Assessing the Sectoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy

Transcription

1 Theoreical and Applied Economics Volume XIX (2012), No. 2(567), pp Assessing he Secoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy Bogdan MOINESCU Buchares Academy of Economic Sudies Adrian CODIRLAŞU Buchares Academy of Economic Sudies Absrac. The paper is an aemp a sudying he mechanisms whereby he economic aciviy dynamics and money marke condiions affec he developmens in nonperforming loan raio across he main aciviy secors, namely agriculure, indusry, commerce and consrucions. The defaul raes are modeled boh on he basis of a linear approach and via a logisic funcion, saring from he mehodological soluion of he repued condiional risk model referred o as Credi Porfolio View. The robusness of he analyical framework is ensured by applying SUR esimaion mehod for simulaneous sysems of equaions in combinaion wih ha of auoregressive vecors. The empirical analysis is based on unique se of quarerly daa, which allows for assessing he qualiy of non-financial companies loan repaymen. The relevan explanaory variables were used in various configuraions and lags for consrucing several macroeconomic credi risk models. Keywords: nonperforming loans; financial sabiliy; macroeconomic credi risk models; real economy; moneary condiions. JEL Codes: G01, G17, G21, G32, G33. REL Code: 11B.

2 70 Bogdan Moinescu, Adrian Codirlaşu 1. Inroducion Finding he deerminans of reimbursing capciy of non-financial corporaions sars from he empirical observaion ha bank credi defaul raes are higher during recessionary periods han during an economic upurn (Williamson, 1985, Kiyoaki, Moore, 1997, Marcucci Quagliariello, 2009). A he roo of such discussions lay economic growh and ineres raes, along wih he exchange rae in he case of counries wih a large share of foreign-currency denominaed loans in he loan sock. This paper follows he mehodological line suggesed by Wilson C. Thomas (1998), who wen ino more deails as regards he heme of condiional credi risk models and proposed he repued Credi Porfolio View model. Accordingly, he relaion beween credi risk and he macroeconomic landscape is bes described by he logisic funcion and he lag is considered wihin he ime frame delineaed by he frequency of observaions (one year). Boss Michael (2002) applies he above-menioned mehodological soluion in order o model he secoral dependence of credi risk o he sae of he Ausrian economy boh a he level of exposures o non-financial corporaions and a he level of exposures o households. Empiric evidence shows ha he macroeconomic facors generaing defaul raes are, generally, he same for boh ypes of counerparies, wih he inflaion rae and he shor-erm ineres rae equally influencing he repaymen abiliy of boh non-financial corporaions and households. Virolainen Kimmo (2004) models he credi risk dynamics for Finland by using macroeconomic variables such as economic growh, iner-bank ineres rae for a mauriy of up o one year and he corporae indebedness level. Unlike previous research works, he analysis of defaul rae sensiiviy relaing o he sock of loans exended o non-financial corporaions vis-à-vis economic performance is broken down by secor, in six ypes of aciviy. A he same ime, in measuring he coefficiens of he equaion sysem he auhor ops for he SUR mehod wih a view o limiing he correlaion beween esimaed regression errors resuling from he significan degree of correlaion beween he dependen variables. Fiori, Robera, Foglia Anonella and Iannoi Simonea (2007) envisage analysing he exen o which he economic picure in Ialy affecs he developmens in credi risk riggered by he exposures o non-financial corporaions broken down by secor in eigh caegories. The mehodology is based on Wilson s approach (1998) and he subsequen developmens designed by Virolainen (2004), whereas he assessmen of he operaional form of he empiric relaion beween he logisically changed values of defaul raes by

