Monday, January 23, 2012

Brief Report on Thomson Reuters Mississippi Prison Status Data

EDITOR Because the Editor of this blog has a keen interest in empirical study of the pardon power, state and federal, it seemed worthwhile to try to look past some previous negative experiences with authors and ask the statisticians at the University of Georgia about their recent analysis of clemency in Mississippi. As expected, they were both very generous with their time and quite open about the details of their effort. So, an additional, brief summary was requested, and they have graciously provided it here (below). The Editor is quite thankful and has renewed faith in the collegiality of scholars.

Prepared for the UGA SCC by Dr. Kim Love-Myers and Dr. Jaxk Reeves:

Analysis: 


We were provided with the data below, regarding the current prison status of 25,480 individuals in Mississippi.

Of the 222 on the pardons list: 142 are white
                                                   68 are black
                                                   12 are unidentified

Of a total population of 25,258: 8572 are white
                                                 16386 are black
                                                    222 are Hispanic
                                                      31 are Native American
                                                      37 are Asian
                                                      10 could not be identified

The 22 individuals who could not be identified were removed from the data, as knowledge of their races was not available.

The emphasis for this analysis was on the disparity between white and black individuals, and so the SCC considered only those individuals in this population who are described by either of these terms. Table 1 below provides the distribution of individuals described as white or black according to whether they are pardoned or currently serving a prison sentence.

Table 1. Prison Status of Mississippi Individuals by Race (White and Black Only)

Table of Prison Status by Race


Table of Prison Status by Race
Status
Race
White
Black
Total
Pardon
142
68
210
1.63%
0.41%

Prison
8572
16386
24958
98.37%
99.59%

Total
8714
16454
25168


From this table, 142 out of 8714 white individuals in this population are pardoned (1.63%); also from this table, 68 out of 16454 black individuals in this population are pardoned (0.41%). The odds of being pardoned for a person described as white are 3.99 times the odds of being pardoned for a person described as black (Footnote 1).  A Fisher's exact test to determine whether Race and Prison Status are independent of one another determined that the probability (P-value) of a population where Race and Prison Status are unrelated producing data with a difference in the two populations this large or larger is 2.14E-22, or 0.000000000000000000000214 (Footnote 2). This is a one in 4,664,179,104,477,610,000,000 chance — much less than a one in a million possibility.

Important Notes

Odds. Note that "odds" are calculated as the probability of an event (such as a pardon) occurring for an individual in a particular population divided by the probability of that event not occurring. This can be written mathematically as p / 1-p, where p is the probability of the event occurring. Therefore, when it is said that the odds of a pardon are 3.99 times higher for whites than for blacks, technically this is not completely synonymous with stating that the probability of a pardon is approximately 3.99 times higher for whites. (In this case, the calculations are not much different, however; the probability of whites being pardoned compared to blacks is 3.94 times higher).

P-value and chance of occurrence. Note that the P-value as listed here (i.e., chance of this distribution occurring randomly) do not take any additional factors beyond race into account. In other words, these analyses compare the probability of pardons for whites to blacks with the underlying assumption that these pardons should be randomly distributed among the races. In reality, there are many other issues to consider and pardons should probably not be randomly distributed. Other factors, which may very well be related to race (not necessarily causally) and influence the rate of pardons are the severity of the crime committed, time spent in prison, number of previous offenses, gender, or any number of additional factors. This analysis does not account for such factors. In particular, it would be wrong to state that this analysis "proves" that race is the "cause" of different rates in pardoning; this type of observational information cannot prove causality, though it does indicate a significant relationship between race and the probability of being pardoned.


[1] See notes section for important information regarding odds.
[1] See notes section for important information regarding this P-value.

No comments:

blogger templates | Make Money Online