To catch a cheat: the pressures of accountability may encourage school personnel to doctor the results from high-stakes tests. Here's how to stop them - Research

Education Next, Winter, 2004 by Brian A. Jacob, Steven D. Levitt

Classrooms identified as "most likely to have cheated" experienced gains on the initial spring 2002 test that were nearly twice as large as the typical Chicago classroom (see Figure 2).

On the retest, however, those excess gains disappeared completely--the gains between the spring 2001 test and the spring 2002 retest were close to the system-wide average. In stark contrast, classrooms identified as potentially having good teachers scored even higher on the reading retest than they did on the initial test. Their math scores fell slightly on the retest, but they continued to post extremely large gains. The randomly selected classrooms also maintained almost all of their gains when retested, as would be expected.

What Predicts Cheating?

Using the indicators of teacher cheating described above, we conducted a series of analyses to examine the relationship between cheating and a variety of classroom and school characteristics. Several striking findings emerged.

On the one hand, classrooms that performed poorly the previous year were much more likely to cheat. For example, a classroom that scored one standard deviation below average the previous year was 23 percent more likely to cheat the next year. Likewise, classrooms in schools with lower achievement, higher poverty rates, and more African-American students were more likely to cheat. Classrooms with a higher proportion of students included in the official test reporting (in other words, fewer kids designated as disabled or limited English proficient and thus excluded from the accountability system) were also more likely to cheat--a 10 percentage point increase in the proportion of students in a class whose test scores "count" increased the likelihood of cheating by roughly 20 percent. Because a greater portion of their students would contribute to the overall assessment of the classroom and school under the accountability policy, these teachers perhaps felt more pressure to ensure that the students in their classrooms scored high on the exams. Teachers who administered the exam to their own students were approximately 50 percent more likely to cheat.

On the other hand, classrooms with students from multiple grades were 65 percent less likely to cheat than classrooms where all students were in the same grade. It is probably more difficult for teachers in such classrooms to cheat since they must administer two different test forms to students, which will necessarily have different correct answers. Classrooms in schools with teachers who graduated from more-prestigious under graduate institutions were also less likely to cheat; classrooms in schools with younger teachers were more likely to cheat.

Reusing a test form that had been administered in a previous year had no statistically significant impact on cheating.

This suggests that teachers' taking old exams and teaching specific questions to students is not an important component of what we are detecting as cheating (though anecdotal evidence suggests that this practice exists).