DELINQUENCY ANALYSIS IN PERUVIAN MICROFINANCE INSTITUTIONS (MFI) Giovanna Aguilar Andía* Gonzalo Camargo Cárdenas** April, 2003.

Transcription

1 DELINQUENCY ANALYSIS IN PERUVIAN MICROFINANCE INSTITUTIONS (MFI) by Giovanna Aguilar Andía* Gonzalo Camargo Cárdenas** April, 2003 Abstract The main objective of this research is to identify the variables that affect the Peruvian Microfinace Institutions (MFI) delinquency. The relative importance of three kinds of variables is analyzed. First, aggregate or macroeconomic variables related to Peruvian economic activity (GDP, inflation, etc.). Second, microeconomic variables related to the managment policy of the MFI (credit policy, credit technology, incentives system to employees, etc.). And finally, variables related to the local dynamics of markets where MFI make their business (regional GDP, dynamics of local finance market, local productive activity, etc.) An unbalanced dynamic model of panel data for 35 MFI with monthly data in the period is estimated using GMM, Arellano and Bond (1991) and Arellano and Bover (1995) methodology. JEL Classification: G20, G21, G29. *Instituto de Estudios Peruanos (IEP) **Instituto de Estudios Peruanos (IEP) y AFP Horizonte (gcamargoc@afphorizonte.com.pe )

2 1. INTRODUCTION It is estimated that the small and micro enterprises (SME) sector 1 accounts for 42% of GDP 2, and gives jobs to 75% of the EAP of the country 3. This sector represents a very dynamic stratum of the national economic activity, which is characterized by short-term activities, with no entry barriers and the presence of several different kinds of links in the productive chain. One of the most serious problems of the SME sector faces is the limited access to adequate financial products which prevents them from taking advantage of some business opportunities in this dynamic sector. Commercial banking lacks adequate credit technology and financial products to meet the needs of the sector (Távara, 1998). In the other hand, the SME portfolio in the financial system is very small, it is estimated that it only reaches 0.8% of its total loans (Portocarrero, 2001). However, during the expansion of the financial system in the mid-nineties, the commercial banking began to be interested in this sector (Trivelli et al. 1999), but this trend reverted in times of the international financial crisis which along with El Niño Phenomenon in 1998, generated a plummeting of the internal activity level and a retraction of the local financial system. In this context, the commercial banking reduced significantly its loans devoted to SME sector 4, leaving these in the hands of microcredit specialized organizations. Microfinance institutions (MFI) in Peru are composed by Municipal Banks for Savings and Loans (Cajas Municipales de Ahorro y Crédito CMAC), Rural Banks for Savings and Loans (Cajas Rurales de Ahorro y Crédito CRAC), Small and Micro Enterprises Development Institutions (Entidades de Desarrollo de la Pequeña y Microempresa EDPYME), Mibanco, Banco del Trabajo and Financiera Solución 5. As to December, 2001, these entities gave loans to the PYME sector for almost 318 million dollars while commercial banking only gave, at the same time, 182 million dollars, that is, a lower amount than that of the specialized institutions According to the Legislative Decree No 705, the small enterprise is defined as a production unit of 20 workers or less and whose annual sales do not exceeds 25 U.I.T. While the microenterprise does not exceeds 10 workers and 12 U.I.T. in sales, and besides, his or her owner must work there. 45% of SME are concentrated in the agricultural sector, 36% in commerce, 11% in services and 8% in industry according to the National Economic Census in National Economic Census in According to the financial data of the commercial banking, published by the Bank Superintendency (Superintendencia de Banca y Seguros SBS) as to December, 2001, SEM portfolio reached only 1.74% of its total loans. Banco del Trabajo and Mibanco are banks while Financiera Solución is a finance company. 2

3 Specialized microfinance institutions or MFI are an important source of funding for the SME sector in Perú since, although they intermediate a lower amount than commercial banking they attend to more customers (Portocarrero et al. 2002), its loans have shown a great dynamics in recent years 6, and use a more efficient credit technology, which reflects in lower transaction costs, shorter paper work and lower arrears rates. Since MFI constitute a very important source of funding for small and medium size companies promoting their solid development and sustainability in the long term will make agile the funding for the SME sector allowing at the same time, gains in growth, efficiency and equity. Although MFI in Peru have shown low levels of delinquency and this has been pointed out as evidence of success, delinquency is a serious problem for any financial institution. A large amount of delinquent loans and/or defaulted loans constitute one of the main causes of insolvency and undercapitalization, which at the end goes against the strength and sustainability in the long run. MFI has shown a delinquency rate near 5% in 2001 (except for the CRAC); a level down what is observed in commercial banking in the same year (9%). This lower level of delinquency is, however, a trend shown in the last four years, which is evidence of a good performance of these institutions specialized in microcredit. Nevertheless, it is noticeable the poor performance showed by the CRAC whose delinquency level is located near 15.6% as to December, It was also observed that there is a large variety of results in the success of the management of the delinquent loans among MFI, which arises a question that, in turn, is the base of this research: What variables affect delinquency in MFI in Peru?. The information suggest there are important differences in the delinquency behavior in MFI, probably, as a result of the differences in their management policies, credit technology and the dynamics of their market-places, something that must be reflected in the differences in the impact of the variables affecting delinquency in this institutions. On the other hand, it is important not to look away from the high levels of delinquency showed by some institutions, probably as a reflect of inefficient administration, that could 6 From 1999 and 2001, the SME loans portfolio denominated in dollars, of these institutions increased in almost 75% according to data published by the SBS as to

