A brief search founds this:
http://www.casact.org/pubs/proceed/proceed51/51073.pdfIt's an old study. But, it does have an age group/accident correlation.
The problem with all statistics is their interpretation bias, collection bias, and mathematical manipulation of the data.
While I haven't made a detailed analysis of the data, at first blush it does not seem to support your assertion that elderly drivers contribute a higher rate of accidents than any other age group. In fact, it seems the contrary is true.
I will certainly agree that getting a DL (Driver's License) is far too easy to obtain and only requires a very minimum skill set. This skill set is expanded afterwards in only a very small population after initial issuance of the DL. Probably the most effective way to reduce accidents would be to make the initial DL test much more stringent and actually test proficiency rather than, can they breathe and pay the fee.
People tend to fall back on their training in reactive situations and DL training does not require performance demonstration or practice in emergency situations. Most extreme performance training is obtained by the DL holder just before the crunch. I'd consider that poor practice.
I do note you are suggesting increased scrutiny on a group of persons of which you are not a part. The laws-for-someone-else-syndrome I call it.
Rather than going after a specific age group based on your personal test case of one, and in order have the largest impact on accident statistics in all age groups, make the demonstrable proficiency of the entire DL population higher. How about doing a 4 wheel skid recovery? Or, control of vehicle under maximum braking effort? Or, a high speed chicane?
These kind of skills are likely to be needed just prior to an accident event, and proficiency of the skills could minimize or avoid the accident itself.
In essence, I feel your scrutiny on a very small population of accident contributors is driven more by revenge than actual concern about the overall accident rate. If you want the most bang for your buck, don't fix the corner cases, apply a fix to the entire population.
My $0.02 FWIW
Cheers,