Implementing Loan-to-Value and Debt Service-To-Income Measures: A Decade of Romanian Experience

Munich Personal RePEc Archive
Implementing Loan-to-Value and Debt

Service-To-Income measures: A decade
of Romanian experience
Neagu, Florian and Tatarici, Luminita and Mihai, Irina
National Bank of Romania
2015
Online at https://mpra.ub.uni-muenchen.de/65988/
MPRA Paper No. 65988, posted 10 Nov 2015 05:53 UTC
Occasional Papers, June 2015
45
Occasional Papers
No. 15
June 2015
Note
The views expressed in this paper are those of the author and do not
necessarily reflect the views of the National Bank of Romania.
All rights reserved.

Reproduction for educational and non-commercial purposes is permitted
provided that the source is acknowledged.
ISSN 1584-0867 (online)

ISSN 1584-0867 (e-Pub)
IMPLEMENTING LOAN-TO-VALUE
AND DEBT SERVICE-TO-INCOME MEASURES:
A DECADE OF ROMANIAN EXPERIENCE1
Florian Neagu*
Luminița Tatarici*
Irina Mihai*2
1
The paper is part of Implementing Loan-to-Value and Debt-to-Income Ratios: Learning from Six Country
Experiences project led by the IMF. The countries participating in the project are: Brazil, Hong Kong SAR, Korea,
Malaysia, Poland and Romania. A summary of the findings of the project is to be found in Jacome and Mitra
(2015). The authors would like to thank Luis Jacome, Srobona Mitra, Ruy Lama, Angela Pîslaru, Alexie Alupoaiei
and the participants in the above-mentioned IMF project for their useful comments.
*2
National Bank of Romania, Financial Stability Department.
Contents
Abstract .................................................................................................................... 7
Executive Summary ................................................................................................. 9
I. Introduction ....................................................................................................... 10
II. Using the instruments ....................................................................................... 11
A. Monitoring systemic risk ............................................................................ 11
B. Taking decisions .......................................................................................... 13
C. Calibration and dealing with “leakages” ..................................................... 14
III. Evaluating effectiveness ................................................................................. 16
A. Credit dynamics .......................................................................................... 16
B. Housing price dynamics ............................................................................. 19
C. Non-performing loan (NPL) dynamics ....................................................... 19
IV. Conclusions ..................................................................................................... 21
References .............................................................................................................. 25
Annex 1 – Regulatory measures taken by the National Bank of Romania ........... 27

Annex 2 – Credit dynamics.................................................................................... 31
Annex 3 – Results of the econometric models on credit growth .......................... 33
Annex 4 – Non-performing loan dynamics ........................................................... 37
Annex 5 – Results of the econometric models on non-performing loan ratios ..... 41
Abstract
We describe an example of designing, implementing and calibrating two macroprudential
instruments – loan-to-value (LTV) and debt service-to-income (DSTI) – based on a decade of
Romanian experience with these tools. We investigate LTV and DSTI effectiveness in trimming
down excessive credit growth and in preserving the quality of banks’ loan portfolios. We find
strong links between DSTI levels and the debtors’ capacity to repay their debt, underpinning
the usefulness of caps for this instrument. We find that an approach based to a large extent on
banks’ self-regulation produces suboptimal results, exacerbating the pro-cyclicality in the system.
A one‑size‑fits‑all approach is less effective than tailoring the DSTI and LTV measures based on
debtors’ disposable income, the currency of indebtedness and the destination of the loan.
Keywords: financial stability, macroprudential instruments, house prices, credit growth, debt
service-to-income (DSTI), loan-to-value (LTV), Romania
JEL classification codes: E44, E58, G21, G28

Executive Summary
A decade of Romanian experience with debt service-to-income (DSTI) and loan-to-value (LTV)
instruments shows that they are relatively effective in: (i) curbing high credit growth and (ii) ensuring
that both debtors and creditors are able to cope with possible adverse shocks in real estate prices,
domestic currency depreciation or interest rate hikes. The second point explains why the DSTI and
LTV caps should be used both in the upswing and in the downswing of the credit cycle.
We find strong negative correlation between the level of indebtedness and the ability to repay the
debt. Moreover, debtors with lower income and high DSTI post a higher NPL ratio, irrespective of
the purpose of lending (consumer or mortgage lending). There are links between the LTV level and
debtors’ capacity to repay their debt: the higher the LTV, the higher the non-performing loan (NPL)
ratio. These arguments underpin the usefulness of using DSTI and LTV caps to foster debtors’
capacity to repay their loans.
We find that banks’ self‑regulation on DSTI and LTV caps, without specific guidance from the
authorities, delivers sub-optimal results. The Romanian banks behaved in a deeply pro-cyclical
manner when an approach based to a large extent on self‑regulation was implemented. We also find
that the one‑size‑fits‑all approach is not very effective. It is more useful to tailor the DSTI and LTV
measures according to the specific patterns of possible risks. The caps for these instruments should
be differentiated based on debtors’ disposable income (low, medium and high), the currency of
indebtedness (domestic or foreign) and the destination of the loan (mortgage or consumer loans).
The empirical evidence shows that the effectiveness of DSTI and LTV instruments in trimming
down the rapid credit growth is higher than that of monetary policy instruments or other prudential
measures. The Romanian specificities during the last decade largely point to this outcome; these
include a high level of euroization, full capital account liberalization, a material share of FX loans,
and significant foreign capital inflows.
There are three main policy messages for improving the effectiveness in using DSTI and LTV
instruments. First, there is a need to strengthen the cooperation between the domestic macroprudential
authority and other domestic and foreign-related authorities in order to preserve the ability of these
instruments to deal with risks. Second, it is more prudent to monitor the total level of household
and corporate sector indebtedness, instead of total domestic credit developments. The differences
indicated by these two approaches are material. Focusing on the latter might delay the authorities’
decisions on DSTI and LTV implementation or calibration, yielding sub-optimal outcomes in
managing systemic risks. Third, higher transparency is needed from the authorities regarding their
macroprudential intermediate objectives and, if possible, the instruments tailored to fulfill these
objectives.
9
I. Introduction
Tentative lists of macroprudential instruments are already available at the international level
(e.g. ESRB, 2013; IMF, 2013). The next step is to learn about the design, implementation,
calibration and effectiveness of these instruments. Empirical findings on these aspects are relatively
scarce, because, prior to the crisis, many such instruments were considered to distort the free
markets, and so countries were encouraged not to use them. Debt service-to-income (DSTI) or
loan-to-value (LTV) caps are among the instruments that, within a few years, changed their status
from administrative to macroprudential instruments. Jacome and Mitra (2015) managed to collect
details on the experience of six emerging economies that used DSTI and LTV both before and after
the crisis, and have provided extremely useful findings.
Our paper falls within the stream of literature that searches for answers about operationalizing
macroprudential instruments. Some papers (e.g. Aikman et al., 2013; ESRB, 2014) focus
on operationalizing the macroprudential regime, as a whole. Detken et al. (2014) target the
countercyclical capital buffer operationalization. We concentrate on the operationalization of
another two macroprudential instruments: debt service-to-income (DSTI) and loan-to-value (LTV).
We provide inputs about their design, implementation and calibration, using the Romanian bank-level
database as a case study. We also investigate their effectiveness in supporting the macroprudential
intermediate objective of mitigating excessive credit growth and leverage. Our paper complements
Neagu et al. (2015), where optimum levels for DSTI and loan-to-income caps are computed.
Romania is among the small open economies that was forced to resort to unorthodox measures
in order to cope with challenges related to capital account liberalization, the euroization of the
economy, credit boom developments, etc. (details in Annex 1). Some of these unorthodox measures
are now considered “the new normal” (e.g. the use of DSTI and LTV caps). After analyzing a decade
of Romanian experience with these instruments, we conclude that DSTI and LTV instruments are
relatively effective in curbing high credit growth and ensuring that both debtors and creditors
are better equipped to withstand possible adverse shocks. However, we find that the instruments’
effectiveness in curbing house price growth is relatively poor in Romania, the result being in line
with other studies (e.g. Lau and Rau, 2015 for Malaysia). On the other hand, Arregui et al. (2013)
consider that the DSTI/LTV measures have an important impact on house price dynamics. Kim
(2015) reaches the same conclusion for Korea, highlighting that the prudential measures were
reinforced by fiscal ones.
We find that, in the absence of specific guidelines from the macroprudential authority, banks’ self‑
regulation on DSTI and LTV yields sub-optimal results. Romanian banks behaved in a deeply
pro-cyclical manner when an approach based more on self-regulation was implemented. However,
self‑regulation might work in the downward phase of the financial cycle, when banks are more
risk averse. This is the conclusion we might grasp from the Brazil (Afanasieff et al., 2015) and
Poland (Bierut et al., 2015) experiences. We also find that it is useful to tailor the DSTI and LTV
measures according to the specific patterns of possible risks. The caps for these instruments should
be differentiated based on debtors’ disposable income (low, medium and high), the currency of
indebtedness (domestic or foreign) and the destination of the loan (mortgage or consumer loan).
10
The first step in implementing DSTI or LTV caps is to know which systemic risks are more likely to
be addressed by using these instruments. High indebtedness of households is one example of a risk
that might impair financial stability, and the level of debt‑to‑service is a good leading indicator for
banking crises (Drehmann and Juselius, 2013). Macroeconomic variables and bank specific factors
are also leading indicators for repayment behavior (Klein, 2013). Real estate negative developments
represent another risk that might trigger a financial crisis; and Davis et al. (2011) provide a survey
of the literature from this perspective. The main systemic risks monitored in Romania when taking
policy decisions regarding DSTI or LTV caps relate to: (i) borrowers’ indebtedness (or excessive
credit growth – from a lender’s perspective), (ii) the sectoral concentration in real estate assets, and
(iii) the macroeconomic imbalances. To this end, the National Bank of Romania has developed a
framework to monitor the challenges stemming mainly from the household and corporate sectors.
The rest of the paper is structured in three parts. Section II highlights the Romanian framework
for DSTI and LTV: how these instruments are used (calibration, enforcement and communication
processes) and how decisions are taken (institutions involved and dealing with “leakages”).
The effectiveness of the instruments from the perspective of credit, house price and non-performing
loan dynamics is presented in Section III. The last section concludes with a summary of the lessons
to be learned from the Romanian experience.
II. Using the instruments

A. Monitoring systemic risk
In Romania, the identification of potential systemic risks is based on both quantitative and qualitative
tools. The main systemic risks that might trigger activation of the DSTI or LTV caps are: the
high level of borrowers’ indebtedness (or excessive credit growth – from a lender’s perspective),
sectoral concentration in real estate assets and macroeconomic imbalances.
At the early age of DSTI and LTV usage, the key indicators monitored by the authorities when
deciding upon the introduction/calibration of these instruments were: credit growth dynamics
(especially in FX), external equilibrium (current account deficit) and the inflation rate. This
monitoring framework evolved in line with the macroprudential analysis agenda. More emphasis
is now put on indicators flagging potential systemic risks stemming from developments in the real
estate market and borrowers’ (household and corporate sector) vulnerabilities, as a complement to
a banking sector analysis.
The household sector is monitored largely by (i) the structure and dynamics of indebtedness and
(ii) the ability to service the debt. The overall level of indebtedness is quantified by a series of
indicators: the level of debt service-to-disposable income, debt-to-assets, debt-to-net wealth and
debt-to-GDP. The indebtedness is also monitored from a structural perspective: by currency (the
share of FX loans in total loans granted to households), by destination (the share of mortgage and
consumer loans, respectively, in total loans), by categories of disposable income (e.g. the borrowers
with lower than average disposable income display a higher level of indebtedness and a higher
11
non-performing ratio), and by tenure (short, medium, and long-term lending). Households’ ability
to service the debt is assessed by computing the non-performing loan ratio (total, by currency,
income, destination, vintages, etc., see Table A4.1, Annex 4) and by the sensitivity analyses of
the NPL response to potential shocks on interest or exchange rates. The monitoring framework is
based on micro-data from the Central Credit Register (within the central bank) and Credit Bureau
(private entity) databases. The analysis of these indicators points to the following conclusions:
(i) lending in foreign currencies is riskier than lending in domestic currency, (ii) consumer lending
should not exceed 5 years (the level above which the NPL ratio would significantly increase);
and (iii) there is a threshold of indebtedness above which the ability of the borrowers to repay
their debts decreases considerably (Chart A4.1, Annex 4). This threshold differs from one cohort
to another (cohorts are grouped by disposable income), but an aggregate 45 percent yardstick
is informally monitored. This value is in line with Neagu et al. (2015), who find a considerable
upsurge in the probability of unsustainable lending if this threshold is crossed.
The corporate sector is monitored from several perspectives. First, the leverage ratio (debt/capital)
is computed for the overall economy, by main economic sectors, and by firm size (micro, small
and medium enterprises). The yardstick for this ratio is 2, above which a closer monitoring is
performed. Second, the structure and dynamics of loans granted to firms by banks and non‑bank
financial institutions (both domestic and foreign lenders) are examined. The focus is on the share of
the FX loans granted to unhedged borrowers. Third, in order to assess the credit risk, the following
indicators are quantified (i) the one‑year probability of default for the corporate sector and (ii) the
NPL ratio (total, by currency, economic sector, firm size, destination of the loans, etc.). Fourth,
the main soundness indicators of the corporate sector (ROE, EBIT/interest expenses, etc.) are
investigated. A stress-test exercise of the corporate sector is performed to assess its ability to cope
with negative macro and financial developments.
The banking sector is assessed mainly through its capacity to adequately cover the expected
and unexpected losses (through provisions and capital) and its ability to shield against the risks
stemming from the concentration in real estate assets (the overall exposures to construction and
real estate sectors, as well as the volume of real estate collateral from banks’ books are computed).
The DSTI/LTV ratios are examined through several databases. First, the quarterly bank lending
survey is used, where explicit questions about the DSTI and LTV levels are listed1 (Chart A4.4,
Annex 4). Second, DSTI and LTV at credit level can be computed since 2012 respectively, by
merging the databases provided by the Central Credit Register, the Ministry of Finance, and the
National Institute of Statistics. Third, one-off questionnaires help collect further information on
DSTI and LTV on banks’ portfolios.
1
The questionnaire is submitted to the largest 10 banks in the system. The figures for LTV and DSTI ratios have
been collected since Q4/2007: (i) for both the existing stock and the new lending activity, and (ii) for the lowest
and the highest limit of LTV and DSTI, as stipulated in the banks’ internal norms.
12
B. Taking decisions
1. Institutional considerations
All the decisions regarding DSTI and LTV are taken and implemented by the central bank.
The National Bank of Romania (NBR) has monetary policy and microprudential responsibilities,
and, starting in 2004, was also given the task of preserving financial stability. These tasks have
proved complementary so far. The inflationary pressures called for measures to tackle high credit
growth. The high level of indebtedness in FX, as well as the rapid transmission of the domestic
currency depreciation into inflation expectations, benefited from measures to reduce volatility and
sudden changes in the exchange rate.
The proposals regarding DSTI and LTV are discussed in the Supervisory Committee and submitted
to the NBR Board for final decision. The NBR Board is made up of 9 members: 4 executives
(a governor, a first‑deputy governor, and two deputy governors) and 5 non‑executives. De facto,
the Supervisory Committee has a semi-hard to hard decision-making power, as a consequence
of its responsibilities to discuss and endorse prudential measures and of its organization (all the
executive members of the NBR Board are acting in the Supervision Committee). Such institutional
arrangement facilitates the swift implementation of measures like DSTI and LTV caps.
There was strong coordination between the Supervisory Committee and the Monetary Policy
Committee in implementing DSTI and LTV measures to manage capital flows. The capital account
liberalization (KAL)2 process was pro-cyclical in nature, because it overlapped with strong external
inflows. The DSTI and LTV measures introduced in 2003 acted as a counterbalancing stance for
the KAL, complementing the prudent stance of the monetary policy.
2. Application and enforcement

The NBR rules on DSTI and LTV usually enter into force 30 days after their publication in the
Official Gazette. Nevertheless, the measures on DSTI and LTV adopted in the last few years allow
for a 60-day period for enforcement. The reason is the increasing complexity of the process of
setting up the caps, which are differentiated based on the risk profile of the debtors and the bank
products (i.e. mortgage or consumer loans).
In order to limit the unintended consequences of the macroprudential measures, several actions
have been implemented. First, restructured loans have been exempted from DSTI and LTV caps,
with a view to not imposing further limitations on these already constrained debtors that would
affect the servicing of their debt. Second, loans granted within the Prima Casă (“First Home”)
government social program are not subject to LTV caps in order to maintain access to the program
for low-income debtors. The government collateral provides a guarantee for banks that such
exposures will be recovered in case the debtors are unable to repay the loan. Third, starting 2009,
DSTI caps for mortgage loans have not been subject to stress testing requirements (regarding
2
An important part of the capital account liberalization process took place during 2003–2006, and was a prerequisite
for Romania’s accession to the EU in 2007.
13
income, exchange rate and interest rate shocks), as this portfolio exhibits a lower risk profile and
residential real estate would be further affected by additional pro-cyclical measures.
3. Communication
The measures implementing DSTI and LTV caps follow the standard procedures applicable to
those regulations that might materially impact the financial system: a preliminary draft regulation
is discussed with the Romanian Banking Association; the final draft is posted on the central bank
website for further comments from other stakeholders; the approved version of the regulation is
published in the Official Gazette.
The Financial Stability Report is a useful communication tool for the NBR’s views regarding the
risks associated with the lending activity and the needed policy actions. The report (released once
a year) provides separate analyses of developments in the household, corporate, and real estate
sectors, including possible measures that the central bank might implement to tackle the identified
systemic risks.
The NBR experience with DSTI and LTV implementation highlights the need for further
transparency of the authorities’ macroprudential intermediate objectives. Such transparency would
enhance the communication between the central bank and market participants, by explaining the
main drivers of the authorities’ actions.
C. Calibration and dealing with “leakages”
1. Calibration
The DSTI and LTV caps have undergone several changes since their first implementation in early
2004. The calibration of these instruments is based on a hybrid approach that combines qualitative
and quantitative analyses. The qualitative assessment looks at potential upside and downside risks
associated with such measures. The quantitative analysis relies on a set of key indicators: (i) risk
indicators (non-performing loan ratios by type of loans, currency and maturity), (ii) the share of
exposures per sub-portfolio and associated growth rates, and (iii) real estate market developments.
The calibration includes both the value of the caps and their applicability across financial institutions.
From their implementation in early 2004, these prudential measures have been applicable to all
credit institutions (Romanian legal entities) and branches3 of the foreign credit institutions present
in Romania. In 2006, the measures were extended to the main non‑bank financial institutions,
while in 2011 payment institutions and electronic money institutions with a significant level of
lending were brought under the umbrella of the DSTI and LTV requirements.
3
Nevertheless, between 2007 and 2011, branches were exempted from these measures, as a consequence of the
implementation of the EU banking rules. In 2011, branches were asked to meet the DSTI and LTV requirements,
for financial stability reasons.
14
The instruments require changes in order to better contain the systemic risks and to drive away the
circumvention phenomenon. In 2003, caps on DSTI differentiated between two types of loans: 30
percent for consumer loans and 35 percent for mortgage loans. The LTV cap was set at 75 percent
for both consumer loans for the acquisition of goods and mortgage loans. In 2006, a maximum
indebtedness limit of 40 percent of the net income of a debtor and his/her family was imposed, and,
for computing the DSTI, the minimum daily living costs4 were deducted from the disposable income.
In 2007 (Romania joined the EU), the NBR moved from a regulatory framework considered to be
administrative (i.e. caps on DSTI and LTV) to an approach based rather on market self-regulation.
The explicit caps on DSTI and LTV were eliminated. In return, lenders were requested to set up
their own rules for establishing maximum indebtedness’ levels, differentiated by classes of risk for
borrowers5. Such change resulted in the pro-cyclical behavior of credit institutions in establishing
adequate prudential values for DSTI caps.
In order to preserve the prudent regulatory stance, the central bank changed its attitude in 2008.
The NBR introduced a new approach based on the mandatory assessment of debtors’ capacity to
repay their debt in a stress scenario and the requirement for lenders to calibrate the DSTI level
(at origination) in such a way that debtors should not exceed the maximum indebtedness level over
the entire life of the loan. The new approach complemented the self-regulation approach, but with
the NBR providing guidelines to ensure that risk was adequately captured in the lenders’ internal
norms. In 2009, loans granted under the newly implemented Prima Casă national program were
exempted from the DSTI requirement.
The European Systemic Risk Board recommendation on foreign currency lending6 found the
NBR well advanced in implementing DSTI and LTV measures. For full compliance, the central
bank tailored the LTV caps7 based on the type of borrower (hedged/unhedged) and by currency:
75 percent for consumer loans denominated in foreign currency or indexed to a foreign currency
rate, 85 percent for mortgage loans denominated in local currency, 80 percent for mortgage loans
to hedged borrowers denominated in foreign currency, 75 percent for mortgage loans to unhedged
borrowers denominated in euro and 60 percent for mortgage loans to unhedged borrowers
denominated in a foreign currency other than euro. The limits were applied to the new loans, with
the exception of those granted through the Prima Casă national program. For setting the DSTI
caps, lenders were required to also take into account the income risk (in addition to interest rate
and currency risks). In 2012, the DSTI rules were extended to non‑financial companies.
4
This measure was part of a prudential package that also included the introduction of a limit on credit institutions’
exposure in FX loans to unhedged borrowers (individuals or firms) of 300 percent of their own funds, and the
extension of the DSTI and LTV requirements to loans granted by non‑bank financial institutions.
5
These changes provided more flexibility to lenders, but a 40 percent cap on total DSTI was still mandatory until
the new internal norms, incorporating the changes introduced by the new regulation, were validated by the NBR.
6
The European Systemic Risk Board Recommendation 2011/1 from 21 September 2011 on lending in foreign
currencies.
7
The figures were established based on a three‑year adverse scenario, where the value of the collateral should
remain sufficient to cover potential losses. The distinction among currencies is done according to the risks
associated with each portfolio of loans and the developments in the real estate markets at the time of implementation
of the measure.
15
2. Dealing with policy “leakages”

The Romanian experience with the macroprudential measures reveals that DSTI and LTV caps
may be circumvented (Georgescu, 2011), similar to the experience of other countries (Jacome and
Mitra, 2015). First, there is the standard circumvention method of using non-regulated entities to
deliver loans. In order to evade the macroprudential measures introduced in 2003, banks started
to transfer their credit activity to non‑bank financial institutions (entities within the same financial
group). In response, in 2006 the NBR brought under its regulatory and supervisory umbrella
the non‑bank financial institutions with significant lending activity and extended the DSTI/LTV
prudential measures to these institutions.
Second, banks apply promotional interest rates in the first period of the loan in order for the debtors
to fulfill the DSTI requirements at the origination of the credit. At the end of the promotional
period, the level of indebtedness may rise considerably. In 2008, NBR asked the lenders to ensure
that debtors would fulfill the DSTI conditions during the entire life of the loan.
Third, banks lengthen the maturity of credits in order to decrease the monthly debt service payment
of the borrowers in order for the indebtedness level to stay within the DSTI caps. As a response, in
2011, NBR limited the maximum tenure of consumer loans to 5 years.
Fourth, loans from the Prima Casă government program are granted an LTV cap of 95 percent
and are not subject to the stress testing conditions in computing the DSTI. These loans have a
government guarantee, and therefore the credit risk in banks’ balance sheets is curbed to a large
extent.8 The non-performing loan ratio for the Prima Casă portfolio is significantly lower compared
with the overall mortgage portfolio. Possible changes might be needed in the event of a new house
price boom and if the volume of Prima Casă transactions remains a significant component of the
total new mortgage loans provided by banks.
III. Evaluating effectiveness

We evaluate the effectiveness9 of the DSTI and LTV caps from the perspective of their ability
to: (1) trim down high credit growth, (2) prevent a housing price bubble, and (3) foster debtors’
capacity to repay the loans.
A. Credit dynamics
We focus on the evolution of household credit, divided into its main components: consumer loans
and mortgage loans10. We employ a bank lending channel model to estimate the effects of the DSTI
and LTV caps on household credit dynamics. This approach, introduced by Bernanke and Blinder
8
For loans granted from 2009 to 2011, the government provided a 100 percent guarantee, while for loans granted as
of 2011, the guarantee has been reduced to 50 percent.
9
We have the same understanding about the effectiveness of a macroprudential instrument as the ESRB (2013):
“<effectiveness of the instrument> means the degree to which the instrument can address market failures and
achieve the ultimate and intermediate objectives”.
10
DSTI and LTV caps for lending to companies were introduced in December 2012. It is too early to grasp pertinent
conclusions from the evolution of credit delivered to companies.
16
(1988) and Kashyap and Stein (1994), considers that a monetary policy shock that is transmitted
to the supply of credit through interest rate and also influences banks’ willingness to lend due to
frictions in switching between different forms of funding. We are interested in assessing the impact
of the prudential regulation on credit developments, and therefore we add in the bank lending
channel model a variable to identify this effect.
The general form of the equation for the lending channel model is:
where i = 1,N banks, t =1,T quarters, rMP = monetary policy rate, rMRR = the minimum
required reserves ratio, π = the inflation rate, y = the real annual GDP growth rate, X =
variables for various bank characteristics variables (the solvency ratio and the loan-to-
deposit ratio, LTD, which measures banks’ structural liquidity position).
D represents the regulation dummy variable and is defined as taking the value 1 when
prudential measures on DSTI and/or LTV are introduced (and/or modified) and zero
otherwise (the measures are detailed in Table A.1, part A and Chart A4.1, Annex 1).
We estimate the model using the GMM dynamic panel data technique11 following the specifications
used by Ehrmann et al. (2001) and Gambacorta et al. (2004, 2011). The model is estimated for the
first 14 banks of the Romanian banking sector12 (cumulating 80 percent of total outstanding loans)
for the period 2004 – 2012, using credit-by-credit information.
For the dependent variable, we use the total loans granted to households (also separated into its
two components: mortgage loans and consumer loans). The growth rate of loans is corrected for
adjustments in the portfolio due to credit sales (externalized loans). A set of data statistics for the
variables employed in the model is presented in Table A2.1, Annex 2.
The coefficient μ from the above equation captures the response of credit to the prudential measures
regarding DSTI and LTV. The identification is not perfect and the results should be interpreted with
care, as we cannot totally separate the effects of the DSTI/LTV measures from other regulations
that are in place during the same period, the most important of which is the NBR norm that limits
foreign currency exposure from credit activity (Table A.1, part B, Annex 1). Also, it is difficult to
disentangle the consequences of the DSTI measures from the LTV measures, as most of the time
these prudential regulations are introduced at the same time.
The period scrutinized (2004 – 2013) covers almost an entire credit cycle. The level of banking
intermediation increases from around 5 percent of GDP to 20 percent at the end of the expansionary
phase (Chart A2.1, Annex 2). Analyzing the impact of the prudential measures only for the
11
Using an OLS method with fixed effects for banks, even in the absence of the autocorrelation in errors due to the
lagged dependent variable, would have most likely produced biased estimators (Nickell, 1981).
12
We select the banks by putting a materiality threshold on their lending portfolio of 1 percent in total aggregate
credit to households and firms.
17
expansionary phase of the cycle (2004 – 2008) is very important, for at least two reasons: (i) the
estimations are not affected by adjustments during the contraction phase of the credit cycle and
(ii) they give an assessment of the measures’ efficiency in the short run, which should help the
macroprudential authorities calibrate their instruments. These measures are set to produce two
types of effects – the first one, in the short run, directly observable, reducing the excessive credit
growth, and, the second one, in the long run, keeping the credit activity within the prudential limits
and, therefore curbing the credit cycle.
Before setting any econometric models, we employ a simple event study for two key moments: (i) 2004
Q1, when the DSTI/LTV regulation was implemented (Chart A2.2, Annex 2), and (ii) 2008 Q3, when
the prudential regulation was tightened by introducing specific guidelines in setting DSTI limits,
thus making the shift from a “self-regulation” framework to a guided “self-regulation” framework
(Chart A2.3, Annex 2). The results indicate that: (i) the macroprudential measures appeared to have
successfully contributed to the reduction of the credit growth rate, and (ii) the direct impact lasts only
for one year, after which the effect on credit growth decreases to near zero.
We estimate the marginal effect of the DSTI/LTV measures on credit to lie between 4.6 and
8.8 percentage points in the first quarter after the measures are introduced (the highest level being
for consumer loans). The incremental impact of the prudential measures is significantly reduced
after one year, and it is close to zero after two years (Chart A3.1, Annex 3). If we look only at
the expansionary phase of the credit cycle, the impact is stronger (around 10 percentage points)
and significantly higher compared to the response of credit to the monetary policy shock (around
6 percentage points).
To test the robustness of our estimations, we add other variables to the specification described
above, in line with other findings (e.g. Barajas et al., 2003): the change in banks’ external funding
and the change in imports as a percentage of GDP. The results are fairly stable across regression
specifications. We also test the model by replacing the monetary policy rate with the interbank
interest rate, following Angeloni et al. (2003). These last estimations are not included in the paper,
as the results are very similar.
The analysis of total household credit dynamics (Table A3.1, Annex 3) indicates that: (i) the prudential
regulations play an important role, as they exhibit a strong anti-cyclical effect (the impact doubles
in the pre‑crisis period, 2004 Q4 – 2008 Q4). The average impact in the first quarter varies between
a 4.9 to 5.6 percentage points decrease in lending and is around 10 percentage points during the
expansionary phase of the cycle; and (ii) the banks’ external funding is also a major determinant factor.
The mortgage credit growth analysis (Table A3.2, Annex 3) shows that: (i) the regulation variable
enters the regression with a lag of two quarters (a possible explanation is that banks adjust their
lending to households more slowly for housing purchases compared to consumer loans); (ii) the
overall result does not change when controlling for the period during which the Prima Casă
government program is in place (the program was introduced during the downturn phase of the
credit cycle, in 2009 Q3); and (iii) the housing price has an important effect on credit growth
(4 percentage points), but its size and statistical significance are reduced when controlling for other
variables like banks’ external funding.
18
For consumer credit (Table A3.3, Annex 3) the results indicate that: (i) prudential regulations play
a major role, higher than in the case of mortgage loans (the marginal effect of DSTI/LTV measures
on credit lies between 7.5 and 8.8 percentage points for the entire period and at 10.5 percentage
points for the first phase of the cycle); (ii) the demand factor is material (the coefficient for the real
GDP growth rate varies between 0.9 and 1.1) and (iii) banks’ external funding and the change in
imports as a percentage of GDP are important drivers of the consumer credit growth. The results
for the consumer credit growth should be considered with additional care, as the consumer credit
indicator is constructed based on the credit-by-credit information from the Central Credit Register,
which does not record exposures per debtor below RON 20,000 (approx. EUR 4,500).
B. Housing price dynamics

We expect no material effects of DSTI/LTV caps on housing prices in Romania, as data indicates
a low recourse to bank credit for funding real estate transactions during the expansionary phase
of the credit cycle. The number of new mortgage-backed loans was considerably lower compared
with the number of total new transactions in the housing market during the expansionary phase
of the credit cycle (less than 20 percent were credit-based transactions, 2005 – 2008). Moreover,
during the downward trend of the cycle, the number of new mortgage-backed loans decreased
more sharply than the number of real estate transactions.
The dynamics of the housing price index support the same conclusion (Chart A3.2, Annex 3), as
there was no meaningful change in the index during the periods when the DSTI/LTV measures
were in place. In order to test the link between real estate prices and bank credit, we apply a
Granger causality test. The result is shown in Table A3.4, Annex 3. According to this test, the
change in real estate prices precedes (Granger cause) the change in mortgage credit, while the null
hypothesis of no Granger causality is accepted for the inverse relationship.
C. Non-performing loan (NPL) dynamics

