The sustainability of credit to low income mortgagors: a risk management perspective

Transcription

1 The sustanablty of credt to low ncome mortgagors: a rsk management perspectve Mchel DIETSCH IEP - Unversté de Strasbourg et ACP Banque de France Joël PETEY IEP - Unversté de Strasbourg Journée OFCE Vlle et Logement 4 décembre 2012

2 Introducton and Objectves To gve access to homeownershp to low ncome households s an objectve of credt polcy for housng n many countres. However, f lower ncome borrowers generate more rsk, the ssue of the sustanablty of homeownershp s rased. A crteron of sustanablty s the ablty to maxmze the success of borrowers,.e. to gve access to homeownershp to selected low ncome borrowers at an acceptable rsk level for the borrowers and the lenders. Low ncome borrowers face two types of fnancal constrants: > Down-payment constrant; > Income constrant lmtng ther repayment capacty. Ther borrowng ablty depends upon the avalablty of nstruments helpng to dsentangle fnancal constrants : > Adjustable rate mortgages; > Publc fnancal assstance mechansms. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 2

3 Introducton and Objectves In ths paper, we adopt a lender s rsk management perspectve, what means measurng rsk at the portfolo level. Thus, to measure credt rsk, we rely on economc captal : > Economc captal s the amount of captal needed to mantan the solvablty of the lendng nsttuton for a gven horzon at a gven confdence level (protecton aganst unexpected losses). The hgher the default rsk of borrowers usng loans wth gven characterstcs, the hgher are economc captal requrements, and the less sustanable s a credt polcy orented towards such borrowers, from the lender s perspectve. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 3

4 Plan The role of loan characterstcs and the poston of the paper n the lterature The multfactor credt rsk modelng The data The results M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 4

5 Plan The role of loan characterstcs and the poston of the paper n the lterature The multfactor credt rsk modelng The data The results M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 5

6 French homeownershp publc assstance In France, publc fnancal assstance programs to promote homeownershp provde varous types of fnancal ad: > Interest rate subsdes whch reduce fnancal charges and deferred amortzaton (Prêt à taux zéro, PTZ): relax repayment as well as down-payment constrants > Access to housng allowances (Prêt conventonnés PC, Prêt à l accesson socale, PAS): relax repayment constrants > state guarantee (Prêt à l accesson socale). They potentally ncrease the solvency of low ncome borrowers. > Housng allowances attached to PC loans help absorbng ncome fluctuatons: reduce credt rsk. > Deferred amortzaton and zero nterest rate: ncrease credt exposure, ambguous effect on rsk. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 6

7 French homeownershp publc assstance Such publc assstance programs are condtonal to ncome celngs: > PAS and PC are only avalable to low ncome borrowers. > PTZ are avalable to medum-ncome borrowers. PC allowance PTZ threshold Baselne scenaro: 2 adults + 2 chldren, 2007 thresholds and allowance s formulas M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 7

8 Fxed vs. adjustable nterest rate loans Theory predcts that ARM borrowers are rsker than FRM borrowers: > Less rsk averson > Relax fnancng constrants: lower monthly repayments. > Hgher exposure to ncome rsk M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 8

9 Relaton to the lterature Default rates could be lnked to loans characterstcs : > Interest rate choce: Campbell and Cocco (2003) show that ARMs are rsker for wealth-constraned borrowers wth hgh LTV loans. Posey and Yavas (2001): rsker borrower have a preference for ARMs. Paella and Pozzolo (2007): lqudty constraned households prefer ARMs. > Loans gvng access to publc fnancal support: Ergungor (2010): default rates are condtonned by the avalablty of state fnancal support to low-ncome borrowers. Avery and Brevoort (2011) fnd that areas served by lenders covered by the Communty Renvestment Act (CRA) experenced lower delnquency rates and less rsky lendng. Goldberg and Hardng (2003) observe that low-ncome borrowers beneftng from a down payment program exhbt hgher default propenstes. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 9

10 Introducton and Objectves The role of loan characterstcs and the poston of the paper n the lterature The multfactor credt rsk modelng The data The results M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 10

11 M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 11 Credt losses wthn a mult-factor framework We consder a multfactor credt rsk model (Lucas et al., 2001, Tasche, 2006). Defnng u as oblgor s loss gven default (LGD) and 1 D hs default ndcator, credt losses are defned as: Default occurs when: Assumng perfect granularty of the portfolo and multvarate normal dstrbuton of rsk factors, credt losses can be approxmated by (condtonal) expected credt losses : ( ) = Φ Φ n R s p u L 1 1 ' 1 ' w w w = = n D u L 1 1 ( ) D p R s U 1 ' 1 ' 1 < Φ + = = ε w w w 1

12 Econometrc methodology : GLMM The credt rsk parameters (defaults thresholds and senstvtes to systematc rsk factors) are estmated usng a GLMM probt/bnomal model. The default rate s modeled as: P random effects Here, the random effects are latent to a segmentaton of borrowers accordng to loans characterstcs, assocatng one rsk factor for each segment. Consder successvely two segmentatons: > Access (or not) to publc fnancal support > ARMs vs. FRMs ( default γ ) = Φ[ x' μ + z' γ ] t t ratng of borrower (dummy varable) r t fxed effect t desgn matrx of random effects (dependence structure) M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 12

