Postby **take5_d_shorterer** » 24 Sep 2017, 21:51

Two comments.

1) The first is that this seems like an example of changing one problem into another. The crucial decision here may be in what order the jury candidates are interviewed.

If, for example, there was suspicion that the prosecutor wanted to exclude black jurors, it would make a big difference whether black jurors were interviewed after white jurors, before, or interleaved.

If they were interviewed after, then assuming they knew Batson v. Kentucky, they could respond in similar ways to the white jurors. In this case, it would be difficult to make the claim that their answers were the cause of their dismissal. (That doesn't mean that the prosecutor couldn't find other reasons such as job occupation to dismiss jurors, but they have no control over what jobs they had prior to being selected.) If the black jurors were all interviewed before the white jurors, then they would have no idea necessarily how the white jurors would answer questions. I'm assuming here that the jurors hear how other jurors answer questions, which may or may not be true. Speaking of that, do you know if jurors in big cases can hear how other jurors are answering questions.

The thing is that what one would expect is that black and white jurors would be interviewed in an order in which both were interleaved. If the jurors were chosen at random, it would be highly unlikely that one ethnicity would be chosen before another. You can estimate the probability with a Monte Carlo simulation.

2)The second is it should be fairly easy to look at whether there is any unusual bias in who is selected initially to be in the potential pool of jurors. This is a fundamental problem in statistics.

Let's say that in a state, there are 1 million registered voters, 500,000 of whom are white and 500,000 are black.

You create a pool of, let's say, 50 possible jurors. Of these 32 are white and 28 are black.

How likely is it that a random selection of voters would result in having at least 32 white candidates? (You could also ask the question, what is the probability that there would be at least 32 white candidates or at least 32 black candidates.)

Using the tool that the normal curve does a pretty good job of approximating the binomial distribution, you look at the expected number of white candidates, which is 50 times (the probability that jury candidate is white, i.e, 0.5). That's 25.

The standard deviation is square root of (n*p*(1-p)) where n=50 for 50 people, p = 0.5. That's about 3.53.

The actual number of white candidates is 32, which is 2 standard deviations away from the expected number of white candidates, which is 25.

7 is approximately 2 times 3.53.

The probability that you would have at least 32 white candidates is not that much, about 2.5 percent.

You could also run a Monte Carlo simulation to get an estimate. Because of computing speed, you could easily run millions of simulations and come up with a pretty good estimate in a few seconds.