Housing Prices, Bank Lending, and Monetary Policy

Transcription

1 Housing Prices, Bank Lending, and Moneary Policy I.J.M. de Greef and R.T.A. de Haas * Sepember 2000 Paper o be presened a he Financial Srucure, Bank Behaviour and Moneary Policy in he EMU Conference, Ocober 5-6, 2000, Groningen. Absrac In order o gain more insigh ino he relaionship beween housing prices and morgage lending, we esimae models for boh he Duch housing and he morgage marke. The empirical analysis presened in his paper offers suppor for he hypohesis ha in he Neherlands housing prices and morgage lending are inerdependen. According o our model, housing prices were influenced by changes in bank lending crieria during he esimaion period, even when we conrol for variables such as disposable household income, morgage ineres rae, demographic developmens and he housing sock. Morgage lending was found o be dependen on housing prices as well as disposable income. Our analysis furher suggess ha in he shor run housing prices can deviae subsanially from heir long-run equilibrium value. JEL-codes: D45, E32, E51, G12, G21, R21 * Secion Banking and Supervisory Sraegies, Direcorae Supervision, De Nederlandsche Bank. The auhors wish o hank Jaap Bikker, Iman van Lelyveld and Klaas Kno for helpful commens on earlier drafs and Miriam Holman- Rijken for excellen research assisance.

2 Inroducion A dependency of households on bank credi poins in he direcion of a credi channel in which bank lending influences real economic aciviy hrough changing housing prices. Exernal shocks, like rapid deregulaion or a change in moneary policy, will influence banks lending condiions and herefore he severiy of households credi consrains. An expansionary moneary policy, and he accompanying lower ineres raes, can for insance lead o higher collaeral values. This will enable households o arac more credi and even o cash par of he surplus value of heir house, since higher collaeral values make he problem of asymmeric informaion less criical o banks. Expecaions abou coninuing housing price rises may induce banks o relax heir lending sandards even furher, hereby leading o a self-perpeuaing process of increasing propery values and higher credi. Such periods of rising propery prices and higher credi graning are no only of consequence for moneary policy, bu also have imporan implicaions for supervisory policies, as boh households and banks become more vulnerable o a fall in housing prices. Lower propery values will again bring o he fore he problem of asymmeric informaion, as banks will no longer rely mainly on (excess) collaeral, bu are forced o focus again on he expeced repaymen capaciy of heir cliens. A squeeze of he credi supply may be he resul. In his paper, we shall examine he currenly soaring Duch housing prices and he accompanying rapid credi expansion. To pu his case ino perspecive we will briefly revisi wo earlier and quie similar European cases of rapidly rising housing prices and he associaed role of bank loans: he Nordic banking crisis, and he boom and bus on he UK housing marke. In paragraph 2 we firs pay aenion o some imporan heoreical noions. Then, in paragraph 3, we briefly discuss boh he Nordic case and he UK case in a descripive manner. Paragraph 4 presens some general informaion wih regard o he modelling of housing prices, followed by an empirical analysis on he relaionship beween housing prices and credi-availabiliy in he Neherlands in paragraph 5. Finally, paragraph 6 concludes.

3 Housing prices, bank credi and consumpion When a household buys a house, i will ypically be dependen on a financial insiuion like a bank from which i can borrow a subsanial porion of he needed funds. Mos households will no be able o raise money from oher sources: banks are special o hem. However, when graning loans, banks will be faced wih asymmeric informaion: hey only have limied insigh ino he financial prospecs and moral reciude of poenial borrowers. To reduce hese informaional shorcomings, banks will screen households before graning credi, hoping ha in doing so hey will be able o avoid he wors risks. Addiionally, hey will demand households o pledge collaeral, usually in he form of he house ha is o be financed (morgage loan). The pledging of collaeral significanly reduces moral hazard behaviour, since households know ha if hey are unable o mee heir redempion and ineres paymens, he bank will be eniled o sell he house and use he proceeds o reclaim is funds. The so-called balance shee channel, which is par of he credi view of moneary ransmission, describes how changing ineres raes can influence he credi supply of banks by affecing he ne worh of boh firms and households. 1 A higher ineres rae, for insance, can decrease he ne worh of households, since households real deb burden will increase and housing prices (collaeral value) will come down. 2 Furhermore, he higher ineres rae will also increase ineres paymens. Togeher, hese effecs may aggravae he consequences of asymmeric informaion, ulimaely leading o a reducion in he supply of credi o households. 3 Exernal shocks, like a change in moneary policy, can hus sar a process in which changing housing prices influence he credi supply by banks. In urn, hese changes in credi availabiliy may influence housing prices. Addiionally, an increase in inerbank compeiion (for example due o deregulaion) can lead banks o focus on gaining marke share raher han profi maximisaion. During such adjusmen processes, overshooing of credi markes is a well-known phenomenon in 1 Bernanke & Gerler, In he remainder of his paper we will focus on households. 2 A household s ne worh is o a large exen made up of he housing value minus is ousanding morgage deb (also ermed surplus value). 3 Besides he balance shee channel, he credi view also highlighs he so-called bank lending channel, which sresses he imporance of he adapaion of banks balance shees in reacion o a shif in moneary policy. Boh he bank

4 - - 3 which he pricing of risk is inadequae. Expeced increases in collaeral values will also lead o a decline in he perceived risk (Herring and Wacher, 1999). Banks may even sar o display disaser myopia behaviour: he probabiliy of low-frequency shocks like a real esae crash is underesimaed, especially when he ime period since he las real esae collapse ges longer (Guenag and Herring, 1984). Periods of rapid credi expansion may no only influence housing prices bu also macroeconomic variables like consumpion and inflaion. Firsly, higher housing prices can lead o wealh effecs: house-owning households will sar o feel wealhier because of he increased housing prices and will herefore spend a higher proporion of heir income on consumpion (decreasing propensiy o save). Secondly, some households will sar o cash par of heir surplus value by aking ou new morgage credi. This so-called wihdrawn equiy can hen be consumed. However, such periods of credi and housing booms can end abruply when a cerain rigger, e.g. a ighening of moneary policy, reverses his vicious circle. Lower collaeral values and increasing uncerainy may hen promp banks o shrink heir credi supply, hereby (furher) depressing housing prices: he disaser myopia urns ino disaser magnificaion (Herring and Wacher, 1999). Wealh effecs and equiy wihdrawal will sar o work in he opposie direcion, resuling in lower consumpion. Ulimaely, economic aciviy will be slowed down by means of he same channels ha were preponderan during he preceding boom period. 3. The macroeconomic imporance of housing and morgages: wo European cases During he 1980s and 1990s a number of European counries were hi by financial insabiliy relaing (a leas in par) o housing and morgage markes, namely he Unied Kingdom and hree Scandinavian counries. While each boom-bus cycle had is own peculiar feaures, hey noneheless shared some remarkable similariies. All four counries experienced a period of rapid financial deregulaion in he 1980s, prior o he boom.4 Before deregulaion, credi raioning had led o subsanial excess demand for lending channel and he balance shee channel sress he way in which moneary policy influences he supply of bank credi, while he radiional money view focuses on he demand for bank credi. 4 In 1980 he Bank of England disconinued he Supplemenary Special Deposis Scheme which had required banks o place a cerain amoun of ineres-free deposis wih he cenral bank. In 1981 he Bank also abolished he Reserve

5 - - 4 credi (Miles, 1992; Callen and Lomax, 1990; Drees and Pazarbasioglu, 1998). In he UK his was he resul of an informal carel of building socieies ha ruled he morgage marke. In Scandinavia, banks were effecively shelered from compeiion from oher domesic and foreign financial insiuions, which allowed hem o be highly selecive in choosing credi risks. Afer deregulaion of he financial markes, boh he amoun of morgage credi and he (real) morgage ineres rae rose in hese counries, reflecing he excess demand ha had been presen for many years. Adding o his effec, which basically refleced he achievemen of a new equilibrium, was an addiional increase in credi demand, pushing up real morgages raes even furher. This credi demand was induced by very opimisic expecaions of households. In he UK an increasing ineres for new producs like small and second morgages also played a role, while in Scandinavia demand for credi was enhanced by a generous ax deducibiliy of ineres expenses. The resuling rapid build-up of household deb was accompanied by a specacular increase in housing prices. Apparenly, he growing value of collaeral subsequenly made he problems associaed wih informaional asymmeries (seemingly) less serious o banks, which promped hem o exend even more credi. Furhermore, fierce compeiion on he morgage marke led o an erosion of he qualiy sandards used: banks sared o use higher loan-o-value raio s. Meanwhile, he combinaion of rising housing prices and less sric credi consrains resuled in a boos of consumpion, as consumers (parly) cashed in he surplus value of heir house. By he lae 1980s or early 1990s he whole process reversed. Moneary ighening increased ineres paymens of households, compounded in Scandinavia by he aboliion of he ax deducibiliy of ineres expenses. Households ne worh decreased or even urned negaive because of he sharp reducion in housing prices. Furhermore, many Asse raio, a liquidiy requiremen. In addiion, in 1983 he Building Socieies Associaion decided o no longer recommend raes o be charged on morgages and sared o allow building socieies o borrow increasingly from he money markes, whereas hey were previously resriced o lend on heir cusomers deposis only (Sargen, 1991, and Muellbauer and Murphy, 1997). In Norway, supplemenary reserve requiremens were abolished in 1984, while ineres rae declaraions were removed in Imporan deregulaion measures in Sweden included he abandoning of he sysem of liquidiy raios for banks in 1983 and lifing he ceilings on bank lending and resricions on bank lending

6 - - 5 households were confroned wih lower han expeced incomes. Households began o consolidae heir financial posiions by cuing back on consumpion. As a resul, he economies plunged ino a deep recession. In addiion o his iniial demand shock he Scandinavian counries, paricularly Finland, suffered from a collapse of he rade wih he former counries of he USSR (Vihriälä, 1997). Personal bankrupcies rose, puing bank s balance shees under pressure. In Scandinavia, he financial problems in he privae secor (including he corporae secor) reached such a level ha a full banking crisis ensued. From hese wo cases i is eviden ha housing and morgage markes play a crucial role in financial sabiliy, in he sense ha problems may spill over ino he real economy. In he remainder of his paper we will concenrae on he Neherlands, where he presen siuaion wih regard o housing prices, morgage lending and consumpion appears o resemble he upward cycle in he UK and Scandinavia in he 1980s o a cerain degree. Our main focus will be he relaion beween housing prices and morgage lending. However, before seing up a model, we will firs discuss he differen ways in which housing prices in general, and he role played by morgage credi in paricular, have been modelled in previous research. 4. Modelling of housing prices and morgage credi There exiss a large body of boh heoreical and empirical lieraure describing housing prices. Tradiional models of housing prices consis of a sock-flow model in which markes clear quickly and prices adjus o equae he demand for housing wih he exising sock. Prices are direcly derived from a demand and supply funcion. 5 However, hese models, like he sock-adjusmen framework developed by Muh (1960), lack a saisfacory micro foundaion. An imporan srand of lieraure herefore sared o view houses as an invesmen asse ha provides he owner wih a sream of housing services.6 raes in In Finland, amongs oher measures, resricions on average lending raes were abolished in 1986 (Drees and Pazarbasioglu, 1998). 5 Wih housing supply assumed fixed (infiniely inelasic) in he shor run, a housing price equaion was ofen consruced by invering he demand funcion (see for insance Hendry, 1984). 6 See for his so-called asse marke approach Breedon and Joyce (1993), Holly and Jones (1997), Brown e al. (1997), Baro and Takala (1998) and he seminal paper by Poerba (1984).

7 - - 6 The fundamenal value of a house can hen be seen as he presen discouned value of he real (expeced) housing services he house will provide over ime. In his lieraure, models generally describe a represenaive household ha solves an ineremporal opimisaion problem involving wo goods: housing services and a composie consumpion good, which are boh par of he uiliy funcion. Apar from housing here is one oher non-housing asse. The household maximises uiliy over ime, aking ino accoun is budge consrain and some echnical consrains, describing he evoluion of he sock of housing and non-housing asses. By doing so, i ensures ha he marginal rae of subsiuion beween housing services and he flow of uiliy from consumpion (u h /u c ) will equae boh he real user cos of housing and he real renal price of housing in capial marke equilibrium (R ): u h u c e h + (1) = ( 1 ) i π δ rp = R h The variables included in he real user coss of housing erm are he ineres rae (i), he marginal rae of income ax (), he expeced real capial gains on housing (π e h ), he rae of depreciaion (δ) and he real housing price (rp h ).7 I is he real housing price which mus now bring abou capial marke equilibrium. The real renal price (R ), which equals he amoun of money ha has o be given o a household o compensae for he loss of one housing uni, is unobservable: i is proxied by he demand for and supply of housing services. Therefore, he real renal price will be a funcion of real income (ry), he housing sock (H) and demographic variables (DEMO)8: (2) R = f ( ry, H, DEMO ) By subsiuing ou for R in (1) we can now express he real housing price in erms of real income, demographic variables, housing sock and he user cos of housing 7 See also Baro and Takala (1998), p. 12. The real user cos of housing includes an expeced price erm and is herefore endogenous. In empirical esing he expeced capial gains on housing are ofen proxied by lags in housing price appreciaion (Muellbauer and Murphy, 1997). 8 Breedon and Joyce (1993) for insance, in he radiion of he Bank of England s model of he early 1990s, represen demography by he proporion of he populaion aged 25-29, a prime house buying group.

8 - - 7 (logarihmic noaion). Noe ha when we assume he exising housing sock fixed in he shor-erm, he resuling model of housing prices is simply an invered demand funcion and is close o he radiional reduced form specificaion ha can be derived from equaing paricular housing demand and supply schedules and solving for housing prices (Pain and Wesaway, 1997). Usually, housing price models include he raio beween some measure of demography and he housing sock, which we represen for he momen as ln(demo/h). Thus, he housing price is relaed o (excess) demand raher han demand for or supply of housing per se. DEMO H (3) ln rp = f (ln ry, ln, ln ( 1 ) i h e { π + δ }) h Noe ha in he discussion unil now, no heoreical reason has been given for including a morgage credi variable in equaion (3). This means ha hus far i has implicily been assumed ha here are no credi marke consrains, meaning ha only he marke price, i.e. he morgage ineres rae, is relevan o households. 9 The morgage ineres rae is already included in (3) as an elemen of he user coss of housing. However, i is very likely ha no only he price bu also he volume of he morgage marke will be of ineres o house buyers. Afer all, as was poined ou in he paragraph 2, households can generally no borrow as much as hey desire a he prevailing morgage rae. Therefore, one can es for he imporance of credi raioning 10 as a resul of informaion asymmeries by including a morgage variable. Hendry (1984) and Hakfoor and Maysiak (1997) use, for insance, he ousanding morgage sock as a proxy for morgage raioning, while Muellbauer and Murphy (1997) proxy changes in morgage lending by he rae of acceleraion of he log morgage sock. Oher credi raioning variables ha are ofen used include loan-o-value raio s for firs ime buyers (Dicks, 1990, and Pain and Wesaway, 1997) and he difference beween an appropriae inerbank rae and he average morgage rae (Pain and Wesaway, 1997). For he ime being we will label he differen morgage variables ln(mor) : 9 Demand influences are assumed o be compleely refleced in he morgage ineres rae. 10 Sigliz and Weiss (1981). Meen (1990) shows ha he exisence of credi marke consrains implies ha he user coss of housing increase wih he raio of he shadow price of he raioning consrain o he marginal uiliy of he

9 - - 8 DEMO H (4) ln rp = f (ln ry, ln,ln ( 1 ) h e { i π + δ },ln MOR) h Boh Muellbauer and Murphy (1997) and Hendry (1984) recognise ha households do no possess all relevan knowledge of he morgage and housing marke and all he variables ha influence i. The raional expecaions assumpion migh herefore be oo srong. On he one hand i can be expeced ha households will have a leas some informaion abou hese markes, housing being one of he bigges expendiures in heir life. On he oher hand, however, he very fac ha mos households buy a house only once or wice in heir lifeime adds o he doubfulness wih regard o he qualiy of heir decision making process. Therefore, sensible (Hendry, 1984) or semi-raional (Muellbauer and Murphy, 1997) expecaions are ofen assumed, which means ha hese expecaions are neiher persisenly wrong nor fully efficien. Available sources of informaion can for insance be ineres raes, he volume of morgage lending and also lagged values of housing prices. DiPasquale and Wheaon (1994) noe ha he housing marke has a somewha predicable cycle wih posiive serial correlaion: housing prices do no follow a random walk bu exhibi significan serial correlaion. This implies ha no only he supply side of he housing marke adjuss gradually, due o he housing sock being fixed in he shor erm, bu also he demand side. 11 Such a gradual price adjusmen process holds when households develop expecaions by looking backward a hisoric prices. Case and Shiller (1988) show for insance ha exrapolaing behaviour (backward looking expecaions) is common in housing markes. During booms home buyers expec furher housing price rises and are worried abou no being able o buy a house in he fuure marke composie good. However, his quaniy is unmeasurable, and herefore excess demand is ofen proxied by measures such as he sock of morgages ousanding. 11 DiPasquale and Wheaon assume he following price adjusmen mechanism, in which P* is he hypoheical equilibrium price and τ is he (quarerly) percenage rae a which acual prices converge o his equilibrium price:?p = {P * P} 12 Levin and Wrigh (1997) proxy expecaions of fuure changes in real housing prices by pas changes in real housing prices and conclude ha ha housing price movemens are sysemaically relaed o hisorical housing price movemens which hey inerpre as being caused by speculaion. Exrapolaing behaviour seems o be confirmed by he resuls of a recen survey held by De Nederlandsche Bank (2000), which shows ha 57% of all Duch households wih a morgage loan hink ha he fundamenal value of heir house is less han he curren marke value, when 60% of his populaion

10 An empirical invesigaion ino he Duch housing and morgage markes 5.1 Inroducion In recen years he Duch morgage marke has displayed very high raes of growh, wih annual increases of 15% or more. 14 On average, banks porfolios of morgage lending doubled in he period beween end-1993 and mid A he same ime, housing prices have shown a remarkable increase of 80% beween 1990 and The combinaion of rapidly increasing housing prices and he relaively high level of ousanding morgage deb as a percenage of GDP means ha he financial risks for Duch households have increased significanly. This is especially so because morgage loans have o some exen been used for consumpion and he purchase of securiies, hereby simulaing boh consumpion and he bull sock marke. Imporan causes for he simulaneous increase in housing prices and morgage lending have been he srong economic growh and low ineres raes. These have ensured ha households borrowing capaciy has increased significanly. Besides hese emporary facors, more srucural causes were also very imporan, noably demographic developmens, ax legislaion (deducibiliy of ineres expenses), cenral governmen policy on spaial planning, as well as changes in banks accepance policies. 15 All hese aspecs have influenced and reinforced each oher, hereby simulaing he rapid growh in morgage credi and housing prices. 5.2 Modelling of he housing and morgage markes In order o gain more insigh ino he relaion beween housing prices and morgage lending in he Neherlands, we explicily model boh housing prices and morgage a he same ime hinks ha housing prices will increase furher. Addiionally, 49% of he populaion hinks ha he morgage ineres rae will rise furher as well. 13 Noe ha slow marke clearing can also be consisen wih housing demand based on fully raional, forward looking forecass. Afer all, when he exogenous variables o he marke move sysemaically, or if he srucure of he marke is so ha only gradual adjusmens o he sock are possible, hen even raionally forecased prices can be highly correlaed over ime (DiPasquale and Wheaon, 1994, p.7). 14 De Nederlandsche Bank (2000 II ). 15 During lae 1980s and early 1990s, banks sared o include second and emporary incomes in deermining borrowing capaciy hereby increasing he permissible morgage deb service/income raio (he maximum proporion of gross income ha may be spen on housing coss). They also increased he amoun of credi graned per uni of collaeral.

11 lending. This allows us o assess he impac of morgage lending on he housing price, as well as he reverse relaionship Housing prices Based on our earlier discussion we expec ha in he long run housing prices are closely relaed o variables such as income and morgage ineres raes. However, in he shor run deviaions from his long-run equilibrium may occur due o backward-looking expecaions and an inelasic housing supply. An Error Correcion Model (ECM) framework would seem a well-suied ool o model his, since i combines informaion abou he shor-erm dynamics (formulaed in firs differences) and he adjusmen process owards he long-run relaionship (in erms of levels). Our model is deermined in wo seps, according o he Engle-Granger procedure. Firs, we esimae a long-erm equaion, he coinegraing regression, relaing he level of he housing price o a number of variables including income and ineres raes. Nex we deermine a shor-erm equaion in erms of firs differences, including an error correcion erm consising of he lagged residuals of he esimaed long-erm equaion. A possible long-erm equaion is: e DEMO (5) ln( rph ) = c + β ln( ry) + φ ln( ry ) + ϕ ( rieff ) + γ ln + η ln( MOR) + ε H In his equaion he dependen variable is he log of he real housing price (rp h ). The housing price is dependen on, firs of all, real per capia income (ry). In as far as agens are forward-looking expeced real per capia income will also play a role. We have chosen o proxy real expeced income (ry e ) by a consumer confidence indicaor referring o he expeced financial siuaion over he nex welve monhs (FINSIT) because his may also be inerpreed in a broader fashion o include wealh (daa on household wealh are hard o come by for he Neherlands). A hird independen variable is he real effecive morgage rae (ri eff ), which is correced for he marginal income ax rae. Our preferred demographic variable o measure agains he housing sock is he number of households (HOUSEH). We experimen wih wo morgage relaed variables (MOR): oal real morgage lending (i.e. he real change in morgage sock, rm) and he raio of double-income households o all households (DIH). This laer variable may serve as a direc measure of changes in credi raioning. Alhough banks formally sared o ake

