Can Workfare Programs Help Mitigate Negative Consequences of the COVID-19 Pandemic on Poor Households? Evidence from Randomized Controlled Trials in the DRC, Egypt, and Tunisia

Last registered on March 21, 2022

Pre-Trial

Trial Information

General Information

Title
Can Workfare Programs Help Mitigate Negative Consequences of the COVID-19 Pandemic on Poor Households? Evidence from Randomized Controlled Trials in the DRC, Egypt, and Tunisia
RCT ID
AEARCTR-0009078
Initial registration date
March 19, 2022

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
March 21, 2022, 1:15 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Primary Investigator

Affiliation
Duke University

Other Primary Investigator(s)

Additional Trial Information

Status
Completed
Start date
2015-05-01
End date
2021-12-31
Secondary IDs
Prior work
This trial is based on or builds upon one or more prior RCTs.
Abstract
The COVID-19 pandemic has had unimaginably tragic human costs and devastating negative effects on countries’ economies, triggering an unprecedented economic crisis worldwide and sinking the global economy into the deepest recession not seen in eight decades. Systematic studies documenting socioeconomic consequences from the pandemic, let alone effective strategies to mitigate these consequences, remain scarce. This study addresses this question head-on, leveraging two large and coordinated cluster randomized controlled trials (RCTs) carried out as part of the rollout of World Bank-funded public works programs in the Democratic Republic of Congo (DRC), Egypt and Tunisia between 2015 and 2021 and at least one endline survey in each country during the COVID-19 pandemic. It provides systematic quantitative evidence on economic conditions as a result of the pandemic and ascertains the extent to which participation in workfare programs helped mitigate the negative socioeconomic consequences of the pandemic and quarantine measures to contain it. The primary outcomes of interest concern potential socioeconomic consequences of COVID-19 on households. The array of any such consequences is broad, but this study focuses on six broad outcome families: (i) exposure to COVID-19; (ii) consumption and food security; (iii) economic shocks (e.g., job loss or drop in productivity and earnings); (iv) coping mechanisms; (v) children’s health and schooling; and (vi) subjective/psychological well-being.
External Link(s)

Registration Citation

Citation
Mvukiyehe, Eric. 2022. "Can Workfare Programs Help Mitigate Negative Consequences of the COVID-19 Pandemic on Poor Households? Evidence from Randomized Controlled Trials in the DRC, Egypt, and Tunisia." AEA RCT Registry. March 21. https://doi.org/10.1257/rct.9078-1.0
Experimental Details

Interventions

Intervention(s)
As with PWPs in general, the project activities varied widely in terms of their location, infrastructure, and implementation modalities, but they all shared key common features and experimental design, which makes rigorous analysis possible. Specifically, the PWPs in our study provided temporary employment opportunities to poor and vulnerable individuals, such as the opportunity to work in the creation and maintenance of local infrastructure, in exchange for a wage. The provision of temporary employment opportunities provided recipients with a source of income to help smooth their income and consumption.

Moreover, the local infrastructure projects benefited not just the workers, but also the broader community indirectly from the upgraded infrastructures and presumably better access to services. For example, newly constructed roads in rural areas can help connect farmers to agricultural markets and thus provide them with an opportunity to sell their products. Further, all PWPs in the proposed research use community-based geographic targeting to reach poor regions.

Our studies are specifically designed to capture both the effects of the PWP infrastructure projects and the voluntary participation in such projects in exchange of wage.

Overall, program features presented in Table 1 are representative of the distribution of PWPs from other low- and-middle income countries (Subbarao et al., 2013), except for those that allow for yearly participation in public works activities like India.
Intervention Start Date
2015-06-15
Intervention End Date
2020-12-31

Primary Outcomes

Primary Outcomes (end points)
I. Exposure to COVID-19 (direct and/or indirect)
II. Economic shocks due to COVID-19 (e.g., loss jobs, decreased hours/productivity or wages/earnings)
III. Food security and general consumption
IV. Shocks and coping strategies during COVID-19
V. Children’s health and schooling/disruptions during COVID-19
VI. Subjective well-being and cohabitation in COVID-19
VII. Shocks and coping mechanisms
VIII. Access to basic services (health and education)
IX. Women’s Empowerment (autonomy, agency and intimate partner violence)
Primary Outcomes (explanation)
I. Exposure to COVID-19 (direct and/or indirect)
a. Number of household members who have been infected with coronavirus.
b. Number of household members who have been exposed to people who were diagnosed with coronavirus.
c. Number of people in immediate social environment who contracted the virus
d. Respondent is worried about contracting this illness in the next 12-months (1 if worried/very worried, 0 otherwise).

