In many, many cases, people have a preference for working and doing business with those who share the same religious beliefs, come from the same geographic region, or have something else in common. If this preference arises from discrimination against other groups – if there is economically inefficient favoritism – the economy will not reach its full potential. But could there also be efficiency gains from transacting with people who are culturally proximate? If so, is it possible for the gains to be large enough to more than offset the losses from discrimination? Surprisingly, the answer to both questions is yes. However, that does not mean the barriers between groups should be reinforced. Policies that break down informational barriers between groups could produce further gains.
Why do, as they say, “birds of a feather flock together”? Very often, a preference for doing business with one’s own kind is seen as a matter of ill-guided tribalistic beliefs about the trustworthiness (or lack thereof) of others, or simply a “taste” for discriminating against those who are different from ourselves.
However, there is another more positive force that can compel individuals to prefer transacting with those who are “culturally proximate” to themselves: because of shared beliefs, a shared language, and a shared social circle, it may be easier to assess the reliability of a prospective partner before agreeing to have him as a supplier, offer her credit, or buy a used car from him. And it may similarly be easier to seek recourse if the buyer finds himself with defective goods or the lender isn’t paid back, via their shared social network (and of course given that both parties know this upfront, hopefully avoid any problems in the first place). Finally, even a personal affinity for others of one’s own kind can have its efficiency benefits – if people suffer less psychic distress from cheating people different from themselves, then the greater trust they have in their own kind is justified.
…a personal affinity for others of one’s own kind can have its efficiency benefits
There are many, many examples in the economics of groups – whether defined by place of origin, religion, or some other feature – showing that people have a preference for working with their own kind. In China, entire industries may be dominated by entrepreneurs who were all born in the same hometown; in East Africa, some sectors are similarly dominated by business owners originally from South Asia. Understanding the extent to which these economic “enclaves” are driven by discriminatory or efficiency considerations (or some combination of both) is important for understanding why they exist, and whether we should aim to do anything about them.
Distinguishing empirically between discriminatory and efficiency motivations for in-group transactions is challenging, however: for the most part, we merely observe that like trades with like, without having any clear benchmark to evaluate the efficiency consequences. If white customers tend to patronize white-owned stores, we do not know how much this increases their purchases, or whether the products they buy end up better (or worse) at suiting their needs.
A deeper problem comes from the fact that the socially similar seller and buyer tend to transact together more often. In the extreme, we may not observe any comparable transactions between dissimilar buyers and sellers to serve as a benchmark against which to compare the in-group transaction. Or, for example, if there is a higher bar for transacting with those outside one’s circle, we may erroneously conclude that out-group transactions are generally more efficient when in fact we simply only observe them when a buyer comes across an “offer he can’t refuse” from an outgroup individual, an offer so enticing that he overcomes his personal repugnance.
So, to study the frequency and consequences of in-group versus out-group transactions, we need something close to random matching between the two sides of the relationship, and we need a way of measuring the outcome’s efficiency. These are the features that make Indian banking such an attractive setting for studying the role of cultural proximity on economic outcomes. There is a clear measure of efficiency – if the borrower defaults, it indicates that the bank’s capital has been misallocated – and because the bank we study (and Indian banks in general) tend to rotate their loan officers around the country at high frequency, we can explore what happens to the quantity and quality of a branch’s loans when the religion or ethnic identity of the branch manager changes. The broader social fabric of modern India makes it a suitable testing ground for theories of in-group versus out-group transactions. The caste system provides a set of well-identified social cleavages across the several government-classified groups that are identified in our data. It is also a country with a recent history of religious conflict – especially between Hindus and Muslims – that could stoke the types of out-group animosity that could affect economic transactions.
We use data from a large state-owned bank in India, which provided us with five years’ worth of detailed credit and personnel records which we used to match all borrowers and branch head officers to their religion and caste, providing a pairwise measure of the cultural “distance” between lender and borrower. We consider each of the country’s main religions, Hindu, Muslim, Christian, Sikh, Parsi, Buddhist, and “Others” as distinct groups, and among Hindus also distinguish among the four government-classified caste groups (General Class, Scheduled Tribes, Scheduled Castes, and Other Backward Castes). In all, we have ten distinct “cultural” groups.
Because officers are rotated every three years, we can account for the possibility that, for example, Hindus are more creditworthy than Muslims in general (or General Caste relative to Scheduled Caste borrowers): if that is what was driving the higher rate of credit access among Hindus, there is no reason to expect any sudden changes when the identity of the branch manager switches from Hindu to Muslim. Because our data also include the information on whether the loan goes into default, we can also examine whether the default rate among Hindus depends upon whether the loan is issued by a Hindu or Muslim branch manager.
