Competition and Loan Delinquency
in the Microfinance Industry
Date: January 23, 2009 Author: Rosienne Martis Student number: s1736736
Abstract
This research is an attempt to assess the existence of a relationship between competition and Loan delinquency in the Microfinance Industry. Based on existing literature on the banking industry, three different theories are tested; the risk-shifting theory, suggesting that competition reduces loan delinquency, the franchise value theory suggesting that competition stimulates loan delinquency and the combined relationship of the risk-shifting effect and the margin effect (competition induces loan delinquency).
Introduction
According to the Millennium Development goals, by the year 2015 extreme poverty should be reduced by 50 percent. The Microfinance industry aims at contributing to this goal by offering the very poor loans for productive purposes. In this matter the very poor are able to integrate in and contribute to their nations’ economy by pulling themselves out of poverty (Montgomery & Weiss, 2004).
Whilst practiced much earlier in history, Microfinance started receiving attention as a poverty reduction tool just in the 1970’s (CGAP, 2006).
Microfinance entails giving credit to groups of people or individuals who would otherwise not have access to capital, because they are considered ‘un-bankable’. These groups of people
belong to the very poor, which according to the World Banks’ standards, are people living at less than $1 a day (World Bank, 2008). They are considered un-bankable, because they do not have salaries or collateral to back up a loan at commercial banks. In the form of loans, credit is given to these people (mostly women) who use it to develop income-generating activities (CGAP, 2006; Byström, 2007).
At present times, Microfinance is undergoing several stages of development. Contrary to the past, Microfinance Institutions (MFI’s) are breaking free from their dependence on donors for capital; they are achieving financial self-sufficiency and are offering a broader set of products (Kiviat & Farooq, 2008; Hishigsuren, 2006). The nexus of these changes is commercialization. Commercialization of Microfinance involves the management of MFI’s, following a financial approach to business (Christen & Drake, 2001).
This research focuses on the effect competition has on Loan Delinquency in the Microfinance industry. Because of the fierce competition commercialization brought about, MFI’s find themselves in an eternal fight to gain more clients in order to survive. By doing this, they incur more risk, as they do not always attract the right clients. These marginal clients’
untrustworthiness increases the risk of non-repayment which results in increasing loan delinquency. Consequently, the MFI has to cope with diminishing returns. Ultimately, MFI’s incurring excessive risk will not be able to survive after all, compromising the achievement of the Millennium Development goals.
Academic literature provides several views of competition and bank risk. The franchise value1 model indicates that an increasing level of competition, decreases franchise value which in turn stimulates the bank to engage in riskier activities in order to recover this value (Berger, Klapper & Turk-Ariss, 2008). On the other hand, under the risk–shifting effect, an increasing level of competition reduces bank risk. Specifically, increasing competition lowers loan rates and thus reduces the probability of loan delinquencies (Byod & De Nicoló, 2005). Additionally, Martinez-Miera and Repullo (2008) presented a combined view. They suggest a “U-shaped” relationship between competition and bank risk. They argue that increasing competition reduces loan rates, which in turn lessens the probability of Loan Delinquency indeed. However, after a certain point, an increasing level competition reduces the buffer to cover for loan losses to such a level, that these banks are forced to engage in riskier activities in order to maintain and increase their revenue.
While several studies have been done on the above mentioned relationships in the banking industry, little has been tested empirically in the Microfinance industry. The aim of this study is to fill that gap in literature. In this research an answer is given to the question whether or not the level of competition has an effect on Loan Delinquency in the Microfinance industry.
The answer to that question will help MFI’s to put their priorities straight in terms of how to deal with the competition. If competition is indeed stimulating Loan Delinquency, action should be
taken in order to ensure that the ultimate goal of Microfinance is achieved: progression of the very poor.
The results of this research show that the level of competition has a negative influence on Loan Delinquency. This result supports the franchise value model, the margin effect and the
competition fragility theory. In addition, forced loans, unchanged high interest rates and multiple loans may be considered key factors contributing to this result.
The remaining of the paper is divided in five sections. The first section sheds light on literature published on the relationship between competition and loan risk; both in the banking industry and in the Microfinance industry. In the second section the research question and the hypotheses are formulated. The third section elaborates on the conceptual model, data and research
Theoretical review
As implied in the introduction, this research acknowledges commercialization as the stimulus behind the threshold of competition in the Microfinance industry. Therefore a brief introduction on the concept of commercialization and its link to competition is presented prior to the
discussion of existing theories on the topic of this research.
In the words of Campion (2001), commercialization of Microfinance is defined as “…the
application of market-based principles to providing financial services to the poor”. A similar
definition is provided by Christen & Drake (2001), who describe commercialization of
Microfinance as “…the movement of Microfinance out of the heavily donor-dependent arena of
subsidized operations into one in which Microfinance Institutions manage on a business basis…”.
This commercial approach of MFI’s in operations has become widely recognized as the best means to achieve sustainability (Charitonenko & Rahman, 2002). Commercialized MFI’s have access to capital markets which allow them to break free from their dependence on donors. With more capital MFI’s are able to attain financial self-sufficiency and can therefore provide a broader range of products to an increasing number of very poor clients (Kiviat & Farooq, 2008). A higher outreach is an indication that more of the very poor are pulling themselves out of poverty and integrating to their nation’s economy. Based on this view, commercialization of Microfinance is to be considered a necessity if the Millennium development Goal of poverty reduction is to be achieved on a timely manner.
commercialized and non-profit), more and more commercial banks are entering the microfinance market as well (Lascelles, 2008; Christen & Drake, 2001). Consequently, the Microfinance industry is becoming saturated, creating fierce competition among MFI’s (Kiviat & Farooq, 2008; Christen & Drake, 2001; Lascelles, 2008; Mc.Intosh, Janvry & sadoulet, 2004 among others). MFI’s find themselves struggling to survive as they now have to compete with an increasing number of new entrants targeting common regions.
The consequences of competition in the Microfinance industry are twofold.
On the one hand competition may stimulate lower interest rates, improve customer service, and encourage wider range of products, higher cost efficiency and ultimately higher outreach in the Microfinance industry.
