Study
Open enrollment was available to middle and high school students in the Charlotte–Mecklenburg, N.C., School District (CMS) and required parents to apply for the lottery-based system, thus creating a self-selecting sample. More than 95 percent of parents submitted at least one choice. However, an open-enrollment policy meant that some schools would have more students who wanted to go there than the schools would have available slots. A policy was developed whereby slots at oversubscribed schools were first filled by students with guaranteed access (i.e., students in the neighborhood) and any remaining slots were allocated by a random lottery-based system. CMS used an algorithm known as “first choice maximizer,” which gave students a higher likelihood of receiving their first choice school than their second, a higher likelihood of their first choice than their third choice, and a higher likelihood of their second choice than their third choice (Deming 2011).
The lottery-based system was structured by an ordered priority grouping, which consisted of the following groups: 1) students who attended school in the previous year and their siblings; 2) free or reduced-priced lunch eligible (FRPL) students applying to schools where fewer than half of the school’s population in the previous year were FRPL; 3) students applying to schools in their choice zone. Applicants were sorted into these priority groups and then assigned a lottery number. Lottery numbers were distributed within each group, so that remaining slots at these schools were offered to those in a higher-ranking priority group, by their lottery number, before slots were offered to applicants in a lower-ranking priority group.
CMS administrative data was matched to arrest records from the Mecklenburg County Sheriff (MCS). The administrative data includes demographic information, attendance and behavioral outcomes, yearly test scores in math and reading for grades 3 through 8, and subject-specific tests for higher grades. The data also includes students’ addresses each year; these were used to create geographic identifiers. The MCS maintains an online searchable database that covers arrests in the county for the previous 3 years, which includes all arrests of adults (age 16 and older in North Carolina) that occurred in the county. Given that open enrollment began in 2002 and the MCS had arrest records only for 2006–09, some of the older students may have been arrested before 2006 and thus would not be captured by this database. As a result, historical arrest records were also obtained directly from the MCS for lottery members only.
The overall sample consisted of 21,132 high school students and 22,896 middle school students. Sixty percent of the total sample were excluded from the analysis because they chose, and thus automatically received, their neighborhood school. Of the remaining 40 percent, 75 percent were excluded because these students were in lottery priority groups where the probability of admission was either 0 or 1; therefore, there was no random assignment to analyze. Further, given that high-risk students are overwhelmingly male, females were excluded from the analysis. This left a final lottery sample of 1,891 high school students and 2,320 middle school students.
Lottery winners and nonwinners were matched on variables such as race, gender, and prior test scores to ensure comparability. Within the final sample, lottery winners were compared with nonwinners on the following outcome measures: all nontraffic arrests; felony arrests; property felonies; violent felonies; drug felonies; total social costs; sentence-weighted, which weights crimes by the expected punishment if convicted, while taking into account criminal intent; and total dates incarcerated.
Total social costs
were defined as the victimization costs, such as lost productivity and medical care, as well as quality of life. Certain crimes are given an estimated financial value, which were used to determine the total social cost of crime.
Sentence-weighted
weighed crimes by the expected punishment if convicted, while taking into account criminal intent. Subgroup analyses were conducted to examine the potential difference in outcomes for middle- and high-school students.
