Social Capital and Access to Credit: Evidence from Uganda

Abstract

We use a nationally representative survey in Uganda to study the links between social capital and financial access. Our results indicate a positive association between individual social capital and access to institutional credit, but no significant relationship between generalised trust and credit access. The effect of individual social capital is more pronounced for poorer people, in rural areas, and in areas where generalised trust is low. Individual social capital seems to promote access especially to semiformal and informal financial institutions.

Acknowledgements

We thank the Editor (Oliver Morrissey), two anonymous reviewers, Simon Cornée, Karen Ellis, Pertti Haaparanta, Ari Hyytinen, Derek Jones, Antti Kauhanen, Päivi Kankaanranta, Ariane Szafarz and Otto Toivanen for useful comments. We also benefited from the comments of the participants of the World Bank Conference on Measurement, Promotion and Impact of Access to Financial Services in Washington DC, Workshop on Wellbeing in Low Income Communities at the Helsinki School of Economics, HECER-WIDER Autumn Seminar in Helsinki, and CERMi seminar in Brussels. This research is supported by the Academy of Finland (Projects No. 122398 and 120234). The data used in the paper and the associated stata code will be available by writing to Dr. Kalmi ([email protected]). The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For recent studies on access to financial services around the world, see Consultative Group to Assist the Poor (CGAP, 2009) and Ardic, Chen, and Latortue (2012).

2. Straightforward aggregation of individual social capital is challenging due to cross-person externalities, as Glaeser et al. (2002) point out. This is why we use completely different measures for individual and aggregate social capital rather than constructing the aggregate social capital variable by aggregating our individual social capital variable.

3. Bonding social capital refers to strong ties within the community, in contrast to bridging social capital, which refers to weaker ties between communities (Woolcock & Narayan, 2000).

4. Guiso et al. (2004) used the aggregate-level social capital variable in examining the interaction between social capital and judicial efficiency. Unlike them, we do not have a direct measure of legal enforcement efficiency in our data set.

5. The Credit Reference Bureau was established in Uganda in 2008 (Tumusiime-Mutebile, 2008), after the collection of the dataset we use.

6. There have been several financial access studies in Africa based on FinScope methodology, and their use in academic research is becoming increasingly common (for example, Aterido et al., 2013; Johnson & Nino-Zarazua, 2011).

7. These data are available from www.worldvaluessurvey.org.

8. It is not uncommon that respondents have taken loans from several types of institutions. For example, 25 per cent of formal institution borrowers in the FinScope data have also taken loans from semiformal institutions, and 18 per cent of semiformal borrowers also have loans from informal groups.

9. The question of whether an individual had been turned down by a financial institution was asked from only those respondents who had actually borrowed at least from non-institutional sources. It is possible that there are individuals who were refused credit and did not borrow at all, but the data do not allow us to identify such individuals.

10. Examples of other attitudinal statements available in the FinScope data include: ‘I participate in communal work’, ‘I attend parties’, ‘I attend cultural functions’, and ‘I participate in sporting activities’. The problem with these statements is that they may also measure things other than individual-level social capital, such as household wealth or consumption of services (see Glaeser et al., 2002).

11. In 2006, there were 56 districts and five provinces in Uganda.

12. The wealth variable is based on a methodology developed by the Uganda Bureau of Statistics (Steadman Group, 2007).

13. Our main results are robust to running the regressions also without the weights.

14. Literacy may promote access, because literate persons are able to check the contents of contractual agreements, which is of importance at least in the formal and semiformal sectors.

15. These results are not presented here but are available from the authors upon request.

16. When a loan is received from a friend, the content of the individual social capital variable is not completely distinct from the dependent variable.

17. We have estimated the regressions reported in Table 3 also by including non-borrowers in our sample. This does not affect the results qualitatively. These results are available from the authors upon request.

18. Another option would be to add an interaction term between individual social capital and wealth indicator to probit specification I of Table 2. We also tried this, and the direction of our results remains the same. However, the calculation of the size and significance of the marginal effects is not straightforward when interaction terms are added to a nonlinear model.

19. Heikkilä et al. (2013) found that social capital influenced the sorting of respondents into semiformal and informal financial institutions in a qualitatively similar way. Therefore and to simplify the analysis, we combine these two categories.

20. It is often customary in this kind of calculation to set all dummy variables to zero: after all, any real observations are either male or female, rather than something in between. However, because the probability changes associated with other than social capital variables are not our primary concern, and setting all dummy variables to zero produces probabilities that are rather different from their actually observed distribution, we think that calculating probabilities in this way gives a better understanding for the resulting marginal effects.

21. The probabilities and their changes have been calculated from the ‘raw’ coefficients reported in Table 4 by using the prvalue command in Stata.

22. This result does not survive tests where non-borrowers are also included in the comparison category.