WP/16/74. Macroprudential and Monetary Policy Interactions in a DSGE Model for Sweden. by Jiaqian Chen and Francesco Columba

Transcription

1 WP/16/74 Macroprudenial and Moneary Policy Ineracions in a DSGE Model for Sweden by Jiaqian Chen and Francesco Columba

2 216 Inernaional Moneary Fund WP/16/74 IMF Working Paper European Deparmen Macroprudenial and Moneary Policy Ineracions in a DSGE Model for Sweden Prepared by Jiaqian Chen and Francesco Columba 1 Auhorized for disribuion by Craig Beaumon March 216 IMF Working Papers describe research in progress by he auhor(s) and are published o elici commens and o encourage debae. The views expressed in IMF Working Papers are hose of he auhor(s) and do no necessarily represen he views of he IMF, is Execuive Board, or IMF managemen. Absrac We analyse he effecs of macroprudenial and moneary policies and heir ineracions using an esimaed dynamic sochasic general equilibrium (DSGE) model ailored o Sweden. Households face a ceiling on heir loan-o-value raio and mus amorize heir morgages. The governmen grans morgage ineres paymen deducions. Lending raes are affeced by morgage risk weighs. We find ha demand-side macroprudenial measures are more effecive in curbing household deb raios han moneary policy, and hey are less cosly in erms of foregone consumpion. A igher macroprudenial sance is also found o be welfare improving, by promoing lower consumpion volailiy in response o shocks, especially when using a combinaion of macroprudenial insrumens. JEL Classificaion Numbers: E44, E52, E58, G28 Keywords: Macroprudenial Policies; Moneary Policy; Collaeral Consrains Auhor s Address: jchen@imf.org, francesco.columba@bancadialia.i 1 Francesco Columba is a Banca d Ialia. The views expressed in his paper are hose of he auhors and do no necessarily represen he views of he IMF, is Execuive Board, or IMF managemen or of he Banca d Ialia. The auhors are graeful o Craig Beaumon, Helge Berger, Fabio Canova, Daria Finocchiaro, Srid Ingvar, Giovanni Melina, Sefano Neri, Pau Rabanal, Ulf Sodersrom and seminar paricipans a he Sveriges Riksbank and IMF for valuable commens. We benefied from discussions wih Kasper Roszbach and Lars Svensson. Thank you also o Srid Ingvar for sharing he daa. All remaining errors are our own.

3 Conens Page 1 Inroducion 5 2 The Swedish Housing Marke 6 3 The Model 12 4 Calibraion and Esimaion 14 5 Model Properies 17 6 Exploring he impac of macroprudenial policies 19 7 Welfare opimal macroprudenial policies 26 8 Discussion and Robusness Checks 33 9 Conclusion 36 Appendix 38 A Addiional Tables and Figures 38 B The Model 41 C Daa and Sources 55 Lis of Tables 1 Calibraed parameers Seady sae raios A1 Prior and poserior disribuion of he srucural parameers I A2 Prior and poserior disribuion of he srucural parameers II.. 39 Lis of Figures 1 Swedish household indebedness and housing price expecaions Household deb as share of disposable income Tax incenives for home ownership, Housing sock o populaion Ineres rae fixaion periods for Swedish morgages Spread beween morgage ineres raes and bond yields Swedish household indebedness and housing price expecaions Moneary policy shock

4 9 Impacs from a permanen reducion in LTV: Impacs from a permanen change in amorizaion requiremen: Impacs from a reducion in morgage ax deducibiliy: Households consumpion responses from a reducion in morgage ax deducibiliy: Impacs from a increase in morgage risk weighs: Welfare: loan-o-value raio Welfare: amorizaion Welfare: ax deducibiliy on morgage ineres repaymens Welfare: morgage risk weighs Household deb: morgage risk weighs Welfare: ineracion beween amorizaion requiremens and LTV Welfare: amorizaion requiremen Welfare: ineracion beween morgage ax deducibiliy and LTV I Welfare: ineracion beween morgage ax deducbiliy and LTV II Aggregae of saver s and borrower s uiliies: ineracion beween amorizaion requiremens and LTV Household deb over amorizaion requiremens and LTV A1 Prior and poserior disribuion of esimaed parameers

5 1 Inroducion In he afermah of he Global Financial Crisis a consensus is emerging around a paradigm ha asks financial sabiliy o macroprudenial policies, wih a role for moneary policy reserved for exraordinary imes (IMF., 215). Moneary policy remains assigned o macroeconomic sabiliy, ofen via an inflaion argeing framework. Curren hinking is ha macroprudenial ools, if deployed in a imely manner, can effecively conain mos vulnerabiliies, alhough heir effecs, especially on welfare, need furher sudy (Claessens, 214). Ye, as macroprudenial policies are sill unproven, risks o financial sabiliy canno be compleely excluded from he consideraions behind moneary policy decisions (Bernanke, 215). Indeed, i is possible ha, hreas o financial sabiliy arise ha canno be adequaely addressed by macroprudenial insrumens, bu given heir macroeconomic impac, any leaning agains he wind by moneary policy agains such hreas can only be jusified afer a horough cos-benefi analysis (Svensson, 216). Coordinaion beween macroprudenial insrumens and moneary policy, which may suppor heir effeciveness in periods of sress, also needs furher analysis (Angelini e al., 214). High and rising house prices and household morgage deb in Sweden have promped financial sabiliy concerns. Moreover, Swedish banks are large, a some 4 percen of GDP, and have a large exposure o morgages increasing he poenial for mouning vulnerabiliies in he housing marke and household balance shees o undermine he resilience of he financial sysem. The banking sysem is dominaed by four large conglomeraes ha are highly inerconneced hrough cross holding of covered bonds. I is herefore imporan o undersand he poenial for macroprudenial polices o curb such risks. Noneheless, he poenial sabiliy benefis of macroprudenial policies from moderaing household deb have o be assessed aking ino accoun he welfare coss from heir impac on he economy. Moneary policy effeciveness in haling such dynamics in housing markes are also examined, even hough here seems o be no scope a his ime for moneary policy o lean agains he wind as low inflaion expecaions have promped he Riksbank o focus on fighing low inflaion o proec he credibiliy of he inflaion argeing framework. This paper conribues o he lieraure in wo ways. Firs, i compares he coss and benefis of macro-prudenial and moneary policies in reducing household deb in a fullyfledged general equilibrium model, aiming o shed some ligh on he difficul rade-offs ha policymakers face. In paricular, we compare he effec of a number of demand-side macro prudenial measures, being loan-o-value (LTV) raios, amorizaion requiremens and ax 5

