Wha nfluences he growh of household deb? Dag Hennng Jacobsen, economs n he Secures Markes Deparmen, and Bjørn E. Naug, senor economs n he Research Deparmen 1 Household deb has ncreased by 10 11 per cen annually snce 2000. In he followng, he facors underlyng he srong growh n deb are analysed usng an emprcal model. The deb growh of recen years s found o be relaed o developmens n he housng marke and o he declne n neres raes snce December 2002. As a resul of he sharp rse n house prces from 1998 o 2001, deb growh remaned a a hgh level whle house prces declned n he laer half of 2002 and no 2003. Ths reflecs ha only a small poron of he housng sock changes hands each year. Even f house prces level off followng a rse, here wll be a long perod durng whch houses change hands a a hgher prce han he las me hey were sold. An ncrease n house prces wll herefore conrbue o deb growh for a long me. Households may ncrease her deb furher by rasng loans o fnance consumpon and nvesmen wh collaeral n he ncreased value of her dwellngs. Ths ype of borrowng has probably ncreased n recen years. Inroducon Norges Bank shall promoe prce sably and fnancal sably. Moneary polcy s orened owards achevng low and sable nflaon, defned as an annual rse n consumer prces of close o 2½ per cen over me. A he same me, moneary polcy can affec fnancal sably, snce he neres rae nfluences prvae secor deb and prces for houses and secures. Srong growh n deb and n asse prces may resul n fnancal mbalances (Boro and Lowe, 2002). Such mbalances may weaken he sably of he fnancal secor and resul n unsable nflaon and employmen. Household deb has ncreased by 10 11 per cen annually for he pas fve years. The srong growh n deb s ofen arbued o rsng house prces and hgh urnover n he housng marke. However, deb growh remaned above or close o 10 per cen even when house prces declned n he laer half of 2002 and no 2003 (see Char 1). Ths ndcaes ha house prces nfluence deb wh a consderable me lag. The fall n neres raes snce December 2002 may explan why deb growh acceleraed n he second half of 2003 and frs quarer of 2004. The purpose of hs arcle s o shed lgh on facors ha nfluence he growh of household deb. In parcular, we evaluae how deb growh hnges on developmens n he housng marke. We esmae a model of household deb on quarerly daa from 1994 Q1 o 2004 Q1. The model conans effecs of house prces, he housng sock, he number of house sales, banks' lendng raes, he unemploymen rae, oal wage ncome n he economy and he number of sudens aged 20 24 as a share of he oal populaon. An earler verson of he model was presened n Inflaon Repor 2/03. Facors ha nfluence household deb Household deb s deermned by demand for loans and banks lendng polcy. In hs secon we dscuss (a) he relaonshp beween households deb and her behavour n he housng marke, (b) demand for loans o fnance consumpon and nvesmen and (c) banks behavour. The relaonshp beween households deb and her behavour n he housng marke Household deb s largely relaed o he purchase of dwellngs. A household buyng a dwellng for he frs me wll normally deb-fnance he purchase o a large exen. Esablshed households wll also normally ncrease her borrowng f hey purchase a more expensve dwellng han he one hey already own. Developmens 103 1 Wh hanks o Arne Bråen, Kjers-Gro Lndqus, Kjel Olsen, Tore Anders Husebø, Ben Vale, Hanne A. Gravnngsmyhr and Jens Olav Sporasøyl for useful conrbuons and commens. The analyss was performed wh he ad of PcGve 10.1 (Hendry and Doornk 2001) and Samp 6.2 (Koopman, Harvey, Doornk and Shepard 2000).
