Micro-financing institutions as a portfolio diversification opportunity : how independent are micro-financing institutions from macroeconomic shocks?

(1)

Master in Science Business Economics, Finance Track

Master Thesis

Micro-financing Institutions as a portfolio diversification opportunity. How

independent are Micro-financing Institutions from macroeconomic shocks?

Handed in by: Jonathan Sieger 27.06.2016

Supervisor: Mr. Dr. Rafael Perez Ribas

(2)

Statement of originality

Amsterdam 27.06.2016

This document is written by Jonathan Sieger who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

I want to thank Mr. Dr. Ribas for his supervision and helpful comments to make the thesis happen. The thesis is dedicated to my Family, Julia, Laura & Oliver, Angelika and especially the support of

my Father Caspar Sieger

(4)

Abstract:

The Microfinancing market achieved continuous growth in the last decades. With its growth

it got enhanced attention by investors. One popular argument for investing in Microfinancing

Institutions (MFIs) is that their financial performance is rather detached from macroeconomic

movements. The paper analyses this relationship by examining 782 MFIs in a twenty year

time horizon. Three key insights emerge with the results: First MFIs are to a similar extent

affected by changes in domestic economy than Commercial Banks, but are less affected than

their entire domestic financial market. Secondly the influence of internationa l market

movements and emerging markets movements are non-existent or respectively slightly

negative. Thirdly higher exposure to market risk is rewarded with higher returns. For

investors the results suggest that MFIs tend to be an attractive diversification opportunity if

an international or emerging market portfolio is hold. For domestic investors it is

questionable if MFIs are indeed an improvement for diversification purposes as long as the

investment is not restricted solely to the domestic financial market.

(5)

1. Introduction

The Microfinancing movement has steadily increased for the last decades and served until 2011 more than 195 million clients worldwide. With its growth it has gotten enhanced attention not only by the developing community but as well by investors (Ahlin & Lin, 2006). MFIs are often considered as an interesting diversification opportunity for investors, because their financial performance tends to be rather detached from common market movements and macroeconomic factors. However there is no clear evidence for this hypothesis in current academic literature.

For the last decade the number of investment funds which are targeting their dollar towards MFIs reached a total investment of more than 9 billion USD and is estimated to continue to grow (Dijk-de Groot & Nijhof, 2015). This augments the necessity of a fundamental understanding of their exposure to macroeconomic shocks.

The purpose of this master thesis is therefore to receive a better insight of how volatile the financial performance of MFIs is to changes in its external environment. By doing so the paper contributes to the understanding of how suitable it is to add MFIs as a diversification asset in a market portfolio to reduce systematic risk.

Micro-financing was initially invented to build a sustainable cornerstone for alleviating poverty (Armendáriz & Morduch, 2010). 30 years after the first micro-credits were issued its founder Muhammed Yunus received the Nobel Peace Price in 2006 for his achievements of reducing poverty through Micro-financing.

From an academic perspective this achievements are subject of a long lasting and controversial discussion. Proponents and opponents of MFIs find contrary empirical results, especially regarding the question if the poverty alleviation is long-term sustainable (Dorfleitner & Oswald, 2016). The MFI approach does not solely promise to reduce poverty, but to simultaneously generate financial returns. This latter aspect of the so called double-bottom line is subject of a growing number of academic publications in Finance.

The major focus lies on the institutional-specific structures and techniques leading to success of MFIs, mainly the organizational structure, lending strategies and contract innovations (Hadi & Kamaluddin, 2015). This focus seems justifiable because institution-specific factors are easily controllable and conveniently applicable for founders of new MFIs, advisers of existing institutions or investors.

In contrast little is known about the influence of external factors on the financial performance. Among the few papers examining this relationship are Krauss and Walter (2006), Ahlin and Lin (2006), Gonzales (2007) and Galema, Lensink & Spierdijk (2011) and the thesis builds up mainly on the considerations of these papers. Though the four studies are the most comprehensive in their field, they receive different results regarding the influence of external factors. The first two find a significant relationship of the macroeconomic context on MFIs' performance, the two others do not.

(7)

2 One explanation for the varying results might be the use of different performance measures and different sample sizes. The thesis explicitly encounters this problem by using different performance measures and an improved dataset.

The data on MFIs are received from MixMarket, the biggest provider of MFI’s investment and information flows. Out of the approximately 10.000 MFIs worldwide Mixmarket contains more than a quarter, namely 2752. The dataset comprises annual data from 1995 until 2015. Even though Mixmarket ensures a certain quality of its data, only MFIs with the two highest reporting rankings are included in the analysis to improve the reliability of the results. Furthermore solely MFIs which occur for three consecutive years in the dataset are included. The final dataset consists of 782 MFIs based in 61 emerging countries.

A major obstacle of measuring MFIs financial performance is the abstinence of market prices for most of them. Only 12 MFIs are currently traded on an exchange market and it is highly questionable if those are a feasible sample of ordinary MFIs.

The thesis follows the argumentation of Krauss and Walter (2006) and uses Returns on Assets (ROA) as a financial performance measure. One major downside of using ROA is that accounting betas are smoothing down the underlying value of an institution. However by missing market prices it tends appropriate substitute. For a more comprehensive overview and to control the robustness of the results other performance measures are used as well, namely Sufficiency Index, Return on Equity,

Profit-Margin, Change in Size and Change in Financial Income.

The Mixmarket dataset is merged with macroeconomic variables from the World Bank database. The variables are the measures of the macroeconomic context and are all expressed in yearly changes. The macroeconomic variables are: Change in GDP, Change in Unemployment rate, Change in Inflation, Change in Added Manufacturing Value (as Percentage of GDP), Change in Foreign Direct Investment.

Furthermore two other datasets are used. Firstly, for the risk return analysis the Morgan Stanley Capital Indices (MSCI) for Emerging Markets and World Markets are used for the approximation of the perfectly diversified market portfolio. Secondly, the OSIRIS database from Bureau van de Dijk is used to gain first insights into how MFIs are affected by macroeconomic changes in comparison to its benchmark. OSIRIS includes 14 million private and public companies worldwide. For the purpose of this paper two benchmarks are constructed, commercial banks as a more narrow defined benchmark and the entire financial market in a country as a broader benchmark.

The analysis obtains two major parts. In the first section macroeconomic key measures are regressed on the MFIs performance indicators. Thereby panel regression is performed with country fixed effect and clustered error terms with each Year by each Country as the group variable. As a second step a modified Capital Asset Pricing Model is performed to get insights into the market movement as well as the risk-return trade-off. For the latter the MFIs are randomly assigned to 100 portfolios of equal size.

(8)

3 As a subsequent step the relationship of return and risk is measured by an ordinal least square regression. The thesis is the first in academic literature to examine the risk-return tradeoff in the MFI-market and therefore contributes to a deeper understanding of risk exposure and the return of MFIs.

