[wenc]

There's a significant tendency on HN to unduly discount good grades. I think this is a mistaken tendency due to univariate, linear thinking. Grades are not everything, but they are not nothing.

The fact is, determining a good hire requires multivariate, nonlinear thinking.

Good grades can be a proxy for metaskills like discipline, cognitive ability, etc. They don't always measure these things perfectly, but the correlation is not negligible.

Doesn't mean say, a 2.5 GPA isn't a good hire -- but in multivariate thinking, there has be other factors that compensate for the low GPA. Otherwise you'd be hiring a 2.5 GPA who is truly mediocre, and my experience is that the majority of 2.5 GPAs are that. Not everyone with a low GPA is pursuing other interests or passions. Also, doesn't mean that everyone with a high GPA isn't (at competitive schools, the best students tend to be active in many extra curricular activities unrelated to their majors)

In the engineering/academic world, grades do matter and are highly predictive of ability. People with poorer grades often struggle a lot, and the amount of time spent on remedial training may not always pay off. We have this ideal of a genius hacker who blew off school but is a 10x coder in real life... but in reality those people are comparatively rare.

[drawnwren]

Yes. To go along with this, the other misstep I see frequently is the inability to differentiate between statistical distributions and the individual. Your friend dropped out of college to become a carpenter and is now worth $100m? That's incredible. I'd take fairly bad odds that the P(Salary>$!00k|Dropped out of college to become a carpenter) is lower than P(Salary>$100k|Graduated with 4.0 gpa). It would be harder to statistically analyze, but I'd take even worse odds that the fellow who dropped out of college would not be too much worse off if he had stayed and graduated with a 4.0. IMO, the cost of a high gpa is fairly low.

[sago]

The GP explicitly reverses the conditional probability

> the most financially successful people I know had mediocre college grades

is saying something about p(mediocre grades|successful) i.e. among the people I know who are successful, a surprising number had bad grades.

where the interesting question, given the chronology of the two events is, as you point out, p(successful|mediocre grades)

[jhanschoo]

Yeah, both the distribution of highly successful people and people with high academic performance are heavily skewed and the slightest decorrelation should result in the phenomenon of heuristically seeming independent.

In an ideally meritocratic world, we should see fewer Harvard grads and magnis cum laudibus occupying the top of the industries.

[WalterBright]

> the cost of a high gpa is fairly low

That's right. I started out nearly flunking out in college, and eventually graduated with honors. The difference was not busting my ass studying, but more effective time management. I had as much free time as before.

[bernardino]

I learned that being able to sit down on your desk and focus on a topic or problem set and learn through trial and error with no distractions for hours is what will determine if you get good grades or not, and it's becoming extremely hard to do so. Most of my friends resort to drugs to focus but for me, that does not prefer drugs because it makes me feel artificial, it's been tough - it's like strengthening muscles in your brain in order to focus for long periods of time and those muscles for me are weak.

[ShabbosGoy]

> Most of my friends resort to drugs to focus but for me, that does not prefer drugs because it makes me feel artificial, it's been tough - it's like strengthening muscles in your brain in order to focus for long periods of time and those muscles for me are weak.

This is desirable. People who use psychostimulants to focus are using it as a crutch.

You can train your mind to focus, it just requires a larger time investment than popping a stim.

[bernardino]

Definitely! I should have started earlier in regard to training my mind to focus. Being in college, it seems I don't have the upper hand because so. It's all good though - I sort of argue that taking psycho-stimulants means you are entangled in that toxic cultural thinking of always going and being in the noise. I suppose then there's something sweet about failing a course without using psycho-stimulants.

[tikhonj]

> In the engineering/academic world, grades do matter and are highly predictive of ability. People with poorer grades often struggle a lot.

This is a purely empirical question to which you seem to have simply assumed an answer. While there are difficulties to studying this question empirically there have been solid studies to this effect.

I've seen a few solid survey papers on personnel selection methods; a solid example is Schmidt and Hunter (1998). Schmidt and a couple of others published an updated version of the survey as a working paper in 2016[1], which has a clear section on GPA:

> The validity value for grade point average (GPA) in Table 1 is for college and graduate level grade point averages. No estimates are available for high school grade point average, which may have validity higher than the .34 in Table 1. Apparently most of the validity of GPA is captured by GMA, because the incremental validity of GPA is negligible (less than .01). GPA has not been studied in relation to training performance, where its validity might be expected to be higher than the .34 for job performance, because of the strong resemblance between training programs and classroom demands.

(GMA in this context is a "general mental ability test"—basically an IQ test.)

This observation does not lend much credence to the idea that grades are "highly predictive of ability" or that people with low grades "often struggle a lot".

