Assessing the financial vulnerability of Italian households: a microsimulation approach Valentina Michelangeli and Mario Pietrunti 1

Transcription

1 Assessing he financial vulnerabiliy of Ialian households: a microsimulaion approach Valenina Michelangeli and Mario Pieruni 1 Sepember 2014 A microsimulaion model o monior Ialian households financial vulnerabiliy is developed using household-level daa from he Survey on Household Income and Wealh. Microeconomic informaion is hen mached wih macroeconomic daa on deb and income in order o generae nowcass and forecass of he pah of households indebedness and deb-service raio. Wihin his framework, where households deb and income are updaed more frequenly han by employing survey daa alone, we find ha he dynamics of income growh are he main driver of households vulnerabiliy. The share of vulnerable households (defined as hose wih a deb-service raio below 30 per cen and income below he median) over he oal populaion is projeced o be abou sable beween 2012 and 2014, wih a sligh decrease in 2015 due o posiive income growh. Their deb is also projeced o decrease in hose years. JEL Classificaion: D14, G10. Keywords: households vulnerabiliy, deb, sress es Inroducion The indebedness of he household secor and is financial vulnerabiliy are acquiring a growing imporance from a financial sabiliy viewpoin, mainly following he boom recorded in several OECD counries in he period (OECD 2010) and he subsequen financial crisis. Wih he aim of closely monior households vulnerabiliy, we build a microsimulaion model using Ialian households daa available from he Survey on Household Income and Wealh (SHIW). The survey offers a comprehensive picure of he household secor, providing informaion on households idiosyncraic characerisics, such as income, balance shee, age, educaion and occupaion. The main issue wih survey daa is heir low frequency and, hence, he delay wih which new informaion becomes available (in he case of he SHIW, he daa is biennial and new daa become available wih abou a year s delay). In his paper we provide a mehodology ha inegraes microeconomic household daa wih higher frequency and more imely macroeconomic daa on income, ineres raes and oal deb. Such a oolki enables us o assess he financial condiions of he household secor in a imely manner and i can be used o evaluae he impac of scenarios of sress or oher possible policy inervenions. In line wih oher sudies (IMF, 2011, 2012, 2013; ECB, 2013), we ake as a reference indicaor of he healh of he household secor he deb-service raio (DSR), defined as he share of deb paymens o income, and we idenify as vulnerable hose households wih a DSR above a given hreshold. Moreover, as households wih low income are more severely affeced by negaive shocks, in his paper we focus on households wih income below he populaion median. Following a previous sudy on Ialian households (Magri and Pico, 2012), we idenify as vulnerable households wih DSR greaer han 30 per cen and income below he median. Our main resuls are as follows. 2 In a baseline scenario, wih almos fla ineres raes and posiive income growh (up o 2.9 per cen in 2015), he share of vulnerable households wih income below 1 Financial Sabiliy Direcorae, Bank of Ialy. We hank Paolo Angelin Anonio Di Cesare, Giorgio Gobb Silvia Magr Maeo Piazza and Luigi Federico Signorini for heir useful commens. The analysis and conclusions expressed herein are hose of he auhors and should no be aribued o he Bank of Ialy. 2 See Michelangeli and Pieruni (2014) for deails on a backesing exercise. 1