4 generate in the long run perverse effects on sustainability and feasibility of the whole microfinance system 7. The general objective of this research is to identify those variables affecting Peruvian MFI s delinquency. It intends to identify the relative importance of three kinds of variables. In the first place, aggregated variables or macroeconomic variables which are linked to the whole Peruvian economic activity (GDP, inflation, etc.). In the second place, microeconomic variables related to the management of the MFI (credit policy, credit technology, incentive systems for their personnel, etc.). And finally, variables related to local dynamics from the markets where MFI operate (regional or department GDP, local market place dynamics, local productive activity structure, etc). Besides this set of variables it must be taken into consideration the presence of shocks and aggregated random events. Since MFI are still small organizations, closely related to high risk and volatile sectors, like agricultural sector and informal sector, it could be expected that shocks, like El Niño Phenomenon in 1998, which affect the dynamics of the economy, would have important effects on the activity of these institutions and so on the delinquency level. With monthly information that covers the period and 35 MFI, a dynamic panel data model is estimated taking into consideration different indicators for portfolio quality and indicators of the proposed variables at the micro and macro level. The paper is divided in six sections. After this brief introduction stylized facts which describe the microfinance market of Peru are presented in order to show the basic characteristics of intermediaries specialized in microcredit. In the third section the goodness and defects of the different indicators of the MFI portfolio quality are discussed and the evolution of the delinquency rate for the past four years at the level of the microcredit market segment and the level of each MFI is presented. In the fourth section the theoretical and conceptual framework which is at the base of the definitions of a set of relevant variables is developed in order to analyze delinquency in MFI. The fifth section shows briefly the methodology to be used in the estimation of the relation between delinquency and the main variables that determine it, some characteristics of the information to be used in the estimations and the main results found in the estimations. Finally, conclusions and recommendations are presented in the sixth section. 7 León de Cermeño and Schreiner (1998), affirm that failed experiences of MFI generate negative externalities in microcredit markets, for instance, trust destruction, damage in 4

5 2. MICROFINANCE IN PERU The introduction of a series of reforms in the financial sphere in early 90 s, became the starting point for credit expansion in Peru and for the emerging of new formal financial intermediaries, both factors in turn, allowed the expansion of microcredit (Trivelli et. al., 1999). The microfinance system in Peru is composed by the Municipal Banks for Savings and Loans (Cajas Municipales de Ahorro y Crédito - CMAC), Rural Banks for Savings and Loans (Cajas Rurales de Ahorro y Crédito - CRAC), Small and Micro Enterprises Development Institutions (Entidades de Desarrollo de la Pequeña y Microempresa - EDPYME), and the banks Mibanco and Banco del Trabajo, and the finance company Financiera Solución. These institutions are specialized in small size loans aiming basically to microenterprises and consumption credit The institutions composing the microfinance system could be classified in two categories according to their institutional characteristics (Portocarrero, 2002). In the first place we have the banks and private finance companies which are specialized in serving low-income customers like Banco del Trabajo, Mibanco and Financiera Solución. These institutions are authorized to perform a number of operations nation-wide. In the second place, we have the non-bank microfinance intermediaries like CRAC, CMAC and EDPYME, which are authorized to perform a limited set of operations and usually operate locally. MFI Loan portfolio During the period the loan portfolio of the formal financial system increased significantly showing an average increase of 9% annually. Although, this trend was reverted at the end of the decade due to the financial international crisis at the end of the 90 s, which along with El Niño Phenomenon in 1998, generated a sharp decline in the internal output level and a retraction in the local financial system. Thus loans in the total system decreased at an average rate of 8.20% annually from 1998 to 2001 (see table 1). Nevertheless, loan portfolio in the MFI system did not follow the same trend, on the contrary, they increased at a 21.7% rate per annum, which demonstrates the expansion this segment of the financial market experienced at the end of the past decade. expectations of the agents, etc. 5

