Posted by Allanfan20:

the league is like 75% black or something like that. It's so obvious that more black players are gonna get calls against them.

the study wasn't about tallying *more* or *less* calls against black versus white players... the key variable was racial composition of officiating crews and whether more calls were made against the same group of players in those 2 or 4 cases (not sure which makeups they considered).

And Masterplan, it's like you said, when conducting a statistical experiment, it's ideal to take a much smaller sample size, how the hell did they go out and use every single game and call as their sample size. It's virtually impossible.

whoa, that's not what i said at all! bigger is definitely better for a sample size in statistics, the ideal is definitely 100%, which they seemed to be going for over that time period... and i'm guessing their claimed numbers were not impossible either... Excel is a wonderful tool, and a few industrious research assistants can cover an awful lot of ground... i'd know, been there ...

If you really really want to conduct a valid study on this, you have to make the league 50% black, 50% white and same with the referees. However, b/c it isn't, and b/c the majority of the league players who actually plays is so overwhelmingly black, you can't conduct that kind of study. A valid study has to be random, and you can't call this random b/c of what the majority of the league is.

i think that's incorrect for a quantitative study... big enough numbers reduce the need for perfect test conditions. while a minority of the league is white, i expect enough players over 13 years have been called for enough fouls to make the method legit, if not necessarily the conclusions.

Posted by tomverve:

Posted by Bippity10:

The best players in the game tend to be black. The guys that get to the line tend to be the best players. The guys defending the best players tend to be black. Just another thought.

That doesn't explain why the race of the referees would factor into things.

Its not difficult, however, to conceive of factors other than subconscious racial "snap judgments" that would differentially affect the foul rate by black and white crews. While the result is "statistically significant," the result itself is miniscule and opens the door for a multitude of confounding or effect modifying factors. Knowing the rather tiny difference that they detect, did they go back and use recursive partitioning to hunt for unknown confounders? Did they look at any of the game tape to obtain an empirical confirmation? From this article and other reports, it looks like a scientific nightmare to me. Here's a quick and dirty list of potential unaccounted factors:

Were the referees controlled for how tightly they call the games? Perhaps the white refs, on average, call a tighter game. Since I would a priori predict more physical encounters per player per minute between black players, such a disparity in calling the game would cause a difference in foul rate due to benign motivations. In other words, black players get called for fouls at a relatively higher rate in a tightly called game due to their tendency towards more physical play.

Were the white/black referee groups personality controlled? Its easy to imagine a disparity in the average "temperament" , in turn, affecting player judgment on the court differentially between white and black players. While the difference in player judgment may be racially motivated, it is a completely different measurement than the snap judgment of officials. And it quite clearly highlights the imprecision of the study.

Was the age of the white/black referee crews controlled? Perhaps white referees are older on average and, hypothetically, maybe white players get along better with older people than black people. And, as a result, white refs have better relationships with white players due to an age confounder. In this scenario, age and not race precipitates the foul rate disparity.

Lets assume for a moment that black refs are more likely to have better personal relationships with other black players and white refs are more likely to have friendly relationships with white players. Even then, that is an aspect of racism which the study doesn't purport to measure. It has nothing to do with "snap judgments" and, unfortunately, the drawbacks of their flawed experimental model are brought out again.

Those (quite obvious) confounding factors took me 30 seconds to come up with. There are a plethora of other compelling effect modifiers. Many of these problems are the result of a study that isn't blinded. This is an absolutely key point and the major reason for the imprecision of the study. It allows for an interplay between referees and the players - and the outcome variable is no longer measuring a specific phenomenon but, instead, a plethora of garbage. They also provide no form of empirical verification. In other words, they should have (but didn't) look a smaller sample of game tape (much smaller than 600,000 obviously) and look at the distribution of bad calls. This is necessary in order to have an inductive argument to supplement your statistical analysis. So, tomverve, even though the volume of data is impressive, the template of the study is on the grade school level and incredibly poorly conceived. The study is not generalizable in the least and, so, it becomes a rather useless piece of literature on 90s NBA basketball.