[1]: https://home.ubalt.edu/tmitch/645/articles/2016-100%20Yrs%20...

[sannee]

> This observation does not lend much credence to the idea that grades are "highly predictive of ability" or that people with low grades "often struggle a lot".

The paper does not seem to control for differences between colleges. It's unsurprising that the predictive value of, say, 3.8 GPA would massively differ between individual schools.

[xyzzyz]

Apparently most of the validity of GPA is captured by GMA, because the incremental validity of GPA is negligible (less than .01).

Just because GPA is basically a worse IQ test doesn’t mean that it’s useless. In fact, if you assume that IQ tests are useful in hiring, this actually makes GPA more useful, since US employers can freely use GPA in hiring, while use of IQ is significantly more complicated.

[dragonwriter]

> In fact, if you assume that IQ tests are useful in hiring, this actually makes GPA more useful, since US employers can freely use GPA in hiring, while use of IQ is significantly more complicated.

Except that's not at all true of the underlying legal standard, and if GPA is essentially an IQ test, it is illegal in the exact same conditions as IQ tests, and any temporary advantage attend from the fact that plaintiff's lawyers haven't yet recognized that GPA is factually nearly identical to an IQ test, and some employer is going to get a big surprise when they first do realize that.

[ChrisLomont]

>it is illegal in the exact same conditions as IQ tests

And IQ tests are legal to give during hiring, in just about any form, if that result correlates to useful for that job. Since we're talking about jobs where IQ/GPA may well correlate to candidate quality for that position, this is not an issue.

You have to first show there is no correlation between GPA or IQ and job performance to claim it's not legal.

[dragonwriter]

> And IQ tests are legal to give during hiring, in just about any form, if that result correlates to useful for that job

No, “correlates to useful” is not the standard; there also needs to not be a less discriminatory alternative available.

> You have to first show there is no correlation between GPA or IQ and job performance to claim it's not legal.

No, you don't: business necessity is an affirmative defense to disparate impact discrimination claims; once the unequal impact is proven, the challenged employer is required to prove the link to job performance. If they succeed, the challenging party has the burden of showing the existence of a less discriminatory alternative.

[ChrisLomont]

Cite a case. Griggs v. Duke Power is what most people claim makes IQ tests not legal, but that's not what the case was about. It was about discrimination against blacks, which is a legal protected class. Intelligence is not a legal protected class, so this ruling does not impact such tests for general intelligence.

>there also needs to not be a less discriminatory alternative available

This was to prevent discrimination against blacks (if you're using the case above), not for intelligence. It's perfectly legal to discriminate based on intelligence, however you want to measure it.

>once the unequal impact is proven

Only against protected classes. Intelligence is not a protected class. Here's [1] the EEOC list of allowed testing. Top of the list is general cognitive tests.

So, what case are you basing your claims on? Are you conflating discrimination against a protected class with discrimination against intelligence?

[1] https://www.eeoc.gov/policy/docs/factemployment_procedures.h...

[xyzzyz]

That’s the theory; in practice hardly anyone does IQ tests, while employers asking for GPA is an extremely common occurrence. What matters for them is not what is legal, but how likely something is going to cause you legal problems.

[joeax]

I was once asked to take an IQ test for a software developer job. I believe it was more of a formality, and they wouldn't tell us our score. I'm not sure it was legal or not, and I was glad to do it.

That being said, in my 20-year career I've never been asked for my GPA. Not once.

[joeax]

> In the engineering/academic world, grades do matter and are highly predictive of ability.

Can you cite some research studies that conclusively show this to be the case?

I'd like to counter by saying that companies like Google have largely ignored GPA as a measure of aptitude. From the article (link below):

"Google doesn't even ask for GPA or test scores from candidates anymore, unless someone's a year or two out of school, because they don't correlate at all with success at the company. Even for new grads, the correlation is slight, the company has found."

http://www.businessinsider.com/how-google-hires-people-2013-...

[Para2016]

You want him to cite a trial or study and you have a business insider article to counter him..?

[sheepmullet]

> I'd like to counter by saying that companies like Google have largely ignored GPA as a measure of aptitude.

I think you have misinterpreted the situation.

What it's actually showing is once you have restricted the pool of applicants to the top 10% of the field GPA does not matter within the restricted subset.

I can't think of anyone I know that could do well in the Google interview process but could not get good grades.

[lsc]

>I can't think of anyone I know that could do well in the Google interview process but could not get good grades.

