SUBSIDIES IN THE MICROFINANCE SECTOR Their composition, allocation and effect on efficiency

(1)

Master thesis

SUBSIDIES IN THE MICROFINANCE SECTOR

Their composition, allocation and effect on efficiency

Evidence from 86 Microfinance institutions

University of Groningen

August 2009

Charlotte van Adrichem

Prins Hendrikkade 134f

1011 AR Amsterdam

Student number: 1344684

Email address: chvanadrichem@gmail.com

Supervisor: Prof. dr. B.W. Lensink

(2)

(3)

Abstract

Given the huge amount of subsidized funds going to microfinance institutions (MFIs), gaining understanding of subsidies in the microfinance sector should be considered a top priority. In my thesis I give insight into the composition of subsidies, their allocation and their effect on efficiency. I do this by analyzing 86 MFIs for which rating reports (PlanetRating) are available for three to four years per MFI between 2002 and 2007.

I identify three subsidy components; donated equity, paid-in capital and indirect subsidies. I find that, in general, donated equity is the most important subsidy component. Paid-in capital turns out to be an important source of subsidy for for-profit MFIs and MFIs that solely grant individual loans. Indirect subsidies form the least important source of subsidy for the MFIs in my sample.

I propose three forces that explain the allocation of subsidies in the microfinance sector, all with a negative influence on subsidy intensity; the MFI’s age, the MFI’s sustainability level and the poverty level of the clients that the MFI serves.

In my sample I find most support that subsidy intensity is decreasing with the age of MFIs, both when only including donated equity and when adding paid-in capital and indirect subsidies to the subsidy intensity measure. This supports the theory that subsidies are used as temporary measures to help MFIs cover start-up costs.

(4)

Introduction

Microfinance is seen as one of the most powerful tools to alleviate poverty. The goal of microfinance is to provide financial services to poor people that have been excluded from the formal banking sector (Morduch 1999). The Microcredit Summit Campaign (2007) reports that the scope of microfinance activities has increased enormously during the last decade. As can be seen in table 1, by the end of 1997 approximately 618 microfinance institutions (MFIs) served over 13 million clients. By the end of 2002 the number of clients had increased to approximately 67 million clients, served by more than 2500 MFIs. By the end of 2006 more than 3300 MFIs reached more than 133 million clients of which 92 million were among the poorest when getting their first loan. Despite the enormous expansion of microfinance activities, future growth is still possible as there are still at least 500 million households excluded from the financial sector (Hudon 2007). Date Number of institutions reporting Total number of clients reached (mlns) Number of poorest clients reported (mlns) End of 1997 618 13.48 7.60 End of 1998 925 20.94 12.22 End of 1999 1,065 23.56 13.78 End of 2000 1,567 30.68 19.33 End of 2001 2,186 54.93 26.88 End of 2002 2,572 67.61 41.60 End of 2003 2,931 80.87 54.79 End of 2004 3,164 92.27 66.62 End of 2005 3,133 113.26 81.95 End of 2006 3,316 133.03 92.92

Table 1: Overview of the scope and expansion of microfinance activities Source: Microcredit Summit campaign (2007)

(7)

thesis I give insight into the composition of subsidies, their allocation and their effect on efficiency. I do this by analyzing 86 MFIs for which rating reports (PlanetRating) are available. These rating reports entail detailed financial information (balance sheet and income statement) and performance indicators of each MFI for three to four years between 2002 and 2007. Understanding the current situation on the composition of subsidies in the microfinance sector, their allocation and their effect on efficiency is also very important in the discussion on the justification of the use of (long-term) subsidies and it is also important when searching for possibilities to improve the use of subsidies in the future.

My thesis has three goals. First of all, I will give insight into the composition of subsidies for the MFIs in my sample. I will identify three different components of subsidies (donated equity, paid-in capital and indirect subsidies through discounts on borrowing expenses) and I will also analyze the relative importance of these components.

Secondly, I will propose three forces that explain the allocation of subsidies in the microfinance sector; the MFI’s age, the MFI’s sustainability level and the poverty level of the clients that the MFI serves.

Young MFIs face high costs associated with the start-up of operations. Temporary subsidies enable MFIs to set interest rates at the long-term sustainable level providing continuity to the organization. I expect a negative influence of the MFI’s age on subsidy intensity, indicating that subsidies are targeted to young MFIs to cover start-up costs.

The sustainability level is the extent to which MFIs are able to cover costs by revenues. Donors should not focus on MFIs that are already sustainable or on MFIs that are able to attract capital on the market at market prices. Instead, donor funds should be allocated to MFIs that are less sustainable. At the same time, donors should encourage MFIs to improve their level of sustainability in order to prevent them from being limited by scarce and uncertain supply of subsidies from donors and governments (CGAP year; Cull, Demirgüç-Kunt and Morduch 2008). I expect a negative influence of the sustainability index on subsidy intensity, indicating that subsidies are targeted to the less sustainable institutions.

(8)

influence of the poverty level of clients on subsidy intensity, indicating that MFIs that serve poorer people receive more subsidies.

I will analyse whether the allocation of subsidies is in line with the previous theoretical arguments by means of scatter graphs and an OLS panel regression.

The third goal of my thesis concerns the effect that subsidies have on an MFI’s efficiency. Some people fear that subsidies encourage inefficient practices and discourage them to improve sustainability. This fear is mainly caused by bad experiences with subsidized credit in the 1980s (Morduch 1999; 2000). I will propose that subsidies have an efficiency-enhancing and diminishing marginal effect.

MicroBanking bulletin (2008) reports that there may exist large regional differences in the expansion of microfinance activities and the amount of subsidies received (in Tajikistan, Azerbaijan, Kenya, Bosnia and Herzegovina and Pakistan for example, the number of borrowers grew impressively (growth rates of above 50 percent)). Differences may also exist among MFIs with different governance structures, different loan products and the access to an (inter)national network. I expand my analysis by assessing whether these different characteristics influence the relationship between subsidy intensity and the MFI’s age, the MFI’s sustainability level and the poverty level of the MFI’s clients and whether they influence the effect that subsidies have on efficiency.

(9)

1 Data

1.1 Data source

I use rating reports of PlanetRating (86 MFIs). These rating reports include financial statements (balance sheet and income statement), performance indicators and information on other characteristics (MFI’s activities, loan products and institutional structure). For every MFI there are data available for three to four years between 2002 and 2007. Appendix 1 gives an overview of the MFIs included in my sample and the country of its operations.

PlanetRating constructs a rating based on the so-called GIRAFE methodology. GIRAFE is a weighted index of several indicators related to Governance (24%), Information (10%), Risk (10%), Activities (20%), Funding and liquidity (14%), and Efficiency and profitability (22%). Every rating domain is further specified in several rating factors (see appendix 2).