6 deducibiliy of morgage ineres paymens, wih a supply-side measure, he morgage risk weigh. This is moivaed by he circumsance ha in he case of Sweden here may be more room for ighening he macro-prudenial sance via demand-side insrumens, as supply-side measures have already been ighened in recen years (Chen, 214). We use a modified form of he dynamic sochasic general equilibrium (DSGE) model of Gerali e al. (21) for his purpose, making i suiable for a small open economy, adding addiional macroprudenial measures, and esimaing i wih Swedish daa. Second, we sudy he welfare implicaions of hese policies o shed some ligh on wheher macroprudenial policies are welfare improving, wheher macroprudenial measures are subjec o decreasing effeciveness, and wheher here are complemenariies among he measures. The remainder of he paper is srucured as follows. Secion II highlighs key developmens in housing and morgage markes in Sweden. Secion III describes he model. Secion IV discusses is calibraion and presens he esimaion resuls and secion V illusraes he properies of he esimaed model. Secion VI discusses he impac of macroprudenial policies and secion VII sudies he welfare implicaions. Secion VIII concludes wih policy implicaions. 2 The Swedish Housing Marke Swedish households deb is high and rising. Deb as a share of disposable income reached 176 percen in June 215 and 195 percen if he deb of enan-owned housing associaions is included. The coninued increase in house prices reflecs he lack of housing supply alongside srong housing demand fosered by hisorically low ineres raes, rising incomes and wealh, and populaion growh especially in he main ciies (Turk, 215). The record high share of households expecing furher house prices increases could suppor furher borrowing. Oher facors have conribued o high and rising household deb. The Swedish ax sysem has favored home ownership, wih very low effecive propery axes since 28, and i incenivizes households o no pay down heir morgage since hey can deduc 3 percen of heir ineres paymens (21 percen above SEK 1, ) from axes due. 6

7 Figure 1: Swedish household indebedness and housing price expecaions Deb (percen of disposable income) House price expecaions, righ 1/ Sources: SEB, Sveriges Riksbank and auhors calculaion. 1/Ne share of households expecs house price o rise. Figure 2: Household deb as share of disposable income 3 25 Deb of enan-owner assosiaions ITA FRA DEU BEL FIN ESP PRT GBR IRL SWE NOR NLD DNK Sources: Saisics Sweden, Naional Saisics Offices and auhors calculaion. Noe:214Q4 or laes available. Sweden s per capia housing sock remains almos unchanged since he early 9s reflecing srucural impedimens in he consrucion secor. For insance, complex and ime consuming land acquisiion and planning sysems have been pulling down supply despie 7

8 Figure 3: Tax incenives for home ownership, FRA ESP GBR IRL ITA DEU DNK FIN NLD SWE Sources: European Commission and auhors calculaion. Noe:Composie ax index range: (Low) 3 (High) rising profiabiliy in he consrucion secor (Emanuelsson, 215). The housing supply issue is mos eviden in he major ciies, where dwellings per capia have been declining over ime, which has been associaed wih a rise in prices relaive o he naional average (Ho, 215). Such supply consrains increase he risk ha house price gains coninue o exceed income growh. Lower morgage raes combined wih ax incenives have made he associaed increase in household borrowing more affordable. Morgage conracs in Sweden ofen run for 3-5 years, bu i is no common pracice o have a fixed amorizaion schedule. The rae of amorizaion varies noably across households, wih Riksbank analysis 1 suggesing ha only abou 6 percen of indebed households reduced heir debs in 213, wih he pace of reducion implying an average remaining amorizaion period of 99 years. 2 More recen borrowers end o amorize more han in he pas, wih 69 percen of all households wih new loans amorizing heir morgage in 214 up from 42 percen in 211. Nowihsanding his recen increase in amorizaion for new morgages, he share of he morgage sock which is being amorized remains largely unchanged from previous years a 62 percen. Amorizaion is also more common for more leveraged loans, 1 See Depuy Governor Skingsley s speech, (Skingsley, 27) 2 In a random sample aken in he auumn of 212 in conjuncion wih FSA s morgage survey, he average repaymen period among households wih a LTV below 75 percen, and which acually amorized, was 14 years. Noe ha he repaymen period in his case refers o he mauriy implied by he amorizaion paymens made and no he mauriy specified in he loan agreemen.(sveriges Riskbank, 214) 8

9 Figure 4: Housing sock o populaion Sources: Saisics Sweden and auhors calculaion. Noe:Raio of dwellings o populaion in housand as 85 percen of households wih LTV above 7 percen amorized heir morgage, bu only 4 percen for loans wih LTV beween 5 and 7 percen. Relaively low amorizaion in Sweden in par reflecs coss of amorizing morgage deb faser han ha sipulaed in he conrac. In paricular, households need o pay compensaion for he ineres rae differenial over he remaining ineres rae fixaion period (Leonhard e al., 212). This is in sharp conras wih oher counries such as he US, Denmark and Germany, where he penaly for early repaymen is eiher very low or does no exis. The ineres differenial compensaion is calculaed as follows. 3 Ineres rae differenial compensaion = (morgage ineres rae (ask rae for a governmen bond wih he same fixed period +1percenage poin)) ousanding deb remaining period The differences beween morgage ineres raes and governmen (or covered since 214) bond yields averaged o abou 2 percen beween 21 and 215. This circumsance ogeher wih he high ousanding deb, implies ha he penaly coss of early repaymen of morgages could be quie high on average, providing he Swedish households wih lile incenive 3 The formula was modified in 214 replacing governmen bond wih covered bond. 9

10 o repay more han he low amorizaion requiremen esablished when he morgage was issued. 4 Figure 5: Ineres rae fixaion periods for Swedish morgages 1% 9% 8% 7% 6% 5% 4% 3% 2% 1% Greaer han 5 years Less han 5 years Less han 3 monhs % Sources: European Commission and auhors calculaion. Boh governmen and covered bond yields have fallen since 211, ye banks have increased he ineres rae margin on morgage loans so ha he differences beween morgage ineres raes and governmen (or covered since 214) bond yields averaged o abou 2 percen beween 21 and 215. (Figure 6) This circumsance ogeher wih he high ousanding deb, implies ha he penaly coss of early repaymen of morgages could be quie high on average, providing he Swedish households wih lile incenive o repay more han wha he low amorizaion requiremen decided when he morgage was issued requires. The composiion of deb has shifed owards variable rae conracs, as abou 75 percen of he new morgages have an iniial ineres rae fixaion period of less han 3 monhs while in 212 abou 5 percen were a variable rae. This preference for variable rae conracs is consisen wih households expecaion for ineres raes o remain low, as he Riksbank has urned o a very accommodaive moneary policy sance and signaled is commimen o 4 The penaly cos of an early repaymen could be around 1.8 imes of disposable income for households wih variable rae morgages, and much larger for households wih fixed rae morgages. This simple calculaion assumes households wih a deb-o-income raio of 366 percen, as observed for he new morgage borrowers in 214, a spread beween morgage rae and bond yields of 2 percen and a residual mauriy of 3 monhs on he conrac. 1

