This study is based on a descriptive analysis and provides qualitative recommendations for improvement. The data in this study is used to describe the distribution of low level arrests in Minneapolis, MN along several dimensions including geography and race. With respect to race, the measure of racial disparity in arrests for low level offenses is calculated as the ratio of the Black/Native American arrest rate to the white arrest rate. So, for example, a Black/white racial disparity measure (or ratio) of eight implies that the rate at which Blacks in Minneapolis are arrested for low level offenses is eight times the rate at which whites are arrested. In other words, Blacks are eight times more likely than whites to be arrested for a low level offense. To calculate the arrest rate for Blacks, the total number of Black arrests for low level offenses is divided by the total Black population in Minneapolis; the corresponding ratio is then multiplied by 100,000 to obtain the arrest rate per 100,000 Black persons. The same procedure is used to calculate white and Native American arrest rates per 100,000. Arrests are either aggregated annually or over the entire study period to produce annual arrest rates per 100,000 or 3-year arrest rates per 100,000, respectively. Arrest rate per 100,000 is standard industry practice used by criminologist, researchers, law enforcement agencies, and federal programs such as the FBI Uniform Crime Reporting (UCR) Program. The reason rates are used (as opposed to total counts) is to allow for comparisons that take account of the fact that the population of Blacks and whites in Minneapolis differ.
This study uses data provided to the ACLU by the Minneapolis Police Department in response to a public records request.
Description of Data Set
The data covers all low level—petty misdemeanor, misdemeanor, and gross misdemeanor— arrests made by the Minneapolis Police Department from January 1, 2012 to September 30, 2014. The data was obtained by the ACLU in October 2014. Because the data does not extend all the way through 2014, it was necessary to weight the 2014 data when calculating annual statistics. Specifically, arrest counts for 2014 were multiplied by 4/3 to enable comparisons with arrest counts for 2012 and 2013—years for which a full year of data was available.
The data on arrests includes all charges brought against each individual arrested. Consistent with the FBI Uniform Crime Reports (UCR), this study “counts one arrest for each separate instance in which a person is arrested, cited, or summoned for an offense.”1 There may be multiple charges within one arrest. The data does not include information on whether a conviction was pursued or obtained by the relevant prosecutor’s office. Importantly, the dataset only contains arrests where the most serious charge is a gross misdemeanor. In other words, arrests where an individual was charged with both a low level offense and a felony offense are not contained in the dataset and not considered in this study.
The original dataset provided by the Minneapolis Police Department was organized by police officer-charge and contained 297,928 charges. Variables in the dataset include date and time of arrest, arrest location, arrestee address, age, race, and sex, charge type (specific offense), charge level (i.e., petty misdemeanor, misdemeanor, gross misdemeanor, and status offense), officer badge number, and disposition of arrest (e.g., cited, booked, hospitalized, and so forth). For youth arrest, data on home address is restricted by rule and was not provided.
Organizing the dataset at the charge level (rather than at the officer-charge level), the final dataset contained 130,930 observations. These observations correspond to 96,975 arrests. There are more charges than arrests because some arrests include more than one charge. Data on home address was only available for 84,467 arrests. Data on home address was not available for youths—there were 8,094 youth arrests. The remaining missing arrest data corresponds to homeless individuals and arrestees for whom home address data was unavailable. Of the 84,467 arrests for which home address data was available, 51,299 arrests are of individuals with a home address within Minneapolis city limits. Further, 73,332 of the 84,467 arrests for which home address data was available are of individuals who live in the Minneapolis-St. Paul-Bloomington Metropolitan Statistical Area (MSA). 1,948 arrests are of individuals who do not live in the state of Minnesota.
In addition, of the 96,975 total arrests, 88,869 were of adults (i.e., individuals age 18 or older). These adult arrests correspond to 71,094 distinct individuals. Of these 71,094 distinct individuals, 85.6% of adults arrested were only arrested once during the study time period. This study follows the FBI UCR convention and counts one arrest for each separate occasion where a person is arrested. “Because a person may be arrested multiple times during a year, [arrest figures in this study] do not reflect the number of individuals who have been arrested; rather, [they] show the number of times that persons are arrested.”2
The dataset contained various kinds of duplicate data. To start, there were a number of observations that contained the same values for all variables contained in the dataset. These observations were considered duplicates and were dropped. This creates an issue for individuals charged with an outstanding warrant. Specifically, the dataset only records the fact that the individual had an outstanding warrant and does not record the charge for which the warrant was issued (although we do know that the charge was not a felony). Thus, for any arrest with more than one warrant charge listed, it is impossible to distinguish if the individual had more than one outstanding warrant for the same level of offense—petty misdemeanor, misdemeanor, and gross misdemeanor—or if the individual had one outstanding warrant and this was erroneously duplicated in the dataset. Our analysis errs on the side of caution by assuming data duplication when more than one warrant charge for the same level of offense is listed for a given arrest.
Relatedly, it was not possible to separately distinguish police officers listed as “PRIVATE” in the dataset. These observations were not included when calculating officer-specific statistics, such as average number of arrests made per officer and booking frequency. These observations were included in the calculation of all other statistics, however.
The charge names, “Flee on Foot” and “Flee Officer on Foot,” were used to denote the same underlying violation and the latter was dropped as a duplicate entry because all other data for the arrest was the same. Similarly, the charge names, “False Name” and “False Name Or Info,” were used to denote the same underlying violation and the latter was also dropped as a duplicate. Further, there were 170 arrests that occurred prior to 2012 and these observations were deleted accordingly.
In addition, independent research revealed that a handful of charges in the original dataset are felony offenses (as opposed to petty misdemeanor, misdemeanor, or gross misdemeanor offenses). Individuals who were arrested for any one of these charges were dropped from the dataset.
The population data relied upon in this study is drawn from the U.S. Census Bureau. The study uses population estimates of the total population for Minneapolis generated by the U.S. Census Bureau and is equal to 394,239 for 2012, 400,070 for 2013, and 405,901 for 2014. The racial demographics of the city are held constant across the relevant study period and were set equal to the 2010 Decennial Census percentages. Specifically, according to 2010 Decennial Census, Minneapolis is 63.8% white, 18.6% Black, 2.0% Native American, 5.6% Asian, and 10.0% other.
The racial demographics of the youth population in Minneapolis differ from the total population (adult and youth combined). Using the online interactive application, American FactFinder, to access the 2010 Census Summary File 1 directly, Minneapolis’ youth population is calculated as 40.4% white, 30.1% Black, 3.2% Native American, 6.9% Asian, and 19.4% other. All arrest rates calculated in this study account for this difference in the racial demographics of youths.
The data set contained both the home address of the arrestee as well as the address of the arrest incident. Having information on the home address allowed us to determine whether or not the arrestee was homeless. Specifically, an individual was recorded as homeless if the home address of the arrestee was listed as anyone of the following: “Homeless,” “In Car Behind Work,” “In & Out Of Treatment,” “Lives On Streets,” “Sleeping On Streets,” “None Homeless, ” “Npa/simpson House,” “Salvation Army,” “Union Gospel Mission,” “Simpson Shelter,” “Higher Ground Shelter,” “Dorothy Day Shelter,” and “Harbor Lakes Shelter.” In addition, an individual was denoted as homeless if the home address of the arrestee was listed as any one of the following: “No Address,” “No Perm Address,” “No Permanent Address,” “N P A” and “Unknown/npa.” An individual was not denoted as homeless if the home address of the arrestee was listed as any one of the following: “Refused,” “Ref,” “Rfd,” “Unk,” “Unknown,” “Unknown/refused,” “Unknwon,” “Unkonwn,” “Unkown,” and “Unnown.” Finally, an address was associated with homelessness if it contained the phrase “1000 Currie” or “1010 Currie,” because 1000/1010 Currie Avenue is the location of the Catholic Charities “Pay for Stay” emergency shelter.
The home and arrest locations were geocoded by first running the addresses on an address locator that was created using the geocoding software installed on ArcGIS. Missing or poorly matched addresses were subsequently geocoded using the Google Geocoding API (which imposes a quota of 2,500 requests per IP address per day).
The racial demographic data used for the maps was drawn from TIGER products provided by the U.S. Census. TIGER products are spatial extracts from the Census Bureau’s MAF/TIGER database that contains features such as roads, rivers, highways, as well as legal and statistical geographic areas. This study uses a TIGER/Line with Selected Demographic and Economic Data product that joins the block group geography from the TIGER/Line Shapefiles to the 2007-2011 American Community Survey 5-year estimates. In other words, the demographic information in all of the maps in this study is aggregated at the Census Block Group level. A Census Block Group is a geographical unit which is between the Census Tract and the Census Block. It is the smallest geographical unit for which the bureau publishes sample data. Typically, Block Groups have a population of 600 to 3,000 people.
Each Census Block Group is assigned a shade of grey which represents the percentage of Black individuals who live in that Census Block Group. White represents the lowest percent Black Census Block Groups; solid black represents the highest percent Black Census Block Groups.
The location of arrests is overlaid on top of the racial demographics layer. Arrests are represented as dots and are scaled according to arrest frequency. In certain instances, arrests are also color coded by race. Note that for less than one percent of arrests, the arrest location provided was outside of Minneapolis city limits, which may be due to irregularities in the geocoding process or in the recording of arrest locations by the police. These irregularities are not reflected on any of the maps presented in this study.
- FBI Uniform Crime Reports, Persons Arrested, available at http://www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/2012/crime-in-the-u.s.-2012/persons-arrested/persons-arrested.
Every month, you'll receive regular roundups of the most important civil rights and civil liberties developments. Remember: a well-informed citizenry is the best defense against tyranny.