IDENTITY AND BIAS: evidence from Israel
12 Feb 2020
Driving test assessors are biased in favour of members of their own ethnic group and the opposite gender.
This is the key finding of new research by Asaf Zussman and Revital Bar published in the January 2019 issue of The Economic Journal, which analyses practical driving tests conducted in Israel between 2006 and 2015.
While the first result – that professional evaluators are biased in favour of members of their own ethnic group – is in line with the typical finding in the academic literature, the second is novel. The authors offer a utility-based interpretation for findings
Looking at the tests – more than 2.5 million tests in total – the study leverages the fact that assignment of students to testers is effectively random. About 25% of the students taking these tests were Arab (the rest were Jewish) and 50% were female. Of the 236 testers who conducted the tests, 9% were Arab and 9% percent were female.
The authors’ analysis provides robust evidence that in the decisions they make, testers – government employees who are expected to assess students’ driving abilities in an impartial manner – favour students from their own ethnic group and from the opposite gender. Specifically, we find that a student is 14 percent more likely to pass a test when assigned a tester from the same ethnic group and 11 percent more likely to pass a test when assigned a tester from the opposite gender (the average pass rate in these tests is about 40%).
The authors argue that these patterns are consistent with a utility-based model in the spirit of that first introduced by economist Gary Becker in The Economics of Discrimination (1957). The key element in the model presented in the book – which focused on racial relations in the United States – is that agents (e.g. employers) incur different levels of utility from contact with members of different racial groups.
When thinking about the issue, it is quite clear that the rationale for racial discrimination described above does not easily carry over to gender discrimination since both men and women usually do not shy away from – and in many situations even prefer – interacting with members of the opposite gender. Thus a simple extension of a Becker-type, utility-based, model to include gender preferences would naturally predict bias in favour of one’s own group in the case of race or ethnicity and in favour of the opposite group in the case of gender. The paper’s findings are consistent with these predictions and suggest that testers seem to reward members of groups whose company they enjoy.
The authors provide several pieces of evidence to support this interpretation. The paper argues that if bias is indeed driven by the different levels of utility testers derive from interacting with members of different groups during the test, it is natural to assume that this effect would decline with physical distance between testers and students.
To explore this hypothesis, we replicate our analysis of bias in private vehicle tests using data on the universe of driving tests for motorcycle licenses, where the student and the tester drive different vehicles and are thus not in close proximity. Since there is only one female tester conducting motorcycle tests, we focus on ethnic bias. Consistent with the hypothesis that physical distance matters, we find no evidence of ethnic bias in motorcycle tests.
Identity and Bias: Insights from Driving Tests by Asaf Zussman and Revital Bar is published in the January 2020 issue of The Economic Journal.
Associate Professor | Hebrew University of Jerusalem | firstname.lastname@example.org
PhD candidate | Hebrew University of Jerusalem | email@example.com