II. Economic shocks due to COVID-19 (e.g., loss jobs, decreased hours/productivity or wages/earnings)
a. Employed =1 if the individual worked for pay for any business, farming, or other activity in the last week. Or was self- /family employed.
b. Has anyone in the household (including respondent) lost their job since the pandemic began (1 if yes, 0 otherwise).
c. Has anyone in the household (including respondent) gained a new job since the pandemic began (1 if yes, 0 otherwise).
d. Individual was not able to work usual in the wage job in the last week (1 if yes, 0 otherwise).
e. Total earnings from the work last week (this will be coded as zero for individuals who did not work in the last week.)
f. Any member of the household (not the respondent) was not able to perform the usual wage job.
g. Household operated a business, including a family business in 2020 (1 if yes, 0 otherwise).
h. The revenue from the business has decreased in the last month (1 if yes, 0 otherwise).
i. In 2020, household members including the respondents has worked on household farm growing crops, raising livestock, etc. (1 if decreased, 0 otherwise).
j. Normal activities in the farm such as raising livestock or fishing has decreased since the pandemic began. (1 if decreased, 0 otherwise).
k. The number of hours worked in the main job since the pandemic have decreased (1 if decreased, 0 otherwise).
l. The wage/earnings since the pandemic from the main job has decreased (1 if decreased, 0 otherwise).
m. Change in time use by head of the household or main breadwinner

III. Food security and general consumption

We will construct household food security/food insecurity and consumption expenditure based on the following categories:
• Food Security
a. Members in the household worried about not having enough food to eat because of lack of resources (1 if yes, 0 otherwise).
b. Members in the household were unable to eat healthy food because of lack of money and resources (1 if yes, 0 otherwise).
c. Members in the household ate less than usual because of lack of money and resources (1 if yes, 0 otherwise).
d. Members in the household were not able to purchase the food due to shortage in supply of food products or restricted hours of grocery stores (1 if yes, 0 otherwise).
e. Number of times respondent made trips to the grocery store in the last 7-days.
• Food Expenditures
a. Data on food purchased was collected by asking households about the values of each of the 10 food items purchased within the past week.
b. Data on the value of own production was collected by asking households about which crops they consumed in the two different seasons (season A= winter, Season B=summer) (based on questions in section C.1.40). We will convert these values into weekly terms.
c. We will compute total household expenditure on food as the sum of value of all food items purchased as well as self-produced during the past week. We will next convert these values to monthly figures by multiplying them by 4.33.
• Non-Food expenditures
a. Data on non-food household expenditure was collected by asking households about the expenses that were incurred during the last month.
b. Expenditure categories include medical, leisure (movie or food in a restaurant, etc.), clothes and decorations, a fee for transportation, electricity/gas/water, landline/mobile phone calls/internet, soap/detergent, hairdresser, etc.)
c. Money households had to pay for monthly rent for house
d. Total non-food expenditure will be constructed as the sum of all aforementioned non-food expenditure categories over a month.

• Non-Food expenditures
a. Expenditures on luxury goods, such as alcohol or tobacco
b. Expenditures on leisure

• Total household consumption expenditures and household per capita consumption expenditures
a. Total household consumption expenditure will be constructed as the sum of food and non-food consumption expenditures in the last 30 days.

b. Household per capita consumption expenditure will be constructed by dividing total household consumption expenditure by household size.
IV. Shocks and coping strategies during COVID-19
• Exposure to shocks

a. This family of outcomes covers shocks and events households experienced in the past 12 months
b. months, which may have disrupted their financial situation, as well as coping mechanisms.
c. Death of household principal income earner in the household due COVID-19
d. Serious illness of principal income earner in the household COVID-19
e. Job loss or business failure among household members COVID-19
f. Loss of livelihood/property due to fire, natural or other disasters COVID-19
g. Failed or bad harvest COVID-19

• Coping mechanisms (negative vs. positive)

1. Received money, food or other social assistance from the government or NGOs during the pandemic
2. Cut down meals consumption or bought cheaper food
3. Received money, food, or other assistance from the family or social network during the pandemic
4. Net amount of money received from friends or family during the pandemic
5. [Sending children away]: How many members were part of this household before the pandemic, but are no longer living in this household during the pandemic
6. Any money saved during the pandemic
7. How long household covered current expenses through savings
8. Household borrowed money and total money borrowed during the pandemic
9. Received money from family members or friends who are not a part of the household in the last 12 months (1 if yes, 0 otherwise).
10. Frequency of money received from family members or friends who do not live in the household have increased.
11. Amount of money received from family members or friends who do not live in the household have increased (this will be coded as zero for households who have not received money.)

