Assessing Proportional Efficacy of Microfinance Institutions As a Function of Consumer Protections

“The poor themselves can create a poverty-free world. All we have to do is to free them from the chains that we have put around them.” ― Muhammad Yunus

Microloans and microfinance were initially posited as the final solution for the woes plaguing many of the world’s impoverished areas. Being somewhat of a cutout of international aid or charity, these new financial avenues allowed for individual entrepreneurs and businesses to aggregate capital and invest in ways that were thought to be improbable due to the absence or inaccessibility of traditional finance. Microloans allowed for smaller, low-capital borrowers— those who were historically cast out by traditional finance– to get loans for the starting or maintenance of some enterprise, oftentimes in smaller denominations that would have a sizable impact in their locality due to massive differences in purchasing power. Microfinance exists as a the in-between of the monolithic machine of finance and the trappings of geo-political charity in the form of loans, taking more of a grassroots, ground-up approach over the massive influx of top-down capital utilized by institutions like the IMF or WorldBank. While such IMF loans can have a massive impact on the macroeconomic wellbeing of a state, as it would be alleged, these loans oftentimes strap an anchor to the ankle of the recipient– something that was loaned under inaccurate or predatory assumptions; in some cases, something that cannot be paid back.

Take, for instance, a recent IMF loan paid to Ecuador. As reported by The Guardian, Ecuador signed an agreement to borrow $4.2bn from the IMF over three years, provided that the government would adhere to a certain economic program spelled out in the arrangement. Then-IMF chief Christine LaGarde was quoted as saying the mutually-agreed upon program was “a comprehensive reform program aimed at modernizing the economy and paving the way for strong, sustained, and equitable growth”. However, certain stipulations of the program called for a massive reduction in Ecuador’s national budget (up to 6%) over the course of three years. This retraction of the economy will result in massive cutbacks to public sector programs and investment. In the same Piece, Guardian writer Mark Weisbrot notes that “the overall impact of this large fiscal tightening will be to push the economy into recession... Unemployment will rise – even the IMF program projections acknowledge that – and so will poverty. One reason that it will likely turn out much worse than the IMF projects is that the program relies on assumptions that are not believable. For example, the IMF projects that there will be a net foreign private sector inflow into the economy of $5.4bn (about 5% of GDP) for 2019–2022. But if we look at the last three years, there was an outflow of $16.5bn (17% of GDP). What would make foreign investors suddenly so much more excited about bringing their money to Ecuador?

Certainly not the recession that even the IMF is projecting.” To build legitimate, strong enterprises that are absent of the geopolitical trappings of institutional investment requires a more touch-and-go approach of small, low risk influxes of capital directly towards those in need and towards the businesses that would benefit the most. Microfinance, while not the end-all-be-all we may have hoped for, is one of the ways of building the capabilities capacities of the global poor who are largely cast away by commercial finance and other banking institutions, and elevating them to sustainable, perpetual self-employment by providing previously-inaccessible financial services like credit, savings, and equity. A bit of metacommentary–my initial goal was to assess microfinance institution performance (represented as Return on Assets) the relationship between performance versus a number of variables outlined by Sen’s thesis on poverty (elements such as access to opportunity, voice in institutions, etc). The exogenous variables did not pan out due to a lack of concrete data, and the determining factors were eventually changed to the UN Sustainable Development Goals; variables reflecting an microfinance institutions focus on funds that dealt with topics such as access to clean water, eradicating poverty, health access, and the like

I felt that using exogenous factors that reflected some sort of altruism or sense of global responsibility allowed me to pursue a “balance sheet” based approach to microfinance assessment, ideally marrying the two fields in a way that could prove some relationship between bettering the world and expanding shareholder returns. However, in my research I encountered an article that completely derailed this line of thinking. It introduced a paradigm shift against the claim that microfinance institutions as “the holy grail of poverty alleviation” is a gross misrepresentation. A recent piece in Bloomberg titled “Big Money Backs Tiny Loans That Lead to Debt, Despair and Even Suicide” details the trapping that many borrowers find themselves in after dealing with many of these microfinance institutions; being taken advantage of by the same institutions that set out to alleviate the grinding poverty of their day to day lives. The authors Finch and Kocieniewski wrote how “as financiers have replaced philanthropists in the microfinance industry, consumer protection has been weakened. Taxpayer-funded development banks, which could fix the problem, are instead channeling hundreds of millions of dollars earmarked for poverty alleviation into some of the most predatory lenders.” This reflects a balance-sheet based approach towards poverty alleviation. The piece continues as follows, “The expansion of the loosely regulated business—all those loans added up to $160 billion in 2020, about twice the amount 11 years earlier, according to industry estimates—has resulted in millions of new borrowers for whom the dream of inclusion has turned into a nightmare of debt— ‘the development banks promoted this personal profit-making version and brought in business investments along with their own investment,” says Muhammad Yunus, the Nobel Peace Prize-winning economist who founded Bangladesh’s Grameen Bank. “The concept of microcredit was abused by some and turned into profit-making enterprises for owners of microcredit institutions. Many went in the inevitable direction of loan-sharking. I felt terrible that microcredit took this wrong turn.’” The piece went on to describe a wave of suicides in India that was attributed to mountains of debt that had been accrued by borrowers seeking to better their lives, and the crushing debt from borrowers that had propped up in microfinance markets all around the globe.