6 Table 1 Financial System loan portfolio, (thousands of dollars) Type of institution Commercial Banking 7 721, , , , ,710 MFI 271, , , , ,041 Rest of the Financial System * 606, , , , ,320 Total 8 599, , , , ,653 Source: Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration * It does not include State owned banks but includes financial companies and leasing companies The share of MFI s loan portfolio in the total system has shown a increasing trend in the period mentioned (see table 2), reaching 5% of total loans in Obviously, commercial banking accounts for the largest portion of loans in the system. Table2 Percentage structure of loan portfolio in the Financial System, Type of institution Commercial banking MFI Rest of the Financial System Total Source: Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration. Even though credit supply from these institutions is small compared to that of commercial banking, it must be noted something very important. And that is that the number of customer these institutions serve is larger. The average loan from MFI is US $ 490 while an average loan from commercial banking is US $ 9,777 (see table 3), which is another evidence of the trend toward small scale loans which are concentrated basically in the SME and farming sectors. While commercial banking, judging by the average amount of the loan it gives, aims to medium-size and large-size companies, and mediumincome and high-income levels. 6

7 Chart 3 MFI : Number of debtors and average loan December 2001 Company No. of Debtors Loan portfolio (thousand of dollars) Average Loan (dollars) Commercial Banking 1 070, ,710 9, MFI 1 244, , CMAC 398, , CRAC 40,480 71,463 1, EDPYME 396,911 51, Banco del Trabajo 247, , Mibanco 78,960 62, Financiera Solución 82,322 59, Source: Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration Most microfinance institutions concentrate their portfolios in loans to small and micro enterprises and in consumption loans, excepting EDPYME, where mortgage loans follow SME s loans. Table 4 Percentage structure of MFI s loan portfolio according to type of loan and economic sector, December 2001 Type of loan and sector CMAC CRAC EDPYME B.Trabajo Mibanco F.Solución Commercial loans and microenterprises loans Agriculture,cattle-rising, hunting and forestry Manufacturing industry Trade Other activities Mortgage loans for housing Consumer loans TOTAL Source: Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration Municipal Banks for Savings and Loans (Cajas Municipales de Ahorro y Crédito CMAC) The Municipal Banks for Savings and Loans System (CMAC) was established as an initiative of the German Government in 1983 It was based on the Banks for Savings and Loans system of that country and received the technical support and participation of the German Cooperation Agency GTZ. The first support came true with the establishment of the Municipal Bank for Savings and Loans of Piura, in 1982, an experience that later expanded to other municipal banks for savings and loans across the country due to an 7

8 agreement between GTZ and SBS. This support was composed by a technical assistance and monitoring of a number of CMAC which were established due to this initiative. CMAC were established with an specific target: democratize credit and reach the sectors the formal financial sector was not interested in, mainly the sector of small and micro enterprises, looking at the same time to cover fully their costs in order to prevent from reducing their equities (Tello, 1995). As to June 2002, there were 13 CMAC which accounted for almost 40% of total loans portfolio of the MFI, with 121 branch offices across the country and 1,974 employees (see table 5). The three most important CMAC according to their share in total loans in this subsystem are: CMAC Piura with 22.02% of total loans of CMAC, 24 branch offices and 363 employees; CMAC Arequipa which concentrates almost 20% of CMAC s total loans, it owns 17 branch offices and 241 employees and; CMAC Trujillo, which accounts for 12% of loans, it owns 14 branch offices and 219 employees. Table 5 MFI S general information (As to June 2002) Nº of Branch Offices Nº Employees Loan portfolio (Thousand of dollars) % CRAC , CMAC , EDPYME Bco. del Trabajo , Mibanco , Fin. Solución , Total 333 7, , Source : Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration Rural Banks for Savings and Loans (Cajas Rurales de Ahorro y Crédito CRAC) Rural Banks for Savings and Loans (CRAC) were established after the disappearance of the Agricultural Bank (Banco Agrario) in The first of these institutions to be established was Caja Rural Del SUR (now Cajasur) beginning its operations in December 12, 1993 in the department of Arequipa. Following Caja Rural San Martín and Los Libertadores in Ayacucho established in March 20, 1994 and May 4, 1994 respectively. In 1995 the remaining CRAC were established across Peru, now they are 12 institutions. CRAC were established in order to meet the financial needs of farm 8

