The opposite is happening of course... women's insurance is steadily going up as the new data, combined with the old, is starting to change the statistics. The problem is that these data points create falsehoods. Speed involved in an accident means speeding is bad to them, they collect data on speeders and charge them more... what really happens though is speed difference and distraction causes most accidents. Complex interactions that are not easily found in the single point data streams.
Anecdotally, I cruise 30% above the speed limit on the highway and have never got in an accident in that situation. I credit this to the fact that I am constantly on the lookout for idiots who pull into the high speed lane without accelerating or properly judging the time it will take me to overtake them. If I see the possibility of one of those situations I'll drop my speed closer to theirs and overtake them more slowly. That kind of data point is not something you can easily gather but it makes me a very safe driver (only accidents I have been in were due to blind corners with snow drifts > 6" deep.) By looking at these single factor issues you can rarely identify the truth of a thing, and those that you can (like drunk driving) have already been weeded out.