In an effort to contain the spread of the COVID-19 pandemic, many countries around the globe adopted social distancing measures. Previous studies have relied on the geographical and temporal variation in the adoption of non-pharmaceutical interventions (NPIs) to show that early adoption of NPIs is correlated to lower infection and mortality rates. However, due to the non-random adoption of NPIs, the findings may not be interpreted as causal. We address this limitation using a different source of identification –namely, the regional variation in the placement on the pandemic curve at the time of a nationwide lockdown. Our results reveal how, relative to regions for which the lockdown arrived 10+ days after the pandemic's outbreak, regions where the outbreak had just started were able to lower their daily fatality rate by 2.5 deaths per 100,000 inhabitants. We also provide suggestive evidence of contagion deceleration as the main mechanism behind the effectiveness of the early adoption of NPIs in lowering the death rate, rather than increased healthcare capacity.