V. Children’s health and schooling/disruptions during COVID-19
a. Number of children stopped going to school in the last three months.
b. Number of children who were able to follow remote classes.
c. Whether household was able to hire a private teacher or tutor for at home lessons
d. Number of children who missed necessary immunization in the last 12 months.
e. Number of children who have been sick in the past 30-days.
f. The household was able to take sick children to the hospital for medical help (1 if yes, 0 otherwise).
g. Household has access to internet and computer for remote learning
h. Parents assist children with remote learning without outside assistance
i. Whether any children in the HH is working on a family farm/ business.

VI. Subjective well-being and cohabitation in COVID-19
a. Individual well-being based on Ladder Happiness Measures
b. This will be created based on the options- Happy, Calm, and Anxious.
c. Tensions between the family members during the pandemic
d. This will be done after reverse coding the questions. This is a five-item scale. Higher scores are desirable in that they indicate the absence of distress amongst the individuals during the last month.
e. Tensions between the family members in the last 2-weeks.
f. This will be done after reverse coding the questions. This is a five-item scale. Higher scores are desirable in that they indicate the absence of distress amongst the individuals during the last month.

VII Access to basic services (health and education)
• Access to healthcare
a. Time it takes to get to closest healthcare facility/services
b. Children 5 and under were sick
c. Children 5 and under received required vaccinations
d. Having a source of potable water in the house member in the past year
e. Household member/felt sick
f. Household used clinic/hospital in last 30 days
g. Medical expenditures

• Access to education
a. Time it takes to get to closest primary school facility
b. Number of school-age children attending school
c. Number of school-age children dropped/taken out of school
d. Household used educational facility in the last 30 days
e. Education-related expenditures

VII Women’s Empowerment (autonomy, agency and intimate partner violence)
• Women’s autonomy
a. Has employment or income generating activity
b. Place of work (i.e., home vs. outside of home
c. Participation in social groups
d. Participation in committees
e. Made a purchase or expense on leisure in the last 30 days
f. Borrowed money or contracted a debt from another person or organization in the past 2 years
g. Plans to borrow money or contracted a debt from another person or organization in the next 12 months

• Attitudes towards gender norms and equality

a. Women should have the same rights and duties as men.
b. Women should have the same chance than men to hold socio-administrative positions in this settlement
c. Women should be eligible to be the president of management committees that exist in this settlement.

• Intimate Partner Violence

a. Emotional violence
b. Physical violence
c. Economic violence
d. Sexual violence
e. Domestic violence

• Involvement in risky behaviors

a. Tobacco and alcohol use
b. Reproductive healthcare

Secondary Outcomes

Secondary Outcomes (end points)
Combined both primary and secondary outcomes, as per the above.
Secondary Outcomes (explanation)
Ditto.

Experimental Design

Experimental Design
We leverage experimental studies and data from three randomized control trials in the DRC, Egypt, and Tunisia. One site follows household/individual-level randomization (DRC rural), whereas the studies in Egypt and Tunisia follow two-stage cluster randomization. In the household level randomized designs, the opportunity to work in a public works program only varies at the household/individual level and hence generates only two groups of households of interest - treatment households and control households– where both belong to the treatment communities. This allows us to estimate only the net effects of the program including both the direct and spillover/general equilibrium effects of the program that would impact beneficiaries and non-beneficiaries alike. In the DRC study, household/individual-level randomization was carried out using a public lottery.

In contrast, in the two-stage cluster randomized design carried out in the Egypt and Tunisia study, the opportunity to work in a public works program is first randomized at the community level and then within treatment communities, at the household level. This generates three groups of households of interest - treatment households in treatment communities, control households in treatment communities, and control households in control communities. A key feature of the two-stage design is that it allows us to separately estimate the direct effect of the public works program from the indirect spillover effects/general equilibrium effects of the intervention.