Hishigsuren (2006) and several other scholars (Littlefield & Rosenberg, 2004 and Poyo & Young, 1999 among others) sustain these positive effects of competition and are convinced that a competitive Microfinance market should make the clients (the very poor) better off. The
increasing competition among MFI’s pressures institutions to differentiate themselves from their peers and triggers them to work more efficiently and offer their clients better fitted and cheaper products. Hence, the Microfinance industry is experiencing a transition from “supply- to
demand-driven development” (Poyo &Young, 1999).
On the other hand, there is evidence of negative consequences of competition for the
Microfinance industry as well. In fact, an article published by M-CRIL (2005)2and a case study done in Mali by Pagura, Graham & Meyer (2001) identified competition among factors causing Loan delinquency in the Microfinance industry.
Moral hazard and Adverse selection problems are here central. More specifically, because of the fierce competition in the industry, asymmetric information poses a major risk for MFI’s. Because of the asymmetric information problem, MFI’s tend to give out loans to risky clients who bear more risk than the average loan rate covers (adverse selection) and who may engage in
2
riskier, unproductive activities after receiving the loan (moral hazard). A few of these problems are discussed below.
The increasing number of MFI’s in remote areas stimulates competition and triggers institutions to pursue as much customers as possible in order to increase their market share. As a result, the very poor are talked into loans they would otherwise reject (Kiviat and Farooq, 2008) and MFI’s approve loans they would otherwise decline. Consequently, the very poor’s inability to pay back forces them to default on the loan. Aside from these forced loans, Kiviat and Farooq (2008) further mention the unchanged high interest rates charged by MFI’s as a possible contributor intensifying the inability of the very poor to repay. Contrary to the positive outlook of the
authors mentioned earlier, competition has yet to reduce interest rates as expected in this industry (Counts, 2008; Lewis, 2008; Gross & Phillips, 2008; Hamm, 2008). Lewis (2008) discusses further that most MFI’s point out the loan size and the accompanied high level of operation costs as justification for these high interest rates. However, as the market becomes more competitive MFI’s are forced to be more efficient. Therefore it should be safe to assume a reduction of the interest rates. Yet, as the market becomes more competitive and commercialized, some regions still bear with high interest rates. In fact, several publications presented Banco Compartamos3as a perfect illustration of this issue when they went public in 2007, became profitable and yet kept their interest rates on an average of a 100 percent annually (Counts, 2008; Lewis, 2008; Kiviat and Farooq, 2008).
Moreover, the lack of communication among MFI’s are considered a factor intensifying Adverse selection and Moral hazards problems as well. The lack of communication between MFI’s because of the growing competition often results in multiple loans (also referred to as double-dipping). According to Lascelles (2008) multiple loans are a result of the fierce competition in the Microfinance industry and are considered a vital explanation for Loan Delinquency. As mentioned earlier, in order to gain more customers, MFI’s lower their standards by giving out loans to excessive risky clients. As a result, MFI’s may give out loans to clients already in debt at a competitor MFI. In this matter, these clients run up debt to several MFI’s simultaneously which they are unable to repay later on. McIntosh, de Janvry & Sadoulet (2004) confirm the latter in their case study on Uganda’s largest MFI (FINCA). They analyze the impact the entry of new
MFI’s in the market has on the clients’ behavior. The results showed that increasing competition caused by the entry of new lenders encourages multiple loans by clients indeed. As a result, these clients are more likely to deteriorate the repayment of their loans and making savings deposits because of their inability to keep up with several payments at the same time.
As noted above, many articles have been published and a few case-studies have been done on the effect of competition on Loan Delinquency in the Microfinance industry. However, very little has been proven empirically.
Luckily, the Banking industry provides abundant theoretical and empirical literature containing evidence from both perspectives.
The theoretical literature on competition contains two dominant paradigms; the
risk-shifting/competition stability effect and the franchise value /competition fragility. Additionally, a combined view of these theories has been developed recently by Martinez-Miera & Repullo (2008). They suggest a combination of the risk-shifting effect and a margin effect.
These three views are discussed below.
Under the first model, competition is seen as negatively correlated with bank risk. More specifically, as competition increases, loan rates decrease, making it cheaper for borrowers to pay back their loans (Boyd & De Nicoló, 2005). Hence, the probability of Loan Delinquency decreases. Accordingly, more market power would increase loan rates making it difficult for borrowers to pay back (Jimenez, Lopez & Saurina, 2007; Berger et al. 2008). Consequently, Loan Delinquency increases as moral hazard problems exacerbate.
The second model converges to the classic franchise value paradigm. Under this model, as competition increases, profit margin decreases reducing the franchise value of banks (Berger et al. 2008). In order to increase returns and recover their franchise value, banks are more likely to take excessive risks. The latter creates more moral hazard and adverse selection problems as these banks might give credit to risky borrowers. As a result, the probability of Loan
Recently, Martinez-Miera and Repullo (2008) published an alternative view of the competition and risk relationship. They suggest a combination of the risk-shifting effect discussed above and a margin effect. The graph below illustrates the relationship suggested by Martinez-Miera and Repullo (2008).
Figure 1: competition theory. Source: Martinez-Miera & Repullo (2008)
The transition from the risk shifting effect to the margin effect is reliant on the type of
correlation between loan risk and bank risk. Specifically, the risk shifting effect theory assumes a perfect correlation between loan risk and bank risk, i.e. the probability of a loan default is
identical to the probability of a bank default (Martinez-Miera & Repullo, 2008). That being said, Martinez-Miera & Repullo (2008) introduce the possibility of an imperfect correlation between these two risks. They suggest two factors conducing to loan risk; a systematic and an
idiosyncratic risk factor4. Perfect correlation between bank risk and loan risk only exists when the loan risk is comprised of solely the systematic factor. In every other situation where there is a probability of idiosyncratic risk, an imperfect correlation holds.
In practice this situation can be explained as follows. Take point A in the graph as being a monopolistic market (none or minimum level of competition). From point A to point B, the risk shifting effect holds, i.e. a rising competition reduces the probability of Loan Delinquency
because of lower loan rates. However, lower loan rates implicate less revenue from the non defaulting loans as well. Consequently, as competition intensifies (and loan rates keep
decreasing) from point A to point B, banks have to cope with a reduced buffer to cover for loan losses (caused by idiosyncratic risk factors). This reduced buffer poses a threat to banks from point B on; indicating that the positive effect competition had on loan delinquency has vanished because of the insufficient buffer to cover for loan losses (caused by idiosyncratic risk factors). From point B to point C the opposite relationship between competition and loan risk holds, i.e. the margin effect. In this situation, the reduced buffer to cover for loan losses forces banks to engage in riskier activities in order to recover and increase their revenue. As the level of competition increases further, these moral hazard and adverse selection incentives increase the probability of Loan delinquency.
Empirically, numerous studies analyzed the relationship between competition and bank risk5. However as this research focuses on loan risk in particular, discussed below are studies that investigated loan risk separately from bank risk.
Berger et al. (2008) tested the competition-stability and the franchise value theories on a sample of 8,235 banks in 23 developed countries. The Z-index and non-performing loans served as proxies for overall bank risk and loan risk respectively. The Lerner Index and the Herfindahl Index were both used as proxies for market power. The results of this research showed support for both the competition stability and the franchise value theories, confirming the combined theory of Martinez-Miera & Repullo (2008). They argue that competition does lower loan rates making it easier for borrowers to repay their loans.However as an increasing number of banks compete, profit margins decline, forcing banks to take excessive risks in order to increase their returns. As these risks entail approving loans to marginal applicants6, the quality of their loan portfolio deteriorates. Hence, the bank becomes more fragile.
On the other hand, Salas & Saurina (2003) found evidence of a positive relationship between the level of competition and loan risk. They analyzed the relationship between deregulation, market
5
Loan risk does not need to be equivalent to overall bank risk. A bank can have high loan risk and still have a low overall risk. Several other methods can be used to influence overall bank risk (Berger et al., 2008)
6
power and risk behavior of 21 Spanish Banks. The analysis included Tobin’s q7as a proxy for market power; time-dependent dummies for regulatory changes in Spain and the ratio of equity over total assets and loan losses over total loans as proxies for bank risk and loan risk
respectively. The results showed that deregulation of the Spanish banking system caused a reduction in market power (thus, increased the level of competition) of local banks, which in turn encouraged higher loan risk. More recently, Jimenez, Lopez & Saurina (2007) tested the
relationship between competition (using the Lerner-Index) and loan-risk (using non-performing loans) on a similar sample. Analyzing data on 107 Spanish Banks, they found support for the franchise value paradigm as well. Their results showed that a decline in the competition level among the sample banks induces a reduction in the amount of nonperforming loans.
Research question
Competition may stimulate MFI’s to gain as much clients as possible to achieve sustainability. However the consequences of doing so may not be worth it as these MFI’s may engage excessive risk. One of these consequences, the subject of this research, is Loan delinquency. In the end, more delinquent loans indicate less outreach as the clients are not able to pull themselves out of poverty with the given loan.
This poses a threat to the Microfinance industry. If because of an increasing level of competition among MFI’s the risk of defaulters increase, the ultimate goal of Microfinance, which is poverty reduction, will not be achieved.
This research is an attempt to empirically test if the level of competition has an influence on Loan Delinquency. Therefore, the research question is:
Does the level of competition influences Loan Delinquency among MFI clients in the Microfinance industry?
To answer this research question a null-hypothesis is formulated:
H (0): the level of competition does not have an influence on Loan Delinquency. And therefore the alternative hypothesis to be tested is:
H (1): the level of competition has an influence on Loan Delinquency.
Methodology
Since the advent of Economic science, more specifically the Industrial Organization (IO) field, many competition theories have been developed in an attempt to understand, measure and assess this phenomenon.
While there are abundant competition theories and assessment tools, statistical measurement of competition remains challenging. To date, various methods to measure competition are
developed by scholars in empirical studies. Discussed below are several methods that can be applied to financial institutions8.
First, the Lerner index (also identified as Price-Cost Margin (PCM)) is discussed. The Lerner index measures the market power of a given firm. The theory behind this method focuses on the price and marginal costs incurred by a given firm. The formula used for this method is:
(1 ) L(ι) = (P (ι) - MC (ι)) / P (ι)
Where L symbolizes the Lerner Index of firm ι, P the price of firm ι and MC the marginal cost of firm ι.
The Lerner Index has a value from 0 to 1, with 0 indicating no market power (fierce competition) and 1 indicating absolute market power (monopolistic market). More specifically, a firm in a highly competitive market is forced to reduce its price to marginal costs in order to survive (Creusen, Minne & van der Wiel, 2006). Holding total cost constant, a lower price would indicate a lower price-cost margin as well. Hence, the Lerner Index would be around 0. On the other hand, in a monopolistic market the firm can raise its price as high as it desires in order to maximize its profit. Consequently, the price level can be raised to such a level that the marginal cost has a minimum impact; hence the Lerner Index approaches 1.
The second measure is the Profit Elasticity (PE) developed recently by Boone, van Ours & van der Wiel (2007). The PE estimates the percentage drop in a firm’s profit due to a percentage increase in its marginal costs (Boone et al. 2007). As elaborated by Boone et al. (2007), in a competitive market, an increase in cost reduces the firm’s profit through either an increase in selling price (reducing demand for the firms product), or a lower price-cost margin (by
maintaining the same price). A fierce competition environment penalizes these inefficient firms severely in terms of profit. To calculate by how many percentages a percentage increase in costs reduces profit, the following formula is used:
(2 ) ln(PR(ι)) = α−βln(MC(ι))
Where, PR is the profit of firm ιand MC is the marginal cost of firm ι. The βis the actual Profit elasticity. It indicates the percentage drop (by the negative sign in front of it) in profits due to a percentage increase in marginal costs. Consequently, a higher level of βindicates a higher level of competition as profits are more affected by a percentage increase in marginal costs. The marginal costs referred to in this formula can be approximated by dividing variable costs by total revenue incurred by the respective firm9.
Third, named after Rosse and Panzar (1977), the H-statistic measures the sum of elasticity’s of the firm’s total revenue with respect to its input prices (Goddard & Wilson, 2006). Somewhat similar to the PE, the H-statistic measures how much an increase in cost affects the level of total revenue (instead of profits as in the PE equation). This method uses the formula:
(3 ) ln(R(ι,t)) = α(ι,t) + β1ln (P(ι,t)) + β2ln (Y(ι,t)) + ε(ι,t)
Where, R is the total revenue (interest income for banks) of firmifor periodt, Pis a proxy for
input pricesof firm ifor periodt, Y is a vector for control variables affecting the revenue of firmi
for periodtand εis the error term. From this equation the H-statistic is then calculated by
summing up the elasticity’s for the input prices (β1) and the control variables (β2). The input
prices referred to in this equation include: interest expenses, personnel expenses and operating and other expenses.
The theory behind the H-statistic assumes a situation of long run equilibrium (Goddard & Wilson, 2006). In this situation, a rise in cost (and therefore price) in a monopolistic market reduces demand and therefore the revenue10. On the other hand, in a market facing perfect competition, there are no long-run economic profits, i.e. price will equal cost at all times. Based on the above, H-statistic
‹
0 indicates a monopolistic market, in which one percentage increase in input prices causes a βpercentage drop in revenue. Moreover, an H-statistic equal to 1 indicates perfect competition. More specifically, as a firm’s price equals its cost in the long run in a market with perfect competition, a percentage change in costs would lead to an equalpercentage change in revenue. Lastly, 1
›
H11›
0 indicates a market facing monopolistic competition. As this market form lies in between the two above mentioned extremes (monopoly and perfect competition), the effect of cost on revenue is a combination of the two extremes as well. In fact, a rise in cost will cause a disproportionate increase in revenue, allowing the firm to generate profit.The fourth and last method is called the Herfindahl Index (HHI). Unlike the above measures, the Herfindahl Index can only be measured at an industry/country level. This index measures a firm’s market share in relation to the industry. The index gives an indication of the concentration ratio. This theory assumes that a higher concentration indicates a lower level of competition in a given market. The Herfindahl index is calculated with the formula:
(4 ) HHI=
∑
(MS1….n) ²Where, MS is the market share of firm1to firmn. In terms of the equation, the HHI is the sum of the squares of each firm’s market share in a specific industry. The value of the index can range
10
In a long run equilibrium, maximum profit can be obtained when Marginal revenue equals Marginal cost
(MR=MC) in a monopolistic market. In addition, a monopolist faces a downward sloping demand curve in the long-run because it faces no competitors. Hence, it can choose a price to achieve desired level of profit. Additionally, in a long run equilibrium, Price equals average cost (P=AC) in a market with perfect competition.
11
from 0 to1, with 1 indicating a firm with 100% of market share (monopoly) and 0 indicating a market with many firms owning equally small market shares (Creusen, Minne & Van der Wiel, 2006).
After considering all measurement options, the Lerner Index (PCM) is chosen for this research. Since this research is done at a firm level, the Herfindahl Index is automatically excluded. Moreover, the choice for the Lerner Index is mainly based on the availability of data. The data used in this research are retrieved from a single database kept by MixMarket12. An assessment of the variables needed for the calculations of the remaining 3 competition measures determined that the Lerner Index is the only measure for which an approximation of its value is present in the database. More specifically, the Lerner Index’s formula roughly coincides with the basic formula generally applied in economics to calculate profit margin, i.e. (price-cost)/price which indicates how much a firm earns with the sale of each unit it sells (difference between selling price and the cost). The data required to calculate the cost-variable and the input-price-variable in the formula of the Profit Elasticity and the H-statistic respectively were lacking and no other variable could be used as an approximation of their values.
12
Conceptual model
Figure 2: Conceptual Model
Illustrated by the above, controlling for the Size, Age, Type and Region of MFI’s, a rising competition as a result of commercialization induces to higher Loan Delinquency rates.
Derived from the conceptual model, the following regression model is created: Size Type of MFI
(5 ) LD ι=αi + β1(W1COMP, ι) - β2(W1COMP, ι)² + β3ln(Y1SIZE, ι) +
β4 (Y2TYPE, ι) + β5 (Y3AGE, ι) + β6(Y4REGION,ι) + ει
Where, LDitis the Loan Delinquency rate of MFI (ι), COMP is the competition level of MFI (ι), COMP² is the quadratic term for the competition level of MFI (ι ) allowing for nonlinear
relationship, SIZE is the control variable for the size of MFI(ι), TYPE is the control variable for the type of MFI (ι), AGE is the control variable for the years of existence of MFI (ι) and REGION is the control variable for the region in which MFI (ι) is established13.
More specifically, the dependent variable in this research is Loan Delinquency. This variable is measured by the write-off ratio of the MFI’s. According to MicroRate (2003), the write-off ratio represents the loans that the MFI has removed from its books because of a substantial doubt that they will be recovered. It symbolizes the risk MFI’s incur in their loan portfolio because of the probability of not recovering the capital they have lent. It is calculated as a percentage derived from the write-offs within a 12 month period divided by the average gross loan portfolio in those 12 months14. In terms of the factors in its formula, it is expected that competition will cause a higher increase in the write-offs within 12 months compared to the increase of the average loan portfolio in those 12 months.
The independent variable which is expected to have an influence on Loan delinquency (the dependent variable) is the level of competition. The level of competition is measured by the Lerner Index. As mentioned earlier, the formula to calculate the Lerner Index is: (p-mc)/p, where p represents the price and mc represents the marginal costs of a given firm. For the purpose of this research, the Lerner Index will be approximated by the profit margin ratio (which as mentioned earlier is roughly calculated by the same formula) of every individual MFI. In the case of a rising profit margin and therefore rising market power, it is assumed that competition is declining as the respective MFI is becoming more powerful within the market. This research is done at a firm-level, which implicates that competition is defined as competition experienced by each MFI individually. When an MFI’s market power declines, the institution is coping with more competition from other institutions and therefore is obliged to reduce profit
13
Noteworthy mentioning is fact that the control variables Type and Region of MFI represent a merely symbolic value in the regression model. They are categorical variables and therefore the direction sign (+, -) is of no particular significance. They are accounted for in a different matter than the other control variables.
14
margin in order to survive and vice versa. In short, rising profit margin symbolizes declining competition and declining profit margin symbolizes rising competition.
To isolate factors that may influence the dependent variable, the following control variables are taken into consideration.
First, the Age of the MFI is measured by the number of years since the establishment of the respective MFI. According to benchmarks established by MicroBanking Bulletin (2008), a bigger part of mature15MFI’s (76.8 percent) are commercialized compared to younger and new ones (62.0 and 68.6 percent respectively). Additionally, write-off ratio’s are on average higher (1.7 percent) for mature MFI’s than for new and younger ones (0.4 and 0.9 percent respectively).
Following the conceptual model of this research, the above would imply that older MFI’s stimulate an increase in Loan Delinquency.
Second is the log value of the Size of the MFI. It is measured by the total assets of each MFI. This proxy for Size is often used in empirical studies on financial institutions (Uchida, Udall & Watanabe, 2006; Berger et al., 2008; Gomez-Gonzalez & Grosz, 2006 among others). Size may have an influence as bigger MFI’s may have more resources to prevent Loan Delinquency than smaller MFI’s.
Third, the Type of MFI is considered a control variable as well, as it may affect the Loan Delinquency level of the MFI. Based on the conceptual model, e.g. NGO’s (Non-Governmental Organization), being the least commercialized type of MFI, may experience lower levels of Loan delinquency than the more commercialized types. In the MicroBanking Bulletin (2008)
publication this division is used when comparing portfolio at risk and the results showed different levels of portfolio at risk for different types of institutions. According to this
publication, NGO’s experience the lowest value of write-off’s, while the commercialized types cope with the highest values. The six types of MFI’s used in this research are; NGOs, Non-bank Financial Institutions, Cooperatives/Credit Unions, Rural Banks, Commercial Banks and Other16. For assessment purposes, they are labeled from 1 to 6 respectively.
Finally, discussed in the MicroBanking Bulletin (2006) as well, the Region in which the MFI is located is controlled for as well. Explained further in a more recent publication of MicroBanking
15MicroBanking Bulletin (2008) classified the age of MFIs as follows; Mature >8 yrs, 4<Young <9 and new <5. 16
Bulletin (2008), not all regions are at the same stage of commercialization in the Microfinance industry. For instance, Latin America is considered the most commercialized Microfinance region in the world. According to the conceptual model discussed before, if commercialization spurred the advent of competition among MFI’s, it can be assumed that regions with a higher level of commercialization have a higher level of Loan Delinquency as well. The regions from which the data sample is comprised of are: Africa, East Asia & the Pacific, Eastern Europe & Central Asia, Latin America & the Caribbean, Middle East & North Africa and South Asia. Similar to the Type of MFI, these regions are labeled from 1 to 6 respectively.
Data
This research is an empirical research. The data for the dependent and the independent variables are collected from the Microfinance Exchange website (www.Mixmarket.org) and afterwards analyzed in order to accept or reject the null hypothesis.
The original sample consisted of 750 MFI’s, accounting for 2250 observations, listed on the MixMarket as 4 and 5 diamonds institutions17. After filtering the data; eliminating MFI’s from which data is incomplete; in addition to MFI’s with competition level in the top and bottom 1 percent of the distribution; and institutions from which Microfinance operations comprised less than 99% of their services, the sample was reduced to 2006 observations.
Based on availability, data is retrieved for the years 2002-2007 from the MixMarket. The years in which data are collected are unfortunately not controlled for. This is because the database the data is retrieved from does not contain data on the majority of MFI’s used in this research in a consecutive manner. That is, for the majority of MFI’s in the sample, data is available for barely half of the time span studied in this research, making a comparison on a year to year basis difficult. Hence, every year represents an independent value.
Data are analyzed by first doing a correlation test to confirm a statistical significant relationship between the dependent, independent and control variables. If the variables are indeed correlated, a regression analysis is done to observe how much of the Loan Delinquency rate can be
explained by the level of competition as well as the direction of such relationship (negative or
17
positive). Based on the existing theories of the relationship between competition level and Loan Delinquency discussed in the literature review, this research allows for non-linear relationship.
Empirical results
Tables 1 and 2 in the Appendix show the descriptive statistics of the dependent, independent and control variables used in this research. As noted in table 1, the total number of observations amount to 2006. From these observations, 17.5 percent are accounted for by the African region; 10.2 percent by East Asia & the Pacific; 20.2 percent by Eastern European & Central Asia; 39.1 percent by Latin America & the Caribbean and 6.5 percent by both the Middle East & North Africa and South Asia.
Throughout the analysis, the relationship between the level of competition and the Loan
Delinquency rate may be confusing. Therefore keep in mind that the measure for competition is the profit margin and that a rise in the latter indicates reduction in the level of competition and vice versa. Thus, in terms of the dependent and independent variable, a negative coefficient (B) value would indicate that a rising level of competition increases Loan Delinquency.
Before the analysis, it is noteworthy mentioning that the control variables Type of MFI and Region of MFI are not included in the correlation test or as independent variables in the
regression analysis. This is because they are categorical variables, hence, their coefficients in the correlation test and the regression analysis would be irrelevant as the numbers given to each category are merely for identification purposes. Instead, regression analyses are done for each type and region of MFI separately afterwards.
correlated (-0.221) with Loan Delinquency. In other words as competition increases (profit margin decreases), the Loan Delinquency level increases as well. The quadratic function of the competition variable indicates the opposite relationship (0.092), confirming a combined relationship. Thus, at a certain level, an increase in competition (reduction in profit margin) induces a reduction in loan delinquencies.
As for the control variables, only the Size of the MFI’s is negatively correlated with Loan
Delinquency at a 1 percent statistical significance level. Thus, as the MFI’s increase in size, Loan Delinquency decreases.
More over, the Age of MFI’s shows no statistically significant correlation with Loan Delinquency. However, both control variables correlated with each other and with the
independent variable (level of competition) creating a data multicollinearity problem. In order to decide whether these control variables should be cut off or not, the Variance Inflation Factor (VIF)18is calculated to assess the severity of the data multicollinearity. The results of this test are shown in table 4. Based on these results, none of the control variables are excluded because most of the VIF values remain around 1indicating no threat to the regression model.
Furthermore, table 5 in the Appendix shows the results of the regression analysis. Statistically significant relationships are found for the two competition measures only. They account for 5.7 percent of the dependent variable, Loan Delinquency. Similar to the correlation results, the first competition measure indicates a negative relationship whereas the quadratic measure indicates a positive one. However, the coefficient of the latter is 0, signifying that it has no influence in the regression19. Therefore, only the first competition measure is taken into consideration.
That being said, the results suggest that for each level increase in competition, the level of loan delinquency increases by 0.043. This result shows support for the franchise value paradigm, competition fragility and the margin effect. Under the pressure of competition and watching their profits erode, MFI’s engage in riskier activities in order to recover their franchise value. As these riskier activities imply lending to riskier clients, a higher level of loan delinquency can be
expected. Less profit symbolizes a lower buffer to cover for loan losses as well, which
18
The VIF measures how much data multicollinearity affects the variance of a coefficient (ß) in a multiple regression model (Stine, 1995).
consequently results in a higher level of loan delinquency as these MFI’s are triggered to take on risky activities in order to recover their stability.
In addition to the above reasoning provided by existing literature in the banking industry, many scholars suggested specific factors contributing to this relationship within the Microfinance industry as well. As discussed in the literature review, unchanged high interest rates, forced loans and multiple loans are considered examples specific to the Microfinance industry intensifying moral hazard and adverse selection problems which result in higher level of loan delinquency (Counts, 2008; Lewis, 2008; Kiviat & Farooq, 2008; Gross & Phillips, 2008; Hamm, 2008; McIntosh, de Janvry & Sadoulet, 2004).
Moreover, table 6 shows the results of the regression analyses for each region separately.
Remarkable for all regions is the fact that similar to the first regression, the quadratic measure of competition is irrelevant to the analysis. For some regions it does not have a statistical significant influence on Loan Delinquency while for other regions the influence is zero.
The only region that showed no statistical significant results for any independent or control variable is the Eastern Europe and Central Asia region. Accounting for 20 percent of the total sample it can be concluded that neither competition nor the control variables have a statistical significant influence on Loan Delinquency in this region. In fact, the publication Benchmarking
Microfinance in Eastern Europe and Central Asia by Microfinance Information Exchange
(2007) provides evidence that could explain the result found. According to this publication, the Eastern European and Central Asian region is in general considered the youngest Microfinance region, with the lowest poverty rate and Loan Delinquency rate among all the other regions. Thus, considering that younger MFI’s are less commercialized (MicroBanking Bulletin, 2007) than mature ones and that the low poverty rate may not attract as much MFI’s as in other regions, it can be safely assumed that competition (and the control variables) does not have a significant effect on the loan delinquency rate in this region.
surprising, as these two regions are considered the most commercialized Microfinance regions in the world (MicroBanking Bulletin, 2008)20. On the other hand, the African region embodies a different scenario. Despite the increasing number of commercialized funding, deposits remain the prominent funding source in the African region (MicroBanking Bulletin, 2008). Therefore, the competition may not be as fierce as in the other regions. The results confirm this as the level of competition in the African region has the least effect on Loan Delinquency compared to the other regions.
Moreover with the highest R² value, approximately 18 percent of loan delinquencies in South Asia can be explained by the regression model. However, unlike the other regions, apart from the competition measure, the Age of the MFI has some influence on Loan Delinquency as well. Based on the results for the South Asian region, older MFI’s experience a higher level of Loan Delinquency compared to the younger ones. This finding confirms the expectations based on the benchmark statistics mentioned earlier regarding the Age of MFI’s and write-off ratio’s
(MicroBanking Bulletin, 2007).
Finally, table 7 shows the regression results by type of MFI.
When categorized by type, the group “Other” shows no statistical significant coefficients for none of the variables. However, in contrast to the situation of the region Eastern Europe and Central Asia, the group “Other” only accounts for 2.1 percent of the complete sample. Perhaps this amount is not large enough to derive meaningful conclusions. Hence, this type of MFI is not discussed.
Similar to the two previous regression analyses, the quadratic competition measure is left out of discussion because it has no statistical significant value for four types of MFI’s whereas for one of them (NGO) it has a value of zero.
From the remaining five types, NGO’s, Non-Bank Financial Institutions, Cooperatives and Commercial banks show a negative relationship between the competition measure and Loan Delinquency.
A negative coefficient for these types of MFI’s indicate that although the level of competition in the more commercialized types of MFI’s are expected to have a stronger effect on Loan
delinquency, all type of MFI’s are affected by this increasing level of competition. For instance, the level of competition of Commercial banks having the strongest effect on loan delinquency simply confirms these expectations. However, the results for NGO’s are somewhat surprising. Specifically, contrary to expectations, the level of competition has the second strongest effect on loan delinquency and the highest value of R². A possible explanation for this result could be associated with the fact that profit margins of NGO’s are already lower than those of
commercialized types of MFI’s21. Additionally, the limited services NGO’s are allowed to offer leave them with almost no other competition strategies to attract and retain clients besides lowering the interest rates. Consequently, an increase in the level of competition forces them to further reduce their profit margins, causing distress and intensifying moral hazard and adverse selection incentives. In short, focusing on the two extremes, NGO’s and Commercial banks, a higher coefficient for commercial banks does confirm the expectations set out in the conceptual model of this research. That is, that the level of competition of more commercialized MFI’s does result in a higher level of Loan Delinquency. Nevertheless, a test22is carried out on both
coefficients (0.047 and 0.081) to determine if they are statistically different from each other. This is done to ensure the validity of the above found results (i.e., that competition level of
commercial banks do have a stronger effect on Loan delinquency than the competition level of NGO’s). With a standard error of 0.005 for the competition coefficient of NGO’s, it can be confirmed that the level of competition of Commercial banks has more effect on Loan Delinquency than the competition level of NGO’s. Thus, although NGO’s have a different pricing policy which may affect the Lerner Index (measured by the profit margin), the results obtained are to be considered accurate.
Noteworthy as well is the result for Rural Banks. In contrast to the other types of MFI’s, competition does not have a statistical significant influence on Loan Delinquency. Instead, the Size of MFI’s shows a strong negative relationship (-0.326) with Loan Delinquency at a 1 percent level of significance. Thus, in the case of Rural Banks, as MFI’s grow larger (in terms of assets), Loan Delinquency is reduced. Confirming the expectations regarding the size of MFI’s,
21This is because of the non-profit business model. For this, it would be expected that the Lerner Index for this type of MFI is among the lowest as the profit margin would be at a minimum.
22
it can be assumed that larger MFI’s are in possession of more resources in order to diminish Loan Delinquency.
In comparison, the Region of the MFI determines more of the effect of competition on Loan Delinquency than the Type of MFI (derived form the higher levels of R² by region).
In general it can be established that the results of this research showed support for the franchise value/margin effect/ competition fragility theories. That is, the level of competition induces an increase in the level of Loan Delinquency. In addition to the reasoning behind the franchise value and the margin effect, some factors specific to the Microfinance industry such as forced on loans, multiple loans and unchanged high interest rates may help explain the obtained results as well (Kiviat & Farooq, 2008; McIntosh, de Janvry & Sadoulet, 2004 Counts, 2008; Lewis, 2008 among others).
To summarize, it can be concluded that the level of competition in the Microfinance industry does have an influence on Loan Delinquency among clients of most types of MFI’s in most regions.
Conclusion
This research is an attempt to assess empirically the existence of a relationship between the dependent variable Loan Delinquency and the independent variable competition.
From an initial correlation test data multicolleanarity problems were indicated among the independent variable and the control variables. However after assessing the severity of the problem (using the VIF), none of the variables were excluded. Subsequently, three regression analyses were presented. One regression analysis was done on the dependent variable (Loan Delinquency), the independent variables (competition and competition²) and two of the control variables (Size and Age). The other two regression analyses were performed in a similar matter as the first one; however they were specified by Region and by Type of MFI.
In general, all three regression results showed the expected relationship between competition and Loan Delinquency. In other words, an increasing level of competition causes Loan Delinquency level to increase as well.
Based on the theories discussed on competition and loan risk, it can be concluded that this
research provided support for the franchise value/margin effect/competition fragility theories. An eroding franchise value and lower buffer to cover for loan losses as a result of competition induces MFI’s to engage in riskier activities which results in higher levels of Loan delinquency. Exclusive to the Microfinance industry; multiple loans, unchanged high interest rates and forced loans can be considered examples of those risky activities (Kiviat & Farooq, 2008; Counts, 2008; Lewis, 2008; McIntosh, de Janvry & Sadoulet, 2004 among others).
Concluding, the null hypothesis has been rejected. Hence, the level of competition does have an influence on Loan Delinquency.
MixMarket. As MFI’s are not obliged to participate and post information on this website, the database may give a skewed view of reality. In fact, it would be logical to assume that only the better MFI’s would volunteer in giving information. Moreover, depending on the information needed, the database is not detailed enough. For instance, only one measure of competition (Lerner Index) could have been used in this research because of data availability. As mentioned earlier NGO’s being non-profit institutions, are likely to have a different pricing policy compared to the profit seeking MFI’s, which may affect the Lerner Index. Perhaps, a comparison of several competition measures could contribute to a more robust analysis.
In addition, the main limitation in this research would be the fact that the years were not
controlled for. Access to additional sources would perhaps provide a more comprehensive set of data from which year to year comparisons can be made.
Finally, the number of MFI’s per type of institution or region is not necessarily representative. As earlier mentioned MFI’s were chosen based on the information disclosure and therefore other representativeness matters were disregarded.
Appendix
Appendix 1
Figure 2: conceptual model on causes and effects of client drop-out. Source: M-CRIL (2005)
Appendix 2
Descriptive statistics
Variable N Minimum Maximum Mean Std.Deviation
Loan Delinquency 2006 0 92.00 2.0970 4.41677 Competition 2006 -252.63 53.70 8.7284 31.29115 Ln Size of MFI 2006 10.03 22.82 15.9192 1.73140 Age of MFI 2006 2 53 14.88 8.540
Descriptive Statistics Africa East Asia & The Pacific Eastern Europe & Central Asia Latin America &The Caribbean Middle East & North Africa South Asia Total NGO 95 62 55 416 86 74 788 Non-Bank Financial Institution 120 61 221 200 25 40 667 Cooperative/Credit Union 87 2 29 80 0 2 200 Commercial Bank 43 11 101 73 0 11 239 Rural Bank 0 69 0 0 0 1 70 Other 7 0 0 14 19 2 42 Total 352 205 406 783 130 130 2006
Appendix 3
This table shows the result of the correlation test performed on the variables in order to confirm a relationship between the dependent variable (Loan Delinquency rate), the independent variable (level of Competition) and the control variables (Size, Type, Age and Region of MFI’s). The * and ** signs indicate statistical significance at the levels 5% and 1% respectively.
Pearson’s Correlation
Loan Delinquency rate
Competition Competition² Ln Size of MFI Age of MFI Loan Delinquency rate 1 Competition -0.221** 1 Competition² 0.092** -0.705** 1 Ln Size of MFI -0.069** 0.225** -0.157** 1 Age of MFI -0.022 0.094** -0.083** 0.149** 1
Appendix 4
The following table shows the result of a regression analysis performed to assess the Variance Inflation Factor (VIF). The control variables which correlated with the level of competition were the independent variables whereas the level of competition represented the dependent variable. A VIF value lower than 5 is considered acceptable23.
Multicollinearity test: Variance Inflation Factor
Competition Competition² Size of MFI Age of MFI
Competition - 1.073 1.995 2.079
Competition² 1.029 - 1.991 1.993
Size of MFI 1.045 1.088 - 1.071
Age of MFI 1.026 1.026 1.009
-N 2006
Table 4: Multicollinearity test of correlated variables
Appendix 5
The table below shows the regression results with Competition as independent variable, the control variables Size of MFI and Age of MFI and Loan Delinquency as dependent variable. The ** sign indicates statistical significance at 1% level.
Regression results Variable N R² B T-statistic Competition 2006 0.057 -0.043** -9.825 Competition² 2006 0.057 0.000** -4.117 Ln Size of MFI 2006 0.057 -0.025 -0.432 Age of MFI 2006 0.057 -0.001 -0.101
Table 5: Regression results of all numerical variables on the dependent variable
Appendix 6
The table below shows regression results broken down by region. The * and ** signs indicate levels of statistical significance of 5 % and 1 % respectively.
Regression results N R² B t-statistic Africa 352 0.062 Competition -0.030** -3.309 Competition² 0.000 -1.772 Ln Size -0.261 -1.579 Age 0.052 1.348
East Asia & the
Pacific 205 0.156 Competition -0.073** -4.747 Competition² 0.000** -3.864 Ln Size -0.280 -1.095 Age -0.021 -0.935 Eastern Europe
& Central Asia 406 0.009
Competition -0.013 -1.100 Competition² 0.000 0.643 Ln Size -0.093 -0.666 Age 0.127 1.321 Latin America & the Caribbean 783 0.096 Competition -0.067** -8.113 Competition² 0.000* -2.023 Ln Size 0.109 1.268 Age -0.029 -1.557
Middle East &
North Africa 130 0.163 Competition -0.049** -3.849 Competition² 0.000 -1.120 Ln Size 0.029 0.096 Age 0.145 1.847 South Asia 130 0.183 Competition -0.036** -3.595 Competition² 0.000 -1.131 Ln Size 0.229 1.902 Age 0.076** 3.504
Appendix 7
The tables below show regression results broken down by type of MFI. The * and ** signs indicate statistical significance level of 5% and 1% respectively.
Regression results N R² B t-statistic NGO 788 0.114 Competition -0.047** -8.537 Competition² 0.000** -3.955 Ln Size of MFI -0.090 -1.009 Age of MFI 0.014 0.818 Non-Bank Financial Institution 667 0.015 Competition -0.023* -2.640 Competition² -9.242E-5 -1.443 Ln Size of MFI -0.160 -1.130 Age of MFI 0.035 1.040 Cooperative/ Credit Union 200 0.142 Competition -0.046** -4.245 Competition² -8.096E-6 -0.062 Ln Size of MFI -0.027 -0.214 Age of MFI 0.022 1.104 Commercial Bank 239 0.087 Competition -0.081** -3.532 Competition² 9.666E-5 0.300 Ln Size of MFI 0.197 0.746 Age of MFI -0.049 -0.671 Rural Bank 70 0.140 Competition 0.009 0.231 Competition² 0.000 -0.801 Ln Size of MFI -0.326** -2.532 Age of MFI -0.003 -0.238 Other 42 0.254 Competition -0.009 -0.259 Competition² 0.000 1.027 Ln Size of MFI 0.345 0.384 Age of MFI 0.515 1.764
Reference list
Berger, A., Klapper, L. & Turk-Ariss, R. (2008). Bank Competition and Financial Stability. The World Bank: Development Research Group.
Boone, J., van Ours, J. & van der Wiel, H. (2007). How (not) to measure competition.
Discussion Paper Center. No. 32. Tilburg University
Boyd, J.H. & De Nicoló, G. (2005). The theory of bank risk taking and competition revisited. The Journal of Finance. Vol. LX, No. 3.
Campion, A. (2001). Challenges to Microfinance Commercialization. Journal of
Microfinanc,. Vol. 4, No.2.
CGAP. (2006). MicroBanking Bulletin: Focus on MFI performance by region. Issue No.
12.
CGAP. (2006). The history of Microfinance. Retrieved from:
http://www.globalenvision.org/library/4/1051/
CGAP. (2008). MicroBanking Bulletin.Issue No. 17
Charitonenko, S. & Rahman, S. M. (2002). Commercialization of Microfinance: Bangladesh. Asian development Bank.
Christen, R & Drake, D. (2001). Commercialization of Microfinance. Working
Draft.USAID.
Counts, A. (2008). Reimagining Microfinance. Stanford Social Innovation Review. Pages 45-53.
de Janvry, A., McIntosh, C. & Sadoulet, E. (2004).How Rising Competition Among Microfinance Institutions Affects Incumbent Lenders.
Goddard, J. & Wilson, J. (2006). Measuring competition in banking: A disequilibrium approach.
Gomez-Gonzalez, J. & Grosz, F. (2006). Evidence of Bank lending channel for Argentina and Colombia.
Gross, D. & Philips, M.(2008). Cheap Loans at Insanely High Rates? Give Us More.
Newsweek.Vol.152. No.13.Business Source premier.
Hamm, S. (2008). Setting Standards for Microfinance. Business Week Online. Business Source Premier.
Jimenez, G., J. Lopez & J. Saurina, (2007), How does competition impact bank risk taking? Working paper, Banco de España.
Kiviat, B. & Farooq, O. (2008). The Big Trouble In Small Loans. Time South Pacific (Australia/New Zealand edition), No. 23.
Lascelles, D. (2008). Microfinance Bananaskins: risk in a booming industry. Centre for
the Study of Financial Innovation, No.80.
Lewis, J. C. (2008). MicroLoan Sharks. Stanford Social Innovation Review. Pages 54-59. Littlefield, E. & Rosenberg, R. (2004). Microfinance and the Poor: breaking down walls between Microfinance and formal finance. Finance & Development.
M-CRIL (2005). Technical Note 1: Client Drop out rate.
MicroRate & Inter-American Development Bank (2003). Performance Indicators for Microfinance Institutions. 3rdEdition. Washington D.C.
Montgomery, H. & Weiss, J. (2004). modalities for Microfinance delivery in Asia and Latin America: Lessons for the People’s Republic of China. Working Paper Series, No.29.
Poyo, J. & Young, R. (1999). Commercialization of Microfinance: A framework for Latin America. Development Alternatives, Inc.
Pagura, M, Graham, D. & Meyer, R. (2001). Determinants of borrower Dropout in Microfinance: an empirical investigation in Mali. The Rural Finance program. Chicago, Illinois.
Ren, Y. & Schmit, J. T. (2006). Franchise Value, Competition and Insurer Risk-Taking. Stine, R. A. (1995). Graphical Interpretation of Variance Inflation Factors. The American
Statisticia, Vol. 49. No. 1.
Uchida, H., Udell, G. F. & Watanabe, W. (2006). Bank Size and Lending relationships in Japan. RIETI discussion paper series.
WorldBank (2008). Extreme Poverty definition. Retrieved from:
