2.1 MFI mission drift: profitability and outreach

In 2000, Morduch argued that higher interest rates do not necessarily imply a reduction in credit allocation. He found evidence that financial sustainability allows MFIs to have a more extensive range. Like Woller, Dunford, and Woodworth (1999), Morduch discussed the debate between institutionists and welfarists, as an example of mission drift. He found that while the objective of the former group was to increase financial sustainability and develop the credit portfolio, the objective of the latter was to eliminate extreme poverty.

In 2006, González and Rosenberg found results that reinforced the arguments for MFI mission drift; they found that 44% of MFIs are more profitable than commercial banks. Likewise, they analyze the relationship between MFI profitability and outreach, which led to evidence of mission drift. It was identified that either MFIs are profitable or wideranging, but not both. Cull et al. (2007) similarly found a weak relationship between profitability and outreach. They found that, even when an MFI is granting small loans, it does not lose profitability. They also found statistical evidence of a relationship between the size of the loan and operating cost. Their findings suggest that the higher the loan, the lower the average operating cost, which is evidence of incentives to mission drift.

Other studies, such as Cotler and Rodríguez-Oreggia (2008), have identified evidence of a significant relationship between profitability and outreach. Through a positive correlation between profitability and average loan size in a Mexican MFI sample. In their study, they suggest two alternatives to resolve the MFI mission drift concern: an increase in productivity, or a reduction in funding costs. Regarding this matter, Alinsunurin (2014) found that MFIs have not yet been able to integrate the dual objectives of profitability and social impact. Alinsunurin also found that non-profit MFIs were more efficient in social impact but less profitable than for-profit MFIs. This result is similar to that of Pop and Buys (2015), which also identified MD in Romanian MFIs. Furthermore, Pop and Buys also found that MFIs are concentrated in developed cities rather than in communities. In regards to the studies that do not support MFI mission drift, Kar (2013) found no evidence that for-profit organizations have dropped their primary objective. However, he does report an existent lack of reliable information in his study.

2.2 Explanatory variables of MFI financial performance and outreach

The literature addressing variables affecting profitability and outreach found that the legal status of MFIs has a significant impact, Cull et al. (2009). As for-profit banks grant more substantial individual loans, in proportion of total portfolio, than non-profit institutions. They also argue that non-governmental organizations have higher operating costs and lower profitability levels compared to for-profit counterparts. In 2011, the same authors established that when for-profit MFIs are regulated, they show significant profitability margins. It was also found that regulation is negatively related to outreach.

Regarding the effect of capital structure, Bogan (2012) found that the amount of equity, measured as a proportion of total assets, is relevant for both profitability and outreach. He also found that donations jeopardize sustainability by reducing operating efficiency. Like Kar (2012), he found that the higher the debt to equity ratio (leverage), the better the financial performance. However, he did not find that leverage had a significant effect on outreach. In regards to this, Pati (2014) found that because MFIs have access to a variety of financial sources, their capital structure has a significant effect on profitability.

When searching for variables that stimulate the core mission of MFIs, Cull et al. (2014) found that the more developed the financial system, measured as bank penetration through branches and ATMs per square kilometer, the higher the impact that MFIs would have on outreach to the low-income population, and in turn this is reflected in an increase of microcredits, including those for women. Pati (2015) found that capital structure, operating expenses, and asset quality are the primary drivers of MFI financial performance and outreach. When analyzing the impact of interest rates on sustainability, Nwachukwu (2014) identified an inverse U-shaped relationship between these two variables, which suggests that there is an optimal MFI interest rate. Also, she did not find any substantial evidence suggesting that small loans imply a higher sustainability risk. Balammal et al. (2016) found that legal status, the regulatory environment, the age of the MFI, number of employees, capital structure, and cost per client have a significant impact on MFI financial performance.

As discussed, in the previous studies we acknowledge a lack of consensus regarding the existence of mission drift. However, none of these studies analyzes an indirect relationship between the variables. In this work, we do not attempt to prove the existence of mission drift. Instead, we attempt to explain how specific independent variables, most commonly cited in MFI literature, affect profitability and outreach.

3.1 Effect of the environment, capital structure, operating efficiency and size on profitability.

In this section, the variables we utilize, allow us to create a construct of profitability, environment, capital structure, size and operating efficiency.

