Here's the Full Report
Naomi Chudowsky and Victor Chudowsky
Center on Education Policy
July 21, 2009
Many in the research and policy worlds have taken for granted the existence of a phenomenon known as the "plateau effect," wherein test scores rise in the early years of a test-based accountability system and then level off. Drawing from our database of reading and math test results from all 50 states going back as far as 1999, we looked for evidence of a plateau effect in 55 trend lines from 16 states with six to ten years of consistent test data. This report outlines those findings.
Our analysis revealed several main findings:
• In the current testing context, one cannot assume the existence of a plateau effect
when trying to predict state test score trends. Although this study found instances of
plateaus in test score trends in the 16 states analyzed, they were not as pervasive as may be commonly assumed. Percentage proficient trends followed a wide variety of trajectories, including some plateau patterns. Of the 55 trend lines we examined from various states and different grade levels in reading and math, 15 exhibited a plateau pattern. We also found 21 trend lines with steady increases in the percentage proficient over time and 19 more with fluctuating “zigzag” patterns that still moved in an overall upward direction.
• The largest gains did not consistently show up in the early years of a testing program. In many of the trend lines, the largest gains occurred between the first and second years of administering a new test. But just as often, the largest gains appeared between the third and fourth years, or between the fifth and sixth. Thus, we concluded that the largest gains are just as likely—and sometimes more likely—to occur after four or even six years of a testing program as they are in the first few years.
• A clear upswing in test results was apparent after the enactment of the No Child
Left Behind Act (NCLB). In many states, the largest gains in percentages proficient
occurred between 2003 and 2004, two years after NCLB took effect. But the early years
of NCLB were not always concurrent with the first few years of a state testing program. In several states, the tests used for NCLB had already been in place for some years, and a bump in scores still appeared after NCLB. This pattern suggests that test results can increase substantially even after a test has been in place for several years if higher stakes are introduced in the accountability system.
• In the three states with the longest trend lines, gains generally did level off after nine or ten years, but the data were too limited to know whether this is a consistent pattern in state test performance. One complicating factor in studying test score trends is that states tend to change their tests quite frequently, so long trend lines are rare. It may well be that by the time a state would start to show a plateau effect, it changes its tests.