The Monetary Policy in the Euroarea, United Kingdom
and the USA: Evidence from financial crisis period
George P. Kouretas*, Nikiforos T. Laopodis** and
Evangelos N. Salachas*
Abstract
The paper examines the monetary policy implementation in the USA, UK and Euroarea
based on a macrofinance framework. We cover an extended period, over two decades, in
order to study effectively the monetary policy transmission mechanism
contemporaneously. The analysis is elaborated on Vector Autoregression procedure and we
include both interest rates and economic activity factors. Our framework involves USA, UK
and Eurozone as the monetary policy entities. We analyze separately Germany, France and
Italy in order to control for any heterogeneity issues after the common monetary policy
implementation. The results suggests that the common Eurozone monetary policy decisions
have different effects on countries macroeconomic variables after the financial crisis
outbreak, especially for Italy. In our model, we treat financial crisis as a structural break in
the overall sample and we examine the effects on economic activity. As an additional
contribution, our model assess the effects of non-standard measures adopted by central
banks against financial turmoil. The results indicate, that unconventional monetary
measures boosted economic activity, mainly for USA. Finally, we incorporate an analysis
about the performance of monetary policy transmission channel comparing the key
monetary policy rates and prime bank lending rates movements. As we found, after the
financial crisis period the channel has significantly distorted.
Keywords: monetary policy, VEC model, financial crisis, unconventional measures,
transmission mechanism, economic activity.
JEL codes: E51,E52,E58
* Department of Business Administration, Athens University of Economics and Business
** Alba Business School
1. Introduction
The aim of the paper is the analysis of the monetary policy and the evaluation of the
monetary policy transmission mechanism before, during and after the financial crisis of
2007-2009. The objective is determined by the need to render the monetary policy more
efficient concerning the financial markets and the real economy. The monetary authorities
must have the sufficient information and influence in order to make optimal monetary
policy choices and attained the desired effects on the macroeconomy. However, due to the
structural differences, the effects of monetary policy decisions may vary across countries.
This paper attempts to investigate the monetary policy implementation and the extent
to which the transmission mechanism altered after the financial crisis. In addition particular
attention is given to the effects of unconventional monetary policy measures in real
economic activity and the financial markets. Also, a further issue is the study of the impact
of monetary policy decisions to country specific heterogeneity and the level of the global
integration in response to monetary policy shocks. In more detail, the analysis of this paper
is focused on the study of European Central Bank (ECB), the Bank of England and the Federal
Reserve actions as central monetary authorities. The countries that we analyze are USA
(fed), UK (BoE), and Germany, France, and Italy (ECB).
Our selection for these countries is based on the fact that their monetary authorities give
the signal for the global monetary policy decision making and their stance affects the global
central banking. Concerning the European Monetary Union (EMU), we choose Germany,
France and Italy, in order to examine the specific heterogeneity in EMU, and how the impact
of monetary policy of ECB has affected by financial and macroeconomic indicators in each
country. For euroarea, the financial crisis of 2007-2009 was followed by the sovereign debt
crisis beginning at the end of 2009.
2
The monetary authorities traditionally use the determination of the level of short term
interest rate to implement their policies, as their key conventional tool. The fundamental
model of the monetary policy decision making is proposed by Taylor (1993). According to
this model, the monetary authorities focus both on minimizing the inflation and output gaps
in order to implement efficiently their policies. However, the information about the
economy may be imperfect and becomes available with a time lag and this may distort the
time consistency of monetary policy decisions.
Our econometric strategy involves the analysis of a Vector Error Correction model on a
time series specification framework that treat the selected variables as endogenous. The
analysis is elaborated in three stages. The first stage includes the main empirical analysis
comprised of the benchmark VEC model, the second stage is focused on the effects on the
monetary policy mechanism after the imposition of the financial crisis structural break, and
in the third stage we set a comparison of some key macroeconomic variables between the
USA, UK, France and Italy with Germany as the benchmark.
By performing the benchmark VEC analysis we observed a high level of integration in
global markets along with increased convergence in Euroarea before the introduction of
euro. We found different behavior of monetary authorities concerning the output and
inflation targeting. Especially, our analysis showed that the inflation rates affect positively
the key monetary policy rates. In case of industrial production, the USA authorities give
greater attention on minimizing the output gap, while the ECB give little significance on
output growth. As we showed, the bank profit spread increased after the rise in key rates
indicating that the monetary transmission worked. In addition, for USA, UK, Italy the
response of inflation to increase in key rates are positive indicating existence of the socalled price puzzle.
3
As a special case, we study the impact of the financial crisis period to our variables
concerning the monetary policy stance. As we found, the transmission mechanism has
changed after the financial turmoil. In addition, our results highlighted the European
heterogeneity in the variables movements which stirred up the Eurozone debt crisis. By
elaborating the study of the effect of unconventional measures of central banks we
observed, that the impact of the non-standard measures in economy was positive and
improved the economic conditions. Furthermore, we follow an innovative measure of the
effect of monetary policy transmission mechanism by performing VAR analysis between the
key monetary policy rates and the prime bank lending rates. Comparing the periods before
and after crisis we observed a significant distortion in the monetary policy transmission
channel due to the effect of financial crisis.
Our paper differentiates from the main literature and includes the following innovations:
First, we included a wide range of macroeconomic and financial variables compared to the
basic literature. Second, by applying our analysis we study the eurozone heterogeneity in
main leading economic factors. Third, our sample covers an extended period (from 1990 to
2012) and this allows us to examine the monetary policy before during and after the
financial crisis. Fourth, we impose a structural break in VEC analysis, in order to have a clear
view of the structural change after financial crisis. Fifth we measure the effect of
unconventional tools implemented by central banks in economy. Sixth, we confirm our basic
results with some innovative robustness tests.
The paper is organized as follows. Section one is the introduction in the general
specification of the monetary policy analysis. Section two gives an overview of the
literature. Section three presents the data and the applied methodology. Section four
contains the preliminary analysis we elaborate and section five conducts the empirical
analysis, reporting the basic estimation results. In section six we analyze the structural break
4
due to financial crisis and section seven controls the main estimation results with
robustness tests. Finally, section eight is the concluding remarks.
2. Literature Review
The prevalent literature examines the relationship with the key monetary policy
measures with other market interest rates (mainly bond market rates), or macroeconomic
factors (such as inflation). The methodological analysis in literature consists of: On one hand
regression models, which defines the exogenous and endogenous variables and control the
effect of exogenous variables to the endogenous ones. On the other hand, Vector
autoregressive (VAR) models which stress all the variables as endogenous and regress one
with another simultaneously. In our specification, we use VAR model as it better predicts
the evolution of the variables and we can perform some structural breaks.
Hamilton, Kim (2000) measure the usefulness of modeling the yield curve in order to
predict future GDP growth. They found that the contribution of the contribution of the
spread can be decomposed into the effect of expected future changes in short rates and the
effect of the term premium. They found that while volatility displays important correlations
with both the term structure of interest rates and GDP, it does not appear to account for
the yield’s spread usefulness for predicting GDP growth.They also showed that yield curve
has flattened or become inverted prior to all seven recessions.
Peersman, Smets (2001) use a benchmark VAR-model to analyze the effects of a
monetary policy shock in the euro area by using a vector of endogenous euroarea variables,
and a vector of exogenous (USA) variables. Their results show that a temporal rise in the
nominal and real short term rates tend to be followed by real appreciation of the exchange
rate and a temporary fall in output. As a result, a monetary tightening leads to the fall in
investments and consumption, but in increase of net exports due to the reduction in
5
internal demand. The authors also investigated the reaction of other macro variables and
the GDP components to a monetary policy shock.
Ang, Boivin, Dong (2008) estimate the effect of shifts in monetary policy using the term
structure of interest rates. The authors use a no-arbitrage model similar to Taylor (1993)
rule, and they use quarterly data from 1952 to 2006. They found that monetary policy
loadings on inflation, but not output, changed substantially over the last 50 years. As they
mention, agents tend to assign a risk discount to monetary policy shifts and are willing to
pay to be exposed to activist monetary policy.
Abassi, Linzert (2011) introduce a regression model for Euribor rate evolution in order to
examine the effectiveness of monetary policy in steering money market rates. As the
Euribor rate, is affected by the EONIA rate, the authors tried to examine if the monetary
policy transmission mechanism applied affectively during financial crisis. The authors use
daily data from 2004 to 2009 and they built the model for euribor evolution based on
money market variables, risk proxies and monetary measures variables.
Peersman (2011) estimated a Structural VAR model in order to examine the
unconventional monetary policy actions in euroarea. The author used some innovations and
showed that the increase in central bank balance sheet as a part of unconventional policy
strategy, had a humped shape effect on economic activity and a permanent impact on
consumer prices.
Herro, Murray (2011) use a model similar to Taylor, by using OLS estimation, they tested
the impact of uncertainty on output inflation with VAR model and ARCH shocks. However,
they failed to find evidence that this uncertainty affects output, inflation, employment
indicating high persistence. But their results, significantly explained the changing volatility of
unemployment, output and inflation.
6
Cecioni, Neri (2012) estimate a Bayesian VAR model over the periods before and after
1999 and suggests that the effects of a monetary policy shock on output and prices have not
significantly changed over time. The estimation of a DSGE model with several real and
nominal frictions over two subsamples indicated that monetary policy has become more
effective in stabilizing the economy as the result of a decrease in the degree of nominal
rigidities and a shift in monetary policy towards inflation stabilization.
2. Data and methodology
Our dataset is comprised of monthly financial and macroeconomic variables for six
countries. We separate the countries set with the criterion if they have independent
monetary authorities or no. So, the one set includes United Kingdom, United States and
Euroarea which we called them “big” countries, and the other set involves Germany, France
and Italy, which we called them “small” countries. The dataset covers the period from
January 1990 to August 2012. The basic sources for data were FRED (Federal Reserve
Economic Database), ECB Statistical Data Warehouse, Bank of England, Banka di Italia,
Banque de France, ECB, Federal Reserve, and Datastream, Bloomberg as databases.
The variables we used as proxies for the key monetary policy rates are, the EONIA for
Euroarea, the Sonia for UK and the Effective Fed funds rate for USA. We apply also the yield
spread as the difference between the ten year bond rate and the three month treasury bill,
as we wish to study the effect of key rates to the bond rates for “big” countries. For “small”
countries we use the yield spread as a proxy monetary policy instrument as the short term
three month Treasury bill rate is closely related to Eonia. Additionally, we use the bank
profit spread (the difference between the bank lending and deposit rates) as main market
rates in order to study the transmission mechanism of monetary policy. The yield spread
and bank profit spread are applied for UK, USA, Germany, France and Italy.
7
We use also the unemployment rate for all the set of countries and the exchange rates of
dollar to one euro, pound to one euro and dollar to one pound to have a broader view for
the results. The variables above are analyzed by their first differences. Furthermore, we
include in our analysis the inflation rate and the industrial production in order to examine
the impact of monetary policy to prices and output respectively. The inflation rate is
computed as the logarithmic change of CPI (Consumer Prices Index), and the industrial
production growth is defined by the logarithmic change in industrial production index. So,
the variables transformations are:
Our econometric strategy is based on time series analysis by applying unrestricted Vector
Autoregressive model. VAR models were popularized in econometrics by Sims (1980) and
Litterman and Weiss (1984) as a combination of univariate time series model and
simultaneous equations models and they are used as they capture the linear
interdependencies among multiple time series.
The VAR models are widely used due to some specific advantages they have. First, the
researcher does not need to specify which variables are endogenous or exogenous, as they
are all treated as endogenous. Second, the unrestricted VAR models examine the impact in
variables from the innovations by other variables and study their behaviour. Third, the VAR
model allows the value of a variable to depend on more than its own lags or combinations
of the white noise term. We choose the VAR model as it can better explain the monetary
policy transmission mechanism, and we can apply shocks by impulse response function to
investigate the effects of the variables innovations. Our model specification based on
Peersman, Smets (2001) has the following representation:
k
Yt  a   biYt i   t
i 1
8
(1)
where Y’s are vectors of endogenous I(1) variables, αis a p*1 vector of constants, b, is a
p*p matrix of parameters to be estimated, the εrepresents an uncorrelated vector of
disturbances, p*1, and k is the order for the vector of variables X. The matrix representation
is:
Y1t 
Y 
 2t 
.. 
Yt  