104 n he housng marke are herefore mporan for deb growh. Snce dfferen ypes of house sales have dfferen effecs on gross deb, s useful o classfy hese sales. We dsngush beween purchases of new homes, frs-me purchases and las-me sales of resale homes, and sales of resale homes beween households ha are neher enerng nor leavng he housng marke. Purchase of new homes If a household rases a loan o buy a new home, s reasonable o assume ha households oal gross deb wll ncrease correspondngly. Ths s because he seller s no normally anoher household ha can use he sales sum o repay deb. For a gven house prce level, growh n he housng sock wll herefore resul n an ncrease n gross household deb. An ncrease n prces for new dwellngs wll furher ncrease hs deb. Frs-me purchases and las-me sales of resale homes When a household eners he resale home marke, anoher household wll of necessy have o leave. Ths household wll free up resources. If he whdrawn equy s used for purposes oher han repayng deb, oal gross household deb wll ncrease f he buyer debfnances par of he purchase. If some of he whdrawn equy s used o repay deb, he oal gross deb wll ncrease f he ncrease n he buyer's deb s larger han he reducon n he seller's deb. A household ha leaves he housng marke wll normally have enered he marke a number of years prevously. The resdual housng loan wll herefore normally be smaller han he loan of he frs-me buyer. Hence oal gross deb wll ncrease when he house s sold. Wha happens o deb f he prce of resale homes rses? We consder a frs-me buyer who enrely loanfnances he purchase of a resale home. If house prces ncrease, wll requre a larger loan o buy he dwellng han he prevous me was sold. The prce ncrease wll herefore conrbue o ncreasng he buyer s deb. The more house prces ncrease, he more resources a household ha leaves he resale home marke wll free up. However, he household s deb s no affeced by he fac ha he dwellng has ganed n value. The prce ncrease wll herefore resul n an ncrease n he gross deb of households as a whole. Sales beween households ha neher ener nor leave he resale home marke We consder a suaon n whch only resale homes are sold, and none of he households eher ener or leave he resale home marke. Some households wsh o purchase a dwellng ha s larger (and more expensve) han he one hey own. In order for hem o be able o do hs, oher households mus wan o buy a smaller (and less expensve) dwellng. We consder a suaon wh consan house prces. Households ha purchase a more expensve dwellng, sell her old dwellng and fnance he dfference by means of a loan. Those buyng a less expensve dwellng wll free up resources. If he morgage equy hey whdraw s used n s enrey o repay deb, oal gross deb wll reman unchanged. However, deb wll ncrease f par of he whdrawn equy s spen on consumpon. Wha happens o deb f house prces ncrease n hs case? Assume ha house prces ncrease by 10 per cen per square mere. If a gven exra number of square meres are purchased, he prce per dwellng wll also ncrease by 10 per cen. The deb of hose loan-fnancng some of he dfference wll ncrease accordngly. Those purchasng a smaller, less expensve dwellng, wll free up more resources han before he prce ncrease. Ther deb wll no be affeced, however. The prce ncrease wll herefore resul n hgher gross deb for households as a whole. The less he freed up resources used o repay deb, he greaer he ncrease n gross deb. The sgnfcance of house sales The examples above show ha oal deb may ncrease when houses change hands. Ths ncrease n deb wll be reversed as he deb s repad. However, assume ha he rae of urnover ncreases permanenly. Then here wll always be more persons han prevously who have recenly aken up loans. Hence he deb level wll ncrease, also n he long erm. Adapaon o he new deb level wll be relavely slow. Assume ha orgnally 10 resale homes are sold each year, and ha one household rases a loan n connecon wh each sale. Assume furher ha he number of sales ncreases o 20 resale homes per year. The number of households ha have aken up housng loans n he las 5 years wll hen ncrease from 50 o 60 he frs year, and from 60 o 70 he followng year. Afer 5 years, 100 households wll have aken up housng loans. Ths s he new equlbrum level. The channels from house prces o deb descrbed above are dependen on dwellngs beng sold. If urnover ncreases, he effec of a hgher house prce wll be amplfed. I s lkely ha hgher house prces move n andem wh hgher urnover: ncreased demand for dwellngs wll resul n a rse n prces and hgher urnover f he supply of resale homes depends posvely on he prce, whch s a reasonable assumpon. In perods of low demand and low prces many wll wa o sell unl prces pck up. Increased urnover may also resul n ncreased borrowng o cover agens fees, ax on legal documens, redecorang and he purchase of furnure and whe goods.
Hgher house prces wll conrbue o deb growh for a long me Now assume ha house prces ncrease (sharply) and hereafer sablse a a new level. Some houses are sold durng he prce rse, and household gross deb herefore ncreases hrough he channels descrbed above. Afer a whle prces wll sablse, bu for a long me here wll be houses ha are sold for a hgher prce han he las me hey changed hands. In prncple, he rse n prces wll conrbue o deb growh unl he enre housng sock has been sold a he new prce level. Abou 4 per cen of he housng sock changed hands n 2001. 2 If 4 per cen of he housng sock s sold each year, a prce rse oday could heorecally conrbue o growh n deb for 25 years. Demand for loans o fnance consumpon and oher nvesmens Some households ake up loans for redecorang and nvesmen n fnancal asses, o purchase cars and oher consumer durables and o purchase houses, cabns and aparmens ha are no used daly. Ths demand for loans depends largely on neres expenses, housng wealh, households ncome and her assessmen of her fuure capacy o pay off her deb. Increased ncome and/or lower neres expenses enable households o servce hgher deb. Moreover, wll be relavely more aracve o borrow han o save f neres raes fall. Demand for loans wll herefore ncrease. Households assessmen of her fuure capacy o pay s probably sensve o changes n he labour marke. Hgher unemploymen may lead o expecaons of lower wage growh and greaer uncerany concernng fuure ncome. Ths wll curb demand for loans. A rse n house prces may resul n ncreased demand for loans o fund consumpon and oher nvesmens va a wealh effec and va a prce effec. 3 Hgher house prces resul n ncreased housng wealh. The expeced fnal wealh (nherance) wll also ncrease f he prce ncrease s expeced o perss. Some households may wsh o whdraw some of hs gan n he form of ncreased consumpon. They wll hen eher reduce her fnancal wealh or ncrease her demand for loans. The prce ncrease may also conrbue o reducng he borrowng rae facng households (he prce effec). Ths reflecs ha () housng loans are secured by collaeral n he dwellng and () oher ypes of loan have weaker or no collaeral and herefore a hgher neres rae. The collaeral value of houses ncreases f banks or oher provders of cred expec he prce ncrease o perss. Ths wll ncrease households possbly of rasng loans secured by collaeral n her dwellng, a lower neres raes han raes on oher loans. The hgher collaeral value may also resul n a lower neres rae on housng loans. Snce house prces have fallen n perods durng he las 20 years, a prce ncrease wll no necessarly be perceved as permanen. Banks and households wll probably wa and see before makng any change n her behavour. Ths mples ha deb s nfluenced wh a me lag when house prces change. Emprcal sudes have produced evdence ha house prces affec prvae consumpon n Norway. 4 Persons of a maure age wh low resdual housng loans may have a parcular endency o rase loans o fund consumpon and oher nvesmens wh collaeral n (ncreased) propery values. The deb of maure age groups has ncreased subsanally n recen years. Char 2 ndcaes ha he ncrease n deb s relaed o he precedng rse n house prces. The ncrease n deb may also reflec a shf n household preferences: may have become more acceped o leave morgaged dwellngs o he nex generaon. The rse n house prces may have conrbued o hs by resulng n ncreased (expeced) fnal wealh for maure households. The ncrease n housng wealh has reduced he need o save fnancal wealh for he nex generaon. Mos sudens ake up suden loans. In addon, persons wh hgher educaon normally ake up hgher housng loans han hose whou hgher educaon (all else beng equal). There s herefore reason o beleve ha gross deb wll ncrease wh he share of sudens n he populaon. An ncrease n he share of sudens wll herefore conrbue o deb growh for a long perod. Frs, mos suden loans ncrease hroughou he sudy perod. Second, for a gven oal populaon, here wll be more new sudens each year han was he case prevously. As 105 2 The esmae s based on he sales fgures used below and daa on he number of occuped homes from he 2001 populaon and housng census. 3 The sgnfcance of hese relaonshps has been dscussed exensvely n recen years. See for example Debelle (2004) and arcles n The Economs. 4 See Brodn and Nymoen (1992), Eka and Nymoen (1992), Harldsad and Nymoen (1993), Brubakk (1994), Frøland (1999), Erhem, Jansen and Nymoen (2002), Boug, Dyv, Johansen and Naug (2002, Chaper 5.3) and Erlandsen (2003).
