Borrowers' characteristics and their impact on repayment behaviour in Sri Lanka. An application of discriminant and logistic models

Table of contents

Abstract

1.0 Introduction
1.1 Overview of SANASA Thrift, Credit and Cooperative Society (TCCS) banks
1.2 Literature Review
1.2.1 Repayment performance and its determinants: Binary logistic model
1.2.2 Repayment performance and its determinants: Linear discriminant function model
1.3 Research questions
1.4 Objectives of the study
1.5 Significance of the study
1.6 Methodology
1.6.1 Measurement of the variables
1.6.2 Analytical tools
1.7 Results and discussion
1.8 Conclusions and recommendations

References

Abstract

This paper identifies the factors that discriminate the borrowers into defaulters and non- defaulters as well as examine how the demographic, farming characteristics and attributes of the farmers affect on the repayment performance among microfinance borrowers in SANASA Thrift, Credit and Cooperative Society (TCCS) banks in Vavuniya district, Sri Lanka. The study involved with a cross – sectional data using multi - stage sampling technique and the relevant primary data were collected from a set of questionnaire and interview. A total of total of 113 small- holder farmers were selected randomly from five villages in the district. The collected data were analysed using independent sample t – test, chi- square test to compare the characteristics related to the demographic, farming activities and attributes among two groups of borrowers which categorize as defaulters and non-defaulters in the study. A linear direct and step wise discriminant function were employed to identify the characters which classify the borrowers into two groups and binary logistic model also used to analyze the demographic, farming characters and farmers’ attributes that influence the loan repayment and the credit worthiness of the borrowers in the district. The stepwise discriminant analysis show that age, income, farm size and family members were significant in discriminating between defaulters and non – defaulters and based on the results further it reveals that, 43% of the borrowers not paid fully whereas 57% of them fully paid which classify them as defaulters and non- defaulters respectively. Empirical results of the binary logistic regression model implies that, out of total fifteen explanatory variables were included in the model, four variables related to demographic characters were important in influencing loan repayment performance while all five farming characters were found to be significant for the probability of being defaulter. In addition to that, among four variables related to attributes of the farmers, three of them factors contribute to microcredit loan repayment in the study area.

Keywords: Repayment performance, Demographic and Farming characters, Defaulters and non –defaulters, Stepwise discriminant model.

1.0 Introduction

The Sri Lankan economy mainly determines on agricultural sector includes plantation and domestic sectors where those sectors play a major role in rural development and livelihood of the people in the country. Agricultural sector in Sri Lanka contributes nearly 7% to the Gross Domestic Product of the country as well as it provides nearly 32.6% of the employment opportunities to the labour force in the economy. (Central Bank Annual Report, 2017). Rural development of a country mainly depends on the growth of agricultural sector and the government or micro financial institutions have more concern on the development of agricultural sector by many ways. Specially, financial supports are necessary for farmers to invest in their farming which help them to improve their living standard. Thus, agricultural loans are the one of the source to improve and enhance living standard and their life style.

