November 30, 2013

Why job testing went out of fashion

As I mentioned in the post below, the Atlantic has a long article about Silicon Valley start-ups attempting to use Big Data for job hiring testing. In the post-WWII era, the article says, American corporations did lots of testing of job applicants, but that fell out of fashion because science. Or something. So for the last generation, firms mostly rely upon resumes and interviews and try to avoid putting much in writing where it can get subpeonaed.

But now in 2013, instead of giant corporations like P&G doing the testing themselves, it's going to be done for the giant corporations by cool little start-ups with cute names like Knack, Evolv, and Gild. So the New Testing won't be like the bad Old Testing of the 1950s when the racist, sexist white male power structure was building a giant middle class with secure jobs and pensions. Or something.

A reader writes:
Saw your post about pre-employment testing.  It's been a while since I've had my head in the selection literature, but historically the general pattern of predictive validity for various selection methodologies has tended to follow a consistent pattern.  The most predictive procedures tend to be those that emphasize biographical data, primarily work history and education, with coefficients in the .5 range or even better.  This is followed by formal testing and is is where it gets a little tricky.  Much of the early selection research was conducted or funded by the military, primarily the Navy through the NPRDC (now the NPRST).  This is because the military can't spend a lot of time messing around trying to figure out what MOS people are suited for.  Therefore, you have the two component tests you describe, an IQ-ish type of test and a group of aptitude-type tests.  This kind of testing has relatively high predictive validities, almost as high as those for biographical data.  The problem with both of these methodologies is disparate impact.  Therefore, employers are going to great lengths to try to avoid correlations with prohibited dimensions.  If you want to read something that strikes fear into the heart of HR managers, peruse section 5 of the Uniform Guidelines on Employee Selection Procedures.   

The whole federal guideline related to job testing is 16,942 words long. From skimming it, I'd say that, yeah, you could get away with testing. After all, P&G still gives a "Reasoning Test" that looks much like the GMAT that MBA applicants take. But this Guideline is intended to make you think long and hard before trying it. Sure, P&G got their test validated, but then P&G is largely staffed by competent people hired in part through testing. Can your staff successfully jump through every hoop in the 16,942 words?

After all, your managers like interviewing. They each think that -- while everybody else is terrible at interviewing -- they are way above average at it. Interviewing lets them hire people they like. So why risk a federal lawsuit to make them hire people they feel less sympatico with just because they'd be better workers?
What you really have above is one methodology that screens for conscientiousness (work history) and one that screens for intelligence (tests). Privately, Industrial/Organizational Psychology types will tell me that this is really the only thing that matters in selection.  The rest is just BS. 
The real scandal in employee selection is the almost zero predictive validity of interviews. No matter how they are constructed they tend to contribute almost nothing to employee selection.  Probably the best they do is identify obvious jerks, but even that is questionable. The reason we continue to do them is mostly cultural, I suppose.  There is also a cult of "personality testing" inside of HR these days, the MBTI seems to be the favored tool, although there are others.  This is obviously quite distinct from the AQFT-type testing referenced above. 
Understanding the HR field today one has to realize it embodies two very different and often contrary functions; selection and compliance.  The first area is dominated by I/O psych types and the second by lawyers.  The academic side of HR, in particular university based research into selection, is almost universally populated by the former.  On the practitioner side, while most HR people are not lawyers, the legal issues tend to overwhelm everything. This has led to growth in formal structured procedures for all of these types of decisions and associated documentation requirements.  Unfortunately, despite all of the attempts to create procedural fairness,  the drift has been back into interview type "fit" exercises, and hence what HR people call the "like-me" problem.

25 comments:

Daniel said...

It's relatively easy to fire someone in the US, though lots of big companies are reluctant to fire someone for simple incompetence, especially if they are a member of a protected class. In France, however, it is notoriously difficult to fire someone. More so than any country I am aware of. Even small enterprises have difficulty getting rid of someone. What methods do the French use to separate desirable from undesirable candidates?

