We quantify the lockdown effect: the extent to which a lockdown limits the spread of COVID-19 infections. For this purpose we focus on Sweden—one of the few countries without a lockdown—and use synthetic control techniques to develop a counterfactual lockdown scenario. Based on a donor pool of European countries, we construct a control unit that behaves just like Sweden in terms of infection dynamics before it is put under lockdown for 8.5 weeks. The outcome in the control unit approximates the counterfactual lockdown scenario for Sweden. Three findings stand out. First, at the end of the lockdown period, COVID-19 infections and deaths in Sweden would have been reduced by one half and one third, respectively. Second, the lockdown effect starts to materialize with a delay of 3 to 4 weeks only. Third, we analyze Google mobility reports and find that actual mobility adjustments in Sweden are similar to those in the control unit—in both cases there is evidence for social distancing, but more so in the control unit.