Effects of Regulating Household Loan on Korean Household
Delinquency Ratios
Ehung Gi Baek* Sangmyung University
Dong Jin Shin** National Assembly Budget Office
Abstract
This paper uses Korean data to analyze whether regulations and monetary policy have
contributed to controlling the financial fragility of household debt. The analysis shows that LTV
and DTI regulations and a low interest rate policy lowered household delinquency ratios. We
suggest that the government should lower household financial fragility and thus enhance the
macroeconomic environment by appropriately applying the policy mix of the Bank of Korea’s
monetary policy and the Financial Supervisory Service’s regulation policies. We also propose
that the government establishes a board to coordinate the policy instruments of these two
independent authorities.
JEL Classification Number E44, E61, G21
Key words: household financial fragility, household delinquency ratios, DTI regulation,
LTV regulation, lowering target interest rate
___________________________________
* First Author, Professor, Department of Economics and Finance, Sangmyung University (Tel: 82-2-2287-5190, E-mail: egbaek@smu.ac.kr)
** Address correspondence to Dr. Dong Jin Shin, Economic Analysis Division, National Assembly Budget Office, 1Uisadang-ro (Yeouido-dong)
Yeongdeungpo-gu Seoul, 150-010, Republic of Korea, Tel: 82-2-788-4654, FAX: 82-2-788-4693, E-mail: shindj@nabo.go.kr
I. Introduction
The U.S. sub-prime mortgage crisis triggered the global financial crisis which happened in
September 16, 2008. At that time, Lehman Brothers’ filed for bankruptcy protection. Thereafter,
the Korean economy was also negatively impacted by the global financial crisis. Its cause was
the declining level of inflation since 2001. In addition, the response of the FRB to the bursting of
the tech bubble at the end of 2000 resulted in excessive global liquidity growth, especially from
2003 to the end of 2006. Therefore, long-term risk-free nominal and real interest rates had been
extraordinarily low since 2003 and hence there was also an explosion of leverage in the nonfinancial sector (see Buiter, 2007). Households leveraged up and so did the financial sector,
namely the US home loan market, where regulatory and supervisory failure happened.
To prevent negative impact from the global financial crisis, governments in most mature
economies prepared a financial stabilization plan to provide liquidity. The financial stabilization
plan in Korea was implemented on December 19, 2008, in which the government guaranteed
payments for foreign liabilities by financial institutions to stabilize the exchange rate. In addition,
the Bank of Korea lowered target interest rates and provided liquidity to reduce uncertainty in
the financial market.
Households in most mature economies facing the global financial crisis started to reduce
consumption in order to alleviate their debt. However, in the case of Korea, the amount of the
real estate mortgages increased, even though there was negative economic growth.1 McKinsey
reported in 2010 that household leverage (measured as the ratio of household debt to disposable
income) in the United States, United Kingdom, Spain, Canada, and South Korea is at historic
peaks. Hence, McKinsey classified these households as having a high likelihood of deleveraging
(see McKinsey Global Institute, 2010).
During the financial crisis in the US the real estate bubble burst and therefore, household
delinquency ratios rose radically,2 while the decline of household delinquency ratios in Korea
1
2
See <Table 1>.
The US household delinquency ratio regarding mortgage loans rose from 1.01% in the second quarter of 2005 to
2.0% in the fourth quarter of 2007. It increased to 3.0% in the third quarter of 2008 and finally reached 4.6% in the
fourth quarter of 2009. However, it declined to 4.4% in the third quarter of 2010.
1
started from 0.96% in the second quarter of 2006 to 0.57% in the second quarter of 2010.3 This
phenomenon should imply that not only the ratio of the loan to value (hereinafter LTV), but also
the ratio of the debt to income (hereinafter DTI)4 contributed to the control of household
financial fragility.5
Also, the declining level of household delinquency ratios implied that the lowering of target
interest rates by the Bank of Korea contributed to the soundness of household debts. Therefore,
this paper analyzes whether the regulations of Financial Supervisory Service and the lowering of
target interest rates by the Bank of Korea actually contributed to controlling household financial
fragility using Korean data. These analyses resulted recommendations for the policies of the
regulatory authority and the monetary authority to strengthen the soundness of household debts.
The paper proceeds as follows. After the introduction in section I, section II describes
household loans, delinquency ratios and repayment capacity. Section III gives a theoretical
model for household delinquency ratios in which household debt depends on demographic
factors, the future income, real interest rates, and so forth, according to the life-cycle model. In
section IV, the model for household delinquency ratios is used to illustrate the effect of
regulating household loans and the lowering of target interest rates on household delinquency
ratios. It gives the empirical results. Derived from these results, policy implications are discussed
in section V. Also, section VI provides conclusions.
II. Household loans, delinquency ratios and repayment capacity
The total amount of household loans reached 725 trillion won in the third quarter of 2010. Its
size expanded by 77 trillion won as compared with the fourth quarter of 2008 (648 trillion won)
in which the global financial crisis happened. The percentage change of household loans
decreased from 10.3% in the second quarter of 2008 to 5.9% in the third quarter of 2009. The
3
4
5
The Korean household delinquency ratio rose from the third quarter of 2008 to the first quarter of 2009 during the
global financial crisis. It declined to the pre-crisis level thereafter.
The loan-to-value (LTV) ratio expresses the amount of a first mortgage lien as a percentage of total mortgaged
property’s appraised value. And the debt-to-income (DTI) ratio is the percentage of a consumer’s annual gross
income that goes toward paying financial debts.
Bernanke (2010) believes that regulatory policy measures are better than monetary policy measures, in order to
prevent immobile bubbles. Shin (2010) also argues that regulatory measures on bank lending such as LTV ratios
and DTI ratios may be important for the macro-prudential policy framework.
2
trend for household loans reversed in 2010. Thus, the percentage change for household loans was
7.6% in the second quarter of 2010 and 7.3% in the third quarter of the same year. The percent
change in mortgage loans increased from 2.7% in the first quarter of 2008 (during the housing
recession) to 10.9% in the second quarter of 2009. In contrast, the percent change decreased from
10.9% in the second quarter of 2009 to 6.8% in the third quarter of 2010. Finally, the amount of
mortgage loans reached 277 trillion won in the third quarter of 2010. It grew by 38 trillion won
as compared to 239 trillion won in the fourth quarter of 2008.