3 Assessing he Secoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy 71 secor and he macroeconomic environmen is achieved via SUR mehod. The informaion on defaul raes a secoral level is achieved by accessing he archives of he Cenral Credi Regiser in Ialy. The preliminary esing of explanaory variables includes an enlarged se of macroeconomic indicaors comprising 22 facors bu, a he end of he mulivariae filering, only seven sysemic facors were elicied, namely economic growh, he degree of financial inermediaion, he shor-erm ineres rae, he real EUR/USD exchange rae, he iner-bank ineres rae spread, he inflaion rae and he price of oil. Ou of hese indicaors, wihin he individual specificaions, he presence of he eigh equaions forming he esimaed sysem is higher for he firs four indicaors. As far as new EU Member Saes are concerned, analyses in his paricular area are more difficul o idenify. Jakubik and Schmieder (2008) give an overview of he macroeconomic facors weighing on credi risk in he Czech Republic, compared o Germany. The paper feaures differen approaches o nonperforming loans generaed by households and hose by non-financial corporaions, wih empirical resuls showing some similariies beween he wo counries in erms of he se of exogenous variables. Boh in he Czech Republic and Germany he annual GDP growh rae and he year-earlier financial indebedness exer an influence on he defaul rae associaed wih corporae loans. In regard o households, one variable from eiher se of indicaors for he labour marke and he counerparies financial burden are included in he funcion form of regressions esimaed for he wo counries. The major deerminans idenified in he paper are he year-earlier unemploymen rae and he real ineres rae recorded hree quarers before in he case of he Czech Republic and he curren disposable income and he year-earlier financial indebedness in he case of Germany. Saring from he mehodological approaches summarised previously, we envisage developing a sysem o predic nonperforming loan raes ha will be able o idenify a secoral level he macroeconomic deerminans of he repaymen behaviour of debors, legal eniies, across he Romanian banking sysem. The operaional objecive of our research paper is o buid a se of economerical models o answer o four inerrelaed quesions: (i) Which are he componens of he macroeconomic variables se ha can singnal in advance he develompmens of he companies nonperforming loans?; (ii) Which are he lags and he individual inensiies of he macroeconomic relevan indicaors when affecing he companies overdue loans rae? (iii) Wha is he hierarchy of he facors wih impac on he dynamics of companies overdue loans?; (iv) Wha are he characerisics of he loans reinbursmen behaviour in case of shocks on real economy or money marke?

4 72 Bogdan Moinescu, Adrian Codirlaşu The paper is composed ou of hree secions and i concludes wih he main findigs and fuure research objecives. In he firs secion we described he mehodological framework for assesing companies overdue loans, saring wih he findings in he inernaional lieraure. The second par comprises of boh presenaion of he daa series ha were used and he economic raional and saisical evidence which lead us in selecing he explanaory variables. In he hird secion we decribed main empirical findings derived by consrucing and esing of he macroeconomic credi risk models. 2. Mehodological framework Mehodological soluions used in idenifying a symplified funcional form of he macroeconomic models for assesing credi risk are based on mulivariae regressions, esimaed independenly or simulaneously (in sysems of equaions). For capuring he general economic conex ha influence he repaymen behavior of companies in differen economic secors, he sensiiviy of rae of nonperforming loans o he developmens of he Romanian economy is assesed individualy on four major componens of economic aciviy. The mehodology used in his research paper is based on he approach used by Virolainen (2004), which developed he condiional credi risk model proposed by Wilson (1998). Consequenly, he concepual model for he forecasing sysem of he nonperforming loans raio is based on he hypohesis ha he funcional link beween he endogeneous variable and he se of explanaory variables is capured by he logi funcion. The funcional link is also invesigaed as a linear equaion. The mehodological soluion used in he research paper is based on he following sysems of equaions: RCN ln 1 RCN RCN ln 1 RCN RCN ln 1 RCN RCN ln 1 RCN AGR, AGR, IND, IND, COM, COM, CONS, CONS, n i1 m j1 n i1 m j1 y i y j y i y j s, j, s, j, AGR IND COM CONS, and (1)

