Across the 48 states and the District of Columbia where felony murder remains a punishable offense (all states except for Hawaii and Kentucky), there is a profound lack of transparency and consistency in the data.

1. For a small handful of states (e.g. Connecticut, Maine, Wisconsin), we were able to identify felony murder through a unique statute in their correctional data. For other states, we had to comb through legal definitions and statutory caveats to find a description or code that referred to felony murder.

2. Yet, for another group of states, we used records from the courts or other criminal justice agencies to get a fuller picture of how many people are charged or convicted under the felony murder rule. This provides an imperfect measure, as we are merging court records with correctional rosters.

3. In the plurality of states, we were not able to obtain any data that would illuminate the true impact of the felony murder rule.

---

Despite these hindrances and nuances, our research and analysis has yielded key findings, including these:

In order to determine racial disparities, we applied the Bayes’ theorem to calculate each person’s likelihood of being incarcerated for felony murder, given their race. According to the Bayes’ theorem, in order to determine the probability of A given B, one must know the probability of B given A, the probability of A, and the probability of B.

For the purposes of our analysis, we are interested in the probability that, given someone is of a particular race, they are incarcerated for felony murder. Thus, our equation for each race is as follows:

In this case, we know the probability that given you are incarcerated for felony murder, you are a certain race (the count of felony murder convictions of that race over the total count of felony murder convictions). We also know the probability that someone is incarcerated for felony murder (the count of felony murder convictions over the total population of the state).

Lastly, we know the probability that a given individual is of a particular race in a state (demographic breakdown of the state’s population). The data on each state’s population and racial breakdown was obtained through the United States Census Bureau’s 2022 estimate. Once we had these probabilities, we could compare each ethnicity’s likelihood with that of white residents of that state. We recognize there are a diverse and contested array of ways to calculate racial disparities in conviction data, and settled on this avenue in consultation with data experts and legal analysts.

To identify county-level racial disparities, we highlighted the counties where there were more than ten felony murder convictions and over 75% of those convictions were Black individuals (for certain states, we used a lower threshold).

As part of our analysis, we sought to uncover what the median age at offense was for all crimes and compare that to the median age at offense for felony murder. In order to calculate these values, we needed the individual’s date of birth and the date of their offense.

Once we obtained these values, we could find the difference between these dates in terms of years, which would give us their age at offense. For states that did not report an offense date, we chose to analyze age disparity through age at admission or age at sentence. If a state did not report date of birth but did report age, we could still calculate how old someone was at the time of their offense.

Each state had a different method of logging their sentence lengths. Some states logged their sentence lengths in year quantities and used decimals to represent months or days added to a sentence. Other states logged sentences separated out by unit such as “10 years, 6 months, and 0 days.” For all states, we normalized sentence lengths to years so, for example, the previous sentence would be represented as 10.5 (years).

Some states had only one column for sentence lengths while others separated into maximum and minimum sentence length. For states with both a maximum and minimum value, we opted to represent the maximum sentence. The language in our analyses is tailored to fit this decision, using phrases such as “up to” to represent that some states’ calculations are based on maximum sentence lengths.

Our data analysts had to navigate significant flaws in the way many states collected and shared their data; the methods described above summarize our best efforts to work with this complex terrain and devise ways of obtaining transparency and reportorial insight in light of those challenges, while recognizing that there are a range of methodological approaches that might be used.