The data provided by PlanetRating is high-quality and very detailed. The selection of MFIs is not random as most of the MFIs in my sample (85 to 90 percent) choose to be rated. The other 10 to 15 percent of the ratings occurs upon request from investors and donors (Otto Wormgoor, PlanetRating). Almost all MFIs in my sample have access to an (inter)national network. Being part of an (inter)national network may have some advantages. It could for example be easier to attract (foreign) capital and MFIs could learn from the experiences of other MFIs in the network, implementing their best practices. The MFIs in my sample also belong to the richest MFIs (high sustainability index). The representativeness of my sample is therefore limited.

In order to conduct a cross-country analysis, I have converted the financial statements of 24 MFIs from Euros (EUR) into US dollars (USD), using the average annual EUR-USD exchange rate obtained from the European Central Bank (ECB).

I estimated the amount of indirect subsidies through a comparison of the interest rates charged to the MFIs and an adjusted base rate (benchmark rate). In section 2.2.3 I give an extensive explanation of the construction of the benchmark rate. Unfortunately there is no single base rate available for all countries in my sample. Appendix 3 gives an overview of the base rate that I have used for each country and the source of the data.

(10)

1.2 Data descriptions

Table 2 gives an overview of the number of MFIs included in my research per year, the total average outstanding loan portfolio of all MFIs, the average outstanding loan portfolio per MFI, the total number of active borrowers, the total numbers of active borrowers per MFI and the average amount disbursed per loan.

For 7 MFIs (UCEC/Z, Microfund for Women, Life Bank Foundation, AMC, ASUSU CIIGABA, JMCC/Tamweelcom and Constanta Foundation) the average outstanding loan portfolio is not available. For these MFIs I have used the loan portfolio measure, which is in general larger than the average outstanding loan portfolio.

For 2 MFIs (MCPEC and Fundamic) the average amount disbursed per loan is not available. For these MFIs I have used the average outstanding loan per client.

Year Number of MFIs included Total average outstanding loan portfolio (mlns USD) Average outstanding loan portfolio per MFI* (mlns USD) Total number of active borrowers** (1000s) Number of active borrowers per MFI Mean average amount disbursed per loan*** 2002 17 30 2.01 135 8459 508 2003 64 221 3.46 732 11614 797 2004 82 367 4.54 1,044 12727 798 2005 85 516 6.07 1,362 16215 899 2006 67 632 9.43 1,285 19773 869 2007 30 311 10.37 679 24231 770

Table 2: Overview of average outstanding loan portfolio and number of active borrowers per year

* in 2002 data is missing for 2 MFIs, in 2004 and 2006 data is missing for 1 MFI; ** in 2003 and 2006 data is

missing for 1 MFI;*** in 2002 data is missing for 2 MFIs, 2003 for 3 MFIs, 2004 for 4 MFIs, 2005 for 3 MFIs

The average outstanding loan portfolio per MFI increases for the MFIs in my sample between 2002 and 2007 from 2.01 million USD to 10.37 million USD.

The number of active borrowers served by the MFIs in my sample grows almost linearly between 2002 and 2007. Note that the total number of clients served by the MFIs in my sample is only a small fraction of the total number of clients served by all MFIs (Microcredit Summit 2007, table 1).

(11)

1.3 Subsamples

I will expand my analysis by assessing whether the allocation of subsidies and the effect that subsidies have on efficiency differs among different governance structures, among different regions, among different loan products and with access to an (inter)national network. I will include dummy variables for each characteristic, which enables me to estimate the regression for each subsample separately.

1.3.1 Governance structure

Following Hudon and Traca (2008a) I distinguish between for-profit organisations (which include (rural) banks and all non bank financial intermediaries), cooperatives and credit unions, non-profit organisations (NGOs), and other (when there is no clear information on the governance structure). Table 3 gives an overview of the number of MFIs in my sample per governance structure. I expect that for-profit MFIs receive the least subsidies and that most subsidies are directed to non-profit organizations as they are more focused on social impact.

Governance structure Number of MFIs

For-profit MFIs 8

Cooperative/credit unions 17

Non-profit MFIs 39

Other 22

Table 3: Overview of number of MFIs per governance structure

1.3.2 Regions

I distinguish between 8 different regions: Middle East and North Africa (MENA), Africa (AF), Central America (CAM), South America (SAM), South East Asia and the Pacific (EAP), Central Asia (CA), South Asia (SA) and Eastern Europe (EE). Table 4 shows the number of MFIs per region. Most of the MFIs in my research are located in Africa. Note that the subsamples for EAP, CA, SA and EE are too small to make general conclusions on these regions. I therefore do not estimate the models for the se regions.

Region Number of MFIs

Middle East and North Africa (MENA) 8

Africa (AF) 36

Central America (CAM) 15

South America (SAM) 15

South East Asia and the Pacific (EAP) 2

Central Asia (CA) 3

South Asia (SA) 3

Eastern Europe (EE) 4

(12)

1.3.3 Loan products

Group lending systems are widely used when offering financial services to the very poor, because it can replace physical collateral. The target group of clients may therefore differ between MFIs with different loan products. This may also have consequences for the willingness of donors to support the MFI and this could translate in differences in subsidy intensity and its relationship with other factors.

I have categorized the MFIs into MFIs that grant solely individual loans, MFIs that grant solely group loans, and MFIs that grant both individual and group loans. Group loans consist of solidarity group loans (three to eight members) and village bank group loans (15 to 20 or more members). In the solidarity group lending system members are jointly responsible for the repayment of loans even though they all borrow individually. In the village bank system, one group member receives a loan and this group member then grants loans to the other members. Both types of group lending allow clients to self-select groups, use dynamic incentives and have regular (often weekly) meetings (Lensink and Mersland 2009).

In my dataset 17 MFIs only offer individual loans to its clients, 9 MFIs only offer group loans and 60 MFIs offer both individual and group loans to their clients (see table 5).

Loan products Number of MFIs

Individual loans 17

Group loans 9

Both individual and group loans 60

Table 5: Overview of number of MFIs categorized by their loan activity

1.3.4 Access to an (inter)national network

I will also consider whether results are different for MFIs that have access to an (inter)national network and for those that have not. This information is recorded in the rating reports. In general I expect that MFIs that have access to a network will have a higher subsidy intensity. Only three MFIs in my sample do not have access to an (inter)national network (see table 6). This subsample is therefore too small to make general conclusions.

Number of MFIs

Access to an (inter)national network 83 No access to an (inter)national network 3

(13)

2 Subsidies

Subsidies are an important source of funding for MFIs. Due to inaccurate recording of financials and the limited amount of information available it is very difficult to calculate the exact amount of subsidies that an MFI has received.

In section 2.1 I will pay attention to different definitions of the total amount of subsidies used by other researchers. In section 2.2 I will define three subsidy components that I will use in my research. In section 2.3 I will construct four different subsidy measures with associated subsidy intensity measures (received subsidies relative to the total assets of the MFI).

