COVID-19 Messaging via Community Leaders and Peer-to-Peer Networks

How to cite

Davis, C. Austin; Shonchoy, Abu; Vernot, Corey; Mobarak, Mushfiq. 2020. "COVID-19 Messaging via Community Leaders and Peer-to-Peer Networks." AEA RCT Registry. AEARCTR-0005877. https://doi.org/10.1257/rct.5877-2.0

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

C. Austin Davis* , Abu Shonchoy , Corey Vernot , Mushfiq Mobarak

* Primary investigator

Registry ID
AEARCTR-0005877
DOI
https://doi.org/10.1257/rct.5877-2.0
First registered
Last updated
Status
On going
Trial dates
2020-05-09 – 2020-12-31
Countries
Private
JEL codes
D83, I12
Registry record
http://www.socialscienceregistry.org/trials/5877

Abstract

The novel coronavirus and the associated illness COVID-19 have spread rapidly across the world, posing grave threats to public health and economic well being. Until testing or a vaccine are available at a global scale, slowing the virus requires reductions in interpersonal contact and increased hygiene like hand washing. We propose a set of experiments that evaluate a social influence campaign intended to spread information and encourage the practice of protective behaviors. The experiments will speak to the underlying mechanisms driving the effects of a social influence campaign, allowing policymakers to design high-impact campaigns in response to this and future crises.

Intervention

Religious leaders, school headmasters, and members of the general public receive a phone-based interactive informational script about coronavirus and protective measures, developed from current WHO guidelines. Recipients are asked to share information and encourage protective behaviors in their communities. While there is substantial overlap, the scripts differ somewhat between the types of recipients, specifically:

- The religious leader script included scriptural basis and practical guidance for adapting religious observance.
- The general public script, hereafter referred to as the peer-to-peer intervention, included several randomized elements: a) the length of the list of protective measures, b) the presence of a statement emphasizing the efficacy of protective measures, and c) an unconditional monetary incentive labeled to encourage respondents to share information in their communities.

Experimental design

We begin with a sample 6,815 households in 264 unions of Bangladesh that was part of a prior study. Treatment is assigned at the union level:

- 50 unions assigned to the headmaster treatment. We attempted to reach all primary and secondary school headmasters in these unions in order to deliver the informational script.
- 50 unions assigned to the religious leader treatment. We attempted to reach all Imams in these unions in order to deliver the informational script.
- 55 unions assigned to the peer-to-peer. Three individuals were randomly selected from the sample of households in these unions to be "treated seeds." Treated seeds received the informational script. Each of the three randomized elements of this script (see above) was randomized independently. Seeds were also asked to share contact information for up to 5 individuals in their social network.
- 109 unions assigned to pure control. Similar to the peer-to-peer unions, three individuals were randomly selected to be "control seeds." Control seeds differ from treatment seeds in that control seeds do not receive the information script. However, we elicit social networks from the control seeds in an identical fashion to the treatment seeds.

Assignment of unions is stratified by division and random assignment in the previous study. Embedded within the main experiment, there are two sub experiments. First, as mentioned above, we randomly vary features of the peer-to-peer script: a) short vs. long list of protective measures, b) the presence or absence of a statement emphasizing the efficacy of protective measures, and c) the presence or absence unconditional monetary incentive labeled to encourage respondents to share information in their communities. Each feature is randomized independently at the individual level. For each feature, the two options are assigned with 50 percent probability.

The second sub experiment involves the religious leader treatment. The 50 unions assigned to this treatment were randomly divided into two groups of 25: Group A and Group B. In an initial phone call, Group A religious leaders received a survey and the intervention while Group B received only a survey. In a follow-up phone call one week later, Group A received a survey only while Group B received a survey and the intervention.

We will various data sources in our analysis. From the main household sample (9 randomly selected from each of the 264 unions):

i) Baseline in-person household surveys conducted in 2017 and 2019 as a part of the previous study.
ii) Baseline household phone surveys conducted prior to the interventions.
iii) Follow-up household phone surveys conducted after the interventions.

From the snowball sample, composed of the elicited contacts from treatment and control seeds:

iv) Follow-up household phone surveys conducted after the interventions.

From the religious leader surveys:

v) Initial phone survey
vi) Follow-up phone survey

From mobile phone providers:

vii) mobility, call volume, and SMS volume.

Design details

Randomization method
In office by computer.
Randomization unit
Union of Bangladesh for religious leader, headmaster, and peer-to-peer treatments; Individuals within the religious leader treatment; Individuals within the peer-to-peer treatment.
Clusters
264 unions
Observations
2,376 households
Arms
50 unions headmaster, 50 unions religious leader, 55 unions peer-to-peer, 109 unions pure control.
Primary outcomes
Knowledge of how coronavirus spreads; knowledge of coronavirus resources like official hotlines; calls to coronavirus hotlines; practice of protective measures.
Secondary outcomes
Mobility as measured by mobile phone records; volume of calls and SMS as measured by mobile phone records.

Cited by

No citation data available for this registration yet.

Keywords