Exonerations in the U.S.

Analyzing race, sentence severity, and time between conviction and exoneration

Project Marvelous Hitmontop
Caleb Chin, Khushi Patel, Alexa Prague, Derrick Chia, Cynthia Lei

5/5/23

Topic and motivation

  • The criminal justice system in the United States has been under scrutiny for its disproportionate impact on minority communities.

  • We want to investigate:

    • how race affects the severity of the sentences exonerees receive

    • how length of time between conviction and exoneration, severity of crime, and sentence are related

Data introduction

  • Observations represent all known U.S. exonerees between 1989 and 2023.
  • Data was collected by The National Registry of Exonerations
  • The data fields consist of exoneree information such as name, race, and sex as well as information about their exoneration.
  • Data has several subpopulations

Highlights from EDA

race num prop
Asian 32 0.0097442144
Black 1724 0.5249695493
Black;#White 1 0.0003045067
Don't Know 7 0.0021315469
Hispanic 400 0.1218026797
Native American 22 0.0066991474
Native American;#White 1 0.0003045067
Other 19 0.0057856273
White 1078 0.3282582217

Asian, Black, White, Hispanic, and Native American exonerees appear the most in the data set

Native American and Black exonerees often spend longer between conviction and exoneration than exonerees of other races.

Predicting sentence severity with race

Significance of results at 5% significance level: Hispanic Exonerees vs White Exonerees

Time falsely convicted, race, and sentence severity

Significance of results at 5% significance level: Black Exonerees vs White Exonerees, Sentence Severity

Conclusions + future work

  • Hispanic exonerees have a stat. significant difference in predicted probabiliy of getting severe sentence than White exonerees
  • How about the rest of the exonerees?
  • Black exonerees have stat. significant difference compared to White exonerees in expected years between conviction and exoneration, holding sentence severity constant.
  • What does this imply?
  • Key Limitation: data only represents those who have already been found innocent, we cannot extrapolate to all falsely convicted people.
  • Future work:
    • Transforming sentence into numeric field to examine how sentence length (in years) is associated with race and/or other factors