The Myth of Yield Protection Forum

[Celtic Bhoy]

This was definitely an interesting read so thanks for providing it! I did have one thought about it though that may or may not support your conclusions. I could be wrong, in which case don't worry about it.

This mostly applies to the T14 schools, but I don't really think that the data sets you used are appropriate for them to disprove yield protection. No matter the T14, people scoring in the bottom deciles on the LSAT are going to find enormous amounts of rejections and in some cases waitlists. The lowest 25th percentile for a T14 is a 165 on the LSAT (for Cornell). This is just outside the top scoring decile (which you noted was 167).

Because of this, the majority of people we'd expect to be "targets" and "over-qualified" for these schools are all crowded in one data set simply listed as the top decile. I know it's not the case, but hypothetically all of the waitlists on the UVA graph (just for example) could be from the 175-180 range and all acceptances below that. In that case, there would still be yield protection occurring, but it could be masked simply by the way the data is shown in your graphs. To get a better picture on yield protection (at least at the T14), I would think you'd have to break down the top two deciles (those where the majority of scores fall for these schools) into small groupings like 178-180, 175-177, etc.

I'm not sure if this would change the results, but I think it would clarify each schools trends a bit. I haven't fully thought it through though so if I'm way off let me know.

Thanks for the comment. You're a bit off base, but I'll let you down gently as it's partly my fault for not clearly defining how I devised the deciles.

Each set of deciles is unique to a school, and is based on all applicants in the pool (inclusive of URM and pending status). Let me unpack that a bit.

For simplicity, let's just look at LSAT scores (it gets more complicated when you weigh them against GPA). At Columbia, the top decile of LSAT scores corresponds to a 176, meaning that among the 1,813 LSN users who applied to Columbia during the two cycles ending '09 and '10, approximately 10% had scores of 176 or above. The next three "deciles" correspond to scores of 174, 173, and 171, and the tenth decile corresponds to anything less than 163. (I use deciles in scare-quotes here to indicate that they don't exactly correspond to 10% increments, due to lumpiness in score frequencies.) As you can see, these deciles are quite tightly bound. Now, compare Columbia to, say GWU. At GWU, the top four LSAT deciles correspond to scores of 171, 169, 167 and 167. (Note: 167 is repeated because the frequency of this score is so great among LSN users who applied to GWU.)

As I alluded to, it gets more complicated when you introduce GPA (leaving you with an Index combining LSAT and GPA). But it suffices to say that the deciles are a way of breaking up the entire applicant pool for a given school into (approximately) equal-size cohorts.

I can affirm that it is not the case that all of the waitlists at UVA are among those with LSATs greater than 175. These account for just 92 out of 868 waitlists.

Hope this helps!

Ahhh I see now. I misunderstood your use of "decile". Thanks for clarifying! It really is an impressive analysis so thanks again!

[dabbadon8]

@ dabbadon8

I've got higher numbers than you and a full ride to Illinois. Doesn't seem to be YP there.

I don't know. I am just curious why applicants who found their way into schools with significantly higher admissions standards were WL with what was most likely an identical application. Especially at a school with no known idiosyncrasies (like needing why essays or W/E). If adcomms are really that holistic that is great and that makes me appreciate my acceptances more. I would be really curious as to what actually is responsible for this.

[tothePAIN]

I would guess that the situation could be more complex and possibly time dependent. For example, a school could practice yield protection after it has already given out the money it would use to attract more qualified applicants.

[notanumber]

It's quite possible that, in the upper end, schools are making decisions based upon soft factors that indicate how likely the student is to attend (e.g. residency, supplemental information, etc...) instead of more typically "significant" soft factors.

I'd also like to see the numbers of students at the upper end and their respective scholarship awards - it's quite possible that schools like UVA and Michigan consistently wait list students with "top" numbers for whom they cannot provide scholarships.

Either of those factors would explain your data while still suggesting that schools make admissions decisions based upon projected yields.

Or perhaps I'm just bitter that UVA didn't admit me.

[beachbum]

I think I was YP'ed by UIUC. Dean Pless is on these boards and may have read me say something stupid ha. Continuing with saying something potentially stupid...

