Wednesday, May 02, 2007

A Little Heavy Reading

A couple articles caught my fancy this morning. One is a New York Times article describing an academic paper published by Justin Wolfers, a Wharton School professor, and Joseph Price, a Cornell grad student, that concludes that NBA referees are affected by subconscious (we hope) racial bias when assessing fouls. The authors conclude that the racial composition of an officiating crew can have a non-trivial impact on player performance, although the effect can really only be seen over the course of the 82-game schedule.

The NBA has done a counter-study that finds no racial bias. I like David Stern and the NBA, but does anyone really take this study seriously? Does anyone think that if the NBA commissioned a counter-study and it replicated the finding that referees have subconscious racial preferences, the NBA would then release that study? The one advantage that the NBA study has is that it can use a data set that takes into account the specific referee making a given call. (Wolfers and Price could only analyze the effect of the racial composition of a crew as a whole, since they did not know which referees made particular calls in a given game.) Naturally, the NBA will not make its data available, citing confidentiality concerns. If the NBA had a real interest in determining whether subconscious racial bias exists on the part of officials, it would make that call-specific data available subject to a confidentiality agreement, a tactic that is used all the time in litigation when competitors produce proprietary information to one another. My guess is that the NBA, in places that David Stern doesn't talk about at parties, will quietly increase the number of African-American referees over the next several years.

What interests me so much about the study, other than the fact that it jibes with my view of modern racism as being very subtle, is the fact that sports can be such a valuable ground for empirical analysis because of its defined outcomes. The results of a basketball game can be tangibly measured, so factors like racial bias can be measured with some degree of precision. This is why I quite enjoyed seeing Rush Limbaugh's demise as an NFL pundit. When Rush is "analyzing" political issues, he can get away with whatever he wants because of the imprecision with which most political issues are analyzed. When he brought his agenda to analyzing the NFL, he was exposed as a fraud within weeks because he went off on Donovan McNabb and his claim that McNabb was a media creation could be easily debunked with numbers. McNabb's merits can be reliably measured in a way that Barack Obama's cannot. This isn't to say that our analysis of the performance of sports figures is perfect (hence Michael Vick's multiple Pro Bowl appearances and Derek Jeter's multiple Gold Gloves), but it's certainly better than the measures of issues that, you know, actually matter.

[Update: John Hollinger makes a pair of good points in response, noting that: (1) the actual disparity in fouls is quite small; and (2) teams with more white players did well because the smarter teams picked up on the fact that European players were undervalued.]

I also enjoyed this opus by Jonathan Chait, a senior editor at The New Republic (and graduate of the University of Michigan), on the Netroots movement, which I view as similar to having Claude Lemieux on my team: their methods might not be totally ethical, but I like having them doing my dirty work for me because the other team has Ulf Samuelsson and if my team doesn't have a pest, then we're going to lose. I'm mainly thinking about the conclusion of the article, which describes how the Netroots have managed to put pressure on the mainstream media from the left that had never existed before.

I was also amused by this paragraph:

In point of fact, the most successful bloggers have been pulled into the warm embrace of the political establishment. Moulitsas consults regularly with influential Democrats in Washington. Presidential candidates hire popular bloggers or court them with private dinners. Last year, numerous top Democrats trekked to Las Vegas to attend YearlyKos, the liberal blog convention, where they sucked up to the attendees as relentlessly as if they were software executives. The climax of the proceedings was a party for bloggers thrown by then-presidential hopeful Mark Warner, costing more than $50,000 and featuring chocolate fountains. None of these things, however, have softened the netroots' sense of grievance and exclusion.

I want to know when I'm going to get to nosh from a chocolate fountain purchased by The Orgeron at Swindlepalooza in Tuscaloosa.


Grandy said...

Wait a second, if the people doing the study lacked the intimate details - which ref called which fouls on which players - I think there might be a problem with their conclusion.