Today, microfinance is an accepted and a highly recognized approach for providing financial needs of the poor across the world. Poor farmers, micro enterprise operators, artisans and other poor groups face numerous challenges when obtaining credit for their financial requirements from other formal and informal sources and this scenario has led to evolving of this type of approach and operational strategies that fall in line with that. (Chandrasiri, J.K.M.D. & Bamunuarachchi, B.A.D.S, 2016). There are many Government institutions and institutional arrangements have been experimented to provide the loans especially to smallholder farmers for agricultural purposes. In Sri Lanka also numerous micro finance services have expanded since 1980s. Some of these have been initiated with the sponsorship of the government institutes and programmes while some have been initiated with the sponsorship of NGOs and INGOs. Samurdhi Banking Scheme, Govijana Bank Pilot Credit Scheme, Women’s Society Based Revolving Fund Credit Scheme, Bhagya and Isura Credit Schemes are some of the micro financing schemes/institutions initiated with the sponsorship of government institutions, banks and development programmes. Credit schemes such as SEEDS (Sarvodaya Economic Enterprises Development Scheme, JBS (Janashakthi Banku Sangam), SANASA and Gami Pubuduwa have been initiated by local NGOs and private commercial banks while the credit programmes operated by Seva Lanka and CARE have been initiated with sponsorship of foreign NGOs. (Chandrasiri, J.K.M.D. & Bamunuarachchi, B.A.D.S, 2016). However, the banks and financial institutions are facing the problems of lower repayment performance from the borrowers and the recovery rates by the banks are a serious issue for them. Some borrowers do not repay their loans in time and thus banks are unable to functioning well. In recent years the number of many financial institutions and got a chance to access the agricultural loans especially after the war in North and East provinces in Sri Lanka. In Vavuniya district, SANASA Thrift, Credit and Cooperative Society (TCCS) banks are providing loans to borrowers who mostly involves in farming activities.

1.1 Overview of SANASA Thrift, Credit and Cooperative Society (TCCS) banks

The Canadian Co-operative Association (CCA), which promotes co-operative development around the world, was asked to practice in a co-operative thrift and credit movement in Sri Lanka called SANASA (SANASA model -ISBN 0–88880–396–6). SANASA Thrift and Credit Co-operative Society is one of the important MFI, and a shining example of how sustainable co-operative development can help to alleviate poverty among poor people. SANASA helps to build a financial background of the poor farmers and provides a support to achieve their hopes. Especially than the other financial institutions the services are provided by SANASA TCCS is easy to access by the rural poor people in Sri Lanka. It was registered on 21.11.1986(303V/21.11.1986) under the co-operative Law. SANASA TCCS mainly focus to promote the living standard of rural people and to support to the poor farmers. Than the individual focus they mostly consider the societies, and through the groups they motivate poor people to practice the Micro Finance systems especially the compulsory saving for the members and provide farming loans and some special loans with the aim to promote the sustainable development among rural farmers.

Before receive the initial loan a new member must be get the membership of the society at least for three to six months, the person must regularly attend the meeting, he requires to have 10 per cent of the loan as a compulsory saving, and to have at least three or five team members as guarantee. Where the SANASA TCCS provides loan mostly on two types one is Farming loans (paddy loan) and other loans such as cow loans and special loans with the low level interest rate than the other financial institutions. In Sri Lanka as parallel the function of SANASA TCCS spreads eventually all around the country.

Vavuniya SANASA TCCS banks have nearly 240 co-operative societies and issuing millions of loan facilities to the Vavuniya farmers and recently they are facing the problems on repayment of loan within the time duration. The inability of the borrowers to repay the loans according to the loan terms will create number of problems to the borrowers and also. In this current scenario there is an instant need for the remedial actions have to be taken in order to reduce the number of defaulters among the farmers in SANASA TCCS and to instruct the borrowers to perform in a correct way to reduce the future burden. In this background, this study intends to identify the borrowers’ characteristics and its impact on the repayment behaviour. Moreover, the study examines the creditworthiness of the borrowers and classify them into two groups like defaulters and non- defaulters based on their characters in the study area.

1.2 Literature Review

There are many researches done by many researchers in many countries on the determinants of repayment performance and its determinants using different methods. Defaulting among the borrowers is the major problems face by the banks and financial institutions in many countries.

1.2.1 Repayment performance and its determinants: Binary logistic model

Mohammad Reza Kohansal, Hooman Mansoori (2009) examined the factors affecting on loan repayment performance of farmers in Khorasan-Razavi Province of Iran using a cross sectional data of 175 farmers of Khorasan-Razavi province in 2008.They used logit model and its results showed that, loan interest rate is the most important factor affecting on repayment of agricultural loans while farming experience and total application costs are the next factors in the study.

