Friday, May 14, 2010

HYPSM Cross-Admits

.
Here are the data of 202 HYPSM cross-admits from different years.

OverallHYPSM
# of Committed:4452433332
# of Admits1131151229986
Cross-admit yield (%):38.9445.2235.2533.3337.21






Class of 2012HYPSM
# of Committed:131515107
# of Admits3734433025
Cross-admit yield (%):35.1444.1234.8833.3328.00






Class of 2013HYPSM
# of Committed:131615910
# of Admits3134372625
Cross-admit yield (%):41.9447.0640.5434.6240.00






Class of 2014HYPSM
# of Committed:182315119
# of Admits4544483624
Cross-admit yield (%):40.0052.2731.2530.5637.50





How many cross-admits are among HYPSM? There should be about 1400, a much smaller number than 2000 of what I initially estimated. A mistake I made when I used Stanford’s numbers -- I did not realize that Stanford not only lost to the rest of HYPSM, but also to other schools in certain percentage that can not be neglected. Class of 2012 was the first class that HP removed their EA, and hence H’s yield dropped to 76% from 80+%, and P’s yield dropped to 59% from 65%. Both lost more to other HYPSM, especially to YS, than before the EA was removed.

For class of 2014, H’s initial yield before taking people from the waitlist was 76+%, or H lost about 2110*(1-76%) = 506. On the CC match game of the class of 2014, H lost 45-18 = 27, or 27/506 = 5.3% of total loss at H. I assume that H did not lose too many to schools other than YPSM (can not say the same for YPSM, especially M). Reasonable assumption would be less than 10% of the total loss. So, there were about 460 loss of cross-admits to YPSM. For three years, H won about 40% of the cross admits, or H admitted 460/60% = 767 cross-admits. This also represented 45/76 = 59% of the total number of cross-admits (H admitted 45 / 76 for class of 2014, and 113 / 202 of past three years. So, I am guessing that H admits 55%-60% each year.) Now the total would be 767 / (59% to 55%) = 1300 to 1400.

The numbers for HYP from CC are more reasonably unbiased than those for SM, as the HYP boards are about equally active. There were no obvious biased observations against any of those 3 schools. The testing ratios are in 30-50% and sample size of 5.3% of the population, which make the common misconception of H winning Y by 70% impossible.

For class of 2014, Stanford admitted 2300, and the initial yield was 72+%, that means it lost about 2300(1-72%) = 644. From the three-year data, Stanford admitted 36 / 76 = 51% of the total cross-admits (class of 2014) to 33% of 99 / 202 = 49% (three years combined). So roughly 50% of the total cross-admits – 1400*50% = 700. The yield of the cross-admits is roughly 11 / 36 = 30% (class of 2014) to 33 / 99 = 33% (three years combined), or 700*33% = 231. A loss of 700 - 231 = 469 to HYPM. The rest of the losses 644 – 469 = 175 was to other schools, a relative large number neglected in the original analysis. The number is also matched with the Stanford's report for class of 2102 that for about 640 Stanford lost -- 173 to H, 117 to P, 80 to Yale and 70 to MIT -- 440 total to HYPM, hence 640 - 440 = 200 to schools other than HYPM.

For class of 2014, Princeton was aggressive as it could to fight for the cross-admits, it even took the ones admitted at Stanford or Yale from its waitlist. It admitted 2148. Roughly 1400*48 / 76 = 884 were HYPSM cross-admits with cross-admit yield of 31%, or 274 committed to come. It lost 884 – 274 = 610 HYPSM cross-admits. A reasonable loss of 150 – 200 to all other schools, like what Stanford did, was likely. This totaled to about 800 losses – a 62% yield – little higher than what it initially reported. This means that either the number of cross-admits was little higher or P lost more to other schools.

At this time, there are no reports for Yale, so I am venturing to guess those numbers. For class 2014, Y admitted 1940. Of those admitted, 1400*44 / 76 = 810 were HYPSM cross-admits, and its cross-admit yield was 52%, or 421 committed, i.e., it lost 810 – 421 =389 to other HYPSM. Plus about 200 to all other schools, it lost about 589, or with (1940 – 589) / 1940 = 69% yield, the actual yield could be a little higher.

MIT’s data should be separated from the rest of HYPSM since it admitted much smaller portion of the cross-admits, and for class of 2014, it had initial yield of 64%. This implies MIT lost much more to schools other than HYPS.

All those numbers are estimates for computation purposes. They serve no purpose to be exact.
.
Before I explain why H did not get the 70% of cross-admits, I need to make some reasonable assumptions that both of us have to agree. Otherwise, we will have no common language to base on.
.
The first assumption is that H did not admit the entire HYPSM cross-admits.
.
I am more familiar with class of 2012 than the other classes, so I will use this class to make the point.Personally, I knew 5 H cross-admits, two went to Y. two YS cross-admits, both went to S. For 7 HYPSM cross-admits, H did not get two and only got 3. There should be no particular reasons that H did not want those cross-admits, the number of cross-admits I believe is about the size of one class at H. The famous Dalton school did not get a single student into H. I am sure there were many HYPSM cross-admits, so H did not get the entire cross-admits.
.
.
Total 59 got in H from New Jersey, more than 10 from the schools with average SAT scores below 1540. Most of the kids in those schools do not know 10-4+6 is.Again, H got enough of those non HYPSM cross-admits.
.
The second assumption is that the Stanford report is accurate and the results do not change drastically from year to year.

.
The second assumption is that the Stanford report is accurate and the results do not change drastically from year to year.
.

The most important info comes from the Stanford's report (refer to MinRpt from now):
.
.

It disclosed too much information about the cross-admits, as they may not even realize what they did.

My following analysis may show how Stanford accepts its students, it is by no means to criticize Stanford’s admission practice, as I believe Stanford will be surpassing Harvard in a very short time period to become the most selective university in the States.

General facts about Stanford Class of 2012:
Admits: 2400
Matriculates -- 1703
Legacies – 345
First in their families (FITF)– 431
Admit rate: 9.49%
Yield: 70.95%

I could not locate the report for the legacies and first in their families, but I believe those numbers are accurate as they were in my original analysis post.

From the MinRpt, Stanford lost to
Harvard -- 27%
Yale – 12.5%
Princeton – 18.2%
MIT – 11%

There were 160 Stanford /Yale cross-admits, 80 went to Stanford, and 80 went to Yale.

This means that there were about 80/12.5% = 640 total non-matriculates. This number contradicts the result from 2400*(1-70.95%) = 697 non-matriculates. This could be due to the MinRpt’s numbers were not final when they were produced. Nevertheless, it is best information I can get. I will stick with 640 non-matriculates for the rest of analysis.

Of the 640 non-matriculates:
173 -- > Harvard
80 -- > Yale
116 -- > Princeton
70 -- > MIT

Or total of about 439 HYPSM cross-admits lost. The CC data showed that Stanford got on average 30% -33% of its cross-admits and admitted 50% of entire cross-admits, this does not mean however others won 70% as it showed on the table. This also implies that there were about (439/67%)/50% = 1310 HYPSM cross-admits in total. Once again, this shows a similar result by using the data from Harvard’s class of 2014. The number could be less, but let us assume that the entire HYPSM cross-admits are between 1300 ~ 1500 every year.

Now I can conclude this way: it is not possible that Harvard won 70% of HYPSM cross-admits for class of 2014 (assume the initial yield of 76%):

If it does, and let the total cross-admits be X, and Harvard did not lose 10% of the non – HYPSM cross-admits, so 70%X+90%*(2110 – X) = 2110*76%, then X = 1480. That means Harvard had to admit ALL HYPSM – cross-admits to make the 76% yield – this contradicts the first assumption.

If Harvard admitted fewer than ALL HYPSM cross-admits, I believe that it was the case, Harvard’s yield should be higher than 76% for class of 2014. And most likely Harvard admitted less HPYSM cross-admits with less than 70% yield as the CC data suggested.
.
.
Updated -- 2010 - 05-25



19 comments:

NYCFan said...

With this dubious type of nonsense you are hurting the reputation of your blog for probity and reliability as a source for current admissions data.

Mathacle said...

My blog is viewed by people from over 80 countries, I will let other people make the comments. I wish that you could pinpoint where the problems are, instead of commenting with "absurd" or "dubious". Silly comments without details only make me feel that your Harvard college education does not make you understand what I think. You should hate this area without me before, as my blog is the only source to make some people think. I may cost Harvard some damages for doing this, which basically I think is not what you like, but I also did something bad to Stanford. In order to let me to reveal the findings nicely, I need to have some reasons to respect you, which is not happening yet, as we can discuss this in a productive way. I can also go to autoadmit.com to do this. Not a problem.

My posts are really for the people who can understand.

NYCFan said...

"As Harvard prepares to confer degrees on yet another batch of graduates Thursday, academic experts scratch their heads at how this institution maintains its reputational dominance in an era of academic parity. But a marketer would understand the Harvard aura in a nanosecond: It's the ultimate brand, at least in the academic world.

"There isn't any doubt that brand matters and that Harvard is the prestige brand," says Stanley Katz, director of Princeton University's Center for Arts and Cultural Policy Studies. "It's the Gucci of higher education, the most selective place."