For the first section the following results emerge: MFIs are affected by the change of the macroeconomic context independently of how MFI’s performance is measured or if it's controlled for. For all performance measures -except for Profit-Margin and Change in Financial Income- MFIs are similarly affected than Commercial Banks. However, the institutions differ significantly regarding their intercepts, where MFIs have a significantly higher intercept for ROA. Furthermore if MFIs are compared to the entire financial market of a country they are less affected by changes in the domestic economy. Finally, the results of the first section show that depending on the past financial performance, targeted market and legal organization MFI are differently exposed to changes of the macroeconomic environment.

The analysis regarding the risk return measure leads to two pivotal insights. MFIs are not significantly affected by movements of the MSCI for World markets. For the MSCI for emerging markets a slightly negative relationship is found. This means that MFIs' returns are complimentary to the emerging market movements. Nevertheless this result has to be interpreted carefully because of its rather small regression coefficient and -value. The focal insight of the r isk return analysis is that higher exposure to risk seems to be rewarded with higher returns, but the relationship is of minor magnitude. A relatively low -value approves that other factors than the market risk are causing higher returns.

In summary the empirical results lead to the following implications for investors. For domestic investors MFIs do not seem a convenient investment opportunity for diversification purposes. Because of similar returns but smaller risk than the entire financial market MFIs tend to be a more attractive investment opportunity for domestic investors who are restricted to financial institutions and aiming for a better diversification. Furthermore the results suggest that investors should consider institutional specific factors, such as legal status or targeted market, because these factors lead to higher exposure to market movements.

For investors who are holding an international market or an emerging market portfolio investments in MFIs seems to be an attractive diversification opportunity.

MFIs seem to be detached from movements of the international markets. Therefore their expected return seems to be generated independently from international market movements. For emerging markets the insights for investors are even more revelatory. The negative correlation with the emerging market portfolio suggests that MFI would be in theory a convenient opportunity for a short-position which is in practice restricted (De Roon et al., 2001). However, given that only the most developed and best reported MFIs currently operating in the market are considered in the analysis, the

(9)

4 Section 2 describes the current literature related to the topic of macroeconomic influences on MFIs performance. Section 3 introduces the variables created for the analysis and the use of data sources. Section 4 explains the empirical methodology of the analysis. Section 5 reports the results. Subsequently section 6 explains the causality behind the results and section 7 concludes finally the paper.

2. Related Literature

The master thesis builds up on the results and contributions of previous studies. The biggest bulk of academic literature, which is evaluating the success and failure of MFIs, focuses on micro-institutional determinates. A comprehensive review about current academic literature regarding micro-institutional determinates can be found in Manos & Tsytrinbaum (2014) and Armendáriz & Morduch (2010).

In contrast to the majority of researches on MFIs this paper focus on the macroeconomic influences and the market risk rather than the micro-institutional determinates which are leading to their success.

Missing market prices for MFIs are the major obstacle of examining the influence of macroeconomic factors on their performance. Only twelve MFIs out of 10,000 worldwide are publically traded on a stock exchange. This small group of MFIs is different in its structure and business model than common MFIs (Brière & Szafarz, 2013) and therefore they are unsuitable for using them as a reference group.

The thesis follows the approach of Krauss and Walter (2006 & 2009) to tackle the problem of missing market prices of MFIs. Krauss and Walter argue that through the absence of market prices the most feasible approach to measure MFIs performance is by accounting betas.

Krauss and Walter follow the argumentation of scholars such as Beaver & Manegold (1975) or Damodaran (2012) that accounting betas can be used as a legitimate estimated, but have to be interpreted carefully and conservative ly. The underlying problem of using accounting betas is that accounting earnings are flattened out relative to the underlying value of the company. This leads to betas which are biased down for risky firms and biased up for relatively safe firms. Thus accounting betas tend to be closer to each other for all companies using this measurement.

Furthermore accounting earnings are exposed by non-operating factors. These factors include beside others depreciation or inventory methods and the allocations of corporate expenses. In other words, companies with a constant performance could have a reported change in their performance solely due to differences in reporting methods.

Furthermore accounting earnings are usually only measured once a year for MFIs, resulting in regressions with only few observations and less predictive power. Given these obstacles Krauss and Walter (2006) suggest that the insight of an empirical examination reveals more a general tendency of

(10)

5 MFIs performance under market and macroeconomic changes rather than a distinctive modeling for precise hedging decisions.

Their research is currently the only published, which is using a modified CAPM to measure market exposure of MFIs. Thereby they are using different market indices, namely S&P500, MSCI World Market Index, MSCI Emerging Market Index, Domestic GDP and Domestic Stock Index. The results show solely a significant impact from Domestic GDP on MFIs performance.

The conclusion of their research is that adding MFIs in a portfolio may be interesting for the diversification for international investors, due to its independence of international market movements.

They only use the 325 Top MFIs1 for their examination and a sample period of nine years. The relatively small sample size leads to fewer observations and leads to possible biases of their results. A further problem -which occurs by including only the Top MFI’s in the dataset- is the correlation of MFI Return on Assets and region they are operating in.

Stephens & Tazi (2006) show that the financial performance of MFIs depends fundamentally on the region they are operating in. They also show that the most successful MFIs are accumulated in certain regions. Therefore the approach of Kraus and Walter could suffer from biases, because it mainly included MFIs from specific regions which may be differently affected in the small time period of the sample.

Beside Krauss and Walter, there are only few other researches which examine the attractiveness of MFIs for investors, due to its relative independence of market movements and macroeconomic shocks. The most often cited are Ahlin and Lin (2006), Gonzales (2007) and Galema, Lensink & Spierdijk (2011).

Ahlin and Lin (2006) are examining the influence of macroeconomic key factors on MFIs. They are using a rather small sample of 112 MFIs from 48 countries in an eight year time horizon. They are using different performance measures, namely self-sustainability, default rates, costs per borrower, and growth in number of clients. Regarding the macroeconomic variables they are using the growth rates of real per capita income, inflation, labor force participation rates, manufacturing’s share in GDP and net foreign direct investment as a fraction of GDP.

The thesis follows their considerations and includes the same variables in its examination of the influence of macroeconomic key variables. The results of Ahlin & Lins and Krauss & Walter lead to a skepticism if investments in microfinance reduce portfolio risk. However their paper shows if the influence of macroeconomic factors is controlled by structural determinants then the influence becomes smaller.