We assess the effectiveness of DSTI/LTV regulations in maintaining the quality of the banks’
portfolio. We analyze the results where DSTI/LTV caps are explicitly provided in the prudential
norms or guided self-regulation is in place versus the periods when an approach based to a large
extent on banks’ self-regulation is implemented.
We apply a GMM dynamic panel data model for the evolution of non-performing loan (NPL)
ratios13 by vintages14. We look at the NPL ratio for the portfolio of loans granted each quarter
13
The NPL ratio is defined as the share of non‑performing loans in the total outstanding amount for each vintage.
A loan is considered non-performing (in default), if it is more than 90 days overdue, at least once since its
origination. We use only the exposures per debtor larger than RON 20,000 available in the Central Credit Register.
14
The vintages are defined as the flow of new loans granted quarterly during 2004 – 2010. We compute the time
series of non-performing ratios for each vintage until the end of 2013. The last vintage is 2010 Q4, in order to have
a comfortable outlook period for analyzing its performance. This approach allows for both time and cross-sectional
dimension in our work and for controlling the persistence in the default state. The rescheduled or refinanced loans
are excluded from the analysis.
19
between 2004 and 2010 and assess the driving factors using variables found in the literature.
Additionally, we run similar regression equations on household portfolios with different types of
loans to test for variations in debtors’ repayment behavior.
We use the following categories of explanatory variables, based on findings in the literature
(e.g. Klein, 2013): (i) macroeconomic variables15 (disposable income growth rate, exchange rate,
unemployment rate, real estate price index, etc.), (ii) financial stance indicators16 (banks’ return on
assets, solvency ratio, leverage ratio, etc.) and (iii) prudential regulation.
We construct a dummy variable for the self-regulation period, as following17:
The equation for the developments of the NPL ratio is the following:
where i = 1, N vintages, and t = 1, T quarters, l = number of lags, Xt-l stands for macroeconomic
variables, Yt-l for the financial stance indicators and D represents the regulation tightening
or self‑regulation dummy variable (defined above). We use the evolution of the NPL ratio
for each vintage as a dependent variable.
Like in the case of credit dynamics, we cannot disentangle between measures targeting DSTI, on
the one hand, and LTV, on the other hand, as the regulations overlap for almost the entire period
analyzed. The impact of prudential regulation is tested separately and also in interaction with
macroeconomic variables (e.g. unemployment rate, index for real estate developments, etc.) or
loan characteristics (interest rate, exchange rate, etc.).
The analysis of the NPL ratio for the household’ portfolio as a whole reveals the following
main results. First, the approach based on banks’ self-regulation led to an increase in the NPL
ratio (see Eq.1, Table A5.1 and Chart A5.1, Annex 5) and also to a higher sensitivity of the
indebted households to macroeconomic developments (such as unemployment rate18) (Table
A5.2, Annex 5). The results are in line with the empirical findings, which indicate that the easing
15
More than 90 percent of respondents from a European Commission study (The over-indebtedness of European
households: updated mapping of the situation, nature and causes, effects and initiatives for alleviating its impact,
2013) indicate macroeconomic factors as being the main determinants of over-indebtedness. Among these factors,
unemployment is the most significant, followed by wages, fluctuations in the interest rate and movements in the
exchange rate.
16
The variables reflecting the financial stance of the banking sector are only tested to control for the sensitivity of the
non-performing loan dynamics to banks’ strategies. The leverage ratio delivers the most stable results.
17
For a robustness check, we also test the impact of regulation tightening. The estimations confirm the results, in the
sense that a tightening of regulations contributes to diminishing NPL ratios for the entire household portfolio, and
also for the sub-portfolios.
18
Other macroeconomic variables were also tested, but the results were inconclusive and they are not presented in
the paper.
20
of the lending standards by credit institutions in a pro-cyclical manner (2007 Q1 – 2008 Q3) is
associated with a higher non-performance of both mortgage and consumer loans (Chart A4.2 and
Table A4.2, Annex 4).
Second, we discover that the real estate price dynamics have an impact on all debtors, regardless of
whether they took a loan in the self-regulation period or not19. The correction of real estate prices
is transferred into higher NPL ratios (see Eq.1, Table A5.1, Annex 5), in line with the empirical
findings that show a link between the LTV level and debtors’ capacity to repay their debt. The
riskiest mortgage loans in terms of NPLs are those granted in the period of excessive credit growth
(2007 – 2008) and that currently have the highest LTV values (90 percent in 2007 and over 110
percent in 2008, Chart A4.3 and Table A4.2, Annex 4). This portfolio accounts for almost 75
percent of the total volume of mortgage NPLs (as of December 2013).
For the mortgage loan analysis, we account for the loans granted through the Prima Casă20
government program. In this sense, we use a dummy variable that takes the value 1 for all loans
granted between 2009 Q3 and 2010 Q4. The results show that these loans contribute to a decrease
in the NPL ratio, compared to other vintages (see Eq.2 from Tables A5.1 and A5.2, Annex 5).
Nevertheless, the dummy variable cannot be specifically applied to the Prima Casă loans, but to
all loans granted during this period. The results should be interpreted with caution, because its
effect might also include other factors such as tighter lending standards, adverse real estate price
developments, etc.
We find that non‑mortgage backed consumer loans21 are the most affected by the self-regulation
approach. The sensitivity to macroeconomic factors (the unemployment rate) of loans granted in
this period is the highest (Eq. 3, Table A5.2, Annex 5) and the impact on the NPL ratio is the largest
(Eq. 3, Table A5.1, Annex 5).
IV. Conclusions
We describe an example of designing, implementing and calibrating two macroprudential
instruments: loan-to-value (LTV) and debt service-to-income (DSTI). We also investigate LTV
and DSTI effectiveness in slowing down excessive credit growth and in preserving the quality
of banks’ loan portfolios. Our work is based on the Romanian experience with the use of these
instruments from 2004 onwards. During this period, Romania witnessed both sides of the credit
and business cycles, and encountered challenges related to full capital account liberalization,
the euroization of the economy, credit boom developments, etc. Some unorthodox measures
19
We also tested the real estate price variable with an interaction, but the sensitivity does not differ substantially.
20
The programme has a significant contribution to the mortgage loan flows and accounts for almost 45 percent of the
total mortgage loan stock (June 2014).
21
The results should be viewed with caution as the NPL ratio for non-mortgage backed consumer credit was, due to
the lack of historical data, constructed based on the information from the Central Credit Register which does not
record loans with values lower than approximately EUR 5,000. This portfolio is among the riskiest category of
loans granted to households (Table A.4.1, Annex 4).
21
implemented a decade ago to cope with these challenges are now among “the new normal” (e.g.
the use of DSTI and LTV caps).
The first step in implementing DSTI or LTV caps is to know which systemic risks are more likely
to be addressed by using these instruments. The main systemic risks monitored in Romania when
taking policy decisions regarding DSTI or LTV relate to: the level of borrowers’ indebtedness (or
excessive credit growth – the lender perspective), the sectoral concentration in real estate assets
and the macroeconomic imbalances. To this end, the NBR has developed a framework to monitor
the challenges stemming mainly from the household and corporate sectors.
We find that the DSTI and LTV caps have been relatively effective in: (i) curbing high lending
activity and (ii) ensuring that both debtors and creditors are able to cope with possible adverse
shocks in real estate prices, domestic currency depreciation, or interest rate hikes. The latter point
explains why the LTV and DSTI caps should be used both in the upswing and the downswing of
the credit cycle.
We learn that a one‑size‑fits‑all approach is not very effective. It is more useful to tailor the
DSTI and LTV measures according to the specific patterns of possible risks. The caps for these
instruments should be differentiated according to the debtors’ disposable income (low, medium
and high), the currency of indebtedness (domestic or FX loans) and the destination of the loan
(mortgage or consumer).
We apply an extension of the Bernanke and Blinder (1989) lending channel model using a GMM
dynamic panel technique to assess the impact of DSTI/LTV measures on lending dynamics
and NPL developments. We use bank-level information and credit-by-credit databases to grasp
the consequences of the DSTI and LTV caps on lenders’ and debtors’ behavior. We find that
these macroprudential instruments have a strong anti‑cyclical effect. We find that banks’ self‑
regulation on DSTI and LTV caps, without any specific guidance from the authorities, would
deliver sub-optimal results. The Romanian banks behaved in a deeply pro-cyclical manner when
an approach based to a large extent on self-regulation was implemented.
We find strong negative correlation between the level of indebtedness and the ability to repay
the debt. Moreover, debtors with lower income income and high DSTI post a higher NPL ratio,
irrespective of the purpose of lending (consumer or mortgage lending). There are links between
the LTV level and debtors’ capacity to repay their debt: the higher the LTV, the higher the non-
performing loan ratio. These arguments underpin the relevance of using DSTI and LTV caps to
foster debtors’ repayment capacity.
The effectiveness of these macroprudential instruments in slowing down the rapid credit growth
seems to have been higher than that of the monetary policy or microprudential instruments. The main
explanations for this outcome are: (i) the high level of euroization, full capital account liberalization
and the material share of FX lending affect monetary policy transmission, (ii) substantial foreign
capital inflows prior to the crisis and significant returns in the banking sector limit the scope of
microprudential instruments; and (iii) in periods of high euphoria among market participants, the price
costs channel (like interest rates) seems to work less effectively than the quantity costs channel (like
22
DSTI and LTV caps) in altering players’ behavior. The econometric analysis supports this conclusion
that the DSTI and LTV measures played an important role in dampening the growth in credit.
We find that DSTI and LTV effectiveness in curbing house price growth is relatively poor in
Romania, the result being in line with other studies. Data indicates a low recourse to bank credit
for funding real estate transactions. The number of new mortgage-backed loans is considerably
lower than that of total new transactions.
The Romanian experience with the macroprudential measures reveals that DSTI and LTV caps
may be circumvented. The experience is similar in other countries (Jacome and Mitra, 2015). First,
there is the standard circumvention method of using the non-regulated entities to extend loans in
order for banks to escape the macroprudential requirements. In response, the NBR has brought
under its umbrella the main non‑bank financial institutions and extended the scope of DSTI/LTV
prudential measures so as to cover these institutions as well. Second, the NBR asked the lenders
to ensure that debtors would fulfill DSTI conditions during the entire life of the loan, not only at
the origination of the credit. Third, the NBR limited the maximum tenure of consumer loans to
5 years, as banks lengthened the loan maturity to decrease borrowers’ monthly debt service in
order for the indebtedness level to fit the DSTI caps.
There are three main lessons to be drawn from the Romanian experience in using DSTI and LTV
caps. First, there is a need for strong cooperation between domestic and foreign authorities in
order to preserve the effectiveness of these instruments. The risk of arbitrage remains, although
these measures are harder to be circumvented compared with others, especially when used in
combination. On the domestic front, the key issue is to ensure that other types of lenders (like
non‑bank financial institutions or some components of shadow banking) fulfill the same lending
conditions as banks. The NBR has addressed this challenge by bringing under its regulatory and
supervisory umbrella the non‑bank financial institutions with significant lending activity. On the
external front, the establishment of the European Systemic Risk Board has created the framework
to deal with a large part of the arbitraging activity that becomes manifest within the EU borders.
The second lesson highlights the need for a change in the macroprudential authorities’ perspective,
by moving from the lender to the debtor side. Monitoring mainly developments in the monetary
surveys (i.e. loans extended by domestic banks), might deliver a substantially incomplete picture
to policymakers, and their decisions would be sub-optimal. In the borrowers’ portfolio there
is a lot of additional debt stemming from: (i) domestic financial institutions other than banks,
(ii) non‑resident financial lenders, (iii) loans originally provided by domestic banks, but currently
externalized to third parties, and (iv) financial loans extended directly by the parent company.
A debtor perspective would focus on the total level of borrowers’ indebtedness, irrespective of the
sources (from domestic or abroad, from banks or non-banks, etc.). A creditor perspective would
concentrate on the challenges to the banks’ portfolio, and the crisis has taught us that such an
approach might deeply underestimate the risks.
The third lesson relates to the importance of higher transparency on the part of the authorities
regarding their macroprudential intermediate objectives. Markets should be informed about these
23
objectives, and, if possible, the instruments should be tailored to fulfill them. Such an approach would
enhance the communication between the macroprudential authority and the financial markets, and
would decrease the resistance from the private sector when such instruments are implemented. The
authorities should build internal risk-monitoring procedures to address the possible impairment of
macroprudential intermediate objectives, including early warning thresholds, and instruments to
be used in case a threshold is breached.
24
References
Afanasieff, T.S., Implementing Loan-to-Value Ratios: The Case of Auto Loans in

