Colleges where National Merit Scholars go

College Search & Selection
1

<p>Link: <a href="http://goo.gl/M62tc1"&gt;http://goo.gl/M62tc1&lt;/a&gt;&lt;/p>

<p>Every year, the National Merit Scholarship Corporation releases their Annual Report (<a href="http://www.nationalmerit.org/annual_report.pdf"&gt;http://www.nationalmerit.org/annual_report.pdf&lt;/a&gt;). In it (amongst other things), they list colleges National Merit/Achievement Scholars will attend in the fall and differentiate between NMSC/corporate-sponsored National Merit Scholars and school-sponsored National Merit Scholars. The latest version is from 2013, which corresponds with the Class of 2017 and the incoming freshmen of Fall 2013.</p>

<p>I went through and took the number of non-school-sponsored National Merit and Achivement Scholars at each listed college and found the number of first-time/full-time freshmen entering the college that same fall (using Common Data Sets when I could find them, and other reliable sources when I couldn't). I excluded school-sponsored National Merit Scholars because not every school offers such sponsorship and because including them would skew the data even more in favor of schools like OU (second-tier flagships and non-flagship state schools) because they tend to offer more aid more frequently for National Merit Finalists. Beyond that, including them would mean the data doesn't track a coherent set of individuals- students' inclusion in the data set would vary based on whether they went to a school like OU or UTD. Instead, I opted to track the decisions of students who'd already received corporate-sponsored (1029 students) or NMSC-sponsored (2500 students) scholarships to see what colleges they picked. This doesn't correspond solely with college quality, of course, since finances are a big factor- especially when you get full rides + stipends thrown at you from some places. There might also be some stipulations in corporate-sponsored scholarships that would skew the data, but I eliminated all the issues I could.</p>

<p>Then I adjusted the numbers based on the size of the incoming freshman class (full-time, first-time only since intending to enroll full-time is a prerequisite for the National Merit Scholarship). You can see the results in the spreadsheet, which you can sort by %Scholar (default), %Merit (excluding Achievement Scholars), and %Achievement (exluding Merit Scholars). The colleges are ranked on the left, of course, and the columns on the far right also tell you what portion of a school's incoming Scholars consists of Merit Scholars or of Achievement Scholars ("Merit/Scholar (%)" and "Achievement/Scholar (%)").</p>

<p>Unsurprisingly, Harvard comes out on top when you look at all Scholars, Caltech when only at Merit Scholars, and Yale when only at Achievement Scholars. Hope this serves as a small resource in your college search.</p>

2

<p>With respect to school-sponsored NM scholarships, including them likely overcounts those schools due to luring students looking for a low cost school who would otherwise go elsewhere (presumably your reason for excluding them), but excluding them likely undercounts those schools, since some NM scholars receiving school-sponsored scholarships actually would choose them over other schools even without the school-sponsored scholarships</p>

3

<p>@ucbalumnus‌ I’m not sure how that would make them “undercount” it. There’s National Merit Finalists who would (and do) pick the schools they attend without any special financial incentive. Plus the National Merit Scholars (non-sponsored) who attend schools with incentives likely are influenced by that incentive- after all, I don’t think UT Dallas would attract more scholars per student than UT Austin without its significant financial incentives.</p>

<p>So here’s how it looks, visually:</p>

<p>Starting Cohort: National Merit Finalists</p>

<p>Attend Colleges that Don’t Make Them Scholars: Less financial incentive, not counted
Attend Colleges that Make Them Scholars: More financial incentive, counted</p>

<p>Starting Cohort: National Merit Scholars</p>

<p>Attend Colleges that Don’t Make Them Scholars: Less financial incentive, counted
Attend Colleges that Make Them Scholars: More financial incentive, counted</p>

<p>The only instance in which a student’s choice does not influence whether or not they’re counted in the data is the latter. Therefore the only way to start with a stable group of people and look at their choices is by looking at people who are already Scholars because we can’t track the choices of Finalists as a whole since we can only get data for the ones who choose colleges that make them Scholars. So, ultimately, we still factor in the financial incentive for an elite group (according to perception) but maintain a consistent definition of what the group is- that is, one that allows us to look at just the choices of the same starting group.</p>

4

<p>So basically you tracked the wealthy and poor NMS and eliminated the middle class students who don’t qualify for aid?</p>

5

<p>@Mom2aphysicsgeek‌ Not exactly. Everyone who would’ve been a National Merit Scholar regardless of what school they went to was included. I know at least one National Merit Scholar in the middle class squeeze who would’ve been included if this data were from 2014, so that’s not quite how the methodology works.</p>

