Framing For Light Instead Of Heat
I.
Ezra Klein uses my analysis of race and justice as a starting point to offer a thoughtful and intelligent discussion of what exactly it means to control for something in a study.
I’m not really going to call it a critique of my piece, because it only applies to two of the six areas I looked at, and in those two areas Klein’s thoughts were already carefully integrated into my conclusion – I described both as showing “ambiguity over the level of racial bias, depending on…how strictly you define racial bias.” The Vox article repeated and expanded on that conclusion rather than contradicting it.
But it’s still an important issue and I’m glad it’s come up since I didn’t have time to deal with at enough length in the original post.
The argument is: any study worth its salt is going to control for things like income level. Therefore, a study that concludes “blacks and whites get arrested at about the same rates” may only mean “blacks and whites of the same income level get arrested at about the same rates”. If blacks on average have lower incomes, then in the real world blacks might still be arrested much more. Blacks being poor and therefore getting uniquely poor treatment from the criminal justice system (Klein says) sounds like exactly the sort of thing we would call “racial discrimination” or “racial bias” or “racism”, but it would be totally missed by the standard methodology of controlling for income.
The solution is terminological rigor, which I foolishly forgot to have. What I should have said at the beginning of my post was “I want to know whether there is any direct bias against black people caused by racist attitudes among police and other officials.” By this definition, all of my conclusions stand.
Klein wants to know whether there is any factor at all that causes disproportionate impact of the criminal justice system on any race. By this definition, my conclusions are only a tiny part of the picture, although at the end I recommend the book Malign Neglect
As long as we keep these two meanings of “racial bias” or “racism” or whatever separate, there’s no problem. Once we start conflating them, we’re going to become very confused in one direction or another.
Ezra Klein and I don’t disagree about any point of statistics. What I think we do disagree about is the terminology.
If we find that much of the overrepresentation of blacks in the criminal justice system is because black people are often poor and poor people often get sucked into the system, should we describe this as “the problem isn’t racism in the criminal justice system, it’s poverty” or as “the problem is racism in the criminal justice system, as manifested through poverty”?
II.
Consider a town with 1000 black people and 1000 white people. 750 black people are poor, and 250 are rich. 750 white people are rich, and 250 are poor. Everyone commits crimes at the same rate – let’s say 10% per year. Rich people have lots of connections and can bribe their way out of trouble in a pinch, so only 50% of rich criminals get arrested. Poor people don’t have any strings they can pull, so 100% of poor criminals get arrested.
We can do the calculations and determine that the black arrest rate will be 8.75% and the white arrest rate 6.25%, a pretty significant difference. The people in the town can do the calculations as well. They correctly observe that in their town, everyone commits crimes at the same rate, so there must be some bias in their system. Using Klein’s definition, they determine that since the system in their town disproportionately affects blacks, their criminal justice system is racist.
The problem is, upon learning that your criminal justice system is racist, what solutions come to mind? The ones I think of include things like increasing the diversity of the officer pool, sending police to diversity training, ferreting out racist attitudes and comments among members of the force, urging officers to consume media that is more positive towards black people, et cetera.
But all of these are unrelated to the problem and will accomplish nothing. We specified the decision algorithm these officers use, and we know it has nothing to do with race and everything to do with class. The townspeople should be attacking the culture of bribery, nepotism, and corruption, not throwing away resources on curing racist attitudes that don’t affect police behavior in the slightest.
Note that this is true even if the poverty is caused by racism. Suppose the town college unfairly admits whites and turns down blacks, which is why the white people in this town are so much richer. I have no problem with saying “the town college is racist”. This suggests the appropriate solutions – educating and/or punishing the people at the college. I have a lot of problems with saying “the town police are racist” as a shortcut for “the town police take bribes, and due to racism somewhere else the people with the cash are all white” because this obfuscates the correct solution.