.. 
.. 



Yn t 

 1t 
 
 2t 
.. 
t  

.. 
.. 



 n t 

 a1 
a 
 2
.. 
a 
.. 
.. 
 

an 

 b1 1
b
 21
 ..
bi  
 ..
 ..


bk 1
Yt i
Y1t i 
Y

 2 t i 
..



..

..




Yn ti 

b1 2
..
..
..
b2 2
..
..
..
..
..
..
..
..
..
bk 2
..
..
..
b1n 
b2 n 

.. 

.. 
.. 

bkn 

Under the regularity conditions, the vector of b’s must satisfy the following orthogonality
conditions:
a) E ( t )  0
Every error term has mean zero.
b) E ( t  t ')  
The contemporaneous covariance matrix of error terms is Ω (positive matrix).
c) E ( t  t  j ')  0
There is no serial correlation across time: no serial correlation of error terms.
In order to decompose the ε vector of errors, we imply the Choleski decomposition which
implies that the b’s matrix is a lower triangular. By performing the Choleski decomposition,
we place the variables in a specific order, with the innovation to the first variable is not
contemporaneously affected by the innovation to any of the other variables, the second
variable is assumed not to be affected from the others and so on. The determination of the
9
optimal lag length is based on the Final Prediction Error (FPE) or Akaike information
criterion (AIC). The lag length we chose is the one which minimizes the FPE or AIC criteria.
The next step in our analysis is to examine if they are cointegrated, that is the existence
of linear interdependencies among the variables used. If the variables are cointegrated then
their relationship can be analyzed in the Vector Error Correction framework. To examine the
cointegration we perform Johansen Cointegration test which yields two likelihood ratio
statistics for the number of cointegrating vectors, the maximum eigenvalue and the trace
statistics. To apply this test, we have to transform the VAR model into VEC model, so the
above VAR model becomes:
k 1
Yt  a  Yt 1    i Yt i   t
i 1
  i 1 bi  I g  i 
k
where
,
 b  I
k
i 1 i
g
, and the above variables are represented
by the vectors below:
Y1t 
Y 
 2t 
.. 
Yt  

.. 
.. 


Ynt 
 11i .. ..
 .. ..
 21i
 ..
..
i  
 ..
 ..

 k1i .. ..
(2)
a1 
 11  12 ..
a 

 21  22 ..
 2
.. .. ..
.. 
a  
.. ..
.. 
.. ..
.. 

 
 k1  k 2 ..
a n 
.. .. ..  1n 

.. .. .. 2 n 
.. 

..
.. 
.. .. 
.. .. ..  kn 
.. .. ..  1ni 
Y1t i 
 1t 



 
.. .. ..  2 ni 
Y2t i 
 2t 
..

.. 
.. 
 Yt  j  
 t   
..
.. 
..

.. 
..

.. 
.. .. 



 
.. .. ..  kni 
nti 
 nt 
10
where, a is a vector of constants, ΔY are the vectors of endogenous variables and εis a
vector of shocks or innovations of the model and Π’s are p*p vectors of coefficients.
Our analysis is based on Peersman, Smets (2001) work but it differentiates as our
specification develops a VEC model as it better explains our results and takes account the
cointegration in the sample variables. By applying the VEC analysis, we determine the order
of the specific variables. Also, our framework includes additional variables such as
unemployment rate and exchange rates in order to have a broader view of our results.
3.1 Johansen Cointegration test
The VEC model contains g variables in first differenced form, and k-1 lags of the
dependent variables, each with a Γcoefficient matrix attached to it. There are two test
statistics for cointegration under Johansen approach which are formulated as:
g
^
trace (r )  T  ln(1  i )
i  r 1
(3)
^
 (r , r  1)  T ln(1   r 1 )
and max
(4)
^

where r is the number of cointegrating vectors under the null hypothesis and i is the
^