Anonymous said...

OT, but Drudge has linked to this AP story that shows Americans don't trust one another as much. The article reports that trust is important for a society.

Does it matter that Americans are suspicious of one another? Yes, say worried political and social scientists.

What's known as "social trust" brings good things.


I read it because I immediately thought about Robert Putnam and his study that shows social trust decreases when diversity increases. Sure enough the article mentioned Putnam.

Putnam says Americans have abandoned their bowling leagues and Elks lodges to stay home and watch TV. Less socializing and fewer community meetings make people less trustful than the "long civic generation" that came of age during the Depression and World War II.
From the article:

But it did not mention his findings that diversity increases distrust. And it did not mention words like immigration, diversity, or demographics.

Instead it reported this nugget, "The decline in the nation's overall trust quotient was driven by changing attitudes among whites.".

So whites are apparently getting selfish or too self absorbed which is leading to a decrease in social trust. No mention of the 800 pound gorilla in the room.

dearieme said...

In France, family and "pull" count for a lot, plus the prestige of your education.

A young French friend told me, after he got his first academic job in Cambridge, that it was very odd: Cambridge was far superior to any institution in France but he'd never have got the equivalent job in any of them. Whereas in Cambridge he was appointed because of the quality of his research; the fact that he'd attended a humdrum French institution for his first degree didn't seem to worry anyone.

Anonymous said...

"...instead of giant corporations like P&G doing the testing themselves, it's going to be done for the giant corporations by cool little start-ups with cute names like Knack, Evolv, and Gild."

I used to like the whimsical names of tech companies, but I feel the cutesy names have become a bit tiresome and reek of arrested development.

Big Bill said...

One way of avoiding employment liability is to cease hiring "employees".

Define a service to be performed and contract that service out to another company.

Since you never "hired" anyone, by definition you cannot have "discriminated" in hiring.

It never shows up on the EEOC radar.

David Pinsen said...

The recent Silicon Valley practice of "acqui-hiring" (buying a startup for its workers rather than its product or tech) is a novel way of getting around tricky hiring regulations.

David said...

The problems of firing people can be significantly reduced through two methods of getting rid of workers without actually firing them.

1. Run them off. Management on all levels has always been adept at doing this, sometimes doing it for no reason besides the perverse pleasure of sadism. There are countless unactionable ways to deliberately break an uncool employee's trust in sanity and humanity and get her running for the exit. In Japan, management gives the target littler and littler offices in the basement and no work. (Occasionally also placing an article about seppuku on the weirdo's desk and saying it's for him to "mark up" or "write a detailed report on" is something I've not seen reported, but it wouldn't surprise me to learn that it has happened.)

2. Load up on temps. A temp can be dispatched instantly, no questions asked. Just make a phone call to the agency at lunch and you won't see him ever again. Temps are an increasing part of the US labor market. Instructive story from Fox.

Cail Corishev said...

So you can't base your hiring decisions on IQ, but you can base them on a parlor game like the MBTI?

Every time I think I can't possibly be surprised by the stupidity of PC thinking, I'm proven wrong.

Anonymous said...

"The problems of firing people can be significantly reduced through two methods of getting rid of workers without actually firing them."

There's another method that's been very big in large tech companies for the last few decades, stack ranking. Just fire the bottom 5% or something like that, every year. I suppose you could call this a variant of "run them off", but it's more along the lines of "just churn through a constant pipeline of lots of people".

How do you identify the bottom? Well, that's a dispassionate metric-drive management exercise, heh, heh. It helps if you have everybody in the group do "360s" on everyone else. That way it just falls out of the numbers. You might have to wait a bit till the right numbers come around.

What happens if you had a superbly functioning team that had no "bottom 5%"? Well, after a few rounds of this, it won't be a superbly functioning team!

