Equilibrium fictions, societal rigidity, and affirmative action

Karla Hoff 24 April 2012

a

A

India’s experience with quotas for women in public office suggests that within a generation, exposure to women leaders can erase the bias in men’s evaluation of female compared to male leaders and lift parents’ and girls’ aspirations by enough to close the gender gap in literacy (Beaman et al. 2009, 2012). Yet affirmative action sits uneasily with the values of individualism. Some argue that affirmative action aims to right a wrong against one group by a policy that wrongs a different group. Since two wrongs don’t make a right, affirmative action makes no sense. However, my research with Joseph Stiglitz suggests a new framework in which affirmative action does make sense (Hoff and Stiglitz 2010). Individuals want to view themselves as intelligent and will process information in a biased way to support that belief. An experiment with college students (Mobius et al. 2011) finds that people systematically discount bad news about their own intelligence, but display no bias when evaluating the intelligence of a robot.

By increasing self-confidence, self-image management can improve performance. Experiments that manipulate individuals’ self-confidence (Smith et al. 2008) find that greater self-confidence improves the ability to perform cognitive tasks.

Gender, race, and caste ideologies crimp self-image management. A study of high-school students (Correll 2001) finds that cultural beliefs about gender differences in math ability produce a gender bias in individuals’ perceptions of their math competence, controlling for ability. In a word-completion task (Steele and Aronson 1995), black Americans who expect to take an intelligence test are twice as likely as whites to fill in the missing letters in ways that indicate self-doubt – for example, to complete DU _ _ as DUMB and LO_ _ _ as LOSER; and self-doubt in turn hurts black people’s performance. An experiment in India (Hoff and Pandey 2006, 2011) finds that low-caste boys can solve mazes just as well as high-caste boys when their caste is not revealed; but in mixed-caste groups when caste is revealed, low-caste boys’ performance drops 20% below that of high-caste boys.

Joseph Stiglitz and I extend the model of confidence-enhanced performance (Compte and Postlewaite 2004) to suggest that stereotypes can be self-fulfilling because of their effect on the way people process information. Suppose that individuals undertake a series of projects, at each of which they can fail or succeed. And suppose that self-confidence, based on perceived success rates in the past, increases the probability of success. Panel A of Figure 1 shows the relationship.

If a person processed information objectively, his self-confidence would be based on the actual frequency of past successes. If he succeeded 70% of the time, he’d perceive a 70% success rate. If he succeeded 60% of the time, he’d perceive a 60% success rate. This idea is captured by the assumption that the 45o line in panel B determines self-confidence. Equilibrium (E) occurs where the confidence curve intersects the production function.

Figure 1. Confidence-enhanced performance and group stereotypes

Most people, as Mobius et al. found, discount bad news about their intelligence. As in Lake Wobegon, “all the women are strong, all the men are good-looking, and all the children are above average.” We capture this idea by supposing people “forget” a fraction of their failures or judge them not to be predictive of future success. Of course, if they never succeeded, they’d perceive a zero success rate (since any fraction times zero equals zero), and if they always succeeded, they’d perceive that too. But for any true success probability between zero and one, they perceive a larger success rate than the one they truly have. The bowed-out confidence curves in panel C captures this idea.

Here is a thought experiment. There are two types of people, reds and greens. They differ only in the way they process information. Compared to a red, a green is more able to forget his failures. This makes a green’s confidence curve more bowed-out than a red’s. Since equilibrium is where the confidence curve intersects the production function, it is clear in the figure that greens outperform reds.

Now reinterpret the experiment: There are two “races,” reds and greens. Reds have historically been characterised as inferior, race remains salient, and race constrains how individuals process information about themselves. Compared to greens, reds less easily forget their failures because they see them as confirming their inferiority. We define an “equilibrium fiction” as a belief that affects perceptions and so changes behaviour in ways that make the belief come true. In the equilibrium in Panel C, the racial gap in performance confirms the fiction of racial inequality even though the groups differ neither in ability nor in the costs and benefits of any economic action.

The cognitive frame created by ideology can sustain inequality because of three features of how the mind works – we think in terms of categories that are given to us by society (such as reds and greens), we have confirmatory bias (modelled in Rabin and Schrag 1999), and self-confidence can improve our performance.

The more the reds are stigmatised and susceptible to believing in their inferiority (e.g. because they see little disconfirmation in the circles in which they live), the wider the performance gap between races. There are many possible equilibria, each corresponding to a different stereotype.

If racist beliefs cause racial divides, what causes racist beliefs?

Given the post-Reformation fundamental belief that all men had rights, colonial powers after the 15th century constructed ideologies that the colonised groups they exploited were by nature inferior, and gave these beliefs precedence over other aspects of belief systems. Historical work finds that doctrines of race came into their own in the colonies that became the US after, not before, slavery (Davis 1999 [1975]); that out of the “scandal of empire” in India emerged a “race theory that cast Britons and Indians in a relationship of absolute difference” (Dirks 2001); and that arguments used by the settlers in Australia to justify their policies towards the Aborigines entailed in effect the expulsion of the Aborigines from the human race (McQueen 1971). As Montesquieu quipped, “It is impossible for us to suppose that [these peoples] should be men; because if we supposed them to be men, one would begin to believe that we ourselves were not Christians.”

