RDisney disc

Comments on:Haliassos-Karamanou-Ktoris-Syrichas (HKKS): mortgage debt, social customs, finance Martins-Villanueva: mortgages and young adults living with parents Grant-Padula: Informal credit markets, judicial costs & consumer credit: 

Comments on: Haliassos-Karamanou-Ktoris-Syrichas (HKKS): mortgage debt, social customs, finance Martins-Villanueva: mortgages and young adults living with parents Grant-Padula: Informal credit markets, judicial costs & consumer credit Finance & Consumption programme, EUI: Consumption and Credit in Countries with Developing Credit Markets: June 16-17, 2006 Richard Disney University of Nottingham and Institute for Fiscal Studies

Common themes….: 

Common themes…. In US (and UK and elsewhere), debt-finance durable purchases driven by financial conditions: Interest rates Downpayment restrictions Availability of credit (e.g. role of credit scoring) But in countries with developing credit markets, access to finance also conditioned on: Social capital (e.g.judicial system) Intrafamiliy transfers (e.g. parental housing gifts, inheritances) Informal credit markets These papers provide evidence from Cyprus, Italy & Portugal on relationship between financial markets & these institutional factors

HKKS – summary of paper: 

HKKS – summary of paper Compare Cyprus 1999 & 2002 Survey of Consumer Finances with US 1998 & 2001 SCFs. Estimate a 2-stage model of prob(owning a home) & prob(having/ever had a mortgage) across 2 countries. Argue that role of parental gifts/inheritance much greater in Cyprus than US, conversely role of household’s own wealth much greater in US. Implication: Sensitivity of US households’ purchases to liquidity constraints and to interest rates means housing market more sensitive to monetary policy in US than Cyprus.

How important are inheritance & parental gifts in Cyprus and US?: 

How important are inheritance & parental gifts in Cyprus and US? In US, one measure: Prob. of inheriting house = 0.231 (Table 7, X column(?)) In Cyprus, two measures: Prob. has received house as a gift = 0.357 (Table 6b) Prob. of inheriting house = 0.357 (Table 6c) In Cyprus, are these numbers supposed to be the same? Are they mutually exclusive or overlap 100%? If not, and ‘true’ prob(inheriting)<prob(gift=0.357) then Cyprus may not be so different from US….

How different are the coefficient estimates between Cyprus and US?: 

How different are the coefficient estimates between Cyprus and US? Look at whether has outstanding mortgage (US), has or has had mortgage (Cyprus)? Examine marginal effects (SEs): Prob. of mortgage if inherited: We can’t reject difference? Gifts probably different but we don’t have US data. Reported loan denials are much higher in US than Europe (Crook and Hochguertel) – not sure how we interpret SCF differences. Other dy/dx (*=significant):

Other points on HKKS: 

Other points on HKKS Estimate ‘probit models with selection’ Presumably (it’s not explained), mortgage finance eqn. is conditioned on owning a home? Something must ‘explain’ home ownership that’s not in mortgage eqn. (identification) – what is it? Samples are pooled across 1999 and 2002. Macroeconomic conditions seem to have differed across years (and been very volatile over this short period). Not a panel (but hard to use panel methods with 2 waves) but concern about coefficient stability with macroeconomic ‘shocks’ – all that’s tested is intercept stability.

Martins-Villaneuva: Summary of paper: 

Martins-Villaneuva: Summary of paper Why don’t young adults leave parental home in S. Europe? Answer: There are credit constraints limiting access to home ownership So look at the impact of a programme in Portugal (Credito Bonificado) that provided subsidies to low income buyers (Question for authors – first time buyers only or could I upsize and still use CB if satisfied eligibility?) Authors use quasi-Diffs-in-Diffs to show that changes to the programme had an impact.

The CB programme had a number of ‘experiments’: 

The CB programme had a number of ‘experiments’ Introduced in 1986, a subsidy to interest payments on mortgages to all with income In 1998-99 amended so that eligibility depended also on house being low price so now In 2002, programme abolished. 3 implicit ‘experiments’: eligibles post-1986 v non-eligibles eligibles post-1998-99 v. eligibles 1986 + non-eligibles All 2002 v. eligibles 1998-99 Identify off local variations in (mix-unadjusted) house prices.