estimated value for the ith ordered eigenvalue from the Πmatrix. The larger the i , the
^
more large and negative will be
ln(1  i )
and hence the larger will be the test statistics.
Each eigenvalue will have associated with it a different cointegrating vector, which will be
eigenvectors. A significantly non zero eigenvalue indicates a significant cointegrating vector.
The λtrace is a joint test where the null is that the number of cointegrating vectors is less
than or equal to r against an unspecified of general alternative that there are more than r.
the λmax conducts separate tests on each eigenvalue and has as its null hypothesis that the
11
number of cointegrating vectors is r against an alternative r+1. If the test statistic is greater
than the critical value from the Johansen’s tables, reject the null hypothesis that there are r
cointegrating vectors in favor of the alternative that there are r+1 for λtrace or more than r
for λmax.
4. Preliminary statistical investigation
4.1 Descriptive Statistics
We start our analysis by the descriptive statistics investigation of some key variables that
include key monetary policy rates, inflation rate and industrial production growth (table 1).
For key rates, the results show that the higher mean value belongs to Sonia rate, due to the
increased observations at the beginning of the 90s. Owing to the higher Sonia rate values its
standard deviation is the greatest among the others. In case of Euroarea and USA the mean
values are around 3 percent. Our proxy variable for inflation is Consumer Price Index. The
higher level of inflation is remarked in Italy with 3.13 percent mean followed by USA 2.56,
UK 2.46, France 1.85 and Germany 1.53 percent. The inflation observations after financial
crisis dropped significantly.
The descriptive statistics for industrial production growth showed that the higher mean
value is remarked in USA with 2.2 percent. It is followed by Germany with 1.33 percent and
overall Euroarea with 1.06 percent. We observe that in France the mean value is 0.15
percent and in case of Italy and UK we observe that the mean values are negative -0.27 and
-0.07 percent respectively. During the financial crisis the industrial production growth was
at its minimum values for all the set of countries, and the standard deviation values are
increased due to this reason. The results show also that in case of Euroarea the industrial
production growth is determined by Germany industry sector. None of the industrial
production rates is normally distributed as the Jarque-Bera statistic show.
12
4.2 Unit root tests
We then proceed to the imposition of unit root tests for all the variables and countries in
order to examine if they are stationary in order to make VAR analysis and multiple
regressions. The unit root tests we elaborate are Augmented Dickey-Fuller and KwiatkowskiPhilips-Schmidt-Shin for comparison (table 2). We choose 6 lags maximum and the Akaike
information criterion. Firstly, we do the unit root tests for the variable levels. If we take the
first differences for all the variables the unit root hypothesis is rejected and the variables are
stationary. For the CPI and M2 variables it was needed to take the second differences so as
to make them stationary. We report the unit root tests for some key variables. We report
the unit root tests for some key variables.
We apply the Augmented Dickey Fuller test in order to examine if the variables do not
have unit root that means they are stationary. The ADF test we impose has the following
representation:
m
X t   0   1   X t 1   ai X t 1   i
(5)
i 1
where
X t 1  X t 1  X t 2
The determination of time lag m in the autoregression procedure
 a X
i
t 1
takes place
with the acceptance of the statistical criterion (Akaike, FPE). The null hypothesis is β=0
against the alternative β<0. If we accept the null the X has unit root and it is non-stationary.
If we reject the null, we accept the alternative and X is stationary. If the test shows that X is
non-stationary we control the first difference of X, ΔX variable (
X t  X t 1
). We also apply
the Kwiatkowski-Phillips-Schmidt-Shin test for comparison, in which the null hypothesis is
β<0 and the alternative is β=0. By performing the ADF unit root tests we observe that all the
13
set of variables for all the included countries are non-stationary. However, if we take the
first differences for all the variables the unit root hypothesis is rejected and the variables
are characterized by stationarity.
We also perform KPSS unit root tests for cross examination of the results. The KPSS for
the level of the variables show that they are non-stationary, as we expected. By elaborating
the KPSS tests for the first differences we observe that nearly for all the set of variables the
null hypothesis of one unit root is rejected, indicating stationarity.
4.3 Johansen cointegration tests results
We test for cointegration in the sample by using the Johansen (1994) multivariate
coingration technique with the null of no cointegration against the alternative of
cointegration (table 3). For Euroarea, USA, Germany and Italy the results show that the
null hyothesis of no cointegration is rejected for all equations in the VEC model as both
trace statistic and maximum eigenvalue show. This fact implies the existence of one
common stochastic trend for the endogenous variables. For UK the trace statistic
indicates strong cointegration, however, maximum eigenvalue reveals no shared
stochastic trend. In case of France the trace statistic shows weak cointegration, but the
maximum eigenvalue depicts strong cointegration.
5. Main empirical findings
5.1 Variance decomposition analysis
In this section, we set up the benchmark VEC analysis by the variance decomposition
analysis. Our goal was to attain the exact evolution of the monetary policy transmission
mechanism and the effect the financial crisis of 2007-2009 on the mechanism.
14
Firstly we choose the variables and we perform the Johansen tests which have showed
that our variables are cointegrated. As a result we proceed to the VEC model specification.
We set up our analysis by the selection of the optimal lag length in order to establish and
estimate the appropriate model for each country. The first group includes the “big”
countries which are USA, UK and the Euroarea as an entity. The second group contains the
“small” countries, Germany, France and Italy, which belongs to the euro currency union. By
applying this framework we are able to test the country specific heterogeneities after the
implementation of monetary policy by ECB.
Our main empirical analysis is based on the reported results from the variance
decomposition analysis we elaborated. The variance decomposition analysis is used to aid in
the interpretation of VAR and VEC models once they have been fitted, and indicates the
amount of information each variable contributes to the other variables in the
autoregression. In other words, it determines how much of the forecast error variance of
each of the variables can be explained by exogenous shocks to the other variables.
Specifically, according to our VEC analysis, the variance decomposition of key monetary
policy rates in “big” countries reveals that their forecast variance derives from their own
evolutions. For Euroarea, a contribution to the variance decomposition of Eonia is given by
the exchange rate dollar/euro.
For the inflation rate, a significant commitment to forecast the variance is offered by the
key monetary policy rates (EONIA for Euroarea, Fed Funds rate for USA and Sonia for UK). In
USA especially, the fed funds rate affects by nearly 12 percent the variation of inflation rate.
The corresponding percentage for the Euroarea is 3 percent. These results indicate that the
monetary policy transmission mechanism performs better in USA than Eurozone. For the UK
the Sonia rate affects nearly by 7 percent the inflation evolutions. For the “small” countries,
the variance decomposition analysis tables showed that the main effects in inflation are
15
derived from their past evolution, except from France, where significant role play the bank
profit spread and the yield spread in the forecasted error of inflation.
The variance decomposition analysis pointed out that in “big” countries, for industrial
production, the forecast error volatility is offered mainly from the own variable fluctuations.
In addition complementary role for the decomposition of the variance play the key
monetary policy rates. More specifically, the largest portion belongs to UK where nearly 10
percent of the variance in industrial production growth comes from Sonia rate. For the
Euroarea the corresponding percentage is 4 percent, but for USA the percentage is almost
one, as in Eurozone, the industry section is more banking financed that in USA. The rise in
key rates, lead to increase in bank lending rates through the monetary policy transmission
mechanism, and as a result the cost of borrowing for industries increase. As the USA
companies are less bank financed, the effect of changes in key monetary policy rates in
industrial production is almost negligible. For the “small” countries, as we mentioned, the
increase in bank lending rates, has negative impact on Eurozone industrial productivity.
Confirming this, the variance decomposition analysis showed that the forecasted error
volatility in industrial production derives from bank profit spread changes, mainly for
Germany (14 percent).
For the yield spread between long term and short term bond rates, the variance
decomposition analysis show that most of the variance in forecast the error comes from the
own variable evolutions. In case of “big” countries, the variance of yield spread error is
determined also from Fed funds rate and Sonia rate movements, as these rates are closely
related to the short term rates (3 month treasury bills), and an increase in key rates leads to
increase in short term rates and as a result the decrease of the yield spread. Furthermore,
for “small” countries, significant role also plays the inflation rate, as it is a main factor for
the determination of long term rates as the term structure of interest rates theory suggests.
16
The variance decomposition analysis, showed also, that the forecasted error volatility in
unemployment rate both for “big” and for “small” countries comes from own variable
changes. In addition, for Eurozone, special role also plays the Eonia rate (22 percent), as the
increases in the key rate pushes up the unemployment rate. As for the exchange rates, the
forecasted error variance is offered by the own lags movements.
5.2
Impacts of shocks analysis
The next step is centralized on the implementation of impulse response functions in the
VECM. Impulse response functions (IRFs) are shocks to the system. The IRFs identify the
responsiveness of the dependent variables in the system when a shock is put to the error
term. We apply a unit (one standard deviation) shock to each variable and examine its
effects on the VEC system. The impulse response periods we select are twelve. As we
observed from the main empirical analysis, some variables responses seems to be
persistent.
For “big” countries, we take the key monetary policy rates as the main monetary policy
instruments and we examine the impact of interest rate innovations to other variables. For
