1. INTRODUCTION

Microfinance institutions (MFI) aim to make a sustainable social impact by providing a credit flow to allow borrowers to gradually develop their entrepreneurial endeavors. In order for these institutions to ensure they can keep offering loans, they also need to worry about their own financial sustainability.

On the one hand, this sustainability requires tracking operating and financial economic indicators, considering income levels, wages, and other expenses the institution incurs to carry out its work. Nevertheless, analyzing loan payback rates-default rates-is also vital, as the loan portfolio quality is generally taken from the largest asset the MFI manages (^{Bekerman and Ozomek, 2003}).

Looking into the features of the borrowers, several questions arise: Are women better at paying loans back than men? How do indicators like borrower's level of poverty, education, degree of access to financing sources, financial education process, amount and/or term of the loan, evaluation and follow-up from each advisor, and geographic proximity of the borrower to the institution influence loan repayment and timeliness? In short, this paper analyzes the empirical results obtained from accessing the database of an institution with a 17-year history of granting micro-loans.

The first section summarizes some overarching concepts related to the topic of microfinance as a tool to access credit for low-income people, with the significance of dealing with loan defaults as a way to ensure that institutions are sustainable in the long term, as well as background information available in the literature and outcomes of other similar experiences in Latin America. The second section outlines the main traits of the Asociación Civil Avanzar. The third section talks about the methodology and the hypotheses guiding this work. The results are presented in the fourth section, analyzing how each of the variables considered behaves. The fifth section contains the econometric analysis, which provides technical support for the compositional study in the earlier section. Finally, we sketch out some conclusions.

2. THE IMPORTANCE OF ACCESS TO CREDIT FOR LOW-INCOME SECTORS AND CONTROLLING DEFAULT

3. ASOCIACIÓN CIVIL AVANZAR FOR HUMAN DEVELOPMENT

The institution was founded in 2000, led by teachers and students from the University of Buenos Aires Faculty of Economic Sciences as a non-profit organization. Its objective set out to help improve living conditions for the neediest micro-entrepreneurs, and it works with people in the southern neighborhood of Ciudad Autónoma in Buenos Aires, especially those most affected by poverty and lack of job opportunities (Villa Soldati, Villa Lugano, Mataderos, and Villa Riachuelo).²

To reach these objectives, Avanzar, on the one hand, extends loans to micro-entrepreneurs in the regions it works to prevent the lack of financing from limiting their growth possibilities, and, on the other, offers training for capacity-building and to great job opportunities for the vast number of people who are unemployed or who are already microentrepreneurs.

In the MFI universe, there are those essentially geared toward attaining sustainability and financial surplus, while others are more focused on socioeconomic impacts. Against that backdrop, although Avanzar does aim to deepen the social impact of the microloans it offers, its other goal is to ensure its own financial sustainability as the only way to make sure the project will last over time. This is one strength that sets Avanzar apart from other smaller non-profit institutions to the extent that it is following up strictly with its borrowers, checking on default and progress made to support financial sustainability. These regular activities are strengthened markedly through course development, mentorship, and encouraging people to participate on a range of virtual commerce platforms, all designed to boost capacity-building and make borrowers more autonomous.

With respect to larger or for-profit institutions, Avanzar’s mission is to reach the poorest of the poor, even if that entails a higher cost when it comes to granting loans. Higher costs are essentially due to the lower amount of the loans these sectors need and the higher levels of default found among people getting their first loan.

The fact that a loan is small (on average, Avanzar’s loans are around 300 dollars) means that the amounts these institutions earn in interest are equally small with respect to the effort required to get borrowers to comply with the minimum requirements to take out a loan. Undoubtedly, this requires, generally speaking, more human capital. The cost is exacerbated by the courses, workshops, and coaching (as already mentioned) offered to improve the quality of the entrepreneurs’ businesses. In Argentina, these human capital costs become much more onerous because wages are higher there than they are in other places throughout Latin America.