<Table 1> Household loans
(Unit: Trillion won, %)
2008
1/4
2008
2/4
2008
3/4
2008
4/4
2009
1/4
2009
2/4
2009
3/4
2009
4/4
2010
1/4
2010
2/4
2010
3/4
5.3
4.2
3.2
-3.2
-4
-2
1.2
6.3
8.4
7.3
4.4
Household loans
604
(9.0)
622
(10.3)
637
(10.2)
648
(8.9)
647
(7.1)
661
(6.2)
675
(5.9)
691
(6.7)
696
(7.5)
711
(7.6)
725
(7.3)
DMBs’ housing
loans
244
(0.5)
248
(2.4)
251
(3.3)
254
(3.7)
260
(6.4)
266
(7.1)
270
(7.2)
273
(7.4)
275
(5.8)
280
(5.3)
283
(4.9)
DMBs’ home
mortgage loans
224
(2.7)
229
(5.3)
234
(7.2)
239
247
254
259
264
(8.1) (10.3) (10.9) (10.5) (10.2)
267
(8.1)
273
(7.4)
277
(6.8)
GDP growth
Notice: DMBs are deposit monetary banks. Year-on-year growth rate is in parentheses.
Source: Bank of Korea
The increasing rate of household loans started to slow down in the third quarter of 2002 as
shown in Figure 1.
<Figure 1> Growth rate of household loans and real lending interest rate
Source: Bank of Korea
3
This phenomenon might have been caused by the policy of LTV regulation in the third quarter
of 2002 (Table 2). The Financial Supervisory Service regulated not only LTV in September 2002,
but also DTI in August 2005. However, the regulatory authority eased the regulations in
November 2008, since the global financial crisis happened. At last, the authority regulated LTV
again in July 2009 and DTI in September 2009 in order to reduce the high risk of asset inflation
due to the excessive liquidity caused by the stabilization plan for the economy recovery.
However, the Korean government decided to temporarily ease the DTI regulation from August
2010 to March 2011 in order to stimulate the housing market.
<Table 2> History of regulation on home mortgage loans
Date
Contents
Applied companies
(LTV introduction) nationwide 60%
Banks, Insurances
2002
9.4
2003
6
(LTV strengthening) speculative area, 3 years maturity,
60% → 50%
Banks, Insurances
10.29
(LTV strengthening) speculative area, 3 years maturity,
50% → 40%
Banks, Insurances
6.30
(LTV strengthening) speculative area, 10 years maturity,
over 600 million won mortgage, 60% → 40%
Banks, Insurances
8.31
(DTI introduction) speculative area, single borrower under 30
All financial companies
years old, speculative area, a spouse who has a home loan, 40%
3.30
(DTI strengthening) speculative area, over 600 million won
mortgage, 40%
All financial companies
11.15
(LTV strengthening) speculative area, over 600 million won
mortgage, 50%
Savings banks,
Specialized financial Ltd.
2
(DTI strengthening), under 600 million won mortgage,
applied to DTI 40~60%
Banks
8
(DTI strengthening) under 600 million won mortgage,
applied to DTI 40~70%
Insurances,
Saving
banks,
Specialized
financial Ltd.
2005
2006
2007
2008
11.3
(Off speculative areas) from 7. November, except Gangnam 3 Gu All financial companies
2009
7.6
(Metropolitan LTV strengthening) over 600 million won
mortgage, applied to LTV 60% → 50%
9.7
(Metropolitan DTI strengthening) Seoul 50%, Metropolitan 60% Banks
8.29
When homeless buy a house (except speculative arrear, under
All financial companies
600 million won mortgage), deregulation of DTI to March 2011
2010
Banks
Source: Financial Supervisory Service, Ministry of Strategy and Finance
4
Figure 2 shows that the delinquency ratios of household loans held the 2% level between the
fourth quarter of 2002 and the first quarter of 2005 and then decreased gradually. This resulted
from the effect of LTV and DTI regulations.
<Figure 2> Household delinquency ratios and home mortgage loan regulations
Source: Bank of Korea
Meanwhile, the Bank of Korea lowered the target interest rates from the fourth quarter of
2008 to the second quarter of 2009 in order to prevent the negative impact from the global
financial crisis on the Korean economy. That period is shown in the unshaded area in Figure 2.
After Lehman Brothers Holdings filed for bankruptcy protection in September 2008, the Bank of
Korea lowered the target interest rate by 325 base points from 5.25% in October 2008 to
February 2009.6 Therefore, the household delinquency ratios remained below 1%, even though
the global financial crisis negatively affected the Korean economy.
However, the ratio of household debt7 to disposable income8 rose sharply from 0.56 in the
first quarter of 2000 to 0.93 in the third quarter of 2002. The introduction of LTV regulation
began to reduce the rate of household loans in the fourth quarter of 2002 as shown in Figure 1,
even though the ratio of household debt to disposable income increased gradually. The easing of
6
7
8
The Bank of Korea lowered the target interest rate by 325 base points during this period; 0.5%p in twice in
October 2008, 0.25%p in November, 1.0%p in December, 0.5%p in January and February 2009. Thereafter, the
central bank raised the target interest rate by 0.25%p 3 times respectively. The increases were carried out in July,
November 2010 and January 2011.
Household debt data is the same one as the household loans in Table1 and Figure 1.
Quarterly disposable income is reconstructed in which annual individual disposable income is distributed
according to the proportion of quarterly GNI.
5
the regulations according to the global financial crisis in 2008 increased remarkably the ratio of
household debt to disposable income. The ratio of household debt to disposable income rose
from 1.06 in the second quarter of 2008 to 1.14 in the first quarter of 2009. Thereafter, the ratio
dropped somewhat until the end of 2009.
The Korean household tended to increase mortgage loans during the global financial crisis as
remarked above. Therefore, we need to investigate the trend of household debt repayment
capacity in order to comprehend credit risk caused by increasing household debt. The ratio of
household debt to disposable income started to increase from 114% in 2004 and reached 143% in
2009. This means that the repayment ability of households declined continuously.
In addition, the ratio of personal interest payments to disposable income increased from 3.1%
in 2004 to 7.5% in 2008 and decreased to 7.3% in 2009. However, its increasing rate kept the
level above 7% after 2006. This means that the debt repayment ability of households declined.
Additionally, the ratio of the personal financial debt to nominal GDP increased from 65.7% in
2004 to 80.4% in 2009. This also means the debt repayment ability of households declined.
<Table 3> Personal household debt repayment capacity
2003
2004
2005
2006
2007
2008
2009
Financial debt to disposable income
118
114
120
129
136
139
143
Interest payments to disposable income
4.5
3.1
5.4
6.3
7.4
7.5
7.3
Financial debt to nominal GDP
67.9
65.7
69.6
73.8
76.3
78.2
80.4
Source: Bank of Korea
III. Theoretical Model for Household Delinquency Ratios
1. Related literature
Kwak and Kim (2006) chose the proportion of delinquent borrowers to economically active
population instead of the consumer default probability as a dependent variable in their model.
The explanatory variables were the growth rate of GDP, the ratio of household debt to GDP and
the household credit conditions index. Chun et al. (2008) used the rate of consumer default as a
substitute for total default rate. Also, they used the economic lagging index, the unemployment
6
rate and the consumer price index as explanatory variables for the model of the consumer default
probability. Kim et al. (2009) used the consumer default rate as a dependent variable and added
household debt repayment ability index to estimate a model. As aforementioned, they estimated
the credit risk on a financial system which might be caused by a crisis.
Davis (1992) set up the consumer default model through research on individual financial
vulnerabilities of the United States, United Kingdom, Germany, Canada, and Japan. The
dependent variable of that model was the individual consumer default rate, while the explanatory
variables were the ratio of household debt to disposable income, the ratio of assets to disposable
income, the unemployment rate and the interest rate. Lawrance (1995) added the consumer
default probability as an option to the life-cycle income model to explain how the possibility of
default influences the level of consumption, its sensitivity to income, and the type of borrowing
constraints likely to emerge. Thereafter, Rinaldi and Sanchis-Arellano (2006) set up a model for
household delinquency ratios, using Lawrance’s simple two-period life-cycle model with a
default option. They used non-performing loans as a dependent variable. In addition, the ratio of
household debt to disposable income, the ratio of financial assets to disposable income, the real
interest rate for household loans, the unemployment rate and the inflation rate were independent
variables. They analyzed the household debt sustainability of European countries. Handale et al.
(2007) estimated a model of the consumer default probability using borrowers’ repayment ability
index in the financial stability report of the Bank of England. The delinquency ratios of mortgage
loans was used as a dependent variable, while the housing price, the ratio of mortgage loans to
GDP and the unemployment rate were used as explanatory variables in their model.
However, there is no study which considers the effects of lowering interest rates, which was
enforced to overcome consumer defaults and to prevent the negative effects of the global
financial crisis. Also, there is no study which considers the effects of LTV and DTI regulation,
which were enforced to stabilize soaring housing prices. Therefore, it is meaningful that this
study analyzes the effects of regulations reinforcement and lowering interest rate policy by using
a model for household delinquency ratios.
2. A theoretical model
7
Ando and Modigliani's (1963) life-cycle model shows that households try to maintain their
marginal utility of consumption constantly in their lifetime under their budget constraints. In
spite of different income levels for all periods, it is only possible for consumers to maximize
their lifetime utility by smoothing their consumption under the assumption of no borrowing
constraint (Hall, 1978). On the other hand, some studies, which are cases of setting the default
risk as an exogenous variable, introduced borrowing constraints.9 Lawrance (1995) applied
consumer default probability into the model as exogenous variables instead of using credit
constraints. Rinaldi and Sanchis-Arellano (2006) used the basic set up of Lawrance (1995),
namely a simple two-period life-cycle model with a default option. Consumers maximize
expected lifetime utility with preferences for consumption described by:
V (C1 , C2 )  U (C1 ) 
1
E[U (C2 )]
1 
(1)
where C is the consumption in period i , and  is the subjective rate of time preference. E[]
is the expectations operator, conditional on information available in period one. U is the oneperiod, constant relative risk aversion utility function with properties that U   0 , U   0 , and
U (0) is infinite. Consumption in period two is uncertain because labor income in period two is
uncertain. In this model, Lawrance (1995) assumes a stochastic process: with exogenous
Y
probability q , period-two income equals L , a low level of total income, while with
probability 1  q , period-two income equals YH , a high level of total income. Consumers borrow
and lend freely at a risk-free rate R . It is assumed that a saver gives up x1 units of 1st period
consumption in return for x2 units of additional consumption in period two, where x2 equals
(1  R) x1 . In a way similar to savers, borrowers give up x1 units of consumption in period one
in return for x2 units of additional consumption in period two with x2  (1  R) x1 . Since
consumption in period two is not certain (because of the uncertain income), consumers maximize
their intertemporal expected utility (2):
V ( x1 , x2 )  U (Y1  x1 ) 
9
1
[qU (YL  x2 )  (1  q)U (YH  x2 )]
1 
(2)
See Tobin (1972), Dolde (1978), and Hubbard and Judd (1986).
8
subject to the budget constraint: x2  (1  R) x1 .
At the optimum, the consumer’s marginal rate of substitution equals 1  R :
MRS 
U (Y1  x1 )
(1   )  1  R
qU (YL  x2 )  (1  q )U (YH  x2 )
with borrowers characterized by x1  0 , x2  0 and savers by
(3)
x1  0 , x2  0 .
That is to say, in equation (2) x1 represents the amount lent ( x1  0) or borrowed ( x1  0)
in period one, and x2 is the amount received as repayment ( x2  0) or given as repayment
( x2  0) in period two. x2 will depend on the market real interest rate( R ) with a perfect capital
market. However banks are only willing to lend freely at the riskless rate, R , when they face no
risk of default.10 Therefore, when consumer's default risk is applied to the model, borrower's
intertemporal trade-off and the loans would change. If the default happens in a low income state,
a borrower who accepts x1 in period one has to give up x2 units of period-two consumption
only with probability 1  q .
