How much does where you go to college affect earnings?

Stu­dents who go to more se­lec­tive col­leges make more money later in life. The 2013-2014 PayS­cale Col­lege Salary Re­port gives start­ing salaries of ~$60k for grad­u­ates of Ivy League schools vs. ~$45k/​year for grad­u­ates of mid-tier state schools, and mid-ca­reer salaries of ~$110k/​year for grad­u­ates of Ivy League schools vs. ~$80k/​year for mid-tier state schools, so the differ­ence is about 30%.

This par­tially re­flects stu­dents who go to more se­lec­tive col­leges be­ing more able and am­bi­tious, as op­posed to at­tend­ing a more se­lec­tive col­lege boost­ing one’s in­come. How much does at­tend­ing boost in­come? The fa­mous pa­per Es­ti­mat­ing the Re­turn to Col­lege Selec­tivity Over the Ca­reer Us­ing Ad­minis­tra­tive Earn­ing Data (2011) by Dale and Krueger raises the pos­si­bil­ity that on av­er­age, at­tend­ing a more se­lec­tive col­lege doesn’t raise earn­ings at all. Speci­fi­cally, con­trol­ling for

  • Stu­dent high school GPA

  • Stu­dent SAT score

  • The av­er­age SAT score of the col­leges that a stu­dent ap­plied for1

  • The num­ber of ap­pli­ca­tions that the stu­dent submitted

  • Var­i­ous de­mo­graphic fac­tors (race, sex, parental ed­u­ca­tion)

they found that as a group, there was no statis­ti­cally sig­nifi­cant differ­ence in in­come later in life be­tween stu­dents who went to more se­lec­tive col­leges and stu­dents who went to less se­lec­tive col­leges. Their find­ing is some­what ro­bust: it’s based on a large (~10k) sam­ple size, it’s true both of the class of 1976 and the class of 1989, it’s true of the class of 1976 from age 25 through age 50 and it’s true both of men and of women.2,3

A cou­ple of caveats:

  • Blacks, His­pan­ics, and chil­dren of par­ents who don’t have col­lege de­grees and who at­tended more se­lec­tive schools earned more than those who did not.

  • The 27 uni­ver­si­ties that the stu­dents were drawn from are highly se­lec­tive: if one were to look at less se­lec­tive uni­ver­si­ties, one might get differ­ent re­sults.

Dale and Krueger’s cen­tral find­ing has been taken to be ev­i­dence that go­ing to a more se­lec­tive col­lege does not in­crease one’s earn­ings, con­trary to con­ven­tional wis­dom.

There are many fac­tors that could give rise to a differ­ence in in­come be­tween those who at­tended more se­lec­tive col­leges, some of which fa­vor those who at­tended less se­lec­tive col­leges, and could coun­ter­bal­ance oth­ers that in­crease earn­ings. I list some po­ten­tial con­tribut­ing fac­tors be­low (some of which were pointed out by Dale and Krueger). In each case, whether these effects are pre­sent is un­clear, and to the ex­tent that they are there, the over­all effect of them is some­times un­clear, but they could give rise to a differ­ence be­tween the two groups.

Selec­tion effects

Th­ese have no bear­ing on whether go­ing to a more se­lec­tive col­lege in­creases ex­pected earn­ings.

Cost

The cost of at­tend­ing a more se­lec­tive col­lege could (af­ter tak­ing a stu­dent’s abil­ity and merit-based fi­nan­cial aid into ac­count) be higher or lower than the cost of at­tend­ing a se­lec­tive col­lege. So the choice of at­tend­ing a more se­lec­tive school could re­flect hav­ing a fam­ily that’s will­ing to pay more for col­lege (and/​or the will­ing­ness of the stu­dent for his or her fam­ily to pay).

Fu­ture ca­reer plans

Stu­dents who choose to at­tend more se­lec­tive schools could be more or less likely to go into academia than stu­dents who don’t. Aca­demics make less money than other peo­ple do (af­ter con­trol­ling for fac­tors such as GPA, SAT scores and con­scien­tious­ness). Similarly, those who at­tend more se­lec­tive schools may have other ca­reer prefer­ences that feed into ex­pected earnings

For stu­dents who go on to pro­fes­sional school (law, medicine or busi­ness), col­lege at­tended doesn’t mat­ter as much for later life prospects. So at­tend­ing a more se­lec­tive col­lege could re­flect lower in­tent to go on to pro­fes­sional school. The earn­ings of peo­ple who go onto pro­fes­sional school are gen­er­ally higher than those of peo­ple who don’t – this points in the di­rec­tion of ex­pected earn­ings be­ing higher for those who go to less se­lec­tive col­leges.