9 producers who were not reached by commercial banking, that is, they emerged as institutions specialized in the agriculture activity, though in recent years it has been observed their bias toward urban SME loans market and commercial loans. These institutions have been interfered for a long time by the Ministry of Agriculture, something that caused some government problems, image and dependency from the political power, something that has long lasting consequences for their proper workings. As to June 2001, the CRAC subsystem accounted for 10.22% of total loan portfolio from MFI (see table 5), they were composed by 12 institutions, with 52 branch offices and 564 employees. The most important are: CRAC San Martín, which accounted for 20% of total loans in the subsystem, also it owns 10 branch offices and 131 employees; CRAC Nor Perú, second position according to total loans, with 6 branch offices and 69 employees and CRAC Cajasur which accounts for 14.6% of total loans, 5 branch offices and 62 employees; it was also the first one that was established in December, Small and Micro Enterprises Development Institutions (Entidades de Desarrollo de la Pequeña y Microempresa EDPYME), The EDPYME (its name in Spanish) emerged due to the specialization and formalization of the so called Non-Governmental Organizations (NGOs), aiming to get funds from local and international capital markets. Under these formalization attempts, at the end of 1994 the Peruvian Government decree a rule which regulated its workings by the Banking Superintendency (Superintendencia de Banca y Seguros SBS). Unlike preceding MFI, EDPYME do not pick up deposits from the public something that made them very dependent on funding sources from the public sector and international cooperation. EDPYME account for 8% of total loan portfolio from MFI (see table 5). Now there are 13 EDPYME which own 41 branch offices and 647 employees. The three most important EDPYME according to their shares in total loans in this subsystem are: EDPYME Edyficar, which accounts for 37% of total loans in all of these institutions, it owns 14 branch offices and 234 employees; EDPYME Raíz accounts for 24% of total loans, 3 branch offices and 108 employees; y finally, EDPYME Proempresa which accounts for 10% of total loans in this subsystem, owns 7 branch offices and 66 employees. 9

10 Banco del Trabajo Banco del Trabajo is a company of Altas Cumbres economic group, with Chilean capital and long experience in financial business. It accounts for 20% of total loans of MFI (see chart 5). It owns 58 branch offices across the country and 2,489 employees. Since its establishment, in August, 1994, it aimed to offer consumption loans to people from medium and low income levels. According to information as to December, 2001, 57% of its loans were consumption loans. Mibanco Mibanco emerges as the result of the transformation of NGO Acción Comunitaria del Perú which, in turn, was supported by Acción Internacional, an NGO with long experience in microcredit businesses. It began its operations as a microfinance intermediary in May, It specialized in serving the micro and small enterprise segment, through giving short-run small-size loans which are commonly used as working capital. As to June, 2002, it accounted for 12.27% of total loans from MFI, it owns a net composed by 28 branch offices and 703 employees (see table 5). Financiera Solución Financiera Solución is a company of the Banco de Crédito del Perú economic group. In 1996 it changed its official name to Solución Financiera de Crédito and later to Financiera Solución. It accounts for 10.34% of total loans in the MFI system as to June, Subsidiary of Banco de Crédito it concentrates its activities in consumption loans and SME s loans (see table 5), it owns 33 branch offices and 1,563 employees. The differences in its origin are evidences, partly, of the different success levels gained in the microloan market. Except for the CRAC, other MFI were based, as a positive element, in the microfinancial experience of other institutions, in some cases, with international operations. In the case of the CRAC such an element did not existed, even more, their relations with the political power has gained for them more prejudices than benefits. 10

11 3. MFI s DELINQUENCY The analysis of portfolio quality of a financial institution needs the use of a proper indicator. However, there is no consensus on the discussion about which is this proper indicator for level delinquency of the credit institution s loan portfolio. From the financial data published by the SBS, we have three portfolio quality indicators which quantify, in relative terms, the level of delinquency in the loan portfolio or the loan portfolio with more credit risk. These indicators are: loan portfolio in arrears or arrears rate, high risk loan portfolio and doubtful loans The loan portfolio in arrears or arrears rate indicator is the ratio of loans in areas and on trial to loan portfolio. The high risk loan portfolio indicator is a more severe asset quality ratio, including in its numerator loans in arrears, on trial, refinanced and restructured; with the same denominator, loan portfolio. Doubtful loans indicator shows more differentiated characteristics. It is defined as the ratio between loans and contingent credits classified as deficient, uncertain and losses, and total direct loans and total contingent credits. Note that in the case of microenterprises loans these classifications are made based solely on days of arrears 8 and based on the total amount of the balance due. In the other hand, this is a fine tuning indicator about the quality of the portfolio since it considers that, after eight days in arrears, the total amount of the loan which shows payments due (and not only these ones) as loan portfolio in arrears. Limits to the indicators Since the record of loans due, refinanced, restructured, as well as the classifications of loans from MFI are based on countable and regulatory criteria since there is no record mechanisms at market values-, the confidence and accuracy of this information is a function of the level of fulfillment of the rules in force for each financial company. However, and with no prejudice to the preceding statements, it is noteworthy that this indicators show limits in concept which we explain here: 8 Normal Category: up to 8 days in arrears, with Potencial Problems Category: up to 30 days in arrears, Deficient Category: up to 60 days in arrears, Uncertain Category: up to 120 days in arrears, and losses category: more than 120 in arrears. 11

