eres2012_012.content

Households, Intermediaries and Originators in
Mortgage Markets
ERES Annual Meeting 2012, Edinburgh
Ruben H.G.M. Cox
Rotterdam School of Management, Erasmus University
Mortgage lending
The compensation structure of intermediaries
Mortgage production (€/year) € 0 - € 2.5 million € 2.5 - € 5 million
Bonus
€ 5 - € 10 million
0%
0.20%
0.25%
Traditional mortgage products
Linear mortgage
0.75%
0.95%
1.00%
Alternative mortgage products
Savings mortgage
Endowment mortgage
Life-insurance mortgage
2.50%
2.25%
2.25%
2.50%
2.45%
2.45%
2.50%
2.50%
2.50%
Source: Ecorys, 2004
Research idea
•
Why would a lender employ differential underwriting criteria for direct written
and brokered mortgages?
•
If lender has risk (and reputational) exposure, then it has screening
incentives;
– E.g. non-insured mortgages
•
If risk is transferred (securitized or insured), then screening incentives might
be weakened (Keys et al., 2010; Avery et al., 2006);
– E.g. insured mortgages
•
So brokers can only work within the acceptability criteria of the lender if
lenders have risk exposure.
Measuring the outcome of underwriting process
•
Driven by data availability;
•
Need a proxy for a ‘suitability’-standard (Inderst and Ottaviani, 2009)
•
Loan-to-value ratio:
– Measure of indebtedness of household
– Proxy for loss-given-default for originator/insurer
•
Debt-service-ratio:
– Measure of affordability for household
– Proxy for probability of default for originator/insurer
Hypotheses
Hypothesis 1
Intermediaries have an insignificant effect on LTV and DSR ratios if the mortgage is
uninsured.
Hypothesis 2
Intermediaries possibly have a positive effect on LTV and DSR for insured mortgages.
Dataset
•
Data are from the DNB Household Survey (DHS):
– Survey yearly administered among a representative sample of 2000
Dutch households;
– The survey started in 1993;
– Wide coverage of topics: work, wealth, income, housing, health,
psychological concepts.
•
Origination channel is administered since 2002
– Only first mortgages originated after 2001 are included in the sample.
•
Excesses based on underwriting standards published by NIBUD (Institute for
Budget Education) and National Mortgage Guarantee (NHG).
Descriptive statistics (1)
Insured
Mean
N
Eligible/not-insured
Mean
N
Non-eligible
Mean
N
Loan-to-value ratio
Intermediated
Direct written
98.5%
91.8%
131
84
87.4%
87.7%
115
98
100.0%
88.0%
424
318
Debt-service ratio
Intermediated
Direct written
28.0%
23.7%
109
70
28.4%
25.6%
97
83
22.4%
18.4%
549
395
Source: DNB Household Survey
Descriptive statistics (2)
Variables
Interest rate
Fixed rate period (1 = yes)
NHG-insured (1 = yes)
Log(propertyvalue)
Log (household income)
Log (amount of mortgage)
Direct written
Mean Std. Dev.
4.7%
0.01
88.8%
0.32
28.6%
0.45
4.87
0.68
10.50
0.43
4.86
0.66
N
660
741
742
737
742
589
Mean
4.6%
90.3%
36.8%
4.85
10.40
5.01
Intermediated
Std. Dev
0.01
0.30
0.48
0.63
0.46
0.53
N
1025
1151
1153
1144
1152
865
T-stat for
mean diff.
2.76***
-1.02
-3.70***
0.68
4.44***
-4.60***
Specification of regression model
•
General OLS-model:
𝑅𝑎𝑡𝑖𝑜𝑖 = 𝛼 + 𝛽𝑑𝐼𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑟𝑦 + 𝜔𝑡 𝑑𝑌𝑒𝑎𝑟𝑡 + 𝛾𝑋𝑖 + 𝜂𝑖
•
Ratioi is either LTV or DSR ratio for household i
•
Main variable of interest is dIntermediary
•
Control variables in X:
– Mortgage characteristics (fixed rate dummy, interest rate, mortgage
guarantee)
– Household characteristics (composition, education, income, value
collateral, wealth, gender, marital status etc.)
