Measuring Student Performance and Academic Growth

Wednesday, March 4, 2015 -- Dave Adams

School reform is a critical, complex, and constantly evolving part of our overall education environment. As state standards, tests, and accountability all continue to evolve toward higher expectations of student performance, so does our method of reporting and evaluating student performance. Since the passing of the NCLB Act of 2001, states have been mandated to assess all students in grades 3–8 and one high school grade, as well as report on grade-level proficiency. This system of measuring and reporting proficiency has been in place now for well over a decade, and some policy makers and education reformers have begun to question this proficiency-based system and consider other options. 

Why consider alternatives to proficiency? 

The key reason to consider alternatives is that proficiency provides data on a single snapshot in time (one test administration). Proficiency-level reporting does not provide information on a student’s academic progress over time or velocity of learning. 

What is proficiency, and why is it limiting?

Proficiency levels are the most common means of reporting state test data today. With proficiency-level data, students are grouped into cut-score ranges based on performance on the grade-level state tests. Most states use between three to five performance levels; for example, a state using five performance levels might look something like:  1) Advanced, 2) Proficient, 3) Approaching Proficient, 4) Basic, 5) Unsatisfactory. Both Smarter Balanced and PARCC are in the process of establishing initial performance levels. Even as states continue using proficiency levels in their data and accountability systems, there is a movement to shift to new models of data reporting and accountability that are focused on measuring student growth.

What is growth in relation to student performance?

First, let’s say what growth is not—it is not a single snapshot of student performance on a single test. Growth is not “status.” Growth is the academic performance of an individual or group of students over time (usually on two or more assessments).   

Why could growth be a better measure of student performance and accountability?

One of the weaknesses of the proficiency model is that its single-point-in-time “status” approach does not consider what level of knowledge and ability students bring with them and, consequently, cannot tell us how much learning has occurred. On the other hand, growth, because it is academic progress over time, does tell us how much learning has occurred from a baseline point. Thus, growth could be a fairer, more informative approach to measuring both individual and group performance. Consider, for example, these two students:

The first student begins seventh grade right on grade level in math and finishes seventh grade with a math test score of 70%.

The second student begins seventh grade behind in math by two grade levels and finishes seventh grade with a math test score of 55%.

Which student has achieved more academic growth during the seventh grade? It is, of course, the second student, who started far below grade on math, even though this student had a lower end-of-year test score and would be placed lower in a traditional proficiency model of accountability.

Now, consider how this applies to two very different schools:

The first school is a high-performing school in a high socioeconomic area. This school has average seventh grade math scores that have increased from 81% to 83% over five years.

The second school is a historically low-performing school in a high-poverty urban area. This school has average seventh grade math scores that have increased from 52% to 65% over five years.

Which school has achieved higher academic growth? Of course, it is the second school, which has achieved tremendous growth, even though traditional proficiency models of accountability would rank this school far below the first school. So, growth has the potential to be a more accurate way of measuring how much individual or group academic development is occurring and, therefore, a fairer way of measuring what’s really happening.

What different types of growth models exist?

There are a number of different technical models of understanding and calculating growth. Here are seven different types of growth models defined in the paper A Practitioner’s Guide to Growth Models, authored by Katherine E. Castellano, University of California, Berkeley, and Andrew D. Ho, Harvard Graduate School of Education:

  • Gain Score models describe growth with simple differences or average gains over time.
  • Trajectory models extend gains or average gains in a predictable, usually linear, fashion into the future.
  • Categorical models define growth by transitions among status categories (e.g., Basic, Proficient, Advanced) over time.
  • Residual Gain models describe growth as the difference between current status and expected status, given past scores.
  • Projection models use past scores to predict future scores through regression equations
  • Student Growth Percentile models consider thepercentile rank of current status in a reference group of students with similar past scores.
  • Multivariate models use entire student score histories, including other subjects and teachers, to detect higher-than-expected student scores associated with particular teachers. 

What’s next?

Based on the increasing interest in growth and the number of states in the process of incorporating some type of growth measures in their state accountability systems, I believe we will continue to see growth becoming increasingly important from the highest level of federal and state education reform to the data driven instruction occurring every day in our classrooms. This means that we all need a common understanding of growth and how it applies to individual student learning as well as school and district performance.

Some additional resources: