Bank Deposit Trends

Interest rate pass-through in Turkey and impact of global financial crisis: asymmetric threshold cointegration analysis.

1. Introduction

The outbreak of global economic and financial crisis starting in
2008 has attracted a lot of studies, which emphasize the importance of
the monetary transmission mechanism. To evaluate the performance of
monetary policy, interest rate transmission channel turns out to be
essential for conducting interest rate pass through analysis. Interest
rate passthrough is described as the degree and speed of adjustment in
banking retail interest rates to changes in monetary policy rates. High
interest rate pass-through defines the close link between monetary
policy rate and retail banking rate. Since individuals and financial
agents shape their investment, saving and consumption decisions with
respect to deposit and lending rates (banking retail rates), it is
important for central banks to influence the banking retail rates
through monetary policy actions. For this, central banks use monetary
policy rates to affect banking retail rates. Efficiency of this
mechanism is important especially for countries using interest rate
channel as the monetary policy transmission mechanism. Complete
pass-through means that changes in policy rates are totally transferred
to banking retail rates so monetary policy decisions can be implemented
successfully by central banks. Such a pass-through mechanism indicates
the effectiveness of interest rate channel in establishing price
stability and strong banking system.

Especially for inflation targeting countries, the relationship and
adjustment degree between policy rate and banking retail rates are
highly examined under the concept of interest rate pass-through (see
e.g. Kwapil, Scharler 2006; De Bondt 2005; Sander, Kleimeier 2004).
Turkey implemented implicit inflation targeting regime between 2002 and
2005 (Kara 2006). After that, explicit inflation targeting regime is
accepted as main monetary policy stance. Thus, there are two studies
focusing on interest rate pass-through in Turkey after the inflation
targeting regime conducted. Aydin (2007) previously analyzed the
interest rate pass-through in Turkey using linear cointegration. Aydin
(2007) uses data for individual banks operating in Turkish banking
system and classifies loan types to include corporate, housing, cash and
automobile loans. The data set includes all public, private, foreign,
investment and development banks. Thus, Aydin (2007) explores
cointegration relationships in panel data and reveals the sources of
heterogeneity in price setting behavior of banks in different types of
loans. The empirical results revealed that monetary policy decisions are
transmitted to loans market within a quarter. Housing loans are the most
responsive rates to changes in the policy rate whereas; cash and vehicle
rates are less responsive. Commercial loans were found to be adjusting
incompletely.

Ozdemir (2009) also investigated pass-through mechanism from money
market to retail rates between 2001 and 2006 in Turkey using symmetric
and asymmetric error correction method. The estimation results indicated
that retail rates adjust completely in the long run however, in the
short run loan rate is flexible compared to deposit rate.

Within this perspective, the aim of this study is to examine
interest rate pass-through mechanism in Turkey using asymmetric
threshold cointegration method to capture the asymmetric behavior (if
exists) of interest rate channel during the period December 2001 to
April 2011. The superiority of this method with respect to traditional
linear cointegration method is that asymmetric cointegration method
takes into account asymmetric adjustment process among the interest
rates, and nonlinearity in the pass-through mechanism due to asymmetric
information or market structure (Wang, Lee 2009). Thus, when we analyze
the interest rate pass-through mechanism in Turkey, we choose to employ
asymmetric cointegration method following Enders and Siklos (2001) to
observe (if there exists) asymmetric long run relationship between the
monetary policy rate and banking retail rates. Our study differs from
two previous studies, Aydin (2007) and Ozdemir (2009). We adopt
asymmetric cointegration methodology while Aydin (2007) applied linear
cointegration. The difference of our study from Ozdemir (2009) is
twofold. First, we differentiate the data set so that instead of one
aggregate loan rate we employed disaggregated retail rates. Second, our
data covers longer time period starting from December 2001 to April
2011, which also enables us to analyze the global financial crisis
period.

In our study, we employed cash, vehicle, housing and commercial
loan rates as well as monthly weighted average of 1-month, 3-month,
6-month and 12-month deposit rates as banking retail rates and interbank
overnight borrowing interest rate as the policy rate for the period
December 2001 to April 2011. The long run relationship between banking
retail rates and the policy rate is investigated using threshold
autoregressive (TAR) and momentum threshold autoregressive (MTAR)
methods. In order to analyze the impact of 2008 financial crisis, we
used dummy variables for the period October 2008 to September 2009
during which Turkey has experienced negative growth rates.

The organization of the paper is as follows: Section 2 gives
literature review. Section 3 presents brief information about Turkish
banking system. Section 4 describes the data and explains the
methodology used. The estimation results are stated in Section 5.
Finally, Section 6 concludes the study.

2. Literature review

It is important to assess nature and speed of adjustment of retail
banking interest rates to changes in policy rates for conducting an
effective monetary policy. Thus, interest rate pass-through attracted
great attention. For instance, De Bondt (2005) examined the interest
rate pass-through in the euro zone by employing vector error-correction
and vector autoregressive model. It was suggested that the pass-through
is complete up to 3 months but incomplete for rates with longer
maturities. After the introduction of euro, pass-through became faster.
Andion et al. (2010) carried out a cointegration analysis among nominal
and spread mortgage rates of member countries of European Union (EU) to
search for the degree of integration among EU money markets. The
estimation results indicated that there are no clear long term
relationships between EU countries’ mortgage series in 1995-2008
periods.

There are some studies focusing on interest rate pass-through
mechanism in individual European countries. For example, Egert et al.
(2007) found low and decreasing interest rate pass-through from policy
rate to retail rates five Central and Eastern European countries–the
Czech Republic, Hungary, Poland, Slovakia and Slovenia. However, the
cointegration and error correction analysis revealed that there are
still higher passthrough between policy rate and some corporate lending
rates. Bruna (2008) investigated the effect of market interest rates on
lending and deposit interest rate in Czech Republic alone during
inflation stabilization period 1999-2006. The results of cointegration
analysis suggested that lending and deposit rates react differently to
changes in market interest rates both in long run and in short run.
Changes in the banking sector influenced the adjustment mechanism among
the market, deposit and lending interest rates.

Similarly, Chionis and Leon (2006) analyzed the properties of
propogation mechanism from policy interest rate to lending and deposit
interest rates in Greece before and after the access to EU. Using
bivariate cointegration and error correction method, the study concluded
that lending and deposit rate react more to policy rate after entering
the EU however, the interest rate pass-through is still incomplete.

The interest rate pass-through process in Argentina is analyzed by
Humala (2005) with employing Markov switching vector autoregression
models. It was shown that financial crises change the adjustment level
of short term lending rates to changes in interbank rate.

Recent literature showed that most of the macroeconomic variables
such as inflation, national income, unemployment, interest rate display
asymmetric adjustment to their long run equilibrium level (see e.g.
Wang, Lee 2009; Sander, Kleimeier 2004; Enders, Siklos 2001). On the
other hand, Haug and Siklos (2006) showed that the nonlinear models are
superior to linear models in explaining the relationship between short
term and long term interest rates. By using several developed country
data and applying exponential smooth transition autoregression, they
showed that especially during the periods of policy changes, nonlinear
models are better in describing the behavior of short term interest
rates. Therefore, there are various studies analyzing either symmetric
or asymmetric adjustment properties of interest rate pass-through in
different countries.

For instance, Sander and Kleimeier (2004) made such an analysis for
the euro area using asymmetric cointegration method. After specifying
structural breaks, it was shown that there is divergence in the
pass-through mechanism of member nations although pass-through improved
in deposit market due to increased competition in post-break periods.
Later, Kleimeier and Sander (2006) studied interest rate pass-through in
the European Union by explicitly describing expected and unexpected
monetary shocks in their analysis. Their model was found to be
successful in describing the behavior of loan interest rates even under
asymmetric adjustment. However, for the deposit rates a slower
pass-through was observed.

Hofmann and Mizen (2004) depicted the nature of interest rate
pass-through in the UK. By allowing asymmetric and nonlinear adjustment
processes, it was shown that the gap between the policy rate and the
retail banking rates is the driving force of the interest rate
pass-through so that when this gap is high the adjustment pace is rapid
and when this gap is small the adjustment speed is slow. Later, Fuertes
and Heffernan (2009) attracted attention to variations in the adjustment
processes of financial firms’ interest rates to changes in policy
rate in the UK. Therefore, it was suggested that different retail rates
set by financial firms may decrease the efficiency of monetary policy
since consumers face distinct interest rates. However, the estimation
results are parallel to the forecasts that there is a long run
relationship between policy rate and retail rates.

Panagopoulos et al. (2010) explored pass-through from central
bank/money market rates to deposit and lending rates in the U.S.,
Canada, the UK and Eurozone. The cointegration analysis allowing also
asymmetric adjustment indicated that the UK and U.S. have positive long
run interest rate pass-through while Canada and Eurozone have negative
long run pass-through.

Gambacorta and Iannotti (2007) explored the asymmetric properties
of interest rate pass-through between policy rate and bank interest
rates in Italy during 1985-2002. The estimation results of asymmetric
error correction model showed that after 1993 degree of adjustment
increased, adjustment is asymmetric only in the short run, loan rates
are adjusted faster than deposit rates during monetary easing periods
and this asymmetry has disappeared since 1990.

Payne (2006) examined the long run relationship between the federal
funds rate and the 30-year mortgage rate in the U.S. using momentum
threshold autoregressive (MTAR) method following Enders and Siklos
(2001). It was documented that there is symmetric incomplete
pass-through between the interest rates in the long run. However, Cook
(2008) used Bardsen transformation to investigate the properties of
interest rate passthrough between federal funds rate and the mortgage
rate in the U.S. Contrary to Payne (2006), Cook (2008) discovered
complete pass-through.

Wang and Lee (2009) used the asymmetric threshold cointegration
method following Enders and Siklos (2001) and the EC-EGARCH (1, 1)-M
model to analyze pass-through mechanism between the money market rate
and the retail interest rate in the U.S. and nine Asian countries. The
complete pass-through was found only for the U.S. deposit rate,
supporting the work of Cook (2008). Asymmetric pass-through was seen in
the deposit interest rate in five Asian countries and in the lending
rate in three Asian countries. The symmetric cointegration relation was
found to be existing in two Asian countries. Wang and Thi (2010) used
asymmetric threshold cointegration test to analyze interest rate
pass-through in Taiwan and Hong Kong. It was reported that there exist
asymmetric and incomplete interest rate pass-through in both countries.

Lastly, Tai et al. (2012) examined the differences in the interest
rate transmission mechanism of six Asian countries, which are Hong Kong,
Indonesia, Korea, Malaysia, Philippines, Singapore and Thailand, using
seemingly unrelated regression equations. They concluded that
transmission from money market rate to deposit and lending rate is
sluggish signaling inefficiency in transmission mechanism of the
mentioned countries.

3. Brief information on the Turkish banking system

In 1999, an exchange rate based stabilization programme with
International Monetary Fund (IMF) was announced to manage inflation
expectations and decrease borrowing cost of Treasury. However,
remarkable credit growth together with foreign exchange risk due to lack
of commitment to structural reforms and delays in privatization
programme of public banks led to liquidity crisis in 2000-2001 in
Turkey. This was the end of 1999 stabilization programme. Devaluation
costed more than 4% of gross national product, more than 10 banks went
bankrupt and overnight interest rate increased beyond 15 000% (Gormez
2008).