Con­cern for prestige

Stu­dents who choose to at­tend more se­lec­tive col­leges plau­si­bly care more about pres­tige. This de­sire for pres­tige could cor­re­spond to more de­sire to make money later in life – this points in the di­rec­tion of ex­pected earn­ings be­ing higher for those who go to more se­lec­tive col­leges.

Signaling

The im­pact of pres­tige of col­lege at­tended on hiring decisions

Ac­cord­ing to re­search by Lau­ren Rivera, high pay­ing elite pro­fes­sional ser­vice firms (in­vest­ment banks, law firms, and man­age­ment con­sult­ing firms) give a lot of weight to whether job ap­pli­cants at­tended top four uni­ver­si­ties when mak­ing their hiring de­ci­sions.

More broadly, em­ploy­ers give weight to the pres­tige of col­lege at­tended. How­ever, the effect size is smaller than it might seem. In a 2013 Gal­lup Poll, 9% of busi­ness lead­ers said that the col­lege a job ap­pli­cant at­tended is “very im­por­tant” to man­agers mak­ing hiring de­ci­sions, and 37% said that it’s “some­what im­por­tant.” Notably,

(i) em­ploy­ers listed col­lege at­tended as the least im­por­tant of the 4 fac­tors that they were asked about

(ii) the Amer­i­can pub­lic gives more cre­dence to em­ploy­ers weight­ing col­lege at­tended when mak­ing hiring de­ci­sions: 30% think it’s “very im­por­tant” and 50% think it’s “some­what im­por­tant.”

(The 2013 data is much more re­cent than the Dale-Krueger data, but still pro­vides rele­vant ev­i­dence.)

Once one is hired, where one went to col­lege won’t play a role in pro­fes­sional ad­vance­ment at that firm (ex­cept to the ex­tent that it built rele­vant skills).

In­fluence on grades

Carl Shul­man sug­gested that go­ing to a more se­lec­tive school re­duces one’s ex­pected GPA, be­cause of higher grad­ing stan­dards. It’s un­clear to me whether this is true, and the effect could even cut the other way, but there’s plau­si­bly an effect in some di­rec­tion. Re­duced GPA re­duces one’s prospects for get­ting into med­i­cal or law school. Perus­ing fo­rums for ap­pli­cants, one finds many peo­ple say­ing that GPA is a dom­i­nant fac­tor in ad­mis­sions, one that’s far more im­por­tant than pres­tige of col­lege at­tended.

In­fluence on ma­jor choice

At­tend­ing a more se­lec­tive school re­duces one’s rel­a­tive stand­ing amongst stu­dents in a given ma­jor. This can lead to stu­dents at more se­lec­tive schools choos­ing a less de­mand­ing ma­jor than they oth­er­wise would have (whether be­cause the ma­jor re­quires more effort than it would have, be­cause one finds it in­tol­er­able to be one of the weaker stu­dents ma­jor­ing in a given sub­ject, or for some other rea­son). In the 2013 Gal­lup Poll cited above, busi­ness lead­ers sur­veyed listed the sub­ject that a stu­dent ma­jored in as sig­nifi­cantly more im­por­tant than col­lege at­tended, in the con­text of hiring de­ci­sions.

Treat­ment effects

Peer group

Hav­ing a more ca­pa­ble peer group can lead to bet­ter learn­ing op­por­tu­ni­ties, and higher ex­pected earn­ings. Ben Kuhn wrote

By watch­ing how more com­pe­tent peo­ple work and think, you can of­ten pick up use­ful study habits and bet­ter tech­niques for the sub­ject you’re study­ing. I’ve found this es­pe­cially true in CS classes, where I’ve had this ex­pe­rience from both sides, e.g. teach­ing class­mates how to use Git and pick­ing up C cod­ing style and tricks from bet­ter pro­gram­mers.

It can also give one ac­cess to bet­ter ad­vice: Ben Kuhn also wrote:

Know­ing tal­ented stu­dents has given me info about sev­eral ex­cel­lent courses, as well as sum­mer op­por­tu­ni­ties, I wouldn’t oth­er­wise have known about.

Th­ese con­sid­er­a­tions gen­er­ally fa­vor more se­lec­tive schools, but not as strongly as might meet the eye: less se­lec­tive schools of­ten have hon­ors courses and hon­ors pro­grams, where one might be able to meet stu­dents as ca­pa­ble as those who one would be in­ter­act­ing with at less se­lec­tive col­leges (though the best stu­dents at more se­lec­tive col­leges will gen­er­ally be stronger than the best stu­dents at less se­lec­tive col­leges).