You can’t just cut links out of a causal chain and preserve meaning. “Blacks are arrested disproportionately often because of gravity” is true, insofar as without the formation of the Earth from the gravitational coalescence of a primordial gas cloud humans and therefore racism wouldn’t exist. But if the natural reaction to hearing the phrase is to solve the problem by attaching hundreds of helium balloons to black people, then say something less misleading.
III.
Klein goes on to say:
An example is research around the gender wage gap, which tries to control for so many things that it ends up controlling for the thing it’s trying to measure. As my colleague Matt Yglesias wrote, the commonly cited statistic that American women suffer from a 23 percent wage gap through which they make just 77 cents for every dollar a man earns is much too simplistic. On the other hand, the frequently heard conservative counterargument that we should subject this raw wage gap to a massive list of statistical controls until it nearly vanishes is an enormous oversimplification in the opposite direction. After all, for many purposes gender is itself a standard demographic control to add to studies — and when you control for gender the wage gap disappears entirely!
The question to ask about the various statistical controls that can be applied to shrink the gender gap is what are they actually telling us,” he continued. “The answer, I think, is that it’s telling how the wage gap works.
Take hours worked, which is a standard control in some of the more sophisticated wage gap studies. Women tend to work fewer hours than men. If you control for hours worked, then some of the gender wage gap vanishes. As Yglesias wrote, it’s “silly to act like this is just some crazy coincidence. Women work shorter hours because as a society we hold women to a higher standard of housekeeping, and because they tend to be assigned the bulk of childcare responsibilities.”
Controlling for hours worked, in other words, is at least partly controlling for how gender works in our society. It’s controlling for the thing that you’re trying to isolate.
Once again, when someone says “women make seventy seven cents for each dollar a man earns”, the response is almost always “That’s outrageous!” and demands that companies stop being so sexist. I don’t even have to speculate here. Google “gender wage gap”, and just on the first page of results you find statements like:
“While some CEOs have been vocal in their commitment to paying workers fairly, American women can’t wait for trickle-down change. The American Association of University Women urges companies to conduct salary audits to proactively monitor and address gender-based pay differences.”
“Our project on sex and race discrimination in the workplace shows that outright discrimination in pay, hiring, or promotions continues to be a significant feature of working life…the Institute for Women’s Policy Research examined organizational remedies such as sexual harassment training, the introduction of new grievance procedures, supervisory training or revised performance management, and reward schemes.”
“Today marks Equal Pay Day, the date that symbolizes how far into the new year the average American woman would have to work to earn what the average American man did in the previous year. With a new executive order issued today, President Obama and Democrats are hoping to peg the gender wage gap as a major issue ahead of the 2014 elections. This week, Senate Democrats also plan to again bring forward the proposed “Paycheck Fairness Act,” a bill that aims to eliminate the pay gap between female and male employees. Both men and women see a need for moves such as this – 72% of women and 61% of men said “this country needs to continue making changes to give men and women equality in the workplace”
Given that the supposed gender pay gap is being used at this very moment to argue for salary audits, sexual harassment training, grievance procedures, and paycheck fairness acts, isn’t it really important to know that a lot of it is due to upstream factors like how men and women are socialized as children to have different values, which wouldn’t be affected by these things at all?
(Given that the entire issue is probably being used to load the term “feminism” with positive affect, isn’t it important to know that it’s mostly unrelated to what we expect feminists to do with their extra trust and power?)
It might be worthwhile to come at this from an ideologically opposite angle. Suppose I state “Professors who identify as feminist give twice as many As to female students as they do to male students.”
(This is true, by the way.)
It sounds like a big problem. So you dig through mountains of data, and you figure out that most feminist professors tend to be in subjects like the humanities, where twice as many students are female as male, and so naturally twice as many of the As go to women as men. If I just give you my best trollface and say “Yes, that’s certainly the mechanism by which the extra female As occur”, you have every reason to believe I’m deliberately causing trouble. Especially if colleges have already vowed to stop hiring feminist professors in response to the subsequent outrage. Especially especially if you know I am a cultural conservative activist whose goal has always been to make colleges stop hiring feminist professors, by hook or by crook.
