SAM2 wrote:Let me clarify: I'm not particularly focused on proving or disproving that legacy affects admissions. There's lots of research showing it does have an effect, and many colleges freely admit it. How much effect legacy actually has is often debated, and I think it has less impact than many people assume. So maybe that's worth exploring, but it's more efficient to rely on others' research into that topic rather than creating my own.

I think what I was envisioning is looking at the question of whether legacy has a big impact at particular schools. In some respects, it would be a backdoor way of assessing whether or not certain schools have lots of legacies skewing their college results, or instead if the college results seem unaffected (which would suggest legacy plays little role on how those particular schools' students do at college admissions).

I acknowledge the confounding variable problem someone pointed out. If one school's college admissions results do not correlate with academic results, I cannot know whether the lack of correlation is due to legacy effects, or affirmative action, or athletics, or something else. That's makes things harder. I suppose if the correlation is about the same for different schools though, then that would suggest all those other potentially confounding factors are really not very significant as a practical matter, or at least cancel one another out.

Someone already did this. I remember seeing it published somewhere.

3:26 again. Now I remember. Somebody at Princeton has done a study along these lines: http://www.princeton.edu/~tje/files/Opportunity%20Cost%20of%20Admission%20Preferences%20Espenshade%20Chung%20June%202005.pdf. This study adjusted for legacy, race and athletic recruitment. I don't have time to reread the study, so I'm going to have to rely on the handy Wikipedia summary of this same study (by all means double-check this against the actual study). So, according to Wikipedia's summary of the study, the following hooks are equivalent to adding/subtracting SAT points (on the old 1600-point scale):

Blacks: +230
Hispanics: +185
Asians: -50
Recruited athletes: +200
Legacies (children of alumni): +160

Another way to think about the legacy effect is just a pretty straight-forward calculation. If Harvard takes 30% of legacy applicants, and 6% of regular applicants, then a kid who is a legacy has a 5x advantage over a kid who is not a legacy.

In any case, I'm not sure you could readily get the dataset you want without cooperation from schools or colleges. And for that, it would probably help to be a tenured professor at a place like Princeton. You would need many years of data in order to eliminate cohort effects. For example, suppose this year STA's graduating class has 30 legacies at whatever "selective college" criteria you decide on, and Sidwell's graduating class has only 5--but maybe next year this is completely reversed. Plus, as I wrote earlier, the impact of legacy would be clearer if you have a dataset that lets you adjust for race and athletic recruits.

Let me clarify: I'm not particularly focused on proving or disproving that legacy affects admissions. There's lots of research showing it does have an effect, and many colleges freely admit it. How much effect legacy actually has is often debated, and I think it has less impact than many people assume. So maybe that's worth exploring, but it's more efficient to rely on others' research into that topic rather than creating my own.

I think what I was envisioning is looking at the question of whether legacy has a big impact at particular schools. In some respects, it would be a backdoor way of assessing whether or not certain schools have lots of legacies skewing their college results, or instead if the college results seem unaffected (which would suggest legacy plays little role on how those particular schools' students do at college admissions).

I acknowledge the confounding variable problem someone pointed out. If one school's college admissions results do not correlate with academic results, I cannot know whether the lack of correlation is due to legacy effects, or affirmative action, or athletics, or something else. That's makes things harder. I suppose if the correlation is about the same for different schools though, then that would suggest all those other potentially confounding factors are really not very significant as a practical matter, or at least cancel one another out.

It would probably only be instructive for people interested in private schools in the DC area - parents and academics interested in this.

The dynamics of admission change if you have a geographic hook, for example. Also, admissions officers probably judge private school kids by slightly different yardsticks than public school kids. I don't want to start speculation on how differently, but a kid who has been given every advantage (which would include private school kids and also kids at magnets and top area publics) is going to be appreciated for the rigor of their education at the same time they are judged for whether or not they used all their advantages.

If you want to know whether legacy status makes a difference, there's probably no better place to start than with what one Dean of Admissions says about it: http://www.thecrimson.com/article/2011/5/11/admissions-fitzsimmons-legacy-legacies/

Why and to whom would this be useful?

