CBMs predict performance on statewide assessments 1-2 years later

iStock_000020536048XSmallHow does a student’s rate of growth, based on curriculum-based measures (CBM), predict how well that student will perform on a state assessment in reading and math 1 or 2 years later?

CBMs are highly predictive of performance on state assessments 1 or 2 years later, says a recent study in School Psychology Review, but cut scores were more useful in predicting performance than student’s growth slope.

“Across CBM skills, as many as 83.57% of students were accurately classified as either passing or failing the high-stakes or standardized measure based on their CBM performance above or below the identified cut score,” the authors write. “Consistently, few students performed worse than expected, with the highest percentage of students predicted to pass who actually failed being 12.1%.”

That slopes were less useful in predicting how a student would perform on the state assessment suggests that it may be more important for educators to consider if the student achieved a particular level of performance than if the student attained a certain rate of growth, the researchers report.

The issue of the diagnostic accuracy of CBMs is important for those who want to allocate resources wisely and who wish to have reliable early warnings so that they will have 1 or 2 years to intervene with students, the researchers say.

Educators working within models of response to intervention potentially use CBM data to make instructional and even special education eligibility decisions, the authors write. Understanding how well CBM cut scores predict how students will do on statewide as well as other large-scale achievement tests is clearly very important, they note.

Some of the findings from this study, which used data from a CBM norming project in a moderately sized urban-suburban district of 14,615 students in eastern Pennsylvania, are:

  • CBM scores from the spring were most highly correlated with performance on state assessments (as opposed to fall and winter)
  • Student growth slopes had much weaker and inconsistent correlations with performance on state assessments 1 or 2 years later (with the exception of reading slope in 1st grade and 2nd grade)
  • Student slopes’ correlation with performance on state assessments decreased the higher the grade
  • Math CBM slopes were much more inconsistently and weakly correlated with outcomes on state assessments than reading slopes.

The authors speculate that math slopes may be less predictive than reading slopes because reading skills are more cumulative whereas math skills are very diverse and student performance may vary based on the math skills at the focus of instruction.

The norming project, which produced the data used in this study, took place across 6 elementary schools (grades 1-5) in the district. The schools were selected to represent overall district performance on reading and math assessments. The resulting normative sample consisted of between 199 and 235 students per grade or 1461 students for reading and 1477 students for math. Data was collected from students in Grades 1-5 over a 10- to 15-day time period in October, February and May.

CBM data were collected for:

  • oral reading fluency
  • math computation
  • math concepts and applications (grades 2-5)

Assessment data were from:

  • TerraNova Achievement Test
  • the PSSA, Pennsylvania’s assessment for educational accountability

Below are some of the demographic characteristics of the district:

  • 32.8% low-income level
  • 58% Caucasian
  • 31% Hispanic
  • 9% African American
  • 3% Asian/Pacific Islander and
  • <1% American Indian students

CBMs can be used with confidence in the screening and early identification of students who are having difficulty with basic academic skills, the authors write. Although CBM is a brief screening tool, it has a fair degree of predictive utility for future high-stakes assessments, they conclude.


“Long-Term Diagnostic Accuracy of Curriculum-Based Measures in Reading and Mathematics,” by Milena Keller-Margulis et al., School Psychology Review, Volume 37, Number 3, 2008, pp. 374-390.

Leave a Reply

  • (will not be published)