12 second incomes (fully) ino accoun in he firs half of he 1990 s, research by DNB (2000 II ) indicaed ha in pracice mos banks sared o include second incomes in heir credi decisions some years earlier. Since he iming of hese changes in informal bank lending is no known exacly, we experimened wih various dummy variables o deermine indirecly in wha year his break in banks behaviour began o have a significan effec on housing prices. Tesing resuls suggesed his o be Therefore, we included a dummy (DUMMY90) in our model seing ln( DIH) zero before Shor-erm dynamics in he housing price are modelled by relaing he change in he log real housing price o he firs differences of he righ-hand variables of he long-erm equaion. The maximum number of lags is four, lagged variables wih insignifican coefficiens being lef ou of he final esimaed equaion: (6) ln( rp ) + 4 m= 1 ϕ h m = c+ 4 g= e ( rph ) + βi ln( ry) i+ φ j ln( ry ) 4 ( ) DEMO 4 rieff + γ k ln + η ln( MOR) n + κε 1 + ν m ρ g ln k= 1 g H i= 1 k n= 1 j= 1 j In addiion, he lagged change in he real housing price iself and an error-correcion erm consising of he lagged residuals of he esimaed long-erm equaion (ε -1 ) are included Morgage lending In order o invesigae he role of housing prices in morgage lending, a separae model for morgage lending is se up. The housing price may ener morgage lending boh as a demand and a supply facor. Housing prices, ogeher wih income and he morgage rae, deermine how much credi households wish o ake ou. A he same ime, banks decisions regarding morgage applicaions may also be influenced by housing prices, which form an indicaor of he collaeral value of a house. Credi may be raioned. In he Neherlands, he morgage deb service/income raio, which is he maximum proporion of curren gross income ha may be spen on housing coss, plays a dominan role in he supply decision, more so han he loan-o-value raio which is no officially limied. This suggess ha he double-income household dummy should also be included in his model, as a measure of changes in credi raioning by banks. Since changes in he deerminans

13 of morgage lending (income, morgage raes, housing price ec) may be expeced o have immediae effec raher han o se off a lenghy adjusmen process, a single-equaion model would seem more appropriae han an ECM. On he basis of he above consideraions a model of morgage lending migh look as follows: ln( rm) 4 (7) + λk ln( rieff ) k + τdummy90 ln( DIH) + ε k= 0 = α + 4 g= 1 ω g ln( rm) 4 m= 0 g + 4 i= 0 θ ln( rp ) i h i m + 4 j= 0 χ ln( ry) j j The dependen variable in his equaion is real morgage lending (rm). (Lagged) forms of he real housing price (rp h ), real income (ry), he real effecive morgage rae (ri eff ) as well as real morgage lending have been included. The equaion is framed in firs differences o cope wih he fac ha mos ime series used conain uni roos (see below). 5.2 The daa In general, quarerly daa for he period 1977:Q2-1998:Q1 have been used (1977:Q2-1999:Q4 in case of he morgage lending model). The appendix gives a more deailed descripion of he daa and heir sources. The ime series used were analysed o deermine he presence of uni roos. Mos variables (real housing price, real per capia income, expeced financial siuaion, real effecive morgage rae of ineres, and share of doubleincome households) were found o be I(1). Only real morgage lending was found o be I(2). However, from an economic raher han a saisical viewpoin one would expec real morgage lending o be inegraed o he same order as he real housing price. Therefore, real morgage lending is reaed as if i were I(1). Finally, he raio of households o he housing sock ln( HOUSEH/H) proved I(0). This suggess ha in he long-run his variable ough no play a role in he deerminaion of housing prices. However, he variaion in he log raio of households o he housing sock suggess ha his long run siuaion is no acually reached (Figure 1).

14 Figure 1 Raio of households o housing sock (logarihm) 0,06 0,05 0,04 0, Raio of oal households o housing sock Since we are in effec concerned wih a medium long period we decided o include his variable in he long erm equaion for he housing price, despie i being inegraed o he order of zero. Mulicollineariy problems are no expeced: no correlaion beween independen variables exceeds 0.7, which as a rule of humb may be aken as a cu-off level (Table 1). 16 Table 1 Correlaion marix?ln(rp h )?ln(rm)?ln(ry)?ln(finsit)?ri eff?ln(dih)?ln(househ/h)?ln(rp h ) ?ln(rm) ?ln(ry) ?ln(FINSIT) ?ri eff ?ln(DIH) ?ln(HOUSEH/H) Empirical resuls Our preferred specificaion is shown in Table 2. When esimaing he long-erm equaion for he housing price we found ha of he wo morgage variables he index of doubleincome households yields he bes resuls. When included ino he esimaion ogeher wih real income, he coefficien on oal real morgage lending proved o have he wrong sign (negaive). The high correlaion beween he log levels of real income and real