2.1 Existing literature

Cull, Demirgüç-Kunt and Morduch (2008) use a non-commercial funding ratio (non-commercial funding divided by the total funds) in order to summarize the importance of non-commercial funding to the MFI. Non-commercial capital includes all capital that is not obtained at market prices. It includes donations, non-commercial borrowing (borrowing at concessional interest rates), and equity (sum of paid-in capital, reserves, and other equity accounts). Donations refer to donated equity from prior years, donations to subsidize financial services and an in-kind subsidy adjustment. Commercial funding on the other hand refers to capital that an MFI attracts at market rates. Total funds include commercial funding, non-commercial funding and deposits (both savings and time deposits).

Many researchers distinguish between direct subsidies and indirect subsidies when estimating the total amount of subsidized funds.

Direct subsidies include all money transfers from donors to the MFI (most of the times with a

predetermined goal of use). It includes for example donations to the loan capital of an MFI, donations to finance fixed assets, donations to help to pay for staff training and other operating or non-operating expenses (Armendariz and Morduch 2005).

Indirect subsidies refer to cost savings that are the result of special rules applied to the MFI.

Indirect subsidies are for example “soft loans”, tax advantages, loan guarantees, “soft equity”, and the assumption of exchange rate risk (Armendariz and Morduch 2005). The calculation of indirect subsidies is often problematic, as it demands very detailed information on the way an MFI finances its activities. This information is often not available.

(14)

between the interest paid by the MFI and the amount of interest it would have had to pay at the benchmark rate. The benchmark rate should account for country risks, political risks, exchange rate risks, and risks with respect to possible deterioration of the loan portfolio quality (Cull, Demirgüç-Kunt and Morduch 2008). Consequently, there is not one single benchmark rate appropriate to all MFIs (even when they are from the same country). A too low benchmark rate underestimates the amount of indirect subsidies and a too high benchmark rate overestimates the amount of indirect subsidies.

Many researchers use the rate that local banks pay on 90-day time deposits (CGAP 2003). Cull, Demirgüç-Kunt and Morduch (2008) argue that a country’s deposit rate is a relatively low benchmark rate, because expanding capital by collecting savings deposits is relatively cheap. Furthermore, using the deposit rate as the benchmark interest rate only makes sense for MFIs that collect savings deposits.

Cull, Demirgüç-Kunt and Morduch (2008) use the prime rate as an MFI’s alternative cost of capital. This is the price for capital between banks and their most trustworthy clients. They add two percent to the prime rate in order to account for the perceived extra risk of lending to MFIs. The addition of 2 percent is rather arbitrary and it assumes that all MFIs are the same and that similar interest rates apply to all MFIs. In reality, MFIs are seldom similar. I therefore do not choose to add a fixed interest rate margin, but to let the margin vary with the risk profile of the MFI (described in more detail section 2.2.3).

Yaron and Schreiner (1999) identify six different forms of subsidized funds and categorize them in equity grants and profit grants.

Equity grants include direct grants, paid-in capital and true profit. Equity grants increase net

worth but do not directly change the accounting profit of the year that the equity grants were received (they are not recorded through the income statement).

Direct grants include both cash gifts and in-kind gifts (computers, trucks and other things that

are recorded in the accounts). The amount of money a donor spends on in-kind gifts, training and technical assistance, may overestimate the amount of subsidies, because the MFI would not have spent a similar amount of money in the absence of the gift. It is for example not very likely that the MFI would have hired western workers at western wages for assistance in the absence of the gift.

(15)

True profit is the accounting profit less profit grants and it represents the MFI’s profit in the

absence of subsidies.

Profit grants include revenue grants, discounts on public debt and discounts on expenses. Profit

grants inflate revenues and/or deflate expenses. They increase the accounting profit of the year in which they are received. They also increase net worth by increasing the retained earnings at year-end.

Revenue grants are cash gifts and are recorded as revenues although they are not the result of the

operations of the business.

Discounts on public debt and discounts on expense are both non-cash gifts to the MFI. Discount

on expenses are for example technical assistance, free deposit insurance, coverage of organization costs, debt guarantees, consultant services, classes for loan officers, and travel for employees (Yaron and Schreiner 1999).

One dollar in every form of subsidy increases the net worth of the MFI by one dollar. To calculate the total subsidies exactly, dividends and taxes should also be included in the analysis. Yaron and Schreiner (1999) ignore these for simplicity.

Hudon and Traca (2008a; 2008b) define subsidy intensity as the ratio of donated equity and total equity. They argue that donated equity entails the main part of the subsidies received by MFIs.

2.2 Subsidy components

In my research I consider three components of subsidies; donated equity, paid-in capital, and indirect subsidies through discounts on borrowing expense. Due to data limitations I do not include in-kind donations that are not recorded in the financial statements, into my analysis.

2.2.1 Donated equity

Donated equity represents accumulated donations to an MFI. It is recorded in the balance sheet of the MFI. Donated equity is received through cash donations that carry no restrictions from sources that do not receive stocks. It includes donations for capital loans and/or the purchase of fixed assets (CGAP 2003).

2.2.2 Paid-in Capital

(16)

which do not act as private owners, it can be seen as an equity grant to the MFI. Paid-in capital should then be included in total subsidies.

2.2.3 Indirect subsidies: Discount on borrowing expenses

Many MFIs are able to attract capital at a cost lower than the market interest rate. The discount on borrowing expenses can be seen as an indirect subsidy. In order to quantify indirect subsidies, I will compare the interest rate that the MFI pays on borrowed capital with a benchmark interest rate. The benchmark rate represents the minimum interest rate that the MFI has to pay when attracting capital from the market. As already stated, it is very difficult to find an appropriate benchmark rate, because it should both account for country risks as well as for risks related to individual characteristics of an MFI (loan portfolio quality, management structure etc.). MFIs from countries with higher risk profiles and/or MFIs with more risky individual characteristics should pay a higher interest rate on its borrowings. Its interest rate should therefore be compared with a higher benchmark rate.

I will construct a benchmark rate by adding a margin, relative to the risk profile of the individual MFI, to a base interest rate.

Not all MFIs in my data set are deposit-taking institutions. Using a country’s deposit rate as the base rate is therefore not appropriate. Instead, I use the annual average 90-day interbank rate. The 90-day interbank rate is the interest rate that local banks charge each other on 90-day loans. Some central banks do not distinct between interbank rates with different maturity periods in their annual reports. In that case I use the general interbank rate. For some countries there is no data available on interbank rates. In those cases I used the most appropriate and available rate, that proxies the cost at which financial institutions can attract capital best (mostly after consultation with a representative of the country’s central bank).

(17)

the market interest rate. Appendix 3 gives an overview of the base rate and the source of the data per country.