the “small” countries, as they have not independent monetary instruments, we treat the
yield spread as the monetary instrument in order to examine the impact in variables to the
specific shocks, as the correlation between the 3month treasury rates and 10year bond
rates with the Eonia rate is high.
We begin our analysis with the responses of key monetary policy rates in shocks from
the other variables. The impact of unit volatility shock by inflation rate affects positively the
key monetary policy rates, as the implementation of monetary policy is based on the Taylor
rule. So, for the three “big” countries, the key rates react positively from an increase in
17
inflation rate. The effects of the other variables on key rates depend on the country. The
impact of industrial production on key rates is negative but relatively low for ECB, BoE,
comparing to the inflation effect and we see that the monetary authorities strategy are
more inflation targeted that output targeted. In case of FED the greater significance is given
to the output level targeting.
As the impulse response functions suggests, in case of the “big” countries, the responses
of bank profit spread to shocks from monetary policy rates are positive. As the transmission
mechanism works, the increase in monetary policy rates leads to the increase in lending and
deposit rates, but the increase is higher on lending rates and as a result the bank profit
spread increases. In addition, the responses of bank profit spread to innovation from the
inflation rate are positive, as the financial institutions do not wish the real lending rates to
fall. However, for the “small” countries, the response of bank profit spread to inflation is
positive only for Italy. The corresponding responses for Germany and France are negative.
As regards to the inflation rate, the impulse responses functions show, that for all the set
of countries, a positive shock from industrial production leads to the increase in inflation
rate. The fact that industries augment their production means that there is adequate
demand by consumers which the industries aim to cover. This suggests inflationary
pressures in the economy. The only exception is UK where the effect is negative. The
responses of inflation to the impact of a unity volatility shock by the key monetary policy
rates highlight the “price puzzle” (introduced by Sims 1992). These reactions of inflation are
positive for USA, and UK which indicates the paradox of increased monetary rates and high
inflation. Intuitively, the price puzzle occurs, as the central banks preemptively raise interest
rates in anticipation of future inflation. For Euroarea, the responses are negative as we
normally expected. In case of the “small” countries, the price puzzle exists only in Italy
18
however, in Germany and France the proxy monetary instrument, yield spread indicates
positive relationship between the spread and inflation rate.
In case of industrial production growth, the impulse responses functions show that the
responses of industrial production to inflation rate innovations are positive. The increase in
price level (money supply) rises total output and widens the profit spread for businesses
and as a result they increase their productivity. Our results confirms the Keynesian approach
at least in the short run, as the increase in money supply (M2) raises aggregate demand and
has positive impact on output. As a result the neutrality of money (classical approach) has
negligible impact on our sample in the short run, as the prices fell to adjust immediately.
Despite the literature findings, the responses of industrial production to shocks from key
monetary rates are positive for the three “big” countries. In addition for the “small”
countries, the effect is similar, as the proxy monetary instrument, the yield spread has
negative relation with output. We suppose, that firms suggest that a rise in interest rates
gives signal for predicted economic overheating. The responses of unemployment rate to
innovations from the industrial production are negative, as we expected. The increases in
productivity lead to more job positions and the fall in unemployment rate. Furthermore, we
have not find evidence of any trade off existence between inflation and unemployment rate
as these variables behavior differentiates depending on the specific country.
5.3 Country specific analysis
We report the country specific VEC analysis results in order to have a full view of the
country idiosyncratic characteristics. The sample period is between 1990 and 2012. After
the application of Akaike and FPE criteria we conclude to the proper number of lags for the
VEC specification of each country. As the information criteria indicated, for Euroarea we
applied VEC with three lags, for USA, UK, Germany, France the VEC model included two lags
19
and finally for Italy we used six lags for the maximum information. We report the main
variance decomposition and impulse response function results for each country. For the
variance decomposition analysis we select 36 periods and for IRFs we apply 12 periods’
horizon shocks.
EUROAREA
Variance decomposition
The variance decomposition table shows the following results (table 4):
For the Eonia rate, most of the variance of the error in forecasting the change in eonia
rate comes from its own innovations (nearly 90 percent). A smaller in significance role plays
the exchange rate of dollar/euro (5 percent), as when the base monetary policy interest rate
raises, the demand for euro also increases and the euro appreciates against dollar.
For the industrial production growth variable, most of the variance of the error in
forecasting the change in industrial production growth comes from innovations in industrial
production index (87 percent). A complementary role to the variance decomposition of
industrial production play eonia rate (3.9 percent), as the increase in key interest rate
increase the cost of capital and the industrial production falls. In addition, inflation rate and
exchange rate dollar to euro provides information for industrial production changes.
Impulse response functions
The impulse response function is based on one standard deviation shock as Cholesky
decomposition implies. We can remark for all the set of variables that the response of the
variables to innovations from the other variables have a remarkable persistence (figure 1).
The responses of eonia rate to unemployment rate are negative.
As IRF show, the responses of industrial production to innovations from inflation,
exchange rate pound to euro and eonia rate are positive. An increase in eonia rate by 15
20
basis points leads to 1.17 percent decline in industrial production at the peak of the effect
after five periods. Contrary, the responses of industrial production to unemployment rate
shocks are negative. A positive unemployment rate shock by 0.04 percent is followed by fall
in industrial production by 0.5 percent.
USA
Variance decomposition
For the fed funds rate, most of the variance of the error in forecasting the change in fed
funds rate comes from its own innovations (table 5). A complementary role for the
decomposition of fed funds rate variance plays also the industrial production growth, nearly
at three percent.
For the bank profit spread variable, most of the variance of the error in forecasting the
variable comes from the innovations in bank lending and deposit rate nearly at5 56 percent.
In addition the fed funds rate movements contributes to the variance decomposition of
bank profit spread. The contribution of fed funds rate to bank profit spread movements is
significant as higher fed funds rate leads to the prompt increase in base lending rates.
Furthermore, the level of industrial production explains significantly the variation in bank
profit spread as the increase in lending rates pushes down the rate of productivity.
Impulse response functions
For USA impulse responses functions show (figure 2): The responses of fed funds rate to
industrial production and yield spread shocks are positive.
An increase in industrial
production by 0.63 percent will lead at the increase in fed funds rate by 0.03 percent, as the
monetary authorities impose contractionary policy against economic overheating. The
21
responses of fed funds rate to inflation and bank profit spread are negative. An increase in
bank profit spread by 0.21 percent leads to the fed funds rate decline after three periods.
The responses of yield spread to inflation and bank profit spread innovations are
positive. The responses of yield spread to shocks from fed funds rate are negative. An
unexpected monetary policy tightening by 14 basis points is followed by 0.05 percent
increase in bank profit spread, as the market participants deal with the effect as temporary.
UK
Variance decomposition
For UK we have (table 6): For the bank profit spread variable, the variance
decomposition table shows that the variations are mainly derived from changes in bank
lending and deposit rates. However, a remarkable contribution to the decomposition of
bank profit spread variance is offered by Sonia rate (above 8 percent), as the Sonia rate
determines the changes in lending and deposit rates. In addition, most of the contribution in
decomposition of bank profit spread variable is offered by the variations in CPI index which
determines the inflation rate.
The variance decomposition table shows that the forecasted error in unemployment rate
comes from its own innovations. in addition, a significant contribution to the variance
decomposition are offered by the inflation rate (12 percent) as the increase in inflation rate
provokes rise in unemployment rate, as the economy is overheating.
Impulse response functions
For UK we have the IRF reveals the following results (figure 3): For the Sonia rate, the
responses of Sonia rate to innovations from inflation rate are positive. A positive shock with
the increase in inflation rate result in Sonia increase by 0.02 percent. As the economic
theory suggests, the increase in inflation rate will lead to the increase in monetary policy
22
key rates. The responses of Sonia rate to bank profit spread, unemployment, industrial
production growth and yield spread are relatively insignificant.
The responses of yield spread to inflation and industrial production shocks are positive.
An increase in inflation rate by 0.4 percent will lead to increase in yield spread by almost
0.04 percent after three periods, as the investors think that the effect will remain in future.
The responses of inflation rate to Sonia rate, bank profit spread, yield spread and
unemployment rate are positive. An increase in bank profit spread by 0.14 percent leads to
inflation rate increase by 0.1 percent after four periods.
GERMANY
Variance decomposition
The variance decomposition for Germany (table 7) shows the following results:
For the inflation rate, the variance decomposition analysis shows that the variance of the
error is derived from innovations in CPI index nearly at 68 percent. In addition, a significant
contribution to the decomposition of the variance in inflation rate is offered by the bank
profit spread movements (13 percent) as the increase in lending rate leads in inflation
decrease. Also, most of the information in the decomposition of inflation is offered by