15 morgage lending as well as beween he firs differences in logs (0.93 and 0.44, respecively) is probably he cause of his. However, when morgage lending was proxied by he index of double-income households, his variable urned ou o have a significan and posiive influence on housing prices, even when he real income variable was included as well. To us, his suggess ha morgage lending influences housing prices mainly as a supply facor, wih demand for morgages adjusing o more fundamenal facors ha co-deermine he housing price, such as income and he morgage rae. Table 2 Esimaed Error Correcion Model of he Housing Price 1977:Q2-1998:Q1?ln(rp h ) = ?ln( rp h ) ?ln(rp h ) ?ln(ry) ? ri eff (7.1) (1.8) (2.7) (3.7) (2.0) [2.1 ln(ry) ln(finsit) ri eff ln(househ/h) (2.7) (4.1) (4.9) (11.3) (1.5) DUMMY90ln(DIH) ] (3.6) (3.9) N = 84 R adj 2 = 0.53 prob. LM (4 lags) = 0.06 prob. Whie = 0.08 ADF residuals (4 lags) = (1% criical value = -2.58) Explanaory noe: The model is shown wihou seasonal dummies (absolue -values in brackes). N = number of observaions, R adj 2 = adjused muliple correlaion coefficien, prob. LM = probabiliy of Breusch-Godfrey es on serial correlaion in residuals, prob. Whie = probabiliy of Whie heeroscedasiciy es, ADF = augmened Dickey- Fuller uni roo es. The lagged residuals of he long-erm coinegraing regression, wrien ou fully beween he large brackes in erms of he fundamenal variables, proved saionary, supporing he noion ha he real housing price is indeed coinegraed wih he oher variables included in he long-erm equaion such as real income and he real ineres rae. Mos long-erm 16 Tes resuls including lagged independen variables (no shown here) also sugges ha here is no significan mulicollineariy.

16 coefficiens are significan a he 1% level, alhough he significance level is likely o be overesimaed due o serial correlaion. Only he raio of households o he housing sock urns ou o be subsanially less significan. However, because of heoreical consideraions we decided o keep his variable in our esimaion. 17 All he coefficiens have he expeced sign and he adjused R 2 equals 0.53 for he oal ECM equaion (0.66 for he long-erm equaion aken separaely). The error correcion erm in he equaion is significanly negaive, indicaing ha in he long run he housing price moves in line wih is fundamenals. However, he adjusmen process akes a considerable ime: each quarer only abou 6% of he deviaion from he long-erm value is correced. 18 Since he model is esimaed in logs, he long-erm coefficiens (beween large brackes) can be inerpreed as long-run elasiciies or, in he case of he morgage rae, semi-elasiciies. The oucomes show han an increase of 1% in real disposable income will resul in a 2.1% increase in he real housing price. Similarly, a 1% increase in he index measuring financial expecaions, he double income households dummy or he raio of households o oal housing sock yields an increase in he real housing price of 0.7%, 0.2%, and 4.6%, respecively. Finally, a rise of 1%-poin of he real effecive morgage ineres rae induces a decrease of he real housing price of 11%. 19 Of he various long-erm variables only real income and he real effecive morgage rae appear o have a significan influence on housing prices in he shor run as well. Addiionally, he lagged real house price provides some more shor-erm dynamics, indicaing a sicky pricing process. Morgage lending appears predominanly dependen on previous developmens in morgage lending iself and on (lagged) housing prices and real income (Table 3). The adjused R 2 shows ha our equaion explains 88% of he variaion in he amoun of 17 Removing his variable would only have a very small impac on boh he magniude of he coefficiens of he remaining variables and heir significance. 18 This comes down o an adjusmen process of approximaely four years. 19 However, i should be kep in mind ha he real effecive morgage rae equals he nominal ineres rae correced for he fiscal deducibiliy of ineres expenses and subsequenly lowered by he inflaion. In mos years he correcion simply implied dividing he nominal ineres rae by 2, as he marginal ax rae for he average Duch household was 50%. This means ha a 1%-poin increase in he real effecive morgage rae roughly corresponds o a 2%-poin increase in he nominal morgage rae, implying ha a 1%-poin increase in he nominal ineres rae would lead o a long-run reducion in real housing prices by approximaely 5%.

