This study contributes to the debate on the effect of Lending Concentration (LC) on credit risk. It is based on data from an online survey of 151 employees from 37 lending institutions, employees with diverse experience in the different sections of their institutions. Following a successful three-factor solution on LC based on an Exploratory Factor Analysis, Binary logistic regression models were implemented to determine the Perceived Lending Risk (PLR) based on three types of LC, namely Social Status Lending Concentration (SSLC), Private Sector Lending Concentration (PSLC) and Public Employee Lending Concentration (PELC). It was noted that over- concentration based on social status provided an explanation for the increase in Non-Performing Loans (NPL) risk among both large and small lenders. Since Lending Concentration reverses the effect of macroeconomic variables, such as credit risk management practices (CRMPs), Credit Processing Considerations (CPCs), as well as collateral types, assessing the degree at which the lender is concentrated across sectors is imperative, prior to any credit risk management initiative. Although LC directly affects lending risks perceptions alongside the traditional and corporate finance theories, the indirect LC effect via CRMPs, CPCs, bank size and originality, as well as the various collateral typologies seems to provide new insights into this area of research.
to agency conflict between equity holders and creditors of the lender (Adzobu et al., 2017; Paravisini et al., 2015; Beck & de Jonghe, 2013; Acharya et al., 2006). Beck and de Jonghe (2013) however, suggests that LC causes higher lending risks among Banks. To date there is no consensus despite the acknowledgement that LC increases information at the cost of higher portfolio risk, thus may not necessarily yield overall lower Perceived Lending Risk (PLR) among lenders (Berger et al., 2017). At least two theoretical perspectives have been used to explain LC risks facing financial institutions. The first pertains to the standpoint of Traditional Portfolio Theory (TPT), which suggests that diversification, whether across lending sectors or name/products, and services, provide an opportunity for lenders to mitigate PLR. From the perspective of the TPT, LC is susceptible to the economic volatility of the sectors of LC than when they are well diversified (Adzobu et al., 2017). The TPT’s suggestions are in favor of diversification, principally drawing from Markowitz (1959) and supported by Diamond (1984). The second theoretical perspective pertains to a tenet of Corporate Finance Theory (CFT) which claims that most of the LC effects accrue from the built-in nurturing of expertise (Francisa et al., 2018). According to the CFT, lenders should specialize by concentrating their activities on specific sectors, or lines of business in order to enjoy the comparative advantage of developing expertise in the areas of their focus (Jensen, 1986; Denis, Denis, & Sarin, 1997; Adzobu et al., 2017). The CFT has suggested that LC can also result in better screening of potential borrowers and loan applications, and there is more efficiency in monitoring, hence leading to a lower NPL risk (Beck & de Jonghe, 2013). LC gives room for experienced expertise to develop, thus facilitating detection of deteriorating businesses early and taking the necessary appropriate measures (Adzobu et al., 2017; Böve et al., 2010). Similarly, it might be possible to prevent risk shifting by borrowers (Beck & de Jonghe, 2013). This study intends to contribute to the debate on the effect of LC on NPL risks by evaluating lender's perceptions on lending risks and the associated degree of LC. Despite the abundant literature on LC and diversification in the developed world, little is known with regard to developing countries, mainly because of the problem of limited data (Abu Hussain & Al-Ajmi, 2012; Adzobu et al., 2017). For example,
Tanzania experienced a credit crunch and lenders suffered from huge losses as a result of the high NPL rates (Richard et al., 2008). Similarly, lending has been concentrated more on the secondary rather than primary sectors (Kira, 2013), whereby the incorporated companies have more chances of getting loans than the unincorporated ones. With regard to the CRMPs, the Bank of Tanzania (BOT) requires each lending institution to prepare a comprehensive Risk Management Programme (RMP) designed to its needs and circumstances under which it operates (BOT, 2010). Despite the treatment provided through the CRMPs, the BOT (2019) has reported that the NPLs ratio declined to 9.8 percent from 10.4 percent, which however, was still above the desirable level. Since the Tanzania’s lending market has been highly concentrated (IMF, 2018), it is possible that LC could be responsible for the persistently higher NPL rates. Efforts to curb the NPL through the CRMPs and credit considerations in Tanzania seemed to be massive, but the NPL’s responses seemed to be very marginal. This study is carried out under the assumption that the NPL does respond to certain specific lending considerations and the CRMPS, but not to all of them despite the fact that all are intended to lower the NPL risk. LITERATURE REVIEW Sector Lending Concentration Sector specific observations in Vietnam and Lithuania have suggested that LC in sectors such as agriculture, manufacturing, transport and communication, construction and real estate and other communitywide sectors (electricity, gas and water) has been associated with higher Perceived Lending Risk (PLR), compared to service sectors such as housing, hotels and restaurants (Le & Diep, 2020; Pham & Lensink, 2008; Skridulytė & Freitakas, 2012). Simpasa & Pla (2016) provided evidence that for sectors where commercial lenders had high LC, i.e., Private Sector Lending Concentration (PSLC), the corresponding NPL risk tended to be lower to reflect monopoly rents accruing from superior information and monitoring efficiency or expertise in a given sector. These observations though contradictory, have pointed to the hypothesis that PLR might be low in the PSLC than in other forms of LC.
Social Status Lending Concentration (SSLC) describes LC in terms of individual characteristics (Hunter & Nixon, 1999). Theoretically, the existence of at least one major customer/borrower who accounts for 10 percent or more of the supplier’s induces a high PLR among lenders (Liu et al., 2019). Beck and de Jonghe (2013) observed that individual lender differentiation would contribute to more systemwide risks in the financial system. Under the SSLC, borrowers might transfer a part of their businesses’ achieved results to reduce the financial expenses associated with loan repayment (Laib, 2013). Similarly, in some western countries an individual’s share of the NPL would be determined by external factors, i.e., institutional quality or lack of adequate CRMPs by lenders, which could contribute further to a higher NPL risk (Hunter & Nixon, 1999; Skridulytė & Freitakas, 2012; Liu et al., 2019). Pham and Lensink (2007) has suggested that in developing countries, an individual’s share of the NPL might be linked to informality, whereby formal employment borrowers were more likely to default than their informal counterparts. These notable findings seem to suggest support for the hypothesis that the SSLC could be associated with a high or low PLR depending on the context. The Public Employee Lending Concentration (PELC) on the other hand, may be considered the safest given the certainty in income flow (Le & Nguyen, 2018; Pham & Lensink, 2008). Liu et al. (2019) and Yang (2017) however observes that, LC is associated with higher interest rate among bondholders to reflect the need to compensate for a higher PLR. Instead of reducing loans, lenders are more likely to increase monitoring through a shorter-term debt when facing a higher PELC (Yang, 2017). Borrowers’ occupation in lending (lenders) also reduces default risk (Pham & Lensink, 2008). As such, it is hypothesized that the PELC is associated with a relatively lower PLR, compared to other sectors. The Macroeconomic Environment In addition to LC, macroeconomic conditions may contribute to increasing the NPL risk (Beck & de Jonghe, 2013; Trautmann & Vlahu, 2013). The channel through which LC can lead to a higher PLR can be via loan losses and difficulties in managing interest rate risk (Paravisini, Rappoport, & Schnabl, 2015; Beck & de Jonghe, 2013; Le & Diep, 2020). Lassoued (2017) notes that while GDP growth generally reflects higher income flows for small borrowers and an increase in the profitability of firms, and therefore, better solvency, higher inflation rates constrain households in their consumption freedom. Similarly, both the GDP and inflation experience affect the NPL because of either lost lending relationship in times of low growth (Trautmann & Vlahu, 2013), or weak institutions or repeated crises where the contagious aggravation of crises are through wordof-mouth (Trautmann & Vlahu, 2013; Iyer & Puri, 2012). In terms of perceptions, psychological studies in the so called ‘Pollyanna Principle’ (Dember & Penwell, 1980), have well established that even with limited information, better fortunes are often perceived better (overstated) while misfortunes are perceived worse (understated) (de Meza & Southey, 1996). Therefore, a booming economy might signal a higher repayment ability, thus attracting more borrowing even where actual repayment is not guaranteed. Lender Specific Characteristics Kira (2013), Al-kayed (2020) and Abu Hussain & Al-Ajmi (2012) have suggested that lender typologies (Islamic and non-Islamic) determine the type and level of credit risk facing a financial institution. Furthermore, relatively larger lenders tend to experience a higher NPL risk from loan loss provisions (Adzobu et al., 2017). However, if a large lender is diversified rather than concentrated, the ability to grab economies of scales is enhanced, leading to a relatively lower PLR than that achieved by smaller lenders (Beck & de Jonghe, 2013). In contrast, Simpasa and Pla (2016) have suggested that there would be a higher NPL risk in response to LC, regardless of a lender’s size. In addition, in another study by Lassoued (2017), it was suggested that lender size had an insignificant effect on the PLR. The effect of bank size based on these studies could be either positive or negative, it depended on other factors. Similarly, the experience of the lender might affect risk perceptions. From the viewpoint of institutional memory hypothesis, whereby the less experienced key decision makers are, the higher the likelihood of lending risks underestimation (Burakov, 2014; Berger & Udell, 2004). Credit Processing Considerations A number of credit processing considerations may be pivotal in reducing the PLR. The first is existing debt (CDebt), which however, is rarely considered important to Microfinance Institutions (MFIs) and