In that context, Avanzar has two nominal annual interest rate systems for the loans it grants to borrowers. One is for the most disadvantaged, at 24%, and the other is for those who have already made progress in managing their businesses, at 65%. These rates may seem high, but several clarifications bear keeping in mind. Number one, inflation in Argentina has hovered around 25% since 2007, with a couple of spikes in recent years, which has had a negative impact on the real value of the institution's portfolio. Number two, the inter-bank rate currently charged in Argentina is more than 50%, which rises to 70% for financing credit cards.

Finally, another distinction is worth remembering about Avanzar versus other MFIs in the country. It is that Avanzar has ties to the University of Buenos Aires, which means that its field activities are closely related to the research done by university students and scholars.

4. METHODOLOGY, HYPOTHESIS, AND CLASSIFYING THE BORROWERS BY DEFAULT CIRCUMSTANCE

When doing an intake for someone interested in taking out their first loan, credit agents capture a ton of data, including both qualitative and quantitative variables. Of 2,500 borrowers at Asociación Civil Avanzar who have taken out microloans, we managed to put together a complete database for a total of 861 borrowers, who had data for all of the variables considered. Then, we cross-checked that database with the internal default database: a monthly record that advisors keep to measure and identify unpaid installments.

There are 13 potential variables classified into three groups: personal, household-related, and credit-related. Likewise, the ex post credit experience analysis divided the participants into four groups pursuant to the degree to which they paid back their agreed installments.

On the one hand, there are those individuals called the “excellent payers" (EP) and the "good payers" (GP), comprising the "low-default borrowers" group (LDB). The first are perfectly compliant paying back their loans, and have never been behind on any payment during the years of analysis. The GPs are admitted with a maximum of a monthly delay, considering a contingent term and without any major administrative issue for the institution in terms of both liquidity and financial solvency.

On the other hand, the two groups comprising the significant delay group were denominated the “late payers” (LP) and the “defaulters” (DR), and both are “high-default borrowers” (HDB). The former group has been late on payment by one to three months; the defaulters have more than a quarterly delay. All of the borrowers began with Avanzar as EPs, but their statuses changed at some point throughout the time period analyzed if one or more payments was late. Accordingly, their status changed to GP, LP, or DR, depending on how late their payments were made.

By recognizing the salience of personal, household, and loan traits for the borrowers, it is possible to determine a set of hypotheses pertaining to the significance of the variables in explaining the financial behavior of borrowers.

5. RESULTS

Crossing the institutional data on borrowers with the information pertaining to late payments on a monthly basis, we classified individuals into their payment categories, which gave rise to a compositional analysis, comparing the proportional distribution of EP, GP, LP, and DR for each variable available, trying to identify any bias due to the distribution of the original sample (see Table 1).

Table 1 Classifying the Borrowers by Their Creditworthiness or Default Level 

Source: Created by the authors based on Avanzar´s records.

Of the 861 borrowers in the sample analyzed, nearly 80% were EPs or GPs, which is to say, LDBs. The HDBs, meaning those more likely to default or pay late, were the LPs and DRs.

Personal Variables

Age. The results made clear that default or late payment is lower among older borrowers, with nearly 85% of them scored as EPs or GPs. Accordingly, the cohorts of older individuals had approximately half of the number of defaulting borrowers as the rest of the categories did, as well as a marked decrease in the number of late payers. As such, the hypothesis on this matter was not valid (see Table 2).

Table 2 Default or Late Payment Circumstance by Age. 

Source: Created by the authors based on Avanzar´s records.

