
Stop Using Interim Tests to Predict Summative Scores!
Choose assessments that support instruction and evaluate programs
In 2009, Center for Assessment researchers identified three main purposes of interim assessments: predicting future performance, evaluating programs or policies, or helping inform instruction. Among the three uses, the authors argued that prediction was the least interesting and helpful. Despite this, in our experience working with most states and many large school districts, prediction remains a primary use for interim assessments.
Interim assessments became nearly ubiquitous during the early years of the No Child Left Behind (NCLB) era, as schools were eager to identify “bubble students” they might boost over the Adequate Yearly Progress (AYP) bar. We understand the desire to identify students most at risk of not meeting grade-level targets and intervene with additional supports. But “prediction” falls short of supporting teachers and schools in implementing practices that advance learning.
As Perie, Marion, and Gong explained, an interim assessment’s chances of achieving one of its main purposes depend on its design and implementation, and no single assessment will serve all three purposes well. Schools’ and districts’ emphasis on prediction has distorted interim assessment design and limited their use for evaluative and, especially, instructional purposes.
What’s the Point of Predicting?
We often hear from educators and school leaders that they need information early in the year to understand how students are likely to do on the end-of-year assessment. As Rich Hill, the Center’s founder, quipped many years ago, “Most of these students just took a good assessment a couple of months ago. Why not just look at that?” If educators want to see where their students stand in the fall, they can use last spring’s data. Children’s knowledge and skills won’t have shifted dramatically between May and September.
Our colleague Nathan Dadey verified Rich’s observations by demonstrating that interim assessment results administered during the year predicted end-of-year summative scores no better, and often worse, than those from a student’s previous state assessment. A recent Educational Assessment article reached a very similar conclusion. If prior-year state assessment results can be used to identify students at risk of not passing the next state test, then what is the marginal utility of administering the additional assessments?
Those who are quantitatively inclined know that prediction simply capitalizes on the statistical relationship between two variables. In this case, between an assessment administered early in the year and the state end-of-year assessment. Almost all academic assessments are correlated with one another. Therefore, most 5th grade math tests administered early in the year, for example, will correlate with the state test given at the end of the year. Just because a test administered early in the year can predict a later score doesn’t mean we should prioritize this use case.

Interim assessments are often judged by how closely they correlate with the end-of-year test. A more useful standard is whether the assessment helps an educator support struggling students, so they outperform what an early test predicted. Fundamentally, we want people to “break” predictions. If an interim assessment predicts that a student will score low on the state test, an educator’s job is to break the prediction by intervening and changing the student’s learning trajectory.
The overemphasis on prediction is particularly concerning because it can lead to misguided actions. Some states, for instance, have added requirements for early literacy screeners to predict Grade 3 outcomes. If a school system administers an assessment in kindergarten or Grade 1 that strongly predicts Grade 3 outcomes, it may indicate that students were not provided appropriate remediation or support.
In other words, schools knew which students were unlikely to perform well on the 3rd grade test, and those kids did not, in fact, do well. The prediction was strong, but the instructional support did not help change the students’ trajectory. In contrast, a kindergarten test that does not strongly predict Grade 3 scores might suggest that students received effective support and thus score higher than their original assessment score predicted.
Choosing Prediction Limits Other Assessment Uses
This obsession with prediction limits other potential benefits of administering interim assessments; namely, evaluative or instructional purposes.
Tests that strongly predict state scores typically have several design features. First, they measure the same or very similar knowledge and skills defined by the state academic standards. Second, they sample across the full content of a school year, since state assessments also use this full-domain approach. Third, they may use item types similar to those in the state assessment.
Each of these features helps make the interim assessment as similar as possible to the state assessment. Unfortunately, these design features often limit the capacity to use the results to make meaningful instructional changes.
Shifting to Evaluative and Instructional Uses
A better approach is to prioritize assessments that help teachers and students identify how students are thinking in relation to the learning target so they can help prepare students to learn what comes next.
Assessments designed to support instruction and learning have different features than those designed for prediction. Such assessments could be instructionally focused interims, or they could be well-designed classroom or formative assessment. Among other features, assessments designed to support instruction typically assess content at a smaller grain size than on end-of-year tests, are closely tied to the curricular sequence, and allow for a greater variety of allowable responses.
Further, these assessments should be administered when teachers and students are still able to act on the results and should provide more detailed information about what to do next to improve learning, rather than a general sense that students are not performing up to standard.
Typically, interim assessments have not had many of these features, since response types are constrained and reporting often provides quantitative scores rather than emphasizing qualitative feedback. However, interim designers are innovating in this space, and some are working to provide more content-focused feedback to assessment users.
Interim assessments can also serve an important evaluative role. Adopting a continuous improvement mindset could encourage the design of assessments appropriate for use as outcome measures in program evaluation studies. In such cases, assessments could be designed to focus on specific skills that are the target of intervention or curricular change.
Focusing on prediction often leads teachers and students to chase scores rather than focus on tangible changes to teaching and learning that can improve student understanding. Ideally, assessment practices should increase the flow of information between students and teachers by identifying content students understand and where they need additional support. Even better, assessment should point teachers to next steps by identifying what students know and making student thinking visible so teachers can help students build on that knowledge.
Teachers probably already have a good sense of which students need additional support. Investing time and resources in assessments that help teachers figure out what to do is more likely to drive the outcomes schools want to see in the long run.