To get good grades, you must be able to do rote work on time. I did really poorly in high school, because if you don't do one thing, and you do brilliantly on the next thing? That averages to a failure. Doesn't matter how brilliant that next thing was. For me? This means I barely cleared 2.0 in high school, and didn't seriously pursue college. But in industry, I seem to do pretty okay. My experience is that if I finish between 2/3 and 3/4 of what I start? I get a positive performance review and a raise. They even talk about it; like "if you are accomplishing all of your goals, you probably aren't being ambitious enough when setting those goals."

(I mean, I've been in industry since 1997, and my impression is that breaking in was a lot easier then than it is now... and it did take me a long time to work up to the point where I could get a job at a top-tier tech company, and even now, I'm a SysAdmin and not a SWE, (I have worked SWE type jobs at less prestigious companies... but here? I'm a SysAdmin.) I would be a better employee, with better job prospects if I had the personality and follow through to get a degree, no question.)

My experience with those interviews (at least for a more senior position) is that they test knowledge of whatever specialty you are dealing with and to a lesser extent, intelligence and problem solving ability. The former, of course, can (and should be) studied for; the latter, less so.

I'm sure intelligence and problem solving ability also help (and to some extent, are required) in academia, but if you aren't the sort of person who does 'good enough' work every time on time, you aren't going to get good grades, as far as I can tell, even if you are brilliant. That sort of plodding follow-through is not tested at all in interviews, and while it's a positive attribute to have as an employee, from experience, it won't kill your career if you are lacking it.

[joeax]

The article states that they don't even ask at all, so you cannot assume that the "restricted subset" has all high GPA candidates. Within that subset could be high school dropouts who are math and/or CS geniuses, for all we know.

> I can't think of anyone I know that could do well in the Google interview process but could not get good grades.

There's a lot of reasons people don't get good grades that has little correlation to their ability. People get bogged down by life, develop entrepreneurial interests outside of school, or have little interest in academics. I've met people who are brilliant software engineers and couldn't or wouldn't complete a semester of school.

[electricslpnsld]

> I can't think of anyone I know that could do well in the Google interview process but could not get good grades.

In most of the Google offices the percentage of employees with PhDs is hirer than the percentage of employees with a bachelor's alone (Seattle and NYC look like the only office where BS holders are the majority). While Google might not be explicitly selecting for GPA alone, academic achievement seems correlated with employment at Google.

[1] https://hackerlife.co/company/google/san-francisco-bay-area-...

[aswanson]

Sample bias.

[eganist]

> We have this ideal of a genius hacker who blew off school but is a 10x coder in real life... but in reality those people are comparatively rare.

I had exactly one of these guys. The reason I hired him was because he demonstrated mid-flight during a code review interview an immense capability to apply newly learned knowledge. I gave him a few (code) classes despite his background up until that point being nothing more than infrastructure-scripting, let him know he could ask absolutely any questions he wished, and would be scored on findings after he confirms he's completed the review.

Killed it. He was a bit rough around the edges in terms of social skills but I'd hire him again given the opportunity and given what he'd produced for me. He got picked up by a rather large firm in the Austin area whose name fits the pattern "A _ _ _ e"

Other hires aside from him were of the good-grades, great-work variety, but he was the exception who produced great work without that background.

[ThrustVectoring]

Yeah, compensating factors are a must. If you have a screening function that correlates with both ability to do the job and GPA independently, though, then conditional on passing the screen, the better candidates have a lower GPA. You've essentially got a noisy test for the sum of ability + GPA, so if the sum is high and they have a low GPA, the source of passing the combined test has to come from higher ability.

This process comes up in other similar contexts. SAT scores are not generally correlated between Math and Verbal, but if you screen based off "incoming Freshman to a particular college", an inverse correlation comes up. This is because admittance is based off the sum of the scores - too high and they go to a better college, too low and they get rejected.

[thaumasiotes]

> SAT scores are not generally correlated between Math and Verbal, but if you screen based off "incoming Freshman to a particular college", an inverse correlation comes up. This is because admittance is based off the sum of the scores

Your second point is correct, but your background is wrong; SAT math and verbal scores in the general population are strongly positively correlated, not uncorrelated.

[ThrustVectoring]

Right, lots of tests end up correlated with the general intelligence factor.

[drb91]

But a low gpa isn’t inherently meaningful at all. What do you mean “make up for” a low gpa when you shouldn’t be hiring around gpa in the first place? Like it’s not some debt to be paid; it just means they attended class sometime in the past. Just discount it entirely. What’s the worst that could happen?

[wenc]

This is why I mentioned nonlinear thinking.

If you imagine a multivariate nonlinear function (I don't believe there exists a definitive function, this is just a framework to help us reason more rigorously) representing the performance of a hire, there will be ranges where the function is insensitive to the x (e.g. say GPAs between 1.0-2.0). If you discount x entirely, you're throwing out the baby with the bath water because the signal might be meaningful in other ranges.