An opposite relationship between the macroeconomic environment and MFIs performance is presented by Gonzales (2007). He analyzes if the change in domestic GDP has an impact on MFIs performance, respectively their portfolio risk. Thereby four indicators are used to measure portfolio

1

(11)

6 risk, namely Portfolio at Risk2 over 30 days, Portfolio at Risk over 90 days3, Loan loss Rate, and

Write-off Ratio. His results show no relationship between changes in GDP and the indicators beside

Portfolio at Risk over 30 days. Therefore his results draw an opposite conclusion about the relationship changes in the Macro-Economy and MFIs assets quality. The study provide some evidence for the wide spread assumption that the MFI-market has a relatively high resilience to macroeconomic changes. In other words MFIs hold an attractive opportunity for portfolio diversification for investors. The sample he uses is assembled of 639 MFIs in 88 countries for the period 1999–2006.

In line with the results of Gonzales is the paper from Galema, Lensink & Spierdijk (2011). They conclude that MFIs may yield to diversification gains for a portfolio of risky international assets. Their sample contains nearly 3000 MFIs from 121 countries in the time period from 1997 to 2007. Their findings investigate that MFIs are an attractive investment opportunity through a better risk-return profile if particular regions and legal-status of MFIs are solely included. Their analysis suggests that investing in MFIs with the legal-status “Rural Bank” or from the region Latin America yield to more efficient portfolios, due to rather high resilience to macroeconomic shocks and higher expected returns. In contrast to that MFIs from Africa and with the legal-status “NGO” tend to be non-beneficial for investors portfolio.

The current literature on the influence of the macroeconomic environment on MFIs performance is ambiguous. Important to mention is that according to Barnjee et al. (2015) the MFI market experienced a major change in the last decade. Because the first three researches mentioned are using data nearly a decade ago, the possible differences of the result to current analyses may be explained by the evolution of the market. Furthermore the quality of available data for MFIs increases constantly and allows, according to Bauchet & Morduch (2010) to retain more precise results for the MFI-market.

3. Data

The dataset is assembled from four different sources. The data for the MFI are received from mixmarket.org. Mixmarket is the biggest provider for MFI’s investment and information flows. The dataset comprises annual data from 1995 until 2015. All numbers and ratios used are adjusted to US Dollar from their domestic currency. The initial data set of Mixmarket contains 2752 MFI’s in 121 countries. Even participation in the Mixmarket database is voluntary, MFIs have to submit substantiating documentation, such as audited financial statements, annual reports, ratings and institutional appraisals.

2

The fract ion of all loan-payments fro m the borro wers, wh ich are a fter 30 days behind sch edule 3

(12)

7 Due to the fact that the dataset from Mixmarket only includes MFIs with a certain report standard, the dataset is not a representative sample of all MFIs. As described by Gonzales (2007) the sample of Mixmarket represents rather the world’s top MFIs than a holistic representation of all MFIs.

Apart from closely monitoring the quality of the data from all participating MFIs, the Mixmarket sample applies certain adjustments, such as accounting for inflation, loan loss provisioning and most importantly subsidies. From the dataset MFIs are dropped, which has less than three consecutive years of financial statements. Furthermore only MFIs with the two best reporting standards are included. The Mixmarket dataset provides an ordinal rating system about the quality of the reported data. As Ahlin and Lin (2006) claim that only MFIs with the two best ratings4 are feasible for a comprehensive analysis. Both constraints leave 782 MFIs based in 61 emerging market countries.

This dataset is the base of the analysis in section 5.1.3 and section 5.2. It provides the most observations given a certain quality of the data. In comparison to the reduced dataset of section 5.1.1 and 5.1.2 this dataset enable to have more observations, for more MFIs and in more countries. As described in the subsequent section more observations tend to decrease the standard deviations and enable better statistical predictions.

At its beginning the paper examines the influence of macroeconomic key factors on MFI-performance. Therefore the Mixmarket dataset was merged with the dataset of the Worldbank, which provides publically accessible data of macroeconomic indicators. The indicators, namely domestic

GDP, inflation rate, unemployment rate, Foreign Direct Investment and added Manufacturing value as a percentage of GDP are transformed to their yearly change by logarithmic

transformation.

The logarithm of the change in GDP or respectively of GDP growth is the pivotal measurement of the analysis of this thesis. It represent the change of the overall macroeconomic environment in a country. The Change in Inflation is included because it clearly affects returns to lending, borrowing, and saving, at least when unanticipated and not indexed for. Because MFIs mainly operate in developing markets -which are often characterized by an unstable monetary environment- it is of interest if MFIs financial performance is affected.

The Change in Unemployment approximates work opportunities in the official economy, since unofficial economic activity is underrepresented in official unemployment figures. Finally the logarithm of Change in Manufacturing value added in the economy and the logarithm of Change in Net Inflows of FDI, both as fractions of GDP are included. These reflect to some degree the availability of wage labor at a level accessible to potential MFI clients. Thus, they

4

(13)

8 allow rough an assessment of the degree of rivalry between MFI-led development and development based on industrialization and wage labor.

Table 3.1 shows the descriptive statistics of the macroeconomic indicators. The table elucidates that for the emerging countries in our sample the standard deviations are considerably high, especially for the variables Change in Unemployment, Change in Inflation and Change in Foreign

direct investment. The consequence of such high standard deviations is the loss of predictive power.

The most pivotal indicator for the first analysis, the Change in GDP, is in respect to its standard deviation from decent quality, which is maybe one explanation for the occurrence of the high significant results in section 5.1.

The third dataset merged is OSIRIS, which is obtained from Bureau van Dijk (BVD). The dataset provides 14 million private and public companies worldwide. The dataset is used to obtain indication how MFIs are performing in comparison to a benchmark. Two different benchmarks are constructed, namely commercial banks and the financial sector of each country as a whole. The first benchmark is more narrowly constructed and includes solely commercial banking institutions.

Commercial banks are from their business operating similar to MFIs. As well as MFIs, they provide saving and lending facilities, but differ substantially in their targeting market. In some developing countries the commercial banking sector was even partially supplanted by MFIs, as e.g. in Bangladesh (Briere & Szafraz, 2013). To omit that institutions are occurring in both datasets, all MFIs, which are occurring in the BVD dataset as well are deleted. The final BVD dataset contains 823 commercial banks in 53 emerging markets. Due to the fact that the BVD dataset does not provide data for all countries the dataset for the comparison between the two different type of institutions in section 5.1.1 and 5.1.2 is further reduced Therefore the benchmark comparison exists of 682 MFI’s in 53 countries, which is nearly a 75% reduction from the initial dataset. Table 3.2 shows that MFIs and Commercial banks do not differ on average in respect to their return on assets and equity. This means that both types of institutions are similar to their relative profitability on Assets and Equity (ROA and respectively ROE). Furthermore Return on Assets has lower intra MFIs

Table 3.1 Descripti ve statistics of macr oeconomic al key indic ators: Macr oeconomic al Vari ables me an Me dian Standar d de viation Change in GDP 3.57% 3.5% 3% Change in Unemployment -1.9% 0.00% 13.71% Change in Inflation -0.31% 3.59% 59.52% Change in Manufacturing value added (% of GDP) -0.81% -1.18% 5.24% Change in Foreign direct investment 18.21% 4.66% 88.27%

(14)

9 differences than Return on Equity. Rather great differences occur for the profit-margin and the

leverage between both kind of institutions.

