Using Information: Data to Drive Decisions

Value-Added Indicators

High Schools only

What you are looking at

This chart shows the relationship between the actual performance of students in this school – expressed as the percentage of students who met the standard on the state tests – and the statistically generated performance range of similar students statewide. This chart uses only the 2005 assessment data.

What you are looking for

You are hoping to see the school’s students performing at or above the performance range of similar students statewide. This computer-generated model is not a standard, and performing as well or even better than similar students across the state is only the beginning of a journey towards 100% proficiency of all students. Over time, as the schools improve, the computer-generated ranges will themselves rise. This model helps us understand that schools do not start on a level playing field, and some will need more time, specialists, resources or any number of supports to help all of their children reach proficiency.

Statistically generated performance models level the playing field

Schools with high concentrations of students living in poverty or students with disabilities have sometimes been unfairly compared with schools whose less challenged children perform at higher levels on standardized tests. So, for example, the achievement demonstrated by schools on this year’s school-performance classifications charts roughly reflects their average socioeconomic background. The schools classified as high performing tend to have children who come from more affluent backgrounds; the reverse is also true.

In general, the public tends to compare schools without considering differences in student characteristics. In fact, poverty is the strongest predictor of student achievement, except for that student’s prior achievement.

The rationale

Increasingly, education researchers are using these models, often called “value-added,” to calculate what results schools are likely to achieve when taking into consideration the characteristics of their student body.

“Value-added” allows us to determine whether a particular school adds more value, or improves the child’s skill set more effectively, than other schools. For more than 40 years, researchers have known that the achievement results of different sets of students, such as those from different schools, vary in association with several specific key factors, including:

1. Poverty (by far the strongest predictor of student achievement, with the exception of prior achievement)
2. Non-English-speaking background
3. Educational background of the parents
4. Having special learning needs
5. Having a minority racial-group identity

Though individuals with one or more of these characteristics can and do perform well on state assessments, the majority tend to perform less well than children who do not have these characteristics. The many reasons for these historic patterns of lower achievement include such things as school expectations, the availability of flexible grouping and different types of instruction, inadequate funding and support to the schools these children attend, individual and family health, and the quality of social services offered to students.

Statistical models allow the public and those evaluating school performance to look at the achievement data through a lens that factors in some of the students’ challenges. This value-added perspective helps us to see to what extent the challenges facing each school influence performance.

These models predict only for groups of students with similar characteristics; they can not predict any individual student’s performance. As always, the unit of accountability in the Rhode Island school-reform agenda is the school and not the individual student.

The Rhode Island model

Rhode Island researchers created a model that considers the five characteristics mentioned above. Because Rhode Island is such a small state, the entire body of students enrolled in public schools serves as the context from which the test and grade-specific ranges were derived. Thus, students within a school are compared with similar groups of students statewide; schools themselves are not sorted for comparisons. The computer-generated ranges will change depending on the test because, for example, a writing assessment is more strongly affected by language-minority status than a mathematics test. The model uses only the 2005 assessment data.

Also see the technical description of the model.