[drb91]

So what would you call the meaning of the low GPA? It is clearly meaningful to you. How would you describe the candidates you’re rejecting with this heuristic? It could mean anything from “i can’t read” to “i dropped out to form a business”.

I don’t see any value for the gpa beyond preparing students for performance reviews and identifying whether or not they give a shit about the material. And why would they? High Schools teach few real life skills and a lot of bad habits. Even in college, you’ll learn more (in some cases) by skipping class and building things.

Without context, gpa is not a meaningful metric

[mcny]

This is basically what a strategic advisor (sic) for ETS told me was the official position of the corporation.

He said ETS does not recommend that schools and employers use test scores as a (edit: SOLE) way to decide whether or not to admit/hire someone. What I heard was ETS is aware that schools and employers do this and wants to cover its behind.

[fucking_tragedy]

> What’s the worst that could happen?

Having to confront the cognitive dissonance that their youth was spent chasing numbers that have little to no bearing on their present or future.

[optimuspaul]

I have never asked or cared what grades anyone ever got. I don't think that there is any correlation between grades and real world performance. I am in the software engineering space though and I think that there is a lot of room for creativity. Having rigor in academic engineering isn't always useful and I've found actually prevents you from getting things to the good enough stage.

I totally agree that "determining a good hire requires multivariate". Maybe grades are a proxy for other meta skills, but I think evaluating those skills directly are a better indicator.

[tomc1985]

Other, more traditional careers place a bigger emphasis on grades as a matter of course. I don't think one just reduce it down to number of variables held in one's head whilst contemplating. All a high GPA should tell you is that the person successfully input things that the academic system accepted as an 'A', or whatever the score might be.

Good grades can be proxy for all those things you describe, and they can also be proxy for corner-cutting, systems-gaming, cheating, laziness, and other chicanery. I graduated sub-3.0 and, at least in my suburban, not-top-tier city environment, I seem to run circles around most of the technology-practicing clowns 'round here -- and it seems their muggle-ness transcends particular bands of GPA scores.

An "A" tells me you're either a cheater, a systems-gamer, or a schmuck.

[SamReidHughes]

Or you can just easily get an A, because you enjoy doing a few fun programming assignments and learn the things taught in class.

[hahnpv]

As an aerospace engineer, grades matter for your first job, maybe, then no one cares. After that, they are looking at your professional success/experience. Personal experience, having graduated with a ~ 2.5 GPA myself. Employers #2, 3 and 4 haven't asked about educational experience beyond formality, and are more interested in post-education body of work.

[bmarkovic]

> In the engineering/academic world, grades do matter and are highly predictive of ability.

Or rather, they would matter if:

- Various different schools used the same standard for grading so that grades could be even compared across candidates

- There were companies that actually care about GPA when hiring, I've never encountered one

[vonseel]

> in multivariate thinking, there has be other factors that compensate for the low GPA.

And the fact that it’s hard for a company to determine what those factors are is a problem in itself. Was the student immature at time of college (partied too much, disinterested in academics/chosen subjects are two problems that come to mind)? Did he work a full/part-time job while taking classes? Did he have a child? Did she care for someone else’s child or a younger sibling? Catch my drift?

Overall, I agree with most of what you are saying, but I also agree with the previous post. You seem eloquent and well-spoken, in fact it’s a bit off-putting.

> The fact is, determining a good hire requires multivariate, nonlinear thinking.

I understand your argument, but this colorful language only weakens your argument, unless you’re writing a paper for class. In conversation, you’d be better off rephrasing in a less formal manner.

[ataturk]

I worked my way through college and still got a 3.75 in a double major of Math and CS. My girlfriend at the time got a 4.0 at her college but had a full tuition scholarship and did it by dropping any hard math course that wasn't absolutely required. We both later got PhDs in our respective fields. I used to help her with her homework a lot.

Who was more well-rounded? I challenged myself with some of the courses I took and was overloaded some of the time due to work obligations, etc. I got a "C" in one math class in my final semester. Grades tell a tiny fraction of a story about discipline and self-sacrifice.

[wenc]

Again, multivariate thinking is important. GPA alone doesn't predict everything, but it is one predictor in a complicated function and it may have varying sensitivities in different ranges as well as have interacting terms with other variables.

One needs to factor in all the other predictors, as you alluded to in your story.

[mrkstu]

The question is how do you reliably gather a rich dataset that encompasses those variables. It is an un-solved problem vs shortcuts like GPA and prestige of their attended institutions.