Pete Goldschmidt, Christy Kim Boscardin, and Robert Linn
National Center for Research on Evaluation, Standards, and Student Testing

The latest release of the California Academic Performance Index (API) indicates that, overall, California schools demonstrated about an 11-point, 1.6%, improvement over last year. This result is consistent with previously released STAR results. Fewer schools (52%) met their growth targets in 2006 than in 2005, but this is not unexpected because additional subgroups are included in the calculations this year as noted in Superintendent O'Connell's comments.

Another challenge facing California schools and schools across the country is that the longer an accountability system is in place, the smaller the gains. California may have reached that plateau, which usually happens between the 3 rd and 7 th year of an accountability program. STAR began in the 1997/1998 school-year and the API was passed into legislation in 1999. Previous CRESST research has long documented this leveling-off phenomenon on almost any test.

All California subgroups, including African American students, Hispanic students, and English Learners, (ELs) have made steady statewide progress. The largest gain for the subgroups was reported by ELs in grades 2 through 6, with a 16-point difference from last year.

However, ELs in grades 9 through 11 made only slight improvements, with a 2-point gain. Although the improvements made by the subgroups are encouraging, the achievement gap among the subgroups continues to be substantial. For example, English Learners (over 32% of students included in the current API growth report) scored about 80 points lower in the overall comparison. The achievement gap is as high as 96 points in grades 9 through 11 for ELs. Based on preliminary data, about 62% of schools have met their EL performance growth target.

African American students also continue to lag behind with an 83-point difference in overall scores. For schools with numerically significant African American subgroups, about 56% of schools met their African American student growth target in 2005-2006. However, only 79% of the schools that have met the school-wide target have also met their African American student growth target. This raises questions about whether the school growth targets are appropriately aligned and targeting the goal of closing the achievement gaps.

In thinking about API results we should consider three related aspects that place school performance into context. These are whether high API scores accurately identify good schools, whether changes in API scores measure student achievement growth, and whether API scores accurately reflect the performance of student subgroups and corresponding performance gaps.

Schools with higher API scores are supposedly doing a better job of educating students than a school with a lower API score. For parents, an API score can help in choosing a good school for their children. The effect of API scores reach beyond parents as research indicates that home values are directly related to API scores. In fact a recent CRESST study, based on ten years of data, indicates that a 50-point difference in the API is associated with a 5.5% difference in house prices. It is critical to examine whether we can validly infer from API results that schools with higher API scores are actually better schools.

In fact research indicates that API scores are related to the student enrollment composition of a school-a factor that schools have little control over. A better indicator of school quality is comparing schools with similar characteristics. Using ranks based on similar schools is a step in this direction and California 's API includes such a ranking.

In evaluating how well a school is performing, it is also important to consider how much student achievement improves over time. To that extent we can track changes in the API and see whether, despite the fact that the school started with a low API, it has shown progress over the past few years. Changes in the API are often more closely related to changes in grade level performance than they are to changes in individual student performance. While it is a good that 2 nd graders this year are performing better than 2 nd graders last year, it says less about how well last year's 2 nd graders perform this year in 3rd grade.

In fact recently released STAR results indicate that while same grade level performance is generally increasing, student performance in successive years, for example 2 nd to 3 rd grade English language arts or 4th to 5th grade mathematics, often is not.

Because the API accountability model is a composite system of scores, results based on the API will not match results based on the No Child Left Behind (NCLB) accountability model. Using achievement levels, NCLB requires that by 2013-2014, 100% of students must be proficient. This means that over time API and NCLB results will become more discrepant. In fact, results of recent research based on a sample from a large urban district suggest that about 75% of schools with "good" API scores and currently meeting NCLB targets will not meet the final target in 2013-3014.

Lemay Street Elementary School in Los Angeles , for example, currently has an API growth of 840. Its EL students are scoring at 743 and economically disadvantaged students are scoring at 778. All of these results are above the state average and the school, as a whole, has surpassed the state target. Lemay is also currently meeting its NCLB targets with 47% proficient in English language arts (ELA) and 72% proficient in mathematics. The current results point to two areas of concern: one, despite meeting the state target, over half of the students are not proficient in ELA; and two, Lemay must demonstrate about 7% growth per year in proficiency in ELA and about 4% growth per year in mathematics to meet the federal targets in 2013-2014.

Overall, Lemay exemplifies a school that is doing a good job but will be challenged to maintain such dramatic growth rates over the next years. We are not advocating that NCLB is the better accountability model-far from it-but it is the higher stakes model that can lead to sanctions for Title I schools that do not make adequate yearly progress targets established for purposes of NCLB.

The overarching assumption behind test-driven accountability is that it will reduce achievement gaps because it serves as an external pressure for academic improvement. However, its effects on academic performance depend largely on school capacity and social support. For example, the achievement gaps among racial and socioeconomic groups have been found to be relatively small in states with more equitable distribution of school resources and classroom opportunity to learn (OTL) indicators. However, as a recent report from Education

Trust suggests, schools and districts with high percentages of lowincome and minority students who need the most qualified teachers are more likely to have teachers who are inexperienced and have lower basic academic skills.

Our analysis and experience with other state accountability systems leads us to conclude that the API and STAR have successfully pushed California schools to improve learning for many of its students.

However, the achievement gaps remain unacceptably high and to reach both state and national accountability goals, California, as well as other states, must confront very substantial hurdles. As we look to strategies for closing achievement gaps and improving the performance of all schools, the focus will need to shift from monitoring outputs toward placing more emphasis on school inputs such as teacher quality, curriculum, and effective data use that improves instruction.