There is significant literature from EU authorities (and each of the member states local DOT).
Bottom line is:
* 0.08 is the last "safe" limit. Performance is already decreased, but It is approximately equivalent to driving with children in the backseat. Not the best, but acceptable risk. About twice as bad as 0.05.
* However, above 0.08, performance decrease sternly and exponentially. At 0.1, chances of death or dismemberment become alarmingly hight. It is not obvious for a driver to make the distinction between 0.08 "happy" and 0.1 "drunk", since one may not feel impaired, but he is, really.
* Anything over 0.12 is classical "drunk driving" as understood by common folks. Chances of accidents are extremely elevated.
* 0.05 is the bottom of the exponential curve. There are still benefits from driving with a lower BAC, but the lions share of the exponential decrease is passed. The difference with 0.08 is significant (half less chances of accident, or more, more pronounced for young drivers). Below that, chances of accident continue to decrease, but not as quickly, so there is little benefits to be reaped going even lower.
Another interesting point is that effect of BAC on drivers is very age related. Being drunk at 0.1 when you are an experienced driver in your 30's puts you back at the same risk as when you were 16 and road racing everywhere and everyone (this is bad, indeed). However, a teen driver at 0.08 is already at extreme risk (as if he was an experienced driver at 1.4 or more from my memory), the statistics I read just showed this result, but didn't explained why. Could be that most 30+ have acquired some sort of higher alcohol resistance, or that it requires more focus from teen drivers, focus that cannot be achieved when intoxicated, even mildly. Anyway, teens that consume alcohol should never drive, even at legal concentrations.