You'd really need to take the level of drunkenness into account, too. If you give me three beers I'll be too drunk to drive, but not "drunk" in the usual sense of the word. You'd have to give me another ten before I start being dangerous at a walking pace.
The biggest problem, though, is the random assumption that the same proportion of miles are walked drunk as are driven drunk. I personally have walked quite a few miles while drunk, but have driven no miles while drunk, and I hope I'm not the only one.
There are a lot of other variables, not the least of which is that the people who pass out drunk in a snowbank and freeze to death (or who ride a rusty Huffy head-on into traffic) are very likely to be homeless and/or mentally ill. (Speaking as somebody who worked in a public library in a poor neighborhood for several years and has dealt with lots of drunken street people...) It's a different segment of society than people who are arrested for drunk driving.
The people that 'walk while drunk' are most likely the people with a short distance between the bar/party and their destination. If their destination is far away, then they are probably drunk driving instead of drunk walking. How many people would walk to a party that was 5 miles away instead of driving? It's a very self-selecting answer and seems intellectually dishonest to me.
update: To me it feels like the entire reason for coming up with such a statistic is to generate controversy by saying, "drunk walking is more dangerous than drunk driving."
It makes more sense to view such statistics on a 'per trip' basis (i.e. x% of drunk walking trips end in death and y% of drunk driving trips end in death). This also doesn't take into account the number of dead people per accident too, but that might be skewed by practices like teens piling into a car with a drunk person behind the wheel.