On the Detrimental Impact of Non-Events on the Efficiency of Risk Communication

How to cite

Sommer, Mey-Ling; Krügel, Jan Philipp; Traub, Stefan; Paetzel, Fabian; Halm, Pauline; Jacobsen, Thomas; Nestoriuc, Yvonne; Felfe, Jörg. 2022. "On the Detrimental Impact of Non-Events on the Efficiency of Risk Communication." AEA RCT Registry. AEARCTR-0009213. https://doi.org/10.1257/rct.9213-2.1

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

Mey-Ling Sommer* , Jan Philipp Krügel , Stefan Traub , Fabian Paetzel , Pauline Halm , Thomas Jacobsen , Yvonne Nestoriuc , Jörg Felfe

* Primary investigator

Registry ID
AEARCTR-0009213
DOI
https://doi.org/10.1257/rct.9213-2.1
First registered
Last updated
Status
Completed
Trial dates
2022-04-20 – 2022-09-30
Countries
Austria (Vienna), Germany (Hamburg)
JEL codes
I12, D81, D83, C92
Registry record
http://www.socialscienceregistry.org/trials/9213

Abstract

Health agencies are responsible for informing the public about (potential) disease outbreaks. If many individuals follow the recommendations of an agency and take preventive measures in case a health risk is identified, the number of actual cases of harm may be low, even if the estimated health risk was high. The literature suggests that individuals who have experienced such a ``non-event'' may question, in hindsight, whether the risk was really as high as communicated and whether (costly) preventive measures were reasonable. Using an online-experiment, we investigate whether the experience of non-events causes participants to perceive a communicating agency as less reliable and to take more risks the next time a risk is communicated, resulting in a less efficient outcome for society. In a repeated task, participants receive information each period about whether their group is in a low- or high-risk scenario from a communication source, and then must decide whether to choose a safe option or invest in a lottery. A non-event occurs if a high-risk scenario was communicated but the actual number of participants who lost the lottery is low. We find that non-events decrease group efficiency in particular if an uninvolved bystander questions the communicated information after a non-event has occurred. Aside from this, communication generally leads to higher efficiency than no communication.

Intervention

We vary the source of communication. In the baseline treatment (T1), risk communication will be announced directly through the computer.
In the first treatment variation (T2), the received information will be announced with uncertainty. In the second treatment variation (T3), an additional person will assume the role of the communicator. The third treatment variation is identical to T3, however, the communicator will have an incentive to announce wrong information to the participants.

Experimental design

Participants are randomly assigned to groups of 5 players to play a lottery. Each player will also be assigned a color (blue or orange) which will affect the outcome of the lottery: If a player has been assigned the color blue, (s)he will win the lottery if (s)he decides to play. If a player has been assigned the color orange, s(he) will lose the lottery. The players have also the option to refuse to play the lottery, and save their potential payoff. Before the players decide to play or not play the lottery, the communicator provides information regarding the state of the world (either low or high probability of risk). Participants make individual decisions in the lottery. After the lottery, the players receive the information regarding their own outcome in the lottery and, in addtion, they can see how many players in their group have lost the lottery. The game will be repeated over 10 rounds.

Design details

Randomization method
Potential participants show their preferences to participate in scheduled sessions and are than randomly assigned to these sessions. Group, color assignment and color association (blue-orange vs. orange blue) randomization are done by a computer.
Randomization unit
Group assignment is randomized at the individual level; position of buttons in the lottery is randomized at the experimental session-level; color assignment is randomized at the group-level; color association (blue-orange vs. orange-blue) is balanced and randomized at the group-level.
Clusters
Clustered at group-level: 40-50 clusters
Observations
250-300 students
Arms
T1 = 50; T2 = 90; T3 = 90; T4 = 90
Primary outcomes
Trust in the information/source (dummy); Risk-Taking depending on to the communicated information (dummy) and the experienced non-events in the previous round(s)

Cited by

No citation data available for this registration yet.

Keywords