Never mind the price tag (upward of $40,000 per year for tuition, room and board), or the fact that guides such as the U.S. News & World Report ranking of colleges and universities say the differences between Harvard and other top-ranked schools are microscopically small. The gulf that separates Harvard from the rest in terms of reputation remains enormous.

"It used to be the case that of students who were admitted to Harvard and Princeton or Harvard and Yale, seven of 10 would choose to go to Harvard," (Princeton's) Katz says. "It may be more now. There is a tendency for the academically best to skew even more to Harvard. We just get our socks beat off in those cases."

http://www.usatoday.com/money/2005-06-06-harvard-usat_x.htm

Mathacle said...

Good. Now, we are talking. Hopefully we can have very unbiased discussions. If you convince me, I will delete this post. But, you need to admit if you are wrong.

First of all, the article was dated in 2005. At that time, H's yield was much higher. I don't have data available, but I am sure that you do.

Second of all, this Princeton guy should not have known about Yale's numbers. Either the reporter or this guy just made it up.

My HYPSM's post is mathematically complicated. I will try to explain in plain language... later. I can mathematically prove to you that it is impossible for H to win 70% cross the board, even just based on this year's yield.

NYCFan said...

The Harvard yield rate has been in the same narrow range for the last 25-30 years, whether with open admissions (as currently) SCEA, or non-binding early action.

Efforts by Yale, Princeton and Stanford - through binding ED and SCEA - to reduce the size of their overlap pool with Harvard and thus raise their respective yield rates, have had little or no impact on the cross-admit rate.

Harvard's longtime dominance in the battle for cross-admits is unlikely to change soon, as the academic pecking order is fairly static. Even slight perceived differences in "prestige", desirability. etc. can have an enormous effect with respect to the cross-admit ratio, given the "winner take all" phenomenon in elite college admissions.

http://www.educause.edu/ir/library/pdf/ffp0001s.pdf

http://www.educause.edu/ir/library/pdf/ffp9901.pdf

Mathacle said...

You actually know much better than I do about Harvard's yields. It was over 80% for class 2010. You should have given me the yields for the past 10 years, instead of quoting another article written by, who knows, people like you.

http://www.thecrimson.com/article/2006/5/12/class-of-2010-yield-rises-to/

Like I said, in order for me to continue on this, I want to see how sincere you are. There is no need to argue with you on how great Harvard is... simply I just don't have time to do so. I will just post what I think is correct, true or false, I will let other people decide.

NYCFan said...

One of your problems, among others, is taking May stories in college papers as gospel with respect to actual, official yield rates at elite colleges.

The highest actual, official yield rate at Harvard in the last 30 years was 79.3% for the Class of 2006.

The actual, official yield rate for any college cannot be known until after, "summer melt", resort to the waitlist, the number of people deferring matriculation a year, etc.

Preliminary ield rate claims trumpeted in April or May should be taken with a grain of salt. This applies whether the school is Harvard, Stanford, Yale, Princeton, Dartmouth or anywhere else.

Mathacle said...

I did not. That is why I want you to supply me the actual yields for the past ten years, as I know that you have them, which I can use to explain. What I said was that H had higher yields before, but I could not know the exact ones. I was just using the crimson's article to show my points.

Mathacle said...

Before I explain why H did not get the 70% of cross-admits, I need to make some reasonable assumptions that both of us have to agree. Otherwise, we will have no common language to base on.

The first assumption is that H did not admit all HYPSM cross-admits.

I am more familiar with class of 2012 than the other classes, so I will use this class to make the point.

Personally, I knew 5 H cross-admits, two went to Y. two YS cross-admits, both went to S. For 7 HYPSM cross-admits, H did not get two and only got 3. There should be no particular reasons that H did not want those cross-admits, the number of cross-admits I believe is about the size of one class at H.

The famous Dalton school did not get a single student into H. I am sure there were many HYPSM cross-admits, so H did not get all the cross-admits.

Private School Rejects

Total 59 got in H from New Jersey, more than 10 from the schools with average SAT scores below 1540. Most of the kids in those schools do not know 10-4+6 is.
Again, H got enough of those non HYPSM cross-admits.
http://mathacle.blogspot.com/2009/03/harvards-class-of-2012-from-new-jersey.html

The most important info comes from the Stanford's report.

....to be continued

Mathacle said...

The second assumption is that the Stanford report is accurate and the results do not change drastically from year to year.

The most important info comes from the Stanford's report (refer to MinRpt from now):

http://news.stanford.edu/news/2008/june18/minutes-061808.html

It disclosed too much information about the cross-admits, as they may not even realize what they did.

My following analysis may show how Stanford accepts its students, it is by no means to criticize Stanford’s admission practice, as I believe Stanford will be surpassing Harvard in a very short time period to become the most selective university in the States.

