A wise Latina

We like to think that in our form of government, political officials represent the citizens at large. Trouble is, it’s hard to know what “represent” means. Often, we talk about representation through two major lenses. “Descriptive” representation refers to whether people in government look like Americans generally (in terms of race, gender, maybe even age, occupation, class). “Substantive” representation refers to whether people in government make the sorts of decisions that Americans generally would make.

In a recent article, Bonneau and Rice take those two concepts into the world of judicial politics. Their basic question: Do black judges make different decisions than white judges? Bonneau and Rice provide a nice empirical answer to the question, but their interpretation of what their findings mean is confusing and less than persuasive.

Do black judges make different decisions than white ones?

Although there’s a large stack of existing articles asking the same question, previous work has produced mixed results. Bonneau and Rice argue persuasively that institutional context matters enough to resolve those mixed results in the literature.  Specifically, they argue that discretion matters. Some judges have discretion over which cases they hear; others do not. This contrast is most apparent when looking at state supreme courts. Some states have a three-tiered court similar to the federal system, with trial courts, intermediate courts of appeal, and then a state supreme court. Other states omit the intermediate courts of appeal. In either system, criminal defendants always have a right to appeal. In the absence of an intermediate court of appeals, then, the state supreme court has no discretion over which cases it will hear; it must hear all appeals.

Bonneau and Rice “contend that descriptive representation is translated to substantive representation when the court has discretion over their docket and thus an intermediate court of appeal is present” (p 3887).  Elsewhere, they state that “the link between descriptive and substantive representation relies on the expectation that minority and non-minority actors behave differently when faced with the same set of facts”–a curious statement that I discuss below, but one that I’ll accept for the time being. Putting it together, these two statements lead us to expect that race will have an effect (“minority and non-minority actors behave differently”) only when an intermediate court of appeal is present.

They test this argument by looking at the voting records of state supreme court judges, using a variety of appropriate controls (judge’s race, ideology, type of crime committed, etc). And what do they find? In states with an intermediate court of appeal, the judge’s race has no effect on voting; in states without an appeals court, race has a strong effect.

Wait–they find exactly the opposite pattern from what their theory implies. Didn’t they say that race should have an effect when there IS discretion (an intermediate court), not when there isn’t?

Perhaps I read too quickly, but nowhere did they acknowledge this curiosity or seek to explain it. We’re left, then, with an unexplained, unforeseen empirical finding. I know a fellow who likes to say, “The world is correlated at 0.3.” In other words, there are all sorts of correlations out there, but without a strong theoretical story explaining a particular correlation, we have no reason to suppose any particular finding is not random.

Bonneau and Rice are smart people, so I’m sure I missed something. But near as I can tell, the data did not support the theory.

Does substantive representation require that white and black judges vote differently?

Perhaps most curious about Bonneau and Rice’s article is this assertion, stated several times in different forms (including in the quotation given above):

“If there are no differences between white and African-American judges, then it also means that there is no substantive representation on the bench. After all, if African-American judges are deciding cases the same way as white judges, then neither group is representing the interests of minorities” (p 382).

Somebody help me understand why that needs to be true. Bonneau and Rice say nothing to back it up. Perhaps if black and white judges are deciding cases the same way, then BOTH groups are representing the interests of minorities. Must it be the case that only a “wise Latina” can make decisions about Latino defendants? Indeed, in their conclusion, Bonneau and Rice give evidence that would seem to indicate that both groups are reasonably fair to everybody:

“The vast majority of judges … are distinguished jurists with years of experience in the judicial system. These judges are also socialized the same way (both in law school and in the legal profession). This homogeneity in socialization and experience–largely unique to the judiciary–may serve to mitigate any racial differences that exist.”

I’m not trying to make an argument either way about whether the legal system is biased against minority defendants. I’m only saying that Bonneau and Rice need to back up their claim rather than merely assert it.

Parting thoughts

So where are we left?

• The authors find exactly the opposite of what their theory predicts–but they should have tried to explain this.
• The authors need to give further justification for one of their major theoretical assumptions: That substantive representation requires black and white judges to behave differently.

This problem is not unique to the states. Some countries have adopted a quota system to combat such underrepresentation. Obviously, if you reserve a quarter of Congressional seats for women, then you’ll end up with more women in Congress. But the question is, do these quotas actually help change the status of women in politics? What if we imposed a quota for a while and then withdrew it–would women see an enduring improvement?

Turns out we would.

The Indian Experiment

I finally got around to reading Bhavnani’s article, published last February, in which he exploits some fascinating data to show that even if you impose a quota for only one election, in the following (quota-free) election, women will still have much better odds of winning after the quotas are removed than they did before the quotas were put in place.

India uses a quota system to ensure representation of women in local offices. Prior to each election, 33% of local seats will be reserved for women; only women may run in these races. These reservations are made through a genuinely random lottery system.¹ So a seat might be reserved for women in one election but then open to anybody in the following election.

These rules create an honest-to-goodness natural experiment. Bhavnani looks at unreserved seats in the 2002 Mumbai elections. He compares those that were reserved in 1997 (the treatment group) to those that were not (the control). He’s not comparing reserved seats to unreserved seats; he’s comparing unreserved seats that were previously reserved to unreserved seats that were not previously reserved. This technique enables him to estimate the long-term effect of quotas after they are withdrawn.

Even after quotas are removed, women are five times more likely to win in post-quota seats than in other seats. Chew on that. Five times more likely.

Here’s some more comparisons between the treatment and control groups. In the control, only 3.7% of winners are female; in the treatment, 21.6% are. In the control, only 35.8% of wards had a female candidate bother to run; in the treatment, 73% did. In the control, only 4.4% of candidates were female; in the treatment, 11.9% were.

There are several reasons that (withdrawn) quotas could have these enduring effects. Bhavnani points to two as most likely. First: Women realize that they can win elections, so they keep running. Not just incumbents, though; you also get more rookie women choosing to run in the treatment than in the control. Something about seeing successful women makes other women realize that they can run. Second: Parties realize that women can win, so they start nominating women to represent them in the election.

Thoughts

Hillary's problem in 2000?

I have no methodological or theoretical critiques of this paper. That’s unusual, if you read my other posts. However, I do have lingering questions that I hope future research addresses.

Bhavnani shows that quotas continue to help women in the first election (only) after they are withdrawn. How long does it take (if ever) for these effects to disappear? Unfortunately, it appears that Bhavnani could not find data to test that. I’d love to see it though. A short term effect is interesting, but not the end of the story.

These quotas apply only to local elections. I’d be curious to look at all the women in state/national office and see how many got their start under (or immediately after) a quota. Does getting your start under a quota make you look like a weaker candidate when you run for office later? Recall that many thought Hillary Clinton had an unfair advantage winning her first Senate election, so she had to work very hard to prove herself as something more than Bill’s shadow. Do women who start out under a quota have a similar problem?