54.9 Million Reasons to Validate Your Test


So, here’s a story that’s been going around in certain circles. Shipping superstar FedEx settles a discrimination lawsuit for $54.9 million. The suit was brought about in part because they used a “Basic Skills Test” that looked valid on the face of things, but which (supposedly) had adverse impact and for which the company could not produce (or decided not to bother producing) any kind of validation evidence. Oops.

I think this is an interesting and possibly influential case because it involves such a recognizable company that everybody is familiar with. I think a lot of decision-makers may shrug their shoulders and turn away when you point out a lawsuit involving a small police department in Hootinvill, Illinois or some other organization that doesn’t seem relevant. But FedEx? Man, we use them all the time! I’ve shared this story with some people not normally involved with employment testing, and it seems to get them to stop and take notice.

What’s also interesting is that if you look at at a sample of the test, it appears to be fairly straight forward, clean, and possessed of face validity. And those are all nice things, but apparently they couldn’t produce any evidence that the test actually resulted in better hires. I say “apparently” since the article linked to above is kind of vague, saying just that they caved in to the accusations without raising evidence to contradict it. Maybe they decided the evidence of other discriminatory acts was too overwhelming to bother.

Regardless, this case makes for a nice example of the dangers of letting subject matter experts develop their own test and running with it. Sure, they can make a test that looks good, and maybe it is valid. But without the help of I-O psychologists or similarly trained experts, you can’t prove it. And when the lawyers or judges or EEOC come a’knocking and ask you to respond to accusations of discrimination, you’d better have something more substantial than “Well, Ted thought it looked good.”

2 Comments on “54.9 Million Reasons to Validate Your Test”

  1. David says:

    I think most companies are getting lazy with validation. Or they rely too heavily on the vendor for support. One way to fix this is to start treating testing data and performance data used for validation as part of the overall HRIS data mart. This way data is current and can be monitored for adverse impact and if matched to performance data containing dates can be used to collect ongoing validation data (I am standing on my soap box).

  2. Scott says:

    I heard of a site call Video Resume http://videoresume.com
    Can this be the one?