17 morgage lending, which is saisfacory. The resuls show ha, when for each variable he differen significan lags are added up, a higher rae of growh in real housing prices or incomes leads o higher growh of morgage lending. Oddly, supposedly fundamenal facors such as he morgage rae or he double-income household dummy urned ou insignifican. 20 We ake his as an indicaion of he complexiy of modelling he morgage lending process (raher han he possibiliy ha facors such as he morgage rae are ruly insignifican). Table 3 OLS esimaion of morgage lending 1977:Q2-1999:Q4?ln(rm) = ?ln(rm) ?ln(rm) ?ln(rp h ) + 0.1?ln(rp h ) -2 (1.8) (4.6) (5.7) (3.3) (2.6) - 0.1?ln( rp h ) ?ln(ry) - 0.5?ln(ry) ?ln(ry) -4 (2.1) (3.9) (3.9) (3.0) N = 91 R adj 2 = 0.88 prob. LM (4 lags) = 0.74 prob. Whie = 0.41 Explanaory noe: The model is shown wihou seasonal dummies (absolue -values in brackes). N = number of observaions, R adj 2 = adjused muliple correlaion coefficien, prob. LM = probabiliy of Breusch-Godfrey es on serial correlaion in residuals, prob. Whie = probabiliy of Whie heeroscedasiciy es. The overall evidence appears o suppor he hypohesis ha housing prices and morgage lending in he Neherlands are muually dependen. Moreover, he evidence suggess ha in he shor run, housing prices may deviae subsanially from heir long-run growh pah. This indicaes ha he Duch economy is vulnerable o he ype of boom-bus cycle which i befell in he 1970s and which hi he UK and Scandinavia in he 1980s and 1990s. 20 One supply facor ha did work was banks solvabiliy (measured as equiy divided by he oal balance shee). However, we have misgivings abou including his variable on heoreical grounds. Reduced solvabiliy may lead o reduced morgage lending, bu more likely so in bad imes han in good imes (in good imes banks may well prefer issuing new equiy).

18 A dynamic forecas of he housing price In order o invesigae how well our model of he housing price performs ou-of-sample, we execued a dynamic forecas wih regard o he recen rapid increase in housing prices. Firs of all, we exrapolaed wo independen variables (HOUSEH and DIH) from 1998:Q2 in order o have a complee daase unil 1999:Q4. 21 Secondly, we esimaed boh he coinegraing equaion and full ECM on he basis of he sample period 1977:Q2-1994:Q4. In order o be able o forecas for he period 1995:1-1999:4 we needed o supplemen he residuals from he coinegraing equaion wih he difference beween he acual housing price and he housing price forecased on he basis of he coinegraing equaion. The resuls of our forecas on he basis of he full ECM are shown in figure 2. Figure 2 Acual real housing price and ou-of-sample forecased real housing price (housands of 1995 guilders) Realised real housing price Forecased real housing price Explanaory noe: he ou-of-sample forecas is based on he sample period 1977:Q2-1994:Q4. Wihin sample he model appears o give a reasonable forecas of he real housing price. A major excepion is he housing price bubble in he lae 1970s. According o our model, he down-urn should have occurred a number of years earlier han i did in realiy. As a consequence he forecased housing price significanly underesimaes acual developmens for abou four years. Wih regard o he ou-of-sample period we noe ha he forecased real housing price follows he same rend as he acual real housing price, albei on average on a slighly lower level (see also figure 3). The average difference in

19 nominal erms equals NLG 12,800 over his period. Our model herefore suggess ha curren housing prices are somewha overvalued. Figure 3 Difference beween acual and ou-of-sample forecased real housing price (housands of 1995 guilders) Conclusions Problems of asymmeric informaion are relevan o he morgage marke. In reacion o such problems banks may raion heir morgage credi supply. According o he credi view of moneary policy, in which credi is special o households, his can influence households financial capaciy when buying a house. Credi availabiliy may hus influence housing prices. Since pledging collaeral is one way of alleviaing he consequences of asymmeric informaion, housing prices may in urn also influence credi availabiliy. To assess he relevance of his view for he Duch economy, we esimaed wo separae models, one for he housing marke and one for he morgage marke. We found ha in he long run housing prices are influenced by variables such as income, he morgage ineres rae, demographic developmens and he housing sock. Addiionally, changes in bank lending crieria appear relevan. In he Neherlands, such changes may be proxied using he number of double income households, since banks sared o ake second 21 Boh he number of households (HOUSEH) and he double income households index (DIH) showed a very sable developmen during our sample period, so ha exrapolaing hese ime series appears accepable.

20 incomes ino accoun only around he beginning of he 1990s. Furhermore, morgage lending was found o be dependen on he housing price as well as disposable income. The empirical work presened in his paper hus offers some suppor for he supposiion ha in he Neherlands housing prices and morgage lending are indeed inerdependen. Our empirical work furher suggesed ha in he shor run housing prices can deviae subsanially from heir long-run equilibrium value. The adjusmen process o shocks o he housing marke may ake up o four years. In EMU, housing prices are no considered of direc relevance o moneary policy decisions (asse prices are no accouned for in he ESCB s inflaion arge). However, he relaionship beween housing prices and morgage lending may be of indirec relevance, since housing price developmens may influence consumpion and herefore inflaionary pressure hrough equiy wihdrawal and wealh effecs. Research suggess ha he inerdependence beween he housing markes in Europe is (sill) limied (Van Rooij and Vos, 1999). Thus, moneary policy decisions aken from he European perspecive may no be opimal on he naional level in case of diverging housing marke cycles. Combined wih he fac ha according o our findings housing prices may deviae subsanially from heir long-erm value, his implies ha naional auhoriies should carefully monior developmens on boh he housing and he morgage marke and heir effecs on consumpion and inflaion. In his ligh, more emphasis should be laid on guaraneeing an even developmen of he housing marke. Tax disorions and undue governmen suppor of house owners should, for example, be prevened. Also, he influence of morgage lending on he financial fragiliy of households should be moniored and banks should be simulaed o adequaely price he risks ha are aached o graning morgage credi. In so doing, idiosyncrasies on he housing markes can be kep o a minimum, hereby reducing he adverse effecs for hese naional markes of a single moneary policy in he EMU.

