key: cord-0688844-cnwmlleg
authors: Blayac, Thierry; Dubois, Dimitri; Duchêne, Sebastien; Nguyen-Van, Phu; Ventelou, Bruno; Willinger, Marc
title: Population preferences for inclusive COVID-19 policy responses
date: 2020-12-07
journal: Lancet Public Health
DOI: 10.1016/s2468-2667(20)30285-1
sha: 6da0709375f20f9bcbbdf0c7fdd31c4fdc936918
doc_id: 688844
cord_uid: cnwmlleg

nan

The web-based survey, developed with the oTree platform, was available for 2 weeks, from 4 th May 2020 to 16 th May 2020, on a dedicated server managed by the research-team. The survey-institute Viavoice 1 made the recruitment: over the 7500 persons that were contacted by telephone, 5331 accepted and received a web link, 1154 responded to the online survey, with a fully completed questionnaire and a signed online informed consent form (response rate 21.6 %). Graph 1 gives the data obtained from our sample, in comparison with data from the National Institute of Statistics and Economic Studies (INSEE).

Statistical tests demonstrated that our sample is representative for regions (chi2-test of independence not rejected at .05), although weakly unbalanced in terms of gender and age composition (chi2-test of independence rejected at .05, but not at .01). Discrete Choice Experiment (DCE) is a method used for individuals' preferences elicitation i . It consists in putting individuals face to hypothetical, but realistic, situations of choice (scenarios) between two or more alternatives (options) which are differentiated by the levels of their attributes. The choice by an individual of one of the options reflects his personal trade-offs between the attributes and makes explicit his preferences. The DCE has been applied in various fields, in particular for adoption of alternative medical treatments with side-effects as attributes ii . In our case, options were described as combinations of anti-COVID-19 national policies.

Respondents were asked to choose one option among a couple presented in the choice set -see Figure  1 below. The choice task was repeated three times with different options, randomly assigned. Attributes of the choice options are the different prophylactic measures (see Figure 1 ), possibly applied at various levels (e.g.: No-Mask (level 0); Mask in public places (level 1); Mask in all circumstances (level 2)).

The list of attributes was determined in April 2020 after an attentive consideration of the debates (anti-covid strategies reported by the press, at the national and the international levels). The initial list of attributes was discussed with public-health experts from regional health agencies (Observatoire régional de santé PACA). Mask (3 levels: non-mandatory; mandatory in public place; mandatory everywhere); Restrictions in bars, restaurants and festive venues (2 levels: open; closed all the summer season); Adaptations in the public transportation system (2 levels: normal; adapted to working hours); Leisure travel (3 levels: no limitations; limited to France only; limited to 100 km around); Digital tracking (2 levels: no tracking; implementation with free access); and, of course, Additional weeks of confinement (3 levels: no extension; extension for one week; extension for 3 weeks). The introduction of a monetary bonus (4 levels: 0; 500€; 1500€; 2200€) was guided by:

i) a theoretical view, from economics, that monetary incentives could have an impact on the population's willingness to accept and follow binding sanitary programs (as founded by Charness and Gneezy iii ). In a DCE study, introducing a financial compensation as a (continuous) attribute allows the researcher to properly estimate the level of monetary rewards that individuals are willing to accept in exchange for their effort. ii) the fact that some governments (ex. Japan) actually decided to give an unconditional compensation, in order to reinject money in the country, so as to curb the effects of the economic crisis, and/or to make the policy more acceptable iv (Ando et al., 2020).

From all the possible combinations of the levels of these 7 attributes (i.e. a full factorial design consists of 864 possible combinations), 84 options were selected (with a D-efficiency of 83% for main effects and first order interactions) and divided randomly into 42 scenarios (each scenario includes 2 options, named as Option A and Option B). Each individual had to choose one option for each of the 3 scenarios that were randomly presented to him (her). Figure 1 shows a screenshot of the decision screen (translated from French to English).

Our strategy was to give all the information on the 7 attributes in a global choice set (see the screen choice proposed), although it can be cognitively demanding. At the time of the survey, France was at the end of the first lockdown: the media were constantly talking about endof-lockdown measures; that makes the task simple for the respondents. A direct question about the difficulty of the task was asked: did you have difficulty understanding the situations presented? (In French: avez-vous eu des difficultés de compréhension des situations qui vous ont été présentées ?). We obtained the following answers: No difficulty: 50.43%; occasional difficulties: 40.38%; frequent difficulties: 7.10%; always: 2.08%.

Statistical analysis: based on the random utility theory v , the determinants of the 3462 binary choices (3 propositions x 1154 respondents) can be studied using a statistical model. After testing for various specifications, we estimated our model using the following functional form:

The model described above 2 corresponds to a conditional logit and can be estimated by maximum likelihood. The first estimation of this model, made on the general population, has been controlled by some characteristics of the respondents: age, gender, and date of the survey. None of these variables changed the sign -or the magnitude-of the coefficients β.

By carrying out the econometric estimation, we noticed that the best statistical fit for variable ₋ was a quadratic form. The effect increases more than proportionally with the number of additional weeks of lockdown. For example, when 1 equals -0.024 for one week (general population strata), it gives a -0.216 for 3 weeks (=-0.024x3 2 ), and -1.54 for 8 weeks (=-0.024x8 2 ).

For building the figure, we selected an estimated point at: − = 8 (8 additional weeks of lockdown) ; = 1000 (1000 euros of Bonus).

Vulnerability. 'clinically vulnerable' has been documented through the survey by the (self-reported) question: 'do you have an underlying medical condition increasing the risk of severe covid-19 illness' (yes/no, for you, for one member of your household).

Young people and the health risk. In the survey, we documented the willingness to take risk in the health domain. Young people have different risk perceptions (p-value < .001).

Applied Choice Analysis

Discrete choice experiments in health economics: a review of the literature

Incentives to exercise

Fiscal Responses to the COViD-19 crisis in Japan: the First six Months