But why would the target of another group’s self-serving ideology believe it? When the French make jokes at the expense of the Belgians, it doesn’t impair the self-confidence of the Belgians. Yet labels that correspond to structural inequalities have power. When a group is represented as distinct and when for situational reasons it is underrepresented in high-status roles, it is human nature to view the group as inherently inferior. The situational factors are discounted. Psychologists (Ross and Nisbett 1999) call this the fundamental attribution error.

Economists (Bertrand et al. 2005) have found evidence that psychologists’ tests of implicit attitudes can predict economic behaviour. By self-reported measures, black Americans and white Americans express a similar in-group preference, as shown in the first two bars of Figure 2, labelled “explicit.” By implicit measures, white Americans show an even stronger in-group preference. But blacks do not. Instead, they show no in-group preference or even a weak preference for whites relative to blacks. Psychologists (Nosek et al. 2002) argue that the contrast between explicit and implicit preferences among black Americans reveals “the deep reach of culturally acquired attitudes that are reflected in behavioural and brain activity”.

Figure 2. Explicit and implicit racial preferences of Black and White American respondents

Can affirmative action change stereotypes? A natural experiment

In 1993, India adopted gender quotas for village governments, with the quotas assigned to randomly selected villages. In villages with little or no experience of quotas, men evaluate the competence of female leaders compared to male leaders in a biased way, and a higher percentage of boys than girls are in school and are able to read and write. Yet exposure for almost ten years to local women leaders eliminates the bias in men’s perceptions and erases the gender gap in educational outcomes. All the evidence (Beaman et al. 2009, 2012) suggests that the material gains to girls and women as a result of the affirmative action policy are explained by changes in the way females are perceived by others and perceive themselves, rather than by any changes in opportunities.

Conclusion

For economists to ignore the factors that affect how we process information as part of the interpretation of economic change would be as wrong as to ignore the evolution of technology itself. Ideology shapes what we see and how well we perform. Ideology can give rise to “equilibrium fictions.” In our framework, changes in power, technology, and contacts with the outside world matter not just directly but because they can lead to changes in ideology. The results of India’s natural experiment with political affirmative action bear out this prediction.

References

Beaman, L, R Chattopadhyay, E Duflo, R Pande, and P Topalova (2009), “Powerful Women: Does Exposure Reduce Bias?”, Quarterly Journal of Economics, 124: 1497-1540.
Beaman, L, E Duflo, R Pande, and P Topalova (2012), “Female Leadership Raises Aspirations and Educational Attainment for Girls: A Policy Experiment in India”, Science, 335:582-586.
Bertrand, M, D Chugh, and S Mullainathan (2005), “Implicit Discrimination”, American Economic Review, Papers and Proceedings, 95(2):94-98.
Compte, O and A Postlewaite. (2004), “Confidence-Enhanced Performance”, American Economic Review, 94:1535-1557.
Correll, SJ (2001), “Gender and the career choice process: the role of biased self-assessments”, American Journal of Sociology, 106(6):1691-1730.
Davis, DB (1999 [1975]), The Problem of Slavery in the Age of Revolution, New York, Oxford University Press.
Dirks, NB (2001), Castes of mind: Colonialism and the making of modern India, Princeton University Press.
Hoff, K and P Pandey (2006), "Discrimination, Social Identity, and Durable Inequalities", American Economic Review, Papers and Proceedings, 96:206-211.
Hoff, K and P Pandey (2011), "Making up people: The behavioral effects of caste”, unpublished.
Hoff, K and JE Stiglitz (2010), “Equilibrium fictions: A cognitive approach to societal rigidity”, American Economic Review, Papers & Proceedings, 100:141-146.
McQueen, HO (1971), “Racism and Australian literature”, in FS Stevens ed., Racism: The Australian Experience, Vol 1: Prejudice and Xenophobia. Sydney, Australian and New Zealand Book Company, 115-122.
Mobius, MM, M Niederle, P Niehaus, and TS Rosenblat (2011), "Managing Self-Confidence: Theory and Experimental Evidence", NBER Working Paper Series 17014, Cambridge.
Nosek, BA, MR Banaji, and AG Greenwald. (2002). “Harvesting implicit group attitudes and beliefs from a demonstration website”, Group Dynamics, 6:101-115.
Rabin, M and JL Schrag (1999), “First impressions matter: a model of confirmatory bias”, Quarterly Journal of Economics, 114(1):37-82.
Ross, L and R Nisbett (1991), The Person and the Situation, Temple University, Chapter 1.
Smith, PK, NB Jostmann, AD Galinsky, and WW van Dijk (2008), “Lacking power impairs executive functions”, Psychological Science, 19(5):441-447.
Steele, CM and J Aronson (1995), “Stereotype threat and the intellectual test-performance of African-Americans”, Journal of Personality and Social Psychology, 69, 797-811.

 

a

A

Topics:  Poverty and income inequality

Tags:  affirmative action, India, race, sexism, caste