Most people lie on accident reports to avoid potential complications with insurance payments. For example, few will admit that it was their fault when the traffic policeman is standing there just waiting for a confession and to deliver an expensive ticket.
Anecdote time. While going to medical skool, a doctor friend worked in the coroners office of a large city. Like all large cities, the coroners office had a steady stream of deadbeats, bums, winos, and homeless that arrived without the benefit of medical attention and records. Not wanting to spend the money on an autopsy and a medical examiner, they quietly guessed at the cause of death with fairly good accuracy. However, after a few embarrassing mistakes, that was deemed unacceptable. Causes unknown were also not a viable option. So, they inscribed "heart failure" on all such cases, which was certainly true, but not necessarily the cause of death. That actually worked well for a few years, until someone ran statistics on what appeared to be a heart disease epidemic centered in this large city. The city now requires either an attending physician report or a mandatory autopsy.
While I'm not in a position to prove or demonstrate this, I think you'll find that such "accident" reports are highly opinionated, are skewed in the direction of smallest settlements, and are rarely corrected.
Right. Big numbers are more accurate.
The theory is that given a sufficiently large number of independent studies, the errors will be equally distributed on both sides of a desired result, and therefore cancel. That has worked well for global warming predictions. Unfortunately, the studies have to be independent to qualify and does not work at reducing the distribution in a single study.
OMG! Do you really trust the government to do anything correctly? I wish I had your confidence and less personal experience. I'll spare you another anecdote illustrating the problem at the city level.
Ok, think about it. You've just crashed your car into an immovable object while texting. You're still conscious and on an adrenalin high. The police are on their way and the last thing you need is for them to find your smartphone on the floor of the vehicle. So, you make a phone call to your wife telling her you'll be late for dinner and by the way, you've decided to buy her a new car. The police walk up, ask you a few questions, and notice you talking on the cell phone. If you're cooperative, nothing happens. If you're a total jerk, the mention the cell phone in their report, and you get nailed for possibly talking/texting while driving. You're screwed if they confiscate the phone for forensic analysis or request a call record from you provider.
In short, the statisics are where they want them. If there's a political or financial benefit to showing huge numbers of talk/text driving accidents, they will magically appear. If they thing that nobody really cares about the numbers, you will have a difficult time finding them. If the numbers accumulate some academic interest, you will see the same wrong information repeated endlessly in statistical surveys and college dissertations. Everyone lies, but that's ok because nobody listens. Incidentally, 87.3% of all statistics are fabricated for the occasion.