Racial bias and NBA referees -- Part 3

This is part 3 of a post about the NBA referee racial bias study. Parts 1 and 2 appeared last week. Here is the New York Times article that started it all.

The study found that significantly more fouls were called when the referees were of different race than the perpetrator, and/or the same race as the victim. Could there be other explanations than racial bias?

Here's one, suggested by Guy in the comments to part 1:

... what if white players disproportionately commit certain types of fouls more than blacks, and other fouls less? And what if black and white refs, for cultural/historical reasons, also employ different standards for different types of fouls -- blacks quicker to call X foul, but slower to call Y? And suppose these both reflect differences in the "black" and "white" game, such that white refs are more sensitive to the kind of fouls committed more by black players. Wouldn't that then create the result the authors found?

I think it would.

It shouldn't be too difficult to check, if data on foul types is available. You can see if whites tend to take different fouls than blacks. You can check if white referees tend to call different fouls than black referees. Then, you can adjust all the numbers to see if the bias still holds up.

In baseball terms, it's like adjusting baseball numbers for park effects, and then for home/road – the arithmetic is pretty simple.

Guy also posted his comments to a post at The Sports Economist. Five minutes earlier, a commenter named Andrew had posted a similar theory:

Let us assume refs have at some point refs played basketball. Let us also assume black player play a more aggressive, foul-inducing basketball. Now, the white refs would have played with a "white" style while playing, and would be more prone to see aggressive "black" play as fouls, while black refs would see that style as normal and be less prone to call fouls.

So, taking a white player, both refs would likely see the same behavior as fouls.

On the other hand, an aggressive black player would more likely be seen as fouling by a white ref who played a less aggressive style than by a black ref who is used tot he more aggressive style.

Andrew's theory requires that blacks play more aggressively in general, where Guy's requires that blacks and whites take different kinds of fouls. But I think they're really the same theory: both say that referees see certain aspects of the game differently because of race, and both say that players play certain aspects of the game differently because of race.

I like this theory, and I bet it figures at least a little bit into the observations. As Andrew says, black referees and white referees grew up with different views of the game of basketball, just as they undoubtedly grew up with different views of a lot of things in life. It would be a huge coincidence if the interaction between referee tendencies and player tendencies turned out to be exactly zero.

By the way, the study considers that black and white players may have different styles. But it argues that style can be predicted from a function of a player's statistics, and several of their regressions considered such factors. (Specifically, they say that trying to predict race based on all those individual statistics yields a decent r-squared of 0.2.) But I'm skeptical that the connection between stats and style is really enough to say they're the same thing. I can easily imagine two players, all with exactly the same stats (including fouls), where the two players foul in different ways.

So should we conclude that referees are racially biased? I think this theory is enough to at least show reasonable doubt.

----

Here's another possibility. There's something economists call "rational discrimination," where people make decisions about a person based on the statistical profile of their racial group, even though they might not have any bias against the group in general. For instance,

Jesse Jackson once commented, “There is nothing more painful for me at this stage in my life than to walk down the street and hear footsteps and start thinking about robbery—then look around and see somebody white and feel relieved.” (

source)
We can assume that Jesse Jackson isn't prejudiced against his own race; he's just noting the unfortunate statistical fact that muggings are disproportionally committed by blacks. His relief is "rational," in the sense that he does indeed have less of a statistical chance of being mugged than he thought.

Often, when two players make contact, it's hard to tell which player is at fault. Suppose that referees have noticed – consciously or unconsciously – that when a certain situation happens, and it's between a white player and a black player, it's usually the white player's fault: say, 80% of the time. But the ref only gets a good view three-quarters of the time. What should he do the other times?

Presumably, he has to call a foul on at least one of the players, and he should make his best guess possible. Should he consider race? If he's completely unsure who's at fault, should he give the foul to the white player, who has an 80% probability of having done the crime? Or should he remain race-neutral, and flip a mental 50-50 coin?

I think he'd have to call it on the white player 80% of the time. Otherwise, the white player will have an incentive to automatically foul the black player, knowing that the opponent will be falsely accused more than he should be.

Now, suppose that black referees have noticed the 80% figure more than the white referees. That would account for the "same-race" effect. It might still be caused by bias, but a different sort of bias. The black referees were able to spot the differences in style of play. But the white referees, perhaps out of unconscious bias, did not.

I have to say that I don’t think this is the true explanation. But it does have the advantage that it predicts same-race favoritism among both perpetrators and victims. Plus, if it is correct, the bias is easy to fix – just share the 80% figure with the white referees.

-----

Ian Ayres, one of the economists the New York Times called on to review the paper, believes that the findings are indeed the result of unconscious bias. He argues that such bias against blacks is pervasive, and that while well-meaning people can consciously overcome it when given time, that's not possible when making split-second judgements, no matter how hard they might try.

Ayres refers to the "Implicit Association Test," or "IAT." That's a little experiment that tries to measure your bias. Roughly speaking, it works like this. Suppose you want to figure out if you're biased against blacks. They give you two timed tests. In the first test, you have to divide a combination of words and faces into two groups. Group 1 contains words that are "good" (like "celebration" or "joy") and faces that are white. Group 2 contains words that are "bad" (like "evil" and "sadness") and faces that are black. So you have to separate "good or white" from "bad or black."

In the second test, the switch you up, so that you have to separate "good or black" from "bad or white."

The idea is that if you already associate white with goodness and black with badness, the first test will be much easier than the second one, and you'll finish it much faster.