After the 2001 crisis, together with inflation targeting regime,
re-structuring of public banks and re-capitalization of banks took place
in a new stabilization programme (Gormez 2008). Implicit inflation
targeting was implemented between 2002 and 2005, after which explicit
inflation targeting was realized (Kara 2006). Owing to firm
macroeconomic policies, the banking sector experienced a credit growth
although currency risk was a threat due to floating exchange rate
regime. However, the Central Bank of the Republic of Turkey was able to
manage capital flows through market operations. As a result, banks in
Turkey started to adapt themselves to low inflation business conditions
together with new financial instruments (Gormez 2008; Ozdemir 2009).

Banks in Turkey become more powerful after recovering from 2001
crisis and growth in banking sector was observed. Financial Stability
Report of 2007 prepared by the Central Bank of the Republic of Turkey
stated that approximately 87% of Turkish financial system was composed
of banks. According to statistics distributed by the Banks Association
of Turkey, there are 31 deposit banks, 13 development and investment
banks and 4 participation banks operating in Turkey as of December 2011.
Of 31 deposit banks, 3 are state-owned, 11 are privately-owned, 16 are
foreign banks and 1 is under the Deposit Insurance Fund. The total
number of branches belonging to deposit banks and development and
investment banks is 9834. Total number of employees working in all
deposit banks and development and investment banks is 181 443 as of
December 2011 (The Banks Association of Turkey Statistical Reports
2012). The Turkish banking sector has TL 1 212 374 million assets, TL
675 575 million loans, TL 287 212 million securities, TL 2109 million
profit at the end of 2011 (Banking Regulation and Supervision Agency
Press Release 2012).

4. Data and methodology

4.1. Data

To examine the interest rate pass-through mechanism in Turkey, we
employed cash, vehicle, housing and commercial loan rates as retail
rates and interbank overnight borrowing interest rate as the policy rate
(1). All data series are taken from Electronic Data Delivery System of
Central Bank of the Republic of Turkey. Our data set consists of the
period from December 2001 to April 2011.

As it is seen in Figure 1, although retail rates and policy rate
seem to have a general decreasing trend in common, some discrepancies
are observed during 2002, in the beginning of 2003, in the beginning of
2004, in the middle of 2006, during the second half of 2006 and in the
beginning of 2009.

4.2. Asymmetric cointegration and error correction (EC)

In order to examine dynamic adjustment properties of retail banking
rates to policy rate, we follow the threshold autoregressive (TAR) and
momentum threshold autoegressive (MTAR) procedures introduced by Enders
and Siklos (2001), to see whether there exists asymmetric long run
relationship between the stated interest rates. The long run
relationship between the banking retail rate and the policy rate is
estimated by

[b.sub.t] = [[mu].sub.0] + [[mu].sub.1][p.sub.t] + [e.sub.t], (1)

[FIGURE 1 OMITTED]

where b is retail banking rate and p is the policy rate. The degree
of interest rate passthrough is measured by the coefficient
[[mu].sub.1]. Empirical studies revealed that speed of interest rate
pass-through is slow in the short run. Similarly, various studies
carried out for different countries imply that there is not a complete
interest rate pass-through in the long run and the transmission speed is
slow. There are different studies indicating that degree and speed of
interest rate pass-through mostly depend on the type of interest rates
used and the nature of the monetary system of the country under
consideration.

Under perfect competition this coefficient is expected to be equal
to 1 however, there are some factors/conditions that violate this
equality and lead to interest rate stickiness and incomplete
pass-through. For instance, depth of financial markets and switching
costs affect the level of [[mu].sub.1] and cause it being less than 1 or
cost of asymmetric information, adverse selection and moral hazard bring
about a coefficient greater than 1. All of these factors result in
incomplete/over pass-through, which highly affects the effectiveness of
monetary policy (Wang, Lee 2009).

To investigate the degree of adjustment including the asymmetric
behavior specification between the banking retail rate and policy rate,
we used the TAR model of Enders and Siklos (2001) as follows:

[DELTA][e.sub.t] = [I.sub.t][[rho].sub.1][e.sub.t-1] + (1 –
[I.sub.t]) [[rho.sub.2][e.sub.t-1] + [[epsilon].sub.t], (2)

where e is the residual series of pass-through relationship,
[epsilon] is a zero-mean, constant-variance, iid random variable. I is
an indicator function specified as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)

where [tau] is the threshold value, which is naturally set to zero
in many economic applications. This way, the adjustment degree will be
[[rho].sub.1][e.sub.t-1] if there is positive deviation from the long
run equilibrium level and [[rho].sub.2][e.sub.t-1] if there is negative
deviation from the long run equilibrium (Enders, Siklos 2001). Then, in
this TAR model, null hypothesis of no cointegration can be tested by
[[rho].sub.1] = [[rho].sub.2] = 0 and null hypothesis of symmetric
adjustment can be tested by [[rho].sub.1] = [[rho].sub.2].

In equation (3), replacing the level of previous period’s
residual ([e.sub.t-1]) with the change in the level of previous
period’s residual ([DELTA][e.sub.t-1]) produces the momentum
threshold autoregressive (M-TAR) model with the following indicator
function:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (4)

TAR model is convenient to examine persistency of positive and
negative departures of the series from their long run equilibrium. The
asymmetric behavior in trough/peak movements of the series are detected
by TAR model. On the other hand, M-TAR model is rather used to capture
asymmetric sharp contractionary and/or expansionary movements of the
series when they diverge from their long run equilibrium.