Seriously though, not just for my ego, but if there is no YP why do you think UIUC waitlisted three people on these boards who will be now attending Michigan, Chicago and Duke? I am not saying that it is not beyond the realm of possibility that we simply were not what they were looking for...it could be, I just find it strange that people who had otherwise successful cycles would get waitlisted somewhere that most people on these boards predicted would give them $$$. My gf, got in with an identical gpa, an LSAT 12 points lower, same UG, same major, very similar softs. Maybe dean pless really is stalking the boards.

Eh, I think the argument for YP can only be made when you have multiple data points that illustrate a trend, kinda like this: http://minnesota.lawschoolnumbers.com/stats

In Illinois's case, I think the prevalence of a few WLs among the admits actually reinforces the idea of a holistic admissions process. As the Duke-admit you referenced, I really feel Pless & co. just didn't think I'd be a good fit for their program, and I wouldn't be surprised if my ho-hum attitude towards the school came through in my app. And the same can be said for my Notre Dame WL. I didn't do the optional essays; I used my generic PS; and I'll be the first to admit that my softs (and UG reputation, for that matter) are not particularly strong.

On the flipside, the two strong programs (Duke and Vandy) that I was admitted to were the ones in which I showed the most interest. Enthusiasm and thoughtfulness, I think, count for a lot.

[Bildungsroman]

@ dabbadon8

I've got higher numbers than you and a full ride to Illinois. Doesn't seem to be YP there.

I don't know. I am just curious why applicants who found their way into schools with significantly higher admissions standards were WL with what was most likely an identical application. Especially at a school with no known idiosyncrasies (like needing why essays or W/E). If adcomms are really that holistic that is great and that makes me appreciate my acceptances more. I would be really curious as to what actually is responsible for this.

Maybe Illinois thought your application sucked.

[jtemp320]

I know you said you aren't doing more graph work, but I think an interesting analysis would be the tendencies of YP in cities where there are multiple law school options. I think geography is by far the biggest factor in whether a school practices YP and the degree at which it does. DC for example, adcomms probably safely assume that any over qualified applicant at American U is also applying to GULC and using American U as a safety school. Law schools that are the best (highest ranked) option in a particular geographic region probably practice a lesser degree of YP or not at all.

Excellent point. You may be onto something. Perhaps I'll try Fordham?

More reliable data then LSN are the applicant profiles on some schools "school description" PDFs on their lsac.org profiles like the one I used of Hastings. For a few schools the evidence of yield protection is pretty stark. Unfortunately many schools (especially most T14) don't include this data. Like I said I agree with you that T14 schools YP less then people think but I dont think its even arguable that yield protection is a myth...

[dabbadon8]

@ dabbadon8

I've got higher numbers than you and a full ride to Illinois. Doesn't seem to be YP there.

I don't know. I am just curious why applicants who found their way into schools with significantly higher admissions standards were WL with what was most likely an identical application. Especially at a school with no known idiosyncrasies (like needing why essays or W/E). If adcomms are really that holistic that is great and that makes me appreciate my acceptances more. I would be really curious as to what actually is responsible for this.

Maybe Illinois thought your application sucked.

Quite possibly, I have no problem with acknowledging that as a more probably cause then UIUC trying to game the rankings.

I would guess that the situation could be more complex and possibly time dependent. For example, a school could practice yield protection after it has already given out the money it would use to attract more qualified applicants.

I actually think this is a good theory and sounds very plausible. It would be YP though. If a school knows an applicant would not likely attend with out offering them a good amount of money they may WL them not because they don't think they are worthy of admission but rather they may feel they are not worthy of admissions with substantial money.

Anyways, not trying to make this about me. I was just trying to use this as an example.

[Sandrew]

How did you choose your index? Seems like your data could go 100 other ways depending on that. 60% gpa and 40% lsat is pretty darn optimistic for GPA folks...

I started with 50/50 weights for all schools. I then iterated on each school until I found the steepest-sloping acceptance curve. You are correct that the graphs can change dramatically if you make dramatic changes in weights. And come to think of it, my methodology is a bit biased. A better methodology would iterate for the steepest-sloping acceptance curve over a range excluding the top deciles (e.g. 4th through 10th deciles). To be fair, I've re-run UVA's with an inverted weighting scheme (60/40 LSAT/GPA). By this weighting, your suspicion is justified. There does appear to be some YP at UVA.

Also, your basic suspicion that the numbers being worth 9 times more than acceptance or yield, etc. is kind of flawed. I would hardly be surprised if over 90% of the people being yield protected would choose "better" schools anyways.