Small and Medium Enterprises (SMEs) who rely on soft information (Abdulsaleh & Worthington, 2016; DeZoort, Wilkins, & Justice, 2017; Lassoued, 2017; Harif, Hoe, & Zali, 2011; Berger, Minnis, & Sutherland, 2017). The second is income, capacity and savings, all of which captures the ability of the borrower to meet the periodic repayment of the interest and principle (Adzobu et al., 2017; Laib, 2013; Abdulsaleh & Worthington, 2016; Koomson et al., 2016). There were cases however, where these capacity-related considerations had not been significant (Lassoued, 2017). With regard to income (CIncome) the PLR tended to be relatively lower among high-income borrowers than low-income ones, thus providing an explanation for credit refusal among the poor (Le & Nguyen, 2018; Pham & Lensink, 2008; Koomson et al., 2016); and Credit history (CredH) (Abdulsaleh & Worthington, 2016). Additional considerations included concerns with collaterals (CColla) and business plan (CBplan) (Abdulsaleh & Worthington, 2016). Both of these reflected the applicant’s creditworthiness and borrower’s character, especially in the context of lending to highly risky sectors, such as the MFIs and SMEs (Abdulsaleh & Worthington, 2016; Laib, 2013; Pham & Lensink, 2007). Other concerns being demographics, such as age whereby, the older are considered less risky (Koomson et al., 2016; Kira, 2013); the number of dependants, i.e., the more the number of dependants, the higher the PLR (Koomson et al., 2016; Pham & Lensink, 2007); and education, thereby those with a vocational or university degree would be associated with a low PLR (Kira, 2013). Further considerations include the purpose of the loan (LoanPur) and borrower’s employment status (CEmploy). Studies by Le and Nguyen (2018), Klyuev (2008) and Harrison et al. (2004) have suggested that the PLR depended on the stability of income and borrowers’ business experience. It was found that those with higher business experience were associated with a low PLR (Kira, 2013), though Abdulsaleh and Worthington (2016) noted that if the loan was for a startup, experience was irrelevant. Credit Risk Management Practices Lenders mitigate credit risk by using several practices, including the following: Risk-Based Pricing (RBP) to diversify risk when lenders charge a high interest rate on risky borrowers (Wood & Kellman,
2010); impose Different Bank Covenants (DBC) to borrowers to reflect borrowers’ creditworthiness (Choppari & Rajeshwar, 2015; Liu et., 2019; Campello and Gao, 2017); use of Credit Insurance for Loan Granted (CILG) to transfer risk from the lender to the seller (insurer) in exchange for payment (Prasad, 2016); the use of Credit Reference Bureau (CRB) to assess credit history of borrowers (Ondabu, 2019); Credit review (Crev) of borrower income to assess current creditworthiness; bank Guarantee to Cover Debt (GCD) in case of default; Unlock loan of potential borrowers upon request (LPB) (Seyram, 2013); transfer of loan to another bank; offer private banking to specific borrowers; and differential consideration of individual borrowers based on their employment status. The last three have however not been investigated with regard to its PLR effect. Different Collateral Types Provision of collateral for many lenders is a policy issue, without it the loan application is most likely to be rejected (Abdulsaleh & Worthington, 2016). Different types of collateral are accepted by lenders including the following: real estate, machinery, equipment and vehicles, products and goods (inventory), personal guarantees and accounts receivable (Abdulsaleh & Worthington, 2016). Banks require a collateral in order to address any adverse selection (Chan & Thakor, 1987; Stiglitz & Weiss, 1981), and the moral hazard problem. It is a means to monitor borrowers’ behavior and ensure the success of the project for which the loan was granted (Aghion & Bolton, 1992). In rare cases, the collateral may indicate the purpose of the loan, thus increasing the safety margin of the lender (Abdulsaleh & Worthington, 2016). The NPL risk reduction effect of these different types of collateral is however, a matter of empirical observations. Theoretical Determinants of Perceived Lending Risks The conceptual model for this study is as provided in Figure 1. Figure 1 shows that enhancing the intensity of the use of the CRMPs have a positive and significant impact on the PLR, while reducing such intensity increases the lender’s PLR. LC also shapes the PLR, however, its effect is not direct as LC is moderated by multiple exogenous variables. Thus, enhanced lender’s LC in a particular sector leads to significant PLR reduction, only if such LC is also associated with favorable perceptions on the Macroeconomic Environment (ME), Borrower Differentiation (BD) via credit processing considerations and Bank/Lender Characteristics (BC).
Conceptual Model of Determinants of Lenders’ Perceived Lending Conceptual Model of Determinants of Lenders’ Perceived Lending Risk (PLR) Risk (PLR) Perceived Lending Risk (PLR)
Credit Processing Considerations (CPCs) Credit Risk Management Practices (CRMPs)
Customer Lending Concentration (LC)
Borrowers’ differentiation (BD)
Macroeconomic Environment (ME)
Credit Risk Management Practices (CRMP)
Lender characteristics
METHODOLOGY METHODOLOGY
To relationships 1, anwas online survey To evaluate evaluate thethe relationships portrayedportrayed in Figure 1,inanFigure online survey distributed to was distributed to lenders via 2020 Google forms August 2020 lenders via Google forms from August to March 2021.from The contact details of allto respondents wereThe obtained based on online social media, specifically LinkedIn and the March 2021. contact details of all respondents were obtained based personal contact details of the authors. The study also utilized a limited secondary data, on online social media, specifically LinkedIn and the personal contact specifically macroeconomic data from the World Bank Development Indicators to compute details authors. The study also utilized a limited secondary data, the GDP of andthe inflation experience. specifically macroeconomic data from the World Bank Development Data Description Indicators to compute the GDP and inflation experience. By 2015, Tanzania had 63 banks and other financial institutions, of which 36 were Regional and Co-operative banks, three (3) were financial institutions, three (3) were Microfinance banks, two (2) were Development Financial institutions, (1) was the Mortgage Company, three (3) were By 2015, one Tanzania hadTanzania 63 banks andRefinancing other financial institutions,
Data Description Commercial banks, 12 were of which 36 were Commercial banks, 12 were Regional and Co7 operative banks, three (3) were financial institutions, three (3) were
Microfinance banks, two (2) were Development Financial institutions, one (1) was the Tanzania Mortgage Refinancing Company, three (3) were Finance Leasing, one (1) was a Representative bank and two (2) were Credit reference bureau11. Major commercial banks, such as the NMB Bank, CRDB Bank Ltd and Akiba Commercial Bank (ACB) have established considerable presence in rural productive sectors such as in agriculture. The TPB Bank has a countrywide network of post offices, which could facilitate savings mobilization and money transfers. In addition to banks and financial institutions, Tanzania also has several MFIs, which are registered by different authorities and legislations22. The total number of employees in these banks, financial intermediaries and Micro-Finance Institutions (MFIs) is however, not determined for the purpose of this study. Hence a convenient sampling strategy was adopted. Similarly, since the focus of this study was on the experience of the lenders’ operatives through employees, it was necessary to capture the experience of the employees, given their current and previous employments. In sum, the experience of the respondents in the lending industry was of prime importance in this study. The questionnaire instrument was divided into four major parts; the first intended to capture basic information about the respondents and lenders, such as their work experience and in which types of employment departments. The second part captured the CRMPs in detail, using a five point Likert scale to gather the associated reasons for practice variation across lenders. The third part paid some specific attention to practices such as the CRB and loan insurance33. The fourth part captured considerations in granting loans, risk mitigation strategies, as well as LC across sectors. The last part captured the collateral types relevant to credit processing. To maximize the probability of getting new insights into the dynamics of the lending market in Tanzania, this study has shown a preference for more than one respondent per lender, with varying experience, although several non-bank financial institutions had provided only a single response. When the survey was ended, a total of 151 respondents from 36 lending institutions had returned the duly completed questionnaire. Out of the 36 lending institutions, 33 were banks while four were non-bank financial institution. The detailed classification of the sample of financial institutions and the number of respondents is provided in Appendix 2, while a summary of the data used for this study is provided in Appendix 1.
