Data System Flags Dropout Risks by 1st Grade
Indicators pave way for intervening early
As tracking data on students grow ever more extensive, some Maryland educators find early warnings of students at risk of dropping out may become visible at the very start of their school careers.
The affluent and tech-savvy 149,000-student Montgomery County public schools, in a suburb of Washington, is building one of the first early-warning systems in the country that can identify red flags for 75 percent of future dropouts as early as the second semester of 1st grade.
"If these kids are always with us, we can do something about this," said Thomas C. "Chris" West, who built the tracking formula as Montgomery County's evaluation specialist and now works as a data specialist in Frederick County, Md., schools. "Remember, these are signs of students who drop out—it doesn't mean they are dropouts."
Montgomery County's initiative comes in the midst of an explosion in the use of longitudinal-data systems to identify students at risk of not graduating from high school on time. According to the most recent count by the Data Quality Campaign, for 2012, 28 states use early-warning systems, with more in development. These systems can be used to target interventions based on profiles of characteristics of students who fail academically and drop out of school, though at this point, relatively few states or districts have reports available to principals and teachers multiple times a year.
Knowing the District
Most modern early-warning systems evolved out of the work of Robert Balfanz, the co-director of the Everyone Graduates Center and a research scientist at the Center for Social Organization of Schools at Johns Hopkins University in Baltimore, and from the University of Chicago Consortium on Chicago School Research. Research from the Johns Hopkins center showed that three red flags—chronic absenteeism, severe disciplinary infractions, and reading or mathematics failures—signal as early as 6th grade a student's disengagement from school and predict his or her risk of dropping out.
Other studies have since replicated the findings at earlier and earlier grades. However, Mr. West, who is also affiliated with Mr. Balfanz, cautioned that researchers and educators must study these risk factors in the context of a specific school system.
The Montgomery County district compared the grades, attendance, and behavior of 723 dropouts from the class of 2011 and 523 dropouts from the class of 2012 with those of their classmates who graduated. The early-warning system reverse-engineers a risk profile based on warning signs at four critical transition points: spring of 1st grade and fall of 3rd, 6th, and 9th grades.
For example, chronic absenteeism is generally defined as missing 10 percent or more days of school, excused or unexcused. In Montgomery County, Mr. West found virtually no pupils in the early- elementary grades missed 20 days of school. But missing as few as nine days of school nearly doubled a student's risk of dropping out later.
"The message for Montgomery County is, our kids are there in school; they just aren't doing well," Mr. West said at a discussion of the data system at the National Center on Education Statistics' annual conference in Washington last month.
Similarly, elementary schools rarely suspend students, but subtler behavior cues, such as report card notations of incomplete homework, more accurately signaled future problems at that age.
Report card grades were the strongest predictor of dropout risk in grades 1 and 3. An overall GPA of 1.2 (roughly a D) in the spring of 1st grade more than doubled a student's risk of dropping out later, and reading or doing math below grade level in 1st grade boosted that risk by 134 percent.
"A parent has the report card, student has a report card, teacher has a report card," Mr. West said, "so if we base our conversation on the report card, at least everybody's talking from the same page."
In later years, lower academic performance was even more predictive, even with higher report card grades. At both the 6th and 9th grades, a student with a GPA below 3.0 and no other risk factors still was more than 3½ times more likely to drop out of school.
All told, a combination of the grades, attendance, and behavior indicators in 1st grade predicted about 75 percent of the students who dropped out in the classes of 2011 and 2012. A quarter to one-third of students who had at least one warning sign in 1st grade had more red flags in the 6th and 9th grades.
While Montgomery County's warning system is not yet being used to track individual students in real time, the district is changing the way it talks about student risk factors. For example, the data showed that more than 60 percent of students who dropped out were not from poor families. English-language learners were overrepresented among dropouts in the class of 2011—16 percent, compared to the 4 percent district average—and special education students accounted for more than one in five dropouts in 2011, higher than their 11 percent share of the class overall. Mr. West said grade and behavior indicators proved more reliable and less discriminatory than looking at socioeconomics or race.
"It's like getting your blood pressure checked; you have to do it often and over time," Mr. West said.
One reason for caution: At early grades, the system can show almost 50 percent more students at risk of dropping out as those who ultimately do. Still, Mr. West noted that it's not certain whether the false positives come from mistakes that make sense in context—for example, a high-performing student who gets chicken pox and misses two weeks of school—or the effect of interventions to help at-risk students in later grades.
"But then you get into stigma; is it good to tell a 1st grader, 'You might be a dropout?' These kids do move in and out of these indicators," Mr. West said.
"You will not reduce dropout rates by [identifying] the students; it's what you do with them," he said. "Early-warning systems are part of an intervention strategy."
The district is working to analyze changes in the indicators from grade to grade to find the trajectories that might be more accurate predictors than at a single grade. It is also analyzing data from its high school graduates to find indicators associated with later college persistence.
Vol. 32, Issue 37, Page 10