5 Assessing he Secoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy 73 RCN RCN RCN RCN AGR, IND, IND, IND, n i1 m j1 n i1 m j1 y y y i y j i j s, j, s, j, AGR IND COM CONS The individual impac of macroeconomic facors on he evoluion of raes of defaul is analysed using firs diference (wih quaerly frequency) of he ime series using SUR (Seemengly Unrelaed Regression) economeric esimaion echnique. This economeric framework improves he efficiency of he esimaion by adjusing he coefficiens of boh equaions so ha o reduce he errors correlaion, as he significan correlaion beween he dependen variables of he sysem can generae a high errors correlaion. The framework for idenifying he deerminan facors is improved by using auoregressive vecors ha can beer idenify he persisence and impac in ime of he macroeconomic variables on he evoluion of raes of nonperforming loans. The relevance of he deerminan facors was compued based on individual conribuion o he adjused R squared for sysems of equaions esimaed using he SUR mehod and on variance decomposiion for a 24 monh ime horison in case of VAR models 3. Daa The dynamics of he overdue companies loans was illusraed as an overdue loans value weigh. A loan is considered overdue if he deb service is higher ha 90 days. The daa used for he empirical analysis cover quaerly informaion during he period of Q Q The beginning of he esimaion period is deermined by he availabiliy of daa for he dependen variables - companies defaul raios for he following secors: agriculure, indusry, services and consrucions. The independen variables were grouped ino wo subgroups: a) real economy; and b) moneary condiions (Table 1).

6 74 Bogdan Moinescu, Adrian Codirlaşu Lis of candidae variables Deerminans Real economy 1 Seasonaly adjused value added expressed boh in consan and curren prices for he above menioned economic secors Table 1 Expeced sign of impac on he dependen variables 2 Gross income - 3 Energy price index + 4 Fuel price index + 5 Quaerly FDI flow - 6 Quaerly expor value flow - Moneary condiions 1 Ineres rae for RON loans + 2 Ineres rae for euro loans + 3 1M and 3M EURIBOR rae + 4 Volume of morgage loans - 5 EURRON exchange rae + These daa were drawn up from he repors of he Saisics Naional Insiue and hose of he Naional Bank of Romania. The variables of he labor marke and value added were seasonally adjused based on he Tramo Seas mehodology in order o eliminae he impac of he seasonal facors on his marke, especially of he public secor. Moreover, as all he daa series were considered non-saionary and inegraed of order one, based on heir economic significance and saisical ess (ADF şi Phillips-Perron), heir saionalisaion was ensured hrough firs-difereniaion. Moreover, excep nonperforming loans raes and ineres raes, all he variables were considered in he economerical analysis in naural logarihms. 4. Empirical analysis For selecing he models specificaions, he saring poin was he esimaion of equaions sysems. Given he complexiy of he analysis, he selecion procedure for he deerminan facors consised in employing a reasonable number of models ha used a large diversiy of exogenous variables. Models specificaions, including number of lags, was chosen so ha o insure funcional sabiliy of he mechanism boh form he economical and economerical poin of view, aking ino accoun specific saisics as adjused R squared, informaional crieria, ess for regression errors. Empirical resuls lead o he selecion of four (relaively) disinc specificaions based on combining of wo mehodologies for regressions esimaion (linear and logisic) and wo measures of added value (in curren and consan prices) and he lags and he individual inensiies of he macroeconomic relevan indicaors when affecing he companies overdue loans rae were esimaed boh wih SUR sysems of equaions and wih VAR models (Table 2). -