MRS 
(1   )U (Y1  x1 )
 1 r
(1  q)U (YH  x2 )
(4)
x1  0 , x2  0 , where MRS B is marginal rate of substitution for borrowing.
The marginal utility of lower income state cannot affect the quantity of individual loan in the
equation (4). Consumption in the low income state is fixed to YL after the loan has been decided.
Even if the quantity of individual loan is increased, the consumption is not changed in the event
of default.
Rinaldi and Sanchis-Arellano (2006) assumed that some part of the loan is invested in real or
financial assets. Likewise in this study, we assume that the loan will be invested in real assets,
namely houses. I is the value of the investment in each period, hence the exogenous
10
Lawrance(1995) assumes that in case of default the bank can only claim income in excess of YL . In other words,
under the condition that each borrower has q probability of receiving YL in period two, banks have a q percent
chance not to receive repayment. Hence, banks will not lend at the risk-free rate. They will charge a competitive
borrowing rate at which the expected profits equals zero. Therefore, 1  R equals (1  R)(1  rp) , where rp is
the risk premium charged by banks and depends on the collateral provided, the probability of default and market
conditions in general. Namely, applying the default option to the model implies that the borrowing rate exceeds
the risk-free rate.
9
probability of default depends on the outcome of income and of net wealth together. If default
occurs, banks claim the income in excess of YL and the real assets. Even if Lawrance assumes
that when in period one people borrow only for consumption purposes, Rinaldi and SanchisArellano extend the model to include the possibility that people also borrow to invest in real
assets. Since I takes the form of housing investment, the return is the real house price. Hence,
the interest charged on the loan shall be different from the expected return of real investment.
Therefore, if the default occurs, borrowers maximize expected utility (5) subject to the budget
constraint:
V ( x1 , x2 )  U (Y1  I1  x1 ) 
1
[qU (YL )  (1  q)U (Y  I ) H  x2 )]
1 
(5)
subject to the budget constraint: x2  (1  r ) x1 .
Equation (5) shows that it is a two-period model. It is assumed that the whole wealth available
in period two has to be consumed then. I1 is the investment to real assets in period one. I1
should be subtracted from Y1 because the income that is not consumed in period one is not
providing any utility in period one. I 2 is the market value in real terms of this investment in
period two, which will depend on the return on the investment. Under the condition that the
whole wealth is consumed, the market value of this investment in period two I 2 will enter the
utility function. The income as return on the investment would enter as income in period two Y2
and the revaluation of the market value will be included in the term I 2 . Both the income and the
market value are state dependent. Again, two states are assumed. One is in good times when
there is
no default ((Y  I ) H   x2 ) .
The other is
in bad times
when
default
occurs ((Y  I ) L   x2 ) . The borrower will earn a high level of income during the good times in
period two, which will be bigger than the refund
On the other hand, the borrower will earn a low level of income during the bad times in period
two, which will not be enough to repay the loan. In the equation (6), with imperfect capital
markets, x2 is the amount to repay for the loan x1 at the real lending interest rate r charged
in this model. Equation (6) shows the 1st order condition of the optimization process for the
10
borrower. This is characterized by the desired loan size of a borrower facing a q percent
probability of default:
MRS B 
(1   )U (Y1  I1  x1 )
 1 r
(1  q)U [(Y  I ) H  x2 ]
(6)
Extracting the probability of default q , from equation (6), we will get equation (7):
q
(1  r )U [(Y  I ) H  x2 ]  U (Y1  I1  x1 )(1   )
(1  r )U [(Y  I ) H  x2 ]
(7)
recalling that (1  r )  (1  R)(1  rp) , and then x1  0 , x2  0 .
Equation (7) shows that the household delinquency ratio q depends on the amount of
household loan taken, x1 , on current income, Y1 , on investment of the first period, I1 . Also the
probability of default depends on the bank lending rate, r , and uncertain future income and
wealth, which hang on the possibility of unemployment and on the development of housing
prices (Rinaldi and Sanchis-Arellano, 2006). Finally, the probability of household default
depends on the time preference,  , which is the expected inflation rate of individuals. There are
five differences between the preliminary study of Rinaldi and Sanchis-Arellano (2006) and this
study. First, real assets are used in this model. These days, most people specify houses as durable
goods.11 According to "Survey of Household Finances in 2010" from Statistics Korea (2011),
the proportion of real estate to total assets is 75.8%. Among them, primary residence to total
assets reaches 42.4%. Therefore, it might be appropriate to explicitly introduce total real housing
assets into the model. Second, this model used total real housing assets against real disposable
income unlike existing models used housing price index. Third, the ratio of house owners is not
used in this study because the data for Korea is not adequate. Fourth, the ratio of real disposable
income to economically active population is added except for the ratio of real disposable income
to the number of households to default rate caused by aging of the population. Fifth, the target
rate of the Bank of Korea is used as a control variable to analyze the effect of lowering target
interest rates. After all, this study aims to analyze the effects of regulating household loans and
lowering target interest rates on household delinquency ratios using Rinaldi and Sanchis-
11
A reason that consumption and debt levels are empirically not perfectly explained by the lifecycle hypothesis is
that a part of households’ consumption is spent on durable goods such as housing (Philbrick and Gustafsson, 2010).
11
Arellano’s model, which is research about the sustainability of European countries' household
debt.
IV. Empirical Analysis
1. Data and variables
The data of quarterly disposable income, which are used in the model, are based on the annual
data from the Bank of Korea. The data set is drawn up by distributing the annual disposable
income in proportion to seasonal adjusted quarterly GNI. Individual household debt in the flow
of funds table from the Bank of Korea can be used in the model.