SECTOR SECTORLENDING LENDINGCONCENTRATION CONCENTRATIONAND ANDCREDIT CREDITRISK: RISK LENDING CONCENTRATION AND CREDIT RISK: SECTOR LENDING CONCENTRATION AND CREDIT RISKA DataSECTOR Analysis EVALUATION EVALUATIONOF OFLENDERS’ LENDERS’PERCEPTIONS PERCEPTIONSIN INTANZANI TANZAN The framework in Figure OF 1 was operationalized through a four-stage EVALUATION LENDERS’ PERCEPTIONS ININ TANZANIA EVALUATION OF LENDERS’ PERCEPTIONS TANZAN analysis: i) The preliminary LC and PLR indices were computed based 𝐴𝐴𝐴𝐴𝐴𝐴 ⁄𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 on𝐼𝐼 𝐼𝐼the mean transformed scores of the Likert scale indicators, 𝐴𝐴𝐴𝐴𝐴𝐴 𝑖𝑖 ⁄𝑖𝑖𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑖𝑖 𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 …………………………………………………………………...... 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 …………………………………………………………………...... ሾͳሾ 𝑖𝑖 = 𝑖𝑖=1 𝑁𝑁 𝑖𝑖 =∑∑ 𝑖𝑖=1 𝑁𝑁 𝐴𝐴𝐴𝐴𝐴𝐴 ∑ ∑⁄𝑁𝑁 ⁄ 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 ∑ ∑𝑁𝑁 𝐴𝐴𝐴𝐴𝐴𝐴 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 and on equation 1. 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑛𝑛=1 𝑛𝑛=1 𝑖𝑖𝑖𝑖 𝐴𝐴𝐴𝐴𝐴𝐴 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑛𝑛=1 𝑛𝑛=1 ⁄ 𝐴𝐴𝐴𝐴𝐴𝐴 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑖𝑖 ⁄𝑖𝑖𝑖𝑖 𝑖𝑖 𝐼𝐼 𝐼𝐼 based 𝑖𝑖 𝑖𝑖 ∑ 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 = …………………………………………………………………...... ∑𝑖𝑖=1 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 …………………………………………………………………...... ሾͳሿ 𝑖𝑖 𝑖𝑖 =𝑖𝑖=1 𝑁𝑁 𝑁𝑁𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 ∑𝐼𝐼𝐼𝐼𝑁𝑁 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 ⁄∑ ∑𝑁𝑁 ⁄∑𝑛𝑛=1 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑛𝑛=1 𝑖𝑖𝑖𝑖𝑛𝑛=1 𝑛𝑛=1 𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 ⁄𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 Where Where [1] 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = � 𝑁𝑁𝑁𝑁 Where ∑𝑖𝑖𝑖𝑖=1 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ⁄∑𝑁𝑁𝑁𝑁 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 Where 𝑖𝑖𝑖𝑖=1 𝑖𝑖𝑖𝑖=1 𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴𝐴𝐴 the𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑖𝑖 𝑖𝑖 Where 𝑖𝑖 𝑖𝑖isisthe 𝐴𝐴𝐴𝐴𝐴𝐴 is the 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 𝑖𝑖 is the Actual factor score 𝐴𝐴𝐴𝐴𝐴𝐴 is the 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 𝑖𝑖 Maximum possible score for indicatoor 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 is the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for indicator 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 is the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for indicator𝑖𝑖 𝑖𝑖 𝑖𝑖 𝑖𝑖 SECTOR LENDING CONCENTRATION N is the total number of respondents in the sample (max 151) CREDIT RISK: A 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 forfor indicator 𝑖𝑖 𝑖𝑖AND 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 is the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 indicator 𝑖𝑖 is𝑖𝑖 the SECTOR LENDING CONCENTRATION AND CREDIT RISK: 𝜋𝜋 𝜋𝜋 𝐿𝐿𝐿𝐿 ( ( 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖) )==𝛽𝛽0𝛽𝛽0++𝛽𝛽1𝛽𝛽𝑀𝑀𝑀𝑀 𝛽𝛽2𝛽𝛽𝐵𝐵𝐵𝐵 𝛽𝛽3𝛽𝛽𝐵𝐵𝐵𝐵 𝐿𝐿𝐿𝐿 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖==𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑗𝑗 𝑗𝑗++ 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖++ 𝑖𝑖𝑖𝑖 …………………………………………… 1 𝑀𝑀𝑀𝑀 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑖𝑖𝑖𝑖 …………………………………………… ሾʹሾ 1−𝜋𝜋 1−𝜋𝜋 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝜋𝜋the EVALUATION OF LENDERS’ PERCEPTIONS IN TANZANIA ii) In second stage, Exploratory Factor Analysis (EFA) was 𝜋𝜋 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖𝐿𝐿𝐿𝐿= 𝑙𝑙𝑙𝑙 (SECTOR + 𝛽𝛽1 𝑀𝑀𝑀𝑀 𝛽𝛽3 𝐵𝐵𝐵𝐵 ሾʹሿ ( ) = ) 𝛽𝛽=0 𝛽𝛽 𝛽𝛽1 𝑀𝑀𝑀𝑀 +2 𝐵𝐵𝐵𝐵 𝛽𝛽2 𝐵𝐵𝐵𝐵 𝛽𝛽3 𝐵𝐵𝐵𝐵 𝑗𝑗 +𝑗𝑗 𝛽𝛽 𝑖𝑖𝑖𝑖 + 𝑖𝑖𝑖𝑖 …………………………………………… LENDING CONCENTRATION AND CREDIT RISK 𝑖𝑖𝑖𝑖 = 𝑙𝑙𝑙𝑙 0+ 𝑖𝑖𝑖𝑖 + 𝑖𝑖𝑖𝑖 …………………………………………… 1−𝜋𝜋 1−𝜋𝜋 𝑖𝑖𝑖𝑖 used to identify three perceived Lending Concentration (LC) EVALUATION OF LENDERS’ PERCEPTIONS IN TANZANIA 𝑖𝑖𝑖𝑖 Where, Where,sectors – i.e., SSLC, PSLC and PELC based on responses on Where, EVALUATION OF LENDERS’ PERCEPTIONS Where,a total of 15,𝛽𝛽 +𝛽𝛽 five𝑀𝑀𝑀𝑀point Likert scale questions on the level forIN TANZAN +𝛽𝛽 +𝛽𝛽 0𝛽𝛽0 +𝛽𝛽 1 1 𝑀𝑀𝑀𝑀 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑗𝑗 +𝛽𝛽 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑗𝑗 +𝛽𝛽 𝑒𝑒 𝑒𝑒 ⁄ 𝐴𝐴𝐴𝐴𝐴𝐴 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 the loans market was sought to be concentrated along 𝑖𝑖𝑖𝑖 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 ⁄𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 ∑ 𝜋𝜋∑ 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖𝑖𝑖 = =which …………………………………………………………………...... ሾͳ 𝜋𝜋𝐼𝐼𝐼𝐼𝑖𝑖=1 𝛽𝛽0 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 𝑖𝑖𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 …………………………………………………………………...... ሾͳ 𝑖𝑖𝑖𝑖== 𝑁𝑁 𝛽𝛽 +𝛽𝛽 𝐵𝐵𝐵𝐵 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝐵𝐵𝐵𝐵 𝛽𝛽0⁄01𝛽𝛽 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 11𝑗𝑗𝑀𝑀𝑀𝑀 22economic 𝑖𝑖=1 𝑁𝑁 𝑒𝑒𝐴𝐴𝐴𝐴𝐴𝐴 𝑗𝑗𝑗𝑗2+𝛽𝛽 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 ∑𝑁𝑁 𝑖𝑖𝑖𝑖3+𝛽𝛽 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 0𝑁𝑁 1𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑗𝑗 +𝛽𝛽 𝑖𝑖𝑖𝑖 𝑛𝑛=1 ∑ 𝑒𝑒𝑒𝑒𝑖𝑖𝑖𝑖 ⁄⁄∑ pre-identified activities (see Figure 2 & 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 1 1𝐴𝐴𝐴𝐴𝐴𝐴 +𝐴𝐴𝐴𝐴𝐴𝐴 + 𝑖𝑖𝑖𝑖𝑒𝑒 𝑖𝑖𝑖𝑖 𝑛𝑛=1 𝑛𝑛=1 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 𝑖𝑖areas 𝑖𝑖 of 𝑖𝑖𝑖𝑖 𝐼𝐼 ∑𝑛𝑛=1 𝜋𝜋𝑖𝑖𝑖𝑖∑𝜋𝜋= 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖 = …………………………………………………………………...... ሾͳ 𝑖𝑖=1 𝑁𝑁 𝐴𝐴𝐴𝐴𝐴𝐴𝛽𝛽0 +𝛽𝛽 𝑁𝑁 𝑖𝑖𝑖𝑖 = 𝑀𝑀𝑀𝑀1𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 +𝛽𝛽𝑗𝑗2+𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 1+𝛽𝛽 ∑3). 𝐵𝐵𝐵𝐵 𝑗𝑗𝑀𝑀𝑀𝑀 Figure 𝑖𝑖𝑖𝑖3+𝛽𝛽3𝑖𝑖𝑖𝑖𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 2𝑖𝑖𝑖𝑖 𝑛𝑛=1 1 +1 𝑒𝑒+𝑖𝑖𝑖𝑖𝑒𝑒⁄𝛽𝛽∑0𝑛𝑛=1 Where 𝜋𝜋𝜋𝜋 𝑖𝑖 ⁄𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 Where iii) the𝐼𝐼 third 𝐴𝐴𝐴𝐴𝐴𝐴 stage the𝑖𝑖 significant determinants of the LC 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼In 𝑖𝑖 = ∑𝑖𝑖=1 ∑𝑁𝑁 𝐴𝐴𝐴𝐴𝐴𝐴 ⁄∑𝑁𝑁 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆 …………………………………………………………………...... 