Dropouts from Durham Public Schools

Abstract

Low high school graduation rates prove a major challenge for policymakers throughout the United States. Durham, North Carolina is no exception. Durham’s graduation rate is currently 77%, compared to the North Carolina average of 80%. This project seeks to answer the following policy questions: who is dropping out from Durham public high schools and what characteristics in ninth grade or before can predict dropout from Durham public high schools? The project is being completed for the Durham chapter of Communities In Schools, a nationwide network of affiliated non-profit organizations focused on “empowering students to stay in school and achieve in life” through integrated school-based services. This project has four main components, which are based on administrative data from the North Carolina Education Research Data Center, and interviews with several staff members of Durham Public Schools. The first two components explore who dropouts are, by examining dropout definitions, and how dropout and graduation rates are measured using event and cohort dropout rates for Durham’s first-time ninth graders of 2006-2007. The second component describes where dropouts are in Durham through a series of maps showing student and dropouts’ home census block groups. The final component estimates a model for predicting graduation for this cohort.

Defining, counting and mapping dropouts High school dropout is a key construct that needs a precise definition, uniformly applied, in order for one to fully understand the dropout problem in Durham and the nation as a whole. Though No Child Left Behind increased the collection of data on student performance, it did not enforce a uniform dropout or graduation measure until 2011. Dropout is defined as a student who leaves school before graduation without transferring to another school or verified homeschool. Students in community college or GED programs are counted as dropouts, whereas students in youth correction centers are not. Students who re-enroll within the same month, according to the state manual, or the following year, according to Durham high school social workers, are not counted as dropouts. I calculate Durham’s dropout count for 2011 at 379, whereas the state reports 371. Though the discrepancy is small, the difference suggests that defining or counting dropouts is not fully transparent or replicable. Two measures are used to capture dropout and graduation: the dropout event rate and the cohort graduation rate. The event rate is the number of dropouts in a given year divided by the enrollment of the school. I calculate the event rate for ninth through 12th grade in Durham in 2010, which is the expected graduation year for the cohort of this study. At 4.2%, the dropout event rate I calculate is slightly higher than the 3.8% rate the state reports for Durham. The cohort graduation rate is the preferred measure to understand dropout, as it captures the number of graduates for a single cohort as a percent of the number of students, factoring in expansion and contraction of the cohort over time. The cohort graduation rate measures the cohort’s full experience in high school. I calculate the four-year graduation rate for Durham’s first-time ninth graders from 2006-2007, at 56.9%, and the five-year rate at 61%. The state reports much higher rates, at 69.8% for the four-year rate and 76.4% for the five-year rate. The discrepancy is difficult to explain. I theorize that the difference lies in the way a student who originally dropped out but re-enrolls is counted, depending on when that student returns to school. Mapping where Durham dropouts live, compared to the student population at large, shows that the dropout population is concentrated in areas of higher poverty. Using ArcGIS, I produced several maps of the 2009 high school student population (the most recent address information was available). Visuals of where dropouts are concentrated in Durham may help direct neighborhood-based resources.

Predicting graduation Using an ordinary least squares regression, I estimate a model that predicts the five-year graduation of the students in the cohort as a function of three main explanatory variables: failing one or more courses in ninth grade, absences in ninth grade, and number of reported offenses in ninth grade. The model also includes background characteristics (race, parent education, being overage for ninth grade, qualification for free or reduced-price lunch, gender) and a measure of previous achievement (third and sixth grade math and reading end-of-grade test results). Summary statistics are provided for the full cohort, as well as graduate and dropout subgroups. The cohort is 52% male, 59% black, 5% overage, 20% with parents with a high school education or less and 66% who have qualified for free or reduced-price lunch. Prior academic achievement, captured by reading and math end-of-grade test scores in third and sixth grade, is significantly higher for graduates than dropouts. The separation between graduates and dropouts widens between third grade scores and sixth grade scores. Students who graduate differ from the full cohort in many expected ways: 71% of graduates never failed a class in ninth grade compared to 49% of the full cohort; 89% of graduates have no reportable offenses in ninth grade compared to 80% of the full cohort; and 58% of graduates have under eight absences in ninth grade compared to 45% of the full cohort. With or without controlling for other characteristics, the results show that course failure, absences and reportable offenses are all significant predictors of graduation. The effect of the predictors does not sharply decrease when background characteristics such as race, which is generally thought to be strongly correlated with high school graduation, are included in the model. Course failure and over 36 absences in ninth grade have the largest effect size as predictors of not graduating from high school. Having failed a course in ninth grade decreases the probability of graduation by 24 percentage points, and being absent 36 days or more in ninth grade decreases the probability of graduation by 25 percentage points, when controlling for background characteristics and previous achievement.

Discussion and conclusion

The results have multiple implications. First, the results demonstrate that the predicative factors generally used for dropout and early warning system indicators—course failure, absences and behavior—hold as significant predictive factors for this Durham cohort. Second, while much attention is paid to race as an important correlate with high school dropout, this study shows that the indicators with the biggest effect sizes are course failure and absences. Third, the discrepancies between the counts and rates that I produce and those published are cause for concern over transparency on dropout counts. For Communities In Schools, I make the following three recommendations: 1) Focus prevention efforts on students who have failed one or more classes in ninth grade, are absent more than 19 times and especially more than 36 times in ninth grade, or have reportable offenses on their record. 2) Position efforts based on ninth grade characteristics, rather than other background characteristics. 3) Consider the discretionary nature of counting dropouts when targeting students for interventions.