industrial production index changes. The increase in productivity leads to rise in inflation
pressures as economy growths.
The forecasted error in yield spread is derived mainly from innovations in components of
the spread, the ten year bond rate and 3month Treasury bill rate. The variance
decomposition table shows also that inflation rate affects the changes in yield spread (5
percent), as when the expected inflation increases, the yield spread also rises. Additionally,
the bank profit spread offers a significant explanatory power to the decomposition of yield
spread variance as the increase in lending rates pushes up the long term rates.
23
Impulse response functions
For Germany the IRFs (figure 4) show: The responses of inflation rate to innovations from
industrial production index and yield spread are positive. One standard deviation shock of
industrial production index with 1.5 percent increase leads to inflation increase by 0.3
percent approximately. The responses of inflation rate to bank profit spread are negative. A
shock of bank profit spread with 0.08 percent increase leads to 0.05 percent decline in
inflation rate as the time lags augments, indicating a significant impact of interest rates in
inflation rate. The responses of inflation rate to unemployment rate innovations are neutral.
The responses of industrial production growth to shocks from inflation and bank profit
spread are positive. A typical rise of inflation rate by one standard deviation 0.32 percent
leads to an increase by 0.3 percent in industrial production, indicating a close relationship in
the two variables. The responses of industrial production in innovations from
unemployment rate and yield spread are negative.
FRANCE
Variance decomposition
In case of France we can remark significant convergence. More analytically, the variance
decomposition (table 8) shows the following results:
For the industrial production growth variable, most of the variance of the error in
forecasting the change in industrial production growth comes from innovations in industrial
production index (91 percent). A complementary role to the variance decomposition of
industrial production play the unemployment rate (4.5 percent), as the increase in
unemployment rate pushes down the productivity.
For the bank profit spread variable, the variance decomposition table shows that the
variations are mainly derived from changes in bank lending and deposit rates (55.7 percent)
24
Also, the yield spread offers significant explanatory power for the forecasted movements in
bank profit spread nearly at 13 percent. As the long term bond rates rise, cause the
increase in lending rates. In addition the inflation rate is key to the decomposition of bank
profit spread variable (almost 20 percent), as the high inflation gives signal for increased
interest rates.
Impulse response functions
For France the IRF (figure 5) results show: The responses of inflation rate to innovations
from yield spread and industrial production are positive. A shock of industrial production by
1.2 percent will lead to 0.3 percent increase in inflation rate at peak of the effect after two
lags. The responses of inflation rate to bank profit spread and unemployment rate shocks
are negative. A typical positive shock of 0.17 percent in bank profit spread leads to 0.06
percent decline in inflation rate after five periods.
The responses of yield spread to inflation and bank profit spread shocks are positive. A
typical rise in inflation rate by 0.26 percent leads to 0.05 percent increase in yield spread.
The response of yield spread to unemployment rate and industrial production shocks are
relatively small and neutral. The responses of unemployment rate to innovations from yield
spread are positive. The responses of unemployment rate to bank profit spread shocks are
negligible.
ITALY
Variance decomposition
The variance decomposition (table 9) shows the following results:
For the inflation rate, the variance decomposition analysis shows that the variance of the
error is derived from innovations in CPI index. In addition a small contribution to the
25
decomposition of the variance in inflation rate is offered by the yield spread movements as
the increase in lending rate leads in inflation decrease.
The forecasted error in yield spread is derived mainly from innovations in components of
the spread, the ten year bond rate and 3month Treasury bill rate. The variance
decomposition table shows also that inflation rate affects the changes in yield spread, as
when the expected inflation increases, the yield spread also rises. Additionally, the bank
profit spread offers a significant explanatory power to the decomposition of yield spread
variance as the increase in lending rates pushes up the long term rates.
Impulse response functions
For Italy we have (figure 6): The responses of inflation rate to industrial production
growth, bank profit spread and unemployment rate are positive. However the impact is
significantly low.The responses of inflation rate to yield spread are negative. A positive
shock by 0.6 percent leads to 0.03 percent decline in inflation rate after two periods.
The responses of yield spread to innovations from bank profit spread are positive and
significant as in the previous case. An increase in bank profit spread by 0.4 percent leads to
the increase in yield curve 0.18 percent. The term structure of interest rates indicates a
close relationship between these spreads. The responses of yield spread to unemployment,
inflation and industrial production shocks are negative. An increase in inflation rate by 0.16
percent is followed by decline in yield spread by 0.12 percent.
5.4 Monetary policy implementation
By applying the benchmark VEC model we observe some useful findings of monetary
policy decision making for the period 1990-2012. Our main empirical results revealed
persistency for some variables. Both the variance decomposition analysis and the impulse
responses functions showed that inflation rate fluctuations are explained more by key
26
monetary rates than the industrial production generally. In case of industrial production, the
USA authorities and less the UK, give greater attention on minimizing the output gap, while
the ECB and give little significance on output growth.
As we observed, the bank profit spread increased after the rise in key rates indicating
that the monetary transmission worked. Additionally, the variance of the yield spread
movements is explained strongly by key monetary policy rates as the VEC findings confirms.
The responses of industrial production and unemployment rate from variables used are
similar for all countries, and between them there exist a negative relationship.
The VEC analysis highlighted the existence of price puzzle in the sample, as the increase
in key rates is followed by increase in inflation, as monetary authorities anticipated high
inflation, but inflation eventually occurs. Especially, for USA, UK and Italy the responses of
inflation to increase in key rates are positive confirming the price puzzle existence. Also, our
benchmark analysis found non stable evidence for the Philips curve existence, as
unemployment and inflation movements depends on the each country factors.
Furthermore, an additional finding is the relative persistence in inflation rate even after the
crisis period which indicates both the relative stickiness of prices and the increased money
supply. Central banks preserve the high level of money supply as the fall in demand
increased the possibility of deflation. Finally, we observed also that the increase in inflation
rate is followed by rise in industrial production level. The fact that inflation were relatively
increased indicated augmented industrial production activity.
6
The monetary policy in financial crisis period
6.1 Financial crisis period
27
The benchmark model includes a period over two decades. As the financial crisis period
lies on this period we apply structural stability (Chow) tests in order to examine any stability
issues in the estimation sample. The period that we apply Chow tests is between 2007 and
2009 when crisis erupted. According to the results, we observed structural breaks for all
countries except France.
Our procedure was to re-estimate the model for the period starting 2007 ending 2012 in
order to examine the impact of financial crisis on our framework. The econometric strategy
we follow involves the analysis of a VAR model after checking of no cointegration in this
period. More specifically we apply a VAR(2) model as both Akaike and FPE criteria indicated
and we proceed in the sub sample analysis.
The analysis we elaborate assets the effect of financial crisis on our macrofinance
framework. In this period, we have to take into account some exogenous changes induced
by central banks. The key monetary policy rates in this period declined, and the yield spread
for the euroarea countries (Germany, France, Italy) increased, as the short term rates
reduces according to the movements of key policy rates. The bank profit spread also
decreased as lending rates declined following the fall in key monetary policy rates. As we
expected, the industrial production for all the set of countries reduced significantly and
unemployment rate increased sharply (except for Germany). Finally, the inflation rate
increased for all countries despite the fall in consumption and money multiplier, as the
central banks increased the monetary base in order to stimulate the economic activity.
The VAR procedure included the variance decomposition and impulse shock analysis as
in the benchmark model. In general, the model showed increased convergence in this
period for all countries. By applying the variance decomposition in Euroarea, USA and UK we
observed significant contribution in variables explanation from the key monetary policy
rates, increased compared to the main model. The results indicated high effect on
28
macroeconomic variables from key policy rates. Furthermore, the model revealed that for
USA most of the forecaster error in variables variance is derived also by industrial
production index. For Germany, France and Italy the variance decomposition results were
similar to the benchmark model, indicating significant effects mainly from yield spread and
inflation rate.
The impulse response analysis highlighted some more interesting results. For Euroarea, a
positive shock in inflation rate is followed by decline in Eonia rate. This result demonstrated
that the objective of the ECB was the economic recovery. Also, a significant result was the
negative relationship between unemployment and Eonia rate. In the benchmark model the
increase in Eonia was followed by increase in unemployment rate. However, in financial
crisis period, the fall in Eonia rate lead to unemployment increase. This was a clear impact
by financial crisis. Also, a countercyclical effect was the positive response in unemployment
after an increase in inflation.
For USA, the results showed a different response of fed funds rate after shocks by
inflation, unemployment and yield spread. Contrary to the main model results, an increase
in the above variables was followed by fed funds rate decline. This result was an indicator of
economic reheating efforts adopted by FED. Also, we remark the positive effect of inflation
