Belief Updating, Risk Information, and Urban Flooding: A Field Experiment in Mozambique

I analyze the impact of two interventions disseminating information through 6-minute videos, a public official video and a experience-based video. The videos are incorporated into questionnaires with the target population and take place from September to November, 2021. The goal of the interventions is to increase the flood risk awareness, worry and preparedness of the participants. The video interventions contain the same information about flood risk, causes, effects, and guidance on preparation. The visuals in the videos alternate between the speakers and (moving) images that help transmit the message. Publicly available images and recordings of local news outlets are used for this purpose. The same visuals are used in both interventions. To promote inclusiveness among all ages and education levels, the standard Portuguese versions of the videos are also dubbed in Chuabo, the main local language. In the public official intervention (treatment 1 or T1) the information is delivered by local government officials (one man and one woman) responsible for disaster risk management. The experience-based intervention (treatment 2 or T2) features two local residents (one man and one woman). The latter actors draw on their recent flooding experience to deliver the information. Finally, to measure the effectiveness of the interventions, I created a placebo video (control or C). This 6-minute video contains general information about the city of Quelimane, its history and main economic activities. The information is delivered by local residents, contains (moving) images, and is also made available in Chuabo.

2021-09-16

2021-11-13

Measurements include multiple sources of data: survey questions and behavioral measures based on SMS technology. The in-depth household survey questionnaire is divided into four parts. The first part includes detailed questions about the household’s composition, economic status, access to public services and financial instruments, flooding experience and trust. The second part is the video and the third part includes detailed questions about risk awareness, worry, preparedness, risk and time preferences. The final part of the questionnaire is the behavioral experiment.

The chief survey questionnaire excludes the first part, since the necessary context variables such as demographic characteristics and flooding experience are collected during a mapping exercise that precedes these surveys. The chief survey questions mostly differ from the household survey since they are focused on the block in general, not necessarily the chief’s household. Nonetheless, the same topics are covered.

The short survey includes questions from across most in-depth household survey sections. Some questions are asked before and others after showing the video. Additionally, it includes questions to determine the respondent's exposure to spillovers.

The survey measures are supplemented with a behavioral measures through SMS. After the survey, the respondents receive two different invitations through text messages. The first invitation, offers some useful information about the cyclone warning system currently active in Quelimane. Respondents have to reply ‘yes’ to receive the information. The second invitation asks respondents to send anonymous feedback and suggestions about disaster risk management to the local government.

The primary outcomes are survey questions on 1) risk awareness, 2) risk worry and 3) preparation intention, and the behavioral measure through SMS.

An important secondary outcome is the behavioral experiment. An insurance game is played to test the hypothesis that past experiences influence preparation behavior.

The experiment has three phases.

First, a mapping exercise is conducted for which all city blocks of Quelimane are visited. During these visits the block limits are established and interviews are conducted with the block chiefs.

Second, the intervention takes place during the in-depth chief and household interviews in 300 city blocks.

Third, the intervention takes place during short interviews with 30 percent of the households in 150 city blocks (subset of original sample).

The city blocks were randomly allocated into the control and two treatment groups stratified by neighborhood using Stata.

City blocks

Yes

300 city blocks

300 city block chiefs, 600 in-depth interviews with household head or spouse and 3562 short interviews with household head or spouse

100 city blocks per treatment arm for the in-depth interviews and 50 city blocks per treatment arm for the short interviews.