Shaik Abdul Majeeb PASHA and Tolosa Negese have evaluated the performance of loan repayment determinants in Ethiopian Micro Finance (2014) during 1994-95. Based on the analysis, researchers are recommended that proper training should be provided, reasonable amount of loan which should be useful to their business.

A Case of Dedebit Credit and Saving Institution done by Assfaw Kebede Asgedomat el to determine the MFIs Group Loan Repayment Performance in Ethiopia (2015). The study has applied binary logistic regression model to analyze the factors influencing group loan repayment performance and its estimated results proved that peer-monitoring, screening, peer-pressure, social ties, loan officer visit to the group, and other sources of credit were found to have statistically significant effect on the group loan repayment performance in the study.

Logistic regression analysis of predictors of loan defaults by customers of non-traditional Banks in Ghana examined by Edinam Agbemava at el (2016) to identify the risk factors that influence loan defaults by customers in the microfinance sector and to develop a model that links these factors to credit default by customers in the sector. A binomial logistic regression analysis was fitted to a data of 548 customers who were granted credit from January 2013 to December 2014 and their results revealed that marital status, type of collateral or security, assessment, duration and loan type were statistically significant in the prediction of loan default payment with a predicted default rate of 86.67% in Ghana.

Another study done by Mohammed Ameen Qasem Ahmed Alnawah, et al (2018) to identify the factors affecting repayment performance in Microfinance banks in Yemen and the data analysed using binary logistic model. Their major findings suggest that, the borrowers of large amounts default over the borrowers of small amounts and older borrowers are more defaulted. Also, compared to public sector employees, private sector employee borrowers are defaulted more in the study.

Anoher study done by Girma Gudde Jote (2018) to determinants the loan repayment in case of Microfinance Institutions in Gedeo Zone, Ethiopia, The collected data were analysed using binary logistic model and the results imply that, out of ten explanatory variables were included in the model, six variables namely educational level, method of lending, nearness of borrower’s residence to the institutions, family size, and income from activities financed by loan and training were found to be statistically significant to influence the probability of loan repayment in Ethiopia.

1.2.2 Repayment performance and its determinants: Linear discriminant function model

Godwin Chigozie Okpara at el (2013) has analyzed the loan repayment performance of small holder oil producers in Nigeria using a credit rating approach. The discriminant analysis technique was employed in analyzing the data and its results proved that, loan-asset ratio ranks top as the most critical positive discriminating variable for credit worthiness, followed by interest rate, income – expenditure ratio and age of beneficiaries as these variables crossed the 1 unit credit worthiness benchmark.

Using discriminant analysis another study done by Ajah, E. O. Eyo and S. O. Abang (2013) on repayment performance among cassava and yam farmers under Nigerian agricultural bank smallholder loan scheme in Cross River State, Nigeria. The results revealed that only 56% of the respondents were creditworthy and farmers with better educational level, larger farm sizes, and longer years of farming, proper loan supervision, and low total operating expenditure to income ratio were credit worthy farmers while farmers with lower loan to asset ratio were said to be non -credit worthy.

Repayment Performance among Cassava Farmer-Beneficiaries of Microfinance Institutions (MFIs) In Abia State, Nigeria (2015) analysed by K.C Obike and C.K Osondu. Data collected was analyzed using descriptive statistics, discriminant function analysis and OLS multiple regression model. The empirical results revealed that the highest average loan recovery rate of 86.0%, while the lowest average loan recovery rate was 78.0%.The result also revealed that 88 out of 90 MFI cassava farmer beneficiaries were grouped as worthy while 28 out of 30 were grouped as unworthy. The discriminant function analysis revealed that, cassava farmers with better education, larger farm sizes, good farming experience, extension supervision and low total operating expenditure to income ratio were credit worthy in the study.

