Performance of microfinance institutions in Vietnam

ften lead to autocorrelation, variance of variation in the model. Therefore, Arellano and Bond (1991) proposed using the GMM regression method to overcome the above phenomena. Besides, Blundell and Bond (1998) suggest that when the dependent variable has a high correlation between current values and previous period values, and the number of periods is not too long, the DGMM estimation method is ineffective because the tool variables used to be rated are not strong enough. Blundell and Bond (1998) extended the DGMM estimation method with the simultaneous consideration of two estimation methods (basic model, GMM, and DGMM model) collectively known as System Generalized method of moments (SGMM). In this study, because the period of 2013-2017 is not too long and the financial data of MFIs often have a high correlation between the present value and the value in the previous period, the author uses the method estimating SGMM. 1.6. New contributions of the thesis The thesis focuses on specific objectives including: (1) Assess the status of operational efficiency of MFIs in Vietnam; (2) Identify factors affecting the 7 performance of MFIs in Vietnam; (3) Assess the impact of women's empowerment on the performance of MFIs in Vietnam; (4) Proposing policy implications to improve the performance of Vietnamese MFIs. Compared with the previous studies, the thesis has made the following new contributions: Based on Data Envelopment Analysis (DEA), the author evaluated the performance of 26 Vietnamese microfinance institutions. The previous studies, when analyzing business operations and assessing business performance in general and MFI in particular, often used the main financial indicators such as ROA, ROE, NIM, ... because of the calculation relatively simple and easy to understand. Each financial indicator expresses the relationship between the two variables, reflecting an aspect of MFI's performance. Therefore, to fully evaluate the performance of MFI, we must use a series of different criteria. This makes it difficult for administrators and other state management agencies to evaluate and compare the performance of MFIs, especially when evaluating the efficiency of using resources to create complex financial products and services such as those of MFI (Manandhar and Tang, 2002). To overcome the disadvantages in the method of analyzing financial indicators, the thesis has used Data Envelopment Analysis (DEA) to evaluate the performance of MFIs. In addition, based on data sources of 26 MFIs in the period of 2013 - 2017, the author has identified factors affecting the performance of MFIs in Vietnam. Compared to previous studies, the thesis more comprehensively examines the impact of women's empowerment on the performance aspects of Vietnamese microfinance institutions by quantitative research methods along with the Stata 15.0 software support. Specifically, the author assesses the impact of women's empowerment on the performance aspects of microfinance institutions by estimating models using Blundell and Bond's SGMM method (1998). This method is commonly used in linear dynamic panel data estimates to overcome endogenous phenomena that often occur in macroeconomic models. Therefore, the obtained results ensure the reliability to draw conclusions. Thus, the research results provide rigorous evidence to support the theory of the impact of women's empowerment on 8 the performance of Vietnamese microfinance institutions. At the same time, the research results also provide a basis for research methods to assess this impact. In practice, stemming from the fact that the majority of MFIs in Vietnam have achieved their self-sustainability performance indicators but the results are not high and uneven (Nguyen Kim Anh and Le Thanh Tam 2013), the author considers the role of women's empowerment as a catalyst, better and more effectively control the performance of MFIs in Vietnam. The results of the research will help policy- makers set out solutions for sustainable development and improve the efficiency of MFIs in Vietnam so that these organizations can develop commensurate with their potential and play an important role in the national strategy of comprehensive finance in Vietnam in the future. 