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑛𝑛=1 𝑛𝑛=1dummified Where 𝜋𝜋were (scores above average =1, 0 𝑖𝑖𝑖𝑖𝜋𝜋𝑖𝑖𝑖𝑖 established and 𝐴𝐴𝐴𝐴𝐴𝐴 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 𝐿𝐿𝐿𝐿 otherwise) from factors loadings in (ii) based on the three-factor 𝐿𝐿𝐿𝐿 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖𝑖𝑖 is is the the 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 Where 𝐴𝐴𝐴𝐴𝐴𝐴 is the 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 Principle Components (PCs) scores as dependent variables 𝐿𝐿𝐿𝐿 𝑖𝑖 𝐿𝐿𝐿𝐿 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 is the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for indicator 𝑖𝑖𝑖𝑖 and the independent variables summarized in Figure 1. Using 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 is 𝑀𝑀𝑀𝑀 thej j𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for indicator 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖 is is the the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 equations binary logistic regression, four have been estimated 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for indicator 𝑖𝑖 𝑖𝑖 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 j j 𝜋𝜋 𝑖𝑖𝑖𝑖 𝜋𝜋 for this The binary logistic for this purpose has 𝑖𝑖𝑖𝑖 )purpose. 𝐿𝐿𝐿𝐿 = 𝑙𝑙𝑙𝑙 ( + 𝛽𝛽3model 𝐵𝐵𝐵𝐵 …………………………………………… ሾʹ 𝑀𝑀𝑀𝑀 𝑗𝑗𝑗𝑗 + 𝑖𝑖𝑖𝑖 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 = 𝑙𝑙𝑙𝑙 (𝑀𝑀𝑀𝑀 ) = = 𝛽𝛽 𝛽𝛽00 + + 𝛽𝛽 𝛽𝛽11 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 + 𝛽𝛽 𝛽𝛽22 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝐵𝐵𝐵𝐵 …………………………………………… ሾʹ 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 + 𝛽𝛽3 1−𝜋𝜋 𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 is𝜋𝜋𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖the 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 for𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖indicator 𝑖𝑖 𝑖𝑖𝑖𝑖 been specified as; 𝑖𝑖 1−𝜋𝜋 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 = 𝑙𝑙𝑙𝑙 (𝑀𝑀𝑀𝑀 ) = 𝛽𝛽 + 𝛽𝛽 𝑀𝑀𝑀𝑀 + 𝛽𝛽 𝐵𝐵𝐵𝐵 + 𝛽𝛽 𝐵𝐵𝐵𝐵 …………………………………………… ሾʹ 𝑀𝑀𝑀𝑀 𝑗𝑗 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 1−𝜋𝜋𝑖𝑖𝑖𝑖 1 1 𝐺𝐺 𝐺𝐺 Where, 𝜋𝜋 𝑖𝑖𝑖𝑖= ∑∑ . 𝑀𝑀𝑀𝑀 Where, 𝑀𝑀𝑀𝑀 . + 𝛽𝛽 𝐵𝐵𝐵𝐵 + 𝛽𝛽 𝐵𝐵𝐵𝐵 …………………………………………… 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑔𝑔=1 [2] 𝑔𝑔=1 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 =𝑀𝑀𝑀𝑀 𝑙𝑙𝑙𝑙 ( 𝑖𝑖𝑖𝑖= 𝛽𝛽 𝐺𝐺 +𝐺𝐺 𝛽𝛽𝑖𝑖𝑖𝑖 𝑗𝑗 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 1𝐺𝐺)𝐺𝐺1= 𝐺𝐺 𝐺𝐺0 1−𝜋𝜋𝑖𝑖𝑖𝑖 Where, ∑ 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = 𝐺𝐺 . ∑+𝛽𝛽 = 𝛽𝛽𝑔𝑔=1 𝐺𝐺 . 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐺𝐺 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑔𝑔=1 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 𝑀𝑀𝑀𝑀𝑗𝑗𝑗𝑗 +𝛽𝛽 +𝛽𝛽2 2 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 3 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 +𝛽𝛽3 𝑖𝑖𝑖𝑖 𝑒𝑒𝑒𝑒𝐺𝐺𝛽𝛽00+𝛽𝛽11 𝑀𝑀𝑀𝑀 Where,𝜋𝜋𝐺𝐺𝐺𝐺 𝑖𝑖𝑖𝑖= 𝑖𝑖𝑖𝑖 Where, 𝛽𝛽 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝜋𝜋𝑖𝑖𝑖𝑖 = 0𝛽𝛽 1 1 𝑀𝑀𝑀𝑀 2 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑗𝑗 𝑗𝑗 +𝛽𝛽 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 +𝛽𝛽 𝑖𝑖𝑖𝑖 𝛽𝛽0 +𝛽𝛽 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 𝑖𝑖𝑖𝑖 𝑒𝑒 𝐵𝐵𝐵𝐵 1+ + 𝑒𝑒𝑒𝑒 0 1 𝑗𝑗 2 𝑖𝑖𝑖𝑖 3 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝐺𝐺 𝜋𝜋𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝐺𝐺= 𝛽𝛽0 +𝛽𝛽1 𝑀𝑀𝑀𝑀𝑗𝑗 +𝛽𝛽2 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 +𝛽𝛽3 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 +the 𝑒𝑒current 𝑔𝑔𝑔𝑔==11to1tothe 𝑔𝑔𝑔𝑔 =+𝛽𝛽 0 +𝛽𝛽1 𝑀𝑀𝑀𝑀year 2 𝐵𝐵𝐵𝐵 3 𝐵𝐵𝐵𝐵 𝑗𝑗 +𝛽𝛽 𝑖𝑖𝑖𝑖 year =𝐺𝐺. 𝐺𝐺.𝑖𝑖𝑖𝑖 𝑒𝑒 𝛽𝛽current 𝜋𝜋 refers to the probability of LC facing a bank as perceived by 𝑖𝑖𝑖𝑖 𝜋𝜋 𝜋𝜋 = 𝑖𝑖𝑖𝑖 𝑔𝑔 𝑖𝑖𝑖𝑖 = to the current year 𝑔𝑔 = 𝐺𝐺. 𝛽𝛽 +𝛽𝛽 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 1 to 1the = 𝐺𝐺. 1 year 2 𝑔𝑔 𝑖𝑖𝑖𝑖 3 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 𝑗𝑗 + 𝑒𝑒current 𝜋𝜋𝑖𝑖𝑖𝑖𝑔𝑔 = i; lender 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝑖𝑖k𝑖𝑖k 𝐿𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 is the dummy for perceived LC score based on factor loading 𝐵𝐵𝐵𝐵 𝜋𝜋 𝐵𝐵𝐵𝐵 𝑖𝑖k greater 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖k than average for lender i in sectors. 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑖𝑖ljj 𝑖𝑖l 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑖𝑖l𝐿𝐿𝐿𝐿𝑖𝑖l𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀 ii j 𝑀𝑀𝑀𝑀 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 i𝑀𝑀𝑀𝑀 i 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 1j 1 (PLR (PLR ),),
𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 j 𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝜋𝜋+𝛽𝛽 𝛽𝛽0𝑖𝑖𝑖𝑖+𝛽𝛽 𝑗𝑗 𝑖𝑖𝑖𝑖 2 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 +𝛽𝛽3 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 𝜋𝜋 + 𝑒𝑒 𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀 j …………………………………………………………………...... ሾͳሿ ∑𝑁𝑁 The International Journal of Banking𝑀𝑀𝑀𝑀 and Finance,𝑀𝑀𝑀𝑀 Vol. 17, Number 2 (July) 2022, pp: 115–151 𝑛𝑛=1 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀j 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 j j 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 j 𝜋𝜋 𝐿𝐿𝐿𝐿𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 1 𝐺𝐺 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 𝑖𝑖𝑖𝑖 𝐺𝐺𝑖𝑖𝑖𝑖 . 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀j ∑𝑔𝑔=1 th 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖variable. = 𝐺𝐺𝑀𝑀𝑀𝑀 𝐿𝐿𝐿𝐿 of a macroeconomic 𝑀𝑀𝑀𝑀j refers 𝑖𝑖𝑖𝑖 1 to𝐺𝐺the j measurement ∑ 𝑀𝑀𝑀𝑀 = 𝐺𝐺 . 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 In this case inflation and 1growth experience. These two 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑖𝑖 𝐺𝐺 𝑔𝑔=1 1 1 𝐺𝐺 𝐺𝐺 𝐺𝐺 𝑀𝑀𝑀𝑀 ∑ ∑ 𝑀𝑀𝑀𝑀 = 𝐺𝐺 ∑𝐺𝐺𝑔𝑔=1 𝐺𝐺𝑖𝑖𝑖𝑖 . ∑ 𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 = = 𝐺𝐺 𝐺𝐺 . . 𝑀𝑀𝑀𝑀 = 𝐺𝐺 . 𝑀𝑀𝑀𝑀 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐺𝐺 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑔𝑔=1 𝑔𝑔=1 measures were the average for the period under j 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖which the𝑖𝑖𝑖𝑖 𝐺𝐺 𝑔𝑔=1 𝐺𝐺 𝐺𝐺 respondent 𝐺𝐺indicator 𝑀𝑀 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 for 𝑖𝑖 has been in the lending industry. For example, the 𝑖𝑖𝑖𝑖 1 𝐺𝐺 1 𝐺𝐺 growth experience 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖is computed 𝐺𝐺𝑖𝑖𝑖𝑖current year 𝑔𝑔 = 𝐺𝐺𝑖𝑖𝑖𝑖 𝐺𝐺𝑖𝑖𝑖𝑖 as 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 = 𝐺𝐺 ∑ 𝑔𝑔 𝑔𝑔=1 = 1𝐺𝐺to 𝑖𝑖𝑖𝑖 .the 𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 = ∑𝑔𝑔=1 𝐺𝐺𝑖𝑖𝑖𝑖 . 𝐺𝐺 𝐺𝐺 𝑔𝑔 = to the current = 𝐺𝐺. theሾʹሿ the GDP growth year rate𝑔𝑔during respondents’ years of 𝛽𝛽1 𝑀𝑀𝑀𝑀𝑗𝑗 + 𝛽𝛽2 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 + 𝛽𝛽is …………………………………………… 3 𝐵𝐵𝐵𝐵 𝑖𝑖𝑖𝑖1 1 𝐺𝐺 𝐵𝐵𝐵𝐵 =𝐺𝐺.1 to the current 𝑔𝑔 𝑔𝑔 = = to to the the current current year year 𝑔𝑔 𝑔𝑔=𝑔𝑔=𝐺𝐺. 𝑔𝑔𝑖𝑖𝑖𝑖=of to the current year 𝑔𝑔 = 𝐺𝐺. experience from the𝑀𝑀𝑀𝑀 year entry to the current year 𝑖𝑖k 𝐺𝐺 ∑ = 𝐺𝐺 𝑔𝑔=1 𝐺𝐺𝑖𝑖𝑖𝑖 . 𝑖𝑖𝑖𝑖 𝐺𝐺𝑖𝑖𝑖𝑖 th 𝐵𝐵𝐵𝐵 being the k measurement of the borrowers’ differentiation 𝑖𝑖k 𝐵𝐵𝐵𝐵current 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵by variables as perceived i;𝑖𝑖kand 𝑖𝑖l 𝑖𝑖k 𝑔𝑔 = 𝐺𝐺. 𝑖𝑖k 𝑔𝑔 = 1 to the 𝑖𝑖k lender year 𝐺𝐺𝑖𝑖𝑖𝑖 𝑔𝑔 = to the current year 𝑔𝑔 = 𝐺𝐺. th 𝑀𝑀𝑀𝑀 +𝛽𝛽 𝐵𝐵𝐵𝐵 +𝛽𝛽 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵𝑖𝑖l being the l measurement of borrower characteristics, as 𝑗𝑗 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 i 𝐵𝐵𝐵𝐵𝑖𝑖l 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 +𝛽𝛽1 𝑀𝑀𝑀𝑀𝑗𝑗 +𝛽𝛽2 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 +𝛽𝛽3 𝐵𝐵𝐵𝐵𝑖𝑖𝑖𝑖 perceived by lender i. 𝑖𝑖l 𝑖𝑖l 𝐵𝐵𝐵𝐵𝑖𝑖k 𝑔𝑔 = 1𝑖𝑖lto the current year 𝑔𝑔 = 𝐺𝐺. 𝐵𝐵𝐵𝐵 𝑖𝑖k final stage, the sector’s i the iv) In prediction of LC in equation 2, i i for each respondent i who hasi worked at 𝑖𝑖llender (PLR1is), were 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵 𝑖𝑖k 𝑖𝑖l (PLRcombined then in another set of binary logistic regressions for is ), 1 1 (PLR1is), low (up to 20% NPL) (PLR and high (over (PLR (PLR is ), isi),20% NPL) perceptions is ), risk 𝐵𝐵𝐵𝐵𝑖𝑖l i(PCRs) as dependent variables. Other levels of the PLR have been included for clarity. (PLR1is), i (PLR1is), 𝜋𝜋𝑖𝑖𝑖𝑖𝐿𝐿𝐿𝐿 = 𝑖𝑖𝑖𝑖
(PLR1is), RESULTS
At the core of this study is the overall effect of LC on the NPL risk, as presented in Figure 2 and Figure 3. The findings presented thus 1 and Lensink (2008), rrent year 𝑔𝑔 =far, 𝐺𝐺.have found support from the study by Pham 1 a where the SLC which was in favor of individual establishments had lower PLR perception, an indicator that the SSLC could be justified as a low-risk lending avenue in Tanzania. The fact that the SSLC accrued a low PLR seemed to suggest that the lending could be along social networks (Pham & Lensink, 2007), which if not trustworthy 1might plunge the banking sector into serious exposure and could provide the reason for the ever-increasing NPL rates. This observation is however, contrary to western studies which have associated individual lending to a relatively higher PLR than corporate borrowers (Beck & de Jonghe, 2013; Liu et al., 2019; Laib, 2013; Hunter & Nixon, 1999).Figure 3 provides the scores of the various PLR per sector. By comparing it with Figure 2, it is evident that high LC and high PLR sectors include SMEs, CBD businesses and non-partner individual borrowers. The low LC and low PCR sectors were DSM private sector organizations, interbank lending, as well as state corporations.
PLR perception, an indicator that the SSLC could be justified as a low-risk lending avenue in Tanzania. The fact that the SSLC accrued a low PLR seemed to suggest that the lending could be along social networks (Pham & Lensink, 2007), which if not trustworthy might plunge the banking sector into serious provide2the reason forpp: the115–151 everThe International Journal of Banking andexposure Finance,and Vol.could 17, Number (July) 2022, increasing NPL rates. This observation is however, contrary to western studies which have associated individual lending to a relatively higher PLR than corporate borrowers (Beck & de Jonghe, 2013; Liu et al., 2019; Laib, 2013; Hunter & Nixon, 1999).
Figure 3 provides the scores of the various PLR per sector. By comparing it with Figure 2, it
Furthermore, theLCobservations have suggested sectors where the is evident that high and high PLR sectors include SMEs, CBDthat businesses and non-partner individual borrowers. Thehigh, low LC low PCR sectors were DSM private sector PLR was relatively hadandincluded the commercialized sector organizations, interbank lending, as state corporations. the observations where expectations based asonwell Simpasa and Pla Furthermore, (2016) were swayed in have suggested that sectors where the PLR was relatively high, had included the favor of the low risk. The observations to the contrary have suggested commercialized sector where expectations based on Simpasa and Pla (2016) were swayed in thefavor possibility of aTheSSLC induced PLR, have whereby risk sectors of the low risk. observations to the contrary suggestedlow the possibility of a SSLCbe induced PLR, whereby risk sectors might be perceived risky if alending might perceived risky iflowlending concentration exceeded certain threshold. This study however, did not venture more into the threshold effect of LC. Thus, this research avenue is reserved for future studies. 10 the PELC, contrary to the longThe PSLC is also a better option than enshrined belief that those with employment and permanent income were perceived to be relatively less risky (Le & Nguyen, 2018; Pham & Lensink, 2008; Koomson et al., 2016). In this study however, income was not observed to be an important consideration during loan processing, while employment considerations have induced a relatively high rather than low PLR, contrary to the findings in Le and Nguyen (2018); Klyuev (2008) and Harrison et al. (2004). This was perhaps due to the fact that the government employee lending market could be oversaturated, or that lenders would consider the employment of the borrower only when they believed or faced a high NPL risk.
concentration exceeded a certain threshold. This study however, did not venture more into the threshold effect of LC. Thus, this research avenue is reserved for future studies.
1 is also a better option than the PELC, contrary to the long-enshrined belief that those with employment and permanent income were perceived to be relatively less risky (Le & Nguyen, 2018; Pham & Lensink, 2008; Koomson et al., 2016). In this study however, Descriptive Statistics for Variables in the Final Models income was not observed to be an important consideration during loan processing, while employment considerations have induced a relatively high rather than low PLR, contrary to Variable MinKlyuev Max Mean Std. Dev. Skewness Kurtosis theSfindings in Lename and NguyenN(2018); (2008) and Harrison et al. (2004). This was perhaps employeeNon-performing lending market could be oversaturated, A due to the fact that the government Perceived loans risk or that lenders would consider the employment of the borrower only when they believed or A1 below 10% 150 .12 .94 .58 .20 -.11 -.96 faced a high NPL risk.
between 10 - 20% 133 .00 1.00 .45 .31 .29 -1.18 150 .13 .98 .64 .24 -.27 -1.06 A4 20 - 40% 128 .00 1.00 .63 .37 -.55 -1.31 Descriptive Statistics the Final Models A5 Over 40% for Variables 142 in.00 1.00 .52 .39 -.16 -1.61 A6 Over 20% 151 .00 1.00 .47 .50 .12 -2.01 Variable name N Min Max Mean Std. Dev. Skewness Kurtosis B S Lender’s/Respondents’ Characteristics A Perceived Non-performing loans risk B0 Size large 151 0 .47 B1 Local_large 151 0 .36 B2 Org. Local 134 0 .54 B3 GIANT_2 151 0 .15 B4 GIANT_3 151 0 .05 A3 1 below 20% Table
(continued)
Variable name GIANT_1 OIB OLB Back Manage Credit Administration RiskF Operations Exper
N Min Max Mean Std. Dev. Skewness Kurtosis .13 .17 .13 .74 .22 .25 .16 .27 .56 143 1.00 25.00 6.97 4.23 1.37 2.78 Macroeconomic Environment (ME) Grow_Exper 143 3.45 1.46 5.64 1.68 .65 -.23 Infl_Exper 143 5.62 6.41 6.14 .22 -1.42 .87 Lenders application of Credit Risk Management Practices (CRMPs) CILG .24 1.21 1.00 .27 -1.20 .56 DCB .00 1.31 1.00 .29 -.85 .40 GCD .30 1.51 1.00 .38 -.27 -.96 Crev .00 1.19 1.00 .28 -1.48 1.65 LPB .00 1.44 1.00 .36 -.60 -.07 PBB .00 1.27 1.00 .30 -.94 .08 RBP .00 1.32 1.00 .31 -.91 .51 Considerations in credit processing Age .00 1.38 1.01 .40 -.96 .10 BusExp .00 1.30 1.00 .36 -1.43 1.52 CCap .00 1.26 1.01 .34 -1.68 2.35 CDebt .00 1.21 1.01 .31 -2.00 3.76 Cdep .00 1.91 1.01 .55 .06 -.81 Charct .00 1.26 1.01 .37 -1.55 1.39 Collat .00 1.22 1.01 .33 -1.84 2.83 Employ .00 1.32 1.01 .37 -1.27 1.06 Loanpur .00 1.25 1.00 .35 -1.57 1.73 CredH .00 1.18 1.01 .31 -2.13 3.93 Lender’s use of different collateral types Land .00 1.47 1.01 .43 -.73 -.30 PCP .00 1.57 1.01 .39 -.65 .41 CP .00 1.80 1.01 .57 .01 -1.28 BSA .00 1.84 1.01 .60 .03 -1.35 Invent .00 1.61 1.01 .42 -.67 .28 Lien .00 1.61 1.01 .42 -.55 .01 Invoice .00 1.57 1.01 .41 -.64 .12 Customer Lending Concentrations (LC) SSLC .01 .99 .57 .32 -.39 -1.21 PSLC .00 .99 .56 .29 -.31 -.92 PELC .01 .99 .56 .26 -.27 -.84
To analyze the relationship between the PLR and LC, it was necessary to first establish a quantitative measure of LC based on the 15 items evaluated in the questionnaire. Factor analysis suggested a threefactor solution, in which their respective principal components were considered as SLC indicators. Prior to further analyses, descriptive statistics have been considered in Table 1. It can be observed that some variables were skewed with a kurtosis value being far away from zero, thus calling for methods other than linear regression in determining the predictors. The dispersion was however not very alarming, as it seemed to suggest some stability.
| Model Fit | Model Classification | ||||
|---|---|---|---|---|---|
| No | Yes | Overall | |||
| NPL <= 10 -2 Log likelihood | 149.98 | No | 30 | 20 | |
| percent | |||||
| Cox & Snell | 0.17 | Yes | 23 | 55 | |
| Nagelkerke | 0.23 | Percent | 56.6 | 73.33 66.41 | |
| NPL = 20 - -2 Log likelihood | 120.09 | No | 59 | 15 | |
| 40 percent | |||||
| Cox & Snell | 0.36 | Yes | 10 | 44 | |
| Nagelkerke | 0.48 | Percent | 85.51 | 74.58 80.47 | |
| NPL > 40 -2 Log likelihood | 131.8 | No | 23 | 9 | |
| percent Cox & Snell | 0.23 | Yes | 21 | 75 | |
| Nagelkerke | 0.31 | Percent | 52.27 | 89.29 76.56 | |
Table 2 and Table 3 provide the summary information for “Low” and “High” PLR models respectively. From Table 2, it can be seen that the pseudo R2s were adequately high and provided an acceptable classification of cases, all being beyond 75 percent for two (2) models. However, the first one was below acceptable standards. This provides an indication of good model fit for the last two (2) models and a justification for further interpretation of the results. On the high NPL, risk perception, the pseudo R2; i.e., correlation and the classification table are provided in Table 3. All the models are generally fit and provide at least 85 percent classification accuracy.
| Model | |||||
|---|---|---|---|---|---|
| Model fit | Classification | ||||
| No | Yes | Overall | |||
| NPL <= 10 -2 Log likelihood 74.71 | No | 40 | 5 | ||
| percent | |||||
| Cox & Snell | 0.52 | Yes | 7 | 76 | |
| Nagelkerke | 0.71 | Percent | 85.11 | 93.83 | 90.63 |
| NPL = 20 - 40 -2 Log likelihood 85.22 | No | 52 | 10 | ||
| percent | |||||
| Cox & Snell | 0.51 | Yes | 9 | 57 | |
| Nagelkerke | 0.68 | Percent | 85.25 | 85.07 | 85.16 |
| NPL > 40 -2 Log likelihood 79.94 | No | 30 | 6 | ||
| percent | |||||
| Cox & Snell | 0.47 | Yes | 11 | 81 | |
| Nagelkerke | 0.65 | Percent | 73.17 | 93.1 | 86.72 |
With regard to employee characteristics, this study would like to suggest that those in management and those with a relatively broader experience in the banking industry have a higher PLR, an observation which is in line with the experience hypothesis, and which has also been suggested in Dember and Penwell (1980) and de Meza and Southey (1996). A study in URT (2019) has suggested that Tanzania is facing a higher NPL risk and that employees in the credit department are more aware of this. Unexpectedly however, it was further noted that employees in the credit department have suggested a significantly lower NPL risk. It is evident here that risk is overstated by those in management, as they view it as bad to their organization (de Meza & Southey, 1996), while those in operations (credit) understate risk as they prefer being optimistic in order to encourage borrowing. Further analysis among those in the credit department has further suggested that they also understated risk. This was because on average, they had less experience across LCs, which is an expected form of behavior (Burakov, 2014; Berger & Udell, 2004).
| Less than 10% NPL Between 10 - 20 % NPL Below 20% NPL | ||||||
|---|---|---|---|---|---|---|
| Exp(B) | Sig. | Exp(B) | Sig. | Exp(B) | Sig. | |
| Constant | 0.53 (0.70) | 0.13 (1.32) | 0.45 (0.77) | |||
| A | Lender/Respondent Characteristics | |||||
| OrigLocal(1) | 4.03 (0.81) | * | ||||
| Sizelarge(1) | 14.89 (0.93) | *** | ||||
| Local_large(1) | 0.03 (1.27) | *** | ||||
| GIANT_3(1) | 19.17 (1.40) | ** | ||||
| Operations(1) | 2.10 (0.42) | * | ||||
| RiskF(1) | 3.36 (0.50) | ** | 5.26 (0.63) | *** | 4.66 (0.56) | *** |
| Credit(1) | 3.64 (0.57) Lender Application of Credit Risk Management Practices | ** (0.56) | 2.60 | * | ||
| C | (CRMPs) | |||||
| RPAI_CILG | 0.05 (1.15) | *** | ||||
| RPAI_LPB | 6.19 (0.88) | ** | ||||
| RPAI_RBP | 31.15 (1.45) | ** | ||||
| RPAI_DBC | 0.01 (1.47) | *** | ||||
| D | Considerations in Credit Processing | |||||
| RCC_loanpur | Exp(B) | Sig. | 11.74 (0.88) Less than 10% NPL Between 10 - 20 % NPL Below 20% NPL Exp(B) | *** Sig. | Exp(B) | (continued) Sig. |
| RCC_CCap | 9.47 (0.70) | *** | ||||
| RCC_Charct | 4.30 (0.59) | ** | ||||
| E | Lender Use of Different Collateral Types | |||||
| RAI_BSA | 0.30 (0.37) | *** | 0.21 (0.50) | *** | 0.13 (0.47) | *** |
| RAI_Invent | 0.07 (0.93) | *** | ||||
| RAI_Lien | 14.70 (0.89) | *** | ||||
| F | Customer Lending Concentrations (LC) | |||||
| SSLC | 7.57 (0.84) | ** | 3.65 (0.72) | * | ||
As reflected in the data displayed in Table 4 and Figure 4, it can be seen that this study points to at least three main observations with regard to the employees of the three lending giants in Tanzania. The first relates to employees of GIANT_3, where the PLR was generally low and this has been enhanced by having most of its loans alongside the SSLC, which seemed to provide an additional advantage in terms of soft information (Pham & Lensink, 2008). It was found to have a relatively high PLR with regard to the PELC and the PSLC. As a result, the ability of GIANT_3 to generate useful credit information from these sectors was lower, in line with the TPT paradigm (Diamond, 1984; Markowitz, 1959). Thus, both public employees and the private sector are not for GIANT_1 an avenue for the SLC. The lender would be better-off diversifying across sectors if lending on this market. Like GIANT_3’s employees, the ones in GIANT_1 will also be able to achieve a lower PLR if the SSLC dominates and when it is against the PELC. It appears that the SSLC can provide important perceived shock absorbers to address the PLR among employees of this Lender. However, unlike GIANT_3’s employees GIANT_1’s employees seemed to associate the PSLC with high risk, an observation that was contrary to the previous observation that LC to commercialized sectors, such agriculture, manufacturing, transport and communication,
Less than 10% NPL Between 10 - 20 % NPL Below 20% NPL Exp(B) Sig. Exp(B) Sig. Exp(B) Sig. Less than 10% NPL Between (0.84) 10 - 20 % NPL Below (0.72) 20% NPL construction and real estate have a relatively lower PLR Exp(B) would Sig. Exp(B) Sig. Exp(B) Sig. (Le & (0.84) (0.72) Diep,As2020; Lensink, Skridulytė &that Freitakas, 2012). reflected Pham in the data& displayed in Table2008; 4 and Figure 4, it can be seen this study points at least three maincould observations regard toto thethe employees of the threelower lending giants in This toAs observation beinwith linked ability of reflected the data displayed Table and Figure 4, relatively it can seen that this study points Tanzania. Theinfirst relates to employees of 4GIANT_3, where the be PLR was generally low and to athas least main observations thefrom employees of the threeinformation lending giants to in and GIANT_1 tothree channel superior this been enhanced bymonopoly having with mostregard of rents its to loans alongside the SSLC, which seemed Tanzania. first relates to employees of soft GIANT_3, where(Pham the PLR was generally low and provide an The additional advantage in terms of information & Lensink, 2008). It was monitoring efficiency expertise into the private sector (Simpasa & this has been aenhanced having its loans seemedthe to found to have relativelybyhigh PLRmost with of regard to thealongside PELC andthetheSSLC, PSLC.which As a result, Pla, 2016). provide an additional advantage in terms of soft information (Pham & Lensink, 2008). It was ability of GIANT_3 to generate useful credit information from these sectors was lower, in line foundthetoTPT haveparadigm a relatively high PLR withMarkowitz, regard to the PELC andboth the PSLC. As a result,and the with (Diamond, 1984; 1959). Thus, public employees ability of GIANT_3 useful credit information from The theselender sectors was lower, in line the private sector are to notgenerate for GIANT_1 an avenue for the SLC. would be better-off Figure with4the TPTacross paradigm (Diamond, Markowitz, diversifying sectors if lending1984; on this market. 1959). Thus, both public employees and the private sector are not for GIANT_1 an avenue for the SLC. The lender would be better-off diversifying across sectors if lending on this market. Figure