11 Figure 6: Spread beween morgage ineres raes and bond yields 3 Governmen bonds Covered bonds Sources: Severige Riksbank, Saisics Sweden and auhors calculaion. Noe:Weighed average of he differences beween morgage raes over bond yields wih corresponding remaining mauriies. We assume ha covered bonds wih less han 1 year mauriy have he same yields as he Swedish governmen bonds. raise inflaion. Overall, some 69 percen of he exising sock of morgages has a variable rae. Household deb and housing prices have coninued o rise from already high levels, as 29 percen of new morgage borrowers in 214 had a Deb-o-Income (DTI) raio of over 45 percen, nowihsanding a range of measures aken in recen years o enhance he financial resilience of banks and households. Macroprudenial policy measures have focused on he credi supply side by srenghening bank capial buffers, as he Financial Supervisory Auhoriy (FSA) has rolled ou Basel III measures ahead of schedule, including he inroducion in 213 of a capial conservaion buffer a 2.5 percen, in 214 of minimum risk weighs of 25 percen for morgages, and in 215 of a sysemic risk buffer a 3 percen, a 2 percen capial surcharge for he four sysemically imporan banks, and a counercyclical risk buffer a 1 percen. 5 Macroprudenial measures on he credi demand side have been aken o a more limied exen. In 21 he FSA esablished an 85 percen cap for he LTV raio. The measure 5 The counercyclical capial buffer is o be raised o 1.5 percen in June

12 produced some effec as he average LTV raio for new morgage borrowers has sabilized a around 7 percen, haling a rising rend which led he average LTV raio o reach abou 72 percen in 21. Noneheless, abou half of he new borrowers in 214 had an LTV raio jus below he cap. In a conex of double digi growh raes for house prices, house purchasers could hereby increase heir DTI raio while sill meeing he 85 percen LTV cap. Figure 7: Swedish household indebedness and housing price expecaions Ineres rae expecaion 1 year's ime Ineres rae expecaions 2 year's ime Sources: NIER and auhors calculaion. 3 The Model To analyze he coss and benefis of macroprudenial and moneary policies in reducing household deb we use a DSGE model wih financial fricions and an imperfecly compeiive banking secor ha is based on he model of Gerali e al. (21). We modify ha model along wo dimensions. Firs, we ailor he model o Sweden s characerisics by dividing he world ino a small open economy and he res of he world or foreign economy. Second, we inroduce hree macroprudenial measures in addiion o he LTV raio: he amorizaion requiremen, he ax deducibiliy of morgage ineres paymens, and morgage risk weighs. The home economy is affeced by he foreign economy while he reverse is no rue. The fac ha Final consumpion goods are raded and home savers can inves in foreign bonds. The home economy is populaed by wo ypes of households, paien P and impaien I and 12

13 by enrepreneurs. Households consume, work and accumulae housing (in fixed supply), 6 while enrepreneurs produce a homogenous inermediae good using physical capial bough from capial-good producers and labor supplied by households. Agens (households and enrepreneurs) have differen degree of impaience refleced in differen discoun facors for heir fuure uiliy. The heerogeneiy in agens discoun facors provides a simple way o generae financial flows in equilibrium: paien households (savers) purchase a posiive amoun of saving asses (deposis a domesic banks and foreign bonds) and do no borrow, while impaien households (borrowers) and enrepreneurs borrow from he domesic banking sysem. When aking a bank loan, borrowers face a borrowing consrain. In he case of he enrepreneurs, hey can only borrow up o a fracion of omorrow s collaeral i.e. he value of privae physical capial. Whereas for households, he model includes wo consrains. We incorporae an LTV ceiling by allowing impaien households o borrow up o a fracion of he value of new housing acquisiions each period. We allow for an amorizaion requiremen by assuming ha he impaien households mus repay a fixed fracion of he loan principal each period. These repaymens of loan principal make he average LTV on ousanding morgages lower han he average LTV of new morgages. In addiion, we model he ax deducibiliy of morgage ineres paymens by having he borrowers receive, from he governmen, a ransfer ha covers a porion of heir morgage ineres paymens. The governmen imposes a ax on enrepreneur s profi o finance is expendiure. For simpliciy, we assume he governmen runs a balanced budge using a lump sum ransfer from/o he households o ensure his balance is respeced each period. Financial flows are channeled hrough an imperfecly compeiive banking secor. Banks supply deposis and loans o he agens, and se ineres raes on boh deposis and loans in order o maximize profis. For insance, a reducion in bank profi would lead o a cu in deposi raes, in urn lowering saver s income. The amoun of loans issued by each inermediary can be financed hrough he deposis hey raise and by reinvesed profis (bank capial). Banks need o respec a minimum risk weighed capial requiremen, and since any deviaion from he required raio would be cosly, hey adjus ineres raes in order o converge back o he requiremen. Households supply heir differeniaed labor services hrough unions which se wages o 6 This assumpion may no be ha implausible given he housing sock per capia has remained largely unchanged since 199s in Sweden. 13