For Tunisia and Guinea there is no interbank rate and no appropriate alternative rate available. I excluded the two MFIs from these countries from the analysis of indirect subsidies (Enda Inte Arabe from Tunisia and Crédit Rural de Guinée from Guinea).

0,0% 5,0% 10,0% 15,0% 20,0% 25,0% 30,0% 35,0% 40,0% 2002 2003 2004 2005 2006 2007 Years B as e ra te ( p er ce n ta g e)

Benin, Burkina Faso, Mali, Niger, Senegal and Togo

Bolivia

Bosnia-Hercegovina Brazil

Cambodia Cameroon and Chad Chile Dominican Republic Ecuador Egypt Ethiopia Georgia Ghana Guatemala Jordan Kenya Mexico Mongolia Graph 1: Base rate for different countries between 2002 and 2007

Graph 1 presents the base rate of each country included in my analysis between 2000 and 2007. Shown in the graph, the base rate differs widely among different countries. Therefore, contrary to Cull, Demirgüç-Kunt and Morduch (2008), I choose not to increase a country’s base rate with a fixed margin, but with a percentage margin that depends on the risk profile of the individual MFI. In my analysis the interest rate margin and the benchmark rate may therefore differ between MFIs from one country.

(18)

rating category. I linearly distributed the percentage increase over the different rating categories. I chose the value of the percentage increases in such a way that the base rate increases on average by two percent in line with the two-percent margin of Cull, Demirgüç-Kunt and Morduch (2008). The average base rate for the MFIs in my sample is 8.4 percent. The benchmark rate is on average 126 percent of the base rate. This implies an absolute increase of 2.2 percent of the base rate to an average benchmark rate of 10.6 percent.

Rating Benchmark rate as a percentage of the base rate

A+ 107% A 110% A- 113.5% B+ 117% B 120% B- 123.5% C+ 127% C 130% C- 133.5% D+ 137% D 140% D- 145% E 150%

Table 7: Rating and the benchmark interest rate margin

(19)

2.3 Subsidy measures

As described above, paid-in capital should only be included in the subsidy measure when it comes from public sources that do not act as private owners. Unfortunately the source of the paid-in capital is not recorded in the rating reports. It is therefore not entirely sure that it should be included in the subsidy measure. I estimate the amount of indirect subsidy according to the methodology described in section 2.2.3. This estimated amount is based on some (strong) assumptions and it is therefore not certain that the amount is entirely accurate. I therefore construct four different subsidy measures (S1, S2, S3, S4) that are different combinations of the subsidy components. All measures include donated equity. I chose not to construct a subsidy measure that excludes donated equity (i.e. only includes paid-in capital and indirect subsidies), because it is certain that donated equity should be included in the amount of subsidies received. S1 only includes donated equity from the balance sheets (following Hudon 2007; Hudon and Traca 2008a; 2008b).

Following Schreiner and Yaron (1999) I added all paid-in capital to donated equity in the second subsidy measure (S2).

The third measure of subsidies (S3) includes donated equity and an estimation of the received indirect subsidy through a discount on the lending expense (soft loans).

Finally, the fourth subsidy measure (S4) includes donated equity, the paid-in capital and the discount on borrowing expenses. Table 8 gives an overview of the different subsidy measures.

Explanation Formula

S1 Donated equity (DE) S1=DE

S2 Donated equity (DE) + paid-in capital (PIC) S2=DE+PIC S3 Donated equity (DE) + discount on borrowing

expense (IS)

S3=DE+IS S4 Donated equity (DE) + paid-in capital (PIC) +

discount on borrowing expense (IS)

S4=DE+PIC+IS

Table 8: Overview of the different subsidy measures

(20)

Subsidy intensity measure Formula SI1 SI1=(DE/TAS SI2 SI2=(DE+PIC)/TAS SI3 SI3=(DE+IS)/TAS SI4 SI4=(DE+PIC+IS)/TAS

(21)

3 The amount of received subsidies

In section 2.2 I identified three different components of subsidy (donated equity, paid- in capital and indirect subsidy though a discount on the borrowing expense). In this section I will pay attention to the amount of subsidies received by the MFIs in my sample and the relative importance of the different subsidy components. I will also analyze whether differences in governance structure, region, loan products and access to an (inter)national network are associated with differences in subsidy intensity.

3.1 The absolute amount of subsidies

Graph 2 represents the mean donated equity, paid-in capital and indirect subsidies between 2002 and 2007. Most MFIs in my sample received the largest amount of received subsidies through donated equity.

Graph 2: Amount of subsidies per subsidy component

(22)

Mean donated equity Mean paid-in capital Mean indirect subsidies Mlns USD Number of observations Mlns USD Number of observations Mlns USD Number of observations 2002 0.44 16 0.33 16 0.055 16 2003 1.40 65 0.28 65 0.052 62 2004 1.34 81 0.39 81 0.11 79 2005 1.47 85 0.49 85 0.13 83 2006 1.69 66 0.87 66 0.22 65 2007 2.75 31 1.42 31 0.40 31 All 1.54 344 0.58 344 0.15 336

Table 10: Overview of amount of subsidies and number of observations

3.2 Subsidy intensity of subsidy components

Table 11 represents the mean subsidy intensity of the donated equity, paid-in capital and indirect subsidies and the number of observations.

Mean

DE/TAS Number of observations Mean PIC/TAS Number of observations Mean IS/TAS Number of observations

2002 0.327 16 0.083 16 0.014 16 2003 0.342 65 0.059 65 0.011 62 2004 0.300 81 0.055 81 0.018 79 2005 0.266 85 0.071 84 0.012 83 2006 0.243 66 0.078 66 0.015 65 2007 0.331 31 0.105 31 0.026 31 All 0.292 344 0.070 343 0.015 336

Table 11: Overview of the mean subsidy intensity per subsidy component and the number of observations

The mean of the donated equity relative to total assets is 0.29 between 2002 and 2007. Four MFIs (Share an Opportunity MFI (U) Ltd, Hofokam and UGAFODE from Uganda and Enda Inte-Arabe from Tunisia) have subsidy intensities of donated equity higher than one. These MFIs have large negative retained earnings without donations and reserves from previous years on their balance sheets. This represents funds that have been used to cover yearly operating losses. Total equity and total assets are negatively influenced by this account, but donated equity is not. With large enough previous losses, donated equity exceeds total assets and the subsidy intensity of donated equity becomes greater than one.

(23)

The mean of indirect subsidy intensity is 0,015 between 2002 and 2007. This means that on average the MFIs in my sample receive less than 2 percent of their total assets through indirect subsidies. ECLOF RD from Dominican Republic has a high indirect subsidy intensity. This is mainly caused by a high average yield on interbank loans (36.1 percent) in 2004 in the Dominican Republic.

