Loan-to-Value Caps, Bank Lending, and Spill-over to ... · LTV cap introduced in 2011 at 75%, and increased to 80% in 2016m9. 2011m1 2016m9 75% 80% Exposed Unexposed Constrained Borrowers

Loan-to-Value Caps, Bank Lending,and Spill-over to General-Purpose Loans

Selva Bahar Baziki & Tanju CapacıogluCentral Bank of the Republic of Turkey

Macro Financial Analysis Division

Housing, Household Debt and PolicyReserve Bank of New Zealand

December 12, 2017Wellington, New Zealand

The views and opinions expressed are those of the author and do not necessarily reflect those of the Central Bank of theRepublic of Turkey or its employees.

Baziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 1 / 52

A rise in General-Purpose Loan Usage for Residential Borrowers

After the LTV cap increases from 75% to 80% at the end of September 2016, weobserve a rise in the median and total GP borrowing for new home owners.

Figure 1: Median general-purpose loans (thousand TL) by residential borrowers.

What explains this increase after the policy easing?Baziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 2 / 52

Questions

What is the impact of the introduction of restrictive LTV caps on bank lendingbehavior?

I Introducing a cap of 75% (2011) should affect banks’ lending and riskoptimization.

I ”Unmet lending appetite” redistributed among other retail? Commercial?

I ”Unmet risk appetite” triggers riskier lending behavior?

What is the impact of an easing change in LTV caps on non-residential borrowing byconstrained home owners?

I Increasing the cap from 75% to 80% (2016) should affect borrowers’ optimization

I Lower the additional borrowing?


Timeframe

LTV cap introduced in 2011 at 75%, and increased to 80% in 2016m9.

2011m1 2016m9

75%

80% Exposed

Unexposed Constrained Borrowers

Unconstrained Borrowers

Constrained Bor.

Unconstrained Bor.

Figure 2: Timeframe of the analysis


Contributions

I Use credit register data and appraisal reportsI Universal coverage on housing loans at the individual-loan levelI In a large emerging economy

I Policy changes were exogenous shocks

I Bank Assets and Risk Preferences: Contrast exposed and non-exposed banks interms of retail, commercial lending and risk preferences.

I Residential Borrowers’ Additional Loan Use: Contrast non-residential borrowing ofLTV cap constrained borrowers before and after the policy change

I Isolating demand and supply side effects


Analysis on Exposed/Unexposed Banks

How has the balance sheet changed?

2011m1 2016m9

75%

80%

• Same Location : Retail Lending • Same Firm : Commercial Lending

Exposed

Unexposed

Figure 3: Retail and Commercial Lending


Analysis on Exposed/Unexposed Banks

How has other lending to residential borrowers changed?

2011m1 2016m9

75%

80%

Constrained Borrowers


Exposed

Unexposed

Figure 4: General Purpose Loan Issuance


Preview of Results - Lending and Risk Taking

Introduction : High exposure banks

I Reduce residential loan amounts,

I Increase GPloans...

I ...and they do so for new residential borrowers ”unmet lending/risk appetite”

I Increase commercial loans...

I ...esp for riskier firms ”unmet risk appetite”

Expansion : Banks with average LTV closer to 75%

I Increase residential and GPloans esp for residential borrowers ”credit spill-over”

I They have lower commercial loan outcomes than banks farther away from thelimit but the effect is for a short period


Analysis on Constrained/Unconstrained Residential Borrowers

How has the additional demand of residential borrowers changed?

2011m1 2016m9

75%

80%



Constrained Bor.


Figure 5: Additional leverage???


Preview of Results - Easing

Higher LTV cap increases access to residential borrowing - should lower the need foradditional funds

But we see that average and total GP borrowing has increased after the easing policy

LTV cap constrained buyers are taking out more non-residential loans thanunconstrained : ”credit spillover”.

LTV cap constrained buyers after the change are taking out more non-residential loansthan before : ”flight to quality”?


Mechanism

Demand Side Factors

Introduction of a cap makes some buyers credit constrained.

Some buyers are discouraged from entering the housing market altogether.

Others will enter purchase contracts:

I Will need additional funds to cover expenses associated with a new home.I These purchases usually occur after the home purchase.

I May still need additional funds to cover the cost of the required downpayment.I These additional funds will be needed leading up to the home purchase.


Mechanism

Supply Side Factors

Introduction of a cap frees up funds

Banks may chose to reallocate these funds within their loan portfolio

I Either by extending residential credit to constrained applicants the bank wouldotherwise not be in business with,

I Or by extending a non-residential loan to residential loan applicants which wouldresult in a higher leverage ratio per the value of the house,

I Or switching sides completely and increase commercial lending.

Either path, the bank engages in riskier lending behavior.


Previous Literature : Macroprudential Policies

Cross-country studies: Kuttner and Shim (2016), Cerruti et al. (2015), and Lim et al.(2011) Micro-level data: Jimenez et al. (2012) and Igan and Kang (2012)

I Most macroprudential policies are linked to lower credit growthI Focus on macroprudential policies and house prices or financial cycles

Claessens et al. (2014) find policies targeting consumers help bank balance sheetsthrough lower leverage.

I But lower leverage in housing loans does not necessarily mean lower risk orleverage on the consumer side

I Riksbanken (2012) announced an increase in unsecured loans after theintroduction of LTV cap

I Furthermore, riskier lending on the commercial side would exacerbate thesituation.


Previous Literature : LTV and Income Related Caps

Hallissey et al. (2014) LTV and LTI together lower credit risk on new lending

I Loan-level data from 3 banksI Sharp increases in losses on defaulted loans for LTV > 85%.

Jung and Lee (2017) find DTI better at stabilizing HP than LTV.

I District-level data on transactionsI LTV alone is subject to higher LGD

Lee (2012) stable NPL on housing despite lower HP due to LTV and DSTI.

Jacome and Mitra (2015) LTV and DTI together lower credit growth (but not alwaysHP).

IMF Staff Guidelines (2014) imply that DSTI increase the effectiveness of LTV as theylimit the use of unsecured loans to meet the downpayment limit.


Data

Credit Registry DataI Bank-individual matched data covering 2,984,655 new housing loans that allow for

identification of 88% of all housing loans. Housing GP AverageDebt

I Interest rate, capital borrowed and remaining, maturity, collateral, location.

LTV DataI Available from 2010m1 on all housing loans with appraisal values of all purchases from

one dataset.I Another dataset from 2011m1 with further details which we use in the easing analysis.

Bank DataI Monthly balance sheet and income statements for 30 banks.I Private, state-owned, foreign-owned and participation banks.I Real assets, loan-to-asset and deposit-to-asset ratios, capital-to-assets ratio, liquid assets

to total assets, NPL ratios, ROA.

Macro-level DataI Monthly macroeconomic aggregates such as interest rates, IPI, CPI, REER. IntRates


Changes in Residential Loans

I Residential loans (county-month controls):

Loanb,i,l,t =β0 + β1Aftert + β2LTVb + β3After ∗ LTVb + ζBankObsb,t−1

+ γMacroObst−1 + αb + ηl + θt + εb,i,l,t . (1)

I BankObs are bank observables : capital, liquidity, NPL, ROA, loan-to-assetratios, and log real assets taken ex-ante.

I MacroObs are macroeconomic observables: interest rates, IPI, CPI, REER takenex-ante.
















Residential Loans

Residential Loans declined after the introduction, and increased after the expansion ofthe cap.

Table 1: Effects of LTV Regulations on the Amount of Housing Loans

Window: 1-Month 2-Months 4-months 6-MonthsTightening 2011m1

After*LTV -0.832*** -0.813*** -0.426*** -0.262***(0.188) (0.130) (0.093) (0.074)

Obs 35,516 80,316 177,573 264,962R-sq 0.240 0.236 0.232 0.232

Loosening 2016m10

After*LTV 1.187*** 0.470** 0.352*** 0.218**(0.320) (0.203) (0.133) (0.102)

Obs 74,896 147,838 268,475 400,175R-sq 0.242 0.238 0.240 0.245Bank Obs Yes Yes Yes YesBank FE Yes Yes Yes YesCounty*Month Yes Yes Yes Yes

RobustnessBaziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 17 / 52

GP Loans

GP Loans are larger, especially for residential loan customers, and in particularfollowing the increase in the cap.

Table 2: Effects of LTV Regulations on the Amount of GP Loans

Policy Tightening LooseningAfter*LTV 1.875*** 10.739***

(0.356) (0.158)R-sq 0.372 0.997GPL of NonResidential BorrowersAfter*LTV 1.884*** 8.473***

(0.373) (0.156)R-sq 0.367 0.997GPL of Residential BorrowersAfter*LTV 1.833*** 11.404***

(0.100) (0.158)Obs 80,316 154,803R-sq 0.246 0.987Bank Observables Yes YesBank FE Yes YesCity*Month FE Yes Yes


Changes in Commercial Loans

I Newly issued commercial loans (same firm borrowing from two banks):

Loanb,f ,t =β0 + β1Aftert + β2LTVb + β3After ∗ LTVb + ζBankObsb,t−1

+ γMacroObst−1 + αb + ηf ,t + εb,f ,t . (2)


















Commercial Loans

Commercial loan growth responds more in the introduction phase, the effect inloosening dies out very fast.

Table 3: Effects of LTV Regulations on the Amount of Commercial Loans

Policy Tightening LooseningLocation ControlAfter*LTV 4.448*** -0.280*

(0.283) (0.145)Obs 76,755 154,803R-sq 0.873 0.967City*Month FE Yes YesFirm ControlAfter*LTV 0.150*** -0.049***

(0.049) (0.010)Obs 206,935 397,820R-sq 0.679 0.739Bank Observables Yes YesBank FE Yes YesFirm*Month FE Yes Yes

Robustness


Changes in Commercial Loan Riskiness


Loanb,f ,t =β0 + β1Aftert + β2LTVb + β3Ratingb,f ,t + β4After ∗ LTVb

+ β5After ∗ Ratingb,f ,t + β6LTV ∗ Ratingb,f ,t

+ β7After ∗ LTV ∗ Ratingb,f ,t + ζBankObsb,t−1 + γMacroObst−1

+ αb,t + ηf + εb,f ,t . (3)



I Rating shows riskiness for firms ranked 1-4, and non-riskyness for banks ranked5,6 by banks.







+ αb,t + ηf + εb,f ,t . (3)










+ αb,t + ηf + εb,f ,t . (3)










+ αb,t + ηf + εb,f ,t . (3)





Risky Commercial Loans

Commercial loan growth in the initial phase is mostly in riskier firms.

Table 4: Effects of LTV Regulations on the Riskiness of Commercial Loans

Policy Tightening Loosening

Rating -0.033*** 0.015(0.004) (0.010)

Rate*LTV 0.139*** -0.076***(0.013) (0.010)

After*LTV*Rate -0.053*** 0.003(0.009) (0.002)

Obs 360,504 445,578R-sq 0.670 0.746Firm FE Yes YesBank*Month FE Yes Yes

Robustness


Recall: Analysis on Constrained/Unconstrained Residential Borrowers

How has the additional demand of residential borrowers changed

2011m1 2016m9

75%

80%



Constrained Bor.


Figure 6: Additional leverage???


LTV Ratio

I From the appraisal reports, we calculate the LTV ratio for each residential loan attime t for each borrower i using the formula:

LTVit =OriginalHousingLoanit

ExpertEvaluationit. (4)

I Where the appraisal report is done by a certified agency by the BRSA or theCapital Markets Board of Turkey.


LTV ratios before and after

LTV cap increase from 75% to 80% shifts LTV ratios in favor of higher leverage.

Figure 7: LTV histograms by loan amount.


Pseudo LTV Ratio

I Purchasing a home is a long term investment, and requires some planning. Takethe GP borrowing in the months leading up to the house purchase to beresidential-related.

I Define a new LTV ratio called Pseudo LTV, to calculate the effective leverage ofthe housing loan:

PseudoLTVit =OriginalHousingLoanit + GPloansit + GPloansit−1

ExpertEvaluationit. (5)

IntRates


LTV to Pseudo LTV

The month of the initial LTV introduction shows a fuzzy implementation and manyabove the official cutoff. Still, possible to see more people at or around the cutoff havehigher Pseudo LTV.

020

4060

8010

0G

ross

Loa

n-to

-Val

ue R

atio

0 20 40 60 80 100Loan-to-Value Ratio

Figure 8: Mapping of individual LTV to Pseudo LTV - January 2011


LTV to Pseudo LTV around the regime change

A sharper implementation and a larger concentration of high Pseudo LTV around thecutoffs.

020

4060

8010

0G

ross

Loa

n-to

-Val

ue R

atio

0 20 40 60 80 100Loan-to-Value Ratio

Figure 9: Mapping of individual LTV to Pseudo LTV - September and October 2016


LTV to Pseudo LTV last data

A sharper implementation and a larger concentration of high Pseudo LTV with longerpractice.

Figure 10: Mapping of individual LTV to Pseudo LTV - April 2017


Methodology - Diff in Diff Strategy

I For bank b, individual-credit i over the policy change:

GeneralPurposeloansibt =αi + ωAftert + θLTVconstit + βLTVconstit ∗ Aftert

+ ζPersonObsit + γBankObsbt + ηMacroObst + εibt .(6)

I ζPersonObsit are person observables : in-house credit rating, number of creditcards, and other outstanding debt.

I γBankObsbt are bank observables : capital, liquidity, NPL, ROA, loan-to-assetratios, and log real assets.

I ηMacroObst are macroeconomic observables: interest rates, IPI, CPI, REER.

IntRates









IntRates









IntRates









IntRates


Results - DID

Table 5: Differences in Differences Estimates

70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80

BaselineAfter 0.318 1.143** 1.150** 1.147** 1.077** 1.028**

(0.528) (0.500) (0.497) (0.493) (0.489) (0.479)LTVconstr 5.225*** 5.645*** 6.051*** 6.617*** 7.355*** 11.657***

(0.564) (0.588) (0.615) (0.657) (0.721) (1.090)LTVconstr*After 2.280** 4.886*** 5.224*** 5.346*** 6.021*** 3.325**

(0.974) (1.057) (1.103) (1.148) (1.246) (1.547)Larger SampleAfter 0.293 0.836 0.865 0.894 0.857 0.891

(0.648) (0.599) (0.592) (0.584) (0.575) (0.554)LTVconstr 3.894*** 4.193*** 4.571*** 5.094*** 5.742*** 9.493***

(0.532) (0.539) (0.549) (0.571) (0.606) (0.849)LTVconstr*After 1.501* 3.784*** 4.030*** 4.067*** 4.605*** 2.186*

(0.895) (0.886) (0.909) (0.928) (0.983) (1.175)

Alternate IntRates


Results - DID


70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80

BaselineAfter 0.318 1.143** 1.150** 1.147** 1.077** 1.028**

(0.528) (0.500) (0.497) (0.493) (0.489) (0.479)LTVconstr 5.225*** 5.645*** 6.051*** 6.617*** 7.355*** 11.657***

(0.564) (0.588) (0.615) (0.657) (0.721) (1.090)LTVconstr*After 2.280** 4.886*** 5.224*** 5.346*** 6.021*** 3.325**

(0.974) (1.057) (1.103) (1.148) (1.246) (1.547)Larger SampleAfter 0.293 0.836 0.865 0.894 0.857 0.891

(0.648) (0.599) (0.592) (0.584) (0.575) (0.554)LTVconstr 3.894*** 4.193*** 4.571*** 5.094*** 5.742*** 9.493***

(0.532) (0.539) (0.549) (0.571) (0.606) (0.849)LTVconstr*After 1.501* 3.784*** 4.030*** 4.067*** 4.605*** 2.186*

(0.895) (0.886) (0.909) (0.928) (0.983) (1.175)

Alternate IntRates


Results - DID Different time brackets

Table 6: Differences in Differences Estimates for various time brackets and LTV cutoffs

70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80

4 months: August 2016 to November 2016After 0.226 1.136** 1.135** 1.127** 1.059** 0.922*