Net­work­ing benefits

Go­ing to a more se­lec­tive col­lege will gen­er­ally ex­pose one to peo­ple who will be in higher places later on in life, and who will cor­re­spond­ingly be able to con­nect one with in­fluen­tial peo­ple in one’s pro­fes­sional field, who may get one a high pay­ing job and so forth. Such peo­ple may also serve as pro­fes­sional col­lab­o­ra­tors, for ex­am­ple, if one wants to do a startup right out of col­lege.

As above, the effect here is smaller than might ini­tially meet the eye, be­cause one might be able to get similar benefits by in­ter­act­ing with the most ca­pa­ble stu­dents at a less se­lec­tive col­lege.

Confidence

Some peo­ple have sug­gested that be­ing a “small fish in a big pond” re­duces stu­dents’ con­fi­dence rel­a­tive to be­ing a “big fish in a small pond.” As­sum­ing this, to the ex­tent that con­fi­dence in­crease later life earn­ings, all else be­ing equal, at­tend­ing a more se­lec­tive school will re­duce ex­pected earn­ings.

Bet­ter learn­ing due to at­ten­tion from professors

Some peo­ple have found that be­ing a “big fish in a small pond” is con­ducive to get­ting more at­ten­tion from the pro­fes­sors in one’s classes, on ac­count of stand­ing out. This can in­crease the amount that a stu­dent learns, be­cause of pro­fes­sors’ greater will­ing­ness to spend time on per­son­al­ized in­struc­tion.

Self-sufficiency

Some peo­ple who I know re­port to hav­ing a sub­jec­tive sense that be­ing in a less elite en­vi­ron­ment was helpful to them, be­cause it forced them to be in­de­pen­dent (on ac­count of be­ing differ­ent from their peers), whereas had they been in an elite en­vi­ron­ment, they would have “gone with the flow” and un­crit­i­cally made their de­ci­sions based on what their peers were do­ing. All else be­ing equal, this fac­tor would in­crease ex­pected earn­ings of those who at­tend less se­lec­tive col­leges.

In­fluence on ma­jor choice

As above, at­tend­ing a more se­lec­tive col­lege can re­duce the prob­a­bil­ity that one will ma­jor in a de­mand­ing sub­ject. Aside from hav­ing sig­nal­ing im­pli­ca­tions, this also re­duces the chances that one will ac­quire tech­ni­cal skills re­quired for higher pay­ing jobs.

In­fluence on ca­reer choice

Go­ing to a more se­lec­tive col­lege could nudge one to­ward or away from academia, go­ing into the non-profit world, get­ting a pro­fes­sional de­gree, etc. Earn­ings vary across these fields, and this could give rise to a differ­ence be­tween the groups.

Implications

Be­fore think­ing se­ri­ously about the pa­per by Dale-Krueger, I had sub­scribed to the con­ven­tional wis­dom that go­ing to a more se­lec­tive col­lege gen­er­ally boosts ex­pected earn­ings. Think­ing it over more care­fully, it’s now gen­uinely un­clear to me whether this is the case, and it could re­duce ex­pected earn­ings in gen­eral.

Rather than tak­ing the Dale-Krueger find­ing to be defini­tive, one should give some weight to con­ven­tional wis­dom, on the grounds that the pa­per might have hid­den method­olog­i­cal er­rors, or have ceased to be rele­vant in the pre­sent day. But in view of

  1. The Dale-Krueger finding

  2. The fact that there are a num­ber of ways in which go­ing to a more se­lec­tive col­lege could de­crease earnings

  3. The fact that and the fact that the salient ad­van­tages of go­ing to a more se­lec­tive col­lege are less sig­nifi­cant than they might ini­tially appear

  4. The fact that con­ven­tional wis­dom is at least par­tially rooted in a con­fla­tion of cor­re­la­tion and causation

it seems rea­son­able to adopt a “best guess” that if go­ing to a more se­lec­tive col­lege does in­crease ex­pected earn­ings, the effect size isn’t high.

What im­pli­ca­tions does this have for stu­dents who are try­ing to de­cide what col­lege to go to, or how much to fo­cus on get­ting into col­lege? First some gen­eral con­sid­er­a­tions:

  • Rather than tak­ing the Dale-Krueger find­ing to be defini­tive, one should give some weight to con­ven­tional wis­dom, on the grounds that the pa­per might have hid­den method­olog­i­cal er­rors, or have ceased to be rele­vant in the pre­sent day. But in view of (i) the Dale-Krueger find­ing (ii) the fact that there are many ways in which go­ing to a more se­lec­tive col­lege could de­crease earn­ings (iii) the fact that and the fact that con­ven­tional wis­dom is at least par­tially rooted in a con­fla­tion of cor­re­la­tion and cau­sa­tion, it seems rea­son­able to adopt a “best guess” that if go­ing to a more se­lec­tive col­lege does in­crease ex­pected earn­ings, the effect size isn’t high (per­haps on the or­der of ~2% – this is an in­tu­itive guess based on the data given in the Dale-Krueger pa­per).