14 maximize members uiliy subjec o adjusmen coss. Labor services are sold o compeiive labor packers which aggregae hem ino a single labor inpu sold o firms. There are wo addiional secors which produce goods, he reail secor and he capial good producer secor. In he monopolisically compeiive reail secor reailers buy inermediae goods from enrepreneurs a he wholesale price, hen re-sale he goods a a mark-up over he purchasing cos and ha is subjec o adjusmen coss. In he capial-goods producing secor, producers buy old capial and urn i ino new producive capials. As in Gerali e al. (21), producers of physical capial goods are used as a modeling device o make explici he expression for he price of capial, which eners enrepreneurs borrowing consrain. In he Appendix we describe he key feaures of he model borrowing largely from Gerali e al. (21) for he common secions. 4 Calibraion and Esimaion Model parameers are derived hrough a combinaion of calibraion and esimaion. The parameers deermining he seady sae are calibraed in order o obain reasonable values for some key seady-sae values and raios. We esimae he parameers ha are difficul o calibrae, or he ones ha we have very lile informaion abou, using a Bayesian approach. Esimaion of he implied poserior disribuion of he parameers is done using he Meropolis algorihm (see Smes and Wouers (23) and Lindé e al. (29)). We use welve observed series: real GDP, real consumpion, real invesmen, ineres rae on morgages, ineres raes on corporae loans, deposi raes, he Riksbank s repo rae, real loans o households, real loans o firms, wage inflaion, CPIF inflaion, and real house prices. 7 The sample period runs from 1996Q1 o 214Q4 and we remove he rend from he variables using he Hodrick-Presco filer. 4.1 Calibraed parameers and prior disribuions Calibraion. Table 1 repors he values of he calibraed parameers. To calibrae he model o resemble he Swedish economy we use parameers ha have already esimaed in he RAMSES model for he Swedish economy, see Adolfson e al. (28). For example, he paien households discoun facor is se o.99631, and hose of impaien households and enrepreneurs a.975, he same as in Gerali e al. (21) and in he range suggesed by Iacoviello (25) and Angelini e al. (214). The mean value of he weigh of housing in 7 See Appendix B for a descripion of he daa. 14

15 households uiliy funcion is calibraed a.2 following Gerali e al. (21). Table 1: Calibraed parameers Parameer Descripion Value β P Paien households discoun facor.9963 β I Impaien households discoun facor.975 β E Enrepreneurs discoun facor.975 φ Inverse of he Frish s elasiciy 2.98 µ Share of unconsrained households 1/3 ɛ h Weigh of housing in he households uiliy funcion.2 α Capial share in he producion funcion.35 δ Depreciaion rae of physical capial.25 ɛ y is he markup in he goods marke 6 ɛ y ɛ y ɛ y 1 ɛ l ɛ l 1 is he markup in he labor marke 5 m I LTV for new morgages.85 m E LTV for firm loans.25 ν b Risk weighed/loan raio in seady sae (requiremen).12 ɛ d ɛ bh ɛ be ɛ d ɛ d ɛ 1 bh ɛ bh ɛ 1 be ɛ be 1 markdown on deposi rae 1.1 markup on loans o households 3.2 markup on loans o enrepreneurs 2.4 For he loan-o-value (LTV) raios, we se i m i a.85 in line wih he curren LTV cap for new morgages se by he Swedish FSA. The calibraion of m e, i.e. he loan-o-value consrain for he enrepreneurs is more challenging; we calibrae i o.25 so ha he raio beween morgages and loans o enrepreneurs is abou 1.3 as indicaed by he lending daa from Swedish moneary financial insiuions. We calibrae he amorizaion period for exising morgages o be 5 years and morgage ineres rae deducabiliy o be 3 percen. The capial share is se a.35 wih a depreciaion rae of.25 which is sandard in he lieraure. Similar o Gerali e al. (21) we assume a markup of 25 percen and se ɛ a 5. In he goods marke, a value of 6 for ɛ y delivers a markup of 2 percen in he seady sae. We calibrae he elasiciy of subsiuion of deposis ɛ d o maching he seady-sae spread beween he deposi rae and inerbank rae ha is in he range of 5-1 bps (see Table 2). We se he minimum bank (risk weighed) capial raio o 12 percen higher han he Basel III requiremens a 1.5 percen reflecing he morgage risk weigh floor. Moreover, we se he bank risk weighs o 5 percen for corporae loans and o 25 percen for morgages o according o he exising risk weighs floor. We assume ha bank capial depreciaes 15

16 (i.e. i is used for managerial purposes) a he rae of.658, which ensures ha he raio of bank capial o risk weighed loans is.19, which is consisen wih he Swedish bank s high risk weighed capial raios. Prior disribuions. Tables A1 and A2 lis our priors. These are guided by previous lieraure, in paricular Adolfson e al. (28), Lindé e al. (29), Gerali e al. (21), and Walenin (214). In cases where we did no find suiable examples we keep he prior relaively uninformaive. In paricular, we choose a bea-disribuion wih a prior mean of.75 and sandard deviaion of.5 for he persisence parameers. We assume ha all agens have he same habi persisence parameers in consumpion (i.e. a h = a P = a I = a E ) wih mean value of.65 and sandard deviaion of.1. For he moneary policy rule, we se he prior mean of ρ R, Φ π and Φ y o.8, 1.7, and.13 respecively, in line wih Adolfson e al. (28). For he LTV, we se he prior mean on ρ mi o.75 implying i akes some ime for any announced LTV o be implemened. 4.2 Poserior esimaes Tables A1 and A2 repor he poserior mean and 9 per cen probabiliy inervals for he srucural parameers, ogeher wih he mean and sandard deviaion of he prior. In addiion, he ables repor he marginal densiy of he parameers and Figure 7 repors he prior and poserior marginal densiies of he parameers in he model. Draws from he poserior disribuion of he parameers are obained using he random walk version of he Meropolis algorihm. We run 2 parallel chains each of lengh 12,,; he small number of chains was in par due o heir lengh. The scale facor was se in order o deliver accepance raes in he neighborhood of.3. Convergence was assessed by means of he convergence saisics aken from Brooks and Gelman (1998) on individual srucural parameers as well as he mulivariae version. We find a relaively high degree of ineres rae ineria in he moneary policy rule, which is consisen wih Adolfson e al. (28), bu our esimaes sugges more responsiveness of moneary policy o inflaion and oupu (i.e. higher esimaed Φ π and Φ y ). One explanaion could be ha he Riksbank has changed is reacion funcion, becoming more aggressive during he crisis, as Adolfson e al. (28) naurally covers he pre-crisis period. Regarding nominal rigidiies, we find ha wage sickiness is much more significan han price sickiness. This may reflec he fac ha wages are in general only re-negoiaed every hree years, and he goal of collecive wage bargaining is o preserve a seady rise in wages based on rends 16