12 - These indicators consider only balance sheet activities. However, some financial companies use to withdraw from its balance the most damaged loans, selling this portfolio in certain periods of time at a very low price to some related entity, reducing thus its delinquency indicator. Something similar happens when the financial entities make accounting punishments to their damaged loans. In the extent these practices are not made homogeneously by the diverse financial entities, some distortion would be arising. - Portfolio exchange programs with the back of bonds issued by the government represents another source of distortion for delinquency indicators, since those loans given temporarily in exchange for bonds are transferred to a trusteeship, leaving the balance of the financial entity. - Delinquency in each loan portfolio has a dynamic behavior and evolves over time; estimating for the financial institutions a maturing term of 18 months approximately. Nevertheless, our portfolio quality indicators are calculated using aggregations and in a static way. Out of the three indicators mentioned here the most used one is loan portfolio in arrears or arrears rate called in the literature delinquency rate, since it can easily be generated from the accounting data of the institutions which, besides, is public. For description purposes of MFI portfolio quality, which follows, we are going to use the delinquency rate since it is the indicator for which there is the largest amount of data available. However, in estimations we are going to use the three indicators of portfolio quality that we mentioned above. Delinquency rate Even though the microfinance industry has had a considerable development and growth, there still are some problems and challenges to address, one of them is delinquency. Table 6 shows information about the delinquency rate in the general banking and in the MFI for the period. Except for the CRAC in 2001, other MFI show delinquency rates which are lower than those of the commercial banking. However, delinquency increases in all MFI during 2001 unlike banking delinquency. 12

13 Table 6 Evolution of Delinquency Rate, Entity Commercial Banking Banco del Trabajo Mibanco Fin. Solución CMAC CRAC EDPYME Source: Banking Superintendency of Perú (Superintendencia de Banca y Seguros SBS) Own elaboration CRAC show the highest delinquency rate in the microcredit system and even the highest in the whole financial system, in spite of the downward trend that was shown in this period, since delinquency went from 17.4% in 1998 to 15.6% in It is important to mention the interesting result gained by the CMAC in terms of a decrease in their delinquency level. Delinquency rate decreased from 7.5% in 1998 to 5.3% in This is a result similar to that shown by Financiera Solución. The delinquency rate in Mibanco stands in a very low level of 2.5% during the analyzed period. EDPYME show a downward trend delinquency rate in the period analyzed and stands near 6.9%. Figure 1 20% TASA DE MOROSIDAD DE LAS IMFS PROMEDIO (Ene-98 a Dic-01) Average delinquency rate, jan.98 - dec.01 15% 10% 5% 0% C. Municipales C. Rurales Epymes B. Trabajo F. Solución Mibanco Source: SBS Own elaboration 13

14 It is noteworthy that there is a large variance in delinquency rate within the CRAC subsystem. While in the CMAC system the portfolio quality is more homogeneous judging by the similarity of delinquency rates that these institutions show. The same happens within EDPYME except for those which two years of operations or less. Table 7 Descriptive Statistics of delinquency rate by kind of institution (Jan. 98-Dec. 01) Mean Standard Deviation. Coef. Var. Asymmetry Kurtosis CMAC CRAC EDPYME MIBANCO SOLUCION TRABAJO There are some differences in the MFI s delinquency behavior, probably, as a result of the differences in their management policies, their credit technology and the dynamics of their spots. Also, it is important to note that, except for the CRAC, MFI show very low delinquency rates, when compared, by instance, to those showed by the general banking. This result let these institutions to be seen as relatively successful companies in the financial intermediation business for the SME sector. However, it is important not to look away from the fact that high delinquency rates shown by some institutions could generate, if maintained in the long term, perverse effects on the sustainability and feasibility of the whole microfinance system 9. Ledgerwood (1999) summarizes the problems that a high level of delinquency causes in the performance of the MFI. Increased monitoring expenses and a deeper tracking of loans reporting arrears in their repayments are needed when delinquency is high, something that at last could affect the institutions liquidity. On the other hand, there is a negative effect on benefits. A delay in them, as a consequence of the lack of credit repayments, generates a lost in capital gains. Finally, we must consider the negative impact of delinquency on the profitability of the institution. This effect happens, both through income and expenses. Delinquency decreases income since the financial income 14

15 is not gained and expenses on provisions and operative expenses (expenses to recover loans due). Westley y Shaffer (1997) affirm also that high levels of delinquency affect the short - run relation of the MFI with their clients, damaging their loyalty and generating an infection effect which makes them adopt a non-repayment behavior. The larger the resources a MFI devotes to fight portfolio in arrears the lower the amount of funds available to service a larger level of demand for loans and thus the lower its level of growth and expansion. 4. THEORY AND CONCEPTUAL FRAMEWORK The term microfinance refers to the provision of financial services to low income clients, including those who has independent jobs. These financial services include savings and loans, though in some cases they also provide insurance (Ledgerwood 1999). The institutions devoted to these activities are called microfinance institutions (MFI). Like any other financial intermediary, MFI are exposed to credit risk problems, that is, they are exposed to face arrears in loan repayments and, in some cases, they could face a non-repayment. In general credit markets are not efficient, due to the problem of asymmetric information they involved (Stiglitz and Weiss, 1981; Stiglitz 1996; Virmani 1982). The nature of financial transactions is different to that of other economic transactions, since in them a contract is established where money is exchange by a promise for a future payment of the money lent. This promise feature in financial transactions makes it necessary that those who give loans and those who receive them have the biggest amount of information available in order to determine the credit risk; an stable economic context where they could establish term of the contracts; prices that flow freely to reflect credit risks and clear accurate rules in order to fulfill contracts and solve problems in benefit of both parties. As mentioned early, a high delinquency rate constitutes a serious problem which involves the feasibility in the log run of the institution and at the end of the system itself. In 9 León de Cermeño and Schreiner (1998), affirm that unsuccessful experiences on MFI generate negative externalities in microcredit markets, for instance, confidence destruction, damage in expectations of the agents, etc. 15