Right again. But in my defense, this is more USNews' fault than mine. Their methodology is bollocks. In fairness though, if you're yield-protecting purely for rankings, it's at best ambiguous whether it can help/hurt.

You also assume that the only factor for yield-protect is rankings, when it also adds a lot of other benefits: easier to dole out scholarships at a pace that suits people that will actually attend, better numbers to publicize as far as yield (regardless of rankings), easier to figure out quartiles earlier in the process, etc.

I assume no such thing. I explicitly acknowledge other (non-rankings) factors and espouse them as justified.

[powerlawyer06]

Almost everyone who claims yield protection is doing it to cushion their ego.

This.

Additionally, it does not make sense for T14 schools to yield protect. Any law school in the T14 has enough prestige to reasonably expect that any law school applicant that has applied to their school will at least consider an offer from them. However even admissions deans and adcomms know there is a hierarchy in the T14 schools (HYS>CCN>MVPB>DNCG). This does not preclude Duke from getting any 175+ 3.75+ applicants this only means that if Duke wants to remain a competitive choice to the 3.75+ 175+ crowd it needs to offer those students scholarships. Yield protecting would mean that ADcomms think that a candidate only cares about rank inside of the T14. I think most people on TLS have all heard of at least one or two people picking a lower T14 for some type of scholarship even though they were offered in the T6.

This is would be the equivalent of someone who is a 10 on the looks or datability scale going for 8s and 7s and rejecting all 9s and 10s because they might turn you down. Grow some balls bottom T14! You are a 10 and you need to start only talking to 9s and 10s. If you want to get a super hot ten then just flash some cash.

I also agree with the geography of a school and applicant theory posed earlier in this thread.

[FiveSermon]

Any law school in the T14 has enough prestige to reasonably expect that any law school applicant that has applied to their school will at least consider an offer from them.

So you are telling me a 180/4.0 AA male will consider an offer from Cornell/Gtown barring a full ride?

[Bildungsroman]

You messed up the quotes, brah. I didn't say that.

[birdlaw117]

One fallacy I can see in this (although I don't dispute a lot of this very good data) is that it excludes "pending" applicants. It seems to me, though I could be wrong, that some schools take the "let's just wait them out" approach. If this is the case, the "pending" applicants become a very relevant group to consider (though a very difficult one to measure accurately). Just my thoughts. Props on the analysis as a whole though, cool stuff.

[powerlawyer06]

So you are telling me a 180/4.0 AA male will consider an offer from Cornell/Gtown barring a full ride?

Yep. I think they will consider it. They may not accept it without money (see my super hot ten analogy above). It would be ridiculous for a T14 school to not make an offer to this candidate. Since there is not a formal interview process in which law schools can gauge a candidates level of interest, a T14 should assume they have a decent shot at this candidate. Reasonably speaking, all T14 schools should offer acceptance to any candidate that they want to attend their school and offer money to the best ones. By yield protecting a law school is shooting itself in the “rankings” foot. If you don’t make offers and pursue the best candidates then you will never get the best candidates and you will never move up in the rankings. In fact a T14 that yield protects too much could fall out of the T14 to be replaced by a more aggressive regional school.

P.S. This is precisely why all MD schools have mandatory pre-offer interviews where the “Why Us?” question is always asked. It helps them gauge an applicant’s level of interest and then they only offer those who they believe will definitely attend.

[JusticeHarlan]

I'd like to see the Michigan data for waitlists with a Why Michigan is awesome versus waitlists without a Why Michigan is awesome. The same could be said for UVA, etc.

I'd like to see the data for UVA broken down into those who applied ED and those who didn't.

Overall, great work OP, you make a very convincing case. I think the higher LSAT weighting would be more instructive though - one of your arguments was that YP is an inefficient way for schools to game the USNWR rankings, and as USNWR weighs LSAT more heavily than GPA, I'd be curious to see if you get more results like UVA when other schools are looked at through the same filter.