on unemployment rate as in Euroarea. An interesting result is the negative effect of fed
funds rate to industrial production and at the same time the positive effect of fed funds rate
to unemployment rate, contrary to the main analysis findings. In the period before financial
crisis the increased economic growth (high industrial production-low unemployment) the
effect of increased interest rates on economic activity has positive. After the crisis outbreak
and the decline in activity, the reduction of fed funds rate lead in positive response by
industrial production but quickly became negative.
29
For UK, the impulse response analysis revealed that positive shocks from inflation and
bank spread were followed by increase in Sonia rate, contrary to the main findings. Also, in
UK case the relationship between inflation and unemployment is positive as in Euroarea and
USA. Furthermore, the positive shock from bank spread in UK lead to positive responses by
yield spread. This result is also appeared in USA case and it is opposite to the main analysis
findings. In the financial crisis period, the increase in risk premia lead to high bank spreads
as the liquidity reduced, and increased yield spreads due to augmented sovereign risk.
For Germany, France and Italy the impulse response analysis showed that increase in
inflation, bank spread and unemployment rate lead to yield spread decline and vice versa.
This result is opposite to the main model findings (except for Italy where they appeared
also). After crisis, the spread increased due to both the increase in long term rates
(Eurozone debt crisis) and the fall in short term rates. As the relationship between yield
spread and inflation is usually positive the financial crisis broke down this relation. Also, the
yield spread positive shock lead to unemployment increase despite the short term rate
decrease, as the countries focused in this period on recession treating.
6.2 Assessment of the monetary policy transmission mechanism
We apply an innovative procedure for the examination of monetary policy transmission
mechanism, where we control the effect of key rates movements in the determination of
the prime bank lending rates. Our approach is similar to Renne (2012), who studied the
transmission mechanism in Euroarea, by examining the main refinancing operations rate
and Eonia rate before and after the crisis on an affine framework. Renne, found a break in
relationship between the two rates after the crisis outbreak. Our framework includes the
key monetary policy rates and prime bank lending rates for UK, USA, Germany, France and
30
Italy. The methodology we follow is based on the analysis of VAR models for each country
by spitting the time period before and after crisis, with 2008 as the benchmark year. By
applying the impulse responses functions and the variance decomposition analysis we
observed significant positive relationship between the monetary policy rates and the prime
bank lending rates. It is clear from the results that the monetary policy transmission
mechanism worked effectively before the crisis.
However, after the crisis outbreak we remarked that the transmission mechanism
dropped down. The VAR analysis indicated that monetary policy rates and prime lending
rates followed opposite paths. The only exception is Germany as the positive relationship
between two rates maintained. As the crisis accelerated, the central banks cut their main
rates nearly to zero. However, the lending rates soared as the risk premia increased sharply.
This situation aggravated the economic conditions, as the central banks decisions for
stimulating the economy became ineffective, and the bank lending channel collapsed.
6.3 The implementation of non-standard measures by central banks
We proceed to the analysis of the effect of unconventional measures that adopted by
central banks during the financial crisis period. Our specification includes the selection of
representative measures as proxies from central banks’ balance sheets and examine their
effect on economy. The non-standard measures we used cover significant part of the central
banks’ balance sheet concerning their magnitude in order to have representative results
concerning the central banks’ actions during the crisis. We select ratios between tools and
total assets to examine the magnitude of their effects.
For USA we choose the U.S. Treasury securities ratio which consist of the treasury notes
held by FED to total assets. After the financial crisis outbreak the FED increased the number
31
of securities held in order to reduce the term premia of long term interest rates.
Additionally, its actions targeted to promote a stronger pace of economic recovery along
with inflation control. In case of UK, we choose the ratio of total reserves to total assets.
After the crisis, the BoE reserves increased significantly as they were used for funding the
key unconventional strategy, the Asset Purchase Facility program implemented by monetary
authorities to economic recovery and credit easing. In Euroarea we select as representative
measure the ratio of securities held by ECB to total assets. The securities held by ECB rose
sharply after the crisis outbreak as part of the Securities Market Programme (SMP)
implemented by monetary authorities. Its objective was to restore the transmission
mechanism and to provide liquidity in public and private debt securities market.
Our econometric procedure involves the application of a VAR model, against benchmark
VEC, as we observed no cointegration in the sample. The analysis of VAR is based on the
variables we used in benchmark analysis with the addition of the unconventional effect
variable in each country which we treat it also as endogenous. The VAR analysis covers the
period from 2007 until 2012 which includes the financial crisis period and then, in order to
have an integrated view of the unconventional measures effect.
The result show that for all the countries the implementation of unconventional
measures lead in inflation increase, as the money supply increased. The monetary
authorities targeted to inflation sustainability with the control in price level (despite the
recession) against the negative effect of a possible deflation, which would further reduce
consumption and investments. Furthermore, the unconventional effect show a generally
increase in bank profit spread, i.e. the rise in lending rates. The increase in money supply fail
to lower the lending rates as the term risk premia augmented (Germany, UK are
exceptions). This result indicated a shrink in credit growth dynamic, which implied fail in
32
stimulating aggregate demand. So, the unconventional strategies did not manage to
promote consumption and investments.
In general, the implementation of non-standard measures improved the economic
conditions in countries. The effect is greater in USA where the stimulation packages were
massive and their magnitude more significant. Contrary, in UK and Euroarea, the
unconventional measures improved the economic activity however, their effect was
relatively temporary. This fact is explained as the Bank of England quantitative easing was
based mainly on asset purchases and the magnitude on real economy was limited. The
European Central Bank was implemented a wide range of unconventional measures that
helped partly the economic conditions, but, its actions were quickly sterilized after their
implementation.
7. Robustness analysis
We apply the robustness analysis for our VEC model in order to verify and assess our
results, by performing four robustness tests. We set up with the control for any endogeneity
bias in our benchmark specification. As the VAR/VEC analyses are endogenous models, they
often face variables ordering problems which rise endogeneity issues. We address the
endogeneity bias issue by testing the behavior of the variables if we change their order and
run again the VEC model. Firstly, we set the first variable, as last and the last variable as
first, and so on. Secondly, we chose randomly the order of the variables. As we observed, all
the results remained the same after the ordering changes we performed, thus confirming
our benchmark analysis and rejecting any endogeneity.
Furthermore, we implement the VEC/VAR specifications with generalized responses
functions against the traditional impulse responses functions as the generalized impulse
responses are invariant to the reordering of the variables in the VAR/VEC (Pesaran, Shin
33
1997). Typically there are many alternative parameterizations that could be employed to
compute traditional impulse responses, and there is no clear guidance as to which one of
these possible parameterizations should be used. In contrast, the generalized impulse
responses are unique and fully take account of the historical patterns of correlations
observed amongst the different shocks. We implement the generalized responses functions
for each country, and we observed no change in the main results. So, we argue that main
results remain robust.
Afterwards, we consider alternative identification of the main monetary policy
instrument and we apply variance decomposition analysis and impulse response functions
(money supply and interest rates). Then, we apply VAR model instead of the VEC model we
used in main analysis (apart from UK, France we elaborate in main analysis).
We change the monetary policy instrument and we equate each monetary policy rates
with money supply growth as an alternative monetary measure (as in Christiano,
Eichenbaum, Evans 1999). For the “big” countries we use M2 growth instead of EONIA,
SONIA, FED funds rate. For the “small” countries, we exclude the yield spread and we
replace it with M2 growth of Euroarea. In case of “big” countries the effect is reverse of the
key rates impact, as the monetary supply increase is followed by interest rates decline. In
addition, by applying this strategy we can assess the existence of liquidity effect. For the
“small” countries the effect is straight as the increase in yield spread reflects the fall in short
term rate.
Our results indicate that for Euroarea, USA and UK the money supply shock leads in the
reverse results of the EONIA, SONIA, FED funds rate shocks respectively, reflecting the
robustness of our results. Furthermore, the results confirm the so called “price puzzle” we
find in the benchmark analysis. For the “small” countries, the money supply shock leads in
34
the straight results of the yield spread shock for each country, and our results remain
robust.
Finally, we change our econometric methodology by performing VAR analysis instead of
the benchmark VEC we elaborated. As our variables are marginal cointegrated the VAR
analysis have to reveals us similar results. We elaborated the impulse response functions for
each country and the variance decomposition analysis. As we observed, the VAR analysis
showed that our results remain exactly the same with the VEC analysis (appendix 3) and this
gives the robustness of our sample results.
8. Conclusion
We examined the monetary policy decision making in Euroarea, USA and UK by
controlling the impact of central banks actions in financial and macroeconomic variables.
After analyzing the monetary policy for extended period, over twenty years, we have a clear
view for the performance of monetary policy implementation. Within this period we studied
