Caste and Credit: A Woeful Tale?

Abstract

This paper examines caste-based differences in farmers’ access to bank loans in rural India. We investigate whether banks practice taste-based discrimination on the basis of caste. In order to identify potential discrimination, we consider loan applications and approval decisions separately. We find significant inter-caste differences in application rates, and evidence of discrimination against Scheduled Tribe borrowers at the approval stage. To rule out the role of statistical discrimination, we simulate unobserved credit histories with various distributions. Evidence for taste-based discrimination persists despite accounting for unobservables. However, we find that this discrimination does not affect small farmers.

Acknowledgements

We are grateful to Shaun Hargreaves-Heap, Arjan Verschoor, Kunal Sen, Ashwini Deshpande, Ashwin Nair, Anaka Aiyer and Amarjyoti Mahanta, and two anonymous referees whose comments helped improve the paper substantially. The usual disclaimer applies. Kumar acknowledges support from UNU-WIDER, and this paper incorporates findings from WIDER Working Paper 86/2016 ‘Why does caste still influence access to agricultural credit?’. The data used in this paper are available in the public domain while the code is available on request from the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

SUPPLEMENTAL DATA

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2018.1425397

Notes

1. Equality in access to economic resources has traditionally been viewed as a vehicle to promote broader social equality. For example, this is the rationale behind passing the Equal Credit Opportunity Act in 1974 (and the subsequent amendment in 1976) in the United States.

2. Akerlof (1976) provides one of the early theoretical explanations regarding the persistence of caste-based discrimination given informational costs in an Arrow-Debreu general equilibrium framework.

3. There are substantial, enduring complexities and political agitation around the categorisation of castes as OBC (Ramaiah, 1992).

4. STs are not technically part of the caste system. Yet tribals are one of the most disadvantaged peoples in India, and most analyses of caste therefore include the ST category, as does the current paper.

5. It is generally agreed that the caste system applies not only to Hindus but to followers of other religions in the Indian subcontinent as well.

6. It is important to distinguish the idea that this demand may vary according to group-belonging and the notion that demand is ‘exogenous’ with respect to discrimination. The latter may not be true if individuals alter their demand as a result of past discrimination faced by members of their group. We thank an anonymous referee for raising this point.

7. See LaLonde (1986); Duflo, Glennerster, and Kremer (2007).

8. Although we refer to credit histories, the approach is more general, and can be interpreted in terms of any other key unobservable which might be considered relevant, for example the quality of land, effort, or productivity.

9. Though there is no equivalent legislation to the United States’ Equal Credit Opportunity Act in India, the RBI has long encouraged banks to lend to ‘priority-sectors’, including SCs, STs and small farmers. For instance, a recent circular on priority lending targets and classification specifies a target of ‘18 percent of Adjusted Net Bank Credit (ANBC) or Credit Equivalent Amount of Off-Balance Sheet Exposure, whichever is higher’ towards the agricultural sector. The corresponding target for ‘weaker sections’, of which SC, ST, small and marginal farmers are a part, is ‘10 percent of ANBC or Credit Equivalent Amount of Off-Balance Sheet Exposure, whichever is higher’. See https://www.rbi.org.in/scripts/NotificationUser.aspx?Id=9688Mode=0.

10. The National Sample Survey Organisation’s All India Debt and Investment Survey is the other main source of information on household’s access to credit. This is a decennial survey, conducted most recently in 2013. In keeping with previous rounds, it asks about the details of existing loans but does not ask about loan applications.

11. In particular, the north-eastern states account for only 37 farmer households who applied for a loan. Since this sample is too small to implement regression analyses, we are unfortunately forced to drop these states from our estimations even as a large proportion of tribal groups in India reside in these states.

12. We combine the caste categories ‘Brahmin’, ‘Others’ and ‘Forward/General’ into one category, which we refer to as Brahmin hereon. Thus, this category includes all those who do not belong to the OBC, SC, and ST categories.

13. This pattern is also borne out by data from the National Sample Survey Organisation. See Table 2 in Mishra (2008).

14. According to the Reserve Bank of India’s guidelines, banks should not demand collateral for loans of up to Rs 50,000, and for larger loans land is the main form of collateral (Reserve Bank of India, 2007).

15. In the case of lending to small businesses, banks are also known to gather ‘soft’ information on creditworthiness in the absence of credit ratings (Berger & Udell, 2002).

16. Kumar (2013) uses data from the National Sample Survey Organisation which lists past repayments, and finds that these rates are not correlated with caste-group, but this analysis excludes STs.

17. The predicted approval rates STs in table 43 are lower than the average of approval rates for Brahmins, OBCs and SCs by 0.051 (S.E. = 0.026, p-value = 0.047).

18. We obtain qualitatively identical results for loan approvals using linear probability models that incorporate district fixed effects, but with less statistical power. The results are available upon request.

19. The choice of a binary confounder is a simplification over a more generalised, continuous confounder. Ichino et al. (2008) show, through Monte-Carlo simulation, that this choice is conservative: if results are in fact vulnerable to the inclusion of a confounder, then a simulation exercise using a binary instead of continuous confounder is less likely to lead to the (mistaken) conclusion that results are robust.

20. These definitions are along the lines of Ichino et al. (2008).

21. Of course everyone with a good credit history need not apply for a loan.

22. The blank spaces arise from unfeasible combinations of {µ_non, µ_ST, b} following Equation (10).

23. To be clear, we are not claiming to empirically test whether this policy guidance has worked. Since our data are cross-sectional, we cannot examine changes over time, and thus relate changes in policy to changes in access. Instead, our focus is on the presence of caste-wise differences at a given point in time, within a wider policy context that has long emphasised access to credit for disadvantaged groups.

24. A possible explanation may be that small farmers are more likely to apply for loans due to the presence of relatively poorer social or credit networks from which they can borrow informally, unlike larger farmers. We thank an anonymous referee for pointing this out.

25. See Kijima (2006) for a detailed background and discussion of the socio-economic status of STs.

26. In our data, population proportions as a whole and not just for farmers, are: Jharkhand (36%), Chhattisgarh (29%), Madhya Pradesh (16%), Orissa (17%) and Gujarat (13%). These are still the five states with the highest proportion of STs.

27. They might instead resort to informal sector borrowing, but these loans typically have higher interest rates and are on adverse terms compared to bank loans. In this paper our focus is on understanding access to formal bank loans.

28. Dréze et al. (1997) have argued that the distinction between lender discrimination and differences in application rates is largely irrelevant, because any hesitation to apply for a loan in fact reflects differences in expectations that are based on experiences that might span several years or more.

Additional information

Funding

Part of Kumar's work for this project was supported by the UNU-WIDER.