I don’t know what makes a ROI list “most authoritative.” Whether Caltech has a top ROI largely depends on whether you control for Caltech having top students, who generally choose high paying majors, work in an expensive section of the country, come from wealthy families, and and are not Black. If you control for these factors, then Caltech and most of the colleges ranked well in the list above have a far more mediocre ROI. That is particular students with the characteristics above tend to have high earnings wherever they attend college, not just ones who choose to attend Caltech or similar.
For example, The Economist college rankings are based purely on expected earnings vs actual earnings. They use the same CollegeScorecard earnings source data as the Georgetown ROI rankings above, yet their ranking order is very different, as listed below.
The Economist College Rankings (Actual Earnings - Predicted Earnings)
1 . Washington and Lee (+$22k)
2. Babson (+$20k)
3. Villanova (+$13k)
4. Harvard (+$13k)
5. Bentley (+$13k)
6. Otis Art & Design ($+12k)
7. Lehigh (+12k)
8. Alderson (+$12k)
9. Texas A&M (+11k)
10. Cal State Bakersfield (+$11k)
…
1260/1275. Caltech (-$8k)
The Economist rated Caltech as 1260th out of 1275 colleges, putting Caltech in the bottom 1% on earnings and similarly abysmally low on ROI. Caltech had the highest average SAT scores among all 1275 colleges, a high percentage tech majors, etc.; so Caltech is expected to have extremely high average earnings. The Economist expected Caltech to have the 2nd highest average earnings among all 1275 colleges… only 2nd to MIT. However, Caltech’s reported earnings were 19th out of 1275 – below tech colleges like MIT, Mudd, Stevens, RPI, Rose Hulman, WPI, and CO Mines. Caltech’s average earnings were certainly high, but they were $8k below expectations, and earnings below expectations results in a low ranking.
The Economist college ranking methodology also has flaws, as is evident in the extreme differences between certain similar colleges such as Harvard in 99th percentile ROI vs Yale in 0th percentile ROI. I think the Economist earnings prediction is better than most due to the increased controls, but this increased accuracy of earnings prediction can make relatively small differences from the earnings prediction have a huge impact on rankings. For example, if Caltech’s very small sample of earnings for federal FA recipients happened to be off by $10k from actual, and this error was corrected, then Caltech’s ranking would skyrocket from 1st percentile to ~77th percentile in the year the error was corrected. So colleges for which earnings happen to be inaccurate on the positive side have huge jumps in rankings, and colleges for which earnings happen to be inaccurate on the negative side have huge declines in ratings. You need more precise earnings to evaluate this degree of small differences from earnings predictions, or the results will be dominated by random noise.
If you could somehow have perfectly accurate earnings information and ideal controls, then I suspect the conclusion would be that earnings primarily depends on student characteristics – not the name of the college attended. If a kid is accepted to highly selective college X and chooses to attend not as selective college Y, that particular student’s earnings potential is likely largely unchanged; even though graduates from college X and college Y may have very different average earnings. This is essentially what the classic Dale and Krueger studies found, as did other similar ones. However, college ROI rankings usually assume earnings are entirely a function of college attended, with no dependence on whether the individual student excels academically or struggles academically, is interested in high paying major vs low paying major, comes from a wealthy family vs comes from a family that struggle financially, etc So long as this assumption exists, the ROI calculation will be severely flawed.