The EC representation for TAR model is given as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (5)

where b is retail banking rate, p is the policy rate, e is the
residual series of pass-through relationship, [epsilon] is a zero-mean,
constant-variance, iid random variable and I is the indicator function
defined in equation (3). Similarly, the EC equation of MTAR model is
stated as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], (6)

where the only difference from the EC representation of TAR model
is the indicator function M. Long term effect of policy rate on retail
banking rates is observed through the coefficient of error correction
term e, and the significance of these parameters can be tested by
[[rho].sub.1] = [[rho].sub.2] = 0 while symmetric adjustment properties
can be examined by [[rho].sub.1] = [[rho].sub.2]. Short term
relationship between the retail banking rate, its lag and the policy
rate can be quantified by the coefficients [[rho].sub.3] and
[[rho].sub.4], respectively (2).

5. Estimation results

5.1. Unit root test

Before starting our analysis, unit root test is done to check
whether the series are stationary or not. The results of the Augmented
Dickey Fuller (ADF) test (see Dickey, Fuller 1979; Dickey, Fuller 1981;
Said, Dickey 1984) are shown in Table 1.

According to Table 1, the null hypothesis of level series have unit
roots cannot be rejected based on McKinnon (1996) p-values. However,
after taking the first differences, the ADF unit root test shows that
the series are stationary at 1% significance level so that all series
are integrated of order 1, I(1).

In addition to ADF test, Phillips-Perron unit root test due to
Phillips and Perron (1988) and Kwiatkowski, Phillips, Schmidt and Shin
(KPSS) test due to Kwiatkowski et al. (1992) are applied. Results of
Phillips-Perron test and KPSS test are given in Table 2 and Table 3,
respectively.

As it can be seen from Table 2, existence of unit root cannot be
rejected for all series however, when first differences of the series
are taken, the null hypothesis can be rejected at 1% significance level
pointing to stationarity.

Results of KPSS test in Table 3 show that the nul hypothesis of
stationarity is rejected at 1% significance level for all series.
However, stationarity hypothesis cannot be rejected for the first
differences of all series. As a result, all three unit root tests
indicate that the series under consideration are stationary after taking
first difference indicating that all are I(1).

5.2. TAR model

In order to examine the interest rate pass-through mechanism in
Turkey, we first regressed each banking retail rate on the policy rate
separately and then modeled residuals according to TAR specification
explained above. The results of cointegration test and symmetric
adjustment test are shown in Table 4 below.

In Table 4, the second and the third columns show the values of
adjustment parameters of [[rho].sub.1] and [[rho].sub.2] with their t
statistics in parenthesis. Larger of the t statistics is compared with
the critical values given in Table 2-Panel A of Enders and Siklos (2001)
(-2.11 for 5% significance level and -2.55 for 1% significance level).
The fourth column displays F statistic for the null hypothesis of no
cointegration ([[rho].sub.1] = [[rho].sub.2] = 0). The F statistic is
contrasted with the critical values given in Table 1-Panel A of Enders
and Siklos (2001) (5.98 for 5% significance level and 8.24 for 1%
significance level). The fifth column shows the test results for the
symmetric adjustment ([[rho].sub.1] = [[rho].sub.2]) by stating the F
statistic with the correspondent probability in parenthesis. It is found
that cointegration exists between the policy rate and the loan rates.
Besides, positive discrepancies from the long run equilibrium adjust at
the same speed with the negative discrepancies, leading to symmetric
adjustment for the loan rates except housing rate. The F statistic for
the symmetric adjustment test indicates that there is an asymmetric
adjustment between housing and policy rates.

As for the EC model, equation (5) is estimated for the loan rates
for which cointegration with the policy rate is found. The following
Table 5 presents the estimation results. Table 5 suggests that error
correction is significant (the coefficients p1 and p2) revealing that
there is long term relationship between loan rates and the policy rate.
Also, positive and negative discrepancies from the long run path
converge almost at the same speed to equilibrium. We cannot say that
positive or negative departures from the equilibrium path are more
persistent than the other. Pace of convergence is about 2-3 months for
all loan rates. As for the individual significance of the parameters
representing short run impact of the policy rate and the lag of retail
rate, only the effect of policy rate is found to be statistically
significant. However, column eight suggests that joint impact of the lag
of loan rate and the policy rate on the retail loan rate is significant
indicating the presence of short term relationship as well as the long
term.

Our results are similar to the findings of Aydin (2007) and Ozdemir
(2009) partially so that lending rates are sensitive to changes in
monetary policy rate. The results of the TAR estimation showed that the
interest rate pass-through from the policy rate to cash, vehicle,
housing and commercial loan rates is significant and speed of
convergence is almost a quarter. The adjustment processes for these loan
rates are symmetric, too.

5.3. MTAR model

We performed the same analysis using the MTAR specification to see
whether there is a difference, i.e. asymmetric behavior, between the
expansionary and contractionary divergences of the series from their
long run equilibriums. For this, we first replaced the indicator
function I with M in equation (2) to analyze the cointegration
relationship between the retail rates and the policy rate via MTAR
model.

Above Table 6 presents the test results of the cointegration
relationship. Similar to the previous TAR analysis, the larger of the t
statistics given in parenthesis in the second and third columns is
compared to the critical values given in Table 2-Panel B of Enders and
Siklos (2001) (-2.02 for 5% significance level and -2.47 for 1%
significance level). The F statistic in the fourth column is contrasted
with the critical values presented in Table 1-Panel B of Enders and
Siklos (2001) (6.51 for 5% significance level and 8.78 for 1%
significance level). Test of t and F statistics lead to rejection of no
cointegration so that, there is a long term relationship between the
policy rate and the loan rates. When we look at the adjustment speed of
expansionary and contractionary movements to the equilibrium path, it
can be said that recovery periods of positive and negative discrepancies
are about the same except commercial rate. This indicates that, either
positive or negative divergences are long-lasting over the other. The
fifth column supports this claim with the test results of symmetric
adjustment.