Lending Giants NPL Risk in Relation to Lending Concentration
Lending Giants NPL Risk in Relation to Lending Concentration
(a) GIANT_1’s PLR in relation to(b) GIANT_1’s PLR in relation to (c) GIANT_1’s PLR in relation to (a) GIANT_1's PLR in (b) GIANT_1's PLR in (c) GIANT_1's PLR in SSLC PSLC PELC (a) relation GIANT_1’s PLR in relation to (b) GIANT_1’s PLR in relation to (c) GIANT_1’s in relation to to SSLC relation to PSLC relationPLR to PELC SSLC
(d) GIANT_2’s PLR in relation to(e) GIANT_2’s PLR in relation to (f) GIANT_2’s PLR in relation to SSLC PSLC PELC (d) GIANT_2’s PLR in relation to(e) GIANT_2’s PLR in relation to (f) GIANT_2’s PLR in relation to (a) GIANT_2's PLR in (b) GIANT_2's PLR in (c) GIANT_2's PLR in SSLC PSLC PELC relation to SSLC relation to PSLC relation to PELC
Lastly, there is ample evidence in this study that GIANT_2’s employees are unlikely to perceive a high NPL risk regardless of the SLC, an indication that GIANT_2 has proven to be a lower/stable NPL risk avenue. Employees from this lending giant have suggested that a declining PLR in response to the SSLC is an indicator of the strong reliance on social relationship in their operatives, just like the other banks while it is an increasing function of the PELC as an indicator of SLC saturation.
relatively of GIANT_1 to channel rents fromcould superior information Lensink, lower 2008;ability Skridulytė & Freitakas, 2012).monopoly This observation be linked to the and monitoring efficiency expertise into to thechannel private sector (Simpasa Pla,superior 2016). information relatively lower ability of GIANT_1 monopoly rents & from and monitoring efficiency expertise into the private sector (Simpasa & Pla, 2016).
there isJournal ample of evidence inand thisFinance, study that GIANT_2’s employees are unlikely to TheLastly, International Banking Vol. 17, Number 2 (July) 2022, pp: 115–151 perceive high is NPL risk evidence regardlessinof this the SLC, indication that GIANT_2 proven to beto Lastly, athere ample studyanthat GIANT_2’s employeeshasare unlikely a perceive lower/stable NPL Employees from lendingthat giant have suggested that a high NPLrisk riskavenue. regardless of the SLC, anthis indication GIANT_2 has proven to abe declining PLR in response to the Employees SSLC is anfrom indicator of the giant stronghave reliance on social a lower/stable NPL risk avenue. this lending suggested that a relationship in their the other while is an increasing function of Figure 5 PLR declining in operatives, response tojust thelike SSLC is an banks indicator of itthe strong reliance on social the PELC as an of SLC saturation. relationship in indicator their operatives, just like the other banks while it is an increasing function of theRisk PELC as an indicator of SLC saturation. NPLFigure 5 in Relation to Lending Concentration in Relation to Originality Figure 5 and Size of Lender NPL Risk in Relation to Lending Concentration in Relation to Originality and Size of Lender
NPL Risk in Relation to Lending Concentration in Relation to Originality and Size of Lender
(a) origin in (b) origin in (c) lender’s origin in (a)PLR PLRacross acrosslender’s lender's (b)PLR PLRacross acrosslender’s lender's (c)PLR PLRacross across lender's relation SSLC lender’s origin inrelation to across PSLC lender’s origin inrelation PELC lender’s origin in (a) PLRto across (b) PLR (c) PLRto across origin in relation to SSLC relation to SSLC
(d) PLR across lender's size in relation to SSLC origin in relation to PSLC relation to PSLC
(e) PLR across17lender's size in relation 17 to PSLC origin in relation to PELC relation to PELC
(f) PLR across lender's size in relation to PELC
This study further notes a relatively lower PLR in favor of lenders with local origin than those originating elsewhere, suggesting some typological effect in the PLR (Abu Hussain & Al-Ajmi, 2012; Kira, 2013). However, SLC in favor of social status or public employees modifies this effect by suggesting that if local lenders’ SLC is in favor of social status or government employees, it will end-up with a relatively high PLR among its employees. It is evident that the effect of origin of bank is stable only when a lender is focused on the private sector. The panels (a) – (d) in Figure 5 provide additional information on the moderating effect of LC on the effect of lender size on the PLR. There is evidence to the effect that local large banks are unlikely to fall prey to the higher PLR among its employees, to reflect a higher ability to grab economies of scale (Beck & de Jonghe, 2013), but high levels of the PLR among international large lenders who may be unable to benefit from the same, if most such economies emanates from social networks.
Effect of the Macroeconomic Environment This study also examined the Macroeconomic Environment (ME) in relation to the PLR and it became evident that both long lived inflationary and growth experiences were unlikely to increase the probability of a relatively higher PLR. Evaluations of the moderations of LC Panels (a) – (c) of Figure 6 seems to suggest that LC in favor of both the SSLC and the PSLC does not modify the effect of inflation experience. High inflation experience is associated with lower PLR across LC levels only if such LC has been in favor of social status, or private sector firms. This observation suggests that a significant LC based on social status and private sectors may cushion the lender from the negative effect of unprecedented inflationary experience on the NPL. Although the inflationary environment could restraint the household and spending behavior firms (Trautmann & Vlahu, 2013), both the private sector and individuals were relatively more flexible to re-adjust their earnings and expenditure to reflect changes in prices. However, with LC in favor of government employees, the lender is likely to suffer a higher NPL risk, since the earnings of borrowers (employees’ salary) are hardly adjustable in the short-run. This study also analyzed the effect of the GDP growth as experienced by lenders throughout their lending career. The observations suggest that although the high GDP growth experience is generally unlikely to be associated with a high PLR, the higher SLC in favor of social status induces a higher PLR in response to the higher GDP growth. As noted in Le and Diep (2020), the channel of effect could be loan loss provisions and difficulties in managing interest rate risk. Lenders operating during boom are not expected to fall prey to a high PLR, as repayment is guaranteed by the certainty of income flow (Lassoued, 2017). However, if the lender is under the SSLC, see Figure 6 (d), then a reversal of LC effect may be anticipated. This is despite the generalization that LC has a lower PLR, a booming economy might cast a negative shadow on the SSLC, leading to a high PLR. The Effect of Credit Risk Management Based on the results displayed in Table 4, three CRMPs appeared in this study to provide intuitive explanations with regard to the PLR. The application of the RBP to mitigate credit risk is unquestionable in as long as LC is in favor of private sector firms, as well as government employees (Wood & Kellman, 2010). However, this study did not find any significant effect when LC favors social status. For a lender who has not been concentrated in any sector, unlocking the loans of a potential borrower (LBP) would an effective Based on the results displayed in Table 4, threeprovide CRMPs appeared in this studymechanism to provide Based on the results displayed in Table 4, three CRMPs appeared this to study toThus, provide intuitive explanations with regard to the PLR. The application of the mitigate credit the to lower the PLR, as was suggested by SeyraminRBP (2013). intuitive explanations with regard PLR. application of the RBPfirms, to mitigate credit risk is unquestionable in as longtoastheLC is inThe favor of private sector as well as effectiveness of the, LBP asas aLCtool to mitigate creditfirms, risk iswellhigher risk is unquestionable as long is 2010). in favor of private as government employees in (Wood & Kellman, However, thissector study did notas find any when lenders are against instatus. its 2010). entirety. government employees (Wood &social Kellman, However, study didconcentrated not find any significant effect when LC favorsLC For a lender who this has not been in significant when the LC loans favorsof social status. For a lender(LBP) who has not provide been concentrated in any sector,effect unlocking a potential borrower would an effective any sector, to unlocking loans of asuggested potential by borrower would an effective mechanism lower thethe PLR, as was Seyram,(LBP) (2013). Thus,provide the effectiveness of Figure mechanism lower PLR,credit as wasrisk suggested Seyram, (2013). Thus, the the, LBP as to a tool to the mitigate is higherbywhen lenders are against LC effectiveness in its entirety.of the, LBP as a tool to mitigate credit risk is higher when lenders are against LC in its entirety.
