High Stakes Push States to Game the System

Status: 
Archived
Subject: 
K-12 Testing

FairTest Examiner, November 2009

Over the years, state assessment systems have repeatedly demonstrated the relevance of "Campbell’s Law" to high-stakes testing and education. Social scientist Donald Campbell concluded: “The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.” New evidence of this comes from New York City, where schools made dubious claims of dramatic improvement that Mayor Bloomberg used heavily in his narrowly successful reelection campaign. South Carolina also found a way to make test results look better, regardless of whether children are learning more.

New York report cards illustrate corruption of indicators

There is no lack of examples of high-stakes tests corrupting both educational indicators and education itself, but New York City stands out. In September, it released school report cards that give letter grades based on a complicated formula that relies heavily on the results of state tests in reading and math, but no other subjects.

The results appear to show astonishing improvement over the previous two years: 97% of 1,058 NYC elementary and middle schools received an A or B for 2008-2009. Last year, 38% percent of elementary and middle schools got A’s and 41 percent B’s. All schools that were graded F last year and stayed open got A’s or B’s this year.

As education historian and former federal education official Diane Ravitch pointed out, the results are bogus, the consequence of dumbed-down state tests, narrow test preparation, and ignoring everything other than math and reading scores, including other subjects and school safety. In a September New York Daily News article, Ravitch wrote, “The report card system makes a mockery of accountability.” During his re-election campaign, Mayor Bloomberg touted the rising scores as evidence of the success of his test-based approach.

Independent studies conclude that test questions have gotten easier and more predictable. Investigators found that the state asks nearly identical questions year after year, and the excessive focus on test preparation has corrupted the results. New York State’s new education commissioner, David M. Steiner, recently reinforced this view, saying, "In too many cases, the assessment becomes the curriculum... If the test is the curriculum, then you're tempted to teach to the test." He pledged to deemphasize testing and institute a more balanced approach.

Fearing failures, SC changes test benchmarks

The South Carolina Education Oversight Committee voted in September to change the way the new state test used for NCLB accountability purposes is scored so more students would be labeled “proficient.” The change was made because the new Palmetto Assessment of State Standards (PASS) produced too many scores below "proficient."

Illustrating how high-stakes exams often lead to gaming strategies such as manipulation of cut scores rather than school improvement efforts, South Carolina officials voted to adjust the level required to meet federal Adequate Yearly Progress benchmarks. Instead of requiring students to score within the top two levels, students will be considered proficient for NCLB purposes if they attain what had previously been termed “basic,” or the third level.

"This is a confusing mess," said Sen. Mike Fair, R-Greenville, about the process of determining the state's new scoring categories, though he could have been describing NCLB-induced education practices in general. The National Center for Education Statistics released a report in late October that found 26 instances of states lowering their proficiency bar from 2005 to 2007. This is a predictable response to the federal requirement that is expected to cause most states to fail 70% to 100% of their schools by 2014. (See the FairTest report on this at  http://www.fairtest.org/sites/default/files/AYPproblemandSolution091807.pdf.)