2 he median is projeced o be abou sable over he nex few years and equal o 2.7 per cen in A small decrease in 2015 in he number of vulnerable households would be mainly due o he expeced income growh. A slow deleveraging process for vulnerable households is expeced o ake place, parially suppored by an inward shif in credi supply and banks seleciveness in graning loans, wih he heir share of oal deb revering o 2010 levels. We also perform wo sress es simulaions: a projeced zero income growh in 2015 and a 100 basis poins increase in ineres rae. We find ha he share of vulnerable households increases more under he former, alhough even such a case does no represen a major change from baseline projecions. We are hen led o conclude ha Ialian households wih deb do no represen a major hrea for he financial sabiliy of he economic sysem. The paper is organized as follows: Secion 2 presens he daa, Secion 3 gives a descripion of he model, Secion 4 ses ou he resuls and Secion 5 concludes. 1. Daa In our exercise we make use of boh microeconomic variables (aken from he waves of he SHIW) and of a se of macroeconomic variables. Microeconomic variables The SHIW conains deailed informaion boh on households individual characerisics such as age, educaion, employmen saus of he head of household, and on income and deb. 3 Deb is classified in hree differen caegories: morgage deb on he primary residence, morgage deb on oher real esae, and consumer credi. For each deb caegory, we observe he ousanding amoun, he iniial amoun borrowed, he year when he loan was graned, he oal lengh of he conrac, he amoun of he annual insalmen, he ineres rae, and in case of morgage deb wheher i is adjusable rae or fixed rae. Our saring poin is he 2012 SHIW wave, he mos recen available. Some descripive saisics of he sample are provided. The fracion of households wih a morgage deb on a primary or secondary residence was abou 12.4 per cen of he populaion. The saring value of he deb on he primary residence was abou 115,000, while households had average ousanding deb of abou 78,000. Wih respec o he ype of morgage conracs, he sample is roughly equally spli in fixed and variable rae morgages. If we exend he analysis o all kinds of real esae deb, ha proporion remains abou he same. Abou 10 per cen of households had consumer credi deb. The SHIW is an unbalanced panel where only half of each wave s sample is reained in he nex wave of he survey. To overcome he difficuly arising from racking households from one wave o he nex, we make use of a simulaion sraegy similar o he one described in Djoudad (2010): o simulae he income process we group observaions according o heir income class, while o simulae new morgage originaions we gaher observaions in groups according o oher socioeconomic characerisics (age, educaion, occupaion). Macroeconomic daa As shown in he nex secion, he model is buil in such a way o mach he aggregae daa on oal income and amoun of oal deb. Table 1 presens he macro variables used in he model. 3 For a general descripion of he survey, see hp:// 2

3 Firs, we exrac he ime series on income growh from he naional accouns (Conabilià Nazionale, CN). According o he CN definiion, income includes impued rens and capures he sandard of living of households. Nominal income growh was negaive in 2013, while i is projeced o be posiive and equal o 2.4 per cen and 2.9 per cen respecively in 2014 and We also use projecions on lending volumes o households for house purchase as inpu of our model. These projecions are based on a macro-economeric model developed a he Bank of Ialy for inernal purposes. The dynamics of variable of ineres, which represens he volume of loans in banks balance shees plus an esimae of securiized loans, imply a negaive growh in oal deb in and 2014 and a slighly posiive growh (below 2 per cen) in As a las macro inpu, we make use of he projecions of daa on he hree-monh Euribor. Hisorical daa and projecions obained from fuures conracs imply an almos fla value for he Euribor rae wih one-period changes close o zero across he whole simulaion. Such a variable affecs he value of he insalmens of variable ineres rae morgage and he ones of new originaions. We choose he hree-monh Euribor rae as morgage raes in Ialy are ypically compued by applying a spread o such rae. 4 Table 1: Macroeconomic aggregaes (percenages) Income growh rae a curren prices Toal deb growh m Euribor The Model This secion describes he evoluion of households income and deb over ime. 2.1 Income growh dynamics In his paper we make use of wo slighly differen definiions of income for wo differen goals. The variable enering in he denominaor of he DSR is he disposable income gross of financial charges and ne of impued rens. Such a variable is he one ha more closely resembles he acual moneary income available o he household for curren expenses. The second definiion of income is disposable income and we employ i o mach he model saisics wih he macro daa. To compue he income growh dynamics we group households disposable income ino four classes of equal frequency (Djoudad, 2010). 5 The process for he income growh for each class j is given by: 4 The bank spread is assumed o remain fixed across simulaion periods. This is generally rue for exising conracs, apar from he case of morgage refinancing. 5 To ge household equalized income we divide household disposable income by a facor ha capures is number of componens. In each period, he hresholds for each class of equalized income is calculaed and each household is assigned o a specific class. 3