Nationality. The distribution of borrowers in terms of nationality points to one third each of Paraguayans, Argentineans, and Bolivians, with a minority share of Peruvians. The compositional analysis by category revealed that despite the fact that the late payment rate held relatively steady across the nationalities, there was a notable difference when it came to the DR variable. Argentineans and Bolivians had the most borrowers at least one quarter late on their payments (12.36% and 11.83%, respectively), as compared to the Paraguayan and Peruvian borrowers, with substantially lower rates (7.66% and 6.66%, respectively) (see Table 3). This is consistent with the opinion of the credit agents who evaluated borrowers of Paraguayan origin and found they are more timely payers, and once again does not fit with the original hypothesis of no difference due to nationality.

Table 3 Default or Late Payment Circumstance by Nationality. 

Source: Created by the authors based on Avanzar´s records.

Gender. It is essential to look at the values pertaining to gender distribution emanating from the study: of the 861 borrowers, 585 were women and 276 were men. In relative terms, 68% versus 32% points to a situation in which there are clearly more women than men involved in microfinance.³ The values obtained, nevertheless, did not display any significant differential in terms of payment compliance.

Education received. The education category was divided into categories by the maximum school level borrowers reached. They were screened by the following attributes: incomplete elementary school, complete elementary school, incomplete high school, complete high school, and university.⁴ The percentage of non-punctual payers rose as education level went up, beginning with 16.36% of individuals who did not complete elementary school to one-quarter of those who had some sort of tertiary-level education.

Civil status. The compositional analysis for this variable revealed that borrowers who entered “married” as their civil status accounted for a lower proportion of the defaulters. “Cohabitants,” in turn, had the highest number of excellent payer borrowers, confirming out hypothesis. In that context, it is worth bearing in mind that the proportion of married or cohabitating men who were borrowers was far higher than that of women. By contrast, all of the "separated" borrowers were women, and the distribution is clearly biased toward defaulting, with 22.58% of borrowers in that category.

Household Characteristics

Residence zone. Six residence zones were classified pursuant to proximity to the Asociación Civil Avanzar headquarters.⁵ Number one, keep in mind that most of the borrowers (77%) were living in homes adjacent to Villa Soldati and Volla Lugano, very close by to the Avanzar offices. It turned out that the three zones closest to the institution happen to be where the best results were found for loan repayment punctuality. The farther away a household was from the institution, the more drastically the results diverged, with late payments and defaulting going up. This confirms our hypothesis.

Housing circumstance. Borrowers living in lent property they do not own had default rates substantially lower than any of the alternatives, representing 4.41% of the individuals who confirmed this housing circumstance. Moreover, this group had the highest rate of excellent payers, at nearly 16 percentage points (51.47%) higher than in the other two housing groups. On another note, more than one-fourth of the borrowers renting their homes fell into the very late payer group, in contrast with 19.49% of the owner borrowers (see Table 4), which tends to confirm our hypothesis.

Table 4 Late Payment or Default Circumstance by Housing Status. 

Source: Created by the authors based on Avanzar´s records.

Household members. One relevant outcome is the high degree of compliance seen in single-member households. Essentially, they had a very low percentage of late payers and defaulters (5.56% and 7.41%), confirming the hypothesis. Another noteworthy outcome, confirming the proposed argument, is that families with seven or more members performed poorly,⁶ with nearly one third of these households classified as HDB.

Total household income. With respect to the family income variable, the choice was made to divide income into four groups, looking at the minimum and mobile wage at the time the research was done.⁷ Households with income less than 440 dollars, between 440 and 660 dollars, between 660 and 880 dollars, and income above two minimum wages.

The most relevant corollary to the analysis of this category refers to the marked drop-off in the number of DRs as total household income rises (see Table 5). Nevertheless, although households earning less than the minimum wage did exhibit high levels of defaulting, they did, at the same time, have a high proportion of EPs as compared to the higher income categories. These results may reflect, for certain borrowers, a lack of credit or business experience that leads to a default, and in others, better performance in light of the reality of a microfinance supply very restricted to them. Thus, the hypothesis is only partially met.

Table 5 Late Payment or Default Circumstance by Total Household Income, in Dollars. 

Source: Created by the authors based on Avanzar´s records.