(0.523) (0.517) (0.516) (0.514) (0.512) (0.506)LTVconstr 4.973*** 5.325*** 5.674*** 6.188*** 6.876*** 10.508***

-0.343 -0.358 -0.378 -0.405 -0.447 -0.67LTVconstr*After 2.399*** 3.967*** 4.221*** 4.315*** 4.913*** 2.845***

-0.592 -0.643 -0.673 -0.706 -0.767 -0.953Obs 92632 92632 92632 92632 92632 926326 months: July 2016 to December 2016After 0.208 1.263** 1.261** 1.250** 1.190** 1.057*

-0.566 -0.564 -0.564 -0.563 -0.561 -0.556LTVconstr 5.435*** 5.822*** 6.210*** 6.766*** 7.538*** 11.559***

-0.55 -0.584 -0.623 -0.676 -0.751 -1.149LTVconstr*After 2.601*** 3.774*** 4.041*** 4.126*** 4.640*** 2.217*

-0.653 -0.702 -0.745 -0.798 -0.881 -1.251Obs 133130 133130 133130 133130 133130 1331308 months: June 2016 to January 2017After -0.123 0.577 0.595 0.606 0.619 0.6

-0.681 -0.664 -0.663 -0.662 -0.66 -0.657LTVconstr 5.069*** 5.438*** 5.814*** 6.345*** 7.080*** 10.967***

-0.398 -0.422 -0.45 -0.488 -0.541 -0.833LTVconstr*After 3.674*** 4.753*** 5.118*** 5.321*** 5.914*** 3.744***

-0.566 -0.624 -0.665 -0.712 -0.789 -1.049Obs 177793 177793 177793 177793 177793 177793

IntRatesBaziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 32 / 52



70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80


(0.523) (0.517) (0.516) (0.514) (0.512) (0.506)LTVconstr 4.973*** 5.325*** 5.674*** 6.188*** 6.876*** 10.508***

-0.343 -0.358 -0.378 -0.405 -0.447 -0.67LTVconstr*After 2.399*** 3.967*** 4.221*** 4.315*** 4.913*** 2.845***


-0.566 -0.564 -0.564 -0.563 -0.561 -0.556LTVconstr 5.435*** 5.822*** 6.210*** 6.766*** 7.538*** 11.559***

-0.55 -0.584 -0.623 -0.676 -0.751 -1.149LTVconstr*After 2.601*** 3.774*** 4.041*** 4.126*** 4.640*** 2.217*


-0.681 -0.664 -0.663 -0.662 -0.66 -0.657LTVconstr 5.069*** 5.438*** 5.814*** 6.345*** 7.080*** 10.967***

-0.398 -0.422 -0.45 -0.488 -0.541 -0.833LTVconstr*After 3.674*** 4.753*** 5.118*** 5.321*** 5.914*** 3.744***

-0.566 -0.624 -0.665 -0.712 -0.789 -1.049Obs 177793 177793 177793 177793 177793 177793




70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80


(0.523) (0.517) (0.516) (0.514) (0.512) (0.506)LTVconstr 4.973*** 5.325*** 5.674*** 6.188*** 6.876*** 10.508***

-0.343 -0.358 -0.378 -0.405 -0.447 -0.67LTVconstr*After 2.399*** 3.967*** 4.221*** 4.315*** 4.913*** 2.845***


-0.566 -0.564 -0.564 -0.563 -0.561 -0.556LTVconstr 5.435*** 5.822*** 6.210*** 6.766*** 7.538*** 11.559***

-0.55 -0.584 -0.623 -0.676 -0.751 -1.149LTVconstr*After 2.601*** 3.774*** 4.041*** 4.126*** 4.640*** 2.217*


-0.681 -0.664 -0.663 -0.662 -0.66 -0.657LTVconstr 5.069*** 5.438*** 5.814*** 6.345*** 7.080*** 10.967***

-0.398 -0.422 -0.45 -0.488 -0.541 -0.833LTVconstr*After 3.674*** 4.753*** 5.118*** 5.321*** 5.914*** 3.744***

-0.566 -0.624 -0.665 -0.712 -0.789 -1.049Obs 177793 177793 177793 177793 177793 177793


GP Demand 2011

Constrained borrowers demand more GP

Table 7: LTV introduction and the effect on GP demand of the constrained

Window: 1-Month 2-Months 4-months 6-Months73 vs 78LTVconst 0.496*** 0.491** 1.463** 1.157***

(0.181) (0.198) (0.656) (0.432)Obs 7,211 15,309 35,199 54,30774 vs 79LTVconst 0.747*** 0.709*** 1.734** 1.364***

(0.189) (0.202) (0.734) (0.484)Obs 7,211 15,309 35,199 54,30775 vs 80LTVconst 0.785*** 0.744*** 1.870* 1.464**

(0.237) (0.226) (1.085) (0.712)Obs 7,211 15,309 35,199 54,307

Individual Observables Yes Yes Yes YesBank Observables Yes Yes Yes YesBank*Month FE Yes Yes Yes Yes

Robustness


GP Demand - 2016

Constrained borrowers demand more GP after the loosening.