[D. H2Oman]

As shown above, acceptance rates are across-the-board monotonic at these T14 law schools. The conventional wisdom (around TLS, anyway) is that UVA and Penn are more liberal with their waitlists (and stingy with their admittances), on the basis of numbers, than other T14 schools. This fact is borne out by the data. But while UVA and Penn waitlist a lot of qualified applicants, they’re no more likely (less likely, in fact) to waitlist their more-qualified applicants over less-qualified ones. These data are entirely consistent with each school’s putting stronger emphasis on so-called soft factors that don’t necessarily correlate with candidates’ hard stats, such as in-state residency at UVA or work experience at Penn. Come to think of it, this is exactly what these schools avow—a holistic approach to admittances. Nothing to get excited about here.

Conclusion: Yield protection at T14 law schools is a myth.

Good work OP. But I do have an issue to point out here. You're making a pretty sizable assumption imo that they're using a legit method of weighing softs factors when making these decisions. I think the biggest "soft factor" schools like UVA and Penn use is "likelihood of attending" based on "Why X" essays, personal statements, etc etc. I have no problem with this at all btw, but I do think that there is a certain degree of what TLS calls "Yield Protection" involved there. I mean when someone gets into HSCCN and gets WL at V (this happens a lot btw, I've seen it many times) I think that tells us something. I think your data definitely shows us a difference between how CLS and GULC operate and how P and V operate. But, I think your data still shows possible yield protection. In fact, that seems more likely to me than just weighing of true "soft factors" such as UG or WE.


edit: oh wait, your index for UVA and Penn 60/65% gpa 40/35% lsat is totally backwards what the hell, that fucks every thing up. Even 35% GPA, is beyond optimistic for gpa people btw. It should probably be somewhere in the 70-75% lsat range.

[Sandrew]

One fallacy I can see in this (although I don't dispute a lot of this very good data) is that it excludes "pending" applicants. It seems to me, though I could be wrong, that some schools take the "let's just wait them out" approach. If this is the case, the "pending" applicants become a very relevant group to consider (though a very difficult one to measure accurately). Just my thoughts. Props on the analysis as a whole though, cool stuff.

Thanks for the comment. The data, being self-reported, are unquestionably skewed, and I agree with your suspicion that pending applicants are likely to include more waitlisted or rejected applicants (perhaps by way of a "discouraged applicant" effect). But that does not cause me to question the conclusions. For the exclusion of pending applicants to affect my conclusions, they would have to disproportionately represent YP'd students. I've analyzed the index distributions of pending applicants, and they don't significantly differ from the non-pending applicant pool. I admit that this is not necessarily conclusive. It could yet be the case that, among pending applicants, those with a high index were more frequently waitlisted than those with a lower index. I just don't see a mechanism to explain why that might be the case. Alas, there is no way to resolve this without robust data.

Thanks all for the thoughts and keep the questions/criticism coming!

[D. H2Oman]

Lol OP, I just saw your LSN in your profile, I say this entirely in good fun because you have ballin numbers.....but the ultimate irony here is that I 100% guarantee that you are going to get YP'd at uva.

[FiveSermon]

Lol OP, I just saw your LSN in your profile, I say this entirely in good fun because you have ballin numbers.....but the ultimate irony here is that I 100% guarantee that you are going to get YP'd at uva.

And probably at more than a few other T14 schools.

[Verity]

Where does GPA fit into your analysis? Analyzing splitters might provide insight.

[D. H2Oman]

Third and final comment I promise

I started with 50/50 weights for all schools. I then iterated on each school until I found the steepest-sloping acceptance curve.

Here's another huge issue, you seem to be unintentionally distorting your own data. We know that Penn is not really a more GPA friendly school than Columbia. However, your looking to steepen the curves as much as possible, and then looking for that big bump in the middle like on the American graph. But UVA does not YP the same way AU YP's. You're not gonna see that huge bump, because UVA doesn't yield protect by accepting a bunch of 171-172 3.7 students and waitlisting all the 175+. UVA waitlists pretty evenly at every LSAT score over 170, (and they waitlist a lot!) so for a school like UvA you're not gonna see a steep curve, unless you mess with the ratios to make one. At a 169 LSAT score you pretty much can not get in. But once you hit 171 your chances won't really improve at all by raising your LSAT score. The lack of a steep curve actually IS the yield protection in action.

[bdubs]

I think people on these boards conflate some of the sheer randomness of admissions with YP and softs. People can throw similarly well written and tailored apps to the whole T14 and get extremely varied results. This does not always correspond to having the right or wrong numbers for a school, or to having a particularly convincing story.