Table 3.2 Descripti ve statistics: MFI and Commercial B ank key figures

(15)

10 .

Return on Assets Bank 2.84% 2.11% 3.48% 714

Cred it Un ion/ Cooperative 2.28% 1.62% 2.84% 700 Non-Ban k financia l Institutes 2.75% 2.89% 4.09% 2,313

NGO 2.84% 2.71% 4.05% 2,087

Other 1.60% 1.84% 5.23% 88

Rura l Bank 2.91% 2.58% 2.49% 170

Total 2.61% 2.4% 3.9% 5152

Commerci al Bank 1.86% 1.74% 1.55% 8035

Return on Equ ity Bank 14.63% 16.04% 16.50% 714 Cred it Un ion/ Cooperative 9.98% 8.59% 13.60% 700 Non-Ban k financia l Institutes 11.00% 11.31% 19.90% 2,313 NGO 9.97% 9.75% 18.73% 2,087 Other 4.61% 5.44% 27.01% 88 Rura l Bank 21.12% 20.48% 16.2% 170

Total 10.8% 10.5% 19% 5149

Commerci al Bank 15.24% 15.40% 10.59% 7896

Debt to Equity rat io Bank 5.47 5.3 3.68 714

Cred it Un ion / Cooperative 4.7 4.31 3.06 700 Non-Ban k financia l Institutes 3.4 2.62 3.17 2,313

NGO 3.32 1.93 3.64 2,087

Other 2.45 1.49 3.11 88

Rura l Bank 6.74 6.19 3.2 170

Total 3.87 2.84 3.57 5618

Commerci al Bank 10.03 9.14 4.81 8058

Profit Margin Bank 12.96% 14.08% 19.36% 712 Cred itUnion / Cooperative 12.89% 11.84% 16.43% 700 Non-Ban k financia l Institutes 7.80% 12.77% 26.05% 2,316 NGO 7.25% 11.09% 23.69% 2,087 Other -7.88% 4.31% 39.36% 88 Rura lBank 15.5% 18.88% 14.25% 170

Total 8.22% 12.03% 24.14% 5555

Commerci al Bank 29.65% 30.95% 21.21% 7702

Table 3.3 Descripti ve statistics: MFI and Commercial B ank fi nancial r atios

Current legal status me an me di an Standar d de vi ation Fre quenc y Change in Bank 24.85% 23.3% 25.76% 693

(16)

11 MFIs are more equity than debt funded and have on average a smaller profit-margin than their counterparts. The relatively lower leverage of MFIs could be seen as an predictor of more stable performance during financial shock. Similar to the ROA and ROE the additional key figures of table 3.3 do not differ between the two types of institutions.

Financia l Revenue

Cred itUnion / Cooperative 22.41% 21.45% 26.41% 657 Non-Ban k financia l Institutes 26.93% 24.11% 33.87% 2,257 NGO 19.63% 18.47% 27.97% 2,054 Other 23.65% 11.96% 44.88% 85 Rura lBank 22.1% 18.39% 21.89% 169 Total 23.74% 21.04% 30.68% 4374 Commerci al Bank 22.77% 4.34% 82.43% 6911

Change in Assets Bank 27.79% 24.76% 28.10% 693 Cred itUnion / Cooperative 27.81% 23.82% 46.07% 657 Non-Ban k financia l Institutes 30.02% 26.42% 44.38% 2,257 NGO 20.63% 17.11% 52.19% 2,054 Other 30.92% 17.93% 62.86% 85 Rura lBank 22.96% 21.66% 25.90% 169 Total 26.4% 21.82% 46.05% 5158 Commerci al Bank 17.30% 11.51% 31.18% 7038 Change Operational self Sufficency Bank 1.37% 0.90% 12.31% 693 Cred itUnion / Cooperative -0.18% -0.13% 13.82% 657 Non-Ban k financia l Institutes 2.78% 1.35% 17.12% 2,257 NGO 1.26% 0.59% 16.10% 2,054 Other 6.59% 2.02% 25.06% 85 Rura lBank 0.64% -0.01% 10.96% 169 Total 1.98% 0.90% 16.27% 4825 Commerci al Bank 1.42% 0.13% 14.65% 6721

Sufficiency index Bank 53.83% 53.79% 6.66% 679 Cred itUnion / Cooperative 54.18% 53.22% 6.00% 631 Non-Ban k financia l Institutes 52.38% 53.41% 9.00% 2167 NGO 52.36% 52.94% 7.94% 1977 Other 45.88% 50.98% 16.17% 80 Rura lBank 54.47% 54.93% 3.82% 165 Total 52.72% 53.31% 8.16% 5699 Commerci al Bank 63.67% 63.92% 7.48% 7849

(17)

12 Worth to mention are the relatively high standard deviations for all variables beside the sustainability index5. The high standard deviations give an insight of a major statistically obstacle of the analysis of this paper. The magnitude of the standard deviations dilutes the statistical predictive power of the independent variables. Due to high outlier for all variables, the variables are winsorized up to five percent. Although the quality of the dataset is attempted to being maximized, this study is still facing some limitations regarding the accounting standards, consistency and precision of the used data. These limitations lead to reduced and higher standard errors. Given that the dataset only provides one observation per year for each MFI the volume of available observation per institution does not exceed 20. However the relatively large number of MFIs in the dataset makes it arguably for statistical examinations.

Beside Commercial banks a second benchmark is included. This includes all companies in the financial sector, such as insurance companies, trading companies, investment banks, etc. This benchmark is used to detect if MFIs are more or even less affected by a change in the macroeconomic environment than the financial market as a whole. This benchmark includes 13.093 companies in 53 countries. Similar to the commercial bank dataset MFIs which appear in both datasets are deleted from the financial sector one. Furthermore the number of company per country is disparate distributed. For some countries less than ten companies are included, where for others more than 600 companies (India & China) occur in the dataset.

Given the unequal distribution of companies in the countries the BVD-dataset is not feasible to create a market index for the risk-return analysis in section 5.2. As Krauss and Walter (2006) describe the Emerging market index and World market indices from Morgen Stanley International Capital (MSCI) tend to be more suitable for such an analysis. Unfortunately the MSCI Emerging Market Index is only available from 2002 onwards. In comparison to the MSCI-World market index this reduces the sample by seven years and less observations for the MSCI-Emerging Market index. The data of both indices are publically available.