Loan Characteristics

Credit status with other financial entities. Analysis of the sample showed that most of the borrowers (54%) did not have any sort of financial contact with any other institution, so Avanzar would represent their sole contact with the lending system.

Of those borrowers who did hold liabilities with third parties, 17% were involved with formal and identified institutions. However, 30% were engaged with institutions not identified as consumer credit, or durable goods, or productive loan bodies. This means that they were facing high interest rates or credit plans that did not reflect their development as microentrepreneurs.

What immediately emanates from this analysis is that borrowers who have some sort of tie to the institutions identified have an extremely low percentage of defaulting (4.17%). This might indicate that borrowers took some kind of financial education after making contact with these credit institutions. By contrast, borrowers involved with unidentified lenders had a much higher rate of default than the average (13.28%), pointing to vulnerability to indebtedness at very high interest rates (see Table 6).

Table 6 Late Payment or Default Circumstance by Credit Status. 

*(Banco Ciudad+ Banco Provincia + FIE + Cordial + Coopel).

Source: Created by the authors based on Avanzar´s records.

Amount of last loan and current loan. The distribution reflects that 25% of borrowers in the database did not have any prior loans, so this loan was their first. For those getting their first loan, they showed the same dual behavior as found in the “Total Household Income” category.

Along these lines, what emerges is that despite the fact that certain first-time borrowers behave excellently, the rest of these borrowers do display a high rate of default (13.7%), and a large number of them display behaviors indicating their loans will be uncollectible (17.81%). In other words, around one third of them fell into the high likelihood to pay late or default category (LP and DR). To the extent that credit amounts increase, and, therefore, the seniority of the borrower with the institution, the proportion of default and late payment falls. Once the loan amount surpasses the threshold of 4,000 pesos (290 dollars), borrowers' behavior changes markedly. In particular, for loans above the value of 440 dollars, borrowers scored as DR fell to one third of those getting a first loan (see Table 7).

Tabla 7 Late Payment or Default Circumstance by Amount of loan, in USD. 

Source: Created by the authors based on Avanzar´s records.

Current loan term. Borrowers were put into one of three groups: those with a loan term less than or equal to six months, between seven and eleven months, and more than twelve months.

One eloquent result that arises is the percentage of excellent payers in the first group, at 46.49%, as compared to 36.69% and 30.45% in the other two. Crossing this variable against the loan amounts corroborated the following: as the amount rises, borrowers prefer to distribute the capital across more installments, aiming to reduce the monthly pressure on their income. Essentially, most of the loans less than 290 USD (47.7%) were granted in six or less installments, but only 16.61% of these loans required payback over one year. As compared to that, most of the loans above 660 USD (70.9%) were given for longer terms, and only 10.05% were amortized over a time period less than six months.

One interesting outcome that surges from analyzing late payment in this variable cross-check is that there as a higher likelihood of default among borrowers with lower loan amounts when the terms are long. When loan amounts are low, what is most salient is to grant a short repayment term, because it underscores the impact of the dynamic incentive by shortening the loan renewal timeline, fostering a positive attitude among borrowers toward punctuality.

6. ECONOMETRIC ANALYSIS OF DEFAULTING BORROWERS

The econometric analysis took borrower default as the dependent variable. To this effect, a Logit model was built, aiming to establish whether the applicants’ traits had a significant impact on the dependent variable.

Model (1) was designed pursuant to the aforementioned hypotheses:

Y _default= α + δ₁nacparagu + δ₂age + δ₃gender + δ₄education + δ₅civilstatus

δ₆home + δ₇zone + δ₈householdmem + δ₉income + δ₁₀liability

δ₁₁lastloanamt + δ₁₂curramt + δ₁₃loantterm + ε (1)

Nevertheless, the lack of individual significance for some variables led to the construction of another reduced model (2). Both appear in Table 8.