Assessment of creditworthiness and repayment among bank of agriculture loan beneficiaries in Cross River State, Nigeria (2016) investigated by Nkem H. Justice, Atturo E. O. and Nwagbo Kingsley. . A total of 120 poultry farmers were used in the study and those data were analysed using discriminant analysis. The discriminant function revealed that, 51.7% of the respondents were credit worthy and also farmers with better educational level and large farm sizes were non credit worthy. While farmers with large total operating expenditure-income ratio, longer years of farming, older farmers with adequate supervision were credit worthy.

Nguyen Thuy Duong, Do Thi Thu Ha & Nguyen Bich Ngoc (2017) evaluated the application of discriminant model in managing credit risk for consumer loans in Vietnamese commercial bank. Their finding indicates that the estimated function is significant at 1 per cent level of significance and could forecast financial health with average 72.3 per cent accuracy. Also, they found that the demographic, socio-economic and loan related variables can be used to determine the expected group membership of the borrowers in Vietnam.

Okey F. Nwanekezie, Iheanyi J. Onuoha (2019) has investigated on homeowners’ perception of the factors affecting access to housing Finance in Owerri, Imo State, Nigeria. The study surveyed a cross-section of 450 respondents consisting 300 of those who had borrowed and 150 of those who did not borrow. The result showed a significant discriminant function separating the two groups based on their perception of the variables. The Wilks’ Lambda’s F - test and the standardized discriminant function coefficients, indicated that there are significant differences in perception between homeowners who borrowed and those who did not borrow.

1.3 Research questions

This study has the following questions:

1. What are the major borrower characters that influence to categorize them into two groups?
2. Do the demographic characters significantly influencing the repayment performance of the clients and classify them into defaulters and non-defaulters?
3. How the farming characters and attributes of the farmers affect on repayment behaviour and classify them into two different groups in the study?

1.4 Objectives of the study

The main objective of the study is to identify the borrower characteristics that discriminate them into defaulters and non- defaulters and examine the determinants of loan repayment and their credit worthiness in Microfinance institutions in Vavuniya district in Sri Lanka. In line with above general objective, this study has the following specific objectives:

1. To identify the borrower characters those classify them into defaulters and non-defaulters in the study area.
2. To evaluate the impact of major demographic characters such as age, gender, levels of education, civil status and family members of the borrowers that impact on their repayment performance and credit worthiness.
3. To investigate how the farming characters like income, farm size, ownership of land, farming experience and availability of non – farm income as well as farmers’ attributes such as purposes of loan, crop failure, weather conditions and knowledge about loans affect loan repayment and discriminate the borrowers into two groups in the study area.

1.5 Significance of the study

Financial institutions and banks have major role in financial sector as well as rural sector of an economy in terms of providing loans to the rural community in developing countries like Sri Lanka. The borrowers especially, farmers are able to get the loans from the microfinance institutions to improve their living standard through agricultural activities and generate their income. Even the borrowers have chances to receive the loans, the microfinance institutions and banks are facing the problems to recover the loans from the borrowers. Thus, default rate among the borrowers has been increasing over time which is the difficult task to manage the banks and financial institutions. There are a number of many factors particularly demographic and farming characters that affect the loan repayment rates. There has not been any empirical research conducted regarding to repayment performance among the borrowers who get the loans from SANASA Thrift, Credit and Cooperative Society (TCCS) banks in Vavuniya district. Therefore, this study tries to provide the relevant information for a better understanding on the determinants of loan repayment performance of the borrowers and the information will be useful for policy makers, other lending institutions and stakeholders for their future decision making on granting the loans for their clients.

1.6 Methodology

The study involved with a cross – sectional data with multi - stage sampling technique and the relevant primary data were collected from a set of questionnaire. At the first stage, Vavuniya district was selected purposely from Sri Lanka where the district has 4 Divisional secretariats (DS) divisions. In the second stage, from the above 4 divisions, only Marukkampalai Grama Sevaka (GS) division was selected. This division has many villages and out of them 5 villages were selected where the major farmers cultivating the crops and getting loans from the SANASA Thrift, Credit and Cooperative Society (TCCS) bank during the period of 2018 December to 2019 January. Finally, from the five villages, a total of 113 small- holder farmers were selected randomly.