Carvalho F.L.C.A., Brazil (2010-11), International Monetary Fund, Monetary and
Eduardo C. de Castro, Capital Markets Department project, forthcoming, Central Bank
Coelho R.L.P., of Brazil, 2015
Gregório, J.
Aikman, D., Operationalising a Macroprudential Regime: Goals, Tools and
Haldane, A., Open Issues, Banco de España Financial Stability Journal, 24,
Kapadia, S. 2013
Angeloni, I., Monetary Policy Transmission in the Euro Area, Cambridge
Kashyap A.K., University Press, Cambridge, 2003
Mojon, B.
Arellano, M., Another Look at the Instrumental Variable Estimation of
Bover, O. Error-Components Models, Journal of Econometrics, 68(1),
29-51, 1995
Arregui, N., Evaluating the Net Benefits of Macroprudential Policy:
Benes, J., a Cookbook, IMF Working Paper 13/167, 2013
Krznar, I.,
Mitra, S.,
Santos, A. O.
Barajas, A., Dollarization of Liabilities: Beyond the Usual Suspects, IMF
Méndez Morales, A. Working Paper 03/11, 2003
Bierut, B., Implementing Loan-to-Value and Debt to Income Ratios: Learning
Chmielewski, T., from Country Experiences, International Monetary Fund,
Glogowski, A., Monetary and Capital Markets Department project, National Bank
Stopczynski, A., of Poland, forthcoming, 2015
Zajączkowski, S.
Bernanke, B.S., Credit, Money, and Aggregate Demand, NBER Working Paper
Blinder, A.S. No. 2534, 1988
Blundell, R., Initial Conditions and Moment Restrictions in Dynamic Panel
Bond, S. Data Models, Journal of Econometrics, 87(1), 115‑143, 1998
Davis, E.P., Housing Market Dynamics and Macro-Prudential Tools, Brunel
Fic, T.M., University Working Paper No. 11-07, May, 2011
Karim, D.
Detken, C., Weeken, O., Operationalising the Countercyclical Capital Buffer: Indicator
Alessi, L., Bonfim, D., Selection, Threshold Identification and Calibration Options,
Boucinha, M., Castro, C., European Systemic Risk Board Occasional Paper, June, 2014
Frontczak, S., Giordana, G.,
Giese, J., Jahn, N., Kakes, J.,
Klaus, B., Lang, J.,
Puzanova, N. and Welz, P.
Drehmann, M., Evaluating Early Warning Indicators of Banking Crises:
Juselius, M. Satisfying Policy Requirements, BIS Working Paper No. 421, 2013
25
Ehrmann, M., Financial Systems and the Role of Banks in Monetary Policy
Gambacorta, L., Transmission in the Euro Area, ECB Working Paper No. 0105,
Martinéz Pagés, J., 2001
Sevestre, P.,
Worms, A.
European Systemic Risk Recommendation on Intermediate Objectives and Instruments of
Board Macro-Prudential Policy, ESRB/2013/1, 2013
The ESRB Handbook on Operationalising Macro-prudential
Policy in the Banking Sector, 2014
Gambacorta, L., Does Bank Capital Affect Lending Behavior?, Journal of Financial
Mistrulli, P.E. Intermediation, 13(4), 436-457, 2004
Gambacorta, L., The Bank Lending Channel: Lessons From the Crisis, Economic
Marques-Ibanez, D. Policy, 26(66), 135-182, 2011
Georgescu, F. Contribuţia politicii prudenţiale a BNR în cadrul mix-ului de
politici economice, National Bank of Romania Occasional Papers
no.29, 2011
International Monetary Key Aspects of Macroprudential Policy, 2013
Fund
Jacome, L., LTVs and DTIs – Going Granular, IMF Working Paper,
Mitra, S. forthcoming, 2015
Kashyap, A.K., The Impact of Monetary Policy on Bank Balance Sheets, NBER
Stein, J.C. Working Paper No. 4821, 1994
Kim, Y.M. Implementing Loan-to-Value and Debt-to-Income Ratios:
Learning from Country Experiences, International Monetary Fund,
Monetary and Capital Markets Department project, forthcoming,
Bank of Korea, 2015
Klein, N. Non-Performing Loans in CESEE: Determinants and Impact on
Macroeconomic Performance, IMF Working Paper 13/72, 2013
Lau, C.C., Implementing Loan-to-Value and Debt-to-Income Ratios:
Rani, M.S.A. Learning from Country Experiences, International Monetary Fund,
Monetary and Capital Markets Department project, forthcoming,
Bank Negara Malaysia, 2015
National Bank of Romania Financial Stability Report, Box 3.2.1. Assessing the value of
real-estate assets, 2006
Neagu, F., Coping with an Unsustainable Lending: Early Warning
Mihai, I., Indicators and the Use of Macro-Prudential Instruments, ECB
Tatarici, L. Macro-prudential Research Network project, mimeo, 2015
Nickell, S. Biases in Dynamic Models with Fixed Effects, Econometrica, 49,
1399-1416, 1981
Windmeijer, F. A Finite Sample Correction for the Variance of Linear Efficient
Two-Step GMM Estimators, Journal of Econometrics, 126(1),
25-51, 2005.
26
Annex 1 – Regulatory measures taken by the National Bank of Romania

Table A.1. Regulatory measures taken by the NBR
Period A. Prudential measures Credit cycle phases

with an impact on lending
February 2004 A maximum indebtedness level* of 30% Early developments on the domestic credit market:
for consumer loans, 35% for mortgage 2003 – 2004
loans. In the case of consumer loans for The macroeconomic environment is characterized by
purchasing goods the down payment strong economic growth (with GDP growth rates of
was set at 25%, but for consumer loans 5.1% to 5.7% in 2001 – 2003 and 8.5% in 2004) and a
for other purposes the guarantees should steady decline in the inflation rate. Economic growth is
cover 100% of the loan. The maximum spurred by a significant increase in domestic demand,
LTV level for mortgage loans was set at especially household consumption (fuelled by the hike
75%. The requirements are applied to in the minimum wage and optimistic expectations about
both lei- and foreign-denominated loans. future income). The relative improvements in the fiscal
position and a slight deterioration of the external balance
(backed by strong foreign direct investment) add to the
overall enhancement of the macroeconomic context. The
banking sector undergoes major restructuring and reform
measures that end up in several banks being closed down
and a new banking law being enacted. At the end of
2003, financial assets represent 30% of GDP, of which
more than 90% come from the banking sector. Financial
liberalization is the defining element of this period. The
large foreign capital inflows create a liquidity surplus in
the banking sector and the National Bank of Romania
(NBR) strives to sterilize it (the minimum reserve
requirements for foreign currency liabilities are raised to
30% in the second half of 2004).
The rapid acceleration of domestic bank credit (the

total stock of non-government credit increases by 48%
in 2003 and 26.2% in 2004, while loans to households
jump by 214% in 2003 and 44.8% in 2004, real terms
figures), in the context of large capital inflows into the
banking sector, raises concerns for the central bank
as regards lending sustainability. Measures to ensure
adequate risk management for banks are envisaged. In
this environment, DSTI and LTV caps are implemented
at the end of 2003.**
**
* The indebtedness level is calculated as the During the 1970’s and 1980’s, several similar measures to rationalize
monthly debt service value (principal, income credit were in place, including a sort of DSTI (where the maturity and
expenses and, since 2006, fees and other credit the interest rates were set by the law) and LTV. The measures on DSTI
expenses) divided by the debtor’s net monthly and LTV discussed in this paper refer to those implemented in post-
income (and, since 2006, the income includes communist Romania. Such measures were approved by the National
only permanent sources of income and excludes Bank of Romania Board in December 2003 and entered into force in
encumbered resources). February 2004.
27

August 2005 A maximum indebtedness level for all The first phase of the excessive credit growth period:
debts (bank credit, leasing, NFI credit) 2005 – 2006
of 40% of the debtor’s net income or, Economic growth remains elevated and above potential
when appropriate, of his/her family’s is output, fuelled by domestic demand. Household
introduced. The requirement is applied to consumption is the main driver, with domestic
both lei- and foreign-denominated loans. investment rapidly picking up. The external balance
deteriorates further, although it is largely by foreign
October 2006 The regulation is extended so as to
direct investment.
cover non‑bank financial institutions
with significant lending activity. The The strong non-government credit growth resumes in
fees and other costs are included in the 2006, after a slight setback in 2005, with the household
calculation of the indebtedness level. sector remaining the driving force of credit dynamics
The requirement is applied to both lei- (out of which consumer credit posts the fastest rise).
and foreign-denominated loans. According to Georgescu (2011), after the implementation
of 2005 measures, a clear shift is observed in terms of
credit dynamics.
The main challenges for the NBR are: (i) to cope with
the inflationary pressures exerted by demand factors;
the NBR used additional, less traditional instruments,
to achieve its objectives (the MRR ratio for foreign-
denominated liabilities was raised to 40% in March
2006), (ii) to improve the transmission mechanism of
the policy rate in the context of a new monetary policy
regime (inflation targeting was adopted in August 2005)
and of large foreign capital inflows (in the context of
completion of the capital account liberalization ahead of
EU accession in January 2007) and (iii) to weather the
rapidly increasing household indebtedness.
March 2007 The regulation is changed from The late phase of the excessive credit growth period:
explicitly defined caps on DSTI/LTV to 2007 – 2008
a more banks’ self-regulation approach. In 2007 and the first three quarters of 2008, the macro‑
Credit institutions are requested to financial conditions continue to be buoyant. Economic
define in their internal norms the growth is strong (6.3% in 2007 and 7.3% in 2008) and
maximum levels for caps, according to the unemployment rate is on a downward trend (to 6.1%
their risk profile, differentiated by the in 2007 and further to 5.8% in 2008). Non-government
borrower’s risk class. credit continues to expand (by 50.5% in real terms in
2007). Foreign-currency denominated loans become
August 2008 The regulation is amended by requiring
the main component of the credit stock (54.3% in 2007,
banks to evaluate debtors’ capacity
growing by 72.6% in real terms).
to repay in a stress scenario, taking
into account the foreign currency risk
and the interest rate risk. Banks are
required to introduce different ceilings
on indebtedness ratios for different
currency denominated loans.
28

October 2011 ▪ Introduction of explicit limits for LTV: The credit contraction phase or post-crisis period:
2009 – present
– For mortgage loans, the LTV limits
are: 85% for loans denominated The intensification of the international financial crisis in
in local currency, 80% for loans September 2008 entails a severe shift in investors’ risk
denominated in foreign currency tolerance for emerging market assets. Material changes
to hedged borrowers, 75% for occur in the domestic macro‑financial conditions:
loans denominated in euro to economic activity contracts by 6.6% and unemployment
unhedged borrowers and 60% for increases to 7.5% in 2009. The changes at the macro-
loans denominated in other foreign level impacted the credit market mainly through the
currencies than the euro granted to income, foreign currency (representing the majority
unhedged borrowers. LTV limits of the credit stock) and wealth (housing prices adjust
are not applied to the Prima Casă significantly) channels.
loans.
Economic growth returns into positive territory in 2011,
– For consumer credit in foreign after two years of contractions, while the unemployment
currency, the value of purchased rate stands above 7%. Consumer demand remains
goods shall not exceed 133%. subdued, owing to a further decline in real disposable
Consumer credit is limited to 5 income, increasing uncertainty regarding the future
years. financial stance, decreasing real estate prices and high
household indebtedness. The unfavorable international
▪ Maintaining the requirement for macroeconomic and financial environment also add to the
banks to calculate the maximum level negative developments.
of indebtedness in a stress scenario by
taking into account foreign currency,
interest rate and income risks.
In the case of consumer loans, the

shock values for the risk factors are
explicitly specified: (i) a depreciation of
the local currency against the EUR (by
35.5%), against the CHF (by 52.6%)
and against the USD (by 40.9%),
(ii) 0.6 percentage points increase for
interest rate risk and (iii) a 6% shock
for income risk.
December Expansion of the measures to non-