17 in produciviy in Sweden. Regarding he degree of sickiness in bank ineres raes, we find ha deposi raes adjus more rapidly han he raes on loans o changes in policy raes. This is no surprising given ha Swedish households can easily swich heir deposis ino oher financial insrumens, hereby banks end o adjus deposi raes more frequenly. Finally, we found morgage raes adjus faser compared wih he lending rae o firms, reflecing he high share of variable rae morgage conracs in Sweden. 8 5 Model Properies To illusrae he broad properies of he model we provide impulse responses, focusing on he impac of a conracionary moneary policy shock. We also analyze how he impac of such a shock varies wih differen levels of macroprudenial requiremens, because he model has a number of feaures besides he radiional ineres rae channel which shape he ransmission of an ineres rae increase: A balance shee channel reflecs he collaeral consrain on household borrowing. A igher moneary policy sance lowers housing prices, resricing household capaciy from borrowing and amplifying he effecs of he moneary policy shock. Amorizaion requiremens imply ha households canno re-opimize he oal morgage deb sock each period as hey mus carry forward he unpaid morgage principal. The inroducion of amorizaion requiremens is a key disincion beween his model and he main models in he lieraure (Gerali e al., 21) which assume ha households repay he morgage deb in full a he beginning of each period and ge a new morgage a he end of he period. The sickiness of he morgage deb in our model provides an addiional amplificaion channel for an increase in moneary policy raes on he macroeconomy, while dampening he impac on household deb. The banking secor dampens he response of reail ineres raes o a moneary policy shock, especially owing o he sickiness of bank lending raes. This moderaes he reducion in lending, consumpion and invesmen following a moneary ighening compared wih Gerali e al. (21). Moreover, he bank capial consrain inroduces a furher wedge beween he bank lending and deposi raes and moneary policy raes. 8 The esimaed lending rae adjusmen cos parameers are higher han oher esimaes, bu using hese lower esimaes is no found o significanly aler he impac of moneary policy shocks. 17

18 % dev from SS Percen % dev from SS Percen % dev from SS % dev from SS The fac ha households can choose beween o save in domesic banks or o purchase foreign bond affecs he ransmission of moneary policy. Figure 8 shows he impacs on oupu, inflaion, and he household DTI raio of an exogenous 1 basis poin hike in he moneary policy rae. The responses of he main macroeconomic variables are qualiaively comparable wih esimaes in he lieraure (Adolfson e al., 28). Hence, our model has he advanage of inroducing new elemens enriching he iner-linkages beween macroeconomic and financial variables, while mainaining properies ha are consisen wih he research on Swedish business cycles. Figure 8: Moneary policy shock Oupu Inflaion Quarer Consumpion Quarer Bank ineres rae spread Quarer hh deb-o-income Quarer Mon. Pol. Rae Baseline: Amor=5yr, LTV=85% Amor=3yr, LTV=8% Quarer Quarer The figure depics he impulse responses o a 1 percenage poin exogenous increase in repo rae. Moreover, he figure compares how would he responses differ beween cases wih higher and lower amorizaion requiremens. Following a 1 basis poin rise in ineres raes, oupu and inflaion fall by abou.6 percen and.2 percen respecively relaive o he seady sae. Loans o boh households 18

19 and firms fall, reflecing he decline in asse prices, i.e. he price of housing and he value of firm s capial, and he increase in he real ineres rae. Bank loan raes increase much less han he policy rae reflecing he imperfec pass-hrough of lending raes; however, he deposi rae increases by almos he full 1 basis poins, implying a decline in ha banks ineres rae margins. The response of bank capial is iniially negaive, reflecing he decrease in bank profiabiliy, bu i subsequenly increases as margins recover. The policy rae responds endogenously o he oupu and inflaion deviaions, hus i would fall below he seady sae value o simulae he economy, and reverse hese impacs over ime. The ineracion wih macroprudenial policies can be bes described by looking a borrower s consumpion responses for a given shock wih varying levels of macroprudenial insrumens. Comparing a scenario wih lower household deb owing o faser amorizaion requiremen and lower LTV cap (red doed line) wih he baseline (blue doed line), he consumpion response o ineres raes shocks is smaller. The inuiion is ha a smaller household deb in equilibrium implies ha household s deb service burden rises less when a given rae shock his he economy, requiring a smaller reducion in consumpion and demand for housing. 9 6 Exploring he impac of macroprudenial policies We use he framework developed above o sudy he effeciveness of macroprudenial policies and assess he poenial side effecs of macroprudenial policies on he macroeconomy. The ransmission mechanisms of he differen macroprudenial measures are discussed, highlighing some imporan feaures of hese measures. The approach is o make a permanen change in one of he macroprudenial policy insrumens which would change he srucure of he economy and o sudy ransiion pahs of he variables in he model from one seady sae o anoher. For example, a permanen reducion in LTV raio would reduce borrower s capaciy o borrow hence he deb level. Such changes in he borrower s behavior would in urn inerac wih he savers, enrepreneurs, banks and oher agens in he model, unil he new equilibrium is reached. However, as he LTV requiremen only affecs new morgages and as amorizaion raes are currenly very low, i can ake some years for any newly inroduced LTV requiremens o work heir 9 I is also ineresing o noe ha banks ineres rae margin is squeezed more by ineres rae hikes as he loan sock increases implying a lower abiliy for banks o pass on he increasing financing coss. Moreover, a larger loan sock requires banks o mainain a higher level of deposis, and given ha households have he opion o inves abroad, deposi raes would have o increase more he larger he morgage sock. 19

20 way hrough he morgage sock implying a long ime o reach he new seady sae. We will herefore discuss he ransiion pah in wo pars: he shor-erm dynamics and he long-erm when he new seady sae is reached. 6.1 Loan-o-Value Raio A cap on he LTV raio consrains how much households can borrow agains heir housing collaeral when a morgage is originaed. Housing prices fall as borrower s effecive housing demand is cu by heir reduced amoun of morgage financing hey can access. Moreover, falling house prices reduce collaeral values, reinforcing he impac of he iniial ighening of he LTV cap. In wha follows, we consider a scenario in which he LTV cap is reduced by 5 percenage poins o 8 percen. Figure 9: Impacs from a permanen reducion in LTV: LTV Deb (max, SS) DTI (max,ss) Cons (max,ss) The figure depics maximum impacs on household morgage deb, deb-o-income(dti) and consumpion(cons) following a permanen reducion in loan-o-value (LTV) raio from 85 o 8 percen. And changes in he hree variables in he new seady sae (LTV= 8) compared wih he baseline (LTV= 85). Over he shor run. Borrowers consumpion falls by abou.2 percen one year afer he LTV is lowered he modes impac in par reflecing ha, by consrucion, only new housing acquisiion is affeced by he change in LTV cap. Savers consumpion falls oo, alhough by less, as he fall in he sock of morgages lowers bank profiabiliy, hence he deposi rae offered o he savers declines and falling house prices imply a negaive wealh effec. Moreover, marke clearance in he housing marke implies ha savers need o increase 2