7 Table 2 Sysems of equaions specificaions for he dynamics of he defaul raes Agriculure Indusry Services Consrucions Facors Models VA No VA No VA R VA R VA No VA No VA R VA R VA No VA No VA R VA R VA No VA No VA R VA R Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Real economy Value added of he secor (3)** (2)** (3)** - (3)** (3)** (2)*** (2)*** (3)*** (2)*** (3)*** (2)*** (3)** (4)*** Gross income (3)*** (3)*** (3)*** (3)*** (2)*** (2)** Fuel prices (1)* (1)*** (1)* (1)*** FDI (4)*** (4)*** (4)*** (4)*** (6)*** (6)* (8)*** (8)*** (2)*** (2)*** Expor (6)*** (6)* (4)*** (4)*** Moneary condiions RON ineres rae (2)*** (2)*** (2)*** (2)*** (2)*** (2)*** (2)* (2)*** (1)*** (1)** (1)*** (1)** EUR ineres margin (4)*** (4)*** (4)*** (4)*** (5)*** (1)* EURRON exchange rae (5)** (5)** (6)*** (5)* (6)** (5)*** Morgage loans (2)*** (2)** (2)*** (2)*** Consan Adjused R 70.4% 68.0% 70.9% 64.8% 54.2% 58.1% 59.4% 67.7% 76.8% 63.0% 76.6% 68.3% 84.8% 53.6% 77.5% 68.3% squared Durbin Wason Noe: (x) represens he impac inerval expressed in number of quarers.

8 76 Bogdan Moinescu, Adrian Codirlaşu Resuls of he saisical ess show ha he se of models comply wih demands of a good economerics performance. The response ime frame of he raes of defaul o a shock on macroeconomic and moneary variables was se based on he lag of he independen variables in he regression equaions. Srucural invesigaion of he deerminans se was followed by deailed invesigaion of he auoregressive profile, which emphasize he imporance of he second round effecs. The auoregressive profile analyzed by using vecorauoregression (VAR) models. These models evaluae he persisence of effecs of shocks on differen macroeconomic and moneary variables on he dynamics of defauls. Based on he resuls of he impulse-response funcions of defaul raes o macroeconomic and financial variables idenified in he SUR sysems of equaions esimaions procedure, 16 alernaive VAR models were esimaed. Thus each VAR model specificaion was consisen wih final specificaion of each equaions in he esimaed sysems of equaions, while he number of lags for each VAR was seleced so ha o insure is sabiliy and in he same ime o be consisen wih economic heory and saisically relevan (Adjused R-squared, Informaional crieria). According o analysis of he roos of characerisic polynomial, all he four VAR specificaions fulfill he sabiliy condiion. Moreover, according o he residuals correlogram, here is no auocorrelaion of he errors. For generaing he impulse-response funcions, he ime horizon aken ino accoun was 8 quarers and he inerpreaion of he resuls was based on he cumulaed impulse-response funcions of he wo raes of defauls o a shock on variables in he sysem. Resuls of he cumulaed impulse-response funcions, for a iome horizon of wo years are presened in Table 3.

9 Assessing he Secoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy 77 Agriculure Indusry Services Consrucions Table 3 Impulse-response funcions for he VAR models (1) a shock on gross income has an impac on he rae of defauls ha lass four quarers. (1) a shock on he RON loans ineres rae has a saisically significan impac on defaul rae afer one quarer; (2) a shock on FDI has an impac on defaul rae saring from he firs quarer and lass four quarers. (1) he impac of a shock on value added in services is saisically significan afer firs quarer and lass 3 quarers; (2) a shock on FDI has an impac on defaul rae saring from he fourh quarer; (3) a shock on RON loans ineres raes has an impac on defaul rae saring wih he fourh quarer and has a high percinency. (1) he impac of a shock on EUR loans ineres rae margin is saisically significan afer one quarer; (2) he impac of a shock on EURRON exchange rae is saisically significan afer hree quarers; (3) he impac of a shock on morgage loans has an impac on defaul rae saring wih he fourh quarer and has a high percinency. The economeric was deepened by variance decomposiion for boh simulaneous equaions and VAR models, in order o emphasizehow much of he variance of raes of defaul is explained by he variance of each deerminan macroeconomic or financial (Table 4). Facorial decomposiions shows ha he moneary variables are more imporan han hose of he real economy in explaining he dynamics of he rae of defauls for consrucion secor, while for agriculure secor he findings are opposie. For indusry and services secors, he models seems no o show a clear conclusion on his regard. The loan reimbursemen capaciy of he companies from he agriculure secor is deermined mainly by he dynamics of he gross income, foreign direc invesmens and ineres rae for RON denominaed loans. For indusry, he deerminans facors are he financing cos, expressed by he ineres rae for RON denominaed loans and he ineres rae margin for EUR denominaed loans and added value measured in consan prices. Facorial decomposiion for he services and commerce secor shows he imporance of he real economy variables, as value added and FDI. Among he moneary facors he conribuion of RON loans ineres rae is especially noiced, exchange rae having only a minor impac. Bu he exchange rae is especially imporan, along wih he risk premium for EUR denominaed loans and he evoluion of morgage loans, in explaining he dynamics of rae of defaul in he consrucion secor.