I think part of this randomness is driven by the way in which law school admissions are done, usually by a small handful of admissions deans that read thousands of applications. When each application gets a very cursory review, it is easy to imagine some very good applicants slipping through the cracks at highly ranked school X while getting in to all of X's peer schools (or more highly ranked schools).

[Sandrew]

Good work OP. But I do have an issue to point out here. You're making a pretty sizable assumption imo that they're using a legit method of weighing softs factors when making these decisions. I think the biggest "soft factor" schools like UVA and Penn use is "likelihood of attending" based on "Why X" essays, personal statements, etc etc. I have no problem with this at all btw, but I do think that there is a certain degree of what TLS calls "Yield Protection" involved there.

Thanks! Reasonable people may disagree as to what constitutes a legitimate soft factor. I believe that likelihood of attending is a legitimate factor. And yes, weighing such factors can result in yield protection (broadly defined). Your comment leaves wondering whether you agree. In one breath you imply that it's not "legit" and not a "true soft"; in another you state that you have "no problem with this at all."

I mean when someone gets into HSCCN and gets WL at V (this happens a lot btw, I've seen it many times) I think that tells us something.

I disagree. But thanks for bringing this up. This is a common fallacy on TLS, and I should have addressed it directly.

UVA may waitlist more people across the board than higher-ranked schools, but that's their prerogative and doesn’t spell yield protection (and certainly not ratings cooking). Again, my definition of yield protection is that a student with a higher index is, contra expectation, less likely to be admitted. By examining the shape of the acceptance curve, I’ve purposely controlled for each school's propensity to waitlist.

Perhaps you define YP differently. If so, then we’re simply talking past one another. But I chose my definition carefully. If you define YP as a propensity (relative to other schools) to waitlist students, I think your mistake YP for something else.

Loads of people are waitlisted at School X who are accepted to higher-ranked schools. This fact doesn't prove yield protection. To mistake this for evidence of yield protection is to presume that schools identically interpret and weigh all factors. Worse, to think this way is to treat the rankings as divine. You seem to want to control for each individual applicant, which is both methodologically mistaken and empirically impossible.

oh wait, your index for UVA and Penn 60/65% gpa 40/35% lsat is totally backwards what the hell, that fucks every thing up. Even 35% GPA, is beyond optimistic for gpa people btw. It should probably be somewhere in the 70-75% lsat range.

Where are your priors coming from?

[D. H2Oman]

UVA may waitlist more people across the board than higher-ranked schools, but that's their prerogative and doesn’t spell yield protection (and certainly not ratings cooking). Again, my definition of yield protection is that a student with a higher index is, contra expectation, less likely to be admitted.By examining the shape of the acceptance curve, I’ve purposely controlled for each school's propensity to waitlist.

Fair enough, but that's not a very good definition of yield protection and it certainly not how it's commonly thrown around on TLS. Yield protection on TLS is generally used to explain how certain school (UVA for example!) use factors in admissions decisions that result in high numbers of "objectively" better candidates (who then often go on to accept spots at better schools) being passed over for candidates that are more likely to attend. Sorry for the confusion in the first post I made, i think this is perfectly acceptable for a school to operate this way. However, it does constitute yield protection using the working definition of TLS posters. A school can do this like American does, but just accepting every 163 out there and waitlisting everyone over a specific LSAT number, or a school can use a more sophisticated method like UVA does in which they waitlist massive numbers of the applicant pool and make extensive use of early decision thereby building much of the class.

Do tons of jackasses on here claim YP when they weren't? Yes.
Does UVA practice YP? Yes, no question about it.

[D. H2Oman]

oh wait, your index for UVA and Penn 60/65% gpa 40/35% lsat is totally backwards what the hell, that fucks every thing up. Even 35% GPA, is beyond optimistic for gpa people btw. It should probably be somewhere in the 70-75% lsat range.

Where are your priors coming from?

Do you really not see why it's a problem to make a index ratio that's 65% GPA 35% LSAT? That's not even close to how admissions work. I'm making those numbers up because it's truly impossible to quantify. There's no question though that LSAT vastly outweighs GPA in numbers importance. So your index numbers are pretty blatantly wrong. Like you said you iterated (ie kept rerunning the data to you hit a specific desired result) the curves for each school until you got the steepest slope. That throws everything off, and gives you wacky index ratios, where penn and UVA are much more GPA-friendly than CLS and GULC, which of course isn't really the case.