What happens if all the good employees get fed up and leave and by now you've got quite a reputation as a cheesy place and can't find anybody to hire? That's what immigration pipelines are for! There's bound to be somebody somewhere who's worried about getting shot or starving and is willing to put up with any of your shenanigans for a few years until you fire them.

All this turnover is good for the company. No targets here, lawyer types!

Anonymous said...

Have you been following the chimp/pig hybrid human origin hypothesis story in the Daily Mail, Steve? The weird thing is, it would explain so many things. Check it out. If he is correct, or at least if the idea comes to be accepted, he'll win a Nobel for sure.

Anonymous said...

There are a couple of problems with "just running them off" and stack ranking. In the first case there is a legal principle called "constructive discharge." This is where you move someone into a crummy office or give them bad work assignments hoping they quit. You might not have fired them technically, but in a practical sense you did, by making their job unbearable. As to stack ranking, whatever method you use to rank is subject to the same validity rules as tests. From an EEO perspective, this is "selection." If the method results disparate impact, you have a potential problem. IMO, the safest method for getting rid of someone is to pay them to go away.

ChrisA said...

On why hiring people like interviews, it creates a contract between the hirer and the hiree. If you get hired after an interview, basically you are under an obligation to the interviewer, and they know that. I would never accept anyone into my team unless I had a chance to interview them first, if they are placed in my team by someone else then their loyalties lie elsewhere.

Dave Pinsen said...

Is stack ranking that common? I can only think of two companies off the top of my head that practiced it, GE and Microsoft, and I think Microsoft just scrapped it.

Anonymous said...

From an EEO perspective, this is "selection." If the method results disparate impact, you have a potential problem.

Makes you wonder why you have all those groups of a dozen Indians working for an Indian manager in silicon valley, with no non-Indians in the group. With perhaps half to 2/3s of the Indians these days being direct from India, not from a US school.

Perhaps to the EEO it's only a "potential problem" if it works out one particular way. Not so much a problem if it works out the other. Asymetrically enforced, like immigration.

The funny thing is how "un-EEO-selected" some parts of silicon valley tech companies are, no matter what the law says. I think Steve has alluded to it in a few stories. If you're a lawyer, there might be easy pickings for you there.

guest007 said...

To those who wonder how a company can be all Indian (or all black). The first trick is to not advertise the open positions but to just hire your friends in family. It is not illegal until someone sues or comlains. If no one know that there was an open job, then no one will complain.

What is odd are the government and government contractor job announcements that are so specific, it is obvious that there has been pre-selection. Who wants to get involved just to get sue someone?

Anonymous said...

"Is stack ranking that common? I can only think of two companies off the top of my head that practiced it, GE and Microsoft, and I think Microsoft just scrapped it."

Seems much more common than that:

"There's a quote floating around the Internet that some 60% of Fortune 500 companies are doing forced rankings. We tracked down the source of the quote and asked him about the practice. It came from human resources expert Dick Grote, president of Grote Consulting and the research he did for his book Forced Ranking: Making Performance Management Work [2005, Harvard Business Review Press]."

This article says GE's Jack Welch made it popular (he called it "Rank and Yank" and applied it to 10% a year).

Why did GE hire so many bad people that they had to get rid of? Maybe because they couldn't test them in the first place? Or maybe they just weren't nearly as good as they thought they were at hiring good workers or managing workers, I guess. Were they just a bad place to work? Or maybe they just wanted an excuse to get rid of lots of people so they could get rid of real troublemakers without having the government down on them. A conundrum.

Looks like this is also called the Vitality model:

"Jack Welch's vitality model has been described as a "20-70-10" system. The "top 20" percent of the workforce is most productive, and 70% (the "vital 70") work adequately. The other 10% ("bottom 10") are nonproducers and should be fired.[1][2] Rank-and-yank advocates credit Welch's rank-and-yank system with a 28-fold increase in earnings...

...Welch admitted that the judgments were "not always precise"."





Enron was a stack-ranker: "...As Enron internally realized it was entering troubled times, rank-and-yank turned into a more political and crony-based system."


