When Mass Media Fail: Blaming Teachers

Reading primary literature requires careful attention to details especially when the paper remains to be peer-reviewed. Working papers are quite common especially in the field of education research. It is therefore important for mass media to be thoughtful and critical when disseminating information based on social science research that is yet to be critically examined by experts.

One recent example comes from the NPR in the United States. The report entitled Hard Evidence: Teachers' Unconscious Biases Contribute To Gender Disparity talks about a research conducted in schools in Israel.

Above copied from NPR
Fortunately, there is a comment section in the online edition of the above report where a real informed remark from a user named Rational Discourse can be read. Obviously, such comment was not part of the original airing. Those who heard the story are therefore quite unlikely to have seen this important comment. The comment from Rational Discourse, first of all, provides a link to the primary literature discussed in the above report. The study cited comes from a working paper from the National Bureau of Economic Research. The idea that teachers hold biases against girls with regard to math is based on the data presented below:

Above copied from: "On The Origins of Gender Human Capital Gaps: Short and Long Term Consequences of Teachers’ Stereotypical Biases". Victor Lavy and Edith Sand. NBER Working Paper No. 20909. January 2015. JEL No. J16,J24.
© 2015 by Victor Lavy and Edith Sand. All rights reserved.
The work starts with scores from two different exams. The National Score exam comes from a standardized exam administered at the end of fifth grade while the School Score exam comes from tests administered and graded by the teachers sometime during the sixth grade. The difference between these two scores is tabulated for both genders. The user Rational Discourse immediately notices that these are two different exams given at two different times. To equate this difference with a teacher's bias is clearly inappropriate.

I teach chemistry and parts of my test require calculations. It is difficult for me to see how my bias can affect how such a test is graded. In my opinion, this likewise applies to elementary mathematics where correct answers do not really leave that much room to subjectivity. Perhaps, these exams have partial credit, but even in such cases, guidelines on how to award partial credit can be made objective.

It is also important to note the magnitude of these differences especially when compared to the standard errors (also shown in the table in parenthesis). A previous post in this blog, Insights from Gender Differences, summarizes what gender studies on education have really shown:
At the early ages, there are cognitive and verbal differences between boys and girls. And when boys and girls grow up, there are likewise differences in experiences. Looking at these differences allows us to see the possible variations among children in general, regardless of gender. One simply has to take note that differences are usually found as mere fractions of a standard deviation. This means that there is indeed a great deal of overlap between the two genders. From a different perspective, this implies that variations within a gender are actually larger than the differences between boys and girls.
Lastly, NPR although it mentions the correlation found between parental education level and the chance that a girl would choose higher levels of math in secondary school, the title of the report simply does not do justice to this other perhaps even more important part of the study. The study finds that the gender disparity in math and sciences disappears if only children from families whose father and mother have comparable educational attainment are considered. As the study shows, this is even a stronger correlation.

Teachers may not be perfect. And there are indeed well founded concerns, but discouraging girls from doing more math by giving lower scores is clearly not.