The COVID-19 pandemic is a global threat that has caused the lockdown of
about one half of the world's population for weeks. Containing the epidemics comes
however at a very large economic and welfare cost, so that lockdown is being re
moved in many countries. We model the dynamics of the epidemics after (full or
partial) lockdown removal, incorporating the equilibrium behavior of individuals.
Individuals optimally choose self-protection efforts (a reduction in contacts) given
their risk aversion, beliefs and perceptions, utility from contacts, effectiveness of ef
forts, and the current prevalence of the disease. We show that contrary to imposed
social distancing, individually optimal self-protection does not lead to a constant
reduction in contacts intensity. In equilibrium, self-protection efforts adjust at each
date in a non-proportional way given the number of reported cases. We calibrate
the model on French data. The general dynamics of epidemics differs markedly from
the one obtained when neglecting to account for equilibrium self-protection efforts.
It involves a rebound in the epidemics, with a peak number of cases that is about
1/10th of the one with business-as-usual, and a cumulated number of deaths that
is about 1/6th lower.