<Table 4> Variables
Name12
Arrears
Variable
Delinquency ratio for the household debt of deposit money banks (%)
Source
FSS
Debtdisincome
Household debt to disposable income
BOK
Realdisincomehh Real disposable income per households (bill. Won)
SK
Realdisincomepop Real disposable income to economically active population (bill. won)
SK
Debtrealasset
Household debt to total real housing assets
MOL
Rinterest
Real home lending interest rate, rate per annum
BOK
Unemployment
Number of unemployed to economically active population
SK
Inflation
Rate of change in CPI from the same quarter of the previous year
BOK
Note: FSS(Financial Supervisory Service), BOK(Bank of Korea), MOL(Ministry of Land), SK(Statistics Korea)
However, the data are not available to be used because the standard of the data has been
changed into SNA of 1993 since the fourth quarter of 2005. According to the consistency
problem, the data on household debt of the deposit bank are used in this study. Furthermore, the
delinquency ratio of the household debt as the dependent variable, which is used in the model, is
based on data from the deposit bank. As Chun et al. (2008) did, we use the delinquency ratio as a
proxy variable for the probability of household default to consider the size of the bank’s credit
risk at a certain period. In other words, there is a positive correlation between the delinquency
ratio of household loans and household financial vulnerability.
12
Variable names are written in italics.
12
Trends for the variables in Table 4 are shown in Figure 3. The analysis period is from the first
quarter of 2000 until the second quarter of 2009. The Korean government withdrew regulations
to stimulate the real estate market in the fourth quarter of 2009 because they restored regulation
in the third quarter of 2009 to prevent real asset inflation. Therefore, we limited the analysis
period to the second quarter of 2009. The limitation of the analysis period avoids the overlapping
effects of the lowering target interest rate policy and the regulation policy. Finally, we analyze
how both of the two policies independently affected the delinquency ratio of household debt.
<Figure 3> Trends of variables
ARREARS
DEBTDISINCOME
1.5
REALDISINCOMEPOP
5.8
-.56
5.6
1.0
-.60
5.4
0.5
-.64
5.2
5.0
0.0
-.68
4.8
-0.5
-.72
4.6
-1.0
4.4
00
01
02
03
04
05
06
07
08
09
-.76
00
01
02
REALDISINCOMEHH
03
04
05
06
07
08
09
00
01
02
DEBTREALASSET
-.20
03
04
05
06
07
08
09
07
08
09
UNEMPLOYMENT
4.2
4.8
4.0
4.4
3.8
4.0
3.6
3.6
3.4
3.2
-.22
-.24
-.26
-.28
-.30
-.32
-.34
3.2
00
01
02
03
04
05
06
07
08
09
2.8
00
01
02
03
RINTEREST
04
05
06
07
08
09
06
07
08
09
00
01
02
03
04
05
06
INFLATION
14
6
12
5
10
4
8
3
6
2
4
1
00
01
02
03
04
05
06
07
08
09
00
01
02
03
04
05
Notice: Debtdisincome, Realdisincomepop, Realdisincomehh, Debtrealasset are multiplied by 100.
The above four variables and Arrears are all logarithmic values.
2. An empirical model for the delinquency ratio
The delinquency ratio for household loans is a function of the rate of household debt to
disposable income, the ratio of real disposable income to economically active population, the
ratio of real disposable income to number of households, and the real lending interest rate. The
13
delinquency ratio as a proxy variable for the probability of consumer default depends on the
inflation rate, the unemployment rate and the ratio of household debt to total amount of housing
assets as we looked at in the previous section. An empirical model of the delinquency ratio for
household debt extracted from the theoretical model is like the equation below.
 ln A  0  1 ln X1,t 3   2 D1 ln X1,t 3  3 ln X 2,t  4 X 3,t  5 D2 X 3,t  6 X 4,t
 7 X 5,t 1  8  ln X 6,t 2  9 D3 ln X 6,t 2  10 X 7  11 D1  12 D2  13 D3   t
(8)
A : the delinquency ratio for household debt (Arrears)
X1 : the ratio of household debt to disposable income (Debtdisincome)
X 2 : the ratio of real disposable income to economically active population
(Realdisincomepop)
X 3 : real home lending interest rate (Rinterest)
X 4 : inflation rate (Inflation)
X 5 : unemployment rate (Unemployment)
X 6 : the ratio of household debt to total value of housing asset (Debtrealasset)
X 7 : base interest rate of the Bank of Korea (Target)
D1 : DTI regulation dummy (3rd quarter of 2005~3rd quarter of 2008, D1 =1, otherwise,
D1 =0)
D2 : FSP (Financial Stabilization Policy) dummy (4th quarter of 2008~2nd quarter of 2009,
D2 =1, otherwise D2 =0)
D3 : LTV regulation dummy (3rd quarter of 2002~3rd quarter of 2008, D3 =1, otherwise
D3 =0)
0 , 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 9 , 10 , 11 , 12 and 13 are regression coefficients.
 is a residual.
Not only the percentage of Arrears as a dependent variable, but also the percentage of
Debtdisincome and Debtrealasset are logarithmically differentiated. In addition, both
Realdisincomepop and Realdisincomehh are multiplied by 100 and are also logarithmically
14
differentiated. We introduce dummy variables to analyze how DTI and LTV regulations
influence Arrears in order to prevent household financial vulnerability caused by expanding
household loans. Introducing a dummy variable as a quality variable can be used to assess the
effectiveness of policies. In addition, we analyze how DTI regulation affects Arrears using a
dummy variable for the period between the third quarter of 2002 and the third quarter of 2008.
The dummy variable multiplies Debtdisincome which reflects on the slope of Debtdisincome. In
a similar way, we analyze how LTV regulation influences Arrears by introducing a dummy
variable for the period between the third quarter of 2005 and the third quarter of 2008. The
variable multiplies Debtrealasset which reflects on the slope of Debtrealasset. In this way, we
can assess how LTV regulation affected Arrears by limiting the size of loans when household
debt grows fast. In addition, this study analyzes how the lowering target interest rate policy of
the Bank of Korea affects Arrears introducing a FSP dummy variable which multiplies the slope
of real home lending interest rate. In other words, we analyzed how Rinterest controls the rising
of household delinquency ratios caused by the credit crunch during the global financial crisis.
