Disaggregate Your Data to Make the Invisible Visible
Data Tip #9:“Disaggregation is a practical, hands-on process that allows a school’s faculty to answer the two critical questions: ‘Effective at what? Effective for whom?’ It is not a problem-solving (process), but a problem-finding process.”
(Lezotte and Jacoby, Sustainable School Reform 1992.)
If you want to tap one of the most powerful uses of data, disaggregate! Disaggregation means looking at how specific subgroups perform. Typically, formal student achievement data is “aggregated,” or reported for the population as a whole—the whole state, school, grade level, or class. Disaggregating can bring to light critical problems and issues that might otherwise remain invisible.
For example, one district’s state test data showed that eighth-grade math scores steadily improved over three years. When the data team disaggregated those data, they discovered that the math scores for boys improved, while the scores for girls actually declined. Another school noticed increased enrollment in their after-school science club. However, disaggregated data indicated that minority students, even those in advanced classes, weren’t participating.
Here are some examples of questions that disaggregated data can help to answer:
• Is there an achievement gap among different demographic groups? Is the gap getting bigger or smaller?
• Are minority or female students enrolling in higher-level mathematics and science courses at the same rate as other students?
• Are poor or minority students over-represented in special education or under-represented in gifted and talented programs?
• Are students at certain grade levels doing better in core subjects?
• Are students whose teachers participate in ongoing professional development
in reading, math, or science doing better in these subjects than students whose teachers do not participate?
• Are the school’s most recent curriculum and instruction adjustments improving the performance of students in the lowest quartile?
To answer these or other questions, carefully consider what disaggregated data is available and what additional data you need. Develop a data collection plan that includes a wide variety of data that can be disaggregated, such as state and local performance assessments, samples of student work, enrollment data for advanced courses, special programs, and professional development, as well as student and teacher survey results.
Following are tips to help you get started with disaggregating test data:
• Thoroughly understandyour school's demographics in order to select the relevant variables for disaggregation. NOTE: Some schools benefit from disaggregating data within demographic groups, such as Hispanic students born in the continental U.S. compared to those who are foreign born.
• Request state and district test data reports that are disaggregated relevant to your student population.
• Explore technology tools that will help collect, analyze, and report disaggregated data more easily.
• Note relevant demographic data as you collect other information about student learning.
• Ask for support from district data experts or the companies that provide your data system. Let them know the types of disaggregated reports that will best serve your needs.
• Drill down - dive into the data using the four-phase data-driven dialogue process described in TERC's previous tips http://usingdata.terc.edu/data_tips/
As noted by Lawrence Lezotte and Barbara Jacoby in their publication, Sustainable School Reform, “Disaggregation . . . is not a problem-solving but a problem-finding process.” Once you have a clear understanding of who knows what and the learning problems that exist, you can make changes to programs and instruction to target these specific learning gaps.
*Segments excerpted from Love, N. Using Date/Getting Results: A Practical Guide for
School Improvement in Mathematics and Science. (2002). Christopher-Gordon Publishers, Inc., p.39-42.
Mary Anne Mather, Facilitator, TERC’s Using Data
For education and data news, follow us on Twitter http://twitter.com/tercusingdata