Table 8: LTV cap increase and the effect on GP demand of the constrained

Window: 1-Month 2-Months 4-months 6-Months73 vs 78LTVconst 5.458*** 5.593*** 6.006*** 5.886***

(0.634) (0.401) (0.499) (0.355)After*LTVconst 4.739*** 4.135*** 5.360*** 5.126***

(1.471) (0.849) (0.805) (0.607)74 vs 79LTVconst 5.953*** 6.202*** 6.718*** 6.597***

(0.697) (0.445) (0.558) (0.396)After*LTVconst 5.646*** 4.828*** 6.009*** 5.743***

(1.559) (0.911) (0.891) (0.673)75 vs 80LTVconst 9.089*** 9.476*** 10.529*** 10.269***

(1.039) (0.679) (0.875) (0.619)After*LTVconst 4.111** 3.140*** 3.964*** 3.782***

(1.756) (1.089) (1.171) (0.871)Obs 41,292 88,448 173,609 262,691Individual Observables Yes Yes Yes YesBank*Month FE Yes Yes Yes Yes


Summary of Results - Introduction

Introduction of the LTV cap

I Reduces residential loans in favor of higher uncollateralized lending to newresidential borowers,

I And higher riskier lending on the commercial side. ”unmet lending/risk appetite”

I Results confirm that there is ”spillover” from housing to non-residential loans,increasingly so for the exposed banks.


Summary of Results - Easing

I Being constrained by the LTV ratio cap increases GP loans by 5 to 6 thousand TL.

I The added premium of being constrained by the LTV ratio cap after the regimechange is an additional 4 to 6 thousand TL.

I per-capita GP loans are at 12 thousand TL.I Surprising result since a larger LTV ratio cap should reduce the need for additional

loans.

I Results confirm that there is ”spillover” from housing to non-residential loans,increasingly so for the credit constrained.

I And that as the cap is increased, borrowers are switching to more expensivehomes, increasing demand for credit.


Robustness

Results are robust to different definitions of LTV, constraint cutoffs, and brackets.

LTV here can take on one of three values: GPlong

I Share of weighted LTV over total lending LTVb,i,t ∗Amountb,i,tTotalAmountb,i,t

I Share of lending above cutoff over total lending ExposureAmountb,i,tTotalAmountb,i,t

I Quantile Approach : compare top/bottom 5pctl

Where the appraisal report is done by a certified agency by the BRSA or the CapitalMarkets Board of Turkey.

General Purpose Use by Home Owners: GPLlong

Commercial Credit Use: CommercialLong

Commercial Credit Demand 2011: GPdemandL11

Commercial Credit Demand 2016: GPdemandL16

Pooled OLS of entire period : OLS Method


Conclusion

I With LTV, credit constrained individuals have a higher likelihood to top off theirresidential loans with other means of borrowing.

I LTV may also result in riskier bank balance sheets through increased riskiercommercial lending (2011) or uncollateralized loans (2016).

I LTV caps alone do not provide sufficient deterrence from residential leverage forhouseholds.

I Implications for credit availability, transmission, asset decomposition, and optimalrisk weight design.

I Best practice may have to introduce income related limits.


Appendix


Data - Average Debt

Table 9: Average Current Debt by Loan Type of Housing Loan Customers 2011m1-2017m4

Average Amount of Debt (Thousand TL)

Total Debt General Purpose Housing Vehicle Credit No ofTotal Debt w/o Housing Loans Loans Loans Cards Banks

04.16 21.27 17.20 22.12 125.35 37.69 8.10 1.4905.16 21.62 17.27 22.14 127.84 37.81 8.36 1.4906.16 22.45 17.53 22.40 130.66 38.25 8.36 1.5107.16 20.43 17.45 22.29 124.66 38.17 8.41 1.5008.16 21.12 17.44 22.15 120.89 38.27 8.59 1.4909.16 24.37 19.59 22.58 128.23 38.42 8.78 1.4610.16 25.60 19.94 23.14 134.96 38.65 8.77 1.4611.16 26.84 20.36 23.71 141.79 39.52 8.82 1.4712.16 26.92 20.76 24.25 144.56 40.08 9.01 1.4901.17 25.64 20.65 24.17 142.68 40.30 8.85 1.5102.17 25.82 20.70 24.26 148.70 40.43 8.76 1.4703.17 28.14 20.89 24.58 167.17 40.84 8.93 1.47

Data


Results - DID with Cluster


(1) (2) (3) (4) (5) (6)70 vs 75 71 vs 76 72 vs 77 73 vs 78 74 vs 79 75 vs 80

Baseline ClusteredAfter 0.318 1.143* 1.150** 1.147** 1.077** 1.028*

(0.466) (0.559) (0.546) (0.527) (0.504) (0.490)LTVconstr 5.225** 5.645** 6.051** 6.617** 7.355** 11.657**

(2.208) (2.356) (2.497) (2.676) (2.925) (4.479)LTVconstr*After 2.280* 4.886* 5.224* 5.346* 6.021* 3.325

(1.144) (2.470) (2.586) (2.684) (2.921) (2.109)Larger Sample ClusteredAfter 0.293 0.836 0.865 0.894* 0.857* 0.891*

(0.440) (0.510) (0.509) (0.495) (0.476) (0.489)LTVconstr 3.894* 4.193* 4.571* 5.094** 5.742** 9.493**

(1.914) (2.059) (2.190) (2.350) (2.561) (3.988)LTVconstr*After 1.501 3.784* 4.030* 4.067* 4.605* 2.186

(1.455) (1.900) (2.022) (2.095) (2.306) (2.142)

Back


Interest Rates and Spread

0

2

4

6

8

10

12

14

16

18

20

0

2

4

6

8

10

12

14

16

18

20

01.10

05.10

09.10

01.11

05.11

09.11

01.12

05.12

09.12

01.13

05.13

09.13

01.14

05.14

09.14

01.15

05.15

09.15

01.16

05.16

09.16

01.17

05.17

Spread btw General P. And Housing LoansGeneral‐PurposeVehicleHousing

Figure 11: Interest Rates and Spread Between Housing and GP Loans.