1.6.1 Measurement of the variables

The relevant information and data were gathered through a set of questionnaire and both dependent and independent variables contain categorical, binary and scale data.

Dependent variable

Dependent variable is the repayment performance which was categorized as 1 for non-defaulter and 0 for defaulter using binary variables in the study.

Dependent variables

The borrower characters considered as independent variables which were divided into three set as, demographic characteristics, farming characteristics and attributes of the farmers.

Demographic characteristics

Age: Age of the borrowers which is a continuous variable measured in years and it is expected that the borrowers who are young may have more energy, productivity and interest on adopting new techniques which may leads to lower default rate. However, older farmers also may have lower defaulting rate because of their maturity and experience enhance their productivity and improve their repayment performance.

Gender: Gender is a dummy variable defined as 1 for male and 0 for female and its impact on default cannot be expected. In some cases it is argued that, lending to female borrowers can enhance them to repay the loans because of their economic empowerment, culture and discipline than male borrowers. Although, male borrowers have more energy and their possibility of hard work may raise their productivity and thus increase the non – defaulting.

Education: Education of the borrowers was measured by ordinal variable where 1 for primary, otherwise 0 and 1for secondary otherwise 0.This variable is expected to have positive impact on repayment rate where the higher educational levels encourage the farmers to repay the loans than uneducated borrowers.

Civil status: it was measured by binary variable where 1 for single and 0 for married and it has unexpected impact on dependent variable. In some cases single borrowers have more probability to repay the loans than married persons. Because single farmers have less financial burden compared to married farmers. But, when married farmers have financial support they also may have more possibility to settle the loans.

Number of family members or household size: It is a numerical variable measured in numbers which expected that positive impact on repayment performance. Larger the household size will have less probability to settle the loans while smaller the household size has higher the repayment possibility.

Farming characteristics

Income: Income of borrowers earns from farming is a continuous variable measured in rupees and it is expected that the higher income motivates the borrowers to repay the loans on time and the borrowers who have more income they belong to non – defaulters. Similarly, the borrowers who have less income they belong to defaulters.

Farm size: It is a continuous variable measured by the acre of cultivated land and assumed that it has positive impact on borrowers’ repayment performance. The borrowers who have more size of land, they belong to non-defaulters than small size land holders.

Ownership of land: This is variable measured by binary as 1 for the farmers who cultivate the crops using own land and 0 for the farmers cultivate using tenant land. The expected outcome of the variable is positive which means that, own land farmers have more likely to repay the loans than tenant farmers.

Farming experience: This variable also continuous variable and also expected that the farmers who have more experience in farming, repayment rates also will be higher than less experience farmers.

Availability of other sources of income or non-farm income: It is a dummy variable coded as 1 for the borrowers who have other income sources and 0 for the borrowers who do not have it. This variable expected to have positive impact on repayment performance where the borrowers who have other income sources may belong to non-defaulters and if not they belong to defaulters.

Farmers’ attributes

Purposes of loan: This variable is a categorical one where 1 for the loans got agricultural purpose and 0 for other purpose. It is expected that the borrowers who receive the loans for agricultural purpose, their non- defaulting is lower than other lenders.

Crop failure: This variable defines as whether the crop failure affects the repayment behaviour or not which is coded as 1 for yes otherwise 0. It is usually has negative impact on repayment ratio where the farmers facing the problem of crop failure they may belong to defaulters and if they do not have such problem, they are then on- defaulters.

Weather conditions: The variable also is measured by binary variable where 1 for weather conditions affect the borrowers’ repayment behaviour whereas 0 for not. It is expected to have negative impact on repayment performance where the farmers says “yes”, the defaulting probability will be higher and if they says “no”, probability of non- defaulting will be higher.