Y _default= α + δ₁nacparagu + δ ₂age + δ ₃home + δ₄zone + δ₅ouseholdmem + δ₆income + δ ₇liability + δ ₈lastloanamt + δ₉curramt + δ₁₀loantterm + ε (2)

Table 8 Results of Econometric Analysis. 

Note: For all dichotomous variables, values are 1 and 10.

References:***significant variable at 1%; **significant variable at 5%; * significant variable at 10 %. In parentheses aret he p values of the individual significance test for the coeficients.

Source: Created by the autors.

Thus, the final model is determined by the following significant variables: a) nationality, because if the borrower is Paraguayan, the likelihood that the debt becomes uncollectable drops; b) age, because if the borrower is under 35 years old or older than 55 years old the likelihood of default falls; c) housing situation, because if the borrower rents, the likelihood of being untimely rises; d) residential zone, because proximity to the institution drives down the likelihood borrowers fall behind; e) number of household members, which drives up the likelihood of falling behind on a payment; f9 household income, because as income rises, the probability of default wanes; g) if the borrower has a loan with other unrecognized institutions, the likelihood of default goes up; h) loan amount, as an indicator of the longevity of the tie to the institution; negatively related to the likelihood of delay; i) loan term, because if it is more than six installments, the likelihood of defaulting rises; and j) if the borrower is taking out their first loan, the likelihood of defaulting is higher.

These econometric results match the results of the compositional analysis in the above section. Likewise, they are aligned with the results found in ^{Clavijo (2016}), who proposed a Probit model and another multinomial Logit model. The author reached the same conclusions, with the exception of the outcomes for gender, education level, and civil status, which turned out to be not significant in this study.

7. CONCLUSIONS

The hypotheses were made pursuant to practical intuition, comments from advisors, and theoretical interpretations emanating from the pertinent literature. However, not all of them were validated by the study experience, and some were ruled out. Such was the case of the variables nationality, age, sex, and education.

The above conclusions refer, in terms of the first variable, to more punctual repayments by Paraguayan and Peruvian borrowers, the former backed by the econometric analysis and comments by credit officers themselves. Where age was concerned, the notion that middle-aged borrowers behave better was ruled out, as they are even worse than the younger borrowers. Nevertheless, older borrowers perform better, which may be tied to receiving benefits from the government or from their families, which guarantee them a more stable income.

As far as the third variable, despite the existence of international experiences indicating that women are better repayers, this study did not demonstrate that they are more punctual than men are. In the econometric analysis, the variable of gender was discarded for having null significance. Even so, although the reason for bias toward women in receiving microcredit could not be justified from the standpoint of sustainability for the MFI, the far-reaching role female empowerment plays in terms of their economic autonomy is still recognized, the role of microfinance as traction for their productive insertion should still be fostered, as they are at a disadvantage compared to men at present.

The final rejected variable, education, reached the conclusion that higher formal school would not be an indicator of better creditworthiness; on the one hand, based on the variables of credit status with other institutions and current credit amount, it emerged that loan-related or financial education is what matters, coming not from school but from the practice of microfinance itself. Experience with loans is a factor that makes people pay them back in a more timely fashion, either because of their education or some tie with the institution. It is no triviality that ties to unrecognized institutions, often other individuals at abnormally high rates, is linked to late-paying and non-paying behavior, given that both the loan and the institution are low quality, in terms of the rates and coaching provided, and this has a negative impact on the entrepreneur's finances.

Other features that matched the hypotheses included civil status and residential zone. The compositional results support the thesis that entrepreneurs in a relationship are less late on payments. Nevertheless, there is no drastic difference between them and single people. On the other end of the spectrum, separated individuals, mainly women, did have payment issues, getting categorized as the most likely to default; this would back up the need for a second income to ensure stability in the face of contingencies. Looking at residence zone, as claimed in the hypothesis, geographic proximity to the institution is positively related to good creditworthiness. The econometric study made a notable contribution to this hypothesis, corroborating the role of distance in the likelihood of non-payment. The conclusions speak to the behavior of the advisors, as they are more able to visit and stay in touch, and provide coaching, to those living closer by.