10 Facors Models Real economy Secor value added Gross income Fuel prices FDI Expor Moneary condiions Ineres rae for RON loans EU loans ineres margin EURRON exchange rae Morgage loans Table 4 Facorial decomposiion Agriculure Indusry Services Consrucions VA No VA No VA R VA R VA No VA No VA R VA R VA No VA No VA R VA R VA No VA No VA R VA R Me. Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi Linear Logi SE 4.4% 3.0% 4.9% 0.0% 17.2% 12.1% 14.4% 17.7% 39.8% 29.0% 38.6% 35.3% 2.8% 0.0% 24.5% 0.0% VAR 14.8% 28.3% 3.7% 8.6% 17.0% 20.4% 19.3% 11.8% 4.2% 13.2% 21.5% 12.2% 29.6% 33.8% 17.5% 44.9% SE 35.4% 39.0% 31.9% 39.8% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 8.6% 0.0% 25.3% VAR 45.3% 32.1% 46.0% 34.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% SE 2.4% 3.0% 1.9% 0.2% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% VAR 1.6% 3.7% 1.4% 2.9% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% SE 51.4% 41.0% 51.9% 33.8% 0.0% 2.1% 14.4% 7.7% 15.8% 0.0% 17.6% 0.0% 0.0% 0.0% 0.0% 0.0% VAR 5.3% 13.0% 5.7% 13.3% 6.1% 1.7% 1.3% 2.5% 18.5% 4.1% 7.3% 0.7% 0.0% 0.0% 0.0% 0.0% SE 0.0% 0.0% 0.0% 0.0% 20.2% 6.1% 11.4% 6.7% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% VAR 0.0% 0.0% 0.0% 0.0% 12.1% 8.9% 0.9% 28.2% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% SE 51.4% 47.0% 52.9% 44.8% 16.2% 26.1% 3.4% 10.7% 18.8% 5.0% 14.6% 3.3% 0.0% 0.0% 0.0% 0.0% VAR 11.4% 13.5% 20.7% 35.3% 11.0% 19.9% 8.6% 18.1% 33.9% 55.8% 40.3% 67.3% 0.0% 0.0% 0.0% 0.0% SE 0.0% 0.0% 0.0% 0.0% 9.2% 20.1% 36.4% 46.7% 0.0% 0.0% 0.0% 0.0% 14.8% 5.6% 0.0% 0.0% VAR 0.0% 0.0% 0.0% 0.0% 5.2% 9.9% 5.4% 8.8% 0.0% 0.0% 0.0% 0.0% 27.5% 29.9% 30.3% 14.2% SE 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 7.0% 0.0% 5.3% 6.8% 0.4% 5.5% 29.3% VAR 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 15.6% 5.9% 9.5% 7.4% SE 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 16.8% 4.6% 23.5% 35.3% VAR 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 11.4% 10.6% 11.0% 24.4% Real economy facors 80.3% 81.5% 73.7% 66.3% 36.3% 25.7% 30.8% 37.4% 39.1% 23.2% 42.5% 24.1% 16.2% 21.2% 21.0% 35.1% Average Moneary condiions facors Average 31.4% 30.2% 36.8% 40.1% 20.8% 38.0% 26.9% 42.2% 26.3% 34.0% 27.4% 38.0% 46.4% 28.5% 39.9% 55.3%

