First- US drivers take note and extra care on those days. 

Second- why do insurers not have this information anyway? Why do they not have past, current and trend data on both accidents and claims by time of day, day, location, category etc. The data is all there. OK- a great deal is "hidden" in text and free-form words but that is no excuse

Hidden there are fraud indicators both of claimants & fellow conspirators such as passengers, legal firms, scarp yard owners, used car sales operations often very conveniently in close proximity.

Combine that with the time of day the accident or incident took place and you unearth the implausible. 

• Minors frequently in accidents (a good psychological trick to make insurers settle early) during school hours
• Specific road junctions used by fraudsters to create crashes by innocent drivers
• Claims with invalid zip codes, national security numbers

This data, only if accessed and analysed of course, reveals legal companies whose claims are all outliers from norms- a good indicator of fraud. Drivers and passengers conspiring to defraud insurers.

Why is this data typically inaccessible and therefore not analysed? It is unstructured or semi-structured. Hidden in claims forms, emails and letters. Laying there without analysis.

And if insurers cannot detect past and current fraud they certainly cannot predict future fraud. Do not rely on AI if the data is still, raw and inaccessible.

Use tools like 360Retrieve that have already detected and had dismantled (by insurers and police not the software) major crime rings.

So, drivers avoid daydreaming at the wheel and insurers avoid daydreaming and missing the value in that inaccessible data.