In looking at how the manager’s identity affected a branch’s loan portfolio we found that there was a discontinuous jump in how much lending is made to, say, Muslims when a Hindu manager is replaced by a Muslim one, and a corresponding drop in the loans received by Muslims when the Muslim manager departs and is replaced by a Hindu one. But we do not see any such changes around branch manager turnover when the transition involves two managers from the same group. If the results themselves aren’t shocking – after all, the most intuitive versions of both the “information” and “discrimination” stories line up with these findings – the cleanness of the patterns we observe in the data, which we show in below set of graphs, is nonetheless striking.
Figure 1: New credit around officer transitions, partitioned by group identity of officer and borrowers before and after the transition
What was more surprising – at least to us – is what we found when we turned to examine the effect of the branch manager’s identity on the quality of lending to different groups. Here, information versus discrimination make opposing predictions. Discrimination in favor of one’s own kind would predict that, when a Hindu manager is replaced by a Muslim one, the default rate among Muslim borrowers should go up, as the new branch head lends to his (in-group) buddies rather than evaluating Muslim borrowers based on credit quality. If his arrival improves the bank’s ability to assess the creditworthiness of Muslim borrowers (and then turn the screws on Muslim borrowers to make sure they repay), then his arrival should result not just in more loans being made to Muslims, but also better quality loans.
We expected the “discrimination” effect to dominate. But we found the opposite – replacing an out-group branch manager with an in-group one decreases the default rate by 0.6 percentage points
Given the tensions between India’s various ethnic groups, as manifested most visibly in Hindu-Muslim riots that have occurred periodically over the past half-century, we expected the “discrimination” effect to dominate. But we found the opposite – replacing an out-group branch manager with an in-group one decreases the default rate by 0.6 percentage points (7 percent, relative to the baseline default rate of 8.6 percent). While we cannot look at interest rates – the bank does not leave this to the branch manager’s discretion – we find that this improved repayment rate occurs despite lower rates of collateral demanded on in-group loans. Discrimination might well still occur on the basis of personal distaste, but its effect on default is, based on the evidence, outweighed by the beneficent effects of shared background or culture.
As we have already observed, cultural proximity might improve lending quality through two distinct mechanisms: by helping the loan officer better evaluate prospective borrowers’ creditworthiness before a loan is approved, and by making it easier to enforce repayment afterward. While it is extremely difficult to distinguish between these two channels empirically, we also offer up several pieces of suggestive evidence which indicate that better screening plays at least some role in the improved credit outcomes for in-group lending. First, loans made by an in-group branch manager continue to perform well even after he is replaced by an out-group one. While it may be that the manager continues to apply pressure from afar, it is perhaps most easily reconciled with better screening than better enforcement from a bank office half way across the country. (It also helps to rule out the “evergreening” of in-group loans, whereby managers keep the loans from borrowers from their own group from going into default by issuing yet more loans to cover the interest.)
A more nuanced test, motivated by earlier theoretical research on the effects of improved lender information, looks at the dispersion of loans to in-group borrowers. The intuition behind this test is as follows: if a lender knows nothing about a group of loan applicants, there is little choice but to offer them all essentially the same loan contract. But if he learns more about each one (for example, through social connections or better knowledge of the community) he can better distinguish between the ones who are more likely to repay bigger loans – and hence are deserving of more credit – and those that can shoulder only a small amount of credit. Consistent with this type of improved screening of in-group borrowers, we find that when a branch manager arrives, the loan dispersion to borrowers from his group increases.
Discrimination might well still occur on the basis of personal distaste, but its effect on default is, based on the evidence, outweighed by the beneficent effects of shared background or culture
We see our results in part as a salient and important counterexample to the main effect of cultural proximity, economically inefficient favoritism. However, this take on cultural proximity as a vehicle for economic efficiency is a static one. One means by which we might be able to break down misunderstandings and animosities between groups is through greater interaction (though this is another case where the intuition seems clear, but the empirical evidence is mixed – exposure to other groups may reduce animosity as we find they’re “just like us,” or aggravate it because we look for the negative features of the other group that made us dislike them in the first place).
It also ignores that fact that in-group preferences – whatever their efficiency consequences – will disadvantage minority groups in a society. Consider, for example, a Christian borrower in our data – he has only a 2.1 percent chance of encountering a Christian branch manager when goes to apply for a loan. For Hindus, the probability is 93.8 percent. The fact that Christians are less likely to get loans because their Hindu lenders have less information on them doesn’t make the discrimination any less real – it is still a means by which a privileged majority gets a further leg up on disadvantaged minorities. This sort of phenomenon can also help explain why we think that minorities are bad credit risks: if we look at the credit access and repayment rates of Christians in our data, we’d conclude they are riskier borrowers than Hindus, if we don’t account for the fact that Christians’ low rates of loan access and repayment result in part from the fact that they deal with out-group lenders, whereas Hindus interact with in-group ones.
The better policy may then be to consider interventions – potentially ones that take place long before beliefs and mores are fully formed – that break down informational barriers between groups, rather than organizing society in a way that reinforces the existing ones, whatever the immediate payoff to economic efficiency might be.