16 fact, the fragility of a financial institution due to its high level of delinquency loans produces at the beginning a liquidity problem which in the long run, if it persists and if the institution has no contingency credit lines, would become in an insolvency problem which would determine, probably, the end of the institution (Freixas and Rochet, 1998). In the specific case of a MFI, studies have demonstrated that high levels of delinquency led to the failure of these entities (Huppi and Feder, 1990) Most of the papers that try to explain which are the factors that determine banking delinquency address the topic from micro and macro views only, with no adoption of a global perspective which could include jointly both aspects. There is no theoretical model that analyzes in a general way the factors that determine banking delinquency since most of the theoretical models analyze the exclusive effect of certain variables separately. The global approach has been made mainly from an empiric view, in order to find those variables that best contribute in the determination of the observed delinquency rate. In that sense, in Saurina (1998) it is empirically demonstrated the jointly importance of aggregated factors (evolution of the economy, aggregated demand, unemployment rate, wages, etc.) and specific factors related to the credit policy of each entity (market share, loans growth rate, incentives policy, etc.) on delinquency level of the Spanish Banks for Savings and Loans. Following the Saurina s (1998) 10 scheme we are going to review the current literature on the determinants of the banking delinquency in order to get an hypothesis which later could be validated empirically. Macroeconomic Factors There is an interesting number of models that explain the macroeconomic determinants that cause the bankruptcy of a company. Saurina (1998) says that, due to the nature of the financial problems that are faced by companies in bankruptcy, delinquency problem is a previous step toward that bankruptcy though not all delinquent companies would bankrupt. Thus, he uses some theoretical conclusions from models explaining company bankruptcy to explain delinquency aggregated determinants. 10 This part of the document is based on the paper of Saurina (1998) 16

17 A conclusion shared by theoretical and empirical models is that there is a negative relation between economic cycle and delinquency. Also, it must be evaluated if the relation between delinquency and cycle is synchronic or includes lags, that is, whether the current growth rate of the economy determines the current delinquency, or if previous expansion periods generated a better future repayment capability in agents and thus lower delinquency rates in the future. Higher constraints in liquidity could generate problems in agent s repayment capability. For instance, a general reduction in wages, in raw materials price, or an increase in active interest rates could reduce repayment capabilities in households and companies. A higher debt (measured as a portion of GDP or mean income) could increase problems in agents facing their duties, both due to a increase in the weight of the debt service or to a lower access to new loans since higher leverage levels makes it harder to get additional funding. At the most become indebted the agents of an economy minor is his capacity of payment therefore greater the probability of than the companies faces delinquency problems. However, Davis (1992) and Petersen and Rajan (1994) demonstrate that, in the Japanese financial system, banks are eager to fund companies in trouble times even when their debt level is high. Wadhwani (1984 and 1986) derives a theoretical model where he assumes that companies work in imperfect markets and got that bankruptcy probability of a company is related to its wages, raw materials prices, real and nominal interest rates, the quotient between debt and market value of the company and the real aggregated demand. Similarly, Davis (1992) uses the GDP growth rate, wages and raw materials real prices, interest rates and the debt-gdp ratio in order to explain company bankruptcy rates in some OCDE countries. Freixas et. al.(1994) introduces expectations measures, money supply, real aggregated demand, inflation and private sector debt over the GDP, real wages, and real interest rates. Brookes (1994) explains the probability of delinquency in mortgage loans repayments as a function of the income level, service of the debt ratio, quotient of the private sector net wealth over the number of mortgage loans, of the unemployment variation rate and the debt-properties value ratio. As well as liquidity constraints faced by agents. 17

