Double bottom lines in microfinance: are they mutually exclusive?

Abstract

Microfinance institutions (MFIs) are generally constrained by double bottom lines: keep the original social mission intact by serving the poor soutenue (the first bottom line) and attain financial sustainability (the second bottom line). However, these two bottom lines can be mutually exclusive in effect. This paper demonstrates how factor analysis, a multivariate statistical tool, can be utilized to examine this conjecture. First, factor analysis is used to construct two simulated performance indicators combining original variables, each one representing a distinct dimension of performance. Then, individual scores ascribed to MFIs on each factor are used as the dependent variables of two simultaneous-equations models that present new evidence on the determinants of MFIs’ performance and mission drift.

Les institutions de microfinancement (IFM) sont généralement limitées par une double mission: conserver intacte la mission sociale d'origine en étant au service des pauvres (la première mission) et parvenir à la viabilité financière soutenue (la deuxième mission). Cependant, ces deux objectifs peuvent en réalité s'exclure mutuellement. Le présent article démontre comment l'analyse factorielle, un outil statistique à plusieurs variables, peut être utilisée pour examiner cette conjecture. L'analyse factorielle est d'abord utilisée pour construire deux indicateurs de rendement simulé combinant des variables d'origine, chacun représentant une dimension différente du rendement. Les scores individuels attribués aux IFM pour chaque facteur sont utilisés comme variables dépendantes de deux modèles à équations simultanées qui présentent de nouveaux éléments probants concernant les déterminants des performances des IFM et de la dérive de leur mission.

Keywords:

• microfinance
• outreach
• sustainability
• panel data estimation
• factor analysis

Mots clés:

• microfinancement
• travail de proximité
• viabilité
• estimation en données de panel
• analyse factorielle

JEL Classifications:

• C15
• C23
• D23
• G21
• G32

Mots clés:

• microfinancement
• travail de proximité
• viabilité
• estimation en données
• de panel
• analyse factorielle

Acknowledgements

The author would like to thank two anonymous referees for their very helpful comments. Any remaining errors are, of course, the author's.

Notes

Source: The author's own calculation from MIX data.

Source: The author's own calculation from MIX data. * p < 0.05.

Source: The author's own calculation from MIX data.

Source: Author's own calculation from MIX data.

Note: All variables are in natural logs. GLP = Gross loan portfolio (this means scale of operation), NAB = Number of active borrowers (breadth of outreach), CPB = cost per borrower, OELP = operating expense to gross loan portfolio ratio, PAR = portfolio-at-risk past 30 days, WOR = write-off ratio, size = natural logarithm of total assets, age = number of years in microfinance operation, EAP = East Asia and the Pacific, EECA = Eastern Europe and Central Asia, MENA = Middle East and North Africa, SA = South Asia, SSA = Sub-Saharan Africa, NGO = Nonprofit non-governmental organization, rated = whether rated by any external authority. SUR = seemingly unrelated regressions with pooled data, FGLS = panel data feasible generalized least squares model with panel-specific AR1 correlation, OLS (P-W) = OLS or Prais-Winsten models with panel-corrected standard errors; FGLS and OLS (P-W) have been used for HPAC (panel heteroskedasticity, panel autocorrelation and contemporaneous correlation) SUR solution for contemporaneously correlated error terms. ***, **, * denotes significance at the 1%, 5%, and 10% levels, respectively.

Note: R-squared G2SLS = R²-Overall, FE2SLS = R²-Within and EC2SLS = R²-Overall. All specifications were two-way error-component models containing time-effects. However, their results are not presented. ***, **, * denotes significance at the 1%, 5%, and 10% levels, respectively.

1. Following Ahlin and Lin (2006), we computed self-sufficiency indices. The formula for the index in terms of operational self-sufficiency is: Sufficiency index = Operational self-sufficiency/(1 + Operational self-sufficiency).

2. Two chi-square values (chi-square with (n + k) degree of freedom and chi-square with n degrees of freedom, where n stands for the number of observations and k stands for the number of parameters) for formal likelihood ratio tests are obtained and high p-value fails to reject the null that the model for any specific number of factors is not significantly worse than the full-fit model.

3. We assumed normality in our data. As a check, however, maximum-likelihood factoring is used.

4. Additional results are available from the author through personal requests.

5. These approaches are suggestive, of course, not inclusive. Other important rules, for example, include: Velicer's (1976) Minimum Average Partial (MAP) test and Horn's (1965) Parallel Analysis. However, it is extremely important to remember that most of the above rules are someway ad hoc and subjective. All methods have several deficiencies. For limitations of these methods see, for example, Ledesma and Valero-Mora (2007). Therefore, it is always better not to rely on any single method or rule, but to evaluate and determine the solution as a complete picture on the basis of all of the methods employed. Hence, we exercised several methods.

6. There are two performance dimensions: social performance and financial performance.

7. Alternatively, a three-stage least-squares model could be used.