Data LTVg DID DIDresults DIDtime


Data - Housing Loans

Table 11: Summary Statistics on New Housing Loans

Date Total Loans Average Loans # of Credit # of People

04.16 3,838,708 114.61 33,495 33,23005.16 4,350,762 115.79 37,576 37,21606.16 4,390,482 118.07 37,184 36,84407.16 2,838,803 115.67 24,542 24,30708.16 4,238,492 113.40 37,378 37,06109.16 5,167,644 118.30 43,682 43,36010.16 6,132,442 122.43 50,089 49,67111.16 6,386,490 125.34 50,953 50,49112.16 6,328,205 126.49 50,029 49,52401.17 5,073,685 129.89 39,061 38,71202.17 5,339,186 131.04 40,745 40,30403.17 7,074,793 134.90 52,443 52,010

Data


Data - Housing Loans

Table 12: Summary Statistics on General Purpose Loans of Housing Borrowers

General Purpose Loans (Thousand TL)

Same Month 1 Month Before 1 Month After 2 & 3 Months Before 2 & 3 Months AfterDate Amount Number Amount Number Amount Number Amount Number Amount Number04.16 44,304 1,272 7,272 165 4,962 437 8,161 381 18,081 1,75305.16 65,253 1,760 4,867 190 11,955 785 10,846 445 17,171 1,93406.16 68,044 1,832 5,616 186 5,698 567 13,334 453 17,280 2,01007.16 55,473 1,055 4,688 171 5,653 432 5,786 266 13,595 1,18508.16 48,830 1,291 3,106 153 6,109 457 10,438 436 21,270 1,53009.16 65,801 1,385 5,251 209 9,264 562 12,001 569 24,761 1,56410.16 88,281 1,852 6,013 251 13,227 657 11,294 535 23,416 1,67711.16 99,374 2,232 10,036 336 17,251 830 14,235 666 27,050 1,82112.16 101,734 2,335 10,949 344 10,078 660 15,138 651 27,901 1,93801.17 98,759 1,807 9,934 293 9,749 621 16,710 679 22,666 1,88102.17 74,834 1,882 6,846 251 9,546 599 15,813 535 8,154 53303.17 106,362 2,475 8,412 323 10,822 791 18,848 672 0 0

Data


Residential Loans

Residential Loans declined after the introduction, and increased after the expansion ofthe cap.

Table 13: Effects of LTV Regulations on the Amount of Housing Loans

Policy Tightening (2011m1) Loosening (2016m10)Window 1-Month 2-Months 4-Months 6-Months 1-Month 2-Months 4-Months 6-Months

After*LTV -0.832*** -0.813*** -0.426*** -0.262*** 1.187*** 0.470** 0.352*** 0.218**(0.188) (0.130) (0.093) (0.074) (0.320) (0.203) (0.133) (0.102)

Obs 35,516 80,316 177,573 264,962 74,896 147,838 268,475 400,175R-sq 0.240 0.236 0.232 0.232 0.242 0.238 0.240 0.245

After*Exposure -0.144*** -0.101*** -0.049*** -0.027* 0.236*** 0.106** 0.144*** 0.150***(0.040) (0.026) (0.018) (0.015) (0.078) (0.053) (0.045) (0.038)

Obs 35,508 80,308 177,559 264,949 74,896 147,771 268,337 399,978R-sq 0.239 0.236 0.232 0.232 0.242 0.237 0.240 0.244

After*Quantile -0.056** -0.059*** -0.045*** -0.032*** 0.039*** 0.049*** 0.064*** 0.070***(0.024) (0.015) (0.011) (0.009) (0.009) (0.006) (0.005) (0.004)

Obs 16,795 36,336 54,869 85,663 68,787 135,133 248,843 371,641R-sq 0.274 0.268 0.249 0.245 0.247 0.242 0.244 0.247Bank Obs Yes Yes Yes Yes Yes Yes Yes YesBank FE Yes Yes Yes Yes Yes Yes Yes YesCounty*Month Yes Yes Yes Yes Yes Yes Yes Yes

GPshort Robustness


GP Loans

GP Loans are larger, especially for residential loan customers, and in particularfollowing the increase in the cap.

Table 14: Effects of LTV Regulations on the Amount of GP Loans


After*LTV 1.685*** 1.875*** 1.659*** 0.940*** 6.872*** 10.739*** 5.396*** 9.554***(0.471) (0.356) (0.257) (0.205) (0.139) (0.158) (0.206) (0.157)

R-sq 0.399 0.372 0.319 0.316 0.999 0.997 0.980 0.978

GPL of NonResidential BorrowersAfter*LTV 1.728*** 1.884*** 0.589** 0.188 3.633*** 8.473*** 4.694*** 8.403***

(0.605) (0.373) (0.261) (0.209) (0.123) (0.156) (0.216) (0.167)R-sq 0.406 0.367 0.322 0.316 0.999 0.997 0.978 0.975

GPL of Residential BorrowersAfter*LTV 4.182*** 1.833*** 1.968*** 1.854*** 10.879*** 11.404*** 11.459*** 11.595***

(0.208) (0.100) (0.087) (0.071) (0.236) (0.158) (0.137) (0.117)R-sq 0.298 0.246 0.240 0.257 0.992 0.987 0.964 0.959Obs 35,516 80,316 177,573 264,962 78,617 154,803 281,779 419,925Bank Obs Yes Yes Yes Yes Yes Yes Yes YesBank FE Yes Yes Yes Yes Yes Yes Yes YesCity*Month Yes Yes Yes Yes Yes Yes Yes Yes

GPLshort Robustness


Commercial Loans

Commercial loan growth responds more in the introduction phase, the effect inloosening dies out very fast.

