Subsidies to microfinance institutions: how do they affect cost efficiency and mission drift?

ABSTRACT

The costs and benefits of subsidized microfinance are still controversial. We utilize a cost-function estimation approach that accounts for the double bottom line (social and financial) of microfinance institutions (MFIs) to evaluate how subsidies affect both cost efficiency and risk of mission drift. We control for endogenous self-selection into the business models of credit-only versus credit-plus-deposit. Our results suggest that MFIs that both supply loans and collect deposits need no subsidies to be cost-efficient. In addition, subsidies to these MFIs are associated with an increase in deposit size, which might hurt the most disadvantaged depositors. In sum, combining subsidized funds from donors with deposits increases the risk of mission drift, and can therefore be socially undesirable.

KEYWORDS:

• Finance
• microfinance
• cost efficiency
• scale economies
• subsidies

JEL CLASSIFICATION:

• O14
• D24
• G21
• O16
• F35

Acknowledgements

We would like to thank Sylvain Abramowicz, Manthos Delis, Marc Labie, Emir Malikov, Denis Nadolnyak, Fotios Pasiouras, Mike Tsionas, the participants at the 6th European Research Conference on Microfinance, Paris (France), and at the 9th International Conference of the Financial Engineering and Banking Society, Prague (Czech Republic), the editor, and two anonymous referees for many insightful comments and suggestions. We are grateful to the Association for Social Economics, which awarded the 2019 Warren Samuels Prize to this paper following its presentation at the ASSA Meetings in Atlanta, USA. This study was carried out within the framework of the Montpellier Business School Social and Sustainable Finance Chair funded by the Caisse dʻEpargne Languedoc Roussillon. Anastasia Cozarenco is a member of the Montpellier Research in Management research center (MRM; EA 4557, Univ. Montpellier). We thank Roxanne Powell for excellent copy-editing.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 For surveys of the literature on the efficiency of financial institutions, see Berger and Mester (2003), Berger and Humphrey (1997) and Berger, Hunter, and Timme (1993). For a recent meta-analysis of the efficiency of MFIs, see Fall, Akim, and Wassongma (2018).

2 This classical empirical approach comes directly from the duality of profit maximization/cost minimization, where the cost is minimized subject to production technology constraints. The cost function is jointly estimated with the conditional factor demand or cost shares (derived by the Shepardʻs lemma). Since each equation is a function of input prices and output quantities, the seemingly unrelated regressions (SUR) method is often used as it produces the most efficient estimates of the system.

3 We start by using three input prices – of capital, labor, and financial capital. We next impose homogeneity on input prices by normalizing the prices of labor and financial capital, and on the total cost by dividing them with the price of physical capital. Thus, we end up with two share equations.

4 The national saving rate is the difference between gross national income (GNI) and public and private consumption, plus net current transfers (% of GNI).

5 We obtained similar empirical results with a larger model (results available upon request) including the six additional Worldwide Governance Indicators (Kaufmann, Kraay, and Mastruzzi 2010), such as control for corruption, which may affect cost efficiency (Al-Azzam 2016).

6 Other social missions claimed by MFIs include serving women and rural borrowers (Mersland, Nyarko, and Szafarz 2019).

7 Compulsory savings are a common feature of the lending technology used by both regulated and unregulated MFIs (Armendáriz 2011). They are often characterized as the “hidden collateral” of microcredit. Voluntary savings, by contrast, drive demand for specific deposit products, which are valuable to poor people (Collins et al. 2009). The difference between all savings and voluntary savings is significant: Voluntary savings only involve 68% of the total number of savers and represent 78% of the total volume of savings in our sample. Please note, however, that 76% of all observations concern MFIs that do not require any compulsory deposits.

8 The detailed MIX dataset that we purchased provides the year of establishment of each MFI. Age is computed as the difference between the fiscal year and the year of establishment.

9 In the cost Equation (2)(2) $\begin{array}{c}lnC={\alpha }_0+{\sum }^{{\alpha }_j}ln{q}_j+{\sum }^{{\beta }_k}ln{p}_k+1/2\sum {\sum }^{{\alpha }_{ij}}ln{q}_{i}ln{q}_j\\ +1/2\sum {\sum }^{{\beta }_{lk}}ln{p}_{l}ln{p}_k+\sum {\sum }^{{\delta }_{jk}}ln{q}_{j}ln{p}_k+\gamma { }^{\mathrm{\prime }}x+lnv\end{array}$(2) and the cost-share Equation (3)(3) $s_k={\beta }_k+{\sum }^{{\beta }_{lk}}ln{p}_l+{\sum }^{{\delta }_{jk}}ln{q}_j+ln{v}_k,k\in \left\{L,K\right\}$(3) , the standard errors of the coefficients are adjusted for selection bias, as in the standard Heckman (1979) model.

10 The demeaning makes computations of the scale economies much easier.

11 The reason for not using average loan size as a proxy for social mission here is methodological. Within the cost-function framework, cost must be an increasing function of outputs. In practice, however, total cost of MFIs tends to depend non-monotonously on average loan size (because of large fixed costs). By contrast, it increases with both the number of clients and the volume of loans/savings. Therefore, we do include the two components (numerator and denominator) of average loan size in the cost function, but as separate outputs. This cost-function design has also the merit of emphasizing the trade-off, if any, between financial and social performance and, ultimately, makes it possible to indicate whether there is any mission drift.

12 This detailed database was only available against payment of a fee.

13 There are about 2.5 observations per MFI and we clustered standard errors at MFI level in all the models. Therefore, the analysis is close to cross-sectional and a time trend may be unreliable. Reliance on mostly cross-sectional data also suggests that the results reflect the long-run efficiency in the industry, while heterogeneity is accounted for within the panel context, thus generating relevant results for policy purposes.

14 Table A9 in the Appendix shows the distributions of both MFIs and MFI-year observations by country.

15 We could have chosen an alternative definition, which equates subsidies with donations reported in the income statement (DʻEspallier, Hudon, and Szafarz 2017), but under this definition, an MFI might be categorized as subsidized given a positive accumulated donated equity in the balance sheet without it having effectively received any donation during the observational period. Therefore, we opted for the conservative approach to make our empirical results more robust.

16 Since the variables are demeaned prior to taking logs and normalized, the coefficient of the derivative with respect to the output/input price is the marginal effect.

17 Yet the production rather than the intermediation approach to MFI cost function specification with deposits as output remains valid because none of the output coefficients are negative, that is to say, cost does not decrease with savings outputs.

18 We estimated a SUR model to assess whether our results depend on CMP accounting for endogenous self-selection into a business model (results available upon request). Overall, the SUR model produces similar economies of scale for unsubsidized and subsidized credit-only MFIs. For credit-plus-deposit subsidized MFIs, however, the SUR model reveals two notable differences when compared to the CMP: First, the number of active borrowers is no longer significant and, second, the relationship between total cost and deposits in USD is significant at the 5% level. These differences confirm the message from our significant Mills ratio, namely, that endogenous self-selection is a major feature in the microfinance industry.