11 Assessing he Secoral Dynamics of Non-performing Loans: Signs from Financial and Real Economy 79 Among he real economy variables i is worh menioning especially he gross income; he added value for he consrucion secor having only a minor impac. The resuls are no surprising aking ino accoun he likely modes represenaion in he official daa of he acual level of new creaed value in his secor han he oher secors of he economy. 5. Conclusions Economeric resuls shows ha he dynamics of he real economy is he main driver of he evoluion of he raes of defaul for nonfinancial companies; he financial pressure induced by he moneary condiions, excep for he consrucions secor, being a secondary facor. The impac ime frames are generally low, he maximum effecs being regisered unil he firs year. Excepion is he FDI and expors volume, which negaively impac he rae of defaul in indusry afer six quarers and also he exchange rae ha have impac on raes of defaul in services and consrucions secors afer five or six quarers. Facorial decomposiion confirms he assumpion of a srong link beween he dynamics of he morgage loans and he dynamics of rae of defaul in consrucions secor. The linear models seems o be he preferred soluion excep for he indusry secor and he way o express he added value seems no o be of grea imporance for agriculure and services secors, bu for indusry he real value (in consan prices) is relevan and for consrucions he nominal value (curren prices) is more imporan. Acknowledgemens This work was suppored by he European Social Fund hrough Secorial Operaional Programme Human Resources Developmen , projec number POSDRU/89/1.5/S/59184 Performance and excellence in posdocoral research in Romanian economics science domain.

12 80 Bogdan Moinescu, Adrian Codirlaşu References Boss, M., A Macroeconomic Credi Risk Model for Sress Tesing he Ausrian Credi Porfolio, Financial Sabiliy Repor 4, OeNB. 2002, pp Boss, M., Fenz, G., Pann, J., Puhr, C., Schneider, M., Ubl, Eva, Modelling Credi Risk hrough he Ausrian Business Cycle: An Updae of he OeNB Model, Financial Sabiliy Repor no.17, OeNB, 2009, pp Fiori, Robera, Foglia, Anonella, Iannoi, Simonea, Esimaing Macroeconomic Credi Risk and Secoral Defaul Rae Correlaions for he Ialian Economy, 2nd Exper Forum on Advanced Techniques on Sress Tesing, Inernaional Moneary Fund and De Nederlandsche Bank, Amserdam, The Neherlands, Ocober 23-24, 2007 Jakubik, P., Schmieder, C., Sress Tesing Credi Risk: Comparison of he Czech Republic and Germany, Financial Sabiliy Insiue, Bank for Inernaional Selemens, FSI Award 2008 Winning Paper Kiyoaki, N., Moore, J., Credi Cycles, Journal of Poliical Economy, Vol. 105(2), 1997, pp Marcucci, J., Quagliariello, M., Is Bank Porfolio Riskiness Procyclical? Evidence from Ialy Using a Vecor Auoregression, Journal of Inernaional Financial Markes, Insiuions & Money, 18, 2008, pp Mwanza, N., Nonperforming Loans and Macrofinancial Vulnerabiliies in Advanced Economies, IMF Working Paper no. WP/11/161, 2011 O Hara, Ph.A., Inernaional Subprime Crisis and Recession: Emerging Macroprudenial, Moneary, Fiscal and Global Governance, PANOECONOMICUS, 1, 2011, pp Virolainen, K., Macro Sress Tesing wih a Macroeconomic Credi Risk Model for Finland, Bank of Finland, Discussion Papers, No. 18, 2004 Williamson, D.S., Financial Inermediaion, Business Failures, and Real Business Cycles, Journal of Poliical Economy, Vol. 95(6), 1985, 1985, pp Wilson C.Th., Porfolio Credi Risk, FRBNY Economic Policy Review, Vol. 4, No. 3, Ocober 1998