18 Muñoz (1999) evaluates, through a panel data fixed-effect model, the impact of economic growth over the solvency of the Peruvian banking system. He found that the evolution of the delinquency rate is counter cyclical, that the exchange rate volatility affects the loans default denominated in foreign currency and that variations in the exchange rate have a direct relationship with the quality of the portfolio. In Peru, the research on the determinants of MFI s delinquency is relatively new and underdeveloped, one of the most recent and known papers is that of Murrugarra and Ebentreich (1999). The objective of that paper is to test the effect of EDPYME s loan policies on the levels of delinquency observed in their agencies, using as explanatory variables the characteristics of the local credit market and loan management policies of each entity and branch office 11. They estimate a panel data model analyzing the presence of fixed effects and random effects in the sample they used 12 and additionally, taking into account the censured nature of the dependent variable 13, estimate a Tobit model with fixed effects 14. Although, for the authors, factors affecting EDPYME s delinquency are microeconomic factors, they add to the estimation the observed delinquency rate in the local credit market. They acknowledge that although the characteristics of the agencies and the entity determine the delinquency rate per agency, market conditions could also affect this result, that is, regional characteristics have an impact on capacity of the agents to repay their loans in the agreed term. In the case of the estimation of fixed effects, they found that an increase of 1% in delinquency of the department where the EDPYME is located affects its delinquency in 0.02%, while in the case of the Tobit model the effect is 0.25% A recent study that analyzes the determinants of the amount of portfolio in arrears of the banks of the Peruvian financial system is that of Guillén (2001). The author finds some evidence that the biggest banks are those which are more affected by external variables like the exchange rate, contractions in the aggregated demand and interest rates, while internal variables of each institution have a more important role in the case of smaller banks. Among these variables we could highlight the credit policy and risk-taking policy of the institution as well as the proportion of loans due Murrugarra y Ebentreich (1999). They find evidence in favor of the presence of fixed effects. Since in the first months of operation agencies has no significative delinquency rates. Authors acknowledge that Tobit specifications are sensitive to specification errors in nonobservable terms (Maddala, 1993) 18

19 In short, the macroeconomic determinants of delinquency could be classified in three large groups: variables which measure the economic cycle o the level of activity, variables which affect the liquidity level of the agents, like the wage level, and those variables which measure the debt level of the economic agents. The way each one of these groups contribute in the determination of the delinquency rate generates a behavior hypothesis which must be evaluated empirically. This way, we could expect negative relations between the economic cycle and delinquency, and between liquidity and delinquency, while it is expected a positive relation between debt and delinquency, though the latter could have a non-determined sign since in some cases financial intermediaries could be determined to fund agents showing a deficit. Microeconomics Factors The behavior of each financial entity is fundamental in order to explain its delinquency level, for instance, those having a very aggressive loan policy are expected to have higher levels of loans in arrears. In this sense, credit growth, type of business and incentives to adopt higher-risk policies is the most analyzed group of variables (Saurina, 1998). One of the most important factors in the determination of the delinquency rate of a banking institution is the credit expansion speed, important increases of loan growth rates could happen along with reductions in the level of requirements asked to applicants. Also, Clair (1992), Soltilla and Vihriala (1994) found some evidence that past loan growth contributes in explaining current delinquency. One possible cause of the increase in delinquency is the incentive managers have to take higher-risk loan policies. Entities facing solvency problems could begin a kind of forward escape seeking expansions in more profitable segments but facing higher-risk scenarios (Saurina, 1998). On the other hand, adverse selection also affects those institutions seeking to increase quickly its share in the credit market, since if one entity tries to grab the clients of other bank, the latter would likely release its worst clients. If this expansion is made in a new area or segment, adverse selection problems could multiply since the first clients to arrive to the new entity would be the worst quality ones. 19

20 The number of branch offices in each entity is used as an proxy indicator of geographical diversification for each institution 15. In principle, the increase in the number of agencies means access to a variety of markets which could generate some trouble in monitoring and control damaging so the capacity to evaluate and recover (Murrugarra y Ebentreich, 1999). However, and contrary to what was said before, we must take into consideration the probability that if institutions follow a policy aiming to seek for the best borrowers in each place, it is possible that the increase in the number of agencies could generate access to segments with better repayment capabilities, which would increase the average quality of the loan applicant, which in turn would reduce the expected delinquency rate 16. The type of business entities perform is also a determinant of the asset portfolio quality. In general, loans to households and companies represent a higher risk than buying government debt papers. In this sense, the structure of the balance and that of the asset portfolio reflect the risk each entity is eager to take. Higher risk levels use to be associated to certain sectors where activities are of higher risk. For instance, Keeton and Morris (1987) found a higher loan risk level in the agricultural sector. Such authors developed measures of risk diversification in order to explain delinquency rate, and found, that for those entities that for the same kind of credit risk charge higher interest rates are those which show higher delinquency rates. On the other hand, in all credit entity a proper loan supervising could be an important determinant of recovering rates. The scarcity of resources devoted to monitoring tasks is a risky practice affecting the capacity of loan control and recovering. A very commonly used indicator is the loan portfolio per employee. It reflects the amount of loans which, in average, each employee must attend and is defined as the ratio between total loan portfolio and the number of employees 17. In principle, the effect of this indicator on the delinquency rate is expected to be positive. However, this relation is not clear since increases in the amount of loan portfolio per employee generate higher We tried to create a geographical dispersion index, but it was not possible due to the lack of information on the places or departments where most of the MFI, used in the sample, operated. In the Fixed Effects Model of Murrugarra and Ebentreich (1999) it is estimated that an additional branch office represents 0.5% more delinquency. In the Tobit Fixed Effects Model this variable is not of significance anymore. For a better approach the number of analyst was tried to use, unfortunately, this information is not made public by SBS. 20