Next, the EC model with MTAR specification, which is given in
equation (6), is estimated for cash, vehicle and housing loan rates.
Table 7 below displays the results. Table 7 demonstrates that the
estimation results of the MTAR model are similar to that of TAR model.
There is a significant long run relationship between the policy rate and
the loan rates under the MTAR model, too. Both positive and negative
divergences from the equilibrium adjust to their long run trends at the
same speed indicated by the test results of symmetric adjustment in the
seventh column. In other words, contractionary movements are not faster
or slower than expansionary movements while returning back to long term
equilibrium path. The speed of adjustment is about 2-3 months. The short
term effect of policy rate on the loan rates is high and significant for
all loan rates. The significance of short term relationship between the
loan rate, and its lag and the policy rate is also implied by the test
results given in column eight.

The above results suggest that loan rates adjust to changes in
policy rate faster than deposit rates. Besides, positive and negative
shocks adjust symmetrically to long run equilibrium. Contrary to our
analysis, Becker et al. (2010) found asymmetric adjustment of mortgage
rates in the UK. Antao (2009) stated that deposit rates in Portuguese
adjust faster than loan rates.

5.4. Effects of financial crisis

In order to see the impact of financial crisis3, a dummy variable
is used for the period October 2008-September 2009, during which Turkey
experienced negative growth rates, and the same analysis is repeated.
However, the same results are observed at the end of this analysis. So,
it can be said that financial crisis does not have significant impact
over the long term course of interest rate pass-through in Turkey. The
estimation results of financial crisis period are not given here to save
space but they can be supplied from the authors upon request. However,
Hansen and Welz (2011) reached a different conclusion for Sweden. They
reported that although the pass-through from money market rates to
retail rates was complete and sluggish before the crisis, it did not
maintain for longer maturities after the crisis.

6. Conclusion

In this study, we examined the asymmetric adjustment properties of
retail banking interest rates to changes in monetary policy rate in
Turkey. Since monetary transmission mechanism is highly related to
efficiency of the implementation of monetary policy, analysis of
interest rate pass-through is important4. In fact, Kwapil and Scharler
(2006) mentioned that the adjustment degree between the monetary policy
rate and retail banking interest rates is important factor in
determining the relationship between monetary policy actions and
aggregate demand and inflation.

The empirical results of our analysis indicated statistically
significant correlation between policy rate and cash, vehicle, housing
and commercial loan rate. In other words, changes in the policy rate
have significant influence on loan rates. As for the adjustment
properties of discrepancies from the long term equilibrium, all retail
banking rate series adjusted symmetrically to changes in policy rate.
There is no asymmetry between positive and negative divergences in
recovering to equilibrium path. The adjustment speed is about 2-3 months
both for trough/peak and expansionary/contractionary movements. As for
the monthly weighted average of 1-month, 3-month, 6-month and 12-month
deposit rates, the same analysis was performed however, due to sluggish
adjustment of deposit rates to policy rate, any significant
cointegration relationship could not be reached for the deposit rates.
Additionally, the speed and size of the pass-through do not show
significant difference during the financial crisis. That is, symmetric
adjustment property of loan rates to changes in policy rate is preserved
during the financial crisis.

In order to make timely adjustments for policy rate and policy
decisions, for easing liquidity in money market and for a smooth
interest rate pass-through, central bank should take into account the
adjustment speed of banking retail rates to changes in policy rate. The
empirical results are consistent with the evidence so that, central bank
fixes a short term policy rate according to a Taylor-type rule and
conducts a short term monetary policy to keep actual interest rate
fluctuations low and money market liquidity high. Short term effect of
the policy rate on loan rates is found to be high and significant
especially for cash and vehicle credit rates indicating that central
bank has power over the loan rate market via changing the policy rate.
Thus, impact of central bank decisions on inflation and output is
considered to be high. On the other hand, in addition to monetary policy
decisions, it should be noted that while setting the loan rates, banks
take into account financial costs and risk premium, which depends on
state of the economic environment.

In fact, our estimation results are particularly pertinent for
banking sector dominated Turkish financial system. Besides, the
relationship between aggregate demand and retail banking interest rates
is important for an efficient monetary policy transmission mechanism and
macroeconomic stability.

doi:10.3846/16111699.2012.671189

References

Agenor, P. R.; Alper, K. 2009. Monetary shocks and central bank
liquidity with credit market imperfections, CBRT Research and Monetary
Policy Department Working Paper No: 09/6.

Andion, C. L.; Sanfiz, J. M. M.; Penabad, M. C. L. 2010.
Co-integration between mortgage markets in the monetary union:
1995-2008, Journal of Economics and Finance 60(1): 40-57.

Antao, P. 2009. The interest rate pass-through of the Portuguese
banking system: characterization and determinants, Banco de Portugal
Economics and Research Department Working Papers No: 5/2009.

Aydin, H. I. 2007. Interest rate pass-through in Turkey, CBRT
Research and Monetary Policy Department Working Paper No: 07/05.

Banking Regulation and Supervision Agency Press Release. 2012.
Turkish Banking Sector Non-Consolidated Main Indicators No: 2012/1.
Ankara.

Becker, R.; Osborn, D. R.; Yildirim, D. 2010. A threshold
cointegration analysis of interest rate pass-through to UK mortgage
rates, Discussion Paper Series 141. Centre for Growth and Business Cycle
Research, Economic Studies.