the effect of financial crisis of 2007-2009 on economy and also we measured the impact of
non-standard monetary policy tools in macroeconomic variables and the extent to which
they helped the economic recovery.
Our preliminary analysis results, showed that before the introduction of euro, the bond
market reveled signs of convergence of the European government notes. In addition, the
period before euro the yields of UK and USA with the European ones are negatively
correlated as these bonds deemed as substitutes in global markets. After the introduction of
euro the bond rates are positive both European and non-European indicating the global
bond market integration.
The main VEC analysis showed that the monetary policy transmission mechanism
generally works effectively. However, by examining the transmission channel before and
35
after the financial crisis we remarked that this channel altered and distorted significantly. As
a result, the conventional monetary policy instruments become ineffective. For example, a
monetary expansion with the decrease in key nominal interest rate, has limited effect
especially for countries in periphery (Italy), as the results show that bank and bond rates
have relatively low correlation with Eonia.
After the crisis outbreak, where the central banks proceeded to the implementation of
non-standard measures with the parallel increase in their balance sheets and the level of
money supply the economic climate changed. As we observed the unconventional measures
improved the economic conditions for all countries but mainly for USA where the
stimulation packages were larger than Euroare and UK. So, the analysis indicate that the
central bank authorities have to preserve the unconventional monetary policy tools or
expand them with increased money supply to boost the anemic economic growth.
References
Abassi P., Linzert T., 2011. The effectiveness of monetary policy in steering money market
rates ECB working paper series.
Akaike H., 1969. Statistical predictor identification, Annals of the Institute of Statistical
Mathematics, 203–217.
Ang A., Boivin J., Dong S., 2008. Monetary policy shifts and term structure, NBER working
paper.
Boivin J., Giannoni MP., 2008. Global forces and monetary policy effectiveness, International
dimensions of monetary policy, NBER, 429-478.
Cecioni M., Neri S., 2012. The monetary transmission mechanism in the euro area: Has it
changed and why?, WP series, Banca d’ Italia.
Chow G., 1960. Tests of Equality between Sets of Coefficients in Two Linear Regressions.
Econometrica 28 (3): 591–605.
Christiano L., Eichenbaum M., Evans C., 1999. Monetary policy shocks: What have we
learned and to what end?, Handbook of Macroeconomics, North-Holland: 65-148.
Dickey D.A., Fuller W., 1987. Distribution of the estimators for autoregressive time series
with a unit root, Journal of the American Statistical Association 74, 427–431.
Hamilton J., Kim DH., 2000. A re-examination of the predictability of economic activity using
the yield spread, discussion paper, University of California.
Herro N., Murrey J. 2011. Dynamics of monetary policy uncertainty and the impact on
macroeconomy, working paper, University of Wisconsin.
36
Johansen, S., 1994, Likelihood-based inference in cointegrated vector autoregressive models
(Oxford University Press, Oxford).
Kuttner K., 2000. Monetary policy surprises and interest rates: Evidence from the Fed funds
futures market. Journal of Monetary Economics 47 (3): 523–544.
Kwiatkowski, D.P., Phillips P.C.B., Schmidt P. Shin Y., 1992. Testing the null hypothesis of
stationarity against the alternative of unit root: how sure are we that economic time series
have a unit root? Journal of Econometrics 54, 159–178.
Litterman RB., Weiss L., 1984. Money, real interest rates and output: A reinterpretation of
postwar US data, Econometrica, 53(1):129-156.
Mishkin F., 2010. The Economics of Money, Banking, and Financial Markets, AddisonWesley, Boston, 10th edition.
Peersman G. Smets F., 2001. The monetary transmission mechanism in the Euroarea:
evidence from VAR analysis, Cambridge University Press, chapter 2, 56-74.
Peersman G., 2011. Consequences of different types of credit market disturbances on the
Euro area economy, manuscript, Ghent University.
Pesaran M.H., Shin Y., 1997. Long-run structural modelling, unpublished manuscript,
University of Cambridge.
Renne JP., 2012. A model of the euro-area yield curve with discrete policy rates, WP series,
Banque de France.
Sims C., 1980. Comparison of interwar and postwar business cycles: A monetarism
reconsidered, American Economic Review, AEA, 70(2):250-257.
Taylor J., 1993. Discretion versus policy rules in practice. Carnegie-Rochester conference
series on public policy 39: 195-214
Table 1. Descriptive Statistics
Key rates
Mean
Std. Dev.
Skewness
Kurtosis
Jarque-Bera
Probability
Observations
EONIA
FEDFUNDS
SONIA
3.097397
1.692281
0.008556
2.315361
4.377557
0.112054
224
3.559963
2.289252
-0.116048
1.914012
13.97671
0.000923
272
5.352246
3.277722
0.826252
4.328639
50.95527
0.000000
272
Inflation
ggd
EUROAREA
Mean
Std. Dev.
Skewness
Kurtosis
Jarque-Bera
Probability
Observations
2.040807
0.735578
-0.620171
4.468113
34.32163
0.000000
223
USA
2.562236
1.108925
-0.897731
5.712316
114.4754
0.000000
271
UK
GERMANY
2.465758
1.494446
1.874056
6.572098
284.1278
0.000000
271
1.537000
0.784031
0.073606
2.898399
0.266615
0.875196
200
37
FRANCE
1.854682
0.816654
-0.232452
3.828658
10.04375
0.006592
267
ITALY
3.131618
1.521613
0.704893
2.642351
23.97461
0.000006
272
Industrial production growth
Mean
Std. Dev.
Skewness
Kurtosis
Jarque-Bera
Probability
Observations
EUROAREA
USA
UK
GERMANY
FRANCE
ITALY
1.062609
5.468350
-2.339925
9.852409
608.2324
0.000000
212
2.023197
4.456114
-2.032208
7.969950
446.5487
0.000000
260
-0.072484
3.175025
-1.397235
6.347666
206.0058
0.000000
260
1.336553
6.547027
-1.650612
7.544534
341.8011
0.000000
260
0.149855
4.364651
-1.947720
9.029654
558.2545
0.000000
260
-0.272280
6.160695
-2.074096
9.826109
688.5435
0.000000
259
Note: the tables above present the descriptive statistics for the key monetary rates in the level form,
the inflation rate computed as the logarithmic change of monthly cpi index year over year, and
industrial production growth computed as the logarithmic change of monthly industrial production
index year over year.
Table 2. Unit root tests
38
VARIABLE
EUROAREA
ADF
USA
KPSS
UK
ADF
KPSS
ADF
KPSS
-1.309101
1.340859
-2.101523
1.163590
-2.774698
1.548008
-7.620498*
-1.921847
0.061848*
0.302467
-5.808373*
-1.533657
0.071808*
2.186594
-5.827515*
0.558256
0.209184*
2.111902
-11.01941*
0.194242*
-10.74880*
0.213578*
-2.891377*
0.285220*
-2.546992
1.222516
-1.906647
1.788228
-0.954258
0.941150
Δlnip
-4.025292*
0.028904*
-5.160556*
0.221576*
-20.60134*
0.103886*
unemp
-1.661277
0.409397
-1.417648
0.492775
-1.686797
0.860169
Δunemp
-3.797354*
0.028154*
-5.375844*
0.141054*
-4.139407*
0.279826*
treasuryrate
-
-
-2.117577
1.250487
-2.806999
1.503637
-
-
-5.478643*
0.067863*
-6.021194*
0.199476*
-
-
-1.774379
1.090527
-2.690539
1.544013
-
-
-6.123584*
0.060467*
-6.616024*
0.199683*
main rate
Δmain rate
lncpi
Δlncpi
lnip
Δtreasuryrate
lendingrate
Δlendingrate
VARIABLE
lncpi
Δlncpi
GERMANY
FRANCE
ITALY
ADF
KPSS
ADF
KPSS
ADF
KPSS
-1.676216
2.110892
0.043397
2.185343
-2.417529
2.136087
-13.06520*
0.122777*
-3.602292*
0.206881*
-4.269557*
0.130927*
-1.147723
1.825030
-1.594868
0.907048
-1.473349
0.427761
Δlnip
-6.917972*
0.059706*
-7.929622*
0.131269*
-7.046338*
0.221014*
unemp
-0.537214
0.792579
-2.349635
0.373499
-1.351123
1.041245
-5.978889*
0.332162*
-5.747542*
0.156362*
-3.295108*
0.341700*
lnip
Δunemp
treasuryrate
Δtreasuryrate
lendingrate
Δlendingrate
-1.549056
1.268559
-1.787484
1.671210
-1.710696
1.754650
-10.74754*
0.098612*
-12.56887*
0.161190*
-16.09294*
0.114276*
-1.305930
1.720515
-2.054005
1.993126
-1.682803
1.889115
-8.299521*
0.103113*
-6.963022*
0.082239*
-6.448607*
0.105444*
Note: The table presents the Augmented Dickey-Fuller tests and Kwiatkowski-Phillips-Schmidt-Shin
tests. The 1% critical values for stationarity are -3.46 for ADF test and 0.74 for KPSS test and (*)
indicates significance at the 99% confidence level.
Table 3. Johansen cointegration test results
Trace statistic
Maximum eigenvalue
123.3003*
one coint. eqn.
47.72265*
one coint. eqn.
188.9893*
one coi nt. eqn.
105.9872*
one coi nt. eqn.
103.2842*
79.71790*
one coint. eqn.
one coint. eqn.
33.66196
36.74204*
no coint.
one coi nt. eqn.
FRANCE
30.97534
three coint. eqns.
45.62352*
one coint. eqn.
ITALY
122.6990*
one coi nt. eqn.
62.96887*
one coi nt. eqn.
EUROAREA
USA
UK
GERMANY
Note: The table shows the Johansen test results for cointegration presented by trace and
maximum eigenvalues tests, (*) values indicate one cointegrating equation at 0.05 level
both for trace and maximum eigenvalue.
Table 4. Variance decomposition for Euroarea
39
Period
DEONIA
2
94.51143
0.235357
0.414961
0.015521
4.614469
3
90.23358
3.375329
1.702466
0.523104
3.801131
12
89.86588
5.031596
1.500718
0.346466
3.080689
24
89.96755
5.262019
1.500836
0.190536
2.984131
36
90.01311
5.342792
1.500640
0.129895
2.948667
DEXCHRATE $/E
INFL
INDPRODGR
Variance Decomposition of DEONIA:
DUNEMP
Variance Decomposition of DEXCHRATE $/E:
2
2.391743
96.10354
0.046046
0.862066
0.564038
3
2.196949
94.65803
0.264790
1.658407
0.740661
12
2.409023
93.84223
1.185808
1.839570
0.448198
24
2.396538
94.12431
1.195377
1.835414
0.282897
36
2.393316
94.22889
1.200264
1.839031
0.217859
Variance Decomposition of INFL:
2
1.357676
0.154537
97.51313
0.250307
0.195284
3
1.686009
0.309555
96.52730
0.239871
0.753488
12
2.547049
0.390002
94.62251
0.788057
0.851013
24
2.763313
0.336429
94.65769
0.610224
0.911186
36
2.852736
0.316694
94.66228
0.544836
0.933072
Variance Decomposition of INDPRODGR:
2
1.014433
0.630209
1.987217
96.33101
0.020396
3
0.995007
0.751697
2.682852
94.73047
0.254957
12
3.163628
0.967571
3.154095
90.04337
0.545277
24
3.526009
1.243805
3.308265
88.54365
0.571067
36
3.897690
1.499427
3.431506
87.12339
0.598164
Variance Decomposition of DUNEMP:
2
7.934807
1.670532
0.522716
1.117552
88.44223
3
16.44915
1.191243
0.396741
0.709654
80.64887
12
21.01552
0.692574
2.184344
1.166518
74.55181
24
22.10696
0.474482
2.203533
1.210266
73.67349
36
22.48179
0.400467
2.211773
1.219566
73.37687
Note: This table reports the variance decomposition of variables deonia: eonia rate (first
difference), dexchrate: exchange rate dollar to one euro (first difference), infl: inflation, inprodgr:
industrial production growth and dunemp: unemployment rate (first difference) for Euroarea.
Each row shows the percentage of the variance of the error in forecasting the variable mentioned
in the table, at its forecasting horizon presented in the first column.
Table 5. Variance decomposition for USA
40
Period DFEDFUNDS DBANKSPREAD DYIELDSPREAD INPRODGR
Variance Decomposition of DFEDFUNDS:
INFL
DUNEMP
2
94.80852
0.047194
0.526164
4.193919
0.000165
0.382372
3
94.14242
0.807095
0.562172
3.610046
0.452719
0.321326
12
94.37191
1.127348
0.607999
3.255473
0.253374
0.213953
24
94.57882
1.139614
0.627572
3.142716
0.180777
0.176568
36
94.65571
1.142955
0.634206
3.102129
0.153945
0.162781
Variance Decomposition of DBANKPROFITSPREAD:
2
4.649171
88.59077
0.615385
0.189389
1.050776
4.365164
3
5.951675
79.23880
0.569762
8.030991
1.353327
