ÌÇÐÄvlog¹ÙÍø

How do people change their behavior after being made aware of bias?

This question is central to efforts aimed at reducing discrimination in education, workplaces, and other settings. A recent paper entitled Revealing Stereotypes: Evidence from Immigrants in Schools by CID faculty affiliates Michela Carlana and Eliana La Ferrara, along with co-authors Alberto Alesina and Paolo Pinotti, investigates how revealing implicit stereotypes to teachers impacts their grading of immigrant and native students in Italian middle schools.

The study combines data from over 1,300 teachers in Northern Italy and two experiments. The first, a field experiment, randomly revealed to teachers their own Implicit Association Test (IAT) scores—measuring bias against immigrants—either before or after they graded their students. The second, an online experiment, compared two types of interventions: a generic debiasing message about stereotypes in society and personalized feedback on each teacher’s own IAT score. Teachers were then asked to grade hypothetical tests with names randomly assigned to sound either native or immigrant.

Key Findings:

• Bias in Grading: Immigrant students receive lower teacher-assigned grades compared to native students with similar standardized test scores. This grading gap is especially pronounced for high-performing immigrant students and is correlated with teachers’ implicit biases.
• Impact of IAT Feedback: Providing teachers with personalized IAT feedback before grading reduced the immigrant-native grade gap by 27%, primarily by increasing grades for immigrant students. The effect is particularly strong around the threshold that determines whether a student passes or fails a subject.
• Mechanisms: On average, providing personalized IAT feedback does not change grading relative to generic debiasing messaging. However, teachers with stronger implicit biases only adjust their behavior when given personalized feedback, particularly if their results are unexpected.

Impact and Relevance:

This study is motivated by the profound impact that teacher bias can have on immigrant students’ educational attainment, with implications for their long-term human capital development and economic opportunities. Both generic messaging and personalized feedback on implicit stereotypes are effective in reducing grading disparities on average, but the latter works best among teachers with stronger biases. The most cost-effective alternative depends on the policymaker’s specific goals and the strength and prevalence of biases among the teacher population.

The results show that IATs, which are relatively simple to implement, may be a powerful tool to collect metrics that may be used to counteract negative stereotypes about certain groups. These findings extend beyond education, offering lessons for reducing bias in domains such as hiring, performance evaluations, and law enforcement.