There is evidence supporting an increase in accidents when switching back vs switching forward, but it's not really about the time switch.
It's the time people are driving at, and the light levels at those times. Falling back leads to more drivers now driving in dimmer light than the day before, and failing to adjust. If it was lack of sleep, then the "spring forward" period should produce more accidents due to lack of sleep, which is the opposite of what the data shows.
This lines up perfectly to explain why places that don't switch, don't see the increases. The amount of daylight decreases naturally by about 4 minutes a day in the fall, which people naturally adjust to. However on that one day the time moves by 60 minutes thus leading to a poor light vs road condition, which people are mentally unprepared for.
"Analogously, the team also found that leaving DST in the Autumn increases minor road accidents by 13%. This is mainly due to more minor accidents happening during the evening hours between 3 pm and 6 pm when an hour of sunlight is reallocated back to the morning. These findings support the hypothesis that drivers are more sensitive to road lighting conditions rather than to the possible sleep deprivation brought about by the artificial DST clock changes."