Knowledge about loans: The borrowers asked prior knowledge about loans affect their loan repayment attitudes or not using binary variable where 1 for they have and 0 for not have. It is assumed there is a positive impact where the borrowers have knowledge, they become as non – defaulters an otherwise defaulter.

1.6.2 Analytical tools

To examine the demographic and farming characteristics as well as attributes of the farmers, basic statistical analysis such as, independent sample t – test and chi- square test were applied in the study. Further, to classify the borrowers whether they belong to defaulter or non- defaulter group, discriminant analysis also employed. The impact of demographic characteristics, farming features and attributes of the farmers, binary logistic model is used which is more appropriate than other regression model.

Independent sample t -test

Independent sample - t test is more useful to examine whether is there any mean differences in selected demographic and farming variables across defaulters and non- defaulters in the study. Based on the results the researcher can decide those two groups are similar or not and if the mean values differ from each other, and then there are two different groups available. Age, family members, income, farm size and farming experience were used in the independent sample t- test.

Chi – square test

This test is more applicable where the variables are categorical in nature in order to access the association between them. The selected demographic and farming characters were taken to identify the association between defaulter and non- defaulter in the study.

Discriminant analysis

Discriminant analysis is a technique for analyzing data when the criterion or dependent variable is categorical and the predictor or explanatory variables are measured on at least interval scales. Thus, discriminant analysis is a multivariate statistical technique used to describe which variables discriminate between two or more groups and as the criterion variable has two categories, the technique is known as two- group discriminant analysis. Since the borrowers have two group like defaulters and non-defaulters, two- group discriminant model used at two methods, direct and stepwise methods in the study. Therefore, discriminant analysis was applied to classify the demographic and farming characteristics into defaulters and non- defaulters based on repayment performance.

The discriminant function is specified as:

Abbildung in dieser Leseprobe nicht enthalten

Where,

Z = Total score on the discriminant function

X1 = Age in years

X2 = Number of family members

X3 = Income in Rs

X4 = Farm size in Ha

X5 = Farming experience in years

In discriminant analysis, the outcome variable should be categorical while explanatory variables better to have in terms of scale data. To classify the borrowers into two separate groups based on certain characters the following conceptual model used in the analysis.

Abbildung in dieser Leseprobe nicht enthalten

Source: Developed by researchers based on literature reviews

Figure 01: Conceptual model in discriminant analysis

The estimation of the derived discriminant function for the study was done using major statistical analytical tools such as Wilks’ Lambda which measure the goodness fit of the model, group centroids which estimates the cut score or cut-off and the standardized canonical discriminant function coefficient measures the relative contribution of each independent variable to the total score of the function.

Binary logistic model

In order to evaluate the impact of features related to demographic, farming and attributes of the farmers on repayment performance and thus whether the borrower belongs to defaulting or non- defaulting, logit model was employed.

Since the dependent variable is a binary variable with 0 and 1, logit model is more suitable which usually estimate based on the method of maximum likelihood. The coefficients of the model could be interpreted in terms of log of odds ratio as well as probability. The cumulative distribution of the logistic model can be expressed as:

Abbildung in dieser Leseprobe nicht enthalten

If convert it into logarithm form, then,

Abbildung in dieser Leseprobe nicht enthalten

When the error term is taken into account, the logit model become as:

Abbildung in dieser Leseprobe nicht enthalten

Where,

Z = Loan default which was coded as 1 for non- defaulters and 0 for defaulters.

Xi = Independent variables.

α and βi are the parameters to be estimated and ε is the error term in the model.

Finally, the modified version of the binary logistic model for three characteristics is defined by:

Model 01:

Abbildung in dieser Leseprobe nicht enthalten

Where,

Y = Loan default where the borrower is non- defaulter coded as 1 and where the borrower is defaulter coded as 0.

X1 = Age of the borrowers

X2 = Gender

X3 = Primary education

X4 = Secondary education

X5 = Civil status

X6 = Number of family members or household size

ε = Error term

[...]