Bruna, K. 2008. Monetary policy stabilisation, changes in the
banking system and interest rate pass-through: the Czech Republic
1999-2006, Post-Communist Economies 20(4): 413-429.
http://dx.doi.org/10.1080/14631370802444575

Cook, S. 2008. Econometric analysis of interest rate pass-through,
Applied Financial Economics Letters 4: 249-251.
http://dx.doi.org/10.1080/17446540701704372

De Bondt, G. 2005. Interest rate pass-through: empirical results
for the euro area, German Economic Review 6: 37-78.
http://dx.doi.org/10.1111/j.1465-6485.2005.00121.x

Dickey, D.; Fuller, W. 1979. Distribution of the estimators for
autoregressive time series with a unit root, Journal of the American
Statistical Association 74: 427-431.

Dickey, D.; Fuller, W. 1981. Likelihood ratio statistics for
autoregressive time series with a unit root, Econometrica 49: 1057-1072.
http://dx.doi.org/10.2307/1912517

Egert, B.; Crespo-Cuaresma, J.; Reininger, T. 2007. Interest rate
pass-through in Dentral and Eastern Europe: reborn from ashes merely to
pass away?, Journal of Policy Modeling 29: 209225.
http://dx.doi.org/10.1016/j.jpolmod.2007.01.005

Enders,W.; Siklos, P. 2001. Cointegration and threshold adjustment,
Journal of Business and Economic Statistics 19: 166-176.
http://dx.doi.org/10.1198/073500101316970395

Fuertes, A. M.; Heffernan, S. A. 2009. Interest rate transmission
in the UK: a comparative analysis across financial firms and products,
International Journal of Finance & Economics 14: 45-63.
http://dx.doi.org/10.1002/ijfe.366

Gambacorta, L.; Iannotti, S. 2007. Are there asymmetries in the
response of bank interest rates to monetary shocks?, Applied Economics
39: 2503-2517. http://dx.doi.org/10.1080/00036840600707241

Gormez, Y. 2008. Banking in Turkey, Bank of Greece Working Paper
No: 83. Hansen, N. J. H.; Welz, P. 2011. Interest rate pass-through
during the global financial crisis, OECD Economics Department Working
Papers No: 855.

Haug, A. A.; Siklos, P. L. 2006. The behavior of short term
interest rates: international evidence of non-linear adjustment, Studies
in Nonlinear Dynamics & Econometrics 10(4): 1-32.
http://dx.doi.org/10.2202/1558-3708.1276

Hofmann, B.; Mizen, P. 2004. Interest rate pass-through and
monetary transmission: evidence from individual financial
institutions’ retail rates, Economica 71: 99-123.
http://dx.doi.org/10.1111/j.0013-0427.2004.00359.x

Humala, A. 2005. Interest rate pass-through and financial crises:
do switching regimes matter? The case of Argentina, Applied Financial
Economics 15: 77-94. http://dx.doi.org/10.1080/0960310042000297908

Kara, H. 2006. Turkish experience with implicit inflation
targeting, The Central Bank of the Republic of Turkey Working Paper No:
06/03.

Kleimeier, S.; Sander, H. 2006. Expected versus unexpected monetary
policy impulses and interest rate pass-through in euro-zone retail
banking markets, Journal of Banking & Finance 30: 1839-1870.
http://dx.doi.org/10.1016/j.jbankfin.2005.07.006

Kwapil, C.; Scharler, J. 2006. Interest rate pass-through, monetary
policy rules and macroeconomic stability, Oesterreichische Nationalbank
Working Paper Series No: 118.

Kwiatowski, D.; Phillips, P. C. B.; Schmidt, P.; Shin, Y. 1992.
Testing the null hypothesis of stationarity against the alternative of a
unit root: how sure are we that economic time series have a unit root?,
Journal of Econometrics 54: 159-178.

McKinnon, J. G. 1996. Numerical distribution functions for unit
root and cointegration tests, Journal of Applied Econometrics 11:
601-618. http://dx.doi.org/10.1002/(SICI)1099-1255(199611)11:6<601::AID-JAE417>3.0.CO;2-T

Ozdemir, B. K. 2009. Retail bank interest rate pass-through: the
Turkish experience, International Research Journal of Finance and
Economics 28: 7-15.

Panagopoulos, Y.; Reziti, I; Spiliotis, A. 2010. Monetary and
banking policy transmission through interest rates: an empirical
application to the USA, Canada, the UK and the Eurozone, International
Review of Applied Economics 24(2): 119-136.
http://dx.doi.org/10.1080/02692171003590096

Payne, J. E. 2006. The response of the conventional mortgage rate
to the federal funds rate: symmetric or asymmetric adjustment?, Applied
Financial Economics Letters 2: 279-284.
http://dx.doi.org/10.1080/17446540600647037

Phillips, P. C. B.; Perron, P. 1988. Testing for unit roots in time
series regression, Biometrika 75: 335-346.
http://dx.doi.org/10.1093/biomet/75.2.335

Said, S. E.; Dickey, D. 1984. Testing for unit roots in
autoregressive moving-average models with unknown order, Biometrika 71:
599-607. http://dx.doi.org/10.1093/biomet/7L3.599

Sander, H.; Kleimeier, S. 2004. Convergence in euro-zone retail
banking? What interest rate pass-through tells us about monetary policy
transmission, competition and integration, Journal of International
Money and Finance 23: 461-492.
http://dx.doi.org/10.1016/j.jimonfin.2004.02.001

Ucal, M.; Ozcan, K. M.; Bilgin, M. H.; Mungo, J. 2010. Relationship
between financial crisis and foreign direct investment in developing
countries using semiparametric regression approach, Journal of Business
Economics and Management 11(1): 20-33.
http://dx.doi.org/10.3846/jbem.2010.02

Tai, P. N.; Sek, S. K.; Har, W. M. 2012. Interest rate pass-through
and monetary transmission in Asia, International Journal of Economics
and Finance 4(2): 163-174. http://dx.doi.org/10.5539/ijef.v4n2p163

The Banks Association of Turkey Statistical Reports. 2012. The
Banking System in Turkey Quarterly Statistics by Banks, Branches and
Employees December 2011 Report Code: DE13. Istanbul.