1.7. Thesis structure. To solve the research objectives of the topic, the thesis is structured with five chapters: - Chapter 1: Introduction - Chapter 2: Theoretical basis and related studies - Chapter 3: Research Methods - Chapter 4: Experimental research results - Chapter 5: Conclusions and policy implications 9 CHAPTER 2: THEORETICAL BASIS AND RELATED STUDIES 2.1. Related concepts 2.1.1. The concept of microfinance In essence, microfinance is a very special economic activity in the field of money - banking when it can combine a very harmonious combination between profit-seeking (to survive) and implementing and have an important social role (contributing to hunger elimination and poverty reduction). Microfinance is not purely a monetary undertaking like commercial banks or any other credit institution, nor does microfinance today constitute social activity such as the services offered by social policy banks. Microfinance is a service aimed at poor people but with interest rates high enough for microfinance to exist, and service beneficiaries are also able to accept it. In short, microfinance is a very important method of sustainable poverty reduction in the current period, especially in developing countries, where the number of poor people is still high. 2.1.2. The concept of microfinance institution Nguyen Kim Anh et al. (2013) also clarified the concept of MFI in terms of value attributes. An MFI is an organization with a developmental base on a non- exploiting basis that primarily serves the poor. Thus, in this view, even a non- governmental organization can be considered an MFI when carrying out microfinance activities as a core activity or having a separate department to control microfinance activities. 2.1.3. The role of microfinance Over the past 50 years, microfinance has made remarkable achievements, confirming its role in changing people's lives. In particular, the MFI is an element and plays an important role in the socio-economic development of rural areas (Helms, 2006; Hulme, 1996; Ledgerwood, 2006). In essence, MFIs play a role both financially and socially: In the financial aspect, through the process of providing financial services, MFIs perform the important functions of (i) mobilizing savings, (ii) reallocating 10 savings for investment, and (iii) facilitating trade in goods and services, becoming an effective tool to reduce poverty and increase income. Socially, MFIs create opportunities for rural people, especially the poor, to access financial services, increase their participation in community life in general, and enhance their social capacity. 2.2. Theoretical basis for the performance of microfinance institutions 2.2.1. Performance concept According to Berger and Mester (1997), the performance of MFIs is reflected in the relationship between revenue and cost of using resources or the ability to turn input resources into the best outputs in business activities. An enterprise is considered to be efficient if it reaches its maximum output in terms of optimal use of given inputs. 2.2.2. Measure the performance of microfinance institutions According to Berger and Humphrey (1997), Heffernan and Fu (2008), performance analysis of MFI often uses two main methods: method of analyzing financial indicators and method of analyzing data envelopes. 2.2.2.1. Method of analyzing financial indicators  Group of indicators reflecting profitability  Group of self-sustainability indicators 2.2.2.2. Methods of Data Envelopment Analysis Data envelope analysis is a method of determining the relative efficiency index based on comparing the distance of units with the best performing unit on the boundary. The advantage of this method is that it allows us to determine the overall performance index of each MFI and rank the effectiveness of MFI based on actual activities. This is also the best performance that an MFI is performing when compared to other MFIs. This information helps administrators evaluate the current performance of MFI and seek ways to improve and enhance the performance of MFI (Nguyen Viet Hung, 2008). Two main methods for empirically estimating the performance of MFI using the marginal efficiency analysis method are commonly used: parametric and non-parametric approaches (Nguyen Minh Sang, 2014). 11 2.3. Factors affecting the operational efficiency of microfinance institutions  Operating time of MFIs  Degree of access (Outreach)  Size and structure of capital (Capital)  Lending costs and loan portfolio risks  Work Performance  Customer growth 2.4. Theoretical foundations of women's empowerment and the impact of women's empowerment on the performance of microfinance institutions 2.4.1. Theory of women's empowerment in microfinance institutions 2.4.2. Analytical basis for empowering women in microfinance institutions 2.4.3. Theoretical basis of the impact of women's empowerment on the performance of microfinance institutions 2.5. Summary of related studies 2.5.1. Studies on the performance of microfinance institutions Studies to evaluate the performance of domestic and foreign MFIs are quite diverse. These include Majune et al. (2013), Njuguna (2013), El-Maksoud (2016), Afude (2017), Sufian (2006), Ferdousi (2013) 2.5.2. Studies on the factors affecting the performance of microfinance institutions Research on the factors affecting the performance of MFIs has received the attention of many domestic and foreign researchers. Foreign studies include: Schäfer & Fukasawa (2011), Dissanayake (2014), Ngo (2015), Abdulai & Tewari (2017). In addition to foreign studies, local studies have also sought evidence for factors affecting the performance of MFIs. Specifically: Research by Truong Quang Thong & Vu Duc Can (2017), Research by Dao Lan Phuong & Le Thanh Tam (2017). 