<p>The reason I excluded the school-sponsored Scholars is because they’re more or less just elevated National Merit Finalists who automatically qualified for those scholarships by choosing the school; the data would end up including only the National Merit Finalists who chose schools where they got automatic scholarships and excluding the National Merit Finalists who chose schools where they didn’t. In other words, if I’d included them, I’d be including some National Merit Finalists in the data- but not all- so it’d be Scholars + some Finalists. I would personally have preferred a data set including all 15,000-odd National Merit Finalists but I can’t find one.</p>

<p>I just did this so we could start with a certain group of X people and track their decisions- including school-sponsored Scholars would mean that we would have ~2000 more people whose membership in the group occurred only as a result of their decision.</p>

<p>Think of it this way: say you want to model the party preference of American moms. You get a data-set listing parties based on how many American moms voted for them. However, suppose that some parties have a broader definition of “American mom” that includes pregnant women. So, it looks like this:</p>

<p>500 moms voted for the Democrats (35 pregnant women)
485 moms voted for the Republicans
30 moms voted for the Libertarians (3 pregnant women)
12 moms voted for the Green Party
1 mom voted for the Socialist Alternative Party (0 pregnant women)</p>

<p>And say you wanted to track the party preference of a coherent, stable group. Would you go with the narrower definition- for which you can get the results for <em>all</em> members? Or would you go with the Democratic/Libertarian/Socialist Alternative definition? Because in that world all 38 pregnant women voted for the Democrats or Libertarians. That’s not an accurate statistic.</p>

<p>Hence my exclusion of Finalists who accepted automatic (or otherwise less strict than the National Merit Scholarship itself) scholarships to become school-sponsored Scholars.</p>

6

<p>

</p>

<p>However, some students who got UT Dallas sponsored NM scholarships would have chosen UT Dallas anyway even if they did not get the UT Dallas sponsored NM scholarships. Removing such scholarship students undercounts the students’ preference for UT Dallas.</p>

<p>There is no way to avoid either overcounting or undercounting here.</p>

7

<p>@ucbalumnus‌ Right, that doesn’t really undercount them though. Because those students come from a different initial cohort- National Merit Finalists. Everyone who’s initially in the National Merit Scholar cohort is counted, so this does track preferences between them.</p>

<p>As I said before, if I included the school-sponsored Scholars, they’d be coming from a different initial group- National Merit Finalists- that I don’t have the full data for. In other words, I’d only have members of the group “National Merit Finalists Whose National Merit Application Didn’t Get Them a Scholarship” that chose schools that made them National Merit Scholars. I wouldn’t have data for the NMFWNMADGTaS who <em>didn’t</em> choose those schools, so the data would make it look as if everyone in that group chose to become a Scholar through school-sponsored means- misrepresenting their preferences.</p>

<p>Everyone in the same initial group that would’ve picked UT Dallas was included. People who would’ve only entered that group <em>by</em> picking UT Dallas were excluded.</p>

<p>UT Dallas/etc. are still overrepresented if you don’t want the data to include financial input (which is impossible, since financial variation occurs between colleges outside of National Merit incentives) because the people who started as NMS would still get the same bonus if they went to UTD over UT.</p>

<p>Yes, it’s tough because there’s multiple sources of bias, <em>but</em> I believe I’m avoiding undercounting- as long as we can both agree that the group we’re tracking is National Merit Scholars and not National Merit Finalists.</p>

<p>That said, since y’all still seem to have issues with it, I’ll spend the 5 minutes needed to include those people even though I think the resulting data would be rather invalid, useless, and misleading. Just to keep people from complaining that I’m excluding the middle class even though I know plenty of people in the middle squeeze (including myself) who would’ve been counted under this methodology if the data from our graduating years was used.</p>

<p>Just to be crystal clear, I’m not excluding UT Dallas National Merit Scholars as a whole. I’m only excluding National Merit <em>Finalists</em> who get included in the Scholar count only because they chose UT Dallas. These people <em>did not</em> become National Merit Scholars by the same criteria as people who got scholarships from the NMSC or from corporations- and I wanted to track a group that’s defined by the same criteria and whose membership does not vary based on its final decisions. The party preference analogy should help. If it doesn’t here’s something more extreme:</p>

<p>I’m tracking ice cream preferences among bears. For some odd reason, a few companies count pandas as bears. So if a panda wants to be counted in the data, they would have to pick one of those companies as their preferred ice cream vendor. Here’s the data:</p>

<p>138 Klondike Bars (99 pandas)
150 Ben & Jerry’s
120 Blue Bell (80 pandas)
100 Dreyers (12 pandas)
40 Nestle
6 Store Brand
2 Breyers (2 pandas)</p>

<p>Here’s the total customer counts for each company:</p>

<p>Klondike Bar: 1000
Ben & Jerry’s: 1050
Blue Bell: 950
Dreyers: 500
Nestle: 600
Store Brand: 5000
Breyers: 750</p>

