bias

Intersectionality Eliminates Measured Hiring Bias

Zachary BartschLeave a comment

The US Equal Opportunity Commission identifies characteristics by which an employee can’t be harassed, hired, paid, or promoted. A challenge with enforcing the non-discriminatory standards is that the evidence must be a slam dunk. There needs to be a smoking gun of a paper trail, recorded conversation, or multiple witnesses. Mere statistical regularities are insufficient for demonstrating that characteristics like race, age, or sex are being considered inappropriately.

If employees are all identically qualified, then we’d expect the employment at a firm to reflect the characteristics of the applicant pool, within a margin of error due to randomness. One difficulty is that plenty of discrimination can occur within that margin of error. A firm may not have sexist policies, but a single manager can be sexist once or even multiple times and still keep the firm-level proportions within the margin of error. This is especially stark if the company managers or officers are the primary positions for which discrimination occurs.

Another difficulty is that randomness can cause extreme proportions of employee characteristics. Having a workplace that is 95% male when the applicant pool is 60% male isn’t necessarily discriminatory. In fact, given a sample size, we can calculate how likely such an employee distribution would occur by randomness. Even by randomness, extreme proportions will inevitably occur.  As a result, lawsuits or complaints that have only statistical evidence of this sort don’t  go very far and tend not to win big settlements.

But this doesn’t stop firms from avoiding the legal costs anyway. Firms generally prefer not to have regulatory authorities snooping around and investigating. Most people break some laws even unintentionally or innocuously, and a government official on the premises increases the expected compliance costs. Further, even if untrue accusations are made, legal costs can be substantial. Therefore, firms have an incentive to ensure that they can somehow demonstrate that they are not being discriminatory based on legally protected characteristics.

However, as I said, extreme proportions happen randomly. If those extremes are interpreted as evidence of illegal discrimination, then the firms have an incentive to hire among identical applicants in a non-random manner. They have an incentive to tilt the scales of who gets hired in favor of achieving a specific distribution of race, sex, etc. People have a variety of feelings about this. Some call it ‘reverse discrimination’ or discrimination against a group that has not historically experienced widespread disfavor. Others say that hiring intentionally on protected characteristics can help balance the negative effects of discrimination elsewhere. I’m not getting into that fight.

Continue reading →

Berkson’s Paradox nay Bias and Spring Break Blogging

mdmakowskyLeave a comment

You may be tempted to observe a negative correlation between the length of my blog posts and fraction of the previous 7 days that can be accounted for as “Spring Break”, but I submit that you may simply be omitting from the sample all of the short blog posts I could hypothetically be writing in crisp fall months.

Do read Lionel’s whole thread though. It’s good.

The Political Center is Endogenous to You

mdmakowskyLeave a comment

I stumbled into a twitter conversation about this relatively innocuous breakdown of news sources:

Now there are plenty of ways to pick this apart. No, I don’t care that you are socially liberal and fiscally conservative. I also do not care that a one-dimensional political spectrum can’t capture the fine nuance of your political ethos. Yes, the farthest left and right bins equivocate between often very different levels of bias, but that’s mostly of a product of only having five bins. Obviously greater accuracy could be had by delineating into 9 bins (and yeah, there is some real weird stuff in tails of this distribution). But coarseness or inaccuracy at the margin is not what grabbed by attention. I don’t really care that the lunacy at OAN and largely evidence-based reporting at Vox are in bins of seemingly reciprocal bias, as I know that’s just an artifact of the 5 bin structure.

Rather, what I was intrigued by were the frustrations and bias attributed in the comments about news sources, such as the BBC and NPR, improperly classified as the center. I’m well aware that many people are annoyed by “centrists”. They are traitors to the cause who should know better than to side with the enemy and will, of course, be first against the wall when the revolution comes. There is, more seriously, a frustration that centrists think they can make a claim to the truth by simply splitting the difference of the political distribution, like using the mean surveyed number of jelly beans in a jar. This aggravates people partly because it is a gross misuse of the wisdom of crowds, but mostly because we often think our position is the truth against which all other political identities should be gauged. It’s an old George Carlin joke – everyone driving slower than you is an idiot, everyone driving faster than you is an asshole. The “right wing” is everyone to the right of me. The “left wing” is everyone to the left of me. As for the crazies, well, that depends on your social identity. If you think of yourself as a right (left)-of-center, well then the far left (right) is full of lunatic socialists (corporatist fascists) out to destroy everything we love. The far right (left)? Well, they are a bit much I admit, but they are just spirited activists doing their best in a hostile environment.

Everyone hates centrists in large part because so many of us, on some level, think of ourselves as the reasonable political center. For some right or left-wing yahoo to plant their flag in to the rich soil of the center and call it their own is not just an affront to our sensibilities, it’s an act of political war.

To be fair to the Twitterverse, one person did manage to bring to this cavalcade of frustration an excellent alternative chart that had 7 bins (!) and a second dimension (!!!) regarding the reliability of information. That should have calmed most people down, obviously social media is neither the time nor place for such things.

It’s in this wonderful figure that so much of the story really comes out. People are rightfully upset that honest news sources are being conflated with tabloid rags. They’re also upset, however, that excellent and reliable sources are being attributed centrist neutrality. How dare they attribute the power of veracity and truth to those well-known right-wing whackos at the BBC! We tell ourselves we ignore the BBC/NPR/Economist/WSJ because of its gross bias, but the reality is we ignore them because they’re boring and never tell us we’re smart and pretty and righteous.

The original post was trying to suggest to people they consider balancing their political diet. My suggestion would not be to balance the bias in your diet (we like what we like), but rather to focus on the most reliable sources (the green bullseye in the second figure) and cut out the fried BS. All of that rage and confirmation bias, it’s nothing but empty calories.