Risk6 in Relation to Lending Concentration, Originality and Size
NPL Risk in Relation to Lending Concentration, Originality and Size of Lender NPL Risk in Relation to Lending Concentration, Originality and Size of Lender
(a) PLR PLR across origin in (a) acrosslender’s lender'sorigin (b) (b) PLR PLR across acrosslender’s lender'sorigin (c) (c)PLR PLRacross acrosslender’s lender's (a) PLR across lender’s origin (b) PLR across lender’s origin (c) PLRtoacross in relation to SSLC in relation to PSLC relation PELClender’s origin in origin to in SSLC relation to SSLC in relation origintoinPSLC relation to PSLC relation origin in relation to PELC in relation to PELC
(d) PLR across lender’s size (e) PLR across lender’s size in (f) PLR across lender’s size in
(d) PLR lender's (e) PLR PLRtoacross across lender'ssize in (f) (f) PLR PLRtoacross (d) PLR across across lender’s size (e) across lender’s size in in relation to SSLC relation PSLClender’s relation PELC lender's in relation SSLC to SSLC relation size intorelation sizetoinPSLC relation to PSLC relation sizetoinPELC relation to PELC With moderate SSLC and with either very low or very high PSLC, the lender Guarantee to With moderate SSLC and to with either or very PSLC, the lender Guarantee Debt (GCD) works lower the very PLR.low A very low high LC into the private sector seemed to to WithCover moderate SSLC with either very low very high PSLC, Cover works was toand lower the PLR. A very low LC into theor private to suggestDebt that (GCD) more lending in favor of other criteria, which could entail a sector strong seemed use of the suggesttothat morethe lending was favor especially ofDebt other criteria, could entail strongasuse of the the lender Guarantee to in Cover (GCD) works to alower the PLR. GCD avoid associated PLR, from which government employees, well as GCD to avoid associated PLR, especially from government as well as individuals. LC is the withinprivate the private sector, the GCD is also very relevant; this more is A very low When LCtheinto sector seemed toemployees, suggest that individuals. When LC may is within the private sector, thedegree GCD is very relevant; thisthe is because concentration be associated with a high of also uncertainty and thus, lending was in favormayofbeagainst other criteria, which could entail and a strong because concentration associated with degree thus, the use lender may require a cushion the NPL riskainhigh the form of of the uncertainty GCD. lender may require a cushion the NPL risk in the especially form of the GCD. of the GCD to avoid theagainst associated PLR, from government employees, as well as individuals. When LC is within the private sector, the GCD is also very relevant; this is because concentration may be associated with a high degree of uncertainty and thus, the lender may require a cushion against the NPL risk in the form of the GCD.
In terms of considerations in credit processing, the data in Table 4 seem to suggest that assessment of character emerged as the unequivocal consideration, an observation well supported in the literature (Abdulsaleh & Worthington, 2016; Laib, 2013). Regardless of LC, lenders ought to consider borrowers’ character, as such a consideration yields relatively lower PLR. These observations are however, contrary to the findings in Pham and Lensink (2008), who observed a limited correlation between character and the PLR. Though the consideration of loan purpose is seen as relevant, this study has noted that such consideration makes sense in reducing the PLR only if the different types of LC are either very low or very high. On the lowest extreme of the SLC, lenders do consider the purpose of the loan as an important criterion in granting loans; this is because borrowers are from diverse sectors for which the lender have limited knowledge (Beck & de Jonghe, 2013; Francisaet al., 2018). Therefore, detailed information on the purpose of the loan would facilitate the lender in getting some insights on the kind of business and the probability of repayment of the loan from the general behavior of the different purposes of loans (Adzobuet et al., 2017; Böve et al., 2010). At the highest end, the lender is concentrated and faces the risk of putting “all eggs in one basket” alongside the TPT postulates (Le & Diep, 2020). Although the lenders are now more knowledgeable about borrower behavior, they still require information in the appropriate basket – the one which lending should be concentrated. As such, details on the loan purpose will provide a mechanism for the lender to identify the best lending option among the many that exists. The observations from this study also seem to suggest that the effect of capital consideration is dependent on both the type and level of LC. To achieve lower PLR in response to capital consideration, as suggested in the literature (Richard et al., 2008; Adzobu et al., 2017; Koomson et al., 2016), both the lower and very high SSLC and PSLC can be targeted. The SLC in favor of social status, as well as LC for private firms both would require capital to achieve a lower PLR, because such borrowers would often have capital to declare. At low levels of such SLC, the lender has limited knowledge of the borrower and capital provision to the lender should provide a guarantee to the lender on the ability of individual borrowers, as well as firms to recover the advanced loan from its assets in case of default. At high SLC, the lender is knowledgeable of the borrower, but capital adequacy is still needed to address moral hazard problems and motivate the borrower to honor its obligation. In terms of SLC to government employees, capital adequacy is unnecessary, as has been suggested by Lassoued (2017), but when availed it can provide mechanisms to address the PLR by ironing out the moral hazards in the borrower’s mind (Aghion & Bolton, 1992). The Effect of Collateral Types Lastly, this study points to the effect of two collateral types on the PLR. The first is the use of the Business Savings Account (BSA), which was seen as inducing a high PLR perception, contrary to expectations (Koomson et al., 2016). However, a high PSLC reverses such an effect. The use of the BSA can be a useful tool to mitigate the PLR only if the lender LC is in favor of the private sector. In this market, business accounts are more prevalent and having one induces a comfort zone to lenders. Similarly, inventories as collateral can reduce the PLR when LC is either very low or when it is very high for all LC typologies. In either case inventories operate in the same manner as capital (Richard et al., 2008; Adzobu et al., 2017): at the lowest level of the SLC, inventories address the information gap (Chan & Thakor, 1987; Stiglitz & Weiss, 1981), while at the highest the SLC provide a cushion against moral hazard behavior among lenders (Aghion & Bolton, 1992). CONCLUSION The findings of this study, based on the contradictory observations in Tanzania that intensive credit crunch via the CRMPs seemed to provide a marginal outcome in terms of lowering the NPL risk, could be explained by the fact that there had been lending overconcentration based on social status across large and small lenders. Furthermore, it is evident that lending concentration reverses the effect of the macroeconomic variables, credit risk management practices (CRMPs), Credit Processing Considerations (CPCs), as well as collateral types. It is evident that only the intensification of “character”, “capital” and “inventory” scrutiny during loan processing has had a direct NPL risk-reducing effect, while all other considerations or CRMPs were dependent on the degree of lending concentration. Therefore, assessing the degree of lender LC across customers or sectors is imperative prior to any credit risk management initiative. Furthermore, it is notable that although LC has directly affected the PLR, as predicted by the traditional and corporate finance theories, the indirect LC effect via CRMPs, CPCs, bank size and originality, as well as the various collateral typologies has provided some new insights into this area of research. In terms of policy implications, it would be advisable that the Central Bank (BOT), instead of requiring lenders to have in place a holistic credit management program, should demand for the adoption of flexible lending policies to reflect the macroeconomic environment, as well as the respective magnitude and direction of LC. Restrictions should be placed on lenders with a SSLC, i.e., lending based on the characteristics of borrowers to the effect that they refrain from excessively expanding loans during high economic growth, and unless they re-direct such loans in favor of other criteria, especially employment and productivity of enterprises. Such restrictions may also extend to specific CRMPs. For example, the findings of this study have suggested that the RBP is an effective strategy to militate against the NPL risk, only if the lending favors the PSLC or the PELC, while the LBP is effective for lenders who are balanced in their lending behavior. Similarly, loan guarantees are effective only when the SSLC and the PSLC are low, while a business savings account scrutiny during loan processing works to reduce the PLR under high PSLC. A policy to restrict or control the use of the RBP should be advocated only in an environment where the lending concentration is not well guided. It should apply only to lenders whose lending is biased in favor of the SSLC, or the use of the LBP by lenders who are heavily concentrated, or the use of a loan guarantee to a lender who faces the PELC. ACKNOWLEDGMENT This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
For the updated list of financial institutions in Tanzania visit https:// www.bot.go.tz/BankSupervision/Institutions Such legislations include Business Registration and Licensing Agency; the National Payment Systems Act, 2015, Cooperative Societies Act, 2013; Companies Act 2002; Societies Act Cap 337 of 1954; NGO Act 2002; Trustees’ Incorporation Act 2002 (Cap 318). This data was however, not the subject of the current study.