21 housing purchases aided by falling house prices. Alogeher, he ighening of he LTV cap has a conracionary effec on he economy lowering consumpion by.2 percen one year afer he shock, and GDP would remain lower by.1 percen hree years afer he shock. In he new seady sae. Borrowers deb declines by abou 1 percen cumulaively, in par because house prices fall abou 1.5 percen. The relaively modes fall in house prices reflecs o some exend he significan price elasiciy of saver s demand for housing demand compensaing for he decline in borrowers demand (as supply relaive o populaion is assumed o be fixed). Noably, borrower s consumpion of goods will be permanenly higher by abou 1.8 percen in he new seady sae, as heir deb service burden is lower, parly offseing he decline in heir consumpion of housing services. Bu saver s consumpion would coninue o decline during he ransiion, and will be.4 percen lower in he new seady sae. This resul is driven by bank profis falling by abou 4 percen, as banks cu back on morgage lending. Lower bank profis also imply ha banks deleverage, cuing loans o firms by abou 1.2 percen implying lower invesmen and producion. As a resul oupu will be abou.5 percen lower in he new seady sae. 6.2 Amorizaion requiremens Wih he inroducion of amorizaion requiremens, a porion of he morgage principal mus be repaid each period, in an amoun se by he amorizaion plan in he loan conrac. Ye, households can borrow more each period, up o he LTV ceiling, for new housing invesmen, implying ha household deb is posiive in he seady sae. Re-wriing impaien household s borrowing consrain (see Appendix A), i is clear ha household deb a any given ime equals he presen discouned value of he porion of he iniial deb principal ha is no amorized (firs par of he equaion 1) plus a sream of new loans ha were aken ou for new housing invesmen (ne of any repaymens of such loans). Thus, by specifying a faser amorizaion plan, household deb is reduced. b = (1 ρ)n n 1 (1 ρ) i n 1 j= π b n + i 1 j i= j= π m q h (1) j A he same ime, i is imporan o noe ha, if a household signed up for a lenghier morgage conrac i would be forced o carry a larger porion of he deb from one period o he nex, and could no reduce he deb sock by more han wha has been defined in he conrac for he amorizaion plan. We choose his modeling approach as early repaymen of morgage deb can be very expensive in Sweden as discussed in secion II. Figure 21

22 1 illusraes he impac of ighening of amorizaion requiremens equivalen o a 5-year reducion in mauriy o 45 years. Figure 1: Impacs from a permanen change in amorizaion requiremen: Amorizaion Deb (max, SS) DTI (max,ss) Cons (max,ss) The figure depics maximum impacs on household morgage deb, deb-o-income(dti) and consumpion(cons) following a permanen reducion in amorizaion requiremen from 5 o 45 years. And changes in he hree variables in he new seady sae (Amorizaion= 45years) compared wih he baseline (Amorizaion= 5years). Over he shor run. The impac on consumpion parly depends on he balance beween a ighening of household s cash flow consrain from larger amorizaion due and from a reducion in heir deb hence lower fuure ineres paymens which would relax household s budge consrain. The model suggess ha borrower s consumpion would fall by a very small amoun, i.e.,.3 percen by 4 years afer he shocks. This resul parly reflecs he significan endogenous policy response as he moneary auhoriy lowers he repo rae by almos 1 percenage poin. These accommodaive moneary condiions imply ha saver s consumpion would increase bu o a lesser exen. In aggregae, a ighening of he amorizaion requiremen has a small negaive impac on growh reducing GDP by.2 percen, and inflaion by almos.5 percen. Moneary policy herefore plays an imporan role in cushioning he adverse impac on he macroeconomy over he shor run. In he new seady sae. Borrower s deb will fall by abou 1 percen, wih household DTI raio falling by he similar amoun, and house prices down by.5 percen. In addiion, borrower s housing sock will be abou.5 percen lower in he new seady sae. Borrower s 22

23 consumpion will be abou 1.7 percen higher permanenly. Saver s consumpion will decline by.4 percen. This again reflecs a lower bank profi by almos 5 percen leading o a decline in credi o firms by more han 1 percen. As a resul, oupu is lowered by abou.4 percen. Despie boh a reducion of he LTV cap and a ighening of he amorizaion requiremen lead o qualiaively similar new seady saes, here are ineresing differences for he ransiional pah of he variables. In paricular, household deb falls almos linearly over ime if he amorizaion requiremen is ighened. Insead, following a reducion of he LTV cap he fall of household deb acceleraes, re-enforced by he falling house prices. This has implicaions for he ransiion pah of he borrower s consumpion which would fall much faser over he shor erm in he case of a reducion of he LTV cap han in he case of a ighening of he amorizaion requiremen. 6.3 Tax deducibiliy of morgage ineres A reducion in he ax deducibiliy of morgage ineres paymens increases he cos of servicing a morgage hereby ighening households budge consrain a negaive income effec. Moreover, such a reducion would make deb-financed housing purchases more cosly relaive o he price of consumpion goods a subsiuion effec. These wo effecs have opposing implicaions on households consumpion. The negaive income effec suggess ha a reducion in ax deducibiliy lowers household s consumpion, while he subsiuion effec implies ha households consume more as he relaive prices have made consumpion relaively cheaper han housing. Figure 11 illusraes he impacs on household deb and deb-o-income raio, consumpion, and borrowers housing sock following a 5 percenage poin reducion in ax deducibiliy o 25 percen. Over he shor run. The negaive income and subsiuion effecs noed above lower borrower s demand for housing, wih house prices falling by abou.6 percen in he near erm, and household deb would also fall. However, borrower s consumpion increases in he shor run as consumpion becomes cheaper relaive o housing, and also because i is assumed he governmen would fully redisribue he savings from he reducion in ax deducions, which offses some of he negaive income effecs. Savers would increase heir housing invesmen as he price declines, bu hey also benefi from he higher ransfers leading o higher consumpion. The overall impac is slighly higher aggregae consumpion, oupu and inflaion. In he new seady sae. Borrower s household deb is lowered by abou 2.2 percen, wih 23

