Methodology

Print More

unequal-justice-black-horizontal

We analyzed 5.8 million criminal and violation cases comprising 8.4 million charges filed in Oregon’s 36 counties and focused on 5.5 million charges filed between January 2005 and June 2016. 

The Oregon Judicial Department did not provide sentencing data or complete dates of birth. Counties did not universally fill in the race and ethnicity fields, limiting our ability to identify disparities between non-Hispanic white defendants and defendants of color. The data required extensive standardization to correct for variations of data entry.

Disparities

When comparing the demographics of defendants to the demographics by regions where they reside, we omitted cases outside of Oregon, and those where the city and ZIP code did not match those listed by the U.S. Postal Service and Oregon Bluebook. The counties did not uniformly list complete addresses, and existing ZIP code and city data required extensive standardization.

To compare the demographics of defendants to the demographics by geographic regions where they were charged and where they reside, we used the five-year American Community Survey demographics and housing estimates for 2014.

In categories where at least 90 percent of defendants could be identified by race or ethnicity, we calculated disparity rates using a relative rate index, which allows for comparison in the rate of charging between defendants of color and non-Hispanic whites. 

To estimate rates of unauthorized immigration and its impact on disparities, we used research by Jeffrey S. Passel and D’Vera Cohn of the Pew Research Center, reports by the Migration Policy Institute, and data from the U.S. Census on Hispanic citizenship status.

In our mapping, we compared charging data to five-year American Community Survey demographic and housing estimates for 2014. Map projections are NAD 1983 – 201, Oregon Statewide Lambert Intl Feet; Background Imagery World Terrain – Esri, USGS, NOAA.

To compare the financial impact on people of color compared to white people charged or ticketed in Oregon, we combined the total fines and fees associated with each case in which a financial charge was levied. We compared cases having a single defendant and single charge, as well as cases having a single defendant without controlling for the number of charges.

Latino defendants

To more accurately identify Latino coded solely as “white” or omitted from racial or ethnic coding, we employed a methodology developed by epidemiologists using Hispanic surnames. We obtained a table of common Hispanic surnames culled from a 1990 Census Hispanic surname list, and refined in the subsequent decennial Census, from Francis Boscoe, research professor in the Department of Epidemiology and Biostatistics in the School of Public Health at the University of Albany, State University of New York. The list is referenced in the 2013 journal article “Heuristic Algorithms for Assigning Hispanic Ethnicity.” Of the 151,671 surnames included in the 2000 Census that occurred 100 or more times, we used names 1 through 6,531, which ensures 90 percent of assignments are correct and identifies 84 percent of the Latino people in a dataset. These results have been found strongest in regions, such as Oregon, where most Latino residents are of Mexican and Central American descent. 

Generating Racial Differences in Charges

To compare the relative impact of the system on different racial groups we created a series of Relative Rate Indexes (RRI) for specific crimes and grouping of crimes. RRIs have become a common way to represent racial and ethnic disparities. The system was first advocated for by the Office of Juvenile Justice and Delinquency Prevention (OJJDP) and has been applied to a variety of criminal justice areas. The generate each RRI was a two-step process.

Step one began with the creation of a series of odds ratios. For each racial group an odds ratio was created that outlined the relative difference between the percentage of cases that were recorded for a specific racial group compared to the expected percentage given the proportion of that racial group in the general population. For example, if the percentage of black residents charged with a specific crime was 21% in Multnomah County their relative odds would be 3 given that this percentage was 3 times the percentage in the general population, which was 7 percent.

Step two involves the dividing the odds ratio for one racial or ethnic group by another. For example, if we also knew that 40% of the people in Multnomah County charged were white for the same crime that 21% of those charged were black we could conclude that black residents were 6 times (or 500%) more likely to be charged with this offense. This is because white residents would have a relative odds of about .5 or about ½ of the charges we would expect. Thus 3 divided by .5 produces an RRI of 6. This measure captures both the over representation of black residents and the underrepresentation of white residents.

One thought on “Methodology

  1. Kudos to InvestigateWest and Pamplin Media Group for giving their “Methodology” a prominent part of their reporting. “Data Manipulation” is much more common than “Fake News” AND much more sinister and destructive. TRANSPARENCY at its best!