First, to measure profitability we build a construct, essentially a linear combination from ROA, ROE and OSS based on the works of Cull, Demirgüç-Kunt and Morduch (2009), Pati (2015) and Gutiérrez-Goiria and Unceta (2015); those studies intend to prove that MFIs have conflicting objectives because they do not represent a sustainable business model. The first independent variable is the Environment because according to Cull et al. (2014), the setting in which the MFIs operate is an essential determinant of both performance and outreach. For this work, based on Kauffmann et al. (2007), we built the construct with corruption control, the rule of law, and government effectiveness. Our second independent variable is the Capital Structure. For this construct, we use equity and interests’ expenses, following Pati (2014), who made a comparative study about capital structures for MFIs around the world. The third dependent variable is the Size of the MFIs. Following the studies of Cotler and Rodríguez-Oreggia (2008) and Cull et al. (2011), for this construct, we use the staff currently employed and the number of active borrowers. Finally, we formed Operation Efficiency construct, as a measure of the operating expenses.

Before proceeding to the analysis, it is necessary to examine whether the constructs meet the previously mentioned criteria from Jarvis et al. (2003) beginning with the exploratory factor analysis. In this sense, according to Nunnally (1978), this criterion is met when the Cronbach alpha is higher than 0.8, which is the case for the entire construct, as shown in Table 3. The next step is to verify the relationship among indicators through the Kaiser-Meyer-Olin (KMO) test and the Bartlett Sphericity Test. The KMO is a reliability test, which measures whether the constructs have a common variance that must be higher than 0.5. In this case, our results in Table 3 show that, except for operating efficiency, all constructs have a common variance. Also, in Table 3 we demonstrate that the Bartlett sphericity test, which measures whether the factorial construction is appropriate, is indeed significant, and the total percentage of accumulated variance as well as the contribution of each factor to it.

The principal component analysis is performed though the Varimax Rotation Method, which is suggested in samples that contain just a few variables in each factor, allowing maintaining a significant proportion of variance in each construct (Abdi, 2003). To evaluate the model, we use a confirmatory factor analysis (CFA). In Appendix three can be seen that for each construct, the correlation is higher than 0.8 and the variance percentage that each factor captures is higher than 80%. This result indicates that each factor is a good representative of its respective variable. In addition, it conveys statistical evidence of multicollinearity among each construct. As we previously mentioned, allowing this assumption in the model is an advantage of SE models, concerning OLS multiple regression analysis. Figure 1 shows the results of the confirmatory factor analysis.

[Figure ID: f1] Figure 1.

Measurement model, financial performance

  —Source: author using AMOS software.

To prove goodness-of-fit in the model, we must use more than one indicator (Feinian, Curran and Bollen, 2008). The first two parameters are the Chi-Square and chi square over degrees of freedom. In Table 5, we show that the chi-square is significant. According to Hu y Bentler (1995), whether the result of the division of the chi-square divided by the degrees of freedom is higher than two, it this implies statistical evidence that the SEM is not valid. However, Lei and Wu (2007) prove that SEM is well specified and valid if the model meets the following conditions: the sample is large enough and NFI, CFI, and GFI are over 0.9 (Bentler y Bonett, 1980; Bentler, 1989; Joreskog and Sorborn, 1986). As we can see in Table 5, there is evidence of model goodness-of-fit in our proposed model. The final goodness-of-fit is the RMSEA, which must be under 0.8, which is the case in our model (Steiger and Lind, 1980).

Following the work of Orozco-Gómez (2016), we apply a convergent and discriminant validity of SEM methodology. First, we extract the average variance (AVE) of each element of the construct. According to Carr (2002) and Fornell and Larcker (1981), this average must be over 0.5. In Table 5, we demonstrate the results of this test. As we can see, our model shows convergent validity. Additionally, we estimate the square of the correlations of each pair of factors and then compare it with the average variance extracted from each factor (Campbell and Fiske, 1959). To prove discriminant validity, the square of correlations must be less than each AVE. In Table 6, we show the AVE in the diagonal and under the respective square correlations. As we see in Table 6, our model meets the discriminant validity criteria

In this section, we analyze financial performance as a dependent variable, while we treat the other variables as an independent. We also establish a causality relationship and a maximum verisimilitude criterion. In Figure 2, we show evidence that capital structure, size, and environment do not have a significant direct effect on financial performance. Instead, the effect is through operating expense.