106 a resul, he share of persons wh hgher educaon (and suden loans) wll ncrease over a perod of years. The conrbuon o deb growh wll aper off when hs share sablses (cf. smlar reasonng n he secon abou changed sales above). Banks behavour Households rase a large proporon of her deb n prvae banks. Banks lendng polcy may herefore be mporan o deb growh. Ths polcy depends on banks' profably, on cusomers' (expeced) capacy o pay and on he value of her collaeral. 5 Banks may become more relucan o exend loans f her profably deeroraes, f he value of he collaeral decreases or f cusomers (are expeced o) become less able o pay off her deb. Some cusomers may hen be raoned or be offered such poor borrowng erms ha hey do no wsh o ake up loans (any longer). The supply of loans wll herefore depend posvely on households housng wealh and ncome, ncludng neres ncome and expenses. As noed above, ncreased unemploymen wll gve rse o expecaons of lower wage growh and ncreased uncerany abou households fuure paymen capacy. Ths wll probably reduce he supply of cred o households. 6 A rse n defauls by boh enerprses and households may also cause banks o be more cauous abou exendng loans o households. Ths dscusson ndcaes ha he supply of cred wll have an ndependen effec on he demand for dwellngs. If hs s he case, house prces and household deb should n prncple be modelled smulaneously. However, we do no fnd sgnfcan effecs of household deb n he house prce model presened n Fnancal Sably 1/04; 7 banks behavour s capured by oher varables n he model (house prce, neres rae, unemploymen, housng sock and wage ncome). The deb equaon below s herefore esmaed usng he mehod of ordnary leas squares. We es for effecs of defauls, however. A model of household deb We model households domesc gross deb as measured by he C2 cred ndcaor. Ths deb consss of loans from domesc banks, morgage companes, fnance companes, governmen lendng nsuons, lfe and non-lfe nsurance companes, prvae and muncpal penson funds, he Governmen Publc Servce Penson Fund and Norges Bank. Household bond and shor-erm paper deb rased n he domesc marke s also ncluded. The C2 fgures for household deb exend back o December 1995. We have exended he me seres backwards wh growh raes for household gross deb as measured n he RIMINI daabase (RIMINI s a macroeconomc model developed n Norges Bank). Ths deb consss of ax deb, foregn deb, deb o non-fnancal enerprses and deb ncluded n C2 less bond and shorerm paper deb. We sared wh a flexble dynamc model ha conaned effecs of house prces, he housng sock, he number of house sales, banks lendng rae afer ax 8, he unemploymen rae, oal wage ncome 9 n he economy, he number of defauled loans (for boh households and he publc n general) and he number of sudens aged 20 24 as a share of he oal populaon. In addon we ncluded a sochasc rend 10 o capure effecs of changed preferences among maure age groups n he esmaon perod. We hen smplfed he general model by placng resrcons on he coeffcens ha were no rejeced by he daa and ha smplfed he nerpreaon of he dynamcs. The number of defauled loans has a sgnfcan negave effec f we sar he esmaon n he second quarer of 1993 or earler (we have defaul fgures for he perod 1990 Q3 o 2003 Q4). If we nsead sar he esmaon n 1994 Q1 or laer, he defaul varables have an nsgnfcan posve effec. These resuls ndcae ha () subsanal defaulng, among boh enerprses and households, conrbued o banks lmng cred growh n he perod mmedaely afer he bankng crss a he begnnng of he 1990s and () developmens n defaulng have no had any major effec on deb growh snce 1993, even hough he number of defauled loans has ncreased n recen years. We herefore choose o sar he esmaon n 1994 Q1. 11 We smplfed he rend o a consan whou he f beng sgnfcanly weakened. In oher words, we dd no fnd sgnfcan effecs due o changed preferences (n he esmaon perod) beyond hose ha are capured by varables n he model. 12 The rse n house prces may have changed he preferences of households of maure age n recen years (see secon 2). The preferred model s specfed n a separae box. 5 See Sglz (1992, secons 6.2 6.3) for a heorecal dscusson. 6 Frøyland and Larsen (2002) esmae a model for banks losses on loans o households. They fnd ha losses ncrease wh deb, he neres rae level and he unemploymen rae; ncreased ncome and housng wealh resul n lower losses. 7 The model has been esmaed on daa from 1990 Q2 o 2004 Q1. Household deb has a sgnfcan effec n models of Norwegan house prces ha are esmaed usng daa from he 1980s and 1990s (see Erhem (1993) and Boug e al. (2002, Chaper 5.5)). 8 We also red o nclude varous measures of he real afer-ax neres rae. Equaons wh consan nflaon expecaons fed bes. We are herefore omng nflaon expecaons from he model (he effecs of consan nflaon expecaons are capured by he nercep). 9 Tax-movaed flucuaons n share dvdends have had a major effec on he measured developmens n household ncome n recen years. We herefore choose o use wage ncome nsead of dsposable ncome as an explanaory varable. 10 A sochasc rend s more flexble han a lnear (deermnsc) rend, and can capure effecs of excluded flucuang varables. The formulaon s descrbed n Koopman e al. (2000). 11 We could nsead have esmaed a model n whch he parameers vary from one regme o he nex. Aron and Muellbauer (2000) esmae such a model for deb growh n Souh Afrcan households. 12 The same concluson s reached f we es for such effecs usng dummy varables.
