Using cross-country COVID and broad economic data for over 160 countries, we estimate most of the models presented in the literature on the impact of non-pharmaceutical interventions (NPI) on the diffusion of COVID across the world. A very large majority of the papers conclude that the aggregate form of NPI’s – lockdowns - were successful in reducing infections, or deaths, or both. When not successful, the literature argues that failure was due to bad implementation i.e. the lockdown was implemented later than optimal, or accompanied by too little stringency. In addition to those present in the literature, we add the following important test of lockdowns – a before and after comparison for 143 countries, and for one, two, and three months from the date of lockdown. No matter what the test, the dominant result is that not only lockdowns were not effective, but that, in a large majority of cases, lockdowns were counter-productive i.e. led to more infections, and deaths than would have been the case with no lockdowns.
The paper also attempts to rank countries in terms of performance i.e. evaluation of what happened versus what should have happened post interventions. Three months after lockdowns, in only 18 countries lockdowns led to “good performance”; with stricter definition of success, the number falls to only 8 countries.
We also test, in some detail, the hypothesis that early lockdowns, and more stringent lockdowns, were effective in containing the virus. We find robust results for the opposite conclusion: later lockdowns performed better, and less stringent lockdowns achieved better outcomes.