<p>Klondike could claim that “138 bears prefer us” but that’s not using the same definition of “bear” as Ben & Jerry’s. I omitted school-sponsored Scholars for that reason- to have a consistent definition. I hope that makes sense. And no, removing those 99 pandas doesn’t mean you’re undercounting their preferences because you’re not even looking at their preferences due to lack of sufficient data.</p>

8

<p>Corporate scholarships are often awarded based on reduced criteria. They are not in the same category as the “refereed” NMSC scholarship winners.</p>

9

<p>@BobWallace‌ That’s true. Unfortunately there’s no way for me to exclude them from the dataset. They also create some bias due to geographic/etc. stipulations. However, the data set is a little bit less tainted/more consistent with these 3500 than it would be with all 5000-odd people labeled “National Merit Scholar.”</p>

<p>Actually, corporate-sponsored scholarships are still not as bad (to include in this data set) as school-sponsored ones because at least they’re decision-independent (which matters a whole lot when decisions are what you’re tracking).</p>

10

<p>There is still undercounting. Suppose a given NM Finalist has as his/her first choice a school that gives a school sponsored NM Scholarship, but would have received a NMSC NM Scholarship if s/he chose a school without a school sponsored NM Scholarship. S/he would be counted as a recipient of a school-sponsored NM Scholarship and therefore excluded from your count.</p>

<p>In any case, net price does become part of most students’ choices. There are likely some students (NM Scholar or otherwise) who choose Harvard or Yale because their net price is lower than that of other schools they are admitted to, even if they may prefer some other school if cost is no object.</p>

11

<p>@ucbalumnus‌ You don’t forfeit a National Merit Scholarship if you choose to attend a school that sponsors Natoinal Merit Scholarships. I know the scholarships are given out well before May 1, and there’s certainly a lot of people on there who would’ve been NMS’s regardless of college choice (UTD has 19 of them, which I believe solidifies my case about overcounting because they shouldn’t have a higher %Scholar than UT if you didn’t factor in financial assistance based on NM status).</p>

<p>True, and finances did impact the decisions of the National Merit Scholars in this cohort. After all, 19 of them chose UTD’s full ride + stipend over other colleges where they would’ve only gotten the $2500 or corporate scholarship. And yes, I did say that that’s a consideration. I’m not trying to exclude financial considerations because that’s impossible. All I’m trying to do is track one group instead of one and whatever portion of the other group I get data for.</p>

<p>Btw I decided not to add in the sponsorship data since I would have to add more rows and dig up more than just number of NMS recipients. That’s a bit time consuming because some schools just don’t know how to fill out a Common Data Set’s C1 section properly (I think 5 or so didn’t even put the numbers in the right place- it’s kind of odd when you accept 2500 women but none of them enroll, but that’s what happens when you accidentally list women as part-time men). So that’s a task for someone who feels it’s actually important and wants to invest their time on it.</p>

12

<p>Where are the biggest discrepancies between top schools in the commercial rankings and the schools preferred by NM scholars?</p>

<p>Below are schools in the Forbes top 100 that appear to be “under-appreciated” by National Merit Scholars (or perhaps over-ranked by Forbes):</p>

<p>School … NM Scholar Rank … Forbes Rank … Difference
University of California, Santa Barbara 308 96 212
University of California, Davis 264 99 165
Brandeis University 213 51 162
Lafayette College 190 48 142
University of Rochester 196 61 135
Boston College 149 35 114
Franklin and Marshall College 181 86 95
Colorado College 164 76 88
Grinnell College 151 64 87
Boston University 170 85 85
Mount Holyoke College 166 84 82
George Washington University 158 77 81
Haverford College 123 43 80
University of California, Los Angeles 110 34 76
Lehigh University 137 70 67
University of Washington 121 55 66
Centre College 127 80 47
Santa Clara University 118 72 46
Middlebury College 82 41 41
Tufts University 60 25 35</p>

<p>Below are schools in the National Merit top 100 that appear to be “over-appreciated” by National Merit Scholars (or perhaps under-ranked by Forbes):</p>

<p>School … NM Scholar Rank … Forbes Rank … Difference
New College, Florida 88 316 -228
Northeastern University 36 236 -200
Baylor University 67 235 -168
SUNY, Geneseo 90 250 -160
Hillsdale College 59 196 -137
Illinois Institute of Technology 93 228 -135
Auburn University 95 225 -130
University of Pittsburgh 64 193 -129
Hendrix College 43 158 -115
University of Tulsa 47 152 -105
University of Dallas 75 169 -94
Fordham University 78 163 -85
Lawrence University (WI) 85 167 -82
Denison University 52 130 -78
University of Miami 54 124 -70
St. Olaf College 53 121 -68
Case Western Reserve University 34 89 -55
Occidental College 63 117 -54
Georgia Institute of Technology 32 83 -51
Furman University 96 144 -48
Harvey Mudd College 8 52 -44</p>