  • Even if the causal effects lead to there be­ing no differ­ence in ex­pected earn­ings on av­er­age and at the mo­ment, that doesn’t mean that there’s no differ­ence for a ra­tio­nal ac­tor who wants to max­i­mize ex­pected earn­ings. For ex­am­ple, if at­tend­ing a more se­lec­tive (or less se­lec­tive) col­lege were to re­duce ex­pected earn­ings on ac­count of in­creas­ing the prob­a­bil­ity that stu­dents go into academia, if one doesn’t want one’s earn­ings to be re­duced, one can sim­ply buck the trend and not go into academia. How­ever, there are cer­tain ca­sual effects over which one doesn’t have con­trol (such as the fact that elite firms look more fa­vor­ably on stu­dents who grad­u­ated from top 4 col­leges)

  • The ca­sual effects will vary from per­son to per­son. For ex­am­ple, go­ing to a col­lege that elite fi­nance firms look on fa­vor­ably will have an effect for those who as­pire to go into fi­nance to a greater de­gree that it won’t for those who don’t as­pire to go into fi­nance. A high school stu­dent can try to as­sess which causal effects will ap­ply to him or her. But it can be hard to tell ahead of time: be­fore en­ter­ing col­lege, one might be un­clear on (or mis­take about) whether one wants to go into fi­nance.

  • Ex­pected earn­ings is not the only met­ric of the value of go­ing to col­lege. There’s also the con­sump­tive ex­pe­rience, as well as well as its im­pact on ca­reer suc­cess that’s not re­flected in earn­ings. It’s plau­si­ble that go­ing to a more se­lec­tive col­lege mat­ters more for pro­fes­sional suc­cess for peo­ple who will be go­ing on to academia than it does in gen­eral.

Ir­re­spec­tive of the con­sid­er­a­tions above:

  • The Dale-Krueger find­ing gen­er­ally ar­gues points in the di­rec­tion of giv­ing greater weight to cost differ­ences be­tween col­leges than to differ­ences in pres­tige. Depend­ing on one’s fam­ily’s in­come and sav­ings, go­ing to Har­vard can be cheaper than go­ing to Berkeley, but for other fam­i­lies there will be a differ­ence be­tween the two on the or­der of $100k (which if in­vested for 30 years would grow to ~$800k).

  • The Dale-Krueger find­ing points in the di­rec­tion of effort spent get­ting into col­lege be­ing less cost-effec­tive than most peo­ple think, which should shift one in the di­rec­tion of spend­ing less effort on it and more effort on build­ing em­ploy­able skills, en­joy­ing life, and con­tribut­ing so­cial value.

Footnotes

[1] You might won­der why the au­thors didn’t con­trol for the av­er­age SAT score of the col­leges at which the stu­dents were ac­cepted. The au­thors did some­thing like this (ac­tu­ally, com­par­ing stu­dents who had been ac­cepted at the ex­act same set of col­leges) in a 2002 pa­per, and ob­tained similar re­sults. From the pa­per: “The matched ap­pli­cant model and self-rev­e­la­tion model yielded co­effi­cients that were similar in size, but the self-rev­e­la­tion model yielded smaller stan­dard er­rors. Be­cause of the smaller sam­ple size in the pre­sent anal­y­sis, we there­fore fo­cus on the self-rev­e­la­tion model.”

[2] The au­thors say this in the text of the pa­per, but when I look at Table 4 of the pa­per on page 32 of the PDF, the data seems to in­di­cate that for women from the 1976 co­hort bro­ken down into age groups, there is a statis­ti­cally sig­nifi­cant differ­ence, in fa­vor of those who at­tended less se­lec­tive schools. But I as­sume that I’m mis­in­ter­pret­ing the table: I’d wel­come any help in­ter­pret­ing the data on this point.

[3] The au­thors qual­ify this by say­ing “The es­ti­mates from the se­lec­tion-ad­justed mod­els are im­pre­cise, es­pe­cially for the 1989 co­hort. Thus, even though the point-es­ti­mates for the re­turn to school qual­ity are close to zero, the up­per-bound of the 95 per­cent con­fi­dence in­ter­vals for these es­ti­mates are some­times size­able.”

Cross-posted from the Cog­nito Men­tor­ing blog