Because there are no market prices for MFIs in general6, accounting measures have to be used to examine the performance of MFIs. The focal key variable for the following analysis is Return on Assets (ROA), which is a profitability measurement. In section 5.1.2 ROA is substituted for other accounting measures to ensure a holistic measurement of financial performance. The variables namely are:

 Return on Equity (ROE)

 Profit-Margin (PM)

 Sufficiency Index (S-I)



5

A description of the variables will follow in the following paragraph of this section. 6

(18)

13

 Financial Income

ROE and PM are both profitability measures as well. The sufficiency index7 is a self-constructed variable, which measures financial health of the institution. Self-Sufficiency is a monotonic transformation from Operational Self Sufficiency, which itself is calculated by financial

revenue divided by financial expense + loan loss provision expense + operating expense . This

transformation is conducted to reduce the outlier problem. The change in size and change in financial income are growth measures. Size is approximated by the amount of active borrowers. Beside the change of size table 3.2 & 3.3 shows the descriptive statistics of all variables.

Operational Self Sufficiency is used in its lagged form in section 5.1.3 as a control variable. Operational Self Sufficiency gives an insight of the financial health of a MFI. Therefore section 5.1.3

controls if the impact of a change in GDP is reduced when controlling for financial self-sufficiency. Beside the Operational Self Sufficiency further control variables in section 5.1.3 are Age, , dummy variables of targeted market and dummy variables of legal status of the entity. As claimed by Ahlin and Lin (2006) Age is an approximation of experience and tend to have a non-linear relationship, which is controlled for by its squared value.

Banerjee, Karlan & Zinman (2015) describe the influence of the targeted group on the performance of MFI. Target group is operationalized by three dummy variables8, which are indicating the average depth9 of the MFI. The last variable for which is controlled is legal-status of MFIs. Brau & Woller (2004) indicate the relationship between the performance of MFIs and the legal status. The six different peculiarities of MFIs are NGO, Non Bank Financial Institution, Credit Union/

Cooperativa, Rural Banks, Banks and Other.

4. Empirical Methodology

4.1 Accounting beta as focal measure of performance

The statistical examination uses different approaches to describe the influence of the macroeconomic changes on MFIs. Thereby the focal measurement of the MFI’s performance is Return on Assets (ROA). By using ROA there are certain disadvantages than using stock-market prices. The accounting beta seems to be an adequate measurement if there are no traded prices available. Thereby ROA is regressed as an independent variable10 against changes in the macroeconomic environment on an annual basis.

7_{Formu la for Financial self-suffic iency:}

8

The three dummy variables are: low e nd = Depth <20% OR average loan size <USD 150 Broad = Depth between 20% and 149%

High end= Depth between 150% and 250% 9

Formu la for Depth: 10

(19)

14 Even though there is a correlation between 30 and 60 percent between accounting and the market betas, there are some major disadvantages by using this accounting measurement (Krauss & Walter 2006). On the one hand accounting earnings are smoothing out the real underlying value of an institution. This leads to betas, which are biased up for safer firms and biased down for rather risky firms. This may result in betas which are closer to one for all institutions. On the other hands accounting earnings can be changed by non-operating factors, such as changes in inventory methods, depreciation and through allocations of corporate expenses at the divisional level. Finally, accounting earnings are measured only once every year for MFIs, resulting in regressions with few observations and a reduction of predictive power (Damodaran, 2012).

But due to the fact that only 12 MFIs are publically traded worldwide and they differ in many regards to their non-traded counterparts, the accounting beta approach is the most decent estimation approach.

For all regressions in section 5.1 an ordinary least square regression model with Country fixed effects and clustered standard errors11. Country fix effects were used The pivotal advantage of using fixed effects estimation is the opportunity to control for unobserved attributes of a country, which are correlated with the macroeconomic environment and important for MFI's performance. Such an attribute of a country could be corruption or financial development. Both attributes would be partially responsible for macroeconomic growth as well as MFI performance. All regressions are facing particular problems concerning the data. As seen as a glance at table 3.2 and 3.3 outlier problem tend to be severe.

4.2 Operationalization of Hypothesis

The aim of this paper is to contribute to the understanding of MFIs performance under changes in their economic environment. The underlying research question is how detached is MFIs performance from external factors. Thereby the paper follows two basic questions. Firstly it is analyzed if external factors have an influence on MFIs financial performance. Secondly it is examined if higher risk exposure to the market is rewarded with higher returns. By doing so four hypothesis are constructed to answer the question from different perspectives.

4.2.1 Hypothesis 1: Does Macroeconomic key factors influence MFIs and its benchmarks?

As described in section 2 the current literature find ambiguous results concerning the impact of external factors on MFIs performance. In section 5.1.1 regression model 4.1 analyze therefore the impact of macroeconomic factors on MFIs financial performance.

11

The standard errors are c lustered per Country and Year

(20)

15 The actual null hypothesis assumes that there is no relationship between the financial performance of the MFI and the change in the macroeconomic key variables. The rationale behind it is that MFIs performance is detached from the macroeconomic context and generates returns independently from it. In statistical terms it is proved if the regression coefficient of the macroeconomic key variables is significant different from zero.

Because the underlying purpose of the thesis is to understand MFIs from an investment point of view, a comparison to MFI’s benchmark is performed as well. As stated above Commercial banks are from their business operating similar to MFIs. As well as MFIs, they provide saving and lending facilities, but differ substantially in their targeting market. Therefore this section is testing if MFIs are differently affected than their benchmarks. The test which is conducted is described in equation 4.2.

4.2.2 Hypothesis 2: Are further accounting measures other than ROA affected by macroeconomic

changes?

As stated above in section 2 one explanation of the different results in current literature is the heterogenic use of performance measures. In this hypothesis the results of the previous hypothesis is stress tested. Thereby the macroeconomic variables with a significant influence on ROA are regressed on other accounting measurements by regression 4.3. This approach examines if other performance measurements, such as Return on Equity, Profit-margin, Change in Size, change in Financial Income and Sufficiency index are similarly affected than ROA. As described in section 3 these measures are not solely profitability measures, such as ROA, but represent different accounting aspects of an institution. These aspects are sustainability/financial health of an institution and growth.

y = Independent variable ROA t=year

i=MFI j=Country

k=institutional type (Commercial bank or MFI) =Intercept

= regression coefficient

X=set of macroeconomic key variables such as described in section 3

=residual

(4.2)

= Regression coefficient macroeconomic key variables on

MFIs ROA

=Regression coefficient macroeconomic key

variables on Benchmarks ROA

=standard deviation MFI

(21)

16 The results are again compared to their counterparts, commercial banks by using equation 4.2. The comparison enable a further insight to which degree the MFI-sector is standing alone or respectively how related it is to commercial banks, towards macroeconomic shocks.

4.2.3 Hypothesis 3: Do institutional-specific factors lead to a higher exposure to macroeconomic

changes?

There is clear evidence in academic literature that institutional-specific factors have an impact on MFIs performance. The rationale for including the control variables is receiving a more accurate estimation for the influence of domestic macroeconomic situation on the MFI performance. Even though there are myriad factors to control for, there are few fundamental factors which seem to be pivotal determinants for financial success (Brau & Woller, 2004). These factors are experience , target market and the legal-status of a MFI.