As can be seen in table 11, the mean of paid-in capital subsidy intensity and indirect subsidies intensity is constant over the entire period. Between 2003 and 2006 donated equity relative to total assets tends to decrease from 0.34 to 0.24. Between 2006 and 2007 it increases again. Table 11 confirms that donated equity is on average the most important subsidy component for the MFIs in my sample throughout the entire period. This corresponds with the findings of Hudon and Traca (2008a; 2008b).

Discounts on borrowing expenses seem to be the least important source of subsidies. This may be a reflection of the trend that commercial sources of funding have become increasingly important with respect to the global microfinance loan portfolio (MicroBanking Bulletin 2008). Table 12 represents the correlations between the different subsidy intensity components. The different subsidy intensity components are weakly negatively correlated.

DE/TAS PIC/TAC IS/TAS

DE/TAS 1 -0.301 -0.079

PIC/TAS -0.301 1 -0.069

IS/TAS -0.079 -0.069 1

(24)

3.3 Differences in subsidy intensity

In this section I will assess whether there are differences in the subsidy intensity associated with differences in governance structure, regions, type of products, and access to an (inter)national network.

3.3.1 Does governance structure matter?

Table 13 represents the mean of the subsidy intensity of donated equity, paid-in capital and indirect subsidies per governance structure. In line with my expectations, non-profits have higher subsidy intensities than for-profits and cooperatives/credit unions. For-profits show high subsidy intensities associated with paid-in capital, whereas this is small for non-profits. Non-profits, on the other hand, receive most subsidies through donated equity.

For profit (n=33) Cooperative/ credit union (n=65) Non-profit (n=162) Other (n=74) Mean (DE/TAS) 0.13 0.09 0.39 0.35 Mean (PIC/TAS) 0.17 0.12 0.03 0.08 Mean (IS/TAS) 0.01 0.007 0.019 (n=157) 0.017 (n=71)

Table 13: Subsidy intensity of different subsidy components per governance structure

(25)

3.3.2 Does region matter?

Table 14 represents the mean intensities from the different subsidy components per region. The amount of subsidy intensity and the form in which subsidies are received is quite different per region. As can be seen, the subsidy intensity of donated equity is especially high in MENA (0.55), South Asia (0.43), Africa (0.34) and Eastern Europe (0.3). The subsidy intensity of paid-in capital is important paid-in Central America (0.19). Indirect subsidy paid-intensities are generally very small compared to the subsidy intensity of donated equity and of paid-in capital. Eastern Europe and Central America receive the highest indirect subsidy intensities (respectively 0.032 and 0.04).

MENA

(n=36) AF (n=130) CAM (n=37) SAM (n=82) EAP (n=8) CA (n=11) SA (n=11) EE (n=16)

Mean (DE/TAS) 0.55 0.34 0.12 0.23 0.09 0.008 0.43 0.30 Mean (PIC/TAS) 0.06 0.044 0.19 0.07 - 0.069 0.053 0.00003 Mean (IS/TAS) 0.01 (n=31) 0.0097 (n=127) 0.04 0.009 0.022 0.008 0.018 0.032

Table 14: Subsidy intensity of different subsidy components per region

Appendix 5 gives an overview of the subsidy intensities of the different subsidy components per region between 2002 and 2007. The donated equity subsidy intensity for Africa shows an upward sloping curve throughout the entire period (DE/TAS increases from 0.1 to 0.5 between 2002 and 2007).

3.3.3 Does the type of products matter?

As can be seen in table 15, MFIs that offer group loans (as their sole product, or in combination with individual loans) receive on average more subsidies, which is mainly caused by a high subsidy intensity of donated equity. MFIs that only grant individual loans receive relatively a lot of paid-in capital (subsidy intensity of paid-in capital of 0.34).

Individual loans (n=64)

Group loans (n=38)

Both group loans and individual loans (n=234)

Mean (DE/TAS) 0.18 0.34 0.32

Mean (PIC/TAS) 0.12 0.03 0.06

Mean (IS/TAS) 0.008 0.012 0.018 (n=226)

(26)

3.3.4 Does access to an (inter)national network matter?

It is very interesting to see that the sources and the amount of subsidies differ a lot between MFIs that have access to an (inter)national network and those that have not (table 16). The subsidy intensity of donated equity is higher for MFIs that have access to an (inter)national network (0.30 compared with 0.15). The subsidy intensity of paid-in capital, on the other hand, is much higher for MFIs that do not have access to a network (0.28 compared with 0.06). The indirect subsidy intensity is small for both categories. In total MFIs that are not engaged in a (inter)national network receive on average more subsidies than MFIs that have access to an international network. This result is rather surprising. Note that the sample of MFIs that do not have access to an (inter)national network is very small (only three MFIs) and it is therefore not possible to make general conclusions about the amount of subsidies received by these MFIs. It would be interesting to gather more data of MFIs that do not have access to an (inter)national network in order to assess whether my results also hold for these MFIs.

No international network

(n=11) International network (n=323)

Mean (DE/TAS) 0.15 0.30

Mean (PIC/TAS) 0.28 0.06

Mean (IS/TAS) 0.019 0.015 (n=315)

(27)

3.4 Subsidy intensity of subsidy measures

In section 2.3 I identified four subsidy measures (S1, S2, S3 and S4) with associated subsidy intensity measures (SI1, SI2, SI3 and SI4).

Graph 4 shows the evolution of the four subsidy intensity measures. The means of SI1, SI2, SI3 and SI4 show a similar pattern, which is caused by a decrease in donated equity between 2003 and 2006. When indirect subsidies are included in the subsidy intensity measures (SI3 and SI4), the mean intensity become lower in 2003. This is caused by the fact that the two MFIs that are excluded from these measures have a very high donated equity subsidy intensity.

Graph 3: Evolution of the mean of the different subsidy measures, SI1=DE/TAS, SI2=(DE+PIC)/TAS, SI3=(DE+IS)/TAS, SI4=(DE+PIC+IS)/TAS

(28)

all paid-in capital in the subsidy intensity measure (i.e. it is not sure that all paid-in capital comes from public sources).

SI1 SI2 SI3 SI4

SI1 1 0.903 0.995 0.901

SI2 0.903 1 0.894 0.995

SI3 0.995 0.894 1 0.902

SI4 0.901 0.995 0.902 1

(29)

4 Methodology

In section 5 I will analyze the relationship between subsidy intensity and the MFI’s age, the MFI’s sustainability index and the degree of social focus (defined by the poverty level of the MFI’s clients measured by the relative loan size). In section 6 I will focus on the effect that subsidies have on efficiency.

For both analyses I will estimate an OLS regression model using panel data. This data has both a cross-section dimension (86 different MFIs) and a time dimension (three to four observations per cross-section between 2002 and 2007). In both models I will check for multicollinearity, I will control for heteroskedasticity, and I will assess whether the regression should be estimated with common effects, fixed effects, or random effects. In this section I will explain these phenomena and the associated tests. The exact regression equations are specified in section 5 and 6.

4.1 Multicollinearity

Multicollinearity exists when an explanatory variable is a linear combination of the other explanatory variables that are included in the model. When multicollinearity exists, the coefficients from the regression cannot be interpreted individually. The output results are very sensitive to exclusion of one of the explanatory variables (Carter Hill, Griffiths and Judge 2001). I will check whether multicollinearity exists in my models by examining the correlation matrix of the explanatory variables. When correlations are high, multicollinearity may be problematic. Note that I only evaluate pairwise correlations between explanatory variables and that I ignore collinearity relationships that involve more than two of the explanatory variables.

4.2 Heteroskedasticity

(30)

4.3 Different intercepts

A common effects OLS regression assumes that all cross-sections have a similar intercept. It may be possible that each cross-section has a similar slope but a different intercept. Without controlling for different intercepts, the OLS regression may be biased. There are two possibilities to model different intercept and to improve the estimation results; a fixed effects model or a random effects model.

The fixed effects model includes a dummy variable for each section and allows the cross-sections to have different intercepts. A disadvantage of the fixed effects method is that you loose degrees of freedom and explanatory variables that do not vary within the cross-section unit (picked up by intercepts).

The random effects model assumes that different intercepts are random drawings from a set of possible intercepts. The fixed effects and the random effects estimator are the same, but the random effects model uses a different transformation of the data, which saves us degrees of freedom and is therefore more efficient. However, it is not always appropriate to use the random effects model. The random effects estimator is biased when the intercepts are not random and when the error term is correlated with x.

I will assess whether fixed effects should be included in my model by the redundant fixed effects test and whether it is allowed to use random effects by the Hausman test.

4.3.1 Redundant fixed effects test

The redundant fixed effects test examines whether it is appropriate to estimate the regression with fixed effects or with common effects, using the following hypotheses:

H0: Fixed effects are redundant (the regression should be estimated with common effects) H1: Fixed effects are appropriate

4.3.2 Hausman test

The Hausman test examines whether the random effects model or the fixed effects model is more appropriate to the estimated regression model, using the following hypotheses:

H0: Random effects would be consistent and efficient (fixed effects and random effects yield similar coefficients)

(31)

5 The allocation of subsidies

In this section I analyze whether the MFI’s age, the MFI’s sustainability index and the poverty level of the clients that it serves, have influence on the amount of subsidies that an MFI receives. In section 5.1 to 5.3 I pay attention to the relationship between subsidy intensity and all three explanatory variables separately. In section 5.4 I present an OLS panel regression model in order to analyze whether the amount of received subsidies is significantly determined by the MFI’s age, the MFI’s sustainability index and the poverty level of the clients that it serves.

5.1 Subsidies and the age of MFIs

The most widely accepted form of subsidization is a subsidy granted to a start-up MFI for a limited period of time in order to cover extra costs related to the beginning of operations. These temporary subsidies are very important, as only few MFIs have been able to start operations without subsidies (Hudon 2007). I section 5.1.1 I pay attention to some theoretical aspects with respect to temporary subsidies to young MFIs. In section 5.1.2 I analyze the relationship between the subsidy intensity and the age of MFIs in my sample by means of scatter graphs.

5.1.1 Theoretical considerations

The justification of temporary start-up subsidies relies on the assumption that average costs of an MFI fall over time. Temporary subsidies enable the MFI to set interest rates at the long-term sustainable level.

Assume that the long-term sustainable interest rate is represented by r* and that this rate covers costs from t* onwards. Before t* costs are higher and revenues fall short of costs when charging r*. In the absence of subsidies the MFI needs to set interest rates at r0 before t* to cover all costs

(r0 > r*). A subsidy that equals all extra start-up costs (all costs greater than r*) enables the MFI

to charge the long-term sustainable interest rate from the beginning of operations (Armendariz and Morduch 2005). After time t* the MFI can continue to charge r* without needing any subsidies. The temporary subsidy gives continuity to the organization and prevents clients from bearing extra costs in the start-up phase.

(32)

which this rate can be implemented. Designing the optimal donors’ exit strategy is therefore also very difficult (Morduch 1999; Armendariz and Morduch 2005).

Short-term subsidies may also be used to temporarily lower the interest rate charged to (new) clients, who are not (yet) ready to borrow from micro lenders at normal rates. These clients may first, for example, need certain trainings, or perhaps time to build a business that reaches a certain minimum scale. At a certain point in time they will be able to borrow at normal rates. It is very important that clients are motivated and encouraged to develop themselves to the group of normal clients (Armendariz and Morduch 2005; Morduch 2005).

(33)

5.1.2 Empirics

According to the theoretical considerations in the previous section, young MFIs should receive more subsidies and the subsidy intensity should diminish, as the MFI gets older. In summary, when subsidies are used as a temporary measure and donors stick to a well-designed exit strategy, there should exist a negative relationship between the subsidy intensity and the age of the MFIs.

I define the MFI’s age as the number of years after the beginning of operations. The average age of the MFIs in my sample is 10 years. Some MFIs were established by other MFIs. For these MFIs I have ignored the possibility of lower start-up costs resulting from knowledge and experience of the mother company. Scatter graphs 8 and 9 show the relationship between subsidy intensity (respectively SI1 and SI4) and the age of the MFIs in my sample between 2002 and 2007.

The graphs suggest that there exists a very weak negative relationship between age and the subsidy intensity (SI1 and SI4) for the MFIs in my sample. This could support the theory that subsidies are directed to younger MFIs in order to help them cover start-up costs. The evidence is however not very convincing.

Graph 4: Scatter graph of subsidy intensity (represented by SI1) and the age of the MFIs

(34)

5.2 Subsidies and sustainability

In this section I analyze the relationship between subsidy intensity and the sustainability index. In section 5.2.1 I discuss some theoretical considerations and in 5.2.2 I assess the relationship for the MFIs in my sample by means of scatter graphs.

5.2.1 Theoretical considerations

Sustainability is an important priority within the microfinance sector. It refers to the ability to cover costs by revenues. In general there are two measures of sustainability: operating sustainability and financial sustainability.

Operating sustainability refers to the MFI’s ability to cover operating costs by revenues. Operating costs consist of administrative costs (including rental costs, utilities, travel costs, depreciation) and personnel costs. When a MFI is not operationally self-sustainable, its capital holdings will be depleted over time (Morduch 1999).

Financial sustainability refers to the MFI’s ability to cover operating and financial costs by revenues. Financial costs are the costs of attracting capital. It includes interest payments on borrowings and on deposits (Hudon and Traca 2008a). Note that this representation of financial costs leaves out a cost of capital to equity. Being financially self-sustainable means that the MFI does not need subsidies to cover operating and financial costs. This means that the continuity of operations is ensured and activities could possibly be expanded without donor funds (Rosenberg et al. 2009). Cull, Demirgüç-Kunt and Morduch (2008) argue that independency from donor funds is crucial for the expansion of microfinance activities in the future.

Despite the increased attention to sustainability, only few MFIs have reached independence from donors’ funds. Hudon (2007) reports that only 100 out of 10,000 MFIs have achieved financial sustainability. From the MIX Market database, less than 300 MFIs have achieved operational self-sufficiency in 2004.

Note that not being financially self-sustainable is not a problem in itself. As long as microfinance activities outperform other investment possibilities in terms of social performance and as long as donors remain committed, there is no reason why subsidization of microfinance activities should not continue (Morduch 1999).

(35)

funds should be allocated to institutions that are not (yet) self-sustainable, while donors continuously motivate the MFI to become sustainable through increased efficiency and innovative practices.

Some donor support is allocated to already sustainable institutions, instead of to more risky MFIs. Hudon (2007) argues that the current concentration of donor funds is not optimal. He finds that nearly 90 percent of the donor money was directed to (often large) regulated specialized institutions in transition economies.

5.2.2 Empirics

PlanetRating defines operating self-sufficiency by the revenue from operations divided by the sum of operating expense, the loan loss expense and financial expense. Financial self-sufficiency includes several adjustments in the amount of total expenses. Adjustments include adjustments for the cost of funds, for inflation, for in-kind donations, for provisions and other adjustments (rating reports of PlanetRating).

In my research I use a sustainability index (SUS) that is almost similar to the operational self-sufficiency index of PlanetRating. The sustainability index is represented by the following formula:

SUS it = rLit +FRIit +Oit

. . PEit + Ait + LLPit + ßBit + ∂Dit + OFit

The nominator represents total revenues. It includes financial revenues from loan portfolio (rLit), financial revenues from investments (FRIit) and operating revenues (Oit).

The denominator represents total costs, which includes operating costs and financial expenses. Operating costs include personnel costs (PEit), administrative costs (Ait) and the net loan loss

provision expense and write-off (LLPit).

Financial expenses include interest paid on borrowed capital (ßBit, where ß is the interest rate

on borrowings and Bit is the total amount of borrowed capital), interest paid on deposits (∂Dit,

where ∂ is the interest rate paid on deposits and Dit is the total amount of deposits) and other

financial expenses (OFit). Note that I do not explicitly incorporate a cost of capital to equity.

(36)

Graph 6 shows the mean sustainability index of the MFIs in my sample. As can be seen, the sustainability index is very high. This strongly suggests that my sample includes relatively rich MFIs. My sample is therefore not representative for the microfinance sector in general.

Graph 6: Mean sustainability index for the MFIs in my sample

Graph 7 and 8 represent the relationship between the sustainability index and SI1 and SI4. The graphs do not indicate a very strong relationship between the sustainability index and subsidy intensity.

(37)

5.3 Subsidies and poverty focus

An important argument justifying the ongoing subsidization of MFIs is that the microfinance sector cannot exist without subsidy and that subsidies are inherent to the mission and targeted population of the microfinance sector (Hudon 2007).

Offering financial services to very poor clients is in general more costly as it involves smaller loans to people living in rural areas that are more challenging to reach. High (transaction) costs make it difficult for MFIs to charge interest rates that are both sustainable and affordable to clients. Many researchers argue that there exists a trade-off between the ability to reach the poor and the ability to perform financially well.

Continuous subsidization can be justified if the extra costs associated with serving the targeted clients outweigh the possibility of increasing interest rates. Only when financially self-sustainable programs cannot serve a particular group of poor, one may justify the use of donor funds to serve this group (Morduch 1999).

In section 5.3.1 I focus on the relationship between the sustainability index and the poverty focus of MFIs and the existence of a trade-off. In section 5.3.2 I will pay attention to the relationship between the subsidy intensity and the poverty level of the MFI’s clients.

In general there is little information available on the exact poverty level of clients. However, average loan size, the fraction of women borrowers and the fraction of clients living in rural areas correlate with each other and with self-reported measures of household poverty (Cull, Demirgüç-Kunt and Morduch 2008). In other words, poorer people tend to opt for smaller loans and a loan size measure can be used as a proxy for the poverty level of the clients.

I use the amount disbursed per loan (LS) relative to GNP per capita (GNP) as a proxy for the poverty level of the MFI’s clients. This relative loan size measure (RLS) corrects for differences in living standards between countries. A high RLS suggests that the MFI serves richer clients. A low RLS indicate that the MFI serves poorer clients and refers therefore to a higher poverty focus.

(38)

5.3.1 Relationship between sustainability index and poverty focus

According to Armendariz and Morduch (2005) microfinance is characterized by having multiple objectives, namely achieving financial self-sufficiency and reducing poverty. This causes difficulties when making decisions (multitask problem). In theory the two objectives should be complementary and self-reinforcing, because expanding scale and pursuing profits enable institutions to reach more low-income people. In practice, these two objectives often conflict. Rhyne (1998) describes having a social goal and a financial goal as the problem of dual maximization in mathematics. Unfortunately, only one of the variables can be maximized, while treating the other as a constraint. When going along the curve that represents the trade-off between the two objectives, a better performance in terms of goal A means a worse performance in terms of goal B. However, inside the curve there is a possibility to improve performance in terms of both goal A and goal B.

The main argument supporting a trade-off between poverty focus and sustainability is that costs per dollar lent are higher when granting small loans to poor clients in combination with the fact that the MFI cannot raise interest rates to compensate for the higher costs (clients cannot afford higher interest rates).

The cost of funds and of loan loss varies proportionally to the amount lent. Administrative costs (including screening and monitoring costs) are for a large part fixed per transaction and the administrative costs per dollar lent diminish therefore with loan size. Assume for example that a 50,000 USD loan requires an administrative cost of 1,500 USD (3 percent). The same amount of total funds (50,000 USD) could also be granted in 200 loans of 250 USD. A 3 percent administrative costs means 7.5 USD per loan. In practice, the administrative costs per 250 USD loan will be much higher. Managing a portfolio of many small loans is more expensive than managing a portfolio that only consists of a few large loans (Armendariz and Morduch 2005; CGAP).

(39)

When the poor clients cannot afford these rates, subsidies may be necessary to ensure outreach (Armendariz and Morduch 2005). Note that in absence of competition MFIs could also try to cover the costs associated with the smaller loans by charging slightly higher interest rates on larger loans (cross-subsidization). However, when competitors steal away the customers that opt for larger loans by offering a lower interest rate, cross-subsidization is not possible.

Mosley (1996) finds that the financial performance of an MFI improves with loan size as the organization can benefit from economies of scale. MFIs that focus on the poorest can only be independent from donor money when they work in areas with for example a high population density, when they have access to a cheap staff or when they charge very high interest rates that put their social mission into question. Hudon (2007) argues that more than 90 percent of all MFIs are not able to achieve independence from donor funds without putting their social mission into question.

Rhyne (1998) does not find a correlation between the poverty level of clients (measured by loan size) and the financial viability of well-performing institutions. Even though it is perhaps more difficult to reach clients that demand smaller loans and that live in more rural areas, these MFIs developed efficient delivery methods resulting in financially viable institutions.

(40)

Graph 9: Scatter graph of the relative loan size (RLS) and the sustainability index

Graph 10: Scatter graph of the relative loan size (RLS) and the sustainability index with RLS<2

5.3.2 Subsidy intensity and the focus on the poor

As described in the previous section, MFIs that focus on the lower segment may have a weaker financial situation. Ideally, MFIs seek sustainability through a reduction in administrative costs through higher staff productivity, a decline in rent-seeking or other innovations that lead to cost reductions (managerial and/or technological innovations) (Hudon and Traca 2008a). There is a fear that MFIs will sacrifice their social mission if subsidies are reduced sharply (Cull, Demirgüç-Kunt and Morduch 2008). Therefore most practitioners agree that MFIs that serve difficult and/or costly to reach populations may benefit from long-term subsidies in order to prevent them from switching to richer clients.

(41)

subsidized as the programs that focus on the “low end” clients (the poorest of the poor). Furthermore, the financial performance of MFIs that grant individual loans is comparable to the financial performance of MFIs that use group lending, even though the clients of the former are twice as rich.

Graphs 11 and 12 represent the relationship between the subsidy indices and the relative loan size for the MFIs in my sample. The graphs provide some evidence of a negative relationship between subsidy intensity and the relative loan size, supporting the theory that subsidies are directed to MFIs that focus to a greater extent on the poor. The evidence is however not very convincing.

Graph 11: Relationship between SI1 (SI1=DE/TAS) and the relative loan size (RLS=LS/GNP)

Graph 12: Relationship between SI4

(42)

5.4 Regression model

5.4.1 Methodology

I assess whether subsidy intensity, measured by SI1 (only donated equity included) and by SI4 (donated equity, paid-in capital and indirect subsidies included) is influenced by the MFI’s age, the MFI’s sustainability index and the poverty level of the MFI’s clients by estimating the following regression models:

SI1it=β0+ β1AGE it + β2 SUS it + β3RLSit +εi (1)

SI4it=β0+ β1AGE it + β2 SUS it + β3RLSit +εi (2)

Based on sections 5.1 to 5.3 I expect that the subsidy intensity is decreasing in all three explanatory variables (i.e. older MFIs receive less subsidies, MFIs with a higher sustainability index receive less subsidies and MFIs that focus to a greater extent on the poor (i.e. grant smaller loans) have higher subsidy intensities).

Table 18 presents the correlation matrix for the explanatory variables in the above regressions. As can be seen the correlations are very low, which indicates that multicollinearity is not problematic in my models.

RLS AGE SUS

RLS 1 -0.100 0.011

AGE -0.100 1 0.015

SUS 0.011 0.015 1

Table 18: Correlation matrix for the explanatory variables

I analyze whether common effects, fixed effects or random effects are appropriate by conducting the Redundant fixed effects test and the Hausman test described in section 4. For almost all the estimated regressions it is appropriate to estimate the model with fixed effects.

(43)

5.4.2 Results and conclusions SI1

The regression output in this section includes fixed effects (unless stated otherwise), and I also correct for heteroskedasticity by including the White (diagonal) coefficient covariance method. The estimation result for the entire sample (including 310 observations from 83 cross-sections, with the standard errors between brackets) is:

SI1it=0.761 - 0.035AGE it - 0.016 SUS it - 0.154 RLSit +εi [0.069] [0.007] [0.048] [0.084]

Age and has a negative influence on the sustainability intensity at a 1% significance level and the relative loan size on a 10% significance level. The negative influence of the sustainability index is not significant.

Governance structure

Table 19 represents the regression results for for-profits, credit unions and cooperatives, non-profits and MFIs with other governance structures separately. As can be seen there are some differences between the regression results. Non-profits and MFIs with other governance structures have the most significant negative influence of the age, sustainability index and the relative loan size on subsidy intensity.

β0 β1AGE i β2 SUS i β3RLSi R-squared Number of observation s(cross sections included) For-profits 0.609** [0.247] -0.060** [0.026] 0.140 [0.137] -0.162 [0.399] 0.874 31 (8) Credit unions/ cooperatives 0.219** [0.086] -0.004 [0.004] -0.103** [0.042] 0.106*** [0.054] 0.977 57 (16) Non-profits 0.853* [0.106] -0.037* [0.009] -0.002 [0.043] -0.198*** [0.115] 0.919 146 (38) Other 0.920 * [0.108] -0.036* [0.009] -0.137*** [0.072] -0.074*** [0.038] 0.973 73 (20)

Table 19: Output results for SI1it=β0+ β1AGE it + β2 SUS it + β3RLSit +εi (fixed effects and White diagonal

(44)

Regions

Table 20 represents the regression results per region. I have excluded East Asia and Pacific, Central Asia, South Asia and Eastern Europe from the analysis as the number of observations and cross sections included are too little. In three out of four of the remaining regions the MFI’s age has a significant negative influence on SI1.

In MENA the relative loan size also has a negative influence on subsidy intensity. Striking to see is that in this region the sustainability index has a positive influence on the subsidy intensity, suggesting that MFIs that are more sustainable receive more subsidies. As can be seen in section 3.3.2 the MFIs in MENA receive a relatively large amount of subsidies. A possible reason could be a weak exit strategy of donors, meaning that when an MFI performs well after having received donations, donors tend to reward the MFI for its accomplishments by supplying more funds instead of granting the donations to other MFIs that are less sustainable.

β0 β1AGE i β2 SUS i β3RLSi

R-squared Number of observation s (cross sections included) MENA 1.204* [0.212] -0.0931* [0.023] 0.124*** [0.066] -0.306** [0.098] 0.907 30 (70) Africa 0.811* [0.099] -0.021* [0.007] -0.134* [0.049] -0.159 [0.099] 0.967 119 (35) Central America 0.531** [0.227] -0.031 [0.020] -0.020 [0.056] -0.093 [0.205] 0.674 37 (9) South America 0.584* [0.154] -0.015*** [0.009] -0.098 [0.106] 0.016 [0.093] 0.964 75 (20)

Table 20: Output results for SI1it=β0+ β1AGE it + β2 SUS it + β3RLSit +εi (fixed effects and White diagonal

SUBSIDIES IN THE MICROFINANCE SECTOR Their composition, allocation and effect on efficiency

Master thesis