4 k y j log k y j, 1 i ~ N( µ j, σ j ), i for j=1,2,3,4 and k=d,g (1) where k alernaively sands for disposable income (d) or disposable income gross of financial charges and ne of impued rens (g). We esimae mean and variance of each class wih he SHIW daa from 2002 o The esimaed parameers are repored for each of he four income classes in Table 2. Table 2: Esimaed mean and sandard deviaion for he income processes y d growh y g growh d µ d σ μ g σ g 1s-25h percenile h-50h percenile h -75h percenile h -100h percenile The dynamics of he wo processes are raher similar: in boh cases means are posiive and slighly larger for low income households, which are he ones who benefi he mos from a gradual economic recovery. The sandard deviaion, in line wih oher sudies (see, for insance, Djoudad, 2010), is highes for lower groups and lowes for he upper groups. We assign a random income shock o each household in each period, while ensuring ha he mean and he variance for each class remain fixed a he values repored in Table 2. In his way we generae heerogeneiy among households while keeping aggregae dynamics under conrol. As a las sep, in order o mach he dynamics prediced by our macroeconomic daa, we correc income growh in each period by an adjusmen facor. 2.2 Deb growh dynamics Households can hold boh morgage and consumer credi deb. By evaluaing exising debs and accouning for loan originaions and erminaions, we can compue he evoluion of oal deb for each household over ime. For exising loans, he insalmen paymen is compued according o a French amorizaion schedule, which is he sandard schedule for morgages in Ialy. Such amorizaion schedule allows he per-period paymen o change following any modificaion in he ineres rae. For each household i=1,, N, wih N equals o he oal number of indebed households, and for each ype of deb y, he ousanding deb evolves as follows: MDeb, + 1 = MDeby, RPy, i (2) where RP y, i, is he annual paymen of he principal. Le R y, i, be he scheduled oal annual repaymen, which incorporaes he paymen of he principal and of he ineres, and is compued according o he formula: 4

5 A ry, R y, = MDeby, (1 + ry, ) (3) A (1 + r ) 1 y, where ineres r y, is he ineres rae, A is he residual duraion of he conrac. The annual paymen for RI is given by: RI, Thus, he principal repaymen could be calculaed as follows: = ry, MDeby, i. (4) RP Ry, RI y, =. (5) While morgage loans MDeb can be eiher variable or fixed rae, we assume ha for consumer deb CDeb he annual paymen R y, i, remains consan in he periods of he simulaion. This assumpion is reasonable given he span of he simulaion period. New originaions induce a composiion effec by modifying he number and average characerisics of indebed households. In order o capure he increased bank seleciveness afer he 2008 financial crisis, o esimae he projeced number of originaions we sar from he las hree waves (2008, 2010, 2012) of he panel componen of he SHIW. A morgage originaion occurs if a household has a morgage deb equal o zero a ime -1 and a posiive morgage deb a ime ( MDeb y, 1 = 0, MDeb > 0 ). Using he household groups consruced previously, we coun he number of originaions a ime -1 for each group k, θ k, 1, and we mainain i consan for he nex period ( θ k, = θ k, 1 ). According o he SHIW hisorical daa, abou half of morgages are variable rae ones. Hence, we mainain he same proporion for new originaions: abou 50 per cen of new originaions are variable-rae morgage and abou 50 of hem are fixed ones. Each household wih a new morgage is assigned an amoun of deb equal o he average deb of he group o which i belongs. Toal deb associaed wih loans originaions has been adjused o mach he macroeconomic daa on oal deb growh. Morgage erminaions, ceeris paribus, bring a reducion in he oal deb and, like morgage originaions, may induce a change in he characerisics of he pool of indebed households. Benefiing from he microeconomic households daa available in he SHIW, we have informaion on he loan lengh for each household. Hence, given our model srucure, we can projec he evoluion for each ype of deb by aking ino accoun is loan erminaion. The oal ousanding deb is given by he sum of morgage loans and consumer credi loans: Deb = MDeb + CDeb ) (6) ( y, y, y Therefore, he annual paymen is given by he oal annual paymens on morgage deb and consumer credi:, = Ry, i. (7) R i, y 5

6 3. Resuls In he baseline scenario, he share of vulnerable households in he oal populaion is expeced o slighly decrease from 4.8 per cen in 2012 o 4.4 per cen in 2015 (Figure 1). Likewise, he share of he mos vulnerable households, namely hose wih income below he median, moves from 2.9 per cen o 2.7 per cen in he same period. In 2013, he reducion in he share of vulnerable households follows from he cu in he ineres rae, which implies lower morgage insalmens boh for households holding a variable-rae morgage and for new borrowers. The reducion in he share of vulnerable households can be parially aribued o negaive credi growh. Insead, in he years , he posiive income growh is he main facor leading o he decrease in he fracion of vulnerable households. Alhough we expec a moderae growh in cred banks are likely o be selecive in graning loans o households; consequenly, hese wo divergen effecs may no affec in a significan way he share of vulnerable households in he economy. Figure 1: Percenage of vulnerable households in he populaion all below he median Noe: resuls are based on 50 simulaions of he model. The solid line represens median resuls; he dashed lines are resuls a boh he 10 h and he 90 h perceniles. Daa for 2009 and 2011 are inerpolaed via cubic splines. We hen consruc wo sress es scenarios o evaluae how he share of vulnerable households changes (see Figure 2 and Tables in he Appendix for deailed resuls). Firs, he effecs of an increase of 100 basis poins in he Euribor rae in 2015 (from 0.3% o 1.3%) have been evaluaed. I induces an increase in he paymen of he morgage insalmens for households wih a variable-rae morgage as well as for new borrowers: he share of all vulnerable households and of hose wih income below he median reaches respecively 4.6 per cen and 2.9 per cen in Second, income growh is se equal o zero in 2015, affecing he income of all households. In his scenario, he all vulnerable households and of hose wih income below he median moves respecively o 4.7 per cen and 3.0 per cen in