Experience is thereby approximated by the years of being active in the micro-lending market. This variable is named Age in equation 4.4a & 4.4b. Due to the assumed non-linear relationship of experience, the variable is included as well. As described in section 3 target market is characterized by four different peculiarities and therefore operationalized in the regression with three dummy variables. Mixmarket provides in its dataset six different peculiarities for the legal-status. For the purpose of the examination five dummy variables are constructed. As the last variable Operational

Self Sufficiency is included. The first and the second lag of that variable describe if historical

sustainability in the past has an influence of the MFIs performance.

In a first step the institutional determinants are regressed on ROA as described in re gression 4.4a. The regression shows less the direct answer to the hypothesis, than rather give a glance if the impact of change in the macroeconomic context diminishes by including controls.

As scholars, such as Ayayi & Sene (2010) claim there is a structural determination of legal types on target market. The legal status tells to a certain extend something about the business structure of a MFI. E.g. MFIs which are organized as NGOs tend to target lower end costumers than MFIs, which have a Rural Bank status. To omit multicollinearity both regressions of 4.4 are performed with and without the dummy variables of legal-status.

(4.3)

y = Independent variables. Set of accounting performance measures

t=year i=MFI j=Country

=Change in domestic GDP

(22)

17 (4.4a)

y = Independent variables. Set of accounting performance measures t=year

i=MFI j=Country

(23)

18 In a second step the interaction effects of the control variables and the change in GDP are analyzed. In regression 4.4b the regression coefficients from the interaction effects explain if change

in GDP has a greater influence for certain structured MFIs or not. The rationale behind is that the

institutional specific determinants detach MFIs from macroeconomic movements. For example experience might lead to better hedging strategies of MFIs as assumed by Ahlin and Lin (2006). Therefore older MFIs would tend to be less affected than their younger counterparts.

.

4.2.4 Hypothesis 4: Are MFIs affected by emerging and international market movements?

The second approach of the analysis is examining market risk of MFIs. Therefore the Capital Asset Pricing Model (CAPM) is used in a modified way. The CAPM estimates the asset's sensitivity

to non-diversifiable risk, which is denoted by , as well as the expected return of the market and the

theoretical risk free rate for an investment12.

12

For further e xp lanations about the Capital Asset pricing model see Sharpe (1964) (4.4b)

y = Independent variables. Set of accounting performance measures t=year

i=MFI j=Country

(24)

19

As mentioned above the CAPM normally uses market prices rather than accounting returns. Due to the absence of publicly traded prices for MFIs the accounting approach seems the most feasible. Empirical studies show a correlation between 30 and 60 percent between the accounting beta and market price approach (Damarodan, 2012). For the market return the paper followed the approach of Krauss and Walter (2008), who argue persuasively that the Morgan Stanley Capital ind ices for Emerging Markets and World Markets are reasonable benchmarks for the market movement. The

market movement and the market risk (denoted by the value) are ascertained by using both indices.

The results of regression 4.5 explain therefore if MFIs are affected by emerging and international market movements.

4.2.5 Hypothesis 5: Are MFI with a higher risk exposure rewarded with higher returns?

In a second step the risk return tradeoff is examined. Therefore Betas and the Expected returns are regressed. To estimate the betas 100 randomly assigned portfolios from equally size are constructed. The construction of a beta for every MFI is not convenient, because there is only a maximum of 18 observations (median 9) for each MFI. The portfolios include more MFIs and hence increase the number of observations. The MFIs are randomly assigned to the portfolios. The estimated betas from

the portfolios are then regressed on the expected value of the portfolios by a common OLS-regression.

5. Results

In this section the results are discussed and interpreted. The focus lies on statistical discussion. The subsequent section 6 interprets the causality and validity of the results. This section is organized in two main parts. In the first part the results of the macroeconomic influence on MFI performance are presented and compared to commercial banks. This includes the statistical examination of hypothesis 1 to 3. In the second part the results of the market risk measurement by the accounting beta and return are presented.

(4.5)

ROA=Return on equity =Risk free rate i=MFI

=Intercept =market risk

(25)

20

5.1 Baseline Results

5.1.1 Influence of Macroeconomic key factors on MFI in Comparison to benchmark

Table 5.1 compares the influence of macroeconomic key figures on MFI and their benchmark

commercial banks. The comparison of regression 1 and 3 shows that on average the change of GDP seems to have a slightly greater influence for MFI than on its benchmark, namely commercial banks.

Nevertheless given the coefficients and their standard errors there is no significant difference between the market movements of MFIs and Commercial banks at a 5 % significance level (p=0.108). There is no statistical evidence that the change in GDP does affect MFI to a similar magnitude than their benchmark commercial banks. One reason for similar market movements could be a high interrelation between Commercial banks and MFI's Return on Assets. To test a possible statistica l relation between them two indices are constructed. The indices include the average Return on Asset for all MFIs and respectively Commercial banks for each individual year and country. The correlation of both (r= -0.03) suggest that statistically there is no clear relationship in the movement of both indices. Nevertheless regression 1 of table 5.1 show that every additional one percent increase in GDP affect the Return on Assets of MFI’s by 13.67%.

By extending the benchmark to all companies operating in the financial industry there is a clear difference regarding the influence of change in GDP and their performance. Regression 5 of table 5.1 show that a one percent change in GDP leads to nearly 30 percent change in Return on Asset. Given the standard deviation this differences are significant on a 5 percent level. Therefore the financial market as a whole seems to be more affected in its magnitude by a change in GDP.

Table 5.1 Comparison of the influence of macroec onomic key vari ables on MFI’s and Benchmark perfor mance

Independent variables MFI Benchmark Commercial

Banks Benchmark Entire Financial Sector (1) (2) (3) (4) (5) (6) 0.1367*** 0.1136*** 0.0805*** 0.0595*** 0.2988 *** 0.1487*** (0.032) (0.0342) (0.0139) (0.0190) (0.0977) (0.0483) 0.0011 0.0001 0.0049*** (0.0013) (0.0006) (0.0018) -0.0016 -0.0014 -0.0213* (0.0049) (0.0036) (0.0114) 0.0120 -0.0070 0.0277* (0.0167) (0.0081) (0.0145) 0.0008 0.0001 0.0071*** (0.0008) (0.0003) (0.0009) Intercept 0.025*** 0.026 0.017*** 0.017*** 0.034*** 0.036*** (0.0029) (0.003) (0.001) (0.001) (0.005) (0.006) 0.1439 0.1403 0.1751 0.2373 0.0598 0.0679 Obs. 4,318 3,051 8,153 4,129 17,391 12,226

Description: Each column reports coefficient estimates from a country fixed effect regression of the column variable on the Return on Assets of the row variables. MFI and benchmarks-clustered bootstrapped standard errors are in parentheses (see section 4, Empirical Methodology ). Coefficients significantly different from zero in the regression, are marked by superscript signs. Significance levels 1%, 5%, and 10% are denoted by *, **, and ***,respectively.

(26)

21 In Regression 2, 4 and 6 of table 5.1 insights from the former regression are e xtended. These regressions add additional macroeconomic key indices. The additional key indices do not show any significant relationship to the ROA for MFI and Commercial banks. Even though there is a decrease for the regression coefficient of the change in GDP, none of these changes are statistically significant. For the emerging market in the sample neither MFI nor Commercial Banks are significantly influenced by these indices, except change in GDP. As stated above this is maybe caused by the character of ROA, which -as an accounting measurement- smoothes out changes more than market price would do. Another obstacle of receiving decent prediction are the relative high standard deviations for all macroeconomic key variables beside change in GDP.

At least for the measurement of ROA it cannot conclude that these indices have an influence on Commercial banks and MFI. This is congruent to the theory that MFI are relatively independent from macroeconomic influences.

In contrast to MFIs the financial sector as a whole is influenced by changes in the macroeconomic indices. All coefficients are statistical significant, but beside the change in GDP the coefficients for the other indices are rather small in their magnitude. The results show that the performance of MFI’s is less affected by macroeconomic changes than the whole financial market. Being rather inelastic two the economical environment makes MFIs rather stable according to their performance. A further insight of the results above is that Commercial banks and MFI does not differ in the change of their Return on Assets caused by change in the domestic GDP.

5.1.2 Macroeconomic influence on further MFI performance measures

Due to the fact that Return on Asset does not incorporate all information over the financ ial performance further accounting measures are used. As mentioned in section 3 the variables

Sufficiency Index, Profit-Margin and Return on Equity are profitability indicators and change in size

and change in Financial Income are growth indicators. The result of table 5.2 shows for the indicators for MFIs as well as Commercial banks a positive influence by the change in domestic GDP. Similar to the Return on Assets there is no significant difference in the magnitude of the movement ( ) between MFI and Commercial banks for the two indicators Sufficiency Index and Return on Equity.

On the other hand this does not hold true for Profit-Margin. The results show that every change in GDP has a higher impact for Commercial banks than MFIs. This implies that the Profit-Margin of MFIs seems to be more stable to changes in the domestic Economy than its benchmark. The two indicators, which are measuring growth show two opposed results. By measuring growth by the increase of active borrowers there is no difference between the two financial institutions.

Thereby the indicator show that every additional increase in GDP let increase the size of the MFI’s between 0.88% and 1.89% (in a 95% confident interval). In contrast to that the measurement of change in size operationalized by the change of financial revenue show a different result.

(27)

22 The impact of the change in GDP is higher for Commercial banks than for MFIs. In statistical terms a one percent increase or decrease affect the change of Financial Income between 1.7% and 2.66% for MFIs and between 5.74% and 13.01% , given a 95% confidence interval. Such as the

Profit-Margin, the change in financial income seems to be more stable for MFIs when a change in

domestic economy occurs.

In line with the results for Return on Assets in the section 5.1.1 a change in domestic GDP has a similar impact on the indicators self-sufficiency, Return on Equity and the change in size. As table 3.1 shows that the profit margin for MFI is on average 21.65% lower than for commercial banks. The relatively low Profit-Margin for MFI can be an explanation for the lower impact which a change in domestic GDP occurred.

5.1.3 Regression results with control variables and interaction terms

As a first glance the control variable Age seems to have a non-linear impact on ROA. Age as an approximation for experience in lending and collection practices has a negative influence on the ROA. As seen in table 5.3 younger MFI perform financially better than their older counterparts. But the squared value of the variable Age2 suggests that there is a non-linear relation. The coefficients of both variables age and Age2 let conclude that younger firms perform better than older, but this effect changes over time and lead that older firms receiving higher returns.

Beside Age and Age2 the control variables do not exhibit any special particularities, which are worth to mention. The most revealing insight of the regression (1) and (2) of table 5.3 is that the control variables do not change the magnitude of the regression coefficient of the change in GDP Table 5.2 Influence of growth in GDP on different perfor mance me asures

Sufficiency Index Return on Equity Profit-Margin Financial Income MFI Beta 0.1375*** 0.4881*** 0.4502*** 1.3913*** 2.1872*** (0.0475) (0.1367) (0.1492) (0.2588) (0.2461) Intercept 0.519*** (0.0019) 0.099*** (0.011) 0.123*** (0.017) 0.121*** (0.0097) 0.139*** (0.014) OBS 5,545 5,078 5,540 5,074 4,273 Commercial banks Beta 0.1561*** 0.5544*** 0.9758 *** 1.6257 *** 9.374*** (0.0585) (0.095) (0.1696) (0.2913) (1.856) Intercept 0.623*** (0.006) 0.128*** (0.004) 0.251*** (0.015) 0.08*** (0.003) 0.2451*** (0.019) OBS 7,967 8,015 7,822 7,233 6,321 p-value for difference in betas 0.9092 0.6904 0.02** 0.5458 <0.001***

Description: Each column reports coefficient estimates from a country fixed effect regression of the column variable on the Microfinancing institutes (MFIs) and its benchmark, Commercial banks Ret urn on Assets. MFI or benchmark -clustered bootstrapped standard errors are in parentheses (see section 4, Empirical Methodology). Coefficients significantly different from the regression, are marked by superscript signs. Significance levels 1%, 5%, and 10% are denoted by *, **, and ***,respectively. The very bottom row describes the p-value of the t-test about the differences in the regression coefficients

(28)

23 significantly. With the given control variables it cannot be concluded that the relationship of ROA and change in GDP is spurious.

Furthermore table 5.3 shows the results of the interaction effects between change in GDP and the control variables on MFI returns. The results show that MFI with a rather better financial health in the past (measured by the first and second lag of the Operational self sufficiency) are more affected by change in GDP than institution with lower values.

The target market of a MFI reveals to multiply the effect of a change in GDP. MFI's, which target low end, broad and high end borrowers are more affected in their return if the domestic economy changes than MFIs, which target Small businesses.

Table 5.3 Comparison of the influence of macroec onomic key vari ables on MFI’s and Benchmark per for mance

Independent variables Without Interaction effects Interaction effects Results presented

Independent variable* growth

(1) (2) (3) (4) 0,1193*** 0,1170*** (0,0315) (0,0315) -0,0014** -0,0015*** 0,0354 0,0304 (0,0006) (0,0006) (0,0277) (0,0268) 0,0001** 0,0001** -0,0019 -0,0017 (0,0000) (0,0000) (0,0018) (0,0018) 0,0574*** 0,0576*** 0,4317** 0,4483** (0,0065) (0,0065) (0,1976) (0,1975) 0,0058* 0,0060* 0,3080** 0,3231*** (0,0032) (0,0032) (0,1248) (0,1218) Low end 0,0142*** 0,0096*** 0,2203*** 0,0972 (0,0026) (0,0026) (0,0490) (0,0556) Broad 0,0074*** 0,0049** 0,1679*** 0,1006** (0,0022) (0,0022) (0,0431) (0,0458) High end 0,0072** 0,0055** 0,1954*** 0,1406** (0,0029) (0,0028) (0,0607) (0,0583) Rural Bank s -0,0017 -0,0512 (0,0029) (0,0636) Credit Union/Cooperativa -0,0023 -0,0554 (0,0018) (0,0425)

Non Bank Financial Institutions 0,0065*** 0,1576***

(0,0015) (0,0307)

NGO 0,0066*** 0,2115***

(0,0015) (0,0349)

Other type of MFI 0,0122 0,2743**

(0,0076) (0,1231)

Intercept 0.014*** 0.013*** 0.019*** 0.018***

(0.005) (0.006) (0.008) (0.008)

OBS 3,976 3,976 3,976 3,976

Description: Each column reports coefficient estimates from a country fixed effect regression on the Microfinancing institutes (MFIs). MFIs -clustered bootstrapped standard errors are in parentheses (see section 4, Empirical Methodology). Coefficients significantly different from zero in the robust regression are marked by superscript signs. Significance levels 1%, 5%, and 10% are denoted by *, **, and ***,respectively. The very bottom row describes the p-value of the t -test about the differences in the regression coefficients. Column (1) and (2) describes the regression from the column variables on Return on Assets. Regression (3) and (4) describe solely the interaction effect of the row variable with the change of GDP . Column (3) and (4) does not display the coefficients of the row variables, but only their interaction effect.

A test which is not presented in the table shows that the remaining group of targetin g market, namely small business has a non significant interaction effect with the change in GDP. One possible

(29)

24 explanation could be that these target markets are more elastic in respect to their supply and demand than the market for small business lending. Another explanation can be found in the results of regression 4. The target market is often correlated with the legal status of an entity.

As Merton (2012) describes, especially the NGO and Non Bank financial Institutions are serving low end and broad borrowers. By adding the legal status in the regression the interaction coefficient of low end and change in GDP gets non-significant. For the broad and high end targeting MFIs the significant level as well as the coefficient decreases. Beside others the legal status explains if a MFI is regulated of gaining profits and the screening of clients. Traditionally MFIs which are organized as NGO and Non Bank Financial Institutions are less strict and reluctant about new borrowers.

In case of a domestic downturn of the economy this can lead to higher default rates and a more severe impact than for other types of MFIs. On the other hand these kind of MFI charge on average smaller interest rates. In case of a general increase in demand for loans -e.g. caused by an increase in domestic GDP- the demand for these loans with lower interest rate increases relatively more.

5.2 Beta approach and market risk measurement

5.2.1 Market Beta

The regression coefficients as well as the values obtained are rather low. As stated above this is partially caused by the accounting beta approach. Accounting betas compared to traded market prices reduce the apparent sensitivity of the market risk (described by ) and the market movement (described by ), because the market impact of change in GDP is not directly observed by the annual accounting results.

The regression of the MSCI Emerging Market leads to a beta of around 0.8%. The influence is statistical significant, but the magnitude rather low. Interesting enough is the direction of the influence. The financial return of MFI's is negative related to the Emerging Market movements. This negative relationship means that MFIs react antagonistic to changes to the index. The result underlines the results of the study of Gonzales (2009) more than a decade ago.

As described in section 2 the his results claim that including MFI in a emerging market portfolio increase the diversification and reduce the risk of the portfolio, due to its negative relationship to emerging market movements. Different to the paper of Gonzales this analysis includes more MFIs and a longer and more recent time-span. Even though the MFI sector became more developed and moved according to Briere and Szafarz (2013) to a more bank related sector, the result show that MFI seems to be rather unique in its market movements.

Different to the MSCI Emerging market Index MFI's does not seem to be affected by movements of the World Market index. This means that MFI`s are not affected by world market

(30)

25 movements, given the sample and the ROA financial performance measurement. The glance of the results of the MSCI world market index is that MFI's are indeed rather independent from world market movements, as many studies deemed. As a consequence of that MFI's could be an interesting investing opportunity for diversification.

5.4 Results of accounting be ta CAPM wi th MSCI Emerging and Worl d as market inde x MSCI Emerging MSCI World

Beta -0.008*** (0.0025) 0.0007 (0.0035) Alpha 0.0133 *** (0.0021) -0.015 0.0053 <0.0000

Description: Each column reports Beta and Alpha from the Capital Asset Pricing Model with accounting earnings, respectively Return on Assets for Microfinancing institutes (MFIs). The abbreviation MSCI stands for Morgan Stanley Capital Index and is for both columns a proxy for the perfectly diversified market portfolio. Standard errors are in parentheses. Coefficients significantly different from zero in the regression are marked by superscript signs. Significance levels 1%, 5%, and 10% are denoted by *, **, and ***,respectively.

5.2.2 Risk and Return comparison

The relationship between risk and return as shown in table 5.5 and graph 5.1 is positive. The result in table 5.5 represents the regression from the portfolio betas and the expected return of the portfolios. The MFIs in the portfolios where randomly assigned to create portfolios with equal observations and numbers of MFIs. Thereby every portfolio is regressed on the market return, which is approximated by the MSCI Emerging market Index. The coefficient which explains the relationship of the portfolio betas on the expected return of the portfolio is highly significant but rather small from its magnitude and predictive power. Nevertheless the result follows the economic rationale that portfolio with higher risk are rewarded with higher returns, even though the relationship seems to be rather small. The risk return regression was performed myriad times with different randomly composed portfolios, to ensure that the relationship was not coincidently achieved. The results presented in table 5.5 are the average of these calculations.

Table 5.5 Regression fr om be tas on e xpec te d re turn,

per for me d by creati ng 100 be tas fr om 100 different portfolios

Beta 0.03482***

(0.005)

Alpha 0.01633***

(0.002241)

R-Square 0.008

Description: The table reports the Beta and Alpha from the regression from portfolio beta on expected return. The underlying returns are received from the accounting earning, Return on Assets. Morgan Stanley Capital Index is thereby the proxy for the perfectly diversified market portfolio. Standard errors are in parentheses. Coefficients significantly different from zero in the regression are marked by superscript signs. Significance levels 1%, 5%, and 10% are denoted by *, **, and ***,respectively. To receive the betas 100 portfolios from equal size are constructed by randomly assign MFIs to each of them.

Micro-financing institutions as a portfolio diversification opportunity : how independent are micro-financing institutions from macroeconomic shocks?

Master in Science Business Economics, Finance Track

Master Thesis