Table 15: Effects of LTV Regulations on the Amount of Commercial Loans


After*LTV 8.796*** 4.448*** 0.659*** 1.922*** -4.560*** -0.280* 0.125 1.383***(0.479) (0.283) (0.180) (0.148) (0.104) (0.145) (0.116) (0.115)

Obs 33,742 76,755 169,285 251,782 78,617 154,803 281,486 375,597R-sq 0.899 0.873 0.863 0.859 0.977 0.967 0.967 0.965Bank Observables Yes Yes Yes Yes Yes Yes Yes YesBank FE Yes Yes Yes Yes Yes Yes Yes YesCity*Month Yes Yes Yes Yes Yes Yes Yes Yes

Credit Registry DataAfter*LTV 0.140** 0.150*** 0.135*** 0.058* -0.041*** -0.049*** 0.016 0.029***

(0.060) (0.049) (0.038) (0.030) (0.016) (0.010) (0.011) (0.009)Obs 115,759 206,935 389,945 635,786 161,186 397,820 655,631 893,937R-sq 0.682 0.679 0.682 0.676 0.742 0.739 0.744 0.740Bank Observables Yes Yes Yes Yes Yes Yes Yes YesBank FE Yes Yes Yes Yes Yes Yes Yes YesFirm*Month Yes Yes Yes Yes Yes Yes Yes Yes

CommercialLong Robustness


Risky Commercial Loans

Commercial loan growth in the initial phase is mostly in riskier firms.

Table 16: Effects of LTV Regulations on the Amount of Risky Commercial Lending


Rating -0.023*** -0.033*** -0.039*** -0.038*** 0.007 0.015 0.003 -0.002

(0.006) (0.004) (0.003) (0.002) (0.016) (0.010) (0.006) (0.004)Rate*LTV 0.104*** 0.139*** 0.162*** 0.155*** -0.057*** -0.076*** -0.066*** -0.064***

(0.019) (0.013) (0.010) (0.007) (0.017) (0.010) (0.006) (0.004)After*LTV*Rate -0.042*** -0.053*** -0.055*** -0.054*** 0.007* 0.003 -0.003* -0.005***

(0.013) (0.009) (0.006) (0.005) (0.004) (0.002) (0.002) (0.001)

Obs 159,592 360,504 783,051 1,298,736 136,032 445,578 1,234,823 1,555,586R-sq 0.699 0.670 0.645 0.627 0.795 0.746 0.711 0.737Firm FE Yes Yes Yes Yes Yes Yes Yes YesBank*Month Yes Yes Yes Yes Yes Yes Yes Yes

RiskyShort Robustness


GP Demand

Constrained borrowers demand more GP

Table 17: LTV introduction and the effect on GP demand of the constrained

(1) (2) (3) (4) (6) (8) (10)

Window: 1-Months 2-Months 4-Months 6-Months70 vs 75LTVconst 0.596*** 0.533*** 0.458** 0.215 0.157 0.923* 0.711**

(0.190) (0.182) (0.222) (0.177) (0.193) (0.534) (0.355)71 vs 76LTVconst 0.698*** 0.632*** 0.573** 0.322* 0.273 1.088* 0.857**

(0.193) (0.186) (0.225) (0.176) (0.196) (0.563) (0.373)72 vs 77LTVconst 0.759*** 0.693*** 0.658*** 0.393** 0.341* 1.224** 0.962**

(0.198) (0.190) (0.231) (0.177) (0.201) (0.602) (0.399)73 vs 78LTVconst 0.868*** 0.802*** 0.770*** 0.496*** 0.491** 1.463** 1.157***

(0.207) (0.198) (0.237) (0.181) (0.198) (0.656) (0.432)74 vs 79LTVconst 1.094*** 1.026*** 1.044*** 0.747*** 0.709*** 1.734** 1.364***

(0.223) (0.213) (0.256) (0.189) (0.202) (0.734) (0.484)75 vs 80LTVconst 1.184*** 1.122*** 1.208*** 0.785*** 0.744*** 1.870* 1.464**

(0.291) (0.277) (0.339) (0.237) (0.226) (1.085) (0.712)Obs 8,481 8,480 7,213 7,211 15,309 35,199 54,307

Individual Obs - Yes Yes Yes Yes Yes YesBank Obs - - Yes Yes Yes Yes YesMacro Obs - - Yes n/p n/p n/p n/pBank FE Yes Yes Yes - - - -Time (Year) FE Yes Yes Yes - - - -Bank*Month - - - Yes Yes Yes Yes

GPdemandS RobustnessBaziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 49 / 52

GP Demand - 2016

Constrained borrowers demand more GP after the loosening.