21 References Baro, B., K. Takala, House Prices and Inflaion: A Coinegraion Analysis for Finland and Sweden, Bank of Finland Discussion Paper, No. 12, Bernanke, B.S., M. Gerler, M., Inside he Black Box: he Credi Channel of Moneary Policy Transmission, NBER Working Paper No. 5146, Breedon, F.J., M.A.S. Joyce, House prices, arrears and possessions: A hree equaion model for he UK, Bank of England Working Paper, No. 14, Brown, J.P., H. Song and A. McGillivray, Forecasing UK house prices: a ime varying coefficien approach, Economic Modelling, 14, pp , Case, K.E., R.J. Shiller, The Behavior of Home Buyers in Boom and Pos-Boom Markes, New England Economic Review, pp , November/December Callen, T., J. Lomax, The Developmen of he Building Socieies Secor in he 1980 s, Bank of England Quarerly Bullein, No. 30, De Nederlandsche Bank, He Nederlandse publiek gepols over gebruik hypohecair kredie, Kwaraalberich Juni 2000, De Nederlandsche Bank (II), He bancaire hypoheekbedrijf onder de loep, Dicks, M.J., A simple model of he housing marke, Bank of England Discussion Paper No. 49, DiPasquale, D., W.C. Wheaon, Housing Marke Dynamics and he Fuure of Housing Prices, Journal of Urban Economics, 35, pp , Drees, B., C. Pazarbasioglu, The Nordic Banking Crises. Pifalls in Financial Liberalizaion?, Occasional Paper 161, IMF, Washingon DC, April Guenag, J., R. Herring, Credi Raioning and Financial Disorder, The Journal Of Finance, Vol. XXXIX, No. 5, pp , Hakfoor, J., G. Maysiak, Housing Invesmen in he Neherlands, Economic Modelling, 14, pp , Hendry, D.F., Economeric Modelling of House Prices in he Unied Kingdom, In: D.F. Hendry and F. Wallis (eds.), Economerics and Quaniaive Economics, Basil Blackwell, Oxford, pp , Herring, R.J., S. Wacher, Real Esae Booms and Banking Buss: An Inernaional Perspecive, Working Paper Wharon Financial Insiuions Cener, No. 27, The Wharon School, Universiy of Pennsylvania, 1999.

22 Holly, S., N. Jones, House prices since he 1940s: coinegraion, demography and asymmeries, Economic Modelling, 14, pp , Levin, E.J., R.E. Wrigh, The impac of speculaion on house prices in he Unied Kingdom, Economic Modelling, 14, pp , Meen, G.P., The measuremen of morgage marke consrains and he implicaions for economeric modelling of UK house prices, Oxford Bullein of Economical Saisics, 52, pp. 1-23, Miles, D., Housing and he Wider Economy in he Shor and Long Run, Naional Insiue Economic Review, pp , February Muellbauer, J., A. Murphy, Booms and Buss in he UK Housing Marke, The Economic Journal, 107, pp , Muh, R.F., The Demand for Non-Farm Housing, In: Harberger, A. (ed.), The demand for durable goods, Universiy of Chicago Press, Chicago, pp , Pain, N., P. Wesaway, Modelling srucural change in he UK housing marke: a comparison of alernaive house price models, Economic Modelling, 14, pp , Poerba, J.M., Tax subsidies and owner-occupied housing: an asse-marke approach, Quarerly Journal of Economics, 99, pp , Rooij, van, M.J.C., L. Vos, De samenhang ussen Europese huizenprijzen!, Maandschrif Economie, Vol. 63, No. 3, pp , June Sargen, J.R., Deregulaion, deb and downurn in he UK economy, Naional Insiue Economic Review, No. 137, pp , Augus Sigliz, J.E., A. Weiss, Credi raioning in markes wih imperfec informaion, American Economic Review, 71, pp , Vihriälä, V., Banks and he Finnish Credi Cycle , Bank of Finland Sudies E:7, 1997.

23 Appendix Daa and sources CPI DIH FINSIT H HOUSEH ri eff rp h rm ry Consumer price index (OECD). Number of double income households (Saisics Neherlands), divided by he oal number of households, HOUSEH. From 1990 onwards on an annual basis. Inerpolaed o obain quarerly daa. (Before 1990 only available for seleced years. However, in he empirical analysis a dummy is used o se DIH a zero before See commens in ex). Confidence index regarding consumers financial expecaions for he nex 12 monhs (Saisics Neherlands). Unil 1983 polls were held hree imes a year (January, May, and Ocober). Quarerly daa for his period obained by means of inerpolaion. Housing sock (Saisics Neherlands). Inerpolaed annual daa. Number of households (Saisics Neherlands). Inerpolaed annual daa. Real effecive 10-year morgage ineres rae. Consruced as he nominal 10-year morgage ineres rae (Saisics Neherlands) adjused for he average marginal ax rae,, and year-on-year changes in he consumer prices index, CPI. Real housing price. Consruced as he nominal housing price (Kadaa) adjused for inflaion by means of he CPI. Real morgage lending. Consruced as he change in he ousanding morgage sock (De Nederlandsche Bank) adjused for inflaion by means of he CPI. Real disposable income per capia. Consruced as nominal per capia income (Saisics Neherlands) adjused for inflaion by means of he CPI. Inerpolaed annual daa. Marginal income ax rae for married couples (wo children) earning wo imes he average income (Neherlands Bureau for Economic Policy Analysis). Quarerly daa se equal o annual rae.