While I was never naive enough to think microfinance institutions were the end-all-be-all, the piece really took me by surprise. How could bad actors be allowed to enter a space many believed was a novel application of altruism? How is it, with all the considerations taken, that borrowers still ended up with mountains of debt, many of whom are strapped to predatory and insurmountable rates of interest? This forced me to change my tune entirely. Being just another purveyor of the “balance sheet approach” to poverty would simply result in turning in a piece that towed the same line that allowed for the microfinance sector to become bastardized, reverted back to it’s predatory roots while absolving it’s malfeasance in the name of “returns on assets”. My research question had evolved from assessing microfinance institution performance in terms of some set of development-centric variables, into measuring microfinance institution efficacy (i.e., how many loans are given out, or how effective at loan distribution is a given institution) based upon consumer protections on a per-MFI basis.

While this new topic is a divergence, is that not what’s most important?Is the purpose of microfinance to line the pockets of shareholders and to increase dividends? And if so, on the backs of whom? Obviously it’s integral for any found’s survival to maintain (atleast) consistent returns, but what does it say about an microfinance institution it’s exhibiting stronger than average returns? Who is actually benefiting? I chose to recalibrate around Gross Loan to Total Asset (GLtTA) ratio as a measure of efficacy, which represents how much an microfinance institution allocates to lending (which in this case, is its most profitable activity— making loans). This, hopefully, represented some change in how not only poverty, but how poverty alleviation is viewed– any microfinance efforts need to protect the consumers more than the institutions, as there is only one party in the transaction with capital, only one who can write off debt or losses, and conversely there’s only one party in the transaction who likely needs this to survive. To be candid, the risk-averse nature of finance in the western world is what keeps out smaller, more poverty-stricken borrowers. If a borrower has no capital, and little income, how would they be able to pay back even a small sum under the standards, interests, and expectations lauded in bank loans? Microfinance has demonstrated that the global poor can handily be viable consumers of financial services, as long as financing is approached in the correct manner, which requires that predatory practices, profit seeking, adverse selection and other problems are mitigated. While no one solution exists that can eradicate poverty at it’s core, the recognition of that fact means that robust, focused efforts on multiple fronts puts forth the best effort in alleviating the negative externalities of poverty, and thus changing the reality for many who are crushed under it’s weight year after year. From a finance perspective, this means looking beyond the traditional avenues of the IMF, the WorldBank, and UN initiatives– it necessitates the changing of one’s perspective on what it means to be poor in today’s world.

Looking at poverty this way, largely absent of traditional metrics, opens up different possibilities for analysis. Using a more nuanced perspective of both poverty alleviation and microfinance institution efficacy may help address the overarching problems that have not been solved using a more traditional, economic approach. This analysis posits the overall question– how much of a microfinance institution’s efficacy (or proportional loan output) be dictated by the existence of fair practices? Do microfinance institutions that display (whether intentional or unintentional) the key elements of fair lending practices loan proportionally more than those who do not?Is there a correlation of these more esoteric elements of consumer advocacy and fairness and microfinance institution portfolio performance? How much of the variation of gross loan portfolio to total assets across a sample of microfinance institutions can be attributed to the presence of viable consumer protections in the sample?

Amartya Sen identifies two distinct “freedoms” that can be achieved by an individualconstitutive freedom and instrumental freedom. Constitutive freedom is seen as the logical end for development, and this constitutive freedom outlines factors that would increase basic necessities of life, such as freedom from starvation and premature death, freedom of speech and expression, and also an opportunity for political participation. Conversely, instrumental freedom is described as the “principal means” of development, where distinct instrumental freedoms operate often in tandem, and where the development and fostering of freedom in one area can interrelate with others. Microfinance and microlending fall under the latter category. It is a means to further develop the lives of those living in poverty by providing greater financial security and the possibility for scaling investment. The benefits of microfinance are not constrained to merely economic development, but also tie into other aspects of life, such as encouraging greater social inclusion and fostering other networks of lenders and borrowers.