4.248835
12
5.156556
69.12589
0.647420
11.59232
2.023075
9.258185
24
4.563247
61.78861
0.716865
13.16436
2.224176
14.92314
36
4.092543
55.96838
0.773363
14.42595
2.375148
19.40706
2
4.095269
1.399133
93.41310
0.001689
0.882158
0.138140
3
5.903544
2.752931
89.89944
0.248012
0.768403
0.309217
12
4.391621
2.487314
91.70727
0.375616
0.723868
0.185027
24
4.110861
2.042856
92.82103
0.277900
0.567917
0.108051
36
4.001485
1.875528
93.24976
0.239548
0.505422
0.078898
Variance Decomposition of DYIELDSPREAD:
Variance Decomposition of INPRODGR:
2
0.868671
1.880276
0.814735
95.92577
0.125805
0.051995
3
0.742784
3.882646
1.159525
93.21300
0.438653
0.208518
12
0.978946
1.832589
0.982039
95.63205
0.238488
0.134866
24
1.014952
1.248781
0.965545
96.42718
0.135059
0.080001
36
1.028038
1.022602
0.959219
96.73635
0.094947
0.058804
2
7.751660
12.58481
0.051775
0.654875
78.16327
0.032522
3
10.74282
12.37573
0.258783
1.701055
74.11082
0.080797
12
11.73993
9.840555
0.154694
0.735481
76.85760
0.068163
24
12.22764
9.449431
0.119621
0.494886
77.08372
0.049130
36
12.40118
9.303055
0.106524
0.406840
77.17470
0.042109
Variance Decomposition of INFL:
Variance Decomposition of DUNEMP:
2
0.576152
3.236131
0.524848
1.374888
0.130474
92.22232
3
1.505175
3.183989
0.693651
1.138644
0.597336
90.49308
12
1.101904
3.607240
0.544333
0.883855
0.513331
91.98229
24
0.915496
3.825267
0.497283
0.814462
0.436549
92.62094
36
0.843412
3.907538
0.479676
0.787881
0.406277
92.87130
Note: This table reports the variance decomposition of variables dfedfunds: fed funds rate (first
difference), dbankspread: lending minus deposit rate (first difference), dyieldspread: 10year rate
minus 3month treasury rate (first difference), inprodgr: industrial production growth, infl: inflation,
and dunemp: unemployment rate (first difference) for USA. Each row shows the percentage of the
variance of the error in forecasting the variable mentioned in the table, at its forecasting horizon
presented in the first column.
Table 6. Variance decomposition for UK
41
Period
DSONIA DBANKSPREAD DYIELDSPREAD
INFL
Variance Decomposition of DSONIA:
INPRODGR
DUNEMP
2
98.74830
0.887554
0.048806
0.298786
0.012774
0.003775
3
98.07905
1.290988
0.051887
0.502104
0.059886
0.016088
12
98.03522
0.737309
0.063250
0.909065
0.186004
0.069156
24
98.27810
0.439717
0.051392
1.060012
0.118746
0.052036
36
98.37927
0.317200
0.046676
1.121915
0.089762
0.045179
Variance Decomposition of DBANKSPREAD:
2
12.55860
75.84250
0.946040
8.881250
0.197134
1.574477
3
10.99146
75.59617
1.368296
10.03510
0.645375
1.363595
12
9.862333
69.63219
1.696926
15.64914
0.889133
2.270277
24
8.831129
68.76326
1.418480
17.73808
0.799404
2.449651
36
8.409812
68.41121
1.303865
18.59727
0.756587
2.521263
Variance Decomposition of DYIELDSPREAD:
2
7.166290
4.008226
88.77884
0.033284
0.005614
0.007742
3
10.88282
7.256926
79.50475
1.733698
0.231340
0.390461
12
13.49068
7.836908
77.09960
1.066750
0.187180
0.318883
24
14.77206
8.336466
75.49475
0.928624
0.157395
0.310712
36
15.27681
8.531300
74.86346
0.874891
0.145997
0.307537
2
1.618174
4.283335
6.786356
86.54495
0.072942
0.694244
3
4.293975
4.924130
7.627618
79.10356
1.142396
2.908317
12
6.330085
12.75416
12.62776
60.28702
0.796942
7.204033
24
7.221396
15.06522
14.72667
53.60745
0.675179
8.704080
36
7.593997
16.06394
15.62768
50.74524
0.623912
9.345239
Variance Decomposition of INFL:
Variance Decomposition of INPRODGR:
2
8.781608
2.367621
0.458686
0.516852
86.88271
0.992523
3
8.047670
2.466646
0.387866
0.572151
85.88420
2.641468
12
9.485104
2.839277
0.505181
0.336831
84.26566
2.567943
24
10.04217
3.036724
0.490526
0.256245
83.63345
2.540887
36
10.25917
3.115962
0.484109
0.223743
83.38828
2.528737
Variance Decomposition of DUNEMP:
2
0.495657
1.265947
1.187263
4.677804
0.317084
92.05624
3
1.495215
1.220182
0.975969
7.344187
0.322358
88.64209
12
1.663991
1.891225
0.843045
10.74815
0.187352
84.66624
24
1.739592
2.002387
0.767720
11.52581
0.139395
83.82510
36
1.766510
2.043658
0.739890
11.81133
0.121745
83.51687
Note: This table reports the variance decomposition of variables dsonia: Sonia rate (first difference),
dbankspread: lending minus deposit rate (first difference), dyieldspread: 10year rate minus 3month
treasury rate (first difference), infl: inflation, inprodgr: industrial production growth, and dunemp:
unemployment rate (first difference) for UK. Each row shows the percentage of the variance of the
error in forecasting the variable mentioned in the table, at its forecasting horizon presented in the
first column.
Table 7. Variance decomposition for Germany
42
Period
INFL
2
95.25406
2.663050
0.519636
1.070778
0.492473
3
90.37006
5.766543
2.434414
0.931480
0.497506
12
76.90336
11.64303
9.658360
0.892353
0.902893
24
71.36686
14.40704
12.42336
0.891236
0.911512
36
68.77240
15.70692
13.71421
0.890496
0.915980
2
6.749104
87.58695
4.177496
0.322413
1.164035
3
7.721058
83.32755
7.059830
0.845895
1.045665
12
19.18457
67.43914
11.68385
0.743025
0.949414
24
21.83531
63.06391
13.60616
0.607881
0.886737
36
22.92902
61.25865
14.39963
0.553415
0.859288
INPRODGR DBANKSPREAD DYIELDSPREAD DUNEMP
Variance Decomposition of INFL:
Variance Decomposition of INPRODGR:
Variance Decomposition of DBANKPROFITSPREAD:
2
0.140843
4.102952
95.66917
0.021457
0.065575
3
0.199071
7.433830
91.74288
0.344293
0.279930
12
3.289667
11.26537
84.99018
0.235396
0.219385
24
3.376305
12.45338
83.85526
0.194110
0.120938
36
3.410733
12.90057
83.42639
0.178697
0.083616
Variance Decomposition of DYIELDSPREAD:
2
2.738169
0.177237
0.956704
96.08966
0.038230
3
3.311263
0.537825
3.628825
92.47693
0.045155
12
4.750039
0.534052
3.270516
91.37406
0.071328
24
5.087525
0.523605
3.298503
91.02250
0.067863
36
5.210792
0.520375
3.308740
90.89360
0.066496
2
2.265661
0.654541
3.282554
0.565883
93.23136
3
2.001995
0.636749
4.009599
0.559600
92.79206
12
0.973864
0.422736
5.400314
0.524024
92.67906
24
0.646242
0.369899
5.736684
0.509081
92.73809
36
0.528484
0.351354
5.858026
0.503559
92.75858
Variance Decomposition of DUNEMP:
Note: This table reports the variance decomposition of variables infl: inflation, inprodgr:
industrial production growth, dbankspread: lending minus deposit rate (first difference),
dyieldspread: 10year rate minus 3month treasury rate (first difference), and dunemp:
unemployment rate (first difference) for Germany. Each row shows the percentage of the
variance of the error in forecasting the variable mentioned in the table, at its forecasting horizon
presented in the first column.
43
Table 8. Variance decomposition for France
Period
INFL
INPRODGR DBANKSPREAD DYIELDSPREAD
Variance Decomposition of INFL:
DUNEMP
2
95.84409
2.350004
1.296475
0.291167
0.218261
3
92.92461
2.375006
3.227709
0.747608
0.725068
12
74.97884
1.818312
13.95281
8.461203
0.788835
24
68.24857
1.475987
18.44312
10.95061
0.881718
36
64.90669
1.308101
20.67634
12.18255
0.926319
Variance Decomposition of INPRODGR:
2
0.920481
93.17718
0.608971
0.318481
4.974886
3
0.943811
93.28337
1.032205
0.472803
4.267813
12
1.144748
91.63233
1.712461
0.874715
4.635743
24
1.260751
91.58299
1.582994
0.992259
4.581011
36
1.321821
91.53789
1.524910
1.053136
4.562247
Variance Decomposition of DBANKPROFITSPREAD:
2
0.345331
0.269904
94.51362
4.766686
0.104458
3
2.078589
0.336281
89.91455
7.061181
0.609399
12
15.02527
0.333538
66.38416
17.74539
0.511638
24
18.13686
0.259746
59.09224
22.03560
0.475561
36
19.57814
0.225341
55.74304
23.99452
0.458953
Variance Decomposition of DYIELDSPREAD:
2
2.519149
0.315514
1.106487
96.02894
0.029907
3
3.648923
0.283396
0.991497
94.91813
0.158057
12
5.015408
0.171880
1.838764
92.82984
0.144105
24
5.464309
0.132686
1.925380
92.34003
0.137593
36
5.637737
0.117201
1.959978
92.14979
0.135298
Variance Decomposition of DUNEMP:
2
0.363728
2.020075
0.373495
0.044377
97.19832
3
0.857641
4.687228
0.428345
0.086614
93.94017
12
1.324871
4.051987
0.260860
0.838983
93.52330
24
1.343281
4.025999
0.151504
0.853861
93.62535
36
1.348691
4.017260
0.109960
0.857400
93.66669
Note: This table reports the variance decomposition of variables infl: inflation, inprodgr:
industrial production growth, dbankspread: lending minus deposit rate (first difference),
dyieldspread: 10year rate minus 3month treasury rate (first difference), and dunemp:
unemployment rate (first difference) for France. Each row shows the percentage of the
variance of the error in forecasting the variable mentioned in the table, at its forecasting
horizon presented in the first column.
44
Table 9. Variance decomposition for Italy
Period
INFL
INPRODGR DBANKSPREAD DYIELDSPREAD
Variance Decomposition of INFL:
DUNEMP
2
94.89847
0.024369
1.665805
3.408352
0.003007
3
94.72130
0.467643
1.811368
2.995872
0.003817
12
94.06687
0.717772
1.012857
3.459059
0.743438
24
95.10583
0.578261
0.616079
3.139689
0.560141
36
95.67542
0.506579
0.446197
2.927860
0.443940
Variance Decomposition of INPRODGR:
2
1.391478
95.62400
1.037358
0.192330
1.754830
3
2.106358
94.55567
0.972277
0.388002
1.977697
12
4.536450
89.11901
2.442059
1.952693
1.949787
24
4.732224
89.70371
2.522717
1.371461
1.669889
36
4.837728
90.04541
2.573966
1.059072
1.483824
Variance Decomposition of DBANKPROFITSPREAD:
2
1.527296
0.989717
96.54983
0.533520
0.399633
3
1.315455
0.605831
56.84921
40.89524
0.334261
12
3.304396
3.790090
49.15061
42.12198
1.632921
24
3.082779
4.149019
45.21765
45.91129
1.639264
36
2.802106
4.296926
42.34263
48.97209
1.586245
Variance Decomposition of DYIELDSPREAD:
2
1.711670
3.400514
6.950740
87.26356
0.673515
3
4.787288
4.702914
7.074013
82.78657
0.649210
12
4.646964
5.409547
11.66812
77.28892
0.986450
24
3.687640
5.298860
14.08196
75.89580
1.035736
36
3.057958
5.283594
15.28137
75.32924
1.047842
Variance Decomposition of DUNEMP:
2
0.186632
1.197051
0.180133
1.387431
97.04875
3
0.570518
1.263935
0.349150
3.120783
94.69561
12
1.146443
1.154846
0.539456
4.558862
92.60039
24
0.916357
1.030961
0.464200
4.182895
93.40559
36
0.734529
0.959462
0.377582
4.024356
93.90407
Note: This table reports the variance decomposition of variables infl: inflation, inprodgr:
industrial production growth, dbankspread: lending minus deposit rate (first difference),
dyieldspread: 10year rate minus 3month treasury rate (first difference), and dunemp:
unemployment rate (first difference) for Itlay. Each row shows the percentage of the variance
of the error in forecasting the variable mentioned in the table, at its forecasting horizon
presented in the first column.
45
Figure 1. Impulse Responses for Euroarea
Impulse Responses of Industrial Production
Impulse responses of Eonia
.16
10
DEONIA
DEXCHRATEDOLLEURO
INDPRODGR
INFL
DUNEMP
.12
DEONIA
DEXCHRATEDOLLEURO
INFL
DUNEMP
INDPRODGR
8
6
4
.08
2
0
.04
-2
-4
.00
-6
-8
-.04
1
2
3
4
5
6
7
8
9
10
11
1
12
2
3
Impulse Responses of Inflation
1.6
5
6
7
8
9
10
11
12
Impulse Responses of Exchrate D/E
DEONIA
DEXCHRATEDOLLEURO
INFL
DUNEMP
INDPRODGR
1.2
4
.035
DEONIA
DEXCHRATEDOLLEURO
INFL
DUNEMP
INDPRODGR
.030
.025
.020
0.8
.015
0.4
.010
.005
0.0
.000
-0.4
-.005
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Figure 2. Impulse Responses for USA
Impulse Responses of Bank Spread
Impulse Responses of Fed Funds rate
.16
DFEDFUNDS
DBANKPROFITSPREAD
DYIELDSPREAD
INDPRODGR
INFL
DUNEMP
.12
.25
DFEDFUNDS
DBANKPROFITSPREAD
DYIELDSPREAD
INDPRODGR
INFL
DUNEMP
.20
.15
.08
.10
.05
.04
.00
.00
-.05
-.10
-.04
1
2
3
4
5
6
7
8
9
10
11
1
12
2
3
DFEDFUNDS
DBANKPROFITSPREAD
DYIELDSPREAD
INDPRODGR
INFL
DUNEMP
.6
5