In terms of housing status, the results demonstrate that renting borrowers are the latest in their repayments, as they exceed in proportion of defaulters and late payers the rest of the housing options, also corroborated by the econometric analysis. One auxiliary outcome of that is that borrowers living in a “borrowed” house behave better. Another variable that proved the hypothesis posed is related to number of people in the household, because beginning with six members or more, there is a trend toward defaulting. In contrast to that, entrepreneurs living alone were the timeliest payers.

Of total household income, conclusions can be drawn pertaining to the hypothesis: the results once again support the arguments about the integral role of microfinance in diminishing poverty, as to the extent that micro-entrepreneurs consolidate their credit experience and gain financial know-how, their standard of living improves, and they become more efficient and reliable for future loans at the institution. Likewise, what can be deduced from the behavior of first-time borrowers is that in light of the lack of formal guarantees, and therefore inability to access regular loans, many low-income borrowers behave impeccably, harnessing the opportunity available to them.

Finally, when it comes to the features of the loans, the hypotheses posited were partially true: in the case of the last loan amount, borrowers taking out their first loan effectively did display a trend to default, with respect to those who already had experience at the institution. There is also radically opposite behavior in many of the first-time borrowers, moving with household income. When it comes to current credit term, even though borrowers with higher installments on their first loans were the most behind on payments, longer terms for installments are very effective to the extent the loan amount is higher. Because the amount to pay is divided over more months, it is easier to pay back because each monthly amount is lower.

In short, the “timely payers” (EP and GP) are the Paraguayan and Peruvian borrowers, older borrowers, those living in a couple, those living close to the institution, those living alone or with fewer members of the household, those living in borrowed homes, those with complete elementary school education, those with higher income, those who have had some loan with another recognized institution, beginning their experience with small loans and short-term installments, to later expand both to the extent their relationship with the institution solidifies. On the other hand, the “very late payers” (LP and DR) tend to be the Argentinean and Bolivian borrowers, with higher levels of education, separated women, living in rented homes or far away from the institution, with lots of members of the household, and low income. They are more frequently first-time borrowers, so following up with them is essential to reducing the financial costs of payment non-compliance.

The hope is that the information and methodology introduced here, about Asociación Civil Avanzar, will fuel the discussion around the moral risk credit advisors face, in an attempt to identify drivers of late payment or default for the microfinance world. This local case is not free from limitations, which could be resolved in large measure with a more exhaustive survey system that would deal with qualitative aspects, which were not taken into account here as they were not available. One such need is to obtain more specific data on entrepreneurship, like the type, longevity, and number of active employees. Even so, it does fulfill the ideal of providing a methodological reference framework for other MFIs to stack up against and use their services. It is vital to recall that although the profiles that emerged as the “timely payers” are not expected to constitute an ideal “package” for the objective borrower, they are a good benchmark for making official loan decisions. Accordingly, even though lower-income borrowers or renter borrowers are more likely to make late payments, it is up to MFIs to find tools to guarantee financial stability and continuity for the most vulnerable borrowers' entrepreneurial endeavors. It is worth acknowledging that the progressive credit strategy and offering workshops and courses are instruments that pave the way to cementing financial experience, strengthening trust in the institution, and developing entrepreneurial capabilities, all key aspects to tackling the vulnerability underlying poverty.

This sort of project leads to reflection on the importance of implementing public policies that would strengthen the work MFIs do. To start, it will be necessary to set up a mechanism to carry out strict auditing of the mission and efficacy of microfinance institutions that need support. From there, the idea would be to encourage, among other measures, coordinated actions between the institutions and public programs, fund provision, networking, public sharing of activities, and the enactment of prudent norms appropriate for institutions of this sort.