For years, microfinance overlapped with microcredit, often without traditional safety nets or guarantees, aimed at improving the lives of business owners, individual, and their families by sustaining small-scale economic activity. Having the poor and impoverished as clients in a world run by traditional finance was viewed as a non-starter, with private sector financial institutions failing to step up to the plate. Nongovernment organizations played a vital part in developing viable financial products for the poor, exemplified by Bangladeshi NGOs BRAC6 and Grameen Bank which introduced the community lending “microfinance model”. By 2010, private sector microfinance institutions reported about 26.7 million clients, particularly women and the poor. While private sector initiatives avoided the pitfalls of corruption and inefficiency, concerns about unregulated lending by microfinance institutions grew. Thought leaders in the space warned that a “balance sheet approach” and a for-profit initiative superseded any idea of altruism, and was incentivizing lenders to overload the poor with loans, something that still burdens many borrowers to this day.

To embellish on the topic of altruism some more, many have asked “why not just give the money away?” This forces us to draw a distinction between what constitutes as charity and what constitutes as business, and microfinance can sometimes fall awkwardly in the middle. In some countries, pure charity (in the form of international aid) is often misaligned and misappropriated to those in power, often with little making it’s way to those who need it the most. In the paper “Universal Basic Incomes versus Targeted Transfers: Anti-Poverty Programs in Developing Countries”, authors Hanna and Olken outline how some countries have “implemented transfer programs that seek to target beneficiaries: that is, to identify who is poor and then to restrict transfers to those individuals.”

This targeted transfer is different from the wide net of UBI, which aims to raise a general level of equity in a given country or jurisdiction. While UBI can, in some kind of feedback effect, benefit the tax base of an area, it ultimately does not address the core problem of eradicating poverty.

The authors came to the conclusion that, while imperfect in it’s own right, targeted transfers had a more substantial effect on eradicating poverty in the areas studied, “evidence from Indonesia and Peru shows that existing targeting methods in developing countries, while imperfect, appear to deliver substantial improvements in welfare compared to universal programs, because they can transfer much more on a per-beneficiary basis to the poor as compared with universal programs. The primary downside of these programs is horizontal equity—because targeting is imperfect, there will be a substantial number of poor households who slip through cracks and are excluded. Nevertheless, for many developing countries, our simulations suggest the welfare gains from targeting may be substantial.” Ultimately microfinance is independent of these findings, but it revolves around the same principle– find those in need, and cater financial vehicles to their needs. In some instances, this can be like a targeted transfer (which has been compared to a type of reimbursement), in other instances, this can manifest as business or operational loans, which moves beyond lining pockets and can help impoverished individuals build something more valuable than cold hard cash– credit.

Specifically, the focus of this paper is what determines microfinance efficacy and how proportional loan output can, in turn, be determined by proper consumer protections. This is a departure from my initial thesis– albeit both revolve around the question of “what non-financial variables (outside of traditional metrics) can be a strong determining factor in microfinance institution performance?” While there are tried-and-true financial metrics that could presumably be tied to portfolio performance (as they can determine performance in a typical finance setting), I’m more interested in qualitative variables that may have some sort of relationship. For both a comparison of methods and for an analytical proving ground, I’m utilizing Determinants of Financial Performance of Microfinance Banksin Kenya (Ngumo, Collins et.al).

This paper sets out to examine the determinants of financial performance of a sample of microfinance institutions in Kenya. The study adopted a descriptive research design and used secondary data from 7 Microfinance banks for a period of 5 years from 2011 to 2015. The data collected was analyzed using correlation and regression analysis. The study found a positive and statistically significant relationship between operational efficiency, capital adequacy, firm size and financial performance of microfinance banks in Kenya. The variables used in this study are:

● Return on Assets used as a proxy for financial performance

● Operation Efficiency Ratio used as a proxy for operating efficiency

● Capital Adequacy proxied using the proportion of microfinance institution equity to total assets

● Loan to Asset Ratio used as a proxy for liquidity position

● Credit risk proxied using the non-performing loans ratio (NPLR)

● Firm size proxied using natural log total assets

For a comparison, I found similar proxy variables in my dataset (with the operative gross loan portfolio to total assets being included), and set out to recreate the analysis with a larger dataset and slightly different proxy variables. With the details surrounding the analysis below, the paper reaches the following conclusion: The findings of the study established that operational efficiency, capital adequacy and firm size significantly and positively influencesthe financial performance of microfinance banksin Kenya. The study therefore concludesthat there is[a] direct relationship between operational efficiency, capital adequacy, firm size and financial performance of microfinance banksin Kenya. The study further found that liquidity risk and credit risk do not have [a statistically]significant relationship with financial performance of microfinance banks of Kenya.

While the analysis and findings of the paper are consistent, I can’t help but wonder if there’s something more to the story. Given that the number of observations was limited to 35, I knew that utilizing the WorldBank’s MIX dataset on microfinance could add some serious substance to this analysis. I believe that a combination of a small sample size and hand-picked variables plays into the favor of the findings of the paper; while the authors’ conclusion does not stretch beyond Kenya, I am curious to find what variables have the strongest impact independent of location– for this, I sourced data from a number of microfinance institutions throughout the world for the timeframe of 2010-2019. I’m building upon this analysis because I believe the sample size is not adequate enough to capture enough detail regarding microfinance institution performance, and I will be using different proxy values than the ones provided since I feel that there is better data available. Mainly, this difference is with the utilization of Operational Self Sufficiency, Portfolio At Risk over 1 Month as a proxy for credit risk, and Gross Loan Portfolio to Total Assets as a proxy for Capital Adequacy. I believe these variables offer a more accurate representation of financial performance as opposed to some of the composite proxies that are used in the Kenya paper.

From a population of over 41,000 records, I sampled approx. 6,000 annual records for the below variables for the timeframe of 2010 - 2019 for 1,656 distinct microfinance institutions, with the average across the given timeframe calculated for each variable. The variables I’m utilizing in my comparison are:

● Return on Assets used as a proxy for financial performance

● Operational self sufficiency

● Capital Adequacy proxied using Gross loan portfolio to total assets

● Capital to Asset Ratio used as a proxy for liquidity position

● Credit risk proxied using Portfolio at risk > 30 days

● (Log) Net Fixed Assets as a Firm Size Proxy

While this comparison will be embellished upon in the final paper, I am also going to add a secondary endogenous variable to see if there’s any further relationship that can be extrapolated between. This plays into the hypothesis of the paper– what else is out there that plays a part in microfinance institution performance? Where does the relationship exist, if at all? One avenue I’ve considered exploring is the use of total Value of transactions as a second Firm Size Proxy, as well as amount of personnel or amount of outgoing loans.

Appendix I represents my replication of the Comparative Statistics table. The data I utilized has a way stronger capital adequacy ratio than the Kenya data. With that, the standard deviation for net fixed assets is wildly beyond the value presented in the initial paper, with the log value for that same variable being more consistent with previous findings.

Appendix II represents my replication of the Correlation Analysis table. With my analysis, the yellow highlighted cells are ones with p-values below the standard threshold of significance (0.05), and thus can be claimed to have a “statistically significant” relationship to one another. The negative values are seen to have a “negative” relationship on the column variable. The variables I found as being statistically significant are drastically different from the ones found in the Kenya paper; I believe this is due in part to the more wide-ranging microfinance institutions that were included (which also means a wider range of values within each variable), and that there could also be a difference in how each of our datasets calculated Return on Assets (the key dependent variable).

Lastly, Appendix III displays the comparison of both summaries of the regression results; from the initial paper: The results on table 4.3 indicate that the R-square value is 0.522, which indicates that the independent variables (operational efficiency, capital adequacy, liquidity risk, credit risk and firm size) explain 52.2% of the variation in the dependent variable (financial performance). Additionally, the results indicate that the F statistics value is 6.347 with a p value of 0.000 which is less that the significance value of 0.05 (0.000<0.05) hence an indication that the regression model is significant and a good predictor of the relationship between the independent variables and the dependent variable. After some trial and error I found that the variables Credit Risk and Net Assets/Log Net Assets, had no statistically significant relationship with ROA, the independent variable for this dataset. The results on the above table indicate that the R-square value is 0.3177, which indicates that the independent variables (operational self-sufficiency, gross loan portfolio, and capital to asset ratio) explain ~32% of the variation in the dependent variable (ROA; financial performance).

In conclusion, there was a positive yet weak correlation between return on assets (a proxy for financial performance) and operational self-sufficiency, gross loan portfolio, and capital to asset ratio for each microfinance institution. While this is a weaker relationship than was identified in Kenya with similar/same proxy variables, I am inclined to believe there are other, perhaps non-financial variables that have a stronger relationship than even the 0.522 r-value that was determined in the paper (which, is not too strong of a correlation to begin with). My initial hypothesis that there may be some other variable at play here may be correct; with a larger sample size and better proxy-values for each variable, the relationship between financial performance and the operational self-sufficiency, gross loan portfolio, and capital to asset ratio was weaker than the paper had claimed (including all independent variables, the R-Value was 21.3%). While this does not answer the initial question, I believe this passes the sniff test enough to begin utilizing other variables in the MIX dataset to find what is the strongest determining factor to financial performance for microfinance institutions.

The comparison across the two analyses outlines that there is a distinction between summarizing microfinance institution performance in a particular region with a small sample and trying to make more broad assumptions across a number of different microfinance institutions (in different areas). Understanding that each microfinance institution has it’s own operations, considerations, areas of focus, and lending capabilities, this calcified my hypothesis on the basis of searching for a larger trend that may transcend locality or other factors. With that said, this analysis has a ceteris paribus assumption– outside of the endogenous and exogenous variables being analyzed, all other factors are held constant (more on this in the conclusion section).

As mentioned before, the data collected for this analysis largely came from the WorldBank Microfinance Information Exchange (MIX) Market dataset. The MIX data is robust and comprehensive in nature, as it can be used to compare and analyze the performance of microfinance institutions in over 100 developing markets. While this data has been collected since 2002, it only recently became available for every-day analysis. The MIX Market data has been reported by different financial services providers focusing on the financially-underserved in global developing markets.

Data points largely come from financial statements (income statements, balance sheets, tax documents), as well as variables relating to microfinance institution operations, financial products, end clients, and social performance. Being a product of the WorldBank, the MIX dataset is in line with broadly recognized reporting standards within microfinance and inclusive finance. The MIX dataset also provides a number of social performance metrics for each microfinance institution, particularly around development goals, business practices, governance, and client protections. Gross loan portfolio to total assets (GLPtTA) will be used in lieu of Return on Assets to assess microfinance institutions– while Return on Assets indicates how profitable an microfinance institution may be, Gross loan portfolio to total assets is a better gauge of the efficacy of an microfinance institution; or, how effective an microfinance institution is at loaning out as much as it can while remaining a solvent entity. GLPtTA measures how much an microfinance institution allocates to its primary business (in this case, lending) which in turn represents an microfinance institution’s most profitable business. Low GLPtTA may indicate inefficient use of assets, and high GLPtTA may indicate insufficient liquidity levels.

Consumer protections exist as the last line of defense preventing borrowers from being fleeced by predatory institutions (not that such protections are always effective). These protections can take many forms– in traditional finance, think of aspects such as credit scores and banking insurance to gauge the borrowing capabilities and to protect external losses of prospective recipients. These aspects are nullified once exiting the world of big banking– credit scores do not mean much in an area where no one has credit to begin with. Running the gamut of available consumer protections within microfinance institutions, while also choosing the most viable elements of analysis, has resulted in the following variables:

● Presence of internal audits to verify over-indebtedness prevention

● Presence of full disclosure of prices, terms, and conditions

● Presence of full disclosure of cost information

● Presence of clear debt collection practices

● Presence of clear sanctions for violations of debt collection practices

● Presence of functioning client complaint mechanism

● Presence of privacy data clause in loan contracts

● Presence of client protection assessment

The data at hand has one unique aspect that makes the analysis tricky– all of the above variables are boolean variables, in that, a 0 confirms the MIF does not have the qualification express by the variable, and 1 confirms that the microfinance institution does. This makes things like collecting summary statistics difficult, as it would distort the meaning. With that said, to get a handle on the dependent variable (gross loan portfolio to total asset ratio), the following represents a table of it’s summary:

This analysis utilized a descriptive research design, similar to the study done for Kenyan microfinance institutions that was referenced earlier. This distinction is made because the purpose of the analysis is to reach beyond that of a single region or state, but to make more broad conclusions on the microfinance sector as a whole. The population of the study was made of 883 microfinance banks, with financial records from 2009 - 2019 from the MIX financial performance dataset. One quick parameter– notice how the sample size decreased compared to the Kenyan comparison analysis. This is because, for this analysis, I chose to remove all duplicate microfinance institution data values from the sample size– meaning, if I had data for XYZ Bank for 2011, 2012, and 2015, I deduped the data to only include the 2015 data from XYZ. Since the core thesis of the comparison analysis was to have a larger sample size, and the hypothesis for this analysis revolves around finding if there is some kind of relationship, I chose to limit the dataset to one set of variables per microfinance institution as to find the most crystal clear connection between microfinance institution efficacy (without bogging down the analysis with what could be redundant numbers) and consumer protections. Also, the duplicate data per microfinance institution across several years was utilized in the Kenyan study as well, keeping with the parameters of the initial analysis.

