Putting the "L" in LDS
by Nancy Smith* (2008)
Reprinted with permission from the Data Quality Campaign (DQC).
So what is an LDS? What makes it longitudinal? Many states and districts think that they have a longitudinal data system because they have a data warehouse that has many years worth of data. Others don't have a data warehouse, but do report many years worth of annual graduation rates. So, they say that they have an LDS. Others believe that they have a longitudinal data system because they have a student identifier system.
Longitudinal means that data on a given student can be connected across years. In photography parlance, it is more like being able to watch a video of a student as (he or she goes) from grade to grade. If you put all those videos of individual students together into a montage, you can usually spot some trends about what happens to students with different types of experiences in the early grades. Usually, though, school districts and state education agencies (SEAs) review "snapshot" data—pictures taken of a given 3rd grade class one year, the 4th grade class the next year, and the 5th grade class the third year. Some of the same students might be in all three pictures, but it is more likely that some students leave and others join the cohort over the three-year period.
In years past, it was more common for school districts to send summary statistics to the SEA—for example, the count of students receiving special education services or free- or reduced-price lunch, the percentage of students passing that statewide exam in the spring, or the number of students in each racial/ethnic category. The SEA could then aggregate or add up all of the school or district numbers to get statewide totals. Aggregated snapshot data are very valuable to educators and policymakers, especially when they need a way to quickly summarize how schools are performing and see which districts serve which types of students.
However, snapshot data alone do not provide enough information to truly evaluate the impact of student mobility or of dropout intervention programs, the relationship between course-taking patterns and college-readiness, or the ability to calculate a graduation rate while taking into account students who transfer to another school, are retained in a grade, leave for private school or drop out. Only a set of robust longitudinal data on the characteristics and experiences of each student—that tracks students across school years and across campuses within a state and connects that enrollment data with other outcome data (course completion, college readiness, assessment and exit data)—provides the ability to thoroughly investigate the patterns of success and struggle that students experience. Student-level longitudinal data can be aggregated to look at school, district and state trends, but they can also be analyzed at a much finer level of detail than snapshot data to fully understand the relationships between the many factors affecting student achievement.
With the snapshot data that is reported per No Child Left Behind requirements, it is possible to say, for example, that 51 percent of African-American students were proficient on the 10th grade mathematics exam, while 83 percent of White students were proficient. With student-level longitudinal data, it is possible to say that of the 51 percent of African-American students who were proficient on the 10th grade mathematics exam, 65 percent of them were also proficient on the 8th grade mathematics exam; and, of those students, 78 percent took algebra I in the 8th grade. With that information, educators and policymakers can understand the importance of preparing students to take algebra I in 8th grade. This type of longitudinal data shows that students who do not take algebra I in the 8th grade are less likely to show proficiency on future exams. With this information, administrators can tailor their curricular activities in earlier grades to prepare more students for algebra I in the 8th grade. Of course, with longitudinal data, the same administrators will also have the data necessary to determine which elementary and middle school students are on track to take algebra I in the 8th grade and provide the necessary intervention to those who are not yet ready but could get there.
Data warehouses (or alternatives) and easily accessible reporting and analysis tools are critical to improving the use of data in education. They are very useful and important tools, even when they are full of snapshot data and statistics. Having these tools, however, does not automatically imply that the state collects student-level longitudinal data, or that they are using longitudinal statistics to inform their decisionmaking.
Longitudinal data implies the ability to collect many key pieces of data on individual students (examples include: campus of enrollment each year, programs in which the student receives services, ethnicity, age, statewide and end-of-course exam scores every year, reasons for not taking statewide exams, college-readiness test scores, and exit status [graduate, dropout, transfer, home school]); connect all those pieces; and then aggregate across students according to a set of key variables in order to analyze the impact of, and relationship between, variables. This ability to analyze and predict performance at the student level is what will ultimately help educators and policymakers at the local and state levels improve the policies that will eventually lead to improved student achievement for all students.