However, it is not clear whether the FSP dummy variable or the global financial crisis affect
Arrears since the global financial crisis and the financial stabilization plan happened during the
same period. Therefore, we added the target interest rates of the Bank of Korea as a control
variable. It aimed to examine whether lowering target interest rate reduced the real home lending
interest rate to control the household delinquency ratios.
We added dummy variables to examine how DTV, LTV and FSP influenced not only the
slope but also the intercept of the dependent variable. This helps us to analyze whether the
financial stabilization effect happened or not. In other words, we analyzed whether the
strengthening of regulation policy or the easing of monetary policy actually reduced the
household delinquency ratios or not. However, the household delinquency ratios can also
influence the independent variables. In such a case, we have to solve an endogeneity problem by
using instrument variables. However, it is difficult to find proper instrument variables. Hence,
this study solves an endogeneity problem by introducing time lags to the data. Since
Debtdisincome and Debtralasset influence the household delinquency ratios after 2 quarters in
cross-correlation analyses, we adjust time lags to solve an endogeneity.
15
3. Estimation results
We tested the stationarity of variables of the model. We could not reject the null hypothesis of
a unit root for all the variables except inflation.13 We also tested stationarity of these variables
which were first order differentiated. The results showed us that they are stationary by rejecting
the null hypothesis of a unit root. Therefore, all the first order differentiated variables except for
inflation rate and the target interest rate are used for regression analysis in this section.
The regression result of Equation (8) is reported in Table 5. Since the adjusted R-squared for
model 1 and model 2 are each 0.685 and 0.680, the model for the delinquency ratios of the
household debt has persuasive power. Also, the null hypothesis that all the coefficients of both
models are 0 is rejected at the 1% level of significance. Hence, all the coefficients appear to be
significant. In addition, the result of model 2 that contains real disposable income per household
is that same as the result of model 1 that contains real disposable income per economically active
population. Therefore, since the regression result of Equation (8) is not changed regardless of
variable used, the result of the model appears to be robust.
Realdisincomepop in model 1 is an explanatory variable which reflects the debt repayment
ability. Rise of Realdisincomepop by 1%p leads to a drop of Arrears by 1.80%p. This result
coincides with intuition. Thus, we can suppose that an increase of real disposable income to
economically active population decreases financial vulnerability of household debt. Similarly a
rise of Realdisincomehh by 1%p in model 2 leads to drop of Arrears by 1.76%p. Hence, it
appears that an increase of real disposable income per household reduces the risk to the financial
system through the rising of household real disposable income.
The t-value of the inflation rate which indicates individual time preference related household
loans is less than 1.69 in both models. It appears that the inflation rate is not a significant
variable for the household delinquency ratios. However, the unemployment rate of the previous
quarter plays a significant role for an index of uncertainty of future income. A rise of
unemployment rate by 1%p increases the delinquency ratios by 0.09%p in the model 1 and
0.11%p in the model 2. An increase in the unemployment rate does not affect the present quarter
13
There were many of studies whether the nominal interest rate or the inflation rate is stationary. Although many of
the studies reported that both of the variables are stationary, some of studies suggested that only the inflation rate
is stationary (Yoon, 2005). This paper assumes that the inflation rate has a order of integration I(0), while the
real interest rate has a order of integration I(1) as the study of Rose (1998).
16
rather the next quarter, because households that lose a job in the 3 months have a problem with
debt repayment and interest payment ability. Thus, the rise of the unemployment rate as a proxy
variable for uncertainty of future income raises the financial vulnerability of household debt.
<Table 5> Estimation results for household delinquency ratios
ΔlnArrears
2001 Q1 ~ 2009 Q2
Model 1
C
ΔlnDebtdisincome(-3)
DTI*ΔlnDebtdisincome(-3)
ΔlnRealdisincomepop
ΔRinterest
FSP*ΔRinterest
Inflation
Model 2
-0.655***
(-3.251)
1.544***
(3.265)
-2.193**
(-2.360)
-1.797**
(-2.022)
0.0932*
(1.767)
-0.126**
(-2.597)
0.0034
(0.247)
C
ΔlnDebtdisincome(-3)
DTI*ΔlnDebtdisincome(-3)
ΔlnRealdisincomehh
ΔRinterest
FSP*ΔRinterest
Inflation
-0.649***
(-3.247)
1.441**
(3.101)
-2.102**
(-2.325)
-1.757**
(-2.147)
0.097*
(1.993)
-0.124**
(-2.720)
0.002
(0.154)
ΔUnemployment(-1)
0.093*
(1.728)
ΔUnemployment(-1)
0.109**
(2.053)
ΔlnDebtrealasset(-2)
4.084**
(2.362)
ΔlnDebtrealasset(-2)
4.312**
(2.370)
-4.587**
LTV*ΔlnDebtrealasset(-2)
(-2.137)
0.054**
Target
Target
(2.445)
-0.096**
DTI
DTI
(-2.683)
0.487***
FSP
FSP
(2.984)
0.433**
LTV
LTV
(2.790)
Adjusted R-squared
0.685
Adjusted R-squared
DW
2.526
DW
F-statistic
6.530
F-statistic
Prob(F-statistic)
0.0001
Prob(F-statistic)
Note : ***, **, * indicate statistical significance at 1%, 5%, 10% levels, respectively.
t-values are in parentheses. ‘Δln’ means log difference and ‘Δ’ means difference.
LTV*ΔlnDebtrealasset(-2)
-4.836**
(-2.157)
0.005**
(2.467)
-0.101**
(-2.858)
0.485***
(2.961)
0.437**
(2.761)
0.680
2.624
6.386
0.0001
17
On the other hand, a rise in Rinterest causes financial vulnerability. This result accords with
the theoretical model. Rising Rinterest by 1%p forced Arrears up by 0.09%p before the FSP
policy period. However, when Rinterest rose by 1%p during the FSP period, Arrears fell by
0.04%p (=0.09% – 0.13%) due to the lowering of the target interest rate of the Bank of Korea.
A rise in Debtdisincome means that household debt increases faster than disposable income.
As one would expect, the result of empirical analysis shows that a rise in Debtdisincome induces
the financial vulnerability of household debt to increase. In other words, if Debtdisincome
increases by 1%p, Arrears also increases 1.54%p. Also, Debtrealasset is another significant
variable for the delinquency ratios. If it increases by 1%p, Arrears also increases by 4.08%p.