General facts about Stanford Class of 2012:
Admits: 2400
Matriculates -- 1727
Legacies – 345
First in their families (FITF)– 431
Admit rate: 9.49%
Yield: 70.95%

I could not locate the report for the legacies and first in their families, but I believe those numbers are accurate as they were in my original analysis post.

From the MinRpt, Stanford lost to
Harvard -- 27%
Yale – 12.5%
Princeton – 18.2%
MIT – 11%

There were 160 Stanford /Yale cross-admits, 80 went to Stanford, and 80 went to Yale.

This means that there were about 80/12.5% = 640 total non-matriculates. This number contradicts the result from 2400*(1-70.95%) = 697 non-matriculates. This could be due to the MinRpt’s numbers were not final when they were produced. Nevertheless, it is best information I can get. I will stick with 640 non-matriculates for the rest of analysis.

Of the 640 non-matriculates:
173 -- > Harvard
80 -- > Yale
116 -- > Princeton
70 -- > MIT

Or total of about 439 HYPSM cross-admits lost. The CC data showed that Stanford got on average 30% -33% of its cross-admits, this does not mean however others won 70% as it showed on the table. This also implies that there were about 439/30% = 1463 HYPSM cross-admits in total. The number could be less. But, let us assume that 1500 is roughly total HYPSM cross-admits every year.

Now I can conclude this way: it is not possible that Harvard won 70% of HYPSM cross-admits for class of 2014 (assume the initial yield of 76%):

If it does, and let the total cross-admits be X, and Harvard did not lose 10% of the non – HYPSM cross-admits, so 70%X+90%*(2110 – X) = 2110*76%, then X = 1480. That means Harvard had to admit ALL HYPSM – cross-admits to make the 76% yield – this contradicts the first assumption.

If the Harvard admitted fewer than ALL HYPSM cross-admits, I believe that it was the case, Harvard’s yield should be higher than 76 for class of 2014. And most likely Harvard admitted less HPYSM cross-admits with less than 70% yield as the CC suggested.

NYCFan said...

For the Class of 2012, Stanford had 1,703 matriculants (after "summer melt") not 1,726 as you assert.

http://ucomm.stanford.edu/cds/cds_2008.html#admission

Mathacle said...

Thanks for the enrolling number, but this would not affect the analysis, as I was using the correct 70.95% yield anyway. Still the 640 non-enrolling number could not match 697 obtained by using the yield.

The discrepancy should not make much of difference in the conclusion.

I still believe that H did win over others, but not by 70% overall. And further down the road, it would be the Stanvard show.

Also, it is not my purpose to bash H anyway. Harvard is still Harvard, and only Stanford may have a chance to challenge Harvard in the future. I think that both of us agree on this.

NYCFan said...

I have no opinion as to what the future may bring, but am interested in current admissions facts. Your recent analysis based on a few random decisions by friends of yours, or a small unrepresentative sample of CC posters, has little or no value.

You would be better off sticking to the admissions news story aggregation business, and leaving the analysis to others.

Mathacle said...

Well, those samples representing 5.3% of what total non-matriculates Harvard had for class of 2014. It is significant in sampling size. If you have had any stats knowledge, we would have not had this discussion. I try not to go down on this road further since we have no common language to base on.

The papers you quoted on autoadmit.com are virtually garbage, they were once again produced from the people who did not know anything about the business. I have been there, made many papers on the IEEE. Please don't use those things to prove anything.

Like my Yield/Admit Ratio index which has been quoted by people from other countries, most of the times, it is counting problem, not full-blown stochastic problem. i guess that college admission research can be a smiple undergraduate project at Stanford, but a $$ grant research at Harvard.

When you can count how many teeth a donky has, not too much different for each donky, you don't lock yourself in a room and imagine how many teeth a donky has in general.

NYCFan said...

You make so many unjustified assumptions that it impossible to recount them all.

Just with respect to Harvard, you fail to note that between 30-40 of the anticipated non-matriculants are not "cross-admit losses" but, rather, admits who have deferred matriculation for a year.

Mathacle said...

I did by Harvard's 76% yield for class of 2014. Sure Harvard has done so many "dirty" tricks to practice admissions.

Let us end the discussions on this chapter, if you don't want me to further bash Harvard, because none of things I said or did were good for Harvard.

NYCFan said...

You can "bash Harvard" all you want too; I care not. But your "cross-admit formula" is based on faulty data and unjustified assumptions, and remains worthless.

Mathacle said...

Another day and another worthless comment as the comment #1 you made. See the comment #2 for the reply.

On my "proof", I made a mistake on purposely to see if you could find the problem or not. Obviously it proved that you could not follow my post. So, I proved my comment #2 correct.

Tony Leng said...

NYCFan just shut up. If I were you Mathacle, I'd delete his comments and block him. You'll never convince him.