13

<p>

</p>

<p>I can tell you one reason F&M is under-appreciated by NMFs. They give no merit scholarships any more. It had been in the back of my mind as a possible safety and/or merit possibility for D2 for a few years, but when they dropped merit scholarships we dropped it from our list. Why should they get my very high stats kid (to beef up their admitted students stats) with no discount in the cost?</p>

14

<p>Really enjoyed your extensive analysis, dividerofzero. Your college is lucky to be getting you (CMU?). Can you do the same analysis of students who got perfect ACT and SAT scores? … although I doubt the raw data is there for you to cull from.</p>

15

<p>@1203southview‌ I wish I had that data. I’m really looking forward to working on my stats (and becoming a better programmer so I can reduce the manual portion of this stuff) at CMU. Thanks.</p>

<p>I’ll look around for more “elite” groups’ data. I want to use it to find “value” colleges that start out with less elite batches but end up creating more positive outcomes, but that’s going to require a bunch more adjustment.</p>

<p>Thanks for doing that, @tk21769‌.</p>

16

<p>Below is a ranking of the 50 colleges that produce the most science and engineering PhDs per capita, relative to the per capita presence of NM Scholars at those schools. That is, I start with the top 50 schools for per capita PhD S&E production. I then measure the difference between that PhD production ranking and the NM Scholar representation ranking. I then sort the list according to that difference. Universities are shown in bold below to distinguish them from LACs. </p>

<p>Schools at the top appear to be over-achievers. They generate many S&E PhDs per capita relative to NM Scholar representation per capita. A small/negative number in the last column may only mean that many of their students choose to pursue other outcomes. Also, for schools like Cal Tech, Harvey Mudd, and MIT it is pretty hard to “over-achieve” in this ranking, since they already are near the top in S&E PhD production. So there’s little or no room to go up. </p>

<p>Note that the top 4 do not appear in dividerofzero’s NM Scholar ranking. I assigned them all a rank of “313” (1 postion lower than the lowest-ranking school in the NM Scholar ranking).</p>

<p>Source:
<a href=“Top 50 Schools That Produce Science PhDs - CBS News”>http://www.cbsnews.com/news/top-50-schools-that-produce-science-phds/&lt;/a&gt;
(If you don’t see a school on this list, it means it was not among the top 50 for alumni per capita S&E PhD production according to this source.)</p>

<p>School…Rank-PhDs…Rank-NM…Difference
Washington College 24 313 289
Beloit College 31 313 282
Trinity University 44 313 269
St. John’s College 46 313 267
Whitman College 48 313 265
Brandeis University 49 213 164
University of Rochester 38 196 158
Franklin and Marshall College 36 181 145
Grinnell College 8 151 143
Bates College 47 163 116
Haverford College 13 123 110
New Mexico Institute of Mining and Technology 15 111 96
Earlham College 35 124 89
Hampshire College 50 136 86
Kalamazoo College 21 92 71
Reed College 4 71 67
Rensselaer Polytechnic Institute 34 94 60
Lawrence University 37 85 48
Macalester College 28 74 46
Bryn Mawr College 12 46 34
Oberlin College 18 49 31
College of William and Mary 45 73 28
Occidental College 41 63 22
Carleton College 6 26 20
Bowdoin College 32 51 19
Swarthmore College 5 22 17
Wesleyan University 26 39 13
Case Western Reserve University 23 34 11
Johns Hopkins University 20 27 7
Harvey Mudd College 2 8 6
University of Chicago 7 12 5
Pomona College 14 19 5
Cornell University 22 25 3
Carnegie Mellon University 27 30 3
Cal Tech 1 3 2
Rice University 9 11 2
Williams College 16 17 1
Hendrix College 42 43 1
MIT 3 2 -1
Wellesley College 33 31 -2
Vassar College 43 40 -3
Princeton University 10 5 -5
Amherst College 29 23 -6
Harvard 11 1 -10
University of California, Berkeley 39 29 -10
Brown University 25 14 -11
Yale University 17 4 -13
Stanford University 19 6 -13
Duke University 30 7 -23
Dartmouth College 40 16 -24</p>

17

<p>Denison places above USN with NMFs because they offer Full Tuition scholarships to them.</p>

18

<p>That’s weird, I swear Davidson has more than 5 National Merit Scholars…</p>

19

<p>@IAmNotBenji‌ They have 8, but 3 of them are only National Merit Scholars because they went to Davidson. So basically the equivalent of a National Merit Finalist at any other university.</p>

20

<p>Ah, so if someone got a National Merit Scholarship from one school but chose to attend another, they wouldn’t be counted in the statistics as a National Merit Scholar at their enrolled school, correct? </p>