13 VAR and allocaton of captal rato Computaton of the dstrbuton of losses by Monte Carlo smulaton (usng the outputs.e. rsk parameters of the econometrc analyss). Then, propertes of Value-At-Rsk (VaR) allow to compute margnal contrbutons of gven portfolo s sub-segments as partal dervatves of the portfolo VaR (Gouréroux et al., 2000, Tasche, 1999). Thus, t s possble to measure whch part of unexpected losses can be attrbuted to each segment of the portfolo. Advantages of a mult-factor approach: > The captal allocaton accounts for the portfolo s heterogenety. > Allows detectng potental concentraton of losses (.e. stuatons of strongly correlated defaults) wthn partcular sub-segments of the portfolo. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 13

14 Implementaton of the multfactor approach Frst portfolo s segmentaton > Only Free market loans > Mx of free market loans and a PTZ > Only regulated loans Second portfolo s segmentaton > All fxed rates > Mx of fxed and adjustable rates > Only adjustable rate M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 14

15 Plan The role of loan characterstcs and the poston of the paper n the lterature The multfactor credt rsk modelng The data The results M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 15

16 Data Large portfolo of French housng loans: > The lender s a major player n the French market of loans beneftng from state fnancal assstance. > borrowers fles representng more than loans > Complete ratngs hstory of borrowers: nternal ratngs system wth 4 grades (3 + default) > On a quarterly bass over the perod: Q Q > Loans dedcated to personal homeownershp only (accesson à la proprété) > We assume the lender has an ncentve to control and to manage hs (credt) rsk. > Loan and borrower characterstcs: Type of loan : loans beneftng from publc fnancal assstance versus loans contracted of the free market Type of nterest rate : FRM versus ARM Varables enterng the nternal ratng: LTV, maturty, ncome, professon, housng status M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 16

17 Descrptve statstcs Segmentaton Interest rate type Type of loan (1) (2) (3) (1) (2) (3) Segments Fxed rate only Adjustable rate only Both types of rates Market loans PTZ loans PC loans % of borrowers Average EAD (euros) Average quarterly PD% Maturty > 20 years LTV at orgnaton>80% Age (% older than 36 years) Prevous poston (% tenants) M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 17

18 Plan The role of loan characterstcs and the poston of the paper n the lterature The multfactor credt rsk modelng The data The results M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 18

19 VaR and margnal contrbutons: results Captal ratos (quarterly) Panel B Publc subsdy segmentaton Sngle factor model Loan type model IRBA model Market loans 0,20% 0,22% 0,86% Mx of market loans and zero nterest rate loans 0,25% 0,27% 1,02% Loans wth publc assstance 0,22% 0,19% 0,90% The captal rato s the hghest n the free market + PTZ segment. Borrowers havng only access to the PTZ n terms of publc fnancal assstance have the hghest credt rsk. The lowest captal rato s n the segment of (low-ncome) borrowers usng regulated loans. > Publc assstance mechansms targeted on the lower ncome borrowers effectvely help to overcome fnancal constrants, wthout leadng to hgher rsk. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 19

20 VaR and margnal contrbutons: results Same credt rsk for borrowers usng fxed rate only and adjustable rate only, although ther default rates are dfferent. Borrowers usng (at least) two loans wth dfferent nterest rates (fxed and adjustable) are rsker. These borrowers face hgher fnancng constrants they try to relax addng an adjustable rate loan to a fxed rate loan. Borrowers wth only adjustable rate loans have a dfferent profle: they borrow on average the same level, but have on average hgher ncome. Dfferent motvatons for adjustable rate borrowng? Captal ratos (quarterly) Panel A Interest rate segmentaton Sngle Interest IRBA factor rate model model model Fxed rate only 0,22% 0,14% 0,91% Adjustable rate only 0,21% 0,13% 0,89% Both types of nterest rates 0,22% 0,31% 0,92% M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 20

21 VaR and margnal contrbutons: results As a robustness check, combne two dmensons: type of loan and type of nterest rate Captal ratos (quarterly) Segment Number of Proporton of Economc borrowers borrowers captal rato Fxed rate/market loan % 0.20% Fxed rate/ptz loan % 0.41% Fxed rate/pc loan % 0.11% Adjustable rate/market loan % 0.11% Adjustable rate/ptz loan 0 0% - Adjustable rate/pc loan % 0.11% Both rates/market loan % 1.08% Both rates/ptz loan % 0.22% Both rates/pc loan % 0.18% M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 21

22 Concluson Results show that borrowers usng regulated loans beneftng from publc assstance are on average safer than those borrowng wthout assstance n a free market, what means that publc assstance helps to solve potental solvency problems. However, borrowers mxng the two knds of loans (PTZ and free market loans) n order to get a larger amount of credt dsplay the hghest level of rsk. > These borrowers, not elgble to PC/PAS loans, buy relatvely more expensve dwellngs wth hgher loan-to-value loans. > What suggests that these borrowers try to overcome hgher fnancng constrants. > These medum-ncome borrowers are exposed to economc shocks when they are constraned to dsentangle fnancal constrants. > Lender-drven results? Lender nvest n medum ncome borrowers n order to capture future savng streams. Adjustable rate loans are not necessarly rsker. They are rsker when borrowers use adjustable rate loans to overcome fnancng constrants. M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 22

23 Thanks M. Detsch J. Petey : Sustanablty of credt polcy to low ncome mortgagors 23