Abstract
Several studies based on the Synthetic Control Method (SCM) link the many COVID-19 deaths in Sweden during the spring of 2020 to Sweden’s decision not to lock down. I show that the estimated effect of lockdown varies greatly between four hypothetical Swedens depending on the spread of COVID-19 in societies before lockdown. This suggests that the effect of lockdowns found in earlier studies using Sweden as a control are likely to be (at least partially) driven by unobserved variables such as the spread of COVID-19 in societies before lockdown, and other misspecifications. Including a proxy for the spread of COVID-19 before lockdown does not improve the results. Suggesting that the very short data series on the spread of COVID-19 before lockdown limits the SCM ability to accurately estimate the effect of lockdowns.