Interesting:

"Evaluation by performance, merit rating, or annual review of performance" is listed among Deming's Seven Deadly Diseases.

It would seem hard to get by without such evals in a modern international company. (On the other hand, you don't have time for formal performance evaluations in small startups; silicon valley startups often ignore all this, but of course everyone often knows who's doing what.)

Anonymous said...

More on the chimp-pig human hybrid hypothesis, Steve. It's probably bigger than a Nobel in scope, as it has the potential to explain a vast amount of things that standard Darwinian evolution does not, e.g:

Absence of missing links. In this case, the jerks that evolution is punctuated by are caused by instances of hybridization. For example, in this article, McCarthy posits that the reason for limits to size (relative to body) of the non-human primate brain is that it doesn't have an adequate cooling system. He goes into great detail discussing why pigs have such an effective cooling system necessary for a big brain.

So in this instance:
1. Pig/chimp hybrid forms (with back-crossing).
2. Natural selection increases human brain size to the limits of new cooling system.
3. Humans evolve.

This is so much more compelling than the idea that huge changes are brought about by random small mutations such as those caused by cosmic rays etc. Instead, the big changes are brought about by huge mashups of existing animals that are relatively far apart on the evolutionary tree that have themselves diverged and perhaps reached the limits of their own evolution... but have individual solutions that are better than the other animal.

He also posits that the gorilla is actually another hybrid between the ancient chimp and the giant forest boar, but with a different set of traits that were mixed. If so, then there is a general phenomena at work here that would explain the development of many species of creatures.

In essence this idea also solves in general the problem evolution has in reaching local maxima. e.g. The chimp reaches a local maxima in terms of brain size, and the solution can only come from another order of mammal such as a pig.

The impact of this will potentially be vast. However, in the short term I expect the film and game industry will be among the first to draw inspiration from it.

e.g What is the limit to the size of the human brain currently? I would say that it is due to the hip construction of the female, (or from a technocratic point of view, the lack of caesarian section technology). One solution around this to breed the next generation of super-humans that will wipe us out is finding an animal that has better hips than we do, and perhaps some even better cooling than the pig for example. If anyone reads this and writes a blockbuster script from it, please donate 10% of the proceeds to iSteve.

(If this reminds me of anything, it is in approaches to CPU cooling. The heat pipe, the CPU fan, all of those things, the race to increase the TDP that can be removed from a processor. And also things from the processor side. For example, turbo technology where the processor can work very fast, but only for a short time. Human brains have very much the same problems - overheat and they cook. But the advantages to a better brain are immense.)

Marlowe said...

"The creatures outside looked from pig to man, and from man to pig, and from pig to man again; but already it was impossible to say which was which."

Anonymous said...

A reader writes: What you really have above is one methodology that screens for conscientiousness (work history) and one that screens for intelligence (tests). Privately, Industrial/Organizational Psychology types will tell me that THIS IS really the only thing that matters in selection. The rest is just BS.

What is "this"?

The "intelligence (tests)"?

The "conscientiousness (work history)"?

Or the "one methodology" [in which case it would be more "these are" rather than "this is"]?

I ask because we were already under the impression that IQ correlated pretty strongly with conscientiousness.

But if you need an independent measure for "conscientiousness", to supplement the testing for the baseline IQ, then it sounds like what you're searching for would be a quality like "interested in [or reasonably enthusiastic about] working on the particular tasks which would be assigned".

Or "has no interest in [or enthusiasm about] working on particular tasks which would be assigned, but economy sucks so badly that applicant would gladly take lowliest job at waste water treatment facility in order to feed family".

Or maybe the independent measure of "conscientiousness" would be trying to rule out health problems, like schizophrenia or bipolar disorder or substance abuse, or even just plain old-fashioned character flaws, like disloyalty or fatuous unseriousness or outright laziness?

Thanks.

Z Blog said...