This analysis used secondary data from the MIX social performance dataset to bring in associated exogenous variables (i.e., the aforementioned variables on consumer protections). Descriptive statistics (see above) summarized the Gross Loan to Total Assets ratio using the mean, variance etc., while correlation and regression analysis were used to determine any sort of relationship between both sets of variables. Before continuing, a quick refresher on the dependent variable, Gross Loan to Total Asset ratio. GLtTA measures the total principal amount of all loan facilities extended by the microfinance institution to its borrowers in the ordinary course of business outstanding as a percentage of total assets. The higher this ratio indicates an microfinance institution is loaned up and its liquidity is low. The higher the ratio, the more risky a microfinance institution may be to higher defaults. The ratio can be expressed as follows:

● GLtTA = P1 + P2 + …Pn / A1 + A2 + …An

○ n - Number of loan facilities

○ P - Principle

○ A - Total Assets

Using this ratio as a proxy value for microfinance institution performance, the estimating equation is as follows:

● GLtTA = β0 + β1 + β2 + β3 + β4 + β5 + ɛ

○ β1 - internal audits to verify over-indebtedness prevention

○ β2 - full disclosure of prices, terms, and conditions

○ β3 - full disclosure of cost information

○ β4 - clear debt collection practices

○ β5 - clear sanctions for violations of debt collection practices

○ β6 - functioning client complaint mechanism

○ β7 - privacy data clause in loan contracts

○ β8 - client protection assessment

To analyze the correlation between the sets of variables, I utilized the Pearson correlation to measure the strength of the linear relationship between two variables; in this instance, it would be measuring the strength of the linear relationship between Gross Loan to Total Assets and each of the individual beta variables (represented above). For the regression, I utilized a linear regression model since it would help predict the value of GTtLA based on the value of an exogenous variable, as well as give a statistical estimate of the strength of that relationship. The p-value for this study sits steadfast at 0.05; meaning there is only a 5% chance that results from the sample occurred due to chance. Again, a reminder that all other factors in this analysis are held constant.

The results of the analysis were surprising to say the least, and there are some key takeaways to address. Below outlines the P-value for the Pearson correlation test for each of the exogenous variables versus GLtTA (for all, the alternative hypothesis is the true correlation is not equal to 0):

● Full disclosure of prices, terms, and conditions

○ t = -0.33598, df = 881, p-value = 0.737

○ 95 percent confidence interval: -0.07723552, 0.05469645

○ sample estimates: cor = -0.01131879

● Internal audits verify over-indebtedness prevention

○ t = 0.70268, df = 881, p-value = 0.4824

○ 95 percent confidence interval: -0.04237315, 0.08950206

○ sample estimates: cor = 0.02366741

● Clear debt collection practices

○ t = 0.29278, df = 881, p-value = 0.7698

○ 95 percent confidence interval: -0.05614746, 0.07578856

○ sample estimates: cor= 0.009863476

● Full disclosure of cost information

○ t = 3.034, df = 881, p-value = 0.002485

○ 95 percent confidence interval: 0.03595361, 0.16654392

○ sample estimates: cor = 0.1016868

● Functioning client complaint mechanism

○ t = 0.9077, df = 881, p-value = 0.3643

○ 95 percent confidence interval: -0.03547893, 0.09634712

○ sample estimates: cor = 0.03056702

● Client protection assessment

○ t = 4.0663, df = 881, p-value = 0.00005203

○ 95 percent confidence interval: 0.07038631, 0.19991464

○ sample estimates: cor = 0.1357304

● Privacy data clause in loan contracts

○ t = 0.52682, df = 881, p-value = 0.5984

○ 95 percent confidence interval: -0.04828455, 0.08362288

○ sample estimates: cor = 0.01774638

● Clear sanctions for violations of debt collection practices

○ t = 3.6765, df = 881, p-value = 0.0002508

○ 95 percent confidence interval: 0.05741519, 0.18737876

○ sample estimates: cor = 0.122924

For starters, it is evident that a majority of the consumer protections variables had little to no impact on the efficacy (proportional loan output) for a given microfinance institution. For the boolean variables privacy data clause in loan contracts, functioning client complaint mechanism, full disclosure of cost information, clear debt collection practices, internal audits verify over-indebtedness prevention, and full disclosure of prices, terms, and conditions— there is no statistically significant relationship for our dependent variable. The average microfinance institution in this dataset can have these protections, or not, and it would ultimately not affect the average microfinance institution's gross loan ratio. However, two variables have p-values far below the threshold of significance (0.05), and that is clear sanctions for violations of debt collection practices and client protection assessment.