I was skeptical, until I actually found some online tests. I took the one that checks if I'm biased towards Canada over the USA. I'm Canadian, but not particularly patriotic (except when it comes to hockey), and I'm more pro-American than most of my friends. But it freaked me out how much easier it was when the pictures of Canada were paired with the "good" words. I came out as having a "strong automatic preference" for Canada, the largest bias possible under the test.

Ayres thinks this test (for black vs. white) could serve as a proxy for estimating referee bias. I think his theory is reasonable, that the IAT probably correlates with decisions made on the court. It would be nice to have some evidence, of course, and Ayres does suggest giving all the referees the test to see how it matches up to the individual referee "bias" scores in the study. (To get the refs to take the test, you might have to have a third party administer it, and guarantee that only the correlations will be released, and not individual results. Also, you would probably have to hold a gun to their loved ones' heads.)

Now, I have to say that while I think the IAT probably measures something real, I don't think it necessarily measures racist attitudes. It could just show preference for one's "In-group" – and that preference is universal. According to Wikipedia,

Experiments in psychology have shown that group members will award one another higher payoffs even when the "group" they share seems random and arbitrary, such as having the same birthday, having the same final digit in their U.S. Social Security Number, or even being assigned to the same flip of a coin."

I believe this is true. I'm as big a skeptic of astrology as anyone, but if you did an IAT by zodiac sign, I'm pretty sure I'd show a strong preference for my own. (Go, Capricorn!) So I think that part of what the Price/Wolfers study found as "racial discrimination" is actually just in-group Bias.

That is: I think if referees do indeed favor players of their own race for non-basketball reasons, they would also favor players of their own religion, or players from the same school, or players from the same city, or players who support the same political causes. (More importantly, I think that referees may favor certain teams over others. Any time I watch a sporting event between two teams I never heard of, it isn't long before I start arbitrarily rooting for one of them. But maybe that's just me.)

Admittedly, I have no proof of this, and a possible counterargument comes from the IAT itself. As expected, white people taking the IAT come out with a preference for white over black. But black people also prefer white over black! (The IAT site suggests that there's more than just in-group preference happening – see answers #8 and #9.)

A different finding supports the idea that the preference for whites comes from something other than irrational racism. Ayres writes,

Simply showing people a picture of Michael Jordan (or Bill Cosby) reduces implicit racial bias in the IAT. This suggests that people so much want to "Be Like Mike" that he transcends racial categories, making us think less about race.

That implies that, to some extent, it wasn't about race in the first place. Real racists don't care if you're Michael Jordan or Bill Cosby – you still can't get into their club, you're still ordered to sit at the back of the bus, and you still have to drink from the "colored" fountain. It seems to me that if looking at Bill Cosby reduces your unconscious bias against blacks, it's really an in-group thing – you identify with Cosby, you feel that he's somehow one of you, and that winds up extending to all blacks.

Anyway, I'm speculating here, and it seems like debates on the IAT have been going on for years, by people who are smarter than me and have more than ten minutes experience with it. Still, when Ayres suggests that the IAT should be used by the NBA when making referee hiring decisions, I'm not convinced.

For one thing, if you trust the IAT, then referee bias must be fluctuating all the time. If just looking at Bill Cosby can make you feel less biased, wouldn't an ugly disagreement with a black player make you feel more biased?

And second, could it really be true that blacks would be less prejudiced against whites because they favor whites on the IAT? Just two days ago, King Kaufman wrote,

An ESPN-ABC News poll has found that black baseball fans are more than twice as likely as white fans to be rooting for Barry Bonds to break Henry Aaron's career home run record, and white fans are more than twice as likely as black fans [76% to 37%] to believe that Bonds knowingly used steroids, more than three times as likely to believe he shouldn't go to the Hall of Fame.

Now, these numbers are so far apart, that you've got to believe that one side must be significantly biased. If you're a white person, and believe that whites are very biased but blacks aren't, you are forced to accept the black consensus on the Bonds issue. You've also got to accept that O. J. Simpson might be innocent (73% of blacks think so, versus 13% of whites). And on any other issue that has to do with race, where bias could come into the equation, you need to figure that black opinion, as unbiased, is much more likely to be correct than white.

I'm not prepared to do that. Sometimes, I agree with the consensus "white" point of view. Sometimes, I agree with the consensus "black" point of view. Sometimes, I think the truth is somewhere in the middle.

So I think it's overwhelmingly likely that both blacks and whites are biased in favor of their own race, regardless of their average results on the IAT.

The important question then becomes: if in-group bias is universal, can referees be trained to overcome it in split-second decisions? I don't know, and I don't even know how you would check.

----

Anyway, I guess I can sum up my evaluation of the study as:

1. I think the effect the study found is real, that players benefit when referees are of the same race.

2. I think the study cannot and does not show whether white referees are biased, whether black referees are biased, or whether both are biased. I don't think you can even tell which way they're biased.

3. The Guy/Andrew theory – interaction between types of fouls taken and styles of fouls enforced – is a possible explanation for the observed effect.

Also, my gut says – without necessarily any evidence:

4. The Guy/Andrew theory probably only accounts for a small part of the effect, if any (although I am open to contrary evidence).

5. There probably is a real bias effect happening, but "in-group" bias is a more likely explanation than race bias.

6. No matter how non-racist or anti-racist the referee, such bias might be impossible to eliminate.

7. There are probably many similar "in-group" biases that referees have, in all sports, but that nobody has studied yet.

For these reasons, knowing about the same-race effect won't really interfere with my enjoyment of the game. While I think the small amount of bias that the study found might be real, I suspect it's just part of human nature. So long as there isn't any conscious racism, and the league and referees are trying their best to improve and be fair – and I believe both conditions to be true – I don’t think the integrity of the game is in jeopardy.