From my IRR from my 2010 LSAT, for 2006-2009, a 170 is 97.5th percentile, so 2.5% of LSAT test-takers get 170+. I assume that this doesn't change radically from moment to moment. According to LSAC's Volume Summary, there were roughly 87,500 applicants to law school for Fall 2010 (the most recent year for which data is available). Of those, 2.5% would mean that 2187.5 should have 170 or higher on the LSAT — or maybe more, because applicants probably are disproportionately higher-scoring than test-takers, but probably not by much.

Just to put an upper bound, LSAC says that about 171,500 tests were administered in the 2009-2010 cycle, and in 2008-2009 68.9% took the test once, 25.3% twice, and 5.9% more than twice. If we take those numbers to be relatively stable (70% once, 25% twice, 5% three times, let's say for simplicity), that means that there were roughly 134,000 independent test-takers in the cycle, which would give 3,350 people with 170 or above, and if we take all of those 171,500 LSATs, there were 4287.5 scores of 170 or above handed out (presumably several to the same people). At the end of the day, I'd guess that there were 2500-3000 people applying with scores of 170 or above, though the number could be higher.

Now, here are the 75th-25th percentiles for top schools (and median where useful), with number of matriculants (from LSAC's Official Guide to ABA schools):

Yale: 176-170, 214

Harvard: 176-171, 559

Stanford: 172-167 (median 170), 180

Columbia: 175-170, 400

Chicago: 169-173 (median 171), 191

NYU: 169-173 (median 171), 450

Berkeley: 165-170, 292

Penn: 166-171 (median 170), 255

Michigan: 167-170, 371

Virginia: 165-171 (median 170), 368

Duke: 167-171 (median 169), 228

Northwestern: 166-172 (median 170), 271

Cornell: 165-168, 205

Georgetown: 168-172 (median 169), 590

UCLA: 164-169, 320

UT Austin: 164-168, 379

Vanderbilt: 164-169, 195

Nothing below this has 170 or above as any part of the percentiles, though all the way down to Washington & Lee, there are schools with a 75th percentile within spitting distance of 170 (in that case, 167 — and that's for a school ranked 34th).

Now the calculations begin. Here's the basic concept of the methodology: Yale, for example, has a 25th percentile of 170. That means that 75% of matriculants have 170 or better. Since it has 214 matriculants, that's 160.5 people with 170's accounted for (round that up to 161 for sanity). At Harvard, the 25th percentile is 171, which means more than 75% have 170 or above, so I'll calculate a

*lower limit*number of 170's by taking the 559 matriculants and finding 75% of that, which is 419.25 (so 420). Here are the lower limits for the relevant schools:

Yale: 161

Harvard: 420

Stanford: 90

Columbia: 300

Chicago: 96

NYU: 225

Berkeley: 73

Penn: 128

Michigan: 93

Virginia: 184

Duke: 57

Northwestern: 134

Cornell: 0

Georgetown: 232 (the median for the FT program is 170, and there are 463 in that)

That accounts for 2193 people altogether, which is pretty darn near the number of people with 170's that I estimated there were in a given application cycle at the beginning. Also, I'm probably underestimating the number of people with 170's at Harvard, Chicago, NYU, and Duke, all of which have a percentile at 171 instead of 170, but that's probably not more than 100 additional people. For that matter, obviously, some people with 170 and above go to Cornell and to the schools outside the t14, though interestingly we can put upper limits on that (for example, it's fewer than 80 people at UCLA, or else the 75th percentile would be higher). If I had to guess randomly, I'd probably put it at 40-50 each at UCLA and UT Austin and maybe 20-30 at Cornell and Vanderbilt, which accounts for another 100+ people. At that point, I've accounted for something like 2400-2500 people.

Thus, I'd estimate that roughly 2500-3000 people apply with scores of 170 or above each year, and at least 85% (and perhaps nearly every single one) matriculates to a school in the t17. If the number really is 2500, something like one in six of them go to Harvard, and nearly half of them go to Harvard, Columbia, NYU, or Georgetown.

I imagine that this is not a surprise to anyone, but this is what happens to people who get 170+ on the LSAT.

EDIT: Note, if you're just coming to this thread for the first time, that these conclusions got somewhat revised in later posts.