A model of household deb deb = 1.00 housngsock 0.29 (deb housngsock) -1 0.29 INTEREST (2.64) (5.24) + 0.02 urnover -2 + 0.01 ( 4 ncome + houseprce ) 0.03 unemploymen (3.01) (1.52) (3.89) 0.07 [deb houseprce housngsock + 1.70 INTEREST 0.17 urnover 0.64 sudenshare] -1 (7.41) (3.16) (1.36) (4.95) Esmaon perod: 1994 Q1 2004 Q1. σ = 0.0019, DW = 2.20. Esmaon mehod: Ordnary leas squares Absolue values are gven n brackes under he esmaes. s a dfference operaor: X = (X X -1 ), 4 X = (X X -4 ) 107 The varables and es sascs are defned as (small leers ndcae ha varables are measured on a logarhmc scale): deb = Households domesc gross deb (Source: Norges Bank, NB) housngsock = Value of housng sock measured a consan prces (Source: Sascs Norway, SN) INTEREST = Banks average lendng rae. (Source: NB) urnover = Number of house sales (Sources: SN and Norwegan Federaon of Cooperave Housng Assocaons) ncome = Toal wage ncome n he economy. Depends on he wage level and employmen (Source: SN) houseprce = Prce ndex for resale homes (prce per m 2 ) (Sources: NEF, EFF, FINN.no, ECON and NB) unemploymen = Unemploymen rae (Source: The Drecorae of Labour) sudenshare = No. of sudens aged 20 24 years as a share of he populaon. Average for 5 quarers (Source: SN) σ = Sandard devaon of regresson resduals DW = Durbn Wason es sasc The expresson n brackes measures he devaon beween deb n he prevous quarer and an esmaed long-erm relaonshp beween deb, house prces, he housng sock, banks lendng raes, house sales and he share of sudens. The model also conans an nercep and effecs of seasonal varaon. I has sable coeffcens and passes sandard ess for auocorrelaon, normaly and heeroscedascy. Deb, he neres rae and he housng sock are measured a he end of each quarer. The oher varables are measured as a quarerly average. The values of INTEREST and ncome for 2004 Q1 are based on esmaes from Inflaon Repor 1/04.
The model s an error correcon model for he logarhm of household gross deb. I conans effecs of house prces, he housng sock, urnover, nomnal neres raes 13, unemploymen, wage ncome and he suden share as dscussed above. Char 3 shows ha he model fs well. 108 How s deb affeced by shfs n he explanaory facors? The model mples ha household gross deb wll ncrease by 1¾ per cen hrough he frs year and by 10 per cen n he long erm f house prces ncrease permanenly by 10 per cen and he oher facors reman unchanged. The resuls confrm ha hgher house prces wll conrbue o deb growh for a long me (see Char 4). Abou half of he effec wll have maeralsed afer 3½ years, and 90 per cen afer 10 years. Afer 25 years, household deb wll have ncreased by 9¾ per cen. A 10 per cen ncrease n he housng sock wll also ncrease deb by 10 per cen n he long erm, for a gven house prce. In keepng wh he dscusson above, we have spulaed ha he long-erm effec of a change n he housng sock wll be acheved already n he frs quarer. Ths resrcon s no rejeced by he daa. Developmens n he housng sock wll also affec deb by nfluencng house prces. The house prce model n Fnancal Sably 1/04 mples ha house prces wll fall by 17 per cen n he long erm f he housng sock ncreases by 10 per cen. An ncreased housng sock wll hus resul n lower deb n he long erm f we ake no accoun ha house prces are also affeced. Accordng o he model, deb wll ncrease by 17 per cen n he long erm f urnover ncreases by 10 per cen (urnover ncreased by 12½ per cen from 2000 o 2003). Adapaon s slow: deb wll only have ncreased by 7 per cen afer 2 years and by 10 per cen afer 4 years. Ths slow adapaon s conssen wh he dscusson n secon 2. The model mples ha deb wll declne by ½ per cen durng he frs year and by 1¾ per cen n he long erm f banks lendng raes ncrease by one percenage pon and he oher varables reman consan. An ncrease n he neres rae wll also affec deb growh va a number of he oher varables n he model. The house prce equaon n Fnancal