The unreliability of the test at the center of Michigan’s House Bill 5111 (why thousands of 3rd graders would be held back arbitrarily, due to test error) — House Bill 4822 (2015)

I sent a letter to the Michigan members of the House of Representatives on 11/8/2013 to warn about House Bill 5111 (2013). The proposed law would hold back 3rd graders (for a year) if they do not pass Michigan’s standardized reading test. By crunching the numbers, I see that this type of law will disproportionately affect (and negatively so) child English language learners (ELLs) in Michigan. Sadly, this all eventually did pass (as House Bill 4822 (2015)). The amendments to the Bill are unsatisfactory, in my opinion. If you have questions about this, please feel free to call or email me.

Here is the original Bill as introduced in 2013:

Here is what has passed:

Here is some news on the issues:

Here is my 2013 letter, sent to all Representatives by me via email, on 11/8/2013

Dear Representative,

I read with interest the proposed Michigan legislation (House Bill 5111) that would require third graders to repeat third grade if they score lower than “proficient” on the Michigan Educational Assessment Program (MEAP) reading test. As a testing researcher, I implore all legislators to consider this proposal from the perspective of the test because the policy, were it to be implemented, assumes scores from the test are accurate for implementing the policy. The test would be retrofitted: the cut-off between “proficient” and “partially proficient” would be used in a new way. But the cut-point may not be as precise as imagined. We need evidence that any child’s score assignment of “partially proficient” versus “proficient” is fail proof and not prone to measurement error (testing error or unexpected performance due to factors other than reading proficiency).
Unfortunately, as testing specialists know, all tests have measurement error. No test is fail proof, especially when children are involved. Children do not all always try their best. They get distracted, tired, anxious, or bored. Some don’t understand the directions, and their teachers—unlike on other days—aren’t allowed to explain, which can be confusing and stressful. Sometimes children just decide to make patterns on bubble sheets. These are, after all, eight and nine-year-olds.
Luckily, how error-prone a test is can be estimated. This statistic is called standard error of measurement. It tells the chances of a child getting “partially proficient” versus “proficient” due to testing error (due to factors other than his or her true reading ability). This is a handy statistic, and it appears in the MEAP’s technical test manual. I used it to estimate how accurate the proposed cut-point is (as a dividing line) using the most recent set of publicly available data, the 2011-2012 third-grade-reading test.
In 2011-2012, 108,800 third graders took the reading test. Test-error accounted for 6.19% of third graders getting the score band of “proficient” when they should have gotten “partially proficient” or lower, and 3.85% were assigned “partially proficient” when they should have gotten at least “proficient.” This means that if the law had been in place in 2011-2012, 6,735 (6%) would have moved on to fourth grade and 4,189 (4%) would have been held back for no reason other than measurement error (and not based on their reading skills). In total, 38,156 (35%) would have failed third grade, and 10,924 (10%) would have passed or failed at random. The test cannot accurately divide students into pass and fail groups.
The third grade MEAP reading test was designed to describe the academic reading level of students (by districts, schools, and student subgroups) in terms of average reading expectations across the state of Michigan. It was not designed to identify students for grade repeating. I say leave the decision to hold a child back to his or her teachers and parents. They will be able to accurately identify a child’s needs to repeat third grade with far greater accuracy, validity, and fairness than any single test ever could.


Paula Winke, Ph.D.

Associate Professor, Department of Linguistics and Languages, B252 Wells Hall, 619 Red Cedar Road, Michigan State University, East Lansing, MI 48824