All we have for each game is:

1. Player Name
2. Number of Fouls on #1
3. The ethnic breakdown of the officiating crew

Correct? Those results can certainly suggest correlation, but there's no way to account for the Joey Crawford's of the world - referees who are just freakishly whistle happy. I think it's likely than on more than a few occasions, fouls whistled by individual refs will be very skewed. And this sort of causes problems with the draw conclusion, I think. No amount of regression analysis is going to account for this, to the best of my knowledge.

Also, did I read wrong or was this done strictly on a white/black basis? Isn't there a problem with that? Other minorties play in the NBA, even if there aren't many of them. Are other minorities referring as well (I don't know)?

I don't think it's an absurd conclusion per se, though I'm no longer sure racism is the right term to apply here. Any number of environmental and cultural factors probably contributed, over a lengthy period of time, to the formation of the brain that's making these spot judgements. People trust what they know, and someone who grew up mostly around white folk may well be subconsciously inclined towards other white folk in some situations.

It's prejudice, sure. But I think this is an entirely different animal than the institutional and cultural racism that has existed in this country. The people of Springfield have never trusted the people of Shelbyville (and vice versa), and that's kind of an ancient story.

Michael said...

A couple responses:

1. The paper was peer reviewed, so professors with no stake in the matter and a lot more statistical knowledge than you or me have apparently vetted it and given the thumbs up.

2. A Joey Crawford would not change the results, as his whistle-happy nature could be affected by racial bias. Alternatively, it isn't affected and his whistles dilute the overall results.

3. The survey threw out the results for a few players who are not obviously black or white. I'm assume that they did this strictly on skin color, so a white-looking Latino would be white and a black-looking Latino would be black. The point was to measure the impact of subconscious bias as based on skin color.

4. I agree that "racism" is probably not the right term, as that assumes some sort of ill-intent. (On the other hand, WEbster's disagrees with me on the intent piece, as they define racism as the belief that one's own ethnic stock is superior to another.) "Subconscious racial bias" might be a better way to put it. In any event, it is different than other forms of racism, but I'm very interested in the studies that have started to analyze subconscious racism. As the article points out, that's a growing field of inquiry and more relevant to modern racism, in my mind.

Grandy said...

Some good points. Rebuttal:

1. The paper has not been formally peer reviewed, as per the article. I'm not seeing where there was mention that it had been informally peer reviewed.

I wouldn't dismiss an informal review, but I wouldn't put credence in it necessarily.

2. On the contrary, a Joey Crawford (or more to the point, several JC's, or just a crappy ref team) *might* change results significantly, and yes it could do so either way. I think conclusions should take this into consideration.

3. I'm not sure I'm comfortable with this, though it's likely not a big issue. I don't think it invalidates the study all by itself or anything. But I also think that this is more complex than a simple white versus black issue, even if the non white, non black segment make up a small portion of the NBA (which is true, as far as I know).

4. I probably should have said "connotatively".

The reason I don't like using that term is because now we're getting into issues like cultural background and how that has shaped people's perceptions and ability to process data. It would seem to be this is a signifcant issue at play. It's probably impossible to thoroughly explore (well, it is impossible as is because we need to know identities before we can do it), and I don't hold that against the authors.

I agree it's an interesting field of study. I think we're doing a disservice if we just drop the vanilla racist tag on top of it though. Maybe I'm too caught up in semantics, but it seems to me the issues at work here are significantly different than they are when dealing with the guys int he local Clan Chapter.

5. It occured to me last night - I'm not sure how significant these results are. I haven't decided - I'll try to crunch some numbers today. I'm not saying their insignificant, mind. But at a glance, it seems like we're not talking about a huge different in foul numbers. It certainly *could* be a factor in a given game.

also. ..

6. The authors couldn't do this, so it's another for the "would have been nice" column, but wouldn't we ideally be differentiating between types of fouls? Even if it's just an early game/late game differentiation, and maybe traciking intentional & technical foul calls (since they both carry harsher penalties). And maybe looking at categories like "calls in the last two minutes of close games". It's the sabermetrician in me talking, I guess.

Just looking at foul calls is ultimately unsatisfying for me. But it isn't like we can do anything about that.

7. Last for the "if only" column, but it seems to be one area that would be most beneficial to look at if we had the data is the ages of the refs making the calls.