COVID-19 poses a major threat to countries around the world. Prior to distribution of vaccines to the general population, efforts to reduce the spread of SARS-CoV-2 are limited to non-pharmaceutical interventions such as social distancing and face coverings. Although conflicting recommendations have been issued, recent evidence suggests that face masks may significantly reduce the spread of SARS-CoV-2. However, questions remain on how to effectively promote mask adoption: Despite masks being mandatory, recent evidence shows less than 20% wear them at markets in Western Kenya. Together with the Ministry of Health in Siaya county, we evaluate a pilot program that distributes a free mask to each citizen, combined with educational interventions informed by insights from behavioral economics.
Intervention
Community health workers in treatment villages will distribute free face masks to every individual over 5 years in treatment villages, together with education on correct use and maintenance of distributed masks (T1).
Layered on top are 3 cross-randomized information treatments, consisting of additional educational information provided at the time of mask delivery (in T1 treatment villages) and routine visit (in T1 control villages), and through weekly text message reminders over 4 weeks.
T2.1 - Information on mask effectiveness T2.2 - Information on covid-19 symptoms, spread and risk factors T2.3 - Behavioral salience intervention to develop anchor points / reminders for mask wearing
Cross-randomized treatment T2.4 incentivizes local role models (identified at baseline by respondents as 'persons that you trust in health matters') to promote mask use in the village through wearing a mask in public, conversations and text messages.
Experimental design
Treatments will be cross-randomized as described below within 72 randomly selected villages in Ugunja subcounty of Siaya County in Kenya.
Design details
Randomization method
Randomization is done by a random-number generator on a computer.
Randomization unit
For T1 (free mask distribution) and T2.4 (role model intervention), the unit of randomization is the village. For T2.1-2.3 (information treatments), the unit of randomization is the individual, clustered at the village level (where all individuals selected for treatments T2.1-3 within a village are assigned the same information treatment).
Clusters
72 villages
Observations
Including baseline, we plan to collect:
a) 4 rounds of phone survey data from 2,160 households (30 households from each of the 72 villages), and
b) 4 rounds of public space observations, 3 slots of observations per round (with roughly 100 individuals observed in each slot) for each of the 72 villages. Total: 86,400 observations.
Arms
T1: 24 villages treated, 48 villages control
T2.1-2.3: 12 villages treated each (50% of households randomly chosen within each village to be treated), 36 villages control
T2.4: 36 villages treated, 36 control
Primary outcomes
We will collect 2 rounds of midline and 1 round of endline data from phone surveys and 3 rounds of public-space observations of mask use and social distancing behavior in villages and markets to test:
H1: whether free mask distribution increases the proper use of masks in public spaces,
H2: which type of messaging / channel is most effective in increasing mask usage,
H3: whether educational interventions spill over within village and social networks,
H4: whether (potential) increased mask use leads to changes in social distancing behavior, health and socio-economic outcomes
Our primary outcome for H1/H2 is the share of individuals wearing masks correctly (i.e. covering both mouth and nose) as observed in public places by our enumerators. Self-reported mask use is secondary, as our earlier research suggest such data may not be reliable due to social desirability bias.
To test channels through which interventions increase mask use (H2), we also collect detailed data on respondents' knowledge of different aspects of covid-19 transmission, symptoms and risk factors, as well as their views on the social desirability of masks.
For H3, we additionally collect data on how information on covid-19 and mask use spreads within villages and social networks, such as who individuals talked to about these issues, how often and when.
For H4, we collect data on social distancing from both self-reports (including the number of in-person interactions, travel patterns, attendance at markets and religious services, etc.) and public observations (including whether an observed individual maintains at least 2m of distance while interacting with others). In addition, we collect detailed data on households' economic activities, including income, self-employment, employment, consumption, food security. Lastly, we collect data on covid-related symptoms, testing and test-outcomes for covid-19 and other health-seeking behaviors and health outcomes.
We will further pre-specify these outcomes in a registered pre-analysis plan before conducting any data analysis.
Cited by
No citation data available for this registration yet.