Fair cross-section

# Fair cross-section

Abstract: The unusual circumstances of Curtis Flowers’ trials make it possible to estimate the probabilities that white and black jurors would vote to convict him, 98% and 68% respectively, and the probability a jury of his peers would find him guilty, 15%.

#### Background

Curtis Flowers was tried six times for the same crime. Four trials ended in conviction; two ended in a mistrial due to a hung jury.

Three of the convictions were invalidated by the Mississippi Supreme Court, at least in part because the prosecution had excluded black jurors, depriving Flowers of the right to trial by a jury composed of a “fair cross-section of the community“.

In 2019, the fourth conviction was invalidated by the Supreme Court of the United States for the same reason. And on September 5, 2020, Mississippi state attorneys announced that charges against him would be dropped.

Because of the unusual circumstances of these trials, we can perform a statistical analysis that is normally impossible: we can estimate the probability that black and white jurors would vote to convict, and use those estimates to compute the probability that he would be convicted by a jury that represents the racial makeup of Montgomery County.

#### Results

According to my analysis, the probability that a white juror in this pool would vote to convict Flowers, given the evidence at trial, is 98%. The same probability for black jurors is 68%. So this difference is substantial.

The probability that Flowers would be convicted by a fair jury is only 15%, and the probability that he would be convicted four times out of six times is less than 1%.

The following figure shows the probability of a guilty verdict as a function of the number of black jurors:

According to the model, the probability of a guilty verdict is 55% with an all-white jury. If the jury includes 5-6 black jurors, which would be representative of Montgomery County, the probability of conviction would be only 14-15%.

The shaded area represents a 90% credible interval. It is quite wide, reflecting uncertainty due to limits of the data. Also, the model is based on the simplifying assumptions that

• All six juries saw essentially the same evidence,
• The probabilities we’re estimating did not change substantially over the period of the trials,