Cash and Compliance with Social Distancing: Experimental Evidence from Ghana

This study tests the impacts of an intervention that provides weekly cash transfers via mobile phone, known widely in Ghana as mobile money. Participating households receive cash transfers of GHS 90 per week. The version of the intervention being tested in the present study lasts for eight weeks, yielding a total of GHS 720 sent to each treatment household over the course of roughly two months (control households in our study receive only a single transfer of GHS 90 at the outset of the program). The quantity of GHS 90 approximates recent social protection transfer schemes designed by agencies within the Ghanaian government to support consumption in the context of the COVID-19 pandemic.

2020-05-19

2020-08-15

-COVID-19 related health behavior
-COVID-19 related attitudes and beliefs
-Food security
-Expenditures
-Allocation of time between home/outside, work/leisure
-Psychological distress

-COVID-19 related health behavior refers to number of days in past seven spent: staying home all day; attending social gatherings; keeping distance of at least one meter from those outside immediate family; other members of household staying in all day; someone outside immediate family visiting the home; wearing a mask when near someone outside the immediate family
-COVID-19 related attitudes and beliefs refers to: beliefs on mortality rates, infection rates, influence on Ghanaian economy, and death rate in Ghana; and support for a nationwide lockdown, whether social gatherings should be canceled; whether people should stop shaking hands; whether shops shold be closed; and how extreme government reaction should be.
-Food security refers to number of days in past 7 in which someone from the household has had to rely on less preferred food, limit portion size, or borrow foods; and also whether the household has bought food for storage in past seven days, and numbers of meals taken respectively by adults and children.
-Psychological distress is measured by the widely used Kessler 10 scale.

-COVID-19 symptoms
-Religious beliefs

-The COVID-19 symptoms designation refers to a set of questions asking whether the respondent or anyone in the household has experienced fever, dry cough, difficulty breathing, or loss of smell, and whether medical attention has been sought for these symptoms.
-Religious beliefs are measured through a module of questions on frequency of prayer, scripture reading, religious service attendance, and beliefs on the connection between religious faith and wealth/health

We will conduct a randomized evaluation in Ghana that addresses the following research questions:
1) What are the effects of weekly mobile money cash transfers on COVID-19 related health behaviors?
2) What are the effects of the transfers on welfare (e.g., consumption, psychological distress, and physical health)?
3) How and to what extend do the transfers influence perceptions of the COVID-19 pandemic, and particularly the extent of risks perceived by the respondent?
4) How do effects differ across rural versus urban status and pre-treatment expenditure level?
Addressing these questions will allow us to contribute to scholarship in the behavioral, development, and health subfields of economics, and also to test the effectiveness of a policy tool (i.e., mobile money cash transfers) within the context of the COVID-19 pandemic. For our sample, we have selected 2,500 households from the nationally representative Ghana Socioeconomic Panel Survey. We anticipate that roughly 1,500 of these households will be eligible for participation.
These households will be randomized evenly into two experimental conditions: a treatment group and a control group. Randomization will be stratified by the rural versus urban status of the community in which the household lives, and the household’s reported monthly food expenditures in 2019. The treatment group will be designated to receive a mobile money cash transfer of GHS 90 per week for eight weeks beginning immediately after the baseline survey. The control group will be designated to receive only a single transfer of GHS 90, to be sent immediately after the baseline survey.
Innovations for Poverty Action staff members will contact a member of each household. Those who agree to participate and who are eligible (they must have an MTN mobile money account) will be added to our sample. Willing and eligible participants will be added to our sample, given a baseline survey, and informed of the cash transfers they should expect (depending on the experimental group they had been randomized into). All households will receive a transfer of GHS 90 following the first survey. Treatment households will then be sent GHS 90 once per week for the following seven weeks, while the control group will not receive further mobile money cash transfers. All respondents receive GHS 5 in phone credit as compensation for their participation in the survey. Biweekly followup surveys will enable us to estimate treatment effects on the study's outcomes of interest.
Finally, we plan to conduct a long-term follow up survey during the spring of 2021.
Our main estimates will be obtained by the following regression specification

y_(i,t)= α + β_1 y_(i,0) + β_2 Treat_(i,t) + β_3 X_i + δ_s + γ_d + μ_t + ϵ_(i,t)

Where y_(i,t) is the outcome of interest for respondent i at time t, y_(i,0) is the outcome for respondent i drawn from the baseline survey, Treat_(i,t) is the respondent’s treatment status, X_i is a vector of controls including gender and age of the household head and number of household members, and where δ_s γ_d , and μ_t are strata, district, and survey wave fixed effects respectively. We will use robust standard errors, clustered at the individual respondent level. Estimates will be obtained from the post-baseline surveys. The first post-baseline survey will capture anticipation effects, since households in both experimental group will have received transfers and the only difference for treatment households at this point is that they will have been informed that they will receive additional transfers. Estimates from subsequent surveys will include the effects of the transfers.

Randomization will be conducted by the PIs’ research team using Stata software. Specifically, the user-written randtreat command will be used with the universal option. Randtreat adjusts for mismatches arising from uneven strata size – experimental group combinations, while the universal option prioritizes overall (rather than within-strata) treatment-control balance.

Respondent

No

1,500

1,500

750 treatment respondents and 750 control respondents

Given our sample design of 1,500 participants, 0.5 probability treatment-control assignment, 80% power, and an alpha-level of 0.05, we calculated a minimum detectable effect size of 0.145 SD without controls and 0.144 SD with controls (i.e., strata, district, gender of household head, age of household head, and number of household members. In unit terms for the specification with controls, this translates to: (0.05 log Kessler (psychological distress) score; 3 percentage point impact on probability of severe distress; 6 percentage point probability of formal employment; 5 percentage point probability of “looking for work”; 0.4 days worked in past week; 0.76 average hours worked per day in past week; and GHS 8.25/month food expenditure.