Window: 1-Month 2-Months 4-Months 6-Months70 vs 75LTVconst 4.460*** 4.460*** 4.262*** 4.522*** 4.699*** 4.756*** 4.577*** 4.734***

(0.530) (0.530) (0.312) (0.332) (0.430) (0.399) (0.301) (0.286)After*LTVconst 1.895 1.895 1.949*** 2.222*** 3.031*** 3.631*** 3.129*** 3.494***

(1.350) (1.350) (0.707) (0.762) (0.650) (0.658) (0.483) (0.492)71 vs 76LTVconst 4.483*** 4.753*** 4.561*** 4.825*** 5.069*** 5.121*** 4.932*** 5.088***

(0.521) (0.558) (0.330) (0.350) (0.457) (0.427) (0.320) (0.305)After*LTVconst 3.526** 3.957*** 3.253*** 3.672*** 3.993*** 4.719*** 4.024*** 4.479***

(1.388) (1.500) (0.765) (0.824) (0.705) (0.717) (0.527) (0.539)72 vs 77LTVconst 4.788*** 5.056*** 4.886*** 5.146*** 5.447*** 5.489*** 5.282*** 5.432***

(0.551) (0.589) (0.351) (0.372) (0.488) (0.458) (0.341) (0.327)After*LTVconst 3.949*** 4.408*** 3.497*** 3.946*** 4.342*** 5.110*** 4.365*** 4.849***

(1.407) (1.513) (0.788) (0.844) (0.748) (0.760) (0.559) (0.572)73 vs 78LTVconst 5.184*** 5.458*** 5.323*** 5.593*** 5.678*** 6.006*** 5.552*** 5.886***

(0.596) (0.634) (0.379) (0.401) (0.476) (0.499) (0.337) (0.355)After*LTVconst 4.241*** 4.739*** 3.668*** 4.135*** 4.850*** 5.360*** 4.804*** 5.126***

(1.371) (1.471) (0.795) (0.849) (0.754) (0.805) (0.572) (0.607)74 vs 79LTVconst 5.698*** 5.953*** 5.938*** 6.202*** 6.382*** 6.718*** 6.253*** 6.597***

(0.658) (0.697) (0.422) (0.445) (0.534) (0.558) (0.377) (0.396)After*LTVconst 5.082*** 5.646*** 4.334*** 4.828*** 5.478*** 6.009*** 5.410*** 5.743***

(1.463) (1.559) (0.856) (0.911) (0.839) (0.891) (0.637) (0.673)75 vs 80LTVconst 8.752*** 9.089*** 9.096*** 9.476*** 10.056*** 10.529*** 9.773*** 10.269***

(0.984) (1.039) (0.648) (0.679) (0.837) (0.875) (0.589) (0.619)After*LTVconst 3.521** 4.111** 2.689*** 3.140*** 3.487*** 3.964*** 3.543*** 3.782***

(1.657) (1.756) (1.029) (1.089) (1.109) (1.171) (0.828) (0.871)Obs 45,286 41,292 96,417 88,448 189,119 173,609 283,896 262,691Individual Obs - Yes - Yes - Yes - YesBank*Month Yes Yes Yes Yes Yes Yes Yes Yes

GPdemandS16 RobustnessBaziki & Capacıoglu (CBRT) LTV Caps and Additional Loans 50 / 52

Methodology - OLS Regressions

I Finally, we investigate if additional GP lending is motivated by housing LTV ratiosfrom January 2011-April 2017 with the following specification for individual-crediti , for bank b, and time t:

GPloansHousePrice ibt

=αi + βLTVit + ζPersonObsit + ωAftert + θLTVit ∗ Aftert

+ ρPersonObsit ∗ Aftert + γBankObsbt + ηMacroObst + εibt .(7)

I PersonObsit : in-house credit rating, number of credit cards, and other outstandingdebt.

I BankObsbt : capital, liquidity, NPL, ROA, loan-to-asset ratios, and log real assets.I MacroObst : interest rates, IPI, CPI, REER.I LTVit ∗ Aftert accounts for the way LTV levels affect the dependent variable

differently after the policy change.I PersonObsit ∗ Aftert allows for different returns to personal observables.


Results - OLS

Table 18: OLS Estimates Jan 2011 - April 2017

(1) (2) (3) (4) (5) (6)

LTV 0.012*** 0.012*** 0.025*** 0.033*** 0.028*** 0.029***(0.003) (0.003) (0.004) (0.004) (0.005) (0.005)

N of CC 0.026*** 0.053*** 0.057*** 0.057*** 0.062***(0.006) (0.008) (0.008) (0.008) (0.008)

Other Debt -0.014*** -0.005*** -0.005*** -0.006***(0.002) (0.001) (0.001) (0.001)

Rating =2 4.642*** 4.730*** 4.782***(0.958) (0.959) (0.954)

=3 5.131*** 5.124*** 5.213***(1.221) (1.221) (1.216)

=4 2.042 2.095 2.254(2.854) (2.845) (2.848)

=5 2.000 1.854 1.919(1.271) (1.271) (1.266)

=6 3.886*** 3.862*** 3.926***(0.788) (0.787) (0.782)

After -0.269 -0.708(0.732) (0.747)

After*LTV 0.026** 0.024**(0.010) (0.010)

After*Other Debt 0.004***(0.001)

After*N of CC -0.036**(0.018)

Obs 102,799 102,799 90,005 70,745 70,745 70,745R-sq 0.189 0.189 0.182 0.153 0.153 0.154

Robustness


Documents

Loan-to-Value Caps, Bank Lending, and Spill-over to ... · LTV cap introduced in 2011 at 75%, and increased to 80% in 2016m9. 2011m1 2016m9 75% 80% Exposed Unexposed Constrained Borrowers