Except being drunk is actually a benefit in a car crash. You are relatively relaxed, which reduces the extent of injury.
This is a 'fact' that I hear quoted a lot, but for which I hear the actual relevant research cited very seldom. Not saying it's not true, but I am more than a little concerned that too much may be read into it in the retelling. Among other things, I suspect that the latter claim - specifically, a reduced risk of (musculoskeletal?) injury - is more plausible and more likely supported than the former assertion: a general, overall 'benefit'.
I'll leave aside the incidence of crashes when drunk; we'll take it as read that drunk drivers are so much more likely to be involved in a collision in the first place that it complete swamps any benefit from reduced injury severity. (There's probably also an increased risk of collision when a sober driver is distracted by drunk passengers, but we'll skip that, too.)
A study that looked just at injury severity versus 'degree of inebriation' probably tried to do an apples-to-apples comparison that controlled for things like vehicle size and speed, type of collision, and whether or not a seatbelt was worn. If there was a significant difference in seatbelt compliance between the 'drunk' and 'control' groups - that is, if the drunks were less likely to wear a seatbelt - the benefit of seatbelt-wearing would have been lost from the final analysis. (In other words, the study would have compared belted sober to belted drunk, and unbelted sober to unbelted drunk--but not looked at belted sober versus unbelted drunk.)
And then there is the post-crash treatment of injuries. Immediately after a collision, the drunk is less likely to be able to provide or effectively receive aid from his companions. Once medical personnel arrive on scene, and after transport to a hospital, the drunk is less likely to be conscious, less likely to be able to communicate problems to doctors, and less likely to be able to follow instructions or answer questions. Potential symptoms of serious injury - like altered mental state, vomiting, loss of bladder control, etc. - may be misattributed to intoxication. (These are all pretty moot in the event of relatively minor injuries, but for a severely injured patient can have a drastic effect on outcome.) Medical interventions (everything from drugs to surgery) may be less effective or more dangerous in an intoxicated patient.
All that may lead to statistically better outcomes for the drunk who suffers mild-to-moderate injuries, but worse outcomes for one with severe injuries--which is why I would put a big question mark over the claim of overall 'benefit'.