2.5.3. Studies on the impact of women's empowerment on the performance of microfinance institutions. 12 Many important studies have been conducted to investigate whether microfinance has successfully achieved poverty reduction goals and empowered poor women (Cohen and Sebstad, 2001; Goyal, 2004; Somasekhar and Bapuji, 2005; George, 2014). Existing studies vary substantially in terms of baseline indicators, women's access to credit (number of borrowers per borrower) and the financial performance of MFIs. Somasekhar and Bapuji (2005) conducted an investigation into empowering women through self-help groups (SHG) in Andhra Pradesh that formed the microcredit network. The findings indicate that the participation of poor rural women in SHGs not only allowed them to meet their credit needs, but also led to common awareness, capacity building, confidence and courage that can be considered empowering. The micro-credit network promotes capacity building, the spirit of ownership of micro-enterprises, and the socio- economic improvement of women. 13 CHAPTER 3: RESEARCH METHODS 3.1. Research design This study applies the procedures of Abdulai & Tewari (2017), Lopatta et al. (2017) and at the same time combined with the studies of Dao Lan Phuong & Le Thanh Tam (2017), Ngo (2015), author identify and evaluate the factors that affect the performance of MFIs in Vietnam and the impact of women's empowerment on the performance of these MFIs through the following steps: - Step 1: Measuring the performance of Vietnamese MFIs on three aspects: self-sustainability, profitability, and allocation efficiency. Specifically, sustainability will be measured through OSS, profitability will be measured through ROA and ROE, and the distributional efficiency will be measured through: technical effectiveness (TE), scale effectiveness (SE) results from Data envelopment analysis (DEA). - Step 2: Based on the relevant studies of Abdulai & Tewari (2017), Lopatta et al. (2017) and concurrently with the studies of Dao Lan Phuong & Le Thanh Tam (2017), Ngo (2015). The model assesses the factors that affect the performance of MFIs in Vietnam and the impact of women's empowerment on the performance of MFIs. - Step 3: Collect data and estimate models - Step 4: Perform necessary tests - Step 5: Analyze, evaluate, and draw conclusions about the factors affecting the performance of MFIs in Vietnam and the impact of women's empowerment on the performance of MFIs. 3.2. Research Methods 3.2.1. Method of evaluating the performance of microfinance institutions in Vietnam. To evaluate the performance of MFIs, the study used Data envelopment analysis (DEA). This research method has been applied a lot in assessing the performance of traditional banking. However, in recent times, many researchers have expanded this method to apply to MFIs (Sufian, 2006; Ferdousi, 2013). 14 In this study, to evaluate the performance of MFI Vietnam, the author chooses the inputs and outputs of MFI based on previous studies, specifically: The input variables selected according to the research of Sufian (2006), Ferdousi (2013), Bolli et al. (2012) include the following two variables: - Operating expenses: including interest on deposits and equivalent, staff salaries, non-interest expenses. - Number of employees: includes all employees working at MFI. The output variables selected by Ferdousi (2013) include two variables that reflect the performance of MFI as follows: - Total outstanding loans: including all outstanding loans of customers. - Number of borrowers: including all outstanding customers at MFI. Table 3.1. Describe input and output variables of MFI in DEA analysis Variable Define Unit The input variables Operating expenses including interest on deposits and equivalent, staff salaries, non-interest expenses VND Number of employees includes all employees working at MFI Person The output variables Total outstanding loans including all outstanding loans of customers VND 15 Number of borrowers including all outstanding customers at MFI Person 3.2.2. Method of assessing factors affecting the performance