Specifically, in the DRC, we targeted 121 villages in which randomization was carried out the household/individual-level only.
In Egypt, the main villages sample consists of an initial sample of 196 matched pairs of control and treatment villages constructed from a larger sample of 760 villages in which projects could be implemented, based the census poverty map within Egypt’s most deprived Governorates (Asyut, Sohag, Giza, Minya, Qena, Beni Suef, Fayoum, Aswan, and Luxor). Matched pairs were constructed based on a set of covariates coming from the poverty map, creating observationally similar pairs of villages coming from same district. One village in each pair was randomly assigned to treatment and the other to the control. Of these, projects could be implemented in 78 villages due to operational constraints, with 86 control villages. We ended up with an effective sample was 164 villages. In addition, there are 70 villages in which randomization was carried out at the individual level only, as SFD had already awarded contract to the NGOs before the design of the study, although before selection of beneficiaries.

In Tunisia, the study sample consisted of 80 communities, formally known as imadas—the lowest level administrative unit—in rural Jendouba Governorate (province), one of the poorest and underserved provinces in the country. Of these, 40 communities were randomly selected to receive the treatment condition and the remaining 40 communities to the control condition.

In both studies, while the construction of lists of eligible participants (and management of the projects) was done by an NGO in charge of project implementation in each village/community, randomization of potential participants in the project was carried out remotely by the research team using Stata.
Experimental Design Details
Same as above. We leverage experimental studies and data from three randomized control trials in the DRC, Egypt, and Tunisia. One site follows household/individual-level randomization (DRC rural), whereas the studies in Egypt and Tunisia follow two-stage cluster randomization. In the household level randomized designs, the opportunity to work in a public works program only varies at the household/individual level and hence generates only two groups of households of interest - treatment households and control households– where both belong to the treatment communities. This allows us to estimate only the net effects of the program including both the direct and spillover/general equilibrium effects of the program that would impact beneficiaries and non-beneficiaries alike. In the DRC study, household/individual-level randomization was carried out using a public lottery.

In contrast, in the two-stage cluster randomized design carried out in the Egypt and Tunisia study, the opportunity to work in a public works program is first randomized at the community level and then within treatment communities, at the household level. This generates three groups of households of interest - treatment households in treatment communities, control households in treatment communities, and control households in control communities. A key feature of the two-stage design is that it allows us to separately estimate the direct effect of the public works program from the indirect spillover effects/general equilibrium effects of the intervention.

Specifically, in the DRC, we targeted 121 villages in which randomization was carried out the household/individual-level only.
In Egypt, the main villages sample consists of an initial sample of 196 matched pairs of control and treatment villages constructed from a larger sample of 760 villages in which projects could be implemented, based the census poverty map within Egypt’s most deprived Governorates (Asyut, Sohag, Giza, Minya, Qena, Beni Suef, Fayoum, Aswan, and Luxor). Matched pairs were constructed based on a set of covariates coming from the poverty map, creating observationally similar pairs of villages coming from same district. One village in each pair was randomly assigned to treatment and the other to the control. Of these, projects could be implemented in 78 villages due to operational constraints, with 86 control villages. We ended up with an effective sample was 164 villages. In addition, there are 70 villages in which randomization was carried out at the individual level only, as SFD had already awarded contract to the NGOs before the design of the study, although before selection of beneficiaries.

In Tunisia, the study sample consisted of 80 communities, formally known as imadas—the lowest level administrative unit—in rural Jendouba Governorate (province), one of the poorest and underserved provinces in the country. Of these, 40 communities were randomly selected to receive the treatment condition and the remaining 40 communities to the control condition.

In both studies, while the construction of lists of eligible participants (and management of the projects) was done by an NGO in charge of project implementation in each village/community, randomization of potential participants in the project was carried out remotely by the research team using Stata.
Randomization Method
In the DRC, randomization was carried out using a public lottery in the field whereas, in Egypt and Tunisia, both community/cluster-level and individual/household randomization was carried out remotely by the research team using Stata.
Randomization Unit
Village-level and individual-level
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
1--DRC: 121
2--Egypt: 186 clusters in matched pairs
3--Tunisia: 80
Sample size: planned number of observations
1--DRC: 125-135 clusters, though assignment to treatment was done at the individual level (i.e., randomly selected workers in all of these villages received the reatment.) 2--Egypt: 186 clusters in matched pairs (78 treatment villages and 86 control villages) 3--Tunisia: 40 treatment villages and 40 control villages
Sample size (or number of clusters) by treatment arms
1--DRC: 121: about 2,000 participants
2--Egypt: 186 clusters in matched pairs (78 treatment villages and 86 control villages.): About 2,500 participants
3--Tunisia: 40 treatment villages and 40 control villages: half and half and about 2,500 participants
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
See PAP document.
IRB

Institutional Review Boards (IRBs)

IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials