Kapoor et al. (2020) uses rainfall on last weekend before the stay-at-home order as an instrument for social distancing. variables. They find that social distancing works and has persistent effects, especially early on:
We test whether earlier social distancing affects the progression of a local COVID-19 outbreak. We exploit county-level rainfall on the last weekend before statewide lockdown. After controlling for state fixed-effects, temperature, and historical rainfall, current rainfall is a plausibly exogenous instrument for social distancing. Early distancing causes a reduction in cases and deaths that persists for weeks. The effect is driven by a reduction in the chance of a very large outbreak. The result suggests early distancing may have sizable returns, and that random events early in an outbreak can have persistent effects on its course.
Hat tip to Marginal Revolution.