Sably 1/04 mples ha house prces wll fall by 3¼ per cen f he neres rae rses by 1 percenage pon and oher explanaory facors for house prces reman unchanged. The analyss ndcaes ha deb wll only fall by ½ per cen n he frs wo years f he unemploymen rae ncreases from 4 o 5 per cen and he oher varables reman consan; he long-erm effec s equal o zero. An ncrease n wage ncome also has a lmed effec on deb for a gven house prce. Deb growh s neverheless sensve o changes n unemploymen and wage ncome, snce hese varables have a srong nfluence on house prces (accordng o he house prce equaon n Fnancal Sably 1/04). Household deb ncreases by 6 per cen f he share of sudens ncreases by 10 per cen. Half of he effec wll have maeralsed afer 4 years (cf. dscusson n secon 2). Decomposon of deb growh The early 1990s were characersed by a bankng crss, hgh neres raes and hgh unemploymen. The banks problems probably conrbued o lmng deb growh (see above). Unemploymen and neres raes have fallen subsanally snce he early 1990s, resulng n a sharp rse n house prces (see Char 1 and he house prce equaon n Fnancal Sably 1/04). We fnd ha an ncrease n house prces wll conrbue o deb growh for a long perod. The growh n deb over he las 10 15 years can herefore be parally vewed as an adjusmen from a suaon wh a bankng crss, hgh neres raes and hgh unemploymen, o a new suaon wh relavely low neres raes, relavely low 13 We exclude he ax deducon for neres on deb, as hs s consan hroughou he esmaon perod.
unemploymen and a smoohly funconng cred marke. Char 5 decomposes he deb growh of he las wo years (up o 2003 Q1) n accordance wh he esmaed model (see appendx for an accoun of he mehod of decomposon). The calculaons show ha he rse n house prces pushed up four-quarer growh by abou 8 percenage pons n he perod 2002 Q1 o 2003 Q1. Ths llusraes ha a change n house prces affecs deb growh wh a consderable me lag: alhough house prces fell durng he las par of 2002 and no 2003, deb growh was mananed a a hgh level by he sharp rse n house prces from 1998 o 2001. The conrbuon from house prces has declned n he pas year as a resul of he sluggsh prce developmens n he las par of 2002 and frs par of 2003. The ncrease n house sales conrbued posvely o deb growh n he perod 2002 Q1 o 2004 Q1. Developmens n neres rae, unemploymen, wage ncome and he housng sock nfluence deb growh drecly and by affecng house prces. Char 5 shows ha new consrucon has pushed up four-quarer growh by 2 percenage pons n he las 2 years, all else beng equal. However, new consrucon may have curbed he rse n house prces by 3 4 percenage pons n he same perod (see box on house prces n Fnancal Sably 1/04). The reducon n neres raes snce December 2002 has pushed up deb growh by ½ 1½ percenage pons hs pas year for gven house prces. Moreover, he declne n neres raes has conrbued o boosng he rse n house prces. These facors n solaon wll resul n hgher deb growh n he years ahead. Increased unemploymen n 2002 and 2003 pushed down deb growh by ¼ ½ percenage pon n he perod 2002 Q1 o 2003 Q4. Developmens n unemploymen made a posve conrbuon o deb growh n 2004 Q1. However, he rse n unemploymen n 2002 and 2003 dampened he rse n house prces. Developmens n wage ncome have prmarly nfluenced deb growh by affecng house prces. Concluson The growh of deb n Norwegan households has been hgher han ncome growh n recen years. The hgh deb growh s found o be relaed o developmens n he housng marke and o he declne n neres raes snce December 2002. As a resul of he sharp rse n house prces from 1998 o 2001, deb growh remaned a a hgh level whle house prces declned n he laer half of 2002 and no 2003. Ths reflecs ha only a small poron of he housng sock changes hands each year. Even f prces sablse followng a rse, here wll be a long perod durng whch houses change hands a a hgher prce han he las me hey were sold. An ncrease n house prces