Twenty years ago I worked for one of the last companies to extensively use testing as an applicant screening tool. They have gone out of business, but they were still doing it ten years ago. I know the U-Haul company requires a basic math and literacy test. The Wonderlich is still used by a lot of companies.

The company I worked for that did extensive testing was a weird combination of 1950's and 1980's business. The tests they used for personality profiles was created in the 50's. One of the questions, I'm not kidding, was about the appropriateness of women in bars.

On the other hand, the workups they did to determine the right career path within the company were interesting. As a guy who has been tested a lot, I had no trouble cracking the test, but it was fascinating to read the results. I gained access to that data and whiled away man hours reading them.

I suspect guys like me were one reason these things went out of style.

Anonymous said...

As Enron internally realized it was entering troubled times, rank-and-yank turned into a more political and crony-based system.

In my experience, a company doesn't even have to be in trouble for rank-and-yank to turn into a political and crony-based system. It's just a natural result.

Stack-ranking companies are among those pushing strongest for open immigration. If you fire 10% of your workforce every year, regardless, you need lots of bodies every year. You can only burn people for so long before they turn on you. Luckily there are billions of people out there, so it doesn't matter. So these companies are probably right that they need immigrants, since they have a hard time getting people who know much about them to work for them.

It also doesn't take a genius to realize that a company that burns through employees at that rate isn't going to be investing in employee training, skills, or careers. Use 'em up like towels, ring 'em out and throw them away.

The big MBA-type consulting companies, like McKinsey, seem to be in the forefront of pushing for stack-ranking and immigration. Like prestigious economists, it's not clear why companies give the McKinsey's of the world so much power over them, but they do. What's in it for the McKinsey's? Perhaps nothing else than churn. Consultants have to drum up business somehow.

Anonymous said...

More on the chimp-pig human hybrid hypothesis, Steve. It's probably bigger than a Nobel in scope, as it has the potential to explain a vast amount of things that standard Darwinian evolution does not, e.g:

My problem with the theory is that McCarthy argues that a male pig mated with a female chimp. Doesn't it make more sense that a male chimp mated with a female pig?

Anonymous said...

"Instead it reported this nugget, "The decline in the nation's overall trust quotient was driven by changing attitudes among whites."." - they did acknowledge, though they went to pains to explain away, that blacks simply don't trust whites.

Whites were the only suckers left that still believed in the system.

Anonymous said...

Weird.

http://youtu.be/ySXkHj1Yp1w

http://www.uwosh.edu/faculty_staff/earns/song.html

http://en.wikipedia.org/wiki/Mort_Dixon

http://en.wikipedia.org/wiki/Harry_Warren

Anonymous said...

@ Anon questioning about conscientiousness and IQ I ask because we were already under the impression that IQ correlated pretty strongly with conscientiousness.

IQ correlates weakly and negatively with trait conscientiousness (as measured by the personality tests that bother with it).

There are various theories about why this is so -

1) Conscientiousness and regulating your own behavior requires brain resources that can't be spent on problem solving.

2) Conscientious self control restricts free thinking in problem solving - conscientious people tend to crimethink themselves a little bit more when challenging the status quo, although about some things maybe a little less than less conscientious people.

3) Conscientiousness is partially a way for less intelligent people to compensate for their lack of intelligence, while more intelligent people can cut loose and coast on their higher intelligence.
etc.

@ Anon Fan of the Pig-Chimp idea This is so much more compelling than the idea that huge changes are brought about by random small mutations such as those caused by cosmic rays etc. Instead, the big changes are brought about by huge mashups of existing animals that are relatively far apart on the evolutionary tree that have themselves diverged and perhaps reached the limits of their own evolution... but have individual solutions that are better than the other animal.

Selection on standing variation may be more important for human divergence from our chumans ancestors than new "random small mutations". I wouldn't be surprised if we could get back to chimps pretty easily just with standing variation within the human species and strong enough selection - Homo Floresiensis looks a bit like that.