If twice as many women as men take English literature classes, that’s compatible with something about gender socialization unfairly making men feel less able to study or less enthusiastic about studying literature. That could be considered biased or discriminatory, I guess. But phrasing it as “feminist professors give twice as many As to women” is calculated to produce maximal damage. It’s the sort of thing you would only do if you wanted to throw a match on a gunpowder keg for s**ts and giggles.
IV.
So I guess I’ve moved on from “poor choice of terminology” to “active misrepresentation”. Let’s stick with that.
This issue makes for the ultimate motte-and-bailey doctrine.
You go around saying “Gender gap! Women making less than men! Discrimination! Sexism!” Everyone puts on their Gricean implicature caps and concludes that they mean what these words mean in everyday speech. The appropriate remedies are trotted out – companies need to raise their female employees’ pay, companies need to hire more discrimination officers, feminists need to talk more about all the ways men talk over women in the workplace and mansplain to them, etc. This is the bailey.
Then someone says “Wait, according to our study, a lot of this is just that women prefer working shorter hours to have time with their families” – and so they retreat to their motte: “Yeah, that’s the mechanism for the gender gap. You mad, bro?”
But the thing about mottes is that nobody actually cares about them when there’s this awesome bailey they can fight over instead. By turning differential socialization into the motte for sexual harassment or something, we’re doing a disservice not only to sexual harassment, but to the principled study of differential socialization.
Anyway, the situation is actually even worse than this. If you hear “The problem with the criminal justice system is disproportionate impact on the poor,” then you’ll probably start coming up with ideas for how to deal with that, and other people will probably start listening. If you hear “the problem with the criminal justice system is racism,” then you will start sharpening your knives.
Racism is a uniquely divisive issue. Minorities hear it and think of Klansmen trying to kill them. White people hear it and think of witch-hunters trying to get them fired. A single death in a random Midwestern town has turned half the country into experts on ballistics because it involved race. Bring up race, and people will change their opinion in the opposite direction suggested by the evidence just to spite you for having a different opinion about it than they do.
Once you’ve said words like “racism” or “racial bias”, this dynamic is already in play and you have lost control of the conversation from then on. If you mention the word and then suggest that we should do something about the police bribery or whatever, then ten percent are going to yell “HOW DARE YOU IMPUGN OUR OFFICERS’ HONOR, YOU POLITICALLY CORRECT FASCIST”, another ten percent are going to yell “HOW DARE YOU DERAIL THE CONVERSATION ABOUT RACE, YOU WHITE SUPREMACIST ASSHOLE”, and the other eighty percent are going to be yelling so loud at each other they can’t even hear you. By the time all the fires have been put out and all the rubble cleared, it’s a pretty good bet that nobody is in the mood to hear about policy ideas for reducing the impact of police on lower-income individuals anymore.
Klein ends his piece by interviewing a professor who states that “Liberals sometimes overstate the extent of overt racism as a direct explanation of justice system disparities.” He acts like this is some sort of inexplicable quirk of the liberal mind. I wonder whether it might have more to do with liberals reading things like the recent Vox article, “America’s Criminal Justice System Is Racist”, which declares the thesis “There is no reason to be subtle on this point: the American criminal justice system is racist”, then goes on to repeat the phrase “America’s criminal justice system is racist” five times in the next five paragraphs. It never mentions that possibility that any of this racism is anything but overt.
If, like Robin Hanson, you believe in the metaphor of tugging policy ropes sideways, then I can’t think of any worse way to ensure that everyone will be tugging against you in every direction than trying to focus the discussion about race.
That’s why I limited my review to direct bias within the justice system itself, and why I think other ways of framing the issue are less productive.
(Comment screening is on again, I guess. Comments that will start flame wars or derail the conversation will vanish into the aether. Unrelated: the book review yesterday got popular and this blog might go down every so often because of too much traffic. It’ll be up again shortly.)