of microfinance institutions in Vietnam To assess the factors affecting the performance of microfinance institutions in Vietnam, the author built a research model based on the main research model of Abdulai & Tewari (2017) combined with factors in the research of Dao Lan Phuong & Le Thanh Tam (2017), Ngo (2015) and Ngo et al. (2014). The general research model of the topic has the following form: Model (1): OSSit = β0 + β1OSSit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + vi + uit Model (2): ROAit = β0 + β1ROAit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + vi + uit Model (3): ROEit = β0 + β1ROEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + vi + uit Model (4): TEit = β0 + β1TEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + vi + uit Model (5): SEit = β0 + β1SEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + vi + uit Where, 16 Dependent variable: OSS is activity self-sustainability, ROA is net income over total assets, ROE is net income on equity, TE is the indicator of technical efficiency, and SE is the scale effectiveness. These indicators are used to evaluate the effectiveness of MFIs in this study. The targets often refer to the ability of MFIs to continually implement microfinance programs in pursuing their regulatory goals (Abdulai & Tewari, 2017; Dao Lan Phuong & Le Thanh Tam, 2017; Ngo, 2015). The independent variables include: AGE (MFIs' age) is the age of MFIs, CPB (Cost per borrower) is the cost per borrower, OEA (Operating expense to assets ratio) is the ratio of operating expenses to total assets, DER (Debt to equity ratio) is the ratio of equity to total assets, PAR30 (Portfolio at risk) is the risk ratio of the portfolio, NAB (Number of active borrowers) is the number of people real loan, GLP (Gross loan portfolio) is the total loan portfolio. In addition, because vi is a characteristic of MFI that cannot be observed, uit is a specific error. 3.2.3. Methods to assess the impact of women's empowerment on the performance of microfinance institutions in Vietnam To assess the impact of women's empowerment on the performance of microfinance institutions in Vietnam, the author adds to the above research models the independent variable PFB showing empowerment. Women are measured by the total number of women borrowing on the total number of MFI borrowers. The research model has the following form: Model (6): OSSit = β0 + β1OSSit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + β9PFBit + vi + uit Model (7): ROAit = β0 + β1ROAit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + β9PFBit + vi + uit Model (8): ROEit = β0 + β1ROEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + β9PFBit + vi + uit 17 Model (9): TEit = β0 + β1TEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + β9PFBit + vi + uit Model (10): SEit = β0 + β1SEit-1 + β2AGEit + β3CPBit + β4OEAit + β5DERit + β6PAR30it + β7NABit + β8GLPit + β9PFBit + vi + uit The method of measuring variables, expected marks and the basis for proposing variables is presented in the following table: Table 3.2. Describe the variables used in the model Variable Symbo l Measure Expecte d sign Base citation Dependent variable Self- sustainable operation OSS Income from operations Total operating expenses x100% Schäfer and Fukasawa (2011), Dissanayak e (2014), Ngo (2015), Dao Lan Phuong and Le Thanh Tam (2017), Abdulai and Tewari (2017) 18 Return On Asset ROA Net income Total asset x100% Dissanayak e (2014), Abdulai and Tewari (2017) Return On Equity ROE Net income Total equity x100% Dissanayak e (2014), Abdulai and Tewari (2017) Technical efficiency TE Calculated from DEA Scale efficiency SE Calculated from DEA Independent variables Women's empowermen t PFB The total number of women borrowed Number of borrowers + Abdulai & Tewari (2017), Lopatta et al. (2017) Equity ratio to total assets DER Total equity Total asset + Ngo (2015), Dao Lan Phuong and Le Thanh Tam (2017) 19 Total loan portfolio GLP Ln(Total loan portfolio) + Abdulai and Tewari (2017) The risk ratio of the portfolio PAR30 Net losses / Total outstanding loans - Schäfer and Fukasawa (2011), Dao Lan Phuong và Le Thanh Tam (2017), Abdulai and Tewari (2017) Number of active borrowers NAB Ln(Number of active borrowers) + Schäfer and Fukasawa (2011), Ngo (2015), Abdulai and Tewari (2017), Operating expense to assets ratio OEA Total operating expenses Total asset - Dissanayak e (2014), Abdulai and Tewari (2017) 20 Cost per borrower CPB Ln(Total operating expenses / Number of borrowers) - Schäfer and Fukasawa (2011), Dissanayak e (2014), Ngo (2015), Abdulai and Tewari (2017), The age of MFIs AGE Number of years of operation of the microfinance institutions at the time of study + Dao Lan Phuong và Le Thanh Tam (2017), Abdulai and Tewari (2017) 3.3. Estimation methods 3.3.1. Fixed-effects model (FEM) 3.3.2. Random-effects model (REM) 3.3.3. SGMM estimation method 3.4. Collect and process data.  