7 Figure 2: Percenage of vulnerable households under alernaive scenarios However, in order o evaluae he hrea posed by he household secor o he financial sabiliy of he sysem i is relevan o projec he deb a risk, namely he deb held by vulnerable households. Given ha households wih income below he median represen he mos vulnerable ones, we will focus on heir deb (Figure 3). In he baseline scenario, he share of oal deb held by households wih income below he median decreases from 20 per cen o abou 16 per cen in 2015, in line wih he 2010 daa. Insead, under he firs and second sress scenarios, he deb a risk is projeced o be respecively 17.1 per cen and 17.3 per cen in Those resuls sugges ha here is no major hrea for he financial sabiliy coming from he Ialian household secor: boh he number of vulnerable households and heir deb are relaively low. Figure 3: Percenage of oal deb held by vulnerable households wih income below he median baseline ineres rae shock income shock

8 Conclusions In his paper, we presened a microsimulaion model o evaluae and monior he financial vulnerabiliy of he household secor. Saring from he microeconomic daa on Ialian households as repored in he SHIW and augmening hem wih macroeconomic projecions on income, deb and ineres rae, we were able o esimae households indebedness and deb-service raio over ime. In he baseline scenario, he mos vulnerable households, namely hose wih DSR greaer han 30 per cen and income below he median, are projeced o equal 2.7 per cen of he populaion in 2015, holding abou 16 per cen of he oal deb of he household secor. All in all and under alernaive and reasonable scenarios of sress, he model resuls sugges ha he household secor does no pose major hrea o he financial sabiliy of he sysem: he number of vulnerable households and heir deb are relaively small and almos sable in he nex few years. 8

9 References Djoudad, Ramdane (2010), "The Bank of Canada s analyic framework for assessing he vulnerabiliy of he household secor." Financial Sysem Review, European Cenral Bank (2013), The Eurosysem Household, Finance and Consumpion Survey, Resuls from he firs wave, Saisics Paper Series, No.2. IMF (2011), Unied Kingdom Vulnerabiliies of Household and Corporae Balance Shees and Risks for he Financial Secor Technical Noe, July IMF (2012), Spain Vulnerabiliies of Privae Secor Balance Shees and Risks o he Financial Secor, Technical Noe for he Financial Secor Assessmen Program Updae, May IMF (2013), Ialy The financial siuaion of Ialian households and non-financial corporaions and risks o he banking sysem, Technical Noe for he Financial Secor Assessmen Program Updae, Augus Magr Silvia, and Raffaella Pico (2012), L Indebiameno delle Famiglie Ialiane Dopo la Crisi del 2008, Banca d Ialia, Quesioni di economia e finanza (Occasional Papers), No.134. Michelangel Valenina, and Mario Pieruni (2014), A microsimulaion model o evaluae Ialian households financial vulnerabiliy, Banca d Ialia, Quesioni di economia e finanza (Occasional Papers), forhcoming. OECD (2010), OECD Facbook 2010: Economic, Environmenal and Social Saisics, OECD Publishing. 9

10 Appendix A1. Deailed resuls 1- Baseline scenario Percenage of vulnerable households over oal households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal Percenage of deb held by vulnerable households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal Noe: households are divided ino classes according o heir equalized income gross of impued rens. The repored values have been approximaed o he firs decimal. 10

11 2 - Sress es scenarios a) Ineres rae shock b) Income shock Percenage of vulnerable households over oal households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal Percenage of deb held by vulnerable households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal Percenage of vulnerable households over oal households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal Percenage of deb held by vulnerable households 1s-25h percenile h-50h percenile below he median h -75h percenile h -100h percenile Toal