The Central Bank of the Republic of Turkey. 2007. Financial
Stability Report No: 2. Ankara.

Wang, K. M.; Lee, Y. 2009. Market volatility and retail interest
rate pass-through, Economic Modelling 26: 1270-1282.
http://dx.doi.org/10.1016/j.econmod.2009.06.002

Wang, K. M.; Thi, T. B. N. 2010. Asymmetric pass-through and risk
of interest rate: an empirical exploration of Taiwan and Hong Kong,
Applied Economics 42: 659-670.
http://dx.doi.org/10.1080/00036840701704444

Ebru Yuksel [1], Kivilcim Metin Ozcan [2]

[1] Department of Industrial Engineering, Hacettepe University,
06810 Beytepe Ankara, Turkey [2] Department of Economics, Necmettin
Erbakan University, Karatay Konya, Turkey E-mails: [1]
eyuksel@hacettepe.edu.tr (corresponding author); [2]
kivilcim.metin@gmail.com

Received 19 August 2011; accepted 27February 2012

(1) At the beginning of this study, we also included monthly
weighted average of 1-month, 3-month, 6-month and 12-month deposit rates
in our analysis. However, it is observed that due to sluggish adjustment
of bank deposit rates (Agenor, Alper 2009) we could not find any
significant cointegration relationship with the policy rate, contrary to
Ozdemir (2009). We did not report the results to save the space. But
they may be requested from the authors.

(2) In TAR models, the adjustment parameters differ depending on
the position of the so-called indicator variable. Nonlinearities enter
only in the ECM equation, where it is assumed that nonlinearities
influence only the speed of adjustment towards the long run equilibrium.
The indicator variables can take various specifications, including the
size asymmetry, sign asymmetry and volatility asymmetry. However, this
paper only considers sign asymmetry and ignores size and volatility
asymmetries.

(3) See Ucal et al. (2010).

(4) See Metin Ozcan et al. (2003).

Ebru YUKSEL is an Instructor at Department of Industrial
Engineering at Hacettepe University, Ankara, Turkey. She received her
PhD from Bilkent University. Her research interests are applied
macroeconomics, time series analysis and finance. Her work has been
published in Physica A and Applied Economics Letters.

Kivilcim Metin OZCAN is a Professor of Economics at Necmettin
Erbakan University, Konya, Turkey. She received her PhD from the
University of Oxford. Her research interests are macroeconomic modeling
and time series analysis of aggregated macroeconomic and financial data.
Her work has appeared in Oxford Bulletin of Economics and Statistics,
Journal of Business and Economic Statistics, European Journal of
Operational Research, Applied Economics, Applied Financial Economics,
and Empirical Economics, Physica A, Emerging Market Finance and Trade.
She is a fellow of Economic Research Forum at Cairo, Egypt and a Board
member of Strategic Thinking Institute at Ankara, Turkey.

Table 1. ADF test results for all series

                            ADF Test Statistic

Level              Lag      Trend and Intercept

Cash                0              -3.01
Vehicle             11             -2.24
Housing             11             -1.82
Commercial          4              -2.51
Policy              1              -3.15

                            ADF Test Statistic

First difference   Lag      Trend and Intercept

Cash                0     -10.15 * (p-value 0.00)
Vehicle             10    -4.89 * (p-value 0.00)
Housing             10    -4.59 * (p-value 0.00)
Commercial          3     -5.11 * (p-value 0.00)
Policy              1     -5.62 * (p-value 0.00)

Note: * denotes the rejection of null hypothesis at
1% significance level. Lag lengths are chosen according
to Akaike Information Criterion (AIC). Critical values
are from McKinnon (1996).

Table 2. Phillips-Perron unit root test results for all series

                           Phillips-Perron Test Statistic

Level          Bandwidth            Trend
                                and Intercept

Cash               5                -3.01
Vehicle            1                -3.55
Housing            6                -2.30
Commercial         2                -3.71
Policy             5                -2.70

First
difference     Bandwidth     Trend and Intercept

Cash               3       -10.14 * (p-value 0.00)
Vehicle            4       -9.55 * (p-value 0.00)
Housing            2       -12.19 * (p-value 0.00)
Commercial         5       -13.44 * (p-value 0.00)
Policy             4       -7.60 * (p-value 0.00)

Note: * denotes the rejection of null hypothesis at
1% siginificance level. Bandwidths are chosen
according to Newey-West using Bartlett kernel.
Critical values are from McKinnon (1996).

Table 3. KPSS unit root test results for all series

                          KPSS LM Test Statistic

                                      Trend
Level           Bandwidth         and Intercept

Cash                8         0.27 * (p-value 0.00)
Vehicle             8         0.27 * (p-value 0.00)
Housing             8         0.26 * (p-value 0.00)
Commercial          8         0.26 * (p-value 0.00)
Policy              9         0.23 * (p-value 0.00)

First                                 Trend
difference      Bandwidth         and Intercept

Cash                5                 0.10
Vehicle             1                 0.08
Housing            14                 0.09
Commercial          0                 0.10
Policy              5                 0.17

Note: * denotes the rejection of null hypothesis at
1% siginificance level. Bandwidths are chosen
according to Newey-West using Bartlett kernel.
Critical values are from Kwiatkowski et al. (1992).