21 delinquency rates as long as this employee has surpassed the credit saturation point he could efficiently monitor 18, that is, up to a certain amount, the employee could increase or keep control efficiency, and from a certain point, and due to the excessive size of the amount he or she must supervise, it is likely to begin the generation of control efficiency losses. In short, the marginal increase in loans lent per employee not necessarily generate higher delinquency rates. In this respect, Murrugarra and Ebentreich (1999) use the number of loans that an average employee must attend and expect that, ceteris paribus, branch offices with more loans per employee show more deficient credit evaluation systems and thus, higher delinquency. The authors conclude the number of loans per person is significant although they have a small effect, which means that an increase in the number of loans per employee in order to have an effect on the arrears rate must exaggerated be great. Also, Berger and de Young (1997) found evidence that reductions of the efficiency cost go accompanied of increases in delinquency. The relation between delinquency and guaranties has an uncertain sign since traditional theories affirm that the best borrowers are eager to give more guaranties in order to show they are low risk and at the same time, a greater number of guaranties limits the moral hazard of the client. However, some theories that maintain a positive relation have been developed, they affirm the presence of guaranties reduces incentives for the institution to make a proper monitoring of loans while it also could generate optimism in excess among borrowers (Padilla y Requejo, 1998). In the peruvian financial system loan guaranties are classified and made public taking into consideration its capability to be exerted as soon as possible. This way there are slow execution guaranties, fast execution guaranties and other which are not classified. However, and as Murrugarra y Ebentreich (1999) say in the case of microloans this indicator is not so accurate, for it does not capture the effect of other kinds of guaranties which are commonly used for different kinds of loans, like group guaranties, personal guaranties, whose nature is different from that of real guaranties. In short, reviewing the literature on microeconomic determinants of delinquency it could be concluded that credit expansion, the kind of sectorial diversification, the 18 We must understand efficently as the capability to lend loans generating a delinquency rate equal or lower than that of the institution. 21

22 objectives and the level of efficiency of the company, the presence of guaranties, the market power and solvency of the entities are important in the determination of delinquency level observed for a credit institution. The kind of relation among these variables and MFI portfolio quality will be empirically evaluated in the following sections of this paper. As mentioned earlier, the objective of this paper is to analyze MFI s delinquency determinants; in this sense, econometrics provide tools to find statistically significant variables for the determination of delinquency. Due to the kind of data available and to the advantages these kinds of structures provide data panel methodology will be used. In the following part of the paper the main characteristics of the models which are going to be used to find MFI s delinquency determinants will be presented. 5. EMPIRICAL EVIDENCE The Model Panel models are econometric models based on observations repeated throughout the time for same individuals, that is, they are models of cross section to each temporal series. These models have two dimensions: Time dimension: observations in the time for each one of the individuals that conform the sample (t=1,...,t) Cross section dimension: observations of all the individuals for each one of the moments of the time. The most important characteristic of these models is that individuals composing the sample are the same for each year (if panel is balanced) or anyhow, the sample composing the transversal section of a given year is not independent from the following year, so individuals can be born or die (not balanced panel). 19 Panel data models permit to study the behavior of different agents throughout the time. One of the greater advantages of these models, as opposed to those of cross 19 Non balanced panel estimation does not imply substantial changes in the estimation. See Baltagi (1995) for details. 22

23 section or as opposed to those of time series, that they provide more flexibility to study the differences in behavior among individuals throughout the time. According to the kind of variable that could be included as explanatory variables in the equation to estimate we could distinguish two kinds of data panel models: panel data models with strictly exogenous variables and dynamic panel data models or panel data models with predetermined variables. The advantage of dynamic models is that they permit to insert lags of the endogenous one as explanatory variables, which generates more realistic representations since they capture the autoregressive component of a several economic series. The estimation of the dynamic panel data model will be made with reference to some works of Arellano and Bond (1991) and Arellano and Bover (1995) A simplified version of the model to estimate is 20 : y it = γ y it 1 + α i + e it (1) This expression assumes individual processes fluctuate with the same autoregressive pattern for each individual in different points of time. In the case of dynamic models by construction Cov( y, ) 0 it α i, so we are in a context of intragroup estimation. However, the inclusion of autoregressive term and its correlation with error terms causes intragroups estimations of parameter γ to be biased. However, Nickel (1981) demonstrates that the greater the number of observations per individual (T large) the smallest the bias. The size of the bias is: Lim N αˆ WG = α ( 1 + α ) h 2α h 1 ( T 1) ( T 1)( 1 α ) 1 20 For simplicity the strictly exogenous component (β X it ) is not included. Its inclusion has no effect on estimation. 23