24 Figure 11: Impacs from a reducion in morgage ax deducibiliy: Tax Deb (max, SS) DTI (max,ss) Cons (max,ss) The figure depics maximum impacs on household morgage deb, deb-o-income(dti) and consumpion(cons) following a permanen reducion in morgage ax deducibiliy from 3 o 25 percen. And changes in he hree variables in he new seady sae (ax= 25) compared wih he baseline (ax= 3). a similar change in he DTI raio. Their consumpion will be higher by abou.1 percen, wih a 2 percen lower holding of he housing sock. Savers consumpion also increases in his case by abou.1 percen, driven by higher ransfers as explained above. The laer would be sufficien o offse he decline in bank profi such ha a decline in saver s consumpion is avoided. Imporanly, he impac on oal consumpion depends on how he governmen uilizes he savings ha came from a reducion in ax deducibiliy. Figure 12 compares he baseline scenario he saving arising from he 5 percenage poin reducion in ax deducibiliy is fully re-disribued back o all households wih he assumpion ha hese savings are no redisribued. The figure illusraes ha he re-disribuion of he savings clearly helps o miigae he negaive impac on consumpion from reducing ax deducibiliy. 24

25 % dev from SS % dev from SS Figure 12: Households consumpion responses from a reducion in morgage ax deducibiliy: 3 25 Saver consumpion Quarer Borrower consumpion Baseline: full re-disribuion no re-disribuion Quarer The figure depics households (savers and borrowers) consumpion responses o a reducion in morgage ax deducibiliy over he iniial 2 quarers. I compares consumpion dynamics beween a full re-disribuion of he savings from reducion in ax deducibiliy via lump sum ransfers wih no re-disribuion. 6.4 Morgage risk weighs An increase in morgage risk weighs raises bank s required capial in proporion o heir morgage exposure. To accumulae more capial hrough profis 1, banks would increase he profi margins by raising he lending raes. A higher borrowing rae hen leads o a reducion in morgage demand hus lowering household deb. Ye, a decline in morgage credi erodes banks profiabiliy suggesing limied impac on household deb, or, in some cases, banks would reduce morgage raes o simulae higher morgage demand. In general, he higher bank capial requiremen leads o less bank credi including lending o firms, reducing oupu and aggregae consumpion. Thus moneary policy acs immediaely o cushion such a demand shock. Figure 13 illusraes a 5 percenage poin increase in morgage risk weighs o 3 percen. Over he shor run. Household deb would decline by abou.5 percen 2 quarers afer he shock. Borrower s consumpion would, by 2 years afer he shock, decline by a marginal 1 In pracice, requiring ha all capial accumulaion come from profis is an exreme case, as par of an increase in capial requiremens could be me by issuing new equiy. A he same ime, he higher requiremen would imply a need for banks o generae addiional profis in order o generae an adequae reurn on equiy, such ha wider margins may be needed even if capial is insead raised in he financial markes. 25

26 .1 percen, and aggregae consumpion by abou.5 percen. Lower aggregae consumpion reduces oupu and inflaion, which riggers moneary policy o reac by reducing he repo rae. The reducion in ineres rae leads morgage raes o decline, as well as firm lending raes. Thus aggregae consumpion begins o recover, and borrower s deb level will sar o rise. In he new seady sae. Borrower s deb remains almos unchanged, similarly wih he house prices. Ye, borrower s DTI raio increases by.5 percen. Boh saver s and borrower s consumpion will decline, implying a 2 percen permanenly lower aggregae consumpion. This resul is largely driven by he fac ha banks need o reduce credi supply o he firms. A lower firm lending affecs he invesmen, capial sock hus oupu. The simulaion suggess oupu would be 2.4 percen lower in he new seady sae. Figure 13: Impacs from a increase in morgage risk weighs: Morgage Risk Weighs Deb (max, SS) DTI (max,ss) Cons (max,ss) The figure depics maximum impacs on household morgage deb, deb-o-income(dti) and consumpion(cons) following a permanen increase in morgage risk weighs from 25 o 3 percen. And changes in he hree variables in he new seady sae (risk weigh = 3 percen) compared wih he baseline (risk weigh = 25 percen). 7 Welfare opimal macroprudenial policies Are borrowers beer off if policy makers ighen macroprudenial policies o reduce he probabiliy of a crisis down he road? More generally, is he sociey beer off? The resuls from he previous secion indicae ha ighening demand side macroprudenial policies can 26

27 reduce household indebedness wih relaively small negaive impacs on consumpion, parly because he borrowing households wih lower deb would have more resources available o consume. Moreover, a lower deb sock implies less volailiy in household s consumpion when shocks hi he economy. However, in he seady saes ha are associaed wih sricer macroprudenial requiremens he borrowers in general own less housing. So borrowers could be worse off if hey valued a higher housing sock more han a higher level of consumpion. One way o address his quesion is hrough a welfare funcion which akes ino accoun consumpion, housing accumulaion and labor supply for all he agens in he economy. Moreover, he welfare funcion also considers he disribuion of he housing sock as well as sochasic shocks o he economy, for insance shocks o household borrowing consrain and shocks o bank funding. 11 Wih a welfare funcion, we can search for he macroprudenial policy seings ha would maximize welfare. For his purpose, we follow Schmi-Grohé and Uribe (27) and Quina and Rabanal (214) in performing a second-order approximaion o he model s equilibrium condiions and o welfare, simulaing he model subjec o he sochasic shocks a he poserior mean of he model s parameers, and reporing he mean of welfare. We assume ha policymakers maximize he welfare funcion of all ciizens in he economy using he populaion weighs of he differen household ypes. We define he welfare funcion as: W = W saver + W borrower + W enrepreneur W i = U i + β i W i +1 i = saver, borrower, enrepreneur (2) Where W i is he welfare of he i h ype of borrowers, U i corresponds o agen s uiliy funcion, which increases wih higher consumpion and housing sock, bu decreases wih more hours of labor supply. Moreover, he disribuion of housing sock among he borrowers and savers, for a given level, maers as he uiliy funcion is concave in housing. 7.1 Macroprudenial measures Loan-o-value requiremen. We calculae welfare over a range of LTV requiremens. There are wo counervailing forces o deermine he LTV cap ha would maximize welfare: on one hand, if he LTV cap is oo low, mos of he housing sock will be owned by savers 11 For insance if he represenaive saver and borrower households each owed half of he sock of housing he level of social welfare would clearly differ wih respec o he case where saver households own all he housing sock. 27