Both of these consumer protections have p-values that indicate there is indeed a statistically significant relationship between them and GLtTA. Another thing to note is that both clear sanctions for violations of debt collection practices and client protection assessment had a correlation coefficient greater than 0 (0.122924 and 0.1357304, respectively), meaning that as their values increased, the value for GLtTA would increase as well. This figure does not represent what one would call a strong correlation, but it’s a correlation nonetheless, and since our variables are boolean, what it’s indicating is that for each variable, as each microfinance institution “gets closer” to 1, the GLtTA increases as well. From this, we can deduce that the implementation of these specific consumer protections has a positive relationship with gross loan to total asset ratio, and the implementation of these practices may in turn increase microfinance institution efficacy. The second statistical analysis that was performed on this data was a linear regression. This will help determine if we can estimate the value of a given variable based on the value of another. Knowing what we know from the pearson correlation, we can estimate that both clear sanctions for violations of debt collection practices and client protection assessment may have a statistically significant relationship in this regression as well.

Using the same dataset as before, and holding all else constant, below is the result of the regression:

● Residuals: ○ Min = -0.7800 ○ 1Q = -0.1011 ○ Median = 0.0153 ○ 3Q = 0.0985 ○ Max = 5.6345

● Coefficients:

○ `Full disclosure of prices, terms, and conditions`

■ Est = -0.02290 Std Error =0.03508 T-Value = -0.653 P-Value = 0.51398

○ `Internal audits verify over-indebtedness prevention`

■ Est = -0.01033 Std Error =0.02305 T-Value = -0.448 P-Value = 0.65414

○ `Clear debt collection practices`

■ Est = -0.01864 Std Error =0.02671 T-Value =-0.698 P-Value = 0.48541

○ `Full disclosure of cost information `

■ Est = -0.06101 Std Error = 0.06207 T-Value = -0.983 P-Value = 0.32596

○ `Functioning client complaint mechanism`

■ Est = 0.01283 Std Error = 0.02342 T-Value = 0.548 P-Value = 0.58402

○ `Clear sanctions for violations of debt collection practices`

■ Est = 0.14841 Std Error = 0.06344 T-Value = 2.340 P-Value = 0.01953

○ `Privacy data clause in loan contracts`

■ Est = 0.01949 Std Error = 0.02655 T-Value = 0.734 P-Value = 0.46307

○ `Client protection assessment `

■ Est = 0.47558 Std Error = 0.12776 T-Value = 3.722 P-Value = 0.00021

For this linear model, there are a number of associated diagnostic plots that may add some insight. Appendix IV represents the model’s Residuals vs. Fitted plot– this shows if residuals have non-linear patterns. If there are evenly spread residuals around some horizon without distinct patterns, that can indicate that there are no non-linear relationships. From the look of the plot, there are a number of clusters and patterns that emerge, suggesting that there is indeed a number of non-linear relationships represented in the data. Appendix V shows a Normal Q-Q plot, which shows if residuals are normally distributed. Given that there’s no severe deviations (and actually a pretty consistent line for ~80% of the chart), it can be deduced that the model’s residuals are more normally distributed. Appendix VI represents a Spread-Location plot, which is how to check the assumption of homoscedasticity. Our Spread-Location plot, similar to Appendix IV, shows a lot of clusters, vertical lines, and wide-spread deviations, meaning the MFI data’s residuals are not spread evenly throughout the range of predictors. Lastly, Appendix VII shows a Residuals vs. Leverage plot, which can help detail “influential” data points.

For this plot, pattern is not so important as is the element called “Cook’s Distance”. When data points are outside of the Cook’s distance (meaning they have high Cook’s distance scores), they are viewed to be influential to the regression. Notice which points in Appendix VII are beyond the Cook’s distance marker. Moving on, the above results from the model are aligned with what was previously thought– the two variables with p-values that indicate a statistically significant relationship are, yet again, clear sanctions for violations of debt collection practices and client protection assessment. To interpret this a bit more, let’s take a look at each of their t-values (we can disregard all the other consumer protection variables since there’s a high likelihood that any relationship between them and GLtTA is due to chance). The t-value (the coefficient divided by the standard error) helps determine how linear, or nonlinear, a given relationship is. Both variables have a t-value of greater than 2; any t-value greater than 2 (or less than – 2) is deemed to be acceptable. The higher the t-value, the greater the confidence in the coefficient as a predictor.

