@Cavitee . . . a lot of the problem with outcomes’ rankings is that they always refer to Payscale.
Payscale is a wonderful service/site that gives incredible in-depth salary information for specific job description, company, geographic region, and industry. I’m not sure if linkedin – another great service, social-media based – actively assents to the use of its data for job placement wrt the colleges, but there are certainly some rankings that use it to try to head count those who are placed within, say, Oracle by college using linkedin, which is faulty. Payscale, however, does actively sell its services to college-ranking services with specificity to those which are outcome driven.
Both Forbes’ and WSJ’s lists and a few others contract with Payscale which purports to have accurate information of salaries of the graduates of specific colleges with both median and mean at certain points of their careers, which is where it gets into trouble.
Payscale has, though, improved its services in general, and in fact probably helps to drive salaries higher because it is disseminating these data more readily and helps to create more salary-competitive industry. And by overstating what the mean and median salaries of all colleges’ graduates are – one of its problems – it might be helping to drive salaries higher because Payscale is so readily referenced.
Before, it would only count salary data of those with baccalaureate degrees to compute the median/mean salary data for each college to find the truer effect of a college’s reputation in enabling its graduates to land high-paying jobs. Now it has finally included the salaries of those with higher degrees and associates them with the persons’ undergraduate colleges, which is certainly more the individuals’ rather than their alma maters’ accomplishment which it was trying to negate.
The essence of the problem with Payscale wrt to associating salaries to specific colleges is its sampling method. There is a selection bias, because it only sees the top side of salaries of college graduates with its passive way of collecting their salary data and then associating them to the individual colleges, because the ones who will partake will seemingly, intuitively be those at the higher end.
Payscale will state that it has, say, 30,000 survey-takers of Ohio State’s alumni base. OSU has ~ 500,000 undergraduate alumni, so 30k/500k is an admirable 6%. However, these are persons who’ve self-selected themselves (the oft-used compound word on these boards) to openly display their salary history, and additionally, unless they send in their tax returns, the data could also be inflated.
The problem here lies in its passive collection of data without regard to scientific sampling methods. And Payscale does consider it to be a sample size, and it will boast how many millions it has on file. However, there is no regard to major, to geographic or demographic background, i.e., to actively obtain a representable makeup of OSU’s alumni base to determine what the median/mean salary are at various stages of their careers. It probably just takes the average of those within the 30k and reports this at, say, the beginning of and at 10 years into their careers.
Since Forbes’ ranking gives a more informational database, I’ll refer to it as an example of outcome bias. This bias is seen in the its ranking of the 450 colleges out of, I think it said, 4,400 four-years in the nation, and describes them as being the top tier of all colleges.
In its outcomes for the 450, there was a mean of $56,508 starting salary, with a standard deviation of $7,369. That’s incredibly high, probably too high and too well grouped together as is a complaint in this thread.
And the average indebtedness including the service academies is $7,590 with an SD of $1,896, which manifests and even greater cookie-cutter group. One could only hope this is true with TCU having the highest indebtedness at $14,693 which is still pretty reasonable – in addition to their full payers who wouldn’t have any debt at all of course.
It appears that the median salary of the 450 colleges’ average salary is ~ $100,000 at ten years into their careers. That would be a top 10% wage-earner in the US. And as others have stated here for the WSJ rankings, there isn’t a lot of separation in the salaries, though I don’t have time to compute it because it would mean opening up all the hyperlinks for the colleges.
With respect to a probable error in comparison between two colleges, I’ll reference UCSD and UCLA. The Forbes’ rankings database manifests an of outcome for the two is as follows:
Salaries…UCSD…UCLA
Starting…$63,400…$62,000
10-years In…$123,700…$116,100
Certainly there is a disparity in WSJ’s and Forbes’ rankings in these figures.
But I posit that UCLA should have a higher salary at 10 years in than UCSD’s because:
– UCLA produces more attorneys compared to UCSD – there are 23,200 licensed barristers in CA for the former, and 6,600 for the latter;
– UCLA sends more undergrads to med school than UCSD / year – ~ 500 to about 250.
– UCLA students are more business-and-commerce inclined than UCSD’s, which is fairly evident.
The mitigating factor to the above is that UCSD’s students are certainly overall more STEM-inclined than UCLA’s. This would probably imply a greater starting salary for the former, but UCLA’s should surpass SD’s within or at the 10 years because of the above.
And indeed the NYT’s database of salaries manifests this with its median salaries at the age of 34 and above, with their salaries being comparable. According to this database, UCLA has 8.2% of one-percenters and SD has 6.3%. And this database would be more scientifically sound than Payscale’s because there was no willful self-selecting by the NYT in collecting its data.
Additionally, if one doubles the salary of the NYT’s medians, then one comes close to matching Payscale’s average for each college, with adjustments to the NYT’s database as women having children and quitting work or having maternity leave. I don’t think this is feasible. The average will certainly be materially and maybe considerably higher, but I don’t think it would be about twice as much.