Test graders show bias

In a study in India, teachers hired as test graders gave higher scores when the student was identified as high caste, lower scores to low-caste students, reports Inside School Research.

The Harvard study tested elementary and middle-school students in mathematics, language and art.

The tests, however, were randomly assigned different student characteristics. One student, for instance, would be listed on a cover sheet as a member of the Brahmin caste, the highest of India’s four social groups, while another might be described as being in the lowest caste, the Shudra. The tests were also graded separately by a research staff member who had no knowledge of any of the students’ characteristics.

As might be expected, the results showed that teachers, on average, assigned scores to students from low castes that were 3 percent to 9 percent lower than those of students who were described as being from a high-caste group. What was particularly interesting, though, was that teachers from low-caste groups were driving most of that discrimination; no evidence of bias could be found for teachers from high-caste groups. And they tended to direct it most often toward the lowest-performing students in the low-caste group—the students who presumably best fit the stereotype.

Gender didn’t influence grades, except for “the highest-performing girls, who were graded slightly more harshly than their high-performing male counterparts,” writes Debbie Viadero.

About Joanne


  1. tim-10-ber says:

    how did these results compare to the actual scores when the caste of the system was not known…sadly the article does not say…

  2. It does say the bias flattened out toward the end of the session; theory is that at the beginning, the scorers were trying to find the range and made some assumptions about who would represent them. That is a sort of bias, but a different problem.

    That’s why many of us skim an entire set of essays before actually grading them. Personally, I then sort them so I’m alternating strong and weak papers — keeps me more consistent.

  3. I misplaced my affirmative action score card.

    Is it OK to belabor Indians for their still very common discrimination on the basis of caste or are Indians a protected specie in the same sense that Muslim treatment of women is an unacceptable subject for polite conversation?

  4. This is why grading should be done “blind”…

  5. I do think this is very interesting, and at the same time not surprising. However, perhaps people should think closer to home before they (we!) judge. Pupils all over the world can be (and often are) labelled, whether it’s for being of the “wrong” caste or social background, or for their intelligence or behaviour. In the UK some recent research suggested that children often get a bad reputation as soon as they start school – based on their parents’ background or some early behavioural problems (inability to sit still etc). This is at age 5 and can be carried throughout their education.

  6. This is particularly interesting as we consider the new education plan to evaluate teachers based on student test scores. Why would an accomplished teacher be willing to work in a high-needs school?

  7. This sort of prejudice in judging appears to be a problem outside India. While I don’t know of any work specifically on teachers, blind auditions for orchestras has been found to make a difference to who is offered a job, and blinding peer reviewers has been found to make a difference to which first-authored papers get published.


  1. […] This post was mentioned on Twitter by kriley19 and JoanneLeeJacobs, Legal Lawyer. Legal Lawyer said: Test graders show bias « Joanne Jacobs http://bit.ly/JuxKf […]