wll herefore conrbue o deb growh for a long me. Household deb may ncrease furher because hgher house prces may resul n hgher fnal wealh and beer borrowng condons for many households. These households wll hen have a greaer ncenve o rase loans secured by collaeral n her dwellng o fnance consumpon and nvesmen. Ths ype of borrowng has probably ncreased n recen years. References: Aron, Janne and John Muellbauer (2000): Fnancal lberalsaon, consumpon and deb n Souh Afrca. Cenre for he Sudy of Afrcan Economes, Workng Paper Seres 132 Boro, Claudo and Phlp Lowe (2002): Asse prces, fnancal and moneary sably: explorng he nexus. BIS Workng Paper no. 114 Boug, Pål, Yngvar Dyv, Per R. Johansen and Bjørn E. Naug (2002): MODAG en makroøkonomsk modell for norsk økonom (MODAG a macroeconomc model of he Norwegan economy). Sosale og økonomske suder 108, Sascs Norway Brodn. P. Anders and Ragnar Nymoen (1992): Wealh effecs and exogeney: he Norwegan consumpon funcon. Oxford Bullen of Economcs and Sascs 54, pp. 431 454 Brubakk, Lef (1994): Esmerng av en konsumfunksjon for kke-varge goder 1968 1991 (Esmang a consumpon funcon for non-durables 1968 1991). Rapporer 94/9, Sascs Norway 109
110 Debelle, Guy (2004): Household deb and he macroeconomy. BIS Quarerly Revew, March 2004, pp. 51 64 Eka, Kar H. and Ragnar Nymoen (1992): Fnansell konsolderng som konjunkurfakor (Fnancal consoldaon as a cyclcal facor). Penger og Kred no. 1/1992, pp. 29 38 Erhem, Øyvnd, (1993): En dynamsk modell for bolgprsene RIMINI (A dynamc model of house prces n RIMINI). Penger og Kred no. 4/1993, pp. 288 297 Erhem, Øyvnd, Elev S. Jansen and Ragnar Nymoen (2002): Progress from forecas falure he Norwegan consumpon funcon. Economercs Journal 5, pp. 40 64 Erlandsen, Solveg (2003): Age srucure effecs and consumpon n Norway, 1968(3) 1998(4). Workng Paper 2003/1, Norges Bank. Frøland, Gsle (1999): Økonomersk modellerng av husholdnngenes konsum Norge. Demograf og formueseffeker (Economerc modellng of household consumpon n Norway. Demography and wealh effecs). Noaer 1999/86, Sascs Norway Frøyland, Espen and Ka Larsen (2002): How vulnerable are fnancal nsuons o macroeconomc changes? An analyss based on sress esng. Economc Bullen no. 2/2002, pp. 92 97 Harldsad, Anders and Ragnar Nymoen (1993): Bolgsalg og blbesand: 1970 1990. En emprsk modell for personblmarkede (House sales and car socks: 1970 1990. An emprcal model for he prvae car marke). Norsk Økonomsk Tdsskrf 107, pp. 23 40 Hendry, Davd F. and Jurgen A. Doornk (2001): Emprcal economerc modellng usng PcGve 10. London: Tmberlake Consulans Koopman, Sem Jan, Andrew C. Harvey, Jurgen A. Doornk and Nel Shepard (2000): STAMP. Srucural Tmes Seres Analyser, Modeller and Predcor. London: Tmberlake Consulans Sglz, Joseph, E. (1992): Capal markes and economc flucuaons n capals economes. European Economc Revew 36, pp. 269 306
Appendx: A mehod for decomposng deb growh We consder he followng smplfcaon of he model n he arcle: (1) y = x + z + y +, <1. 1 Here y s he logarhm of he deb level; x and z are explanaory varables;, and are parameers and s an error erm. The subscrps ndcae he perod. By backdang he varables and he error erm n (1) by one perod we ge: (2) y x + z + y. 1 = 1 1 2 + 1 Equaon (2) nsered n equaon (1) gves: 111 2 = 1 1 2 1 (3) y x + x + z + z + y + +. We ge he followng expresson by connung he nseron backwards: (4) y = x + z +, where: = 0 x = 0 = 0 = oal conrbuon from x o y (conrbuons from x, x -1,, x - ) = 0 z = oal conrbuon from z o y (conrbuons from z, z -1,, z - ) = 0 = oal conrbuon from oher (omed) explanaory facors o y Snce <1, he conrbuons from x -, z - and - o y wll decrease gradually when. We use esmaes for (,, ) and values of (x, x -1,, x -80 ) and (z, z -1,, z -80 ) o calculae he conrbuons from x and z o y. Then we decompose deb growh over four quarers by ransformng he esmaed conrbuons. = 0