•
‘Excess’ analysis is based on a logit-model
Results (1): Full sample
(1)
LTV
0.037**
[2.21]
(2)
DSR
0.000
[0.02]
Controls
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
1333
0.303
-
1521
0.509
-
1333
0.164
-700.6
1521
0.303
-402.0
Intermediated (1 = yes)
N
R2-adj/Pseudo R2
Log-Likelihood
(3)
(4)
ExcessLTV ExcessDSR
0.331**
-0.235
[2.15]
[-1.15]
Mortgage controls: fixed rate-period, interest rate, mortgage-guarantee,
mortgage type
Demographic controls: household composition, education, income, value
collateral, wealth, gender, marital status, retired
Results (2): Results by originator
Banks
Non-Banks
LTV
0.019
[0.92]
DSR
0.008
[1.10]
LTV
0.045
[0.94]
DSR
-0.005
[-0.32]
Controls
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
N
R2-adj
731
0.339
812
0.493
602
0.260
709
0.536
Intermediated (1 = yes)
Insignificant results are also found for the logit-analysis (underwriting-excesses)
Mortgage controls: fixed rate-period, interest rate, mortgage-guarantee, mortgage type
Demographic controls: household composition, education, income, value collateral, wealth, gender,
marital status, retired
Results (3): Results by originator for insured mortgages
Intermediated (1 = yes)
Loan-to-value ratio
Full sample
Banks
Non-Banks
0.067**
0.060*
0.091
[2.51]
[1.93]
[1.32]
Debt-service-ratio
Full sample Banks
Non-Banks
0.014
0.010
0.030
[1.56]
[0.85]
[1.25]
Controls
Yes
Yes
Yes
Yes
Yes
Yes
Year dummies
Yes
Yes
Yes
Yes
Yes
Yes
N
R2-adjusted
454
0.151
241
0.154
213
0.160
483
0.586
242
0.604
241
0.597
Involvement intermediary increases household LTV-ratio by 6-7 percent
Results (4): Results by originator for uninsured mortgages
Intermediated (1 = yes)
Controls
Year dummies
N
R2-adjusted
Loan-to-value ratio
Full sample Banks Non-Banks
0.026
0.004
0.023
[1.15]
[0.17]
[0.41]
Yes
Yes
Yes
Yes
879
0.320
Yes
490
0.365
Yes
389
0.298
Debt-service-ratio
Full sample
Banks Non-Banks
-0.006
0.007
-0.020
[-0.75]
[0.73]
[-1.02]
Yes
Yes
Yes
Yes
1038
0.489
Involvement of broker is insignificant when mortgages are not insured
Yes
570
0.472
Yes
468
0.528
Economic significance
•
Is the monetary incentive large enough for the broker?
•
Assume 7 percent increase in debt-level for an insurable savings mortgage;
•
If 10 mortgages are 7 percent levered up, say from 250K -> 268K;
•
Additional broker income is 5400 euro or 8 percent increase;
•
Additional searching- and screening costs are small.
Robustness checks
•
We account for endogenous channel selection by households:
– IV-model yields similar results
•
We check for heterogeneity in risk-preferences, that might influence indebtedness:
– No effect was found
•
It is examined whether intermediaries have an effect on the pricing of the credit
(LaCour-Little, 1999):
– No effect was found
Conclusions and implications
•
Originator screening incentives and broker monitoring are maintained for uninsured
mortgages but appear to weaker for insured mortgages;
•
Limited evidence that households are expropriated by brokers;
•
Mortgage insurer is the ‘victim’ of the compensation incentives;
•
Redesign the insurance contract to a deductibles/coinsurance structure with the
lender.