Introduction

Standardized testing, a key component of public education, is about to become even more important. The Elementary and Secondary Education Act (ESEA) reauthorization of 2002, better known as the No Child Left Behind Act, mandates that children in grades 3-8 be tested every year in both reading and math. It requires schools to ensure that all students meet minimum proficiency standards on these tests and to reduce the achievement gap between disadvantaged and minority children and other students. Schools that fail will lose federal money, and their students will have the option of going to other public schools. In short, schools, school districts, states, and indeed, the nation as a whole will now find themselves in the position, known so well to every student, of nervously awaiting their final exam grades each year.

This emphasis on accountability begs the question of whether the testing regimes now in place are up to the task at hand. Many people question the validity of any kind of standardized test. Clearly, no tests are perfect. Nevertheless, they are necessary. Without some sort of systematic evaluation process, there is simply no way to determine which approaches are working and which ones are not. In that environment, ineffective schools can persist without change ad infinitem, which is essentially the situation we've had in public education for many years. The ESEA seeks to change this. For it to work, we have to ensure that the tests we use are accurate, that they actually measure how much and what children are learning in school.

The ESEA allows schools to choose their own testing regimes. Every regime has its own strengths and weaknesses. However, virtually all of them suffer from one major flaw. They focus only on achievement levels, and ignore changes.

The testing regimes, and probably most people, assume that good schools are those with students who have high test scores and bad schools are those with students who have low test scores. But that isn't entirely true. Not all students enter a school at the same level. A school that admits students with low scores and raises them up to average is better than one that admits high-scoring students and merely keeps them at the same level. Ultimately, the best way to measure school quality, or teacher quality for that matter, is to determine how much they change their students' test scores each year. This is nothing more than the simplest and most direct way to measure how much students are learning. This approach is called "value-added," because it focuses on how much value a school or a teacher is adding to what students bring with them.

This does not imply that achievement levels are irrelevant. They are useful in determining what material students are ready for at any given point in time, and where they stand in relation to proficiency standards. But they do not tell the whole story.

Using a value-added testing regime in addition to the standard one could yield at least three important benefits:

• It would make ESEA more effective by providing a more accurate picture of which schools, school districts, and states are and are not making progress.
• It would generate objective measures of teacher quality that could be used to improve teaching.
• It would lend itself more readily to evaluating school reform programs.

Value-added testing is not perfect. It has a harder time distinguishing among students, teachers, and schools in the middle than among those at the extremes. It cannot necessarily distinguish between true learning and "teaching to the test." And there is some disagreement among proponents of value-added testing about its power as a teaching tool. But standard testing regimes suffer from these same problems.

This paper will examine both the promise and the problems of value-added testing. It will show that there are still a number of issues that need to be addressed, so we should be cautious in its application. We should use it in combination with other approaches. Nevertheless, it is clear that the potential benefits of value-added testing are significant.