The only caveat is the R-squared value from this analysis, which sits at 0.03257 for multiple r-squared and 0.02371 for adjusted r-squared. R-squared gives us an idea of how well the model explains the data in the sample we used. For example, in this analysis, an r-squared of 3.25% reveals that 3.25% of the variability observed in the target variable is explained by the regression model. This is not very good, so let’s re-calibrate our model to only include the variables we know have a statistically significant relationship with GLtTA. Calling the following formula:

lm(formula = `Grossloan portfolio to total assets` ~ `Clearsanctionsfor violations of debt collection practices > Yes` + `Client protection assessment > Yes`, data = mix_df2)

We can actually see that the R-squared value went down to 2.92%. This may be a fact of life that has to be baked into our findings— while both variables have a statistically significant relationship, only about 2-3% of the variability observed in GLtTA is explained by the regression model, even after limiting it to the variables that had the best chance for success.

This study sought to establish factors of financial efficacy and output, along variables determining consumer/borrower protections, for Microfinance banks in underserved markets. What we've been able to find is that there are at least two consumer protections that a given microfinance institution can have in place that may have some sort of positive relationship with the overall efficacy, as GLtTA, of the microfinance institution. The findings of the study established that an institution having clear sanctions for violations of debt collection practices, as well as some sort of client protection assessment in place, positively influences the financial efficacy (using Gross Loans to Total Assets as a proxy value) of microfinance banks independent of location. The study therefore concludes that there is a direct relationship between the aforementioned variables and financial efficacy of microfinance banks. The study further found that privacy data clause in loan contracts, functioning client complaint mechanism, full disclosure of cost information, clear debt collection practices, internal audits verify over-indebtedness prevention, and full disclosure of prices, terms, and conditions do not have a statistically significant relationship with financial efficacy of microfinance banks in the terms previously defined. This leads to the conclusion that the existence of the aforementioned consumer protections do not affect the financial efficacy of microfinance banks.

Now let’s unpack these variables just a bit. To embellish a bit more, what clear sanctions for violations of debt collection practices means is that for a given microfinance institution, there are clear rules and bylaws in place that would punish the institution for violating their predetermined debt collection practices (the definition of such practices was it’s own variables, albeit one that had no verifiable relationship). Based on t-values from the linear regression, the presence of a client protection assessment for a given microfinance institution actually had more of a determined relationship than any other variable. What this variable means is that for a given microfinance institution there is some sort of ranking or assessment on how well an MFI shields their clients from risk, or conversely, how “ethical” they may be.

Obviously, there are a number of factors that ultimately affect microfinance institution performance and viability outside of the ones that were assessed here, and even if there is a relationship between performance/efficacy and some number of consumer protections, there may be externalities at play that our analysis simply won’t pick up on. For client protection assessments, this could mean the following:

● Only the most capable or properly scaled MFIs use client protection assessments

● CPAs are publicly available, and MFIs that are looking for customers use them as a means of advertising or market differentiation.

● CPAs are only used in tandem with other best practices, correlating with a more viable institution (absent of the assessment)

None of these factors could be deduced from the analysis that was performed, nor do they relate at all to any sort of “altruism” outside of a drive to keep an MFI competitive in the marketplace. The same can be said for the other correlated variable, clear sanctions for violations of debt collection practices, as some other externalities could be at play:

● Clear sanctions for violating debt collection practices necessitates that there needs to be debt collection practices in the first place, and the sanctions (no matter how clear) do not speak to the ethics of the debt collection practices themselves.

● Having clear sanctions in place does not necessitate that they will be implemented upon any sort of debt collection, just that they are clear and published.

Again, these are aspects that are unique to a given institution and would not be deduced from a linear regression. Ultimately, what we’ve found here is that there is some relationship between certain consumer protections and the overall efficacy of an average MFI.

The relationship may not be the strongest, but it’s positive, indicating that the presence of these protections can be influential in the success of a given institution. These types of protections, whether they have some sort of correlation or not, are extremely important to the world of microfinance, and they exist as the barrier between true development and predatory lending. Microfinance is not something that can be exploited for money, nor should it be viewed as a scheme similar to traditional finance. It sits in a unique position, helping competent and determined individuals and business people bypass the harsh reality of living under the thumb of global poverty, even if it is a dollar at a time. To view these individuals as simply charity cases (or worse, something to be exploited) is to remove all nuance and humanity from something that looks beyond pure profit motivations and the “balance sheet approach” to solving our world’s problems. Global poverty, as defined by Amartya Sen, is multifaceted in nature and goes far beyond what can be described in a set of summary statistics. In order to even begin to make some kind of impactful change, it takes one to recognize the scope and scale of the problems ahead of us.

Previous
Previous

PyTorch for Aerial Land Recognition Pt 1 "Building a Bad Model"

Next
Next

ETL with Python: Redfin Real Estate Data