Primary outcomes (end points) We will study the above mentioned hypotheses by looking at treatment effects on the following families of outcomes. 1. Expenditure plans 2. Savings plans 3. Job-search plans 4. Expenditure outcomes 5. Savings outcomes 6. Job-search outcomes 7. Employment status 8. Employment quality 9. Migration outcomes 10. Reservation wages 11. Psychological welfare (mental health + life satisfaction) 12. Female empowerment and friendship networks Additional primary endpoints measured in the endline survey: 13. Incentivized willingness-to-pay for displacement insurance. Based on analysis of the high-frequency data, we will focus our analysis on the following primary outcome families: (4) expenditure outcomes, (7) Employment status, and (9) migration outcomes. We will additionally focus on the analysis of informal transfers to and from others as we found this to be an important dimension of how workers react to the insurance payments in the high frequency surveys. We treat the other outcome families as secondary for the sake of the endline analysis. Estimation We will focus on the following comparisons: (i) monthly payment vs control, (ii) lump-sum payment vs control, (iii) monthly payment vs. lump-sum payment. To estimate the overall effect of displacement insurance, we will also estimate pooled treatment effects across both treatment arms on primary outcome families 4. (expenditure), 6. (job-search), 7. (Employment status), 10. (Reservation wages) and 11. Psychological welfare. The level of observation for the main specification will be at the month-individual or survey-individual level. When estimating treatment effects we will consider the following time periods in addition to estimating average effects across time: 1. Treatment effect in the first two months after the start of the payments to study the effect of the lump sum treatment against the monthly payment. 2. Treatment effect in the months three to six after the start of the payments to study the effect of the lump sum treatment against the monthly payment. 3. Treatment effect after the end of the pay-out period of monthly payments (months seven to ten) to study the persistence of treatment effects. 4. We will also estimate the impact on endline measures where applicable. We will also estimate more disaggregated treatment effects over time where applicable. We will estimate treatment effects using an ANCOVA specification with further control variables. We will select control variables using LASSO algorithms for each outcome separately. To limit concerns related to multiple hypotheses testing when there are multiple outcomes, we will construct an index for all primary outcome families in addition to estimating impacts on the individual outcomes. We also plan to leverage the exogenous variation created by the experiment to estimate a structural model of consumption and job search decisions. Heterogeneity We will explore treatment effect heterogeneity with respect to the following primary dimensions: (i) policy preferences: a dummy for preferring the monthly payment scheme, (ii) above median baseline savings. We will also explore heterogeneity along the following secondary dimensions: (iii) welfare at baseline: being above the median level of the psychological welfare index and (iv) empowerment at baseline: a dummy for having experienced an improvement in empowerment over the course of the previous employment spell. We will also investigate heterogeneity using endogenous stratification using leave-one-out estimators to split our sample into groups based on predicted job-search, expenditure and migration outcomes. We will also use causal forests to analyse further heterogeneity dimensions.
Secondary outcomes
We will also consider treatment effects on the following secondary outcomes 1. Expenditure sub-categories a. Total expenditure b. Food expenditure c. Non-food non-durable expenditure d. Rent expenditure e. Expenditure on durables f. Expenditure on investments 2. Job-search plans and expectations a. Reservation wage for formal jobs for September 2022. b. Reservation wage for informal jobs for September 2022. 3. Job-search outcomes a. Reservation wage for formal work for each high frequency survey round. b. Reservation wage for informal work for each high frequency survey round. c. Time spent on job-search in an average week for each of the ten months from September 2022 to June 2023. d. Expected time to find formal job (weeks) for each high frequency survey round. e. Expected earnings in formal job for each high frequency survey round. 4. Employment outcomes a. Time to first formal job b. Number of job-offers per application for each high frequency survey round c. Current number of jobs for each high frequency survey round 5. Employment characteristics (all measured for each high frequency survey round) a. Expected tenure in main job b. Job-satisfaction c. Dummy indicating the use of skills acquired in Hawassa Industrial Park. 6. Self-employment (all measured for each high frequency survey round) a. Stock of assets b. Total sales c. Number of employees d. Revenue e. Profits 7. Household finances (all measured for each high frequency survey round) a. Total amount of outstanding loans b. Maximum amount of money that respondent would be able to borrow within 3 months. c. Able to borrow 4000 Birr if needed. d. Able to access formal sources to borrow 4000 Birr if needed. 8. Consumption-based Welfare (all measured for each high frequency survey round) a. Number of usual meals per day b. In the past 7 days, how many times did you go to bed hungry? c. Dummy for worrying about not having enough food in the last seven days. 9. Prevalence of sex work and transactional sex (measured using randomized list experiment at endline)

Cited by

No citation data available for this registration yet.

Keywords