The Interactions Between Beliefs, Behavior and Epidemiology

Principal Investigator: Jérôme Adda

 

This project aims to provide a comparative cost-effectiveness analysis (CEA) of public policies undertaken in real-life conditions during the Covid-19 epidemic in France. We will compare three main measures: i) confinement measures ii) testing with a specific focus on undocumented infections iii) reminders on the so-called “barrier gestures”. To do so we will build a theoretical model of disease diffusion and test the model using quasi-experimental variation during the outbreak to evaluate the respective importance of policies and the timing of their implementation.


Primary objectives

Some of the direct questions related to this goal are:

  1. Are confinement measures more cost-effective than testing or simpler prevention methods such as reminders on best practices? Does the timing of intervention matter?

  2. What was the role of undocumented infections in France during the outbreak?

  3. How do individuals and policy makers adapt and learn during a crisis?

We will provide a number of methodological and theoretical contributions to the analysis of the cost-effectiveness of public policies during an epidemic outbreak. The first one will be to provide a diffusion model that will include individual trade-off under belief-based uncertainty and the learning adaptive process during the outbreak (Bennet, 2015). The second one will be to provide reliable estimates of the cost-effectiveness using quasi-experiments. Lastly we will provide a heterogeneity analysis. Doing so we will explore the correlation of the share of undocumented cases with population characteristics and heterogeneity of the treatment effect of the above mentioned policies. 

 

Secondary objectives

  1. Provide a theoretical model including behaviors. Our theoretical model will start from Adhvaryu (Review of economic studies, 2014) and Li et al. (Science, 2020) and will aim at quantifying the role of undocumented infections in France during the outbreak. Our focus will be on individual behaviors under belief-based uncertainty and the learning adaptive process during the outbreak.

  2. Provide an empirical test for validation. Our empirical test will be based on Adda (Quarterly Journal of Economics, 2016) to estimate the relative cost-effectiveness of the above mentioned measures, while taking uncertainty and learning into account.  Data on infections will come from two cohorts of patients: the "cohort of infected patients" (since the first patient treated on 01/24 in Bichat) and the "contact case cohort" managed by Santé Publique France and REACTing. This data will be combined with school closure, transport limitation, public announcements dates on compulsory confinement, and Twitter volume peaks as a proxy for information campaigns. We will use a simple event-study approach combined with a fuzzy difference-in-difference to assess the effectiveness of such measures. To calculate the expected monetary benefits, cost estimates will be taken from the literature and data we will collect.

 The main outcome of the study will be the costing and comparative cost-effectiveness of public policy measures undertaken during. The secondary outcomes will be:

  1. A diffusion model including behaviors, uncertainty and learning.

  2. Quantification of the role of undocumented infections in France in transmission.

  3. High frequency database combining several French cohorts with real-time policies and economic activity. A short description is provided here: the "cohort of infected patients" (since the first patient treated on 24/01 in Bichat) and the "contact case cohort". These cohorts will be supplemented by data from the sentinel network and eventually GripeNet on cases of Acute Respiratory Infections (ARI) seen in consultation, according to the following definition: sudden onset of fever (or feeling of fever), and respiratory signs (such as coughing, shortness of breath or a feeling of tightness in the chest).This data will be combined with school closure, transport limitation, public announcements dates on confinement measures and sanitary emergency, data on population movements and Twitter volume peaks on preventive messages.

  4. Quantification of the impact of preventive measures and public policies with a fuzzy difference-in-difference framework. This estimate of the treatment effect will be combined with costs to reach the primary endpoint.