Sample construction: 

Sample construction We want to find how CB ‘experiments’ affect probability of young adults leaving home. Cleverly, the authors match data from a household labour force survey on changes in household structure to data on whether individuals acquire mortgage debt. I assume that there are not major statistical problems in doing this non-random selection of matches in data problems with matching individuals to households there are clearly problems of identifying ‘exits’ from households that are genuinely young adults leaving What happens if households dissolve for other reasons e.g. separation, divorce? Etc.

Specific comments: 

Specific comments Frequently-presented paper, value added of new comments probably low? Implicitly diffs-in-diffs but ‘common trend’ assumption not tested explicitly in specification 1 (uses post-2002 as check on 1998-99 reform prediction) But standard diffs-in-diffs gave ‘noisy’ results (footnote 25) In fact some of the key estimates here are not very precisely determined. Take care in interpreting ‘treatment effects’ in non-linear models (Blundell et al). Bivariate probit ‘not identified by functional form’. But surely excluding an interaction from eqn. is identifying off functional form?

‘Big picture’ comments: 

‘Big picture’ comments Don’t overgeneralise from a single reduced form quasi-treatment result e.g. to rest of Europe. This is a ‘treatment effect for the treated’. An obvious point: housing subsidies get capitalised into house prices. Authors (p.23) ‘not much evidence of compression [of house prices]’. Does this mean no evidence or ‘not much’? Why don’t young adults who want to leave home rent instead of buying where borrowing constraints? Role of parental finance of house purchase (Cypriot paper) not discussed. So, is it really financial constraints or social constraints?

Grant-Padula: Informal credit, judicial costs and consumer credit: Summary: 

Grant-Padula: Informal credit, judicial costs and consumer credit: Summary Examines sample of unsecured debt contracts from Italian bank’s administrative data Theory suggests that probability of default on debt contracts depends: Negatively on threat of judicial enforcement Positively on access to ‘outside’ lending e.g informal sector But for uncollateralised debt, threat of judicial enforcement (proxied by regional differences in judicial efficiency) is weak. And so it proves. Matching regional variations in access to informal sector is more successful. Testing for adverse selection and moral hazard…..

A nice treatment of unobservability in loans!!: 

A nice treatment of unobservability in loans!! Well known that modelling loan defaults should be conditioned on the acceptance condition by lender But standard selection models aren’t plausible: Selection on observables implies lenders screen on variables that don’t affect repayment – why do this? Selection on unobservables implies lenders ignore variables in screening that do affect repayment – why do this? (legal requirements – can’t screen on race, gender, postcode?) Use variant of Manski procedure to bound results: ‘loose’ bounds mean refused loans must have had repayment probability between 0 and 1. ‘tight’ bounds implies rejected applicants cannot be better than accepted loans (screening efficiency assumption) So we can bound estimated coefficients.

Results: 

Results Regional variations in ‘judicial efficiency’ in Italy – a well-used variable in Italian studies (e.g. see MIT Book!). As authors point out, for lending without collateral, judicial efficiency unlikely to have an impact. Not surprisingly, it doesn’t. We need a proxy for the adverse impact on future loan applications (via credit score) of default. Match SHIW data on local variations in access to informal loans (‘friends and family’) – how plausible? In areas where these ties are strong, maybe people who apply for formal loans are non-random precisely because they don’t have access to family and friends. And maybe supply of formal and informal finance are closely related – e.g. one crowds out the other?

On moral hazard: 

On moral hazard Evidence of moral hazard? Outside option (family & friends) affects repayment probability and is known only to borrower not lender, so this is moral hazard (G-P) But G-P do not know whether household has access to family & friends – only a probability by matching from SHIW. Does Findomestica not have access to SHIW also? (or not know that family & friends are more important in the South?!) Maybe bank prices credit with expectation of default that varies across households – this is how our UK sub-prime lender behaves. So (in my view) not sufficient to argue for evidence of moral hazard (though may well exist)

(Finally!!…) On adverse selection: 

(Finally!!…) On adverse selection G-P match probability of having access to informal credit from SHIW to: SHIW sample Findomestica sample If (they argue) probability greater among Findomestica sample than among general population, then there is adverse selection. This argument looks a little stronger since I argue that Findomestic should know about local access to informal finance: Again, it may accept risk-return trade-off so not clear this is an informational asymmetry 0.3% is ‘small’ effect (although large relative to total usage of informal finance – 3% - but isn’t this itself rather low?)