Curbing the spread of COVID-19 through effective messaging

How to cite

Falco, Paolo; Zaccagni, Sarah. 2020. "Curbing the spread of COVID-19 through effective messaging." AEA RCT Registry. AEARCTR-0005582. https://doi.org/10.1257/rct.5582-1.0

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

Paolo Falco* , Sarah Zaccagni

* Primary investigator

Registry ID
AEARCTR-0005582
DOI
https://doi.org/10.1257/rct.5582-1.0
First registered
Last updated
Status
In development
Trial dates
2020-03-24 – 2020-07-31
Countries
Private
JEL codes
D03, I12, I14, C93
Registry record
http://www.socialscienceregistry.org/trials/5582

Abstract

We will run an experiment to test different ways of framing a message that promotes social distancing at the time of the COVID-19 crisis. The different variants of the message will differ in their focus, from an emphasis on personal risk to risk for family, other people, and for society as a whole. We will test what are the impacts of receiving different messages on self-reported compliance with social-distancing recommendations, as well as on indicators of behaviour from register data. We will run the experiment with a large sample of Danish residents who will receive the messages through the official channel of communication used by public authorities.

Intervention

Public authorities across the globe are focusing on lowering the transmission rate of COVID-19 to prevent the health system from being overwhelmed and collapsing. Since, COVID-19 spreads through coughing, sneezing and close contact (e.g., see the information provided by the National Center for Immunization and Respiratory Disease), social distancing can reduce the chances of catching the virus and spreading it by minimising the amount of close contact among people.

Methodology

We will contact a large sample of Danish residents through e-Boks (the official system of communication used by public authorities in the country) with a message describing the benefits of social distancing at the time of the COVID-19 crisis. Specifically, the treatment aims to incentivize people to “stay at home unless they have essential reasons for going out”. We will test 5 different ways of framing the recommendation, ranging from an emphasis on personal risk, to risk for family, other people, and society as a whole. The outcomes of interest will be measure by means of a survey the respondents will receive together with the treatment, and of a follow-up survey they will receive two days after responding to the first message. We will therefore be able to measure both intentions to comply with the message and compliance since receiving the message. In addition, by means of register data to be matched with data from the experiment, we will be able to measure impacts on objective indicators (i.e., not self-reported and verifiable by a third party), such as electricity consumption and expenditure on transport/fuel (this will be subject to data availability). Depending on the availability of funding, it may also be possible to add a second follow-up survey, two weeks after the first one.

Experimental design

The experiment consists of sending 5 types of messages to Danish residents between the age of 18 and 69. The messages will all recommend “staying at home as much as possible”, but will provide different motivations. A sixth message will be sent to a control group and will contain no recommendation (only the accompanying survey).

The FIRST MESSAGE focuses on personal risk. The SECOND MESSAGE focuses specifically on the risk for one’s own family. The THIRD MESSAGE focuses on risk for others in general. The FOURTH MESSAGE focuses on the risk that not complying with social distancing may contribute to the collapse of the public health system. The FIFTH MESSAGE will serve as a placebo and will only tell people to stay at home as much as possible, without providing a specific reason.

The first four messages will be further split into two variants, one framed in the loss domain (focusing on the risk of incurring a negative outcome by not complying with the recommendation) and another one framed in the gain domain (focusing on the positive implications of complying with the recommendation). Since the contents of the two variants are identical except for the framing, it will be possible to pool the two variations in the final analysis.

Design details

Randomization method
The different messages will be sent to randomly selected individuals across Denmark. The randomisation will be carried out by Statistics Denmark (and, more specifically, by the branch in charge of conducting projects of this kind, called “DST Survey”). They will draw a random sample of 30,000 Danish residents between the age of 18 and 69 from the general population, and will contact them through e-Boks (the official channel of communication used by public authorities, utility companies, banks, etc., accessible to everyone in the population of interest).
Randomization unit
Randomisation will be conducted at the individual level across the country.
Clusters
N/A (Randomisation at the individual level)
Observations
30,000
Arms
The 30,000 respondents will be split in 10 groups of equal size (8 treatment groups and 2 control groups, one receiving a placebo message, the other receiving no message at all).
Primary outcomes
To measure the outcome of interest (e.g., whether a person chooses to stay at home), we employ two strategies. First, the survey received together with the treatment will ask respondents for how long they expect they will go out of their homes the following day. This will give us a chance to test whether people’s intentions are impacted by different treatments. As a follow-up, we plan to contact the respondents after 48 hours from their first reply and ask what their actions actually were. This will enable us to verify which of the treatments proved to be most effective in changing behaviour (possibly bridging the intention-to-action gap among subjects). In addition, we will ask about the bahaviour of other people in the household. This will allow us to capture family spillover effects. In addition, by matching data from the experiment with information linked to people’s national ID numbers, we will test whether the treatments affect tangible non-self-reported outcomes that may proxy relevant dimensions of behaviour (such as, for instance, usage of public transport, expenditure on fuel, working from home, electricity usage at home, etc.). Specifically, the primary outcomes are the following: • Time the respondent expects to spend outside his/her home the day after receiving the message • Distance from home the respondent expects to travel the day after receiving the message • Time the respondent spent outside his/her home the day before receiving the follow-up message • Distance from home the respondent travelled the day before receiving the follow-up message • Time the other members of the family spent outside their home the day before receiving the follow-up message (average)
Secondary outcomes
• Emotional status of the respondent Subject to the availability of relevant data: • Usage of public transport • Expenditure on fuel • Electricity usage at home • Working from home • Internet usage • GPS data from mobile phones • Usage of delivery services

Cited by

No citation data available for this registration yet.

Keywords