2012 financial companies unhedged to
currency risk by requiring lenders to
apply tighter conditions on foreign
currency-denominated loans.
29
Period B. Other measures with an impact on lending
September 2005 Overall exposure limits for credit institutions (Romanian legal entities) regarding foreign currency-
denominated loans to households and companies equal to at most 300 percent of their own funds
(after deducting credit risk provisions).
October 2005 Credit institutions may include the borrowers who do not earn steady income in the currency in
which their loan is denominated in the ‘B’ financial performance category at most.
February 2006 The regulation and supervision of non‑bank financial institutions by the central bank.
March 2008 The October 2005 restriction regarding the possibility to classify an unhedged borrower in the ‘B’
financial performance category at most is eliminated. Additionally, a new requirement regarding
distinct provisioning coefficients for loans in foreign currency or linked to another currency
exchange rate granted to unhedged borrowers, as compared to hedged borrowers is introduced.
April 2009 Improvements in the provisioning reporting and allowing for a maximum 25% deduction of
provisions for the principal amounts of property-backed loans.
June 2009 Launch of the Prima Casă government program.
Period C. Changes in the minimum reserve requirements ratios (MRR)
August 2004 MRR ratio on foreign currency-denominated liabilities: increase from 25% to 30%
August 2005 MRR ratio on lei-denominated liabilities: decrease from 18% to 16%
January 2006 MRR ratio on foreign currency-denominated liabilities: increase from 30% to 35%
March 2006 MRR ratio on foreign currency-denominated liabilities: increase from 35% to 40%
July 2006 MRR ratio on lei-denominated liabilities: increase from 16% to 20%
November 2008 MRR ratio on lei-denominated liabilities: decrease from 20% to 18%
July 2009 MRR ratio on foreign currency-denominated liabilities: decrease from 40% to 35%
MRR ratio on lei-denominated liabilities: decrease from 18% to 15%
August 2009 MRR ratio on foreign currency-denominated liabilities: decrease from 35% to 30%
November 2009 MRR ratio on foreign currency-denominated liabilities: decrease from 30% to 25%
April 2011 MRR ratio on foreign currency-denominated liabilities: decrease from 25% to 20%
January 2014 MRR ratio on foreign currency-denominated liabilities: decrease from 20% to 18%
MRR ratio on lei-denominated liabilities: decrease from 15% to 12%
30
Chart A1.1 Macroprudential measures in Romania relative to the household loan dynamics
200
Maximum aggregate DSTI level
of 40% (total indebtedness)
(August 2005) Introduction of the
Limits on banks’ aggregate
“Prima casă” program
150 (June 2009)
exposures to legal persons
Limits on banks’
unhedged against FX risk
exposure to FX risk
Maximum DSTI (December 2012)
(September 2005)
is established by
yoy dynamics (%)
100 taking into account

interest rate risk
and FX risk Restrictions on FX loans to households:
(August 2008) LTV varies between 60% for other
DSTI/LTV measures currencies and 85% for lei – denominated
50
Imposing a maximum are extended to loans. Maximum DSTI is established
indebtedness level of non-bank financial by taking into account interest rate risk,
30% on consumer loans, institutions with FX risk and income risk (October 2011)
35% on mortgage loans significant lending
0 activity (October 2006)
The maximum LTV contraction of economic activity
Easing of lending terms by
level was set at 75% total credit (real, yoy)
eliminating the restrictions on
(February 2004) LTV and DSTI (March 2007) FX credit (real, yoy)
-50
2004Q1
2004Q2
2004Q3
2004Q4
2005Q1
2005Q2
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2007Q3
2007Q4
2008Q1
2008Q2
2008Q3
2008Q4
2009Q1
2009Q2
2009Q3
2009Q4
2010Q1
2010Q2
2010Q3
2010Q4
2011Q1
2011Q2
2011Q3
2011Q4
2012Q1
2012Q2
2012Q3
2012Q4
Source: NBR, NBR calculations
Annex 2 – Credit dynamics

Chart A2.1 The evolution of household loans Chart A2.2 The change in household loan
as a share in GDP during 2004 – 2013 dynamics (real annual growth rate) before and
after 2004* prudential regulations
25 percentage points
100
20 50
15 0
-50
10
-100
5
-150
0 -200
Dec.04
Jun.05
Dec.05
Jun.06
Dec.06
Jun.07
Dec.07
Jun.08
Dec.08
Jun.09
Dec.09
Jun.10
Dec.10
Jun.11
Dec.11
Jun.12
Dec.12
Jun.13
Dec.13
Jun.03
Dec.03
Dec.04
Mar.04
Jun.04
Mar.05
Sep.03
Sep.04
credit – lei
consumer loans/GDP credit – foreign currency
mortgage loans/GDP credit – total
household loans/GDP
Note: Rates are normalised at 2004 Q1.
Source: NBR, NIS, own calculations Source: NBR, own calculations
* Due to data availability issues, the impact of the prudential

measures on credit dynamics by loan destination, i.e. mortgage and
consumer loans, could not be calculated for the measures
introduced in 2004.
31
Chart A2.3 The change in household loan dynamics (real annual growth rate) before and after 2008
prudential regulations
by destination by currency
percentage points percentage points, yoy
40 60
20 30
0 0
-20 -30
-40 -60
-60 -90
Mar.08
Mar.09
Jun.08
Jun.09
Dec.07
Dec.08
Dec.09
Sep.08
Sep.09
Mar.08
Mar.09
Jun.08
Jun.09
Dec.07
Dec.08
Dec.09
Sep.08
Sep.09
credit – mortgage credit – lei

credit – consumer credit – foreign currency
credit – total
Note: Rates are normalised at 2008 Q3. Note: Rates are normalised at 2008 Q3.
Source: NBR, own calculations Source: NBR, own calculations
Table A2.1 Summary statistics
Variable Observations Mean Std. Min Max
Growth rate of credit, (total) 490 11.0 19.9 -39.9 151.2

Growth rate of credit, (mortgage) 475 9.5 17.2 -36.3 177.1
Growth rate of credit, (consumer) 487 13.4 25.3 -43.9 151.6
Change in monetary policy rate 490 -0.4 1.1 -5.0 2.0
Change in interbank market rate 490 -0.4 2.2 -5.6 5.6
Change in the minimum reserve
490 -0.1 1.0 -3.0 4.0
requirements rate (MRR, lei)
Change in the minimum reserve
490 -0.3 2.7 -10.0 10.0
requirements rate (MRR, FX)
Loan-to-deposits 490 141.0 97.9 15.2 592.1
Solvency ratio 490 16.3 6.4 8.3 49.2
Real GDP growth rate 490 2.8 4.8 -7.9 9.8
Bank external debt 490 2.6 0.3 1.7 2.9
Imports/GDP 490 3.6 0.1 3.4 3.8
Growth rate of real estate price 490 4.5 0.3 3.7 5.0
32
Annex 3 – Results of the econometric models on credit growth
Table A3.1 Total loans to households
Eq. 1 Eq. 2 Eq. 3 Eq. 4

2004 Q4 – 2012 Q4 2004 Q4 – 2008 Q4
*** *** ***
Growth rate of credit, total 0.532 0.519 0.535 0.411***
(t-1) (0.00) (0.00) (0.00) (0.00)
0.688** 0.619*** 0.719*** 1.410***
Real GDP growth rate (t-1)
(0.03) (0.01) (0.00) (0.00)
Change in monetary policy 0.924 1.214 1.245 -0.812
rate (t-1) (0.79) (0.73) (0.72) (0.84)
rate (t-2) (0.73) (0.48) (0.90) (0.92)
Change in monetary policy -3.608*** -3.793*** -3.590*** -3.645***
rate (t-3) (0.00) (0.00) (0.00) (0.00)
Change in monetary policy -0.171 0.051 -0.157 -0.420
rate (t-4) (0.79) (0.93) (0.79) (0.59)
Regulation Dummy -5.497* -4.867* -5.612** -10.003***
(t-1) (0.05) (0.06) (0.03) (0.00)
LTD (t-1)* change in 0.008 0.007 0.007 0.010
monetary policy rate (t-1) (0.40) (0.44) (0.44) (0.32)
Solvency ratio (t-1)* change -0.193 -0.183 -0.186 -0.167
in monetary policy rate (t-1) (0.16) (0.19) (0.17) (0.24)
Change in minimum reserve 0.665
requirements rate, RON (t-1) (0.41)
requirements rate, FX (t-1) (0.68)
Inflation -0.316
(t-1) (0.14)
Change in bank external debt 19.704*
(t-2) (0.09)
Change in imports/GDP 5.475*
(t-1) (0.09)
Observations 406 406 406 182
Instruments 18 16 16 17
Hansen t-stat 7.148 9.835 9.430 9.801
Hansen p-val 0.307 0.132 0.151 0.279
AR(1) 0.079 0.078 0.076 0.069
AR(2) 0.990 0.955 0.976 0.886
The estimation method is Arellano and Bover (1995). We use the Windmeijer (2005) correction. The small number of groups limits
the number of instrumental variables (Hansen probability increases to very high values). The endogenous variables used are the
lagged dependent variable and the real economic growth. The instrumental variables are entered into the regression with two to five
lags. The p-values are displayed in parentheses, where * p<0.10, ** p<0.05, *** p<0.01.
33
Table A3.2 Mortgage credit portfolio
Eq. 1 Eq. 2 Eq. 3 Eq. 4

2004 Q4 – 2012 Q4 2004 Q4 – 2008 Q4
*** *** ***
Growth rate of credit, 0.390 0.327 0.351 0.290***
mortgage (t-1) (0.00) (0.00) (0.00) (0.00)
Real GDP growth rate 0.887*** 0.893*** 0.944** 1.233**
(t-1) (0.00) (0.00) (0.01) (0.02)
Change in monetary policy -4.331 -4.877** -4.870** -5.249**
rate (t-1) (0.47) (0.02) (0.04) (0.04)
Change in monetary policy 1.086 0.840 0.807 0.002
rate (t-2) (0.22) (0.39) (0.42) (1.00)
rate (t-3) (0.00) (0.00) (0.00) (0.00)
LTD (t-1)* change in 0.012
monetary policy rate (t-1) (0.21)
Solvency ratio (t-1)* change -0.121
in monetary policy rate (t-1) (0.71)
Regulation Dummy 0.488
(t-1) (0.76)
Regulation Dummy -4.820* -5.109* -5.123** -10.922**
(t-2) (0.05) (0.06) (0.04) (0.04)
Change in minimum reserve -0.032
Inflation (t‑1) -0.159
(0.62)
Growth rate of real estate 4.099 2.919 3.490 1.050
price (t-1) (0.10) (0.45) (0.16) (0.85)
Change in bank external debt 9.167
(t-2) (0.21)
Prima Casă Dummy -1.826
(0.44)
Hansen t-stat 1.446 11.896 12.841 6.252
Hansen p-val 0.963 0.156 0.117 0.619
AR(1) 0.018 0.019 0.024 0.020
AR(2) 0.911 0.873 0.870 0.772
lagged dependent variable and the real economic growth. The instrumental variables are entered into the regression with two to six
34
Table A3.3 Consumer credit portfolio
Eq. 1 Eq. 2 Eq. 3 Eq. 4

2004 Q4 – 2012 Q4 2004 Q4 – 2008 Q4
Growth rate of credit, 0.520*** 0.488*** 0.513*** 0.325***
consumer (t-1) (0.00) (0.00) (0.00) (0.01)
Real GDP growth rate 0.930*** 1.035*** 1.134*** 2.068***
(t-1) (0.00) (0.00) (0.00) (0.00)
Change in monetary policy -1.396 -1.269 -1.249 -2.509**
rate (t-1) (0.23) (0.25) (0.26) (0.04)
rate (t-2) (0.77) (0.40) (0.88) (0.99)
rate (t-3) (0.00) (0.00) (0.00) (0.01)
rate (t-4) (0.62) (0.42) (0.57) (0.52)
Regulation Dummy -7.603** -7.510** -8.847** -10.472**
(t-1) (0.02) (0.03) (0.01) (0.05)
Change in minimum reserve -0.462
requirements rate, RON (t-1) (0.54)
Inflation (t‑1) -0.034
(0.91)
Change in bank external 21.842
debt (t-2) (0.13)
Change in imports/GDP 9.430***
(t-1) (0.01)
Hansen t-stat 8.587 11.772 11.037 11.251
Hansen p-val 0.198 0.162 0.200 0.188
AR(1) 0.021 0.019 0.017 0.035
AR(2) 0.652 0.739 0.604 0.863
lagged dependent variable and the real economic growth. The instrumental variables are entered into the regression with two to six
35
Chart A3.1 The impact of DSTI/LTV regulation on credit growth rate
(a) Total household credit (b) Mortgage credit (c) Consumer credit
percentage points percentage points percentage points
0 0 0
-5 -5 -5
-10 -10 -10
-15 -15 -15
-20 -20 -20

After one Cumulative After one Cumulative After one Cumulative
quarter 2 years quarter 2 years quarter 2 years
total cycle first phase of cycle
Note: The results are derived from regression equations 1 and 4 in Tables A3.1, A3.2 and A3.3.
Table A3.4 Granger causality test
Null Hypothesis: Observations F Statistics Probability
Changes in Mortgage Credit do not Granger Cause changes in

31 1.37 0.28
Real Estate Price Index
Changes in Real Estate Price Index do not Granger Cause
31 3.71 0.02
changes in Mortgage Credit
Note: The number of lags used is 4 and was set based on Akaike information criterion (AIC) and on Hannan-Quinn information
criterion (HQ). The sample used is 2002 Q4 – 2013 Q4.
Chart A3.2 The impact of DSTI/LTV regulation on real estate prices
160 160
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
Dec.02
Dec.03
Dec.04
Dec.05
Dec.06
Dec.07
Dec.08
Dec.09
Dec.10
Dec.11
Dec.12
Sep.04
Sep.05
Mar.03
Jun.03
Sep.03
Mar.04
Jun.04
Mar.05
Jun.05
Mar.06
Jun.06
Sep.06
Mar.07
Jun.07
Sep.07
Sep.08
Sep.11
Sep.12
Sep.13
Mar.08
Jun.08
Mar.09
Jun.09
Sep.09
Mar.10
Jun.10
Sep.10
Mar.11
Jun.11
Mar.12
Jun.12
Mar.13
Jun.13
Note:
Note:TheThelineline
represents the real
represents the estate price price
real estate index,index,
the redthe
areared
represents the periods
area represents thewhen the when
periods most important
the most regulations
important regulations on
on DSTI/LTV were in place. For the real estate price data, we use two different indexes. For the period 2002-2008,
DSTI/LTV were in place. For the real estate price data, we use two different indexes. For the period 2002-2008, the index
the index is calculated based on a hypothetical value for residential real estate prices (the methodology is described
in is
thecalculated based onReport,
Financial Stability a hypothetical
2006) andvalue for residential
for 2009 onwards, wereal useestate prices Institute
the National (the methodology
of Statisticsis(NIS)
described
index.in the Financial
TheStability Report,
gap between 2006) and
December 2008forand2009
March onwards, we to
2009 is due usethethe NationalinInstitute
differences the data of Statistics
sources (NIS)
used by index. The gap between
the two
December
methods 2008 and
employed. The March 2009
NIS index is due to
measures the differences
changes in apartment
in prices for the data sources used
buildings, by the two
individual methods
buildings and employed. The NIS
residential property changes
index measures overall. in prices for apartment buildings, individual buildings and residential property overall.
Source: NIS, NBR, own calculations
36
Annex 4 – Non-performing loan dynamics

Table A4.1 Developments of NPL* ratio by type of loan and by currency for the households’ portfolio
NPL ratio* (percent)
Date Mortgage
backed Non-mortgage Local currency
Mortgage loans consumer loans consumer loans loans FX loans
Sep.08 0.69 0.60 3.57 3.07 1.45
Dec.08 1.07 1.02 4.34 3.66 1.97
Mar.09 1.92 1.84 5.62 4.58 3.03
Jun.09 2.37 2.45 6.99 5.63 3.83
Sep.09 2.90 3.66 8.53 7.05 4.83
Dec.09 3.06 4.58 9.88 8.29 5.45
Mar.10 2.97 5.13 10.07 8.45 5.65
Jun.10 3.35 6.27 10.95 8.87 6.53
Sep.10 3.68 7.19 11.89 9.33 7.29
Dec.10 3.90 7.84 11.08 8.70 7.54
Mar.11 4.18 8.45 10.77 8.50 7.87
Jun.11 4.41 9.19 10.13 7.68 8.10
Sep.11 4.51 9.67 10.70 7.95 8.43
Dec.11 4.53 10.36 10.08 7.28 8.62
Mar.12 5.18 10.99 11.54 7.97 9.45
Jun.12 5.69 12.07 12.11 8.15 10.11
Sep.12 5.87 12.56 11.49 7.80 10.12
Dec.12 5.96 12.73 12.01 7.77 10.32
Mar.13 5.99 13.26 13.01 8.80 10.69
Jun.13 6.05 14.03 13.62 8.89 11.12
Sep.13 5.80 13.89 13.53 8.75 11.01
Dec.13 5.70 13.31 13.17 8.95 10.83
Mar.14 5.85 14.36 13.36 8.65 11.39
Jun.14 5.73 14.83 12.06 7.36 11.46
*
The NPL ratio is calculated as the share of non-performing loans (more than 90 days overdue) to outstanding loans.
Source: Central Credit Register, Credit Bureau, MPF, NBR calculations
37
Chart A4.1 Developments of NPL ratio by monthly net income (June 2014)
percent lei bn
30 3
NPL for mortgage loans NPL for consumer loans

20 2
stock of NPL mortgage loans (rhs) stock of NPL consumer loans (rhs)
10 1
0 0
<=700
>7,000
(1,500;2,500]
(2,500;3,500]
(,3500;5,000]
(5,000;7,000]
(700;1,500]
Net monthly income
Note:
Note:The
The information includesonly
information includes onlyhouseholds
households with
with bankbank
loansloans
and and themonthly
the net net monthly
incomeincome does
does not not include
include co-borrowers. The
co-borrowers.
data
The on
dataincome is ofisDecember
on income 2013,
of December and
2013, onon
and credit
creditexposures
exposuresof
ofJune
June 2014. The coverage
2014. The coverageratio
ratioisisaround
around 75 percent of total
exposures
75 percentand 60 percent
of total of and
exposures non‑performing loans (in June 2014).
60 percent of non-performing loans (in June 2014).
Source: Central Credit Register, Credit Bureau, MPF, NBR calculations
Chart A4.2 Developments of NPL ratio for credit vintages (cut‑off date: June 2014)
percent
16
NPL ratio – total loans
(by vintage)
12
NPL ratio – mortgage loans
(by vintage)
8
0
2003Q1
2003Q2
2003Q3
2003Q4
2004Q1
2004Q2
2004Q3
2004Q4
2005Q1
2005Q2
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2007Q3
2007Q4
2008Q1
2008Q2
2008Q3
2008Q4
2009Q1
2009Q2
2009Q3
2009Q4
2010Q1
2010Q2
2010Q3
2010Q4
2011Q1
2011Q2
2011Q3
2011Q4
2012Q1
2012Q2
2012Q3
2012Q4
2013Q1
2013Q2
2013Q3
2013Q4
Note:
Note: The
The NPL ratioreflects
NPL ratio reflectsthe
theshare
shareofofnon-performing
non‑performing loans
loans to total
to total loansloans
and isand is calculated
calculated forquarterly
for each each quarterly
vintage vintage (starting
2003 Q1).
(starting ForQ1).
2003 comparability amongamong
For comparability vintages, a loana is
vintages, considered
loan non-performing
is considered non-performingif the borrower
if the borrowerdefaulted
defaultedon its loans at
least
on itsonce
loansinataleast
3‑year
onceperiod sinceperiod
in a 3-year the origination. Nevertheless,
since the origination. starting June
Nevertheless, 2011
starting Junethe interval
2011 of analysis
the interval of has decreased,
with the has
analysis NPL ratio reflecting
decreased, the developments
with the NPL until
ratio reflecting the June 2014until
developments (theJune
cut‑off
2014point). The point).
(the cut-off data have been corrected for
The data
have been corrected for rescheduled loans starting March 2005, when the identification became available. The dummy
rescheduled loans starting March 2005, when the identification became available. The dummy for the prudential regulation
for the prudential regulation refers only to measures related to LTV and DSTI.
refers only to measures related to LTV and DSTI.
Source: Central Credit Register, NBR calculations
38
Graph A4.3 LTV and NPL values for vintages of mortgage loans (June 2014)
percent percent
110 22
100 20
90 18
80 16
70 14
60 12
50 10
40 8
30 6
20 4
10 2
0 0
2005 2006 2007 2008 2009 2010 2011 2012 2013
LTV for mortgage loans (median values) NPL for mortgage loans (rhs)
Note: LTV values are calculated as of June 2014 (for all annual vintages) and therefore reflect the current collateral values.
Source: Central Credit Register, NBR calculations
Table A4.2 NPL* ratio by vintage and loan characteristics for households’ portfolio
DSTI* LTV**
NPL ratio for NPL ratio for NPL ratio for (median value for (median value for
Date all loans mortgage loans consumer loans all loans) mortgage loans)
2005 1.93 1.48 2.15 27.43 65.36
2006 4.93 2.79 5.60 29.57 51.96
2007 11.30 7.30 12.43 32.11 58.53
2008 14.81 11.72 15.75 41.66 75.27
2009 5.93 2.84 7.90 30.93 58.12
2010 4.14 1.22 6.43 32.88 62.37
2011 3.46 1.40 4.64 33.05 75.06
2012 2.87 0.68 5.67 30.45 64.07
2013 1.16 0.27 2.40 24.88 72.89
Note: The NPL ratio is calculated for each annual vintage and reflects the share of non‑performing loans to total loans. A loan
is considered non-performing if the borrower defaulted (more than 90 days overdue) at least once in a 3-year period since
the origination. Nevertheless, starting June 2011 the interval of analysis has decreased, with the NPL ratio reflecting the
developments until June 2014 (the cut‑off point). These figures are available for exposures higher than RON 20,000 alone.
The data are corrected for rescheduled loans.
*
DSTI represents the median value calculated for all loans originated every year between 2005 – 2013. It is calculated based on
the assumption of constant annuities, with outliers being excluded from the database.
where r is the annual interest rate; P is the credit value at origination; n is the original maturity of the loan (number of months), I is
the average monthly net income per debtor.
**
LTV (loan-to-value) represents the median value calculated for all loans originated every year between 2005 – 2013 (Eq.2). Due
to data limitation1, we use the collateral values from 2013 and correct them going back to the origination, using the residential real
estate price index. We assume that December 2013 figures represent the market value and all collateral values behave in a similar
way to the market (the same correction of real estate prices is applied to all loans):
Source: Central Credit Register, MPF, NIS, NBR calculations
1
The collateral data have been available in the Central Credit Register starting 2012.
39
Chart A4.4 LTV and DSTI from banks’ mortgage loans to households
LTV for new household mortgage loans granted LTV for outstanding household
in each quarter mortgage loans
percent percent
120 120
100 100
80 80
60 60
40 40
20 20
0 0
2008Q1
2008Q3
2009Q1
2009Q3
2010Q1
2010Q3
2011Q1
2011Q3
2012Q1
2012Q3
2013Q1
2013Q3
2014Q1
2008Q1
2008Q3
2009Q1
2009Q3
2010Q1
2010Q3
2011Q1
2011Q3
2012Q1
2012Q3
2013Q1
2013Q3
2014Q1
DSTI for new household mortgage loans granted DSTI for outstanding household
in each quarter mortgage loans
percent percent
80 80
60 60
40 40
20 20
0 0
2008Q1
2008Q3
2009Q1
2009Q3
2010Q1
2010Q3
2011Q1
2011Q3
2012Q1
2012Q3
2013Q1
2013Q3
2014Q1
2008Q1
2008Q3
2009Q1
2009Q3
2010Q1
2010Q3
2011Q1
2011Q3
2012Q1
2012Q3
2013Q1
2013Q3
2014Q1
Note: The figures represent the DSTI and LTV values from banks’ portfolios (and not the caps from the regulations).
The red line represents the average value of the indicators and blue lines represent the interval band (min, max)
of variation.
Source: National Bank of Romania (Bank Lending Surveys)
40
Annex 5 – Results of the econometric models on non-performing loan ratios
Table A5.1 Developments in the non-performing loan ratio for the households’ portfolio and for
sub-portfolios: the impact of self-regulation
(Eq. 1) (Eq. 2) (Eq. 3)

Aggregate NPL for NPL for non-mortgage
Aggregate NPL NPL ratio mortgage loans backed consumer loans
NPL ratio 0.987*** 1.011*** 0.927***
(t-1) (0.00) (0.00) (0.00)
Growth rate of 1.504*** 0.601*** 8.123***
unemployment (t-2) (0.00) (0.01) (0.00)
Growth rate of real estate -1.958*** -0.879**
index (t-2) (0.00) (0.02)
Financial expectations over -0.008*** -0.049***
the next year (t-4) (0.00) (0.00)
Change in local currency 0.517*** 5.081***
interest rate (t-2)1 (0.00) (0.00)
Change in FX interest rate 0.042 0.075
(t-2) (0.80) (0.76)
Leverage ratio 0.047*** 0.029*** 0.142***
(t-4) (0.00) (0.00) (0.00)
Self-regulation dummy2 0.359*** 0.279*** 0.765***
(0.01) (0.00) (0.00)
Dummy for -0.080***
Prima Casă loans3 (0.01)
Observations 658 658 658
Instruments 44 43 42
Hansen t-stat 25.650 24.279 26.350
Hansen dof 36.000 36.000 36.000
Hansen p-val 0.900 0.932 0.881
AR(1) 0.000 0.001 0.000
AR(2) 0.903 0.099 0.054
The estimation method is Blundell and Bond (1998) in a difference GMM approach. The endogenous variables used
are the lagged dependent variable and banks’ characteristic variable, together with the macroeconomic variables. No
restriction is imposed on the number of lags for the instrumental variables introduced into the regression. The p-values
are displayed in parentheses, where * p<0.10, ** p<0.05, *** p<0.01.
1
The impact stemming from the interest rate on loans denominated in local currency reflects also the increase in
margins associated with the period at the outbreak of the crisis. This variable does not influence the overall results.
2
The self-regulation dummy accounts for the changes in the prudential regulation that the NBR introduced in March
2007, moving from explicit limits on DSTI and LTV to self-regulation (March 2007 - September 2008).
3
The dummy for Prima Casă loans accounts for the different risk profile of this category as compared to other real
estate loans. It takes a value of 1 for all mortgage loans granted since 2009 Q2.
41
Table A5.2 Developments in the non-performing loan ratio for the households’ portfolio and for
sub-portfolios: the impact of self-regulation and of its interaction with macroeconomic variables
(Eq. 1) (Eq. 2) (Eq. 3)

Aggregate NPL NPL for mortgage NPL for non-mortgage
ratio loans backed consumer loans
NPL ratio 0.990*** 1.011*** 0.928***
(t-1) (0.00) (0.00) (0.00)
Interaction between unemployment 1.187*** 0.535** 7.672***
rate and lack of self-regulation in (0.00) (0.01) (0.00)
2007 (t-2)
Interaction between unemployment 2.786*** 0.846*** 9.910***
rate and self-regulation in 2007 (t-2) (0.00) (0.01) (0.00)
Growth rate of real estate index -1.894*** -0.871**
(t-2) (0.00) (0.02)
Change in local currency interest 0.524*** 5.076***
rate (t-2)1 (0.00) (0.00)
Change in FX interest rate 0.059 0.077
(t-2) (0.72) (0.75)
Financial expectations over the -0.008*** -0.049***
next year (t-4) (0.00) (0.00)
Leverage ratio 0.045*** 0.029*** 0.140***
(t-4) (0.00) (0.00) (0.00)
Self-regulation 0.338*** 0.276*** 0.744***
dummy2 (0.01) (0.00) (0.00)
Dummy for -0.081***
Prima Casă loans3 (0.01)
Observations 658 658 658
Instruments 45 44 43
Hansen t-stat 23.374 23.757 25.951
Hansen dof 36 36 36
Hansen p-val 0.948 0.942 0.892
AR(1) 0.000 0.001 0.000
AR(2) 0.968 0.092 0.054
We use the Blundell and Bond methodology (1998) in a difference GMM approach. The endogenous variables used are the lagged
dependent variable and banks’ characteristic variable together with the macroeconomic variables. No restriction is imposed on
the number of lags for the instrumental variables introduced into the regression. The p-values are displayed in parentheses, where
* p<0.10, ** p<0.05, *** p<0.01.
1
The impact stemming from the interest rate on loans denominated in local currency reflects also the increase in margins associated
with the period at the outbreak of the crisis. This variable does not influence the overall results.
2
The self-regulation dummy accounts for the changes in the prudential regulation that the NBR introduced in March 2007, moving
from explicit limits on DSTI and LTV to self-regulation (March 2007 – September 2008).
3
The dummy for Prima Casă loans accounts for the different risk profile of this category as compared to other real estate loans.
It takes a value of 1 for all mortgage loans granted since 2009 Q2.
42
Chart A5.1 The impact of self-regulation (of DSTI/LTV) on the NPL ratio for different types of loans
percentage points
5
0
After Cumulative After Cumulative After Cumulative
one quarter 2 years one quarter 2 years one quarter 2 years
Household loans Mortgage loans Non-mortgage backed
consumer loans
Note: The values are calculated based on the regression equation in Table A5.1. The results are similar to those calculated based on
figures in Table A5.2.
43
44

Implementing Loan-to-Value and Debt Service-To-Income Measures: A Decade of Romanian Experience

Uploaded by

Copyright:

Available Formats

Implementing Loan-to-Value and Debt Service-To-Income Measures: A Decade of Romanian Experience

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Implementing Loan-to-Value and Debt Service-To-Income Measures: A Decade of Romanian Experience

Uploaded by

Copyright:

Available Formats

Munich Personal RePEc Archive

Implementing Loan-to-Value and Debt

Neagu, Florian and Tatarici, Luminita and Mihai, Irina

National Bank of Romania

All rights reserved.

ISSN 1584-0867 (online)

Executive Summary ................................................................................................. 9

II. Using the instruments ....................................................................................... 11

A. Monitoring systemic risk ............................................................................ 11

B. Taking decisions .......................................................................................... 13

C. Calibration and dealing with “leakages” ..................................................... 14

III. Evaluating effectiveness ................................................................................. 16

A. Credit dynamics .......................................................................................... 16

B. Housing price dynamics ............................................................................. 19

C. Non-performing loan (NPL) dynamics ....................................................... 19

IV. Conclusions ..................................................................................................... 21

Annex 1 – Regulatory measures taken by the National Bank of Romania ........... 27

JEL classification codes: E44, E58, G21, G28

II. Using the instruments

2. Application and enforcement

C. Calibration and dealing with “leakages”

2. Dealing with policy “leakages”

III. Evaluating effectiveness

B. Housing price dynamics

C. Non-performing loan (NPL) dynamics

We construct a dummy variable for the self-regulation period, as following17:

Afanasieff, T.S., Implementing Loan-to-Value Ratios: The Case of Auto Loans in

Annex 1 – Regulatory measures taken by the National Bank of Romania

Period A. Prudential measures Credit cycle phases

The rapid acceleration of domestic bank credit (the

Period A. Prudential measures Credit cycle phases

Period A. Prudential measures Credit cycle phases

In the case of consumer loans, the

December Expansion of the measures to non-

Period B. Other measures with an impact on lending

June 2009 Launch of the Prima Casă government program.

Period C. Changes in the minimum reserve requirements ratios (MRR)

100 taking into account

Annex 2 – Credit dynamics

* Due to data availability issues, the impact of the prudential

credit – mortgage credit – lei

Table A2.1 Summary statistics

Variable Observations Mean Std. Min Max

Growth rate of credit, (total) 490 11.0 19.9 -39.9 151.2

Annex 3 – Results of the econometric models on credit growth

Table A3.1 Total loans to households

Eq. 1 Eq. 2 Eq. 3 Eq. 4

Table A3.2 Mortgage credit portfolio

Eq. 1 Eq. 2 Eq. 3 Eq. 4

Table A3.3 Consumer credit portfolio

Eq. 1 Eq. 2 Eq. 3 Eq. 4

Chart A3.1 The impact of DSTI/LTV regulation on credit growth rate

-10 -10 -10

-15 -15 -15

-20 -20 -20

Table A3.4 Granger causality test

Null Hypothesis: Observations F Statistics Probability

Changes in Mortgage Credit do not Granger Cause changes in