How to Have a "Great School District"

Scores from the 2006-07 STAR test of California schools were released on Tuesday. This is major event in the life of local schools. Not only does it effect their status under No Child Left Behind, but the scores are published in the newspaper as a source of public pride and shame. This is intentional. The decade-long movement for standardized testing and school accountability is based on the idea that we should know which schools are doing a good job and which are failing. Jack O'Connell, State Superintendent of Public Instruction, for example, called the tests a “bright light on school performance.”

Test scores are supposed to shine a light on the teachers and administrators. Yet, inevitably, someone will note that some of the light is shining past schools to the demographic profile of their students. Assistant Superintendent for the Kern High School District, Joe Thompson, for example, told the Californian that kids who are poor and learning English tend to bring Kern scores down.

Thompson's point seems reasonable enough, but how much difference do poverty and language make? I took this as an invitation for study and found that they actually matter a great deal of difference, even within Kern County. Have a look at this chart below. The dots are scores from Kern elementary districts on STAR tests (called API Base) in the last three sets of scores plotted by the percentage of students receiving free or reduced lunch. There is a pretty clear relationship here. As the percentage of students that qualify for free or reduced lunch goes up by one, it knocks down the average API by 2 points. On this basis alone, then, we should expect the difference in the scores of Richland (~80% free and reduced lunch) and Fruitvale or Norris (~5%) to be about 150 points. The actual difference is typically just a little more, about 185 points.

Surprisingly, not speaking English isn't quite as powerful a drag on test scores (see the chart below). However, note how few districts with more than a quarter of their students learning English score over 700. Consider too the fact that all of the Kern districts with scores better than 750 in the past three years have had fewer than 15% of their students considered "English Learners."

At the elementary level, differences in Kern district test scores are primarily differences of demography. Poverty alone, measured by participation in the free or reduced lunch program explains almost 70% of the differences in scores, according to a regression model. That is not to say that some schools are not better than others. Most of the districts either consistently outperformed or consistently underperformed a prediction model based on subsidized lunch participation and the percentage of English Learners. Fruitvale and Norris did well by this standard, but so did Delano. Richland was a modest underachiever after taking these factors into account. But these differences are small by comparison. The most direct, though perhaps not the best, way to win test score honor in Kern County seems to be reducing the number of poor, non-English speakers in your schools.