Reading it, I think it is essentially a 'before October 2015' vs 'after October 2015' comparison of overall accident rates, given the caveats there. So the -40%

Reading it, I think it is essentially a 'before October 2015' vs 'after October 2015' comparison of overall accident rates, given the caveats there. So the -40% here should be a lower bound, since it includes a time period with 0% Autopilot usage (before October 2015) and a time period with an unknown but <100% Autopilot usage (after October 2015). If people after October 2015 used Autopilot for 50% of their mileage, then you'd expect the true effect of Autopilot to be more like -80%. This strikes me as suspiciously high: what fraction of airbag-worth accidents are supposed to happen on the highways where one can use Autopilot at all? So maybe there's some other trend going on which is driving this decrease.