Therefore, Arrears will be controlled even if the ratio of household debt to disposable income
increases if DTI regulation is enforced. If the ratio of household debt to real disposable income
increases by 1%p, Arrears decreases by 0.65%p (= 1.54% – 2.19%) under the condition that DTI
regulation is enforced. On the other hand, if the ratio of household debt to the total value of
housing assets increases by 1%p, Arrears decreases by 0.51%p (=4.08%-4.59%) while LTV
regulation is enforced with DTI regulation.
V. Policy Discussion
Phelps (2010), at the conference “Post-Crisis Economic Policies” (on 11-12 December 2009
in Berlin), argued that a central bank should be responsible for macro-prudential regulation, and
stability concerns should influence monetary policy, even though the central bank should not be
institutional supervisor. In that context, Catte et al. (2011) simulated a model to answer the
question of whether US financial supervision or the use of a macro-prudential policy instrument
could have helped to prevent the housing bubble, if they would have been stricter. The results of
this simulation show that tighter US monetary policy and additional credit restraint resulting
from an aggressive use of macro-prudential policy tools could have dampened the housing price
boom. However, this study does not specify a role for housing LTV ratios. Hence, the model has
a limitation, since the regulations on the household loan market such as LTV ratios and DTI
ratios may be relevant for a macro-prudential policy framework (Shin, 2010). Additionally,
Bernanke (2010) suggested that the best response to the housing bubble would have been
18
regulatory, not monetary. Therefore, we need to analyze the effect of monetary and regulatory
policies on the financial fragility of household debt.
1. Effect of monetary policy
According to the results of this study, a 1%p rise in the real lending interest rate of
household debt increased the delinquency ratios by 0.09%p. However, increase of the real
lending interest rate by 1%p during the FSP period reduced the delinquency ratios by 0.04%p.
Therefore, the lowering of target interest rates contributed to reducing levels of financial
fragility during the global financial crisis. Thus, we concluded that the monetary authorities
could effectively control the rise of Arrears. In other words, lowering the target interest rate
reduces financial vulnerability through the rise of household debt caused by the credit crunch
and unemployment. To confirm such a financial stabilization effect, we tried to introduce the
target interest rate as a control variable. The result showed that if the target interest rate were
lowered, Arrears would have fallen. Therefore, it appears that the monetary authorities could
effectively control the rise of Arrears by lowering the target interest rate. However, in the FSP
period Arrears rose because the period of FSP is same as the period of the global financial
crisis. In other words, the rise in delinquency ratios was not avoidable because the global
financial crisis influenced the Korean economy negatively in spite of the financial
stabilization policy.
2. Effect of DTI and LTV regulations
A 1%p rise in the ratio of household debt to household disposable income would have
increased the delinquency ratios 1.54%p from the first quarter of 2001 to the third quarter of
2008, if there were no DTI or LTV regulations. In addition, a 1%p rise in the ratio of household
debt to total value of housing assets would have raised the delinquency ratios of household debt
by 4.08%p in the same period. However, the rise in the delinquency ratios was restrained, even
though the ratio of household debt to household disposable income increased, since there was
DTI regulation. In other words, the delinquency ratios decreased by 0.65%p due to DTI
regulation even though the ratio of household debt to household disposable income increased by
1%p. Meanwhile, the delinquency ratios decreased by 0.51%p, when the ratio of household debt
19
to total value of housing asset increased by 1%p under LTV regulation. It means that the effect
of DTI or LTV regulation on the household delinquency ratios is significant.
Table 6 reports the household debt profile by region. Total DTI, 18.5%, is not very high. DTI
for the capital area is 19.4%, and DTI of non-capital area is 17.7%. Thus, the difference in DTI
between the capital and non-capital areas is not very large. While the share of DTI exceeding
40% is 12.6% in the capital area, it is 10.9% in the non-capital area. Even though these
differences between capital and non-capital areas are not big, it appears that the policy of DTI
regulation decreases Arrears in Table 5. We reason that it results from the effect of public
announcement of DTI regulation. The public announcement by the regulatory authority might
affect the behavior of the households. In other words, the households that have less repayment
capacity (or lower income group) give up on borrowing new household loans. For instance,
Southern Seoul, which has higher income than other areas, also has higher a DTI rate. Real estate
prices of this region boomed before the global financial crisis. Thus, this is the reason why we
guess that there is an effect of the public announcement.
<Table 6> Household debt profile by region (year 2009)
Average
Share of
Debt-todebt amount Average income DTI(%) Share of Share of
debt
income
DTI>40% DTI>100%
obligors (%) (mil. Won)
ratio
Total
100
52.4
32.7
1.5
18.5
11.8
1.6
Capital area
50.9
63.7
33.2
1.7
19.4
12.6
1.6
21.3
66.4
34.3
1.7
18.8
12.3
8.1
Northern Seoul
10.1
57.9
33.0
1.6
18.0
11.3
7.3
Southern Seoul
11.2
74.0
35.5
1.9
19.4
13.1
1.8
49.1
40.7
30.2
1.2
17.7
10.9
1.6
Seoul
Non-capital area
Source: Hahm et al., (2010), Table 6.
Even though the effect on loan-to-value through the policy of LTV regulation reduces Arrears,
the delinquency ratios for household debt increased since the policy of LTV regulation was
enforced according to increasing household debt. However, since DTI regulation as a policy
instrument, which is stronger than LTV regulation, was enforced, Arrears started to decrease.
Finally, the regulations on the household loan market were able to stabilize the financial market.
20
Therefore, it appears that the policy of DTI and LTV regulations is available to control financial
fragility.
3. Implications
Policy instruments primarily have not only advantages but also disadvantages. The
lowering of the target interest rate is available to reduce the financial fragility of household
debt through the strengthening of the interest payment ability in the short term, but it causes
inflation in the long term. Thus, the Bank of Korea must judge whether to cut target interest
rates in order to reduce financial fragility, even though the policy causes inflation as a side
effect. This means that the Bank of Korea has to choose either macro-financial stability or
price stability as a policy goal, since financial stability as a policy goal was newly added in
the Act of the Bank of Korea on 16 November, 2011.