[Figure ID: f2] Figure 2.

Model for financial performance

  —Source: author with AMOS software.

In Table 6, as is expected, the coefficient of the relationship between operating efficiency and financial performance is negative. The capital structure and environment coefficients are negative and significant but only for operating efficiency, not for financial performance. The same is valid for size, but with a positive sign. The latter is statistical evidence that capital structure, size, and environment do not have a significant direct effect on financial performance. Instead, the effect is through operating expense. We also tested the model without the environment measure, although the results, concerning significance, coefficients are very similar to those presented in Table 6. Our significant negative relationship between financial performance and operating efficiency result is consistent with the results of Cull, Demirgüç-Kunt and Morduch (2011 and 2014), Nwachukwu (2014) and Pati (2015). However, our result differs from the findings of Bogan (2012) and Kar (2012), as it demonstrates an indirect relationship between capital structure and financial performance through operating efficiency, as opposed to their result, which shows a direct link between these variables.

Regarding size, our result is consistent with the works of Cull et al., (2011), Bogan (2012), Pati (2014 and 2015), Kar (2012) and Gutiérrez-Goiria and Unceta (2015), Cull, Demirgüç-Kunt, and Morduch (2011 y 2014), who find a positive but not significant relationship between size and financial performance. However, unlike our case, they did not relate this variable indirectly with financial performance.

3.2 Effect of the environment, capital structure, operating efficiency, and size on outreach

In this section, we develop an analysis equivalent to the one presented in the former section, but instead of financial performance, we utilize outreach as a dependent. As a measure of outreach, we use an average loan size. According to VanRoose and DEspallier (2013), this is an indicator of the segment that MFIs serve. For example, when the loan size is small, MFI works primarily with the low-income segment. For this analysis, we use the variable Average Loan per Borrower (LOANBORR). In this section, we test whether the environment affects outreach, capital structure, operating efficiency, and size. Furthermore, we test whether this relationship is direct or indirect using operating efficiency as a mediator. Because the steps of the methodology were explained in the previous section, in this section, we present and explain the results only.

First, in the exploratory factor analysis, we found that the factorial construct is appropriate and that the individual variance of each factor is reflected in the main variance of the model (see Table 7).

To prove the goodness of fit, we have indicators that show that the factorial model proves to be adequate for the outreach model (see Figure 3 and Table 8).

[Figure ID: f3] Figure 3.

Measurement model, outreach

  —Source: author.

The convergent and discriminant validity are positive; thus, we decided to develop the structural equation model. In Figures 4.a and 4.b, we show the model testing the relationship between independent variables and outreach. In these figures, we show the coefficients and significance. As shown in the case of the financial performance analysis, we found both a significant direct and indirect relationship between outreach and independent variables. Figure 4.a shows the result without environment and 4.b with it. As we see, when we add environment (KKM 3, 5 and 6), we find a negative and significant relationship with operating efficiency, and with outreach (LOANBORR). However, in this case, both the coefficient and significance of the effect of operating efficiency on outreach decreases. We validate the results of Figure 4.b in Table 9.

[Figure ID: f4] Figure 4.a.

Model for outreach, without environment

  —Source: author using AMOS software.

[Figure ID: f5] Figure 4.b.

Model for outreach, with environment

  —Source: author using AMOS software.

We also tested the model by removing operating efficiency as a mediator variable, but coefficients and significance are very similar, as we see in Table 9.

As we note in Table 9, operating efficiency proves to be, as expected, negative and significant related with outreach, which is consistent with the results of Cull et al. (2009 y 2014), Nwachukwu (2014) and Pati (2015). Also, our result for the capital structure is both positive and significant, as stated by Cotler and Rodríguez-Oreggia (2008) and Kar (2012), but in our work, this only occurs when the effect is through operating efficiency, as opposed to a direct effect.

Regarding the size of the MFI, we find a positive and significant relationship with outreach, through operating efficiency, unlike Cull et al. (2011), Kar (2012) and Gutiérrez-Goiria et al. (2016). Finally, our results suggest that the environment has a significant adverse effect on outreach, which differs from the findings of Cull et al. (2011) and (2014).