I happen to track this in a home tournament I run; here are the numbers for our last 12 tournaments, shown as # of initial players (shouldn't really be a factor, but shown for completeness) and the BB when the tournament ended, as a percentage of the total chips in play -- i.e., BB / (startstack * Nplayers)
<font class="small">Code:</font><hr /><pre>
# players BB% at end
8 4.3%
9 1.3%
9 1.9%
9 1.3%
8 2.9%
8 8.6%
13 7.0%
8 5.7%
11 8.3%
6 2.9%
11 4.2%
6 5.7%
</pre><hr />
The tournament format we use has no ante.
As you can see, there's a lot of variance, which presents a challenge in trying to get the tournament to end by a certain time. In particular, the longest tournaments (the entries with 8% BB factors) I remember as being one of those all-in fests where the shortstack just always won and the chips just kept going back and forth ad nauseum.