On the other hand, regulations such as DTI and LTV are effective to reduce the financial
fragility of household debt in the short term, but they can distort the loan market. Therefore,
the Financial Supervisory Service must judge whether to implement regulatory policies in
order to reduce the financial fragility of household debt, even if the policy may cause loan
market distortion. Hence, the results of this paper imply that the Bank of Korea and the
Financial Supervisory Service have to find an optimal policy mix which can minimize not
only financial fragility but also side effects.
Estimation results from the analyses have policy implications for both the regulatory
authority and the monetary authority. Both authorities are independent institutions and have
their own policy instruments to control the financial fragility of household debt. We
recommend that the government establishes a board in order to find an optimal policy mix
that can minimize financial fragility as well as side effects. The coordination of the two policy
instruments through the board may enhance the efficiency to control the financial fragility of
household debt.
VI. Conclusion
21
This paper, using Korean data, analyzes whether regulation policies and monetary policy have
contributed to the control of financial fragility of household debt. The analysis shows that the
rise of household debt to disposable income or the rise of real total housing assets to disposable
income increases household delinquency ratios, whereas the LTV and DTI regulations and the
low interest rate policy lowered the household delinquency ratios. In addition, the rise of real
disposable income to the economically active population decreases the ratio. Thus, the
government is expected to reduce financial fragility by preparing measures of employment,
which will increase real disposable income to the economically productive population. Hence, it
is not desirable that the government withdraws the regulations unless the household debt
repayment capacity significantly improves. In contrast, the government needs to considerably
manage the rapid expansion of household debt, since not only the personal financial debt to
disposable income, but also the personal financial debt to nominal GDP increase continuously.
We propose that the government establishes a board in order to coordinate each policy measure
of the two independent authorities. The policy coordination would enhance the efficiency of
controlling the financial fragility of household debt.
22
References
Ando, A. and F. Modigliani. (1963) “The Life Cycle” Hypotheses of Saving: Aggregate
Implications and Tests, American Economic Review, 53(1): 55-84.
Bernanke, Ben S. (2010) Monetary Policy and the Housing Bubble, Speech at the Annual
Meeting of the American Economic Association, Atlanta, Georgia, January 3.
Buiter Willem H. (2007) Lessons from the 2007 Financial Crisis, Policy Insight No. 18, Centre
for Economic Policy Research.
Catte, Pietro., Pietro, Cova., and Ignazio, Visco. (2011) The role of macroeconomic policies in
the global crisis, Journal of Policy Modeling, 33(6): 787-803.
Chun, Heung Bae, Jung-Jin Lee and Woon Youl Choi. (2008) A Study of the Stability of Korean
Banks’ Household Lending Sector: A Stress Test under a Macroeconomic Credit Risk
Model, Economic Analysis, the Bank of Korea, 14(2): 71-99.
Davis, E. P. (1992) Debt, Financial Fragility, and Systemic Risk, Clarendon Press, Oxford.
Dolde, Walter. (1978) Capital Markets and the Short-Run Behavior of Life Cycle Savers, The
Journal of Finance 33: 413-428.
EViews 7. (2009) User's Guide, Quantitative Micro Software, LLC.
Haldane, A., Hall, S. and Pezzini, S. (2007) A new approach to assessing risks to financial
stability, Financial Stability Paper no. 2, Bank of England.
Hall, R. (1978) Stochastic implications of the life cycle-permanent income hypothesis: theory
and evidence, Journal of Political Economy, December, 971-987.
Hahm, Joon-Ho, Jung In Kim and Young Sook Lee. (2010) Risk Analysis of Household Debt in
Korea: Using Micro CB Data, KDI Journal of Economic Policy, 32(4), 1-34.
Harris, R. and Sollis, R. (2003) Applied Time Series Modeling and Forecasting, Wiley,
Chichester, Great Britain.
Hubbard, R. Glenn, and Kenneth L. Judd. (1986) Liquidity Constraints, Fiscal Policy, and
Consumption, Brookings Papers on Economic Activity 1: 1-50.
Kim, Young-Sun, Jong Lim Ha and Ja Hae Kim. (2009) Measurement and Evaluation of Credit
Risk of the Financial System by Using Debtor’s Repayment Capability, Monthly Bulletin,
the Bank of Korea, December, 24-55.
23
Kwak, dong Chul and Myung-Jig Kim. (2006) Macroeconomic Stress Testing of Retail Credit
Sector, Journal of Economic Research, 27 (22): 93-112.
Lawrance Emily C. (1995) Consumer Default and the Life Cycle Model, Journal of Money,
Credit, and Banking, 27(4): 939-954.
McKinsey Global Institute. (2010) Debt and deleveraging: The global credit bubble and its
economic consequences, January.
Phelps, Edmund S. (2010) Post-Crisis economic policies, Journal of Policy Modeling, 32(5):
596-603.
Philbrick, P. and Gustafsson, L. (2010) Australian Household Debt – an empirical investigation
into the determinants of the rise in the debt-to-income ratio, Master’s Essay, Lund
University.
Rinaldi Laura and Alicia Sanchis-Arellano. (2006) Household Debt Sustainability What explains
household non-performing loans? An empirical analysis, Working Paper, No. 570, European
Central Bank.
Rose, A. K. (1988) Is the real interest rate stable? , Journal of Finance, 43: 1095-1112.
Shin, Hyun Song. (2010) Macroprudential Policies Beyond Basel Ⅲ, Policy Memo, Princeton
University.
Statistics Korea. (2011) Survey of Household Finances in 2010.
Tobin, James. (1972) Wealth, Liquidity, and the Marginal Propensity to Consume, In Human
Behavior in Economic Affairs (Essays in Honor of George S. Katona), edited by B. Strumpel,
Amsterdam: Elsevier Scientific Publishing Co. 36-56.
Yoon, Jong In. (2005) Testing the Stationarity of the Inflation Rate and Nominal Interest Rate,
Journal of Korea Money and Finance Association, 10(5): 143-164.
24