^^ if there were no adjustment for them

I appreciate the interest in this and the impulse to understand what's going on. I't's the sort of thing I love myself! I've done econometrics at different points during my career. I'm also a parent who attended an Ivy whose kid just got accepted at a better, different Ivy, as a non-legacy obviously (and from an area public HS). So, with that introduction....

I think there are a lot of possible variables here, and the risk is that some variables may be capturing more than one thing. So many different factors go into acceptance at a selective college: legacy, GPA, NMSSF status, whether you even apply because you are confident you can afford it without merit aid and/or financial aid will be sufficient, whether your ECs are at a state or national level of excellence, and even the mood of the AD the day she reads the app.

So, a variable for legacy status (yes/no) may be confounded with things like family income. In particular, there is probably a correlation between legacy status and family income, and this probably affects the family's ability to buy one-on-one SAT tutoring, the ability to buy high-quality soccer coaching and participate on a travel team (better EC's and/or athletic recruitment), the decision about whether the family can afford Princeton (whether to even apply), and (sadly) URM status is still catching up with legacy status.

These other factors would inflate or deflate the importance of a single variable called "legacy status" in different ways, if there was no adjustment for them. But it would seem almost impossible for you to get HHI data so you could hold income constant.

I think there's definitely something to the lgegacy argument. Heck, so long as Harvard acknowledges that they accept 30% of legacies vs. 6% of regular applicants (actually probably less), there has to be something there.

But as long as you know this, and you provide the proper caveats with your numbers, I say go for it!

Anonymous wrote:If this is simply an academic exercise, go for it.

If you think it will in any way inform your kid's college acceptance, you are nuts.

Have no fear, I don't do any of this stuff thinking it's a meaningful guide in how I educate/raise my child. It's just a hobby, albeit perhaps a silly one.

Anonymous wrote:I am fairly confident that someone has done this analysis before in a more rigorous manner than what you are proposing, unless you are a statistician or academic.

Yes, good point. I'd like to think someone with access to better data (and likely even insider data) would have done this better than I ever could. Also, since I'm just doing it as a hobby, I'd expect someone who actually gets paid to do this sort of thing would do it better. I'll check Google Scholar before I embark on anything.

The ranking of the 8 Ivy League schools are: 1, 2, 3, 4, 7, 10, 14 and 16. While all are top schools, grouping the schools as "ivies" is of little value since there are 8 schools that are ranked above the lowest Ivy league school. Moreover, I'm not buying that Cornell is better than Vanderbilt, Georgetown, Notre Dame or any school in the top 30 or so.

I am fairly confident that someone has done this analysis before in a more rigorous manner than what you are proposing, unless you are a statistician or academic.

Just because you find an correlation with objective non-legacy factors, doesn't mean having the legacy doesn't help. If you don't have strong GPA/scores, being a legacy won't get you in. Also what about other factors like athletic recruiting.

legacy and minority status are all important in getting into the ivies. if you dont have these factors, then stem achievement can help.
obviously this is on top of good grades and boards.
good grades means 3.7 and above, unweighted
good scores means 750 and above in everything
national awards and sports captain helps
NCS 2013 4 out of 6 harvard acceptances were legacy
2/6 were legacy + minority
this isnt by chance
go for it

If this is simply an academic exercise, go for it.

If you think it will in any way inform your kid's college acceptance, you are nuts.

DCUM, I need your help thinking through an approach to some data crunching. I was thinking this morning about college admissions and the legacy factor. Many people here question whether college admissions from top high schools are driven by a strong academic profile (SAT, GPA, etc), or instead by legacy. I was wondering whether it's feasible to construct a data set that tests the correlation between college admissions and other objective (non-legacy) scores. I have pretty good data on college admissions to most area high schools (about 10 yrs of data), and I have good data on NMSF success (which is a pretty good proxy for academic success) (15 yrs of data), so I should be able to test the correlation, right? If the correlation is strong, then that suggests academic factors are the big drivers to college admissions success. If the correlation is low, then that can give some better sense of whether other factors like legacy or athletics plays a bigger role.

I suspect there is lots of correlation between college admissions and academics, but it's not perfect, because surely there are other factors that affect college admissions. I'm wondering whether the correlation is stronger or weaker for particular schools.

This is all half-baked and might not really work. I need constructive feedback on how best to develop such an analysis, and whether it might show anything useful. Fire away.

Sam2