28 Welfare as borrowers are more credi consrained; on he oher hand, if he LTV cap is oo high, more indebed households will need o go hrough deeper deleveraging process in response o adverse shocks, reducing mean welfare. We find ha he welfare improvemens become very small afer he LTV cap reaches 6 percen, bu ha mean welfare coninues o improve as he LTV cap ighens. Figure 14: Welfare: loan-o-value raio # LTV (percen) The figure depics welfare over a range of loan-o-value raios Amorizaion requiremen. Nex, we examine welfare over a range of amorizaion requiremens which apply o he exising morgage sock. Higher amorizaion requiremens imply households make a larger morgage principal paymen in every period, in proporion o o heir morgage sock. We find a highly non-linear relaionship beween required amorizaion period and welfare. Saring from an amorizaion plan ha requires households o repay heir morgages in 1 years, ighening he required amorizaion would increase welfare indicaing ha he benefi from lower deb for mean welfare ouweighs he coss from larger repaymens ha lead o lower housing socks for borrowers. The relaionship reaches a local maximum around an amorizaion requiremen of 6 years, hen welfare sars o decline unil he amorizaion requiremen reaches 3 years. Afer ha, i becomes opimal o furher ighen amorizaion requiremens. Tax deducibiliy on morgage ineres paymens. We hen invesigae he welfare implicaions of varying he degree of ax deducibiliy for morgage ineres raes. The baseline model is calibraed for a 3 percen deducion of morgage ineres paymens, and any reducion of his raio would imply ha he borrowers need o pay higher ineres on he exising morgage sock, inducing hem o borrow less and shif owards goods consumpion 28

29 Welfare Welfare Figure 15: Welfare: amorizaion # Amorizaion (year) The figure depics welfare over a range of amorizaion requiremens. Figure 16: Welfare: ax deducibiliy on morgage ineres repaymens # Tax (percen) The figure depics welfare over a range of ax deducibiliy. as discussed in he previous secion. We find ha welfare would decrease from lowering deducibiliy even hough lowering deducibiliy would lead o a reducion in household deb in our baseline model, which aids mean welfare in case of shocks. Risk weighs on morgages. Finally, we analyze supply-side measures invesigaing wheher higher risk weighs on morgages improve welfare as banks wih sronger capial 29

30 Household deb (index) Welfare buffers should have greaer abiliy o preserve funding inermediaion funcion during periods of sress (i.e. shocks o bank funding coss and profi margin) hereby reducing macroeconomic volailiy and improving welfare. On he oher hand, higher morgage risk weighs may lower morgage credi for households reducing borrower s housing sock, which may reduce welfare (see figure 18). We find ha welfare improves as risk weighs on morgages increase wih he marginal improvemens diminishing noably when he risk weigh exceeds 4 percen. Figure 17: Welfare: morgage risk weighs # Morgage risk weighs (percen) The figure depics welfare over a range of morgage risk weighs. Figure 18: Household deb: morgage risk weighs Morgage risk weighs (percen) The figure depics household real deb over a range of morgage risk weighs. 3

31 LTV 7.2 Ineracion beween macroprudenial measures Amorizaion and loan-o-value raio. As illusraed above, welfare displays a nonlinear relaionship wih amorizaion requiremens for a given level of he LTV cap. However, his relaionship can change wih differen levels of LTV. For insance, a high LTV cap implies ha households have more capaciy o borrow when shocks hi he economy suggesing ha i migh be opimal o have a higher deb and amorize lile. A he same ime, a lower LTV cap would limi he negaive impac on households when he amorizaion requiremen is ighened, as a igher LTV implies a lower deb level in seady sae, and he cos of a shorer amorizaion plan is proporional o he deb level. Thus he welfare gain from ighening amorizaion requiremens in he conex of a lower LTV cap could be larger. Indeed, our simulaion suggess ha welfare sricly increases wih a ighening of amorizaion requiremens when morgage loans have LTV cap of less han 8 percen. However, above ha hreshold, he welfare maximizing amorizaion period is in he neighborhood of 6 years. More ineresingly, our resuls sugges ha policymakers can achieve higher welfare using a combinaion of he wo measures. Figure 19: Welfare: ineracion beween amorizaion requiremens and LTV # Amorisaion (years) The figure depics welfare over a combinaion of amorizaion requiremens and loan-o-value (LTV) raios. The dark red color corresponds o he highes level of welfare, and dark blue represens he opposie. The scale is displayed by he verical bar on he righ. 31

32 Welfare Figure 2: Welfare: amorizaion requiremen #1 4-1 Baseline, LTV=85 LTV= Amorizaion (year) The figure compares he welfare over he same range of amorizaion requiremens, bu wih differen level of loan-o-value raios: baseline associaes a LTV of 85 percen, and anoher scenario considers a LTV of 8 percen. Tax deducibiliy and loan-o-value raio. Similarly, we are ineresed o analyze how welfare varies wih differen combinaions of ax deducibiliy for morgage ineres paymens and LTV raios. Ineresingly, we find ha when LTV raio is relaively loose, a abou 9 percen, i is welfare improving o have high ax deducibiliy. The resuls could reflec he fac ha a relaively loose LTV cap is associaed wih higher deb level in he seady sae. Thus a reducion in ax deducibiliy would be oo cosly for he households, which is sub opimal. However, for morgages wih LTV lower han 75 percen, i becomes sricly welfare improving o have lower ax deducibiliy. Similarly, i is found ha he highes welfare is achieved hrough a combinaion of lower ax deducibiliy and igher LTV cap. 32

33 LTV (percen) Figure 21: Welfare: ineracion beween morgage ax deducibiliy and LTV I # Tax deducibiliy (percen) The figure depics welfare over a combinaion of morgage ax deducibiliy and loan-o-value (LTV) raios. The dark red color corresponds o he highes level of welfare, and dark blue represens he opposie. The scale is displayed by he verical bar on he righ. 8 Discussion and Robusness Checks 8.1 Model esimaion Model fi. The esimaed model does a good job in maching some of he key raios in he daa as shown in able 2, ye, i does no perform as well in maching some of he second momens from he daa. In paricular, he esimaed model overpredics he volailiy of household credi, boh in absolue erms and relaive o he sandard deviaion of GDP or consumpion. In order o check he robusness of our policy experimens in secions 6 and 7, we repeaed all he policy experimens wih a calibraed version of he model ha mached well some of key second momens in he daa. Specifically, we calibraed he sandard deviaion of he shock processes o mach he sandard deviaions of GDP, consumpion, invesmen, household credi, corporae credi, and house prices. The resuls for he policy experimens remain qualiaively unchanged These resuls are available upon reques. 33