Table 4. Tests of cointegration and symmetric adjustment according
to TAR model ([tau] = 0)
                          [[rho].sub.1]      [[rho].sub.2]

Cash and policy           -0.30 (-3.38)     -0.23 ** (-2.44)

Vehicle and policy       -0.30 * (-3.19)     -0.36 (-3.72)

Housing and policy       -0.27 * (-3.09)     -0.68 (-9.03)

Commercial and policy    -0.34 * (-3.38)     -0.44 (-4.12)

                          [[rho].sub.1] =     [[rho].sub.1] =
                         [[rho].sub.2] = 0     [[rho].sub.2]

Cash and policy                8.70 *           0.25 (0.62)

Vehicle and policy            11.99 *           0.21 (0.65)

Housing and policy            45.54 *             12.96 *

Commercial and policy         14.19 *           0.56 (0.46)

                               Conclusion

Cash and policy           Cointegration exists
                          Symmetric adjustment
Vehicle and policy        Cointegration exists
                          Symmetric adjustment
Housing and policy        Cointegration exists
                          Asymmetric adjustment
Commercial and policy     Cointegration exists

Note: * indicates significance at 1%.

Table 5. Coefficients of EC models--TAR specification ([tau] = 0)

                  [[rho].sub.1]    [[rho].sub.2]    [[rho].sub.3]

Cash                 -0.35 *          -0.20 **          -0.12
  and policy         (-3.15)          (-1.99)          (-1.24)
Vehicle              -0.28 **         -0.44 *           -0.07
  and policy         (-2.19)          (-4.10)          (-0.67)
Housing              -0.43 *          -0.33 *            0.08
  and policy         (-4.16)          (-2.87)           (1.13)
Commercial and       -0.34 *          -0.23 **          -0.17
  policy             (-3.16)          (-2.14)          (-1.95)

                  [[rho].sub.4]   [[rho].sub.1] =     [[rho].sub.1] =
                                  [[rho].sub.2] = 0     [[rho].sub.2]

Cash                 0.88 *             6.85                0.99
  and policy         (4.41)            (0.00)              (0.32)
Vehicle              1.06 *             10.98               0.93
  and policy         (4.28)            (0.00)              (0.34)
Housing              0.50 **            12.94               0.36
  and policy         (2.53)            (0.00)              (0.55)
Commercial and       0.54 *             7.17                0.54
  policy             (3.62)            (0.00)              (0.46)

                   [[rho].sub.3] =       Conclusion
                   [[rho].sub.4] = 0

Cash                    11.71          EC exists and
  and policy            (0.00)           symmetric
Vehicle                 13.42          EC exists and
  and policy            (0.00)           symmetric
Housing                  7.13          EC exists and
  and policy            (0.00)           symmetric
Commercial and           6.72          EC exists and
  policy                (0.00)           symmetric

Notes: * indicates significance at 1%, ** indicates significance at 5%.

Table 6. Tests of cointegration and symmetric adjustment according
to MTAR model ([tau] = 0)

                           [[rho].sub.1]       [[rho].sub.2]

Cash and policy            -0.26 (-2.97)      -0.29 * (-2.96)

Vehicle and policy        -0.32 * (-3.16)      -0.30 (-3.24)

Housing and policy         -0.31 (-3.52)      -0.28 * (-2.64)

Commercial and policy      -0.16 (-1.44)       -0.43 (-5.02)

                          [[rho].sub.1] =      [[rho].sub.1] =
                         [[rho].sub.2] = 0      [[rho].sub.2]

Cash and policy                8.81 *            0.07 (0.80)

Vehicle and policy            10.24 *            0.02 (0.88)

Housing and policy             9.68 *            0.04 (0.85)

Commercial and policy         13.62 *            3.77 (0.06)

                               Conclusion

Cash and policy          Cointegration exists
                          Symmetric adjustment
Vehicle and policy       Cointegration exists
                          Symmetric adjustment
Housing and policy        Cointegration exists
                          Symmetric adjustment
Commercial and policy     Symmetric adjustment

Notes: * indicates significance at 1%, ** indicates significance at 5%.

Table 7. Coefficients of EC models - MTAR specification (t = 0)

                        [[rho].sub.1]    [[rho].sub.2]   [[rho].sub.3]

Cash and                    -0.26*          -0.27*           -0.13
policy                     (-2.61)          (-2.61)         (-1.36)
Vehicle                     -0.36*          -0.39*            0.06
and policy                 (-2.87)          (-3.78)         (-0.59)
Housing                     -0.45*          -0.30*            0.08
and policy                 (-4.64)          (-2.80)          (1.15)
Commercial and policy      -0.29**          -0.29*           -0.15
                           (-2.47)          (-3.07)         (-1.81)

                        [[rho].sub.4]     [[rho].sub.1] =
                                         [[rho].sub.2] = 0

Cash and                    0.96*               6.29
policy                      (5.07)             (0.00)
Vehicle                     0.96*              10.45
and policy                  (4.16)             (0.00)
Housing                     0.51*              13.51
and policy                  (2.80)             (0.00)
Commercial and policy       0.57*               6.87
                            (3.95)             (0.00)

                        [[rho].sub.1] =      [[rho].sub.3] =
                          [[rho].sub.2]     [[rho].sub.4] = 0

Cash and                      0.00                15.99
policy                       (0.97)               (0.00)
Vehicle                       0.05                13.62
and policy                   (0.82)               (0.00)
Housing                       1.26                 8.03
and policy                   (0.26)               (0.00)
Commercial and policy         0.00                 7.81
                             (0.99)               (0.00)

                          Conclusion

Cash and                EC exists and
policy                    symmetric
Vehicle                 EC exists and
and policy                symmetric
Housing                 EC exists and
and policy                symmetric
Commercial and policy   EC exists and
                          symmetric

Notes: * indicates significance at 1%, ** indicates significance at 5%.