Inter-Group Contact, Racial Stereotypes and Academic Performance
Contemporary societies have become increasingly diverse, and prejudice and stereotypes towards certain groups are at the forefront of public debate. Stereotypes may induce distortions in economic and social behavior, generating adverse consequences for groups subject to negative stereotypes, including unequal access to economic, social and political opportunities.
Can increased interaction between groups reduce stereotypes about out-group members?
The answer to this question is theoretically unclear. Allport’s (1954) ‘contact hypothesis’ maintained that, under certain conditions, exposure to outgroup members would allow individuals to better understand their characteristics and points of view, thus diminishing reliance on stereotypes and eventually ameliorating inter-group relations. On the other hand, a recent literature on the ‘negative contact hypothesis’ contends that negative contact makes categories more salient than positive contact, potentially leading to an increase –as opposed to a reduction–in negative outgroup stereotyping (Paolini et al., 2010; Barlow et al., 2012).
To answer the above question, we test whether a policy designed to exogenously generate exposure to members of different racial groups induces changes in attitudes and stereotypes, and whether this translates into sizeable academic performance gains for the individuals involved.
Students in mixed-race rooms at University of Cape Town, South Africa
We study inter-ethnic interactions in the context of South Africa, a country where the experience of apartheid made people relatively prone to stereotyping and led to the economic marginalization of black South Africans. In particular, we take advantage of a policy implemented by the University of Cape Town (UCT) with the aim of promoting racial integration. The policy randomly allocates students to roommates, thus providing a unique opportunity to examine the causal effect of a roommate’s race on individual attitudes and behaviour. To measure the impacts of the policy, we interviewed students at the beginning and end of their freshman year. By comparing stereotypes and academic performance at the beginning of the academic year - that is, before students are allocated either to a mixed race room or to a same race room – and at the end of the academic year, we can investigate whether interacting with someone of a different race can change individual stereotypes towards that race and influence academic outcomes.
The University of Cape Town (UCT) is a public university and it enrols approximately 5000 incoming freshmen every year, more than half of whom live in university residences. We surveyed students in double rooms living in residences that applied the random allocation policy. Out of the 499 students who participated in our study, 155 were allocated to a roommate of a different race and 344 were sharing the room with a student of their own race. The racial composition of this sample was as follows: 332 respondents are black, 117 are white, 18 are Coloured and 32 are Indian, Asian or other race.
Implicit Association Test to measure implicit stereotypes
Since individuals may be reluctant to disclose prejudice or may not be fully aware of their own stereotypes, we measured racial stereotypes using the Implicit Association Tests (IAT). The IAT is an experimental method based on the idea that respondents who more rapidly pair two concepts in a rapid categorization task more strongly associate those concepts (e.g., how fast people pair images of black versus white people with descriptions of leadership roles). Slower speed in associating certain pairs denotes mental processes that tend to perceive those pairs as less common. This tool is a more objective measure of stereotypes and it has been widely employed in social psychology to understand implicit cognition, that is, cognitive processes of which an individual may not be aware and that include, among others, perception, stereotyping, and memory. For our purposes, IATs have the advantage of avoiding social desirability bias in self-reported answers and implicitly reveal attitudes that individuals may be uncomfortable disclosing or not fully aware of, such as racial prejudice. This is particularly relevant in the South African context. Thus, we use IATs (in our case called Race IAT) to complement subjective and self-reported perceptions of inter-ethnic attitudes with more `objective' measures of racial stereotypes.
Negative value of the IAT means higher prejudice against blacks. As shown in the figure below, prior to the random allocation of students across rooms (at baseline), implicit association tests revealed that all participants held negative stereotypes about black students relative to white students. On average, white students held more stereotypes against black students when compared to students of other races.
Figure: Stereotypes as measured by Race IAT at baseline
Inter-racial contact reduce prejudice for whites and improves academic performance for black students
The results point to a number of positive effects from the inter-racial interactions. For white students, living with a roommate of a different race during their first year in college reduced negative stereotypes, as measured by the IAT and increased positive attitudes towards other races.
Students who are paired with roommates from different races also tended to have more positive interactions with members of other races outside the room. For students in mixed rooms, we observe a higher value of an index which captured individuals’ friendships with members of other races. The effect on white students’ index of friendships was stronger than the effect on black students. Students also reported talking more frequently about race and feeling more comfortable discussing about it.
Finally, our results show that the policy significantly increased academic achievement among black students. On average, black students’ grade point average improved, closing one third of the gap in performance between black and white students observed before the policy. Further, black students in mixed race rooms were also more likely to pass exams and continue on to the following year of college. The positive effect on the likelihood of passing exams and continuing university persisted into black students’ second year at university, when they were in a different residential setting.
These results on academic performance of black students are not due to the fact that a black exposed to a white is on average exposed to a higher ability peer. Interestingly, we find that inter-racial dialogue and better knowledge of the roommate turn out to be important ingredients in explaining academic gains, consistent with a reduction in stereotype threat.
To what extent our findings generalize beyond the context we study? Our experiment takes place in South Africa, a place that is certainly at the higher end of the distribution if one considers its history of inter-racial conflict and the significance of racial stereotypes. If anything, we think that this should have reduced the ex-ante probability of success of a policy like the one we study. First, many would assume that such stereotypes were so deeply ingrained that it would have been difficult to affect them; second, because a large number of `transformation' initiatives have been put in place in South Africa since the end of apartheid, the marginal effect of the roommate allocation policy we study may have been assumed to be small. The fact that in-depth interaction with the other group can reduce negative stereotypes even in historically charged contexts is a very encouraging message in this respect.
A second issue to consider in assessing external validity is that our sample is made of university students who are certainly not representative of the entire South African population. On the one hand, these students are likely to be the political and economic elites of tomorrow's South African society, so that a change in their attitudes and behavior can bring a certain significance for the country as a whole. On the other hand, if we consider how this may have affected our estimates, the direction of a potential bias is not obvious. If these students are more `malleable' than the general population, then our estimates would be an upper bound. But if they were already more open-minded and less reliant on preconceptions when they entered university, then there would have been a lower margin for the intervention to have an effect and the bias could be in the opposite direction.
Overall, we believe the specificity of the South African context should not be seen as a limitation of our results but, if anything, a strong proof of concept. More work to investigate the generalizability of our findings to different social and economic environments will help improve our ability to design integration policies in our increasingly diverse societies.