Many of the schools have multiple, differing entries for ‘Engineering’, so each one may be a specific Engineering major with a bad label making it somewhat meaningless. For example, Berkeley has 4 different entries for ‘Engineering’.
2 Likes
This is indeed what it is. College Scorecard has subsections for engineering, but this data harvest didn’t get that granular. The engineering listed that is very high is certainly EE/CS combined, and not a classic engineering discipline.
1 Like
So this is only useful for degrees like CS and engineering when you look at their methods. THey excluded everyone who had an advanced degree, so all other majors do not appear to have any value since. That is probably why degrees like math and science are not shown to have much value.
That’s not what their data says. I’m not sure what it does say (maybe @Data10 can dig in better than I), but there are 40 programs with an ROI over $1M for Math and Statistics. 
Really not even helpful for engineering at individual schools. You’ll see “engineering” listed multiple times, with no indication of which "engineering’ each set of numbers represents.
1 Like
Other majors the same, they had biological science multiple times too with different results, all terrible though.
1 Like
You’re right. I saw that also.
Putting Columbia as the second worst ROI for math and statistics is a pretty damning red flag. I mean, there is an entire industry in their back yard that cares about that, and is often willing to pay big bucks for it.
1 Like
Are you saying that San Luis Obispo is an inexpensive area?
Less expensive than San Jose.
Oh, I agree! That’s why I said I don’t understand what their data is saying. It isn’t saying though that a math and statistics degree is bad. Au contraire, from the “right” school, it’s gold.
“Less expensive” isn’t the same as saying “inexpensive.”
I’m not sure what metric you’re using to make that “SLO is less expensive” statement, but just googling average rent and apartment size in both locations:
San Jose $2,735 for 887 SF or $3.09 per SF
San Luis Obispo $2,257 for 774 SF or $2.92 per SF.
Negligible IMO. SLO is not a “less expensive” city. The crappy Starbucks coffee ain’t any cheaper in SLO than San Jose. How much is rent is Houston, Pittsburgh, Ann Arbor, Ithaca, etc.?
And SJSU is much more of a commuter school than SLO.
It’s based on College Scorecard reported earnings +additional computations. The ROI table listed the name of the major grouping in College Scorecard, rather than the name of the actual major. You can generally figure out what the actual major is by matching the listed salary on College Scorecard with the major name in the ROI table. For example, they list 3 ROIs for Computer and Information Science for Washington in the table as summarized below:
ROI Table Majors for University of Washington
Computer and Information Sciences – Earnings = $101k, ROI = $1.4m
Computer and Information Sciences – Earnings = $74k, ROI = $0.8m
Computer and Information Sciences – Earnings = $33k, ROI = -$0.3m
College Scorecard shows the following 3 majors in their Computer and Information Sciences grouping. The numbers don’t match up exactly. However, they are close enough to determine what major is being referenced.
College Scorecard Computer + Information Science Majors for University of Washington
Computer Science – Earnings = $100k and 101k
Computer + information Sciences, General: Earnings = $72k and $74k
Computer Software and Media Applications: Earnings = $30k and $36k
As noted in my post above, it’s based on College Scorecard. College Scorecard lists the following median earnings for Columbia. Perhaps the few federal database mathematics majors were dominated by persons working as research assistants or similar. According to College Scorecard, the sample size was 18 kids.
College Scorecard: Columbia
Mathematics: Earnings = $38k and NA (too small sample size in one reference period)
Applied Mathematics: Earnings = $85k and $92k (average = $89k)
ROI Table: Columbia
Mathematics and Statistics: Earnings = $38k, ROI = -0.3m
Mathematics and Statistics: Earnings = $89k, ROI = $1.2m
1 Like
Thanks for heeding the call! Do you know if they factor out graduate students in the salary database. A PhD stipend 2 years out of undergrad would certainly screw up a ROI average.
Looking at the College Scorecard documentation, they started excluding graduate students in 2020. As of 2020, the following groups are excluded:
- Students who did not receive federal FA or otherwise do not appear in federal database
- Students who did not work during measurement year
- Students who were enrolled in school during measurement year
- Students who received a higher credential level
- Students who died
1 Like
That last one was a punch in the gut. I know it can happen but to see it in black and white makes it seem so much more of a real possibility…Wish I could hug my college student right now.
1 Like
The methodology seems quite involved with many calculations. However, this does not mean the result is meaningful or accurate. The ROI calculation has numerous severe limitations, some of which are listed below. In short, such lists can be amusing, but I wouldn’t take them too seriously.
-
Decades long career earnings are extrapolated based on earnings from an often small sample of students 1-2 years out of college. Earnings 1-2 years out of college are often not representative of career trajectory.
-
Assumes chance of graduating is entirely dependent on school attended and not individual student. For example, if student a student chooses to take a full ride scholarship to Alabama over Harvard to save money, the model assumes his chance of graduating college drops from Harvard’s average of 98% to Alabama’s average of 68%, greatly reducing his ROI.
-
Assumes individual student achievement, ability, and similar individual student characteristics have no impact on earnings. I expect results would be quite different, with a control for SAT score or similar.
-
As noted above, the major category grouping is listed in the table, rather than name of actual major. For example, the list might have 5 majors called “engineering” with large variation in listed earnings and ROI, further muddying conclusions.
-
Assumes all students have the same college cost, rather some students paying an $80k/year sticker price and others paying near $0 after FA or scholarship. An inaccurate estimate of particular student’s college costs makes resulting ROI inaccurate for that student.
-
Doesn’t sufficiently control for location of particular students, which is especially relevant for the CS earnings listed in first post. Colleges with the highest CS earnings tend to be ones where the largest portion of CS majors move to high cost of living areas such as Silicon Valley, Seattle, NYC, and Boston after college. Brown is a good example.
Like most lists ranking colleges by outcome, the list assumes that the name of college attended is the primary driver in individual student outcomes rather than individual student characteristics. So long as this assumption exists without sufficient controls for the individual student, the output of the ranking list will be fundamentally inaccurate.
That said, I don’t think the source data is useless. I think the issue is more trying to calculate and rank ROI based on that source data. For example, one might use the College Scorecard source data to estimate which majors are likely to have higher early career earnings and which majors tend to have lower early career earnings, paying attention to the sample size and other conditions. There is a huge variation in typical early career earnings between different majors, including at the most highly selective colleges.
6 Likes
You can’t compare salaries based on university attended, because salaries are based entirely on local markets, especially in the tech world. A $100,000 salary in CA doesn’t mean the same thing as a $100,000 salary in GA. Plus, graduates get jobs out of state, which skews those numbers exponentially. Colleges in small towns rely on out of state recruiters, skewing numbers higher, while a school in TX or GA already have strong tech markets, skewing the numbers down. The only thing they have in common is that they’re ALL entry level jobs.
If you’re going to study CS, please…don’t rely on this. Find a good school you can afford.
1 Like
I don’t think you can completely say this. There are certainly regional variations in employment, but a degree from CMU and Arkansas Tech don’t offer the same employment opportunities.
1 Like