Sample size: As a rule of thumb, the sample size must be at least 5 times the number of variables in the model (Hair et al., 2006). The experimental research model consists of a maximum of 9 variables, so the minimum sample size is 45 observations. With the panel data covering 26 MFIs collected from 2013 to 2017, the sample includes 26 x 5 = 130 observations and meets the requirement for conformance. 21  Methods of collecting and processing data: According to statistics of the State Bank of Vietnam as of June 30, 2019, there are four official MFIs, which are M7 Microfinance Institution Limited, Tinh Thuong Microfinance Institution Limited, and Thanh Hoa microfinance institution, Capital Aid Fund For Employment of The Poor. In addition to the official MFIs, there are 30 semi-formal MFIs in microfinance programs and projects operating in Vietnam (Microfinance Directory, 2018). However, the information of these MFIs is not complete, so the author conducted the study with the 26 most complete MFIs. Research data is the annual financial statements of 26 MFIs in Vietnam for the period 2013-2017 provided by MIX Market. MIX Market is a website operated by Microfinance Information Exchange (MIX). The MIX Market website allows microfinance programs to post information, including audited financial statements and performance indicators, to receive ratings based on the transparency of the information. Regarding the study time, the author conducted at 26 MFI in the period of 2013 - 2017. This stage was selected by the author to carry out the study because to ensure that 26 MFIs have enough data to calculate the variables in the research model. 22 CHAPTER 4: EXPERIMENTAL RESEARCH RESULTS 4.1. Actual situation of Vietnamese microfinance institutions:  Customer growth Unit: Person Figure 4.1: Number of customers of MFIs Vietnam from 2013 to 2017 Source: Mix market  Total outstanding loans Unit: VND billion Figure 4.2: Total outstanding loans to customers of MFI Vietnam in the period of 2013 - 2017 Source: Mix market  Number of employees 7,746,045 7,790,466 7,816,377 7,554,032 7,480,392 7,300,000 7,400,000 7,500,000 7,600,000 7,700,000 7,800,000 7,900,000 2013 2014 2015 2016 2017 160,445 168,590 146,196 162,880 180,614 - 50,000 100,000 150,000 200,000 2013 2014 2015 2016 2017 23 Unit: Person Figure 4.3: Number of employees of MFIs Vietnam from 2013 to 2017 Source: Mix market  Operating expenses Unit: VND billion Figure 4.4: Operating expenses of Vietnamese MFIs in the period of 2013 - 2017 Source: Mix market 11,750 12,634 12,934 13,169 13,469 10,500 11,000 11,500 12,000 12,500 13,000 13,500 14,000 2013 2014 2015 2016 2017 12,152 13,090 14,033 13,580 16,310 - 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 2013 2014 2015 2016 2017 24 4.2. Descriptive statistics of research samples and correlations between variables: Descriptive statistical results measuring characteristic quantities for the study variables are shown in Table 4.1. Table 4.1. Statistical results describe the variables in the model Variable Number of observations Mean Standard error Min Max Dependent variable ROA 130 .040561 .0430161 -.1781117 .1482542 ROE 130 .1279397 .1616091 -.5182695 1.069242 OSS 129 1.389881 .3978275 .3436663 2.900789 Independent variables AGE 130 12.76923 6.654879 2 27 CPB 130 13.32789 1.042988 10.5491 17.0967 OEA 130 .2770734 1.797112 1.11e-09 20.57519 DER 130 .4830661 .8091353 .0181009 9.066581 PAR30 128 .007266 .0233176 0 .1928 GLP 130 24.75121 2.498789 21.61813 32.68964 25 NAB 130 9.198301 1.973082 5.817111 15.77561 Source: Calculation results from STATA 15 software  Correlation coefficient matrix: Table 4.2. Matrix correlation coefficient Source: Calculation results from STATA 15 software  Multicollinearity test: Table 4.3. Multicollinearity test between independent variables Variable VIF 1/VIF GLP 3.47 0.288184 NAB 2.89 0.346021 CPB 2.87 0.348475 PFB 2.84 0.352234 se -0.0882 -0.1965 0.1130 -0.2926 0.2168 -0.1257 0.0802 0.0681 0.0420 0.3837 0.4443 0.6278 1.0000 te 0.0434 -0.1051 0.3378 -0.3710 0.3656 -0.2958 0.1259 0.0459 -0.0326 0.3704 0.4358 1.0000 nab -0.0850 -0.1036 -0.1455 -0.6683 0.5522 0.3400 0.0981 -0.2213 0.0487 0.9435 1.0000 glp -0.0955 -0.0751 -0.1833 -0.7655 0.5637 0.5568 0.0163 -0.2881 0.0047 1.0000 par30 -0.4061 -0.3390 -0.1783 -0.1197 -0.0134 0.0444 -0.0241 0.0379 1.0000 der 0.2219 -0.2268 0.4231 0.1153 -0.1661 -0.3544 0.0052 1.0000 oea -0.0416 0.0195 -0.0741 0.0242 -0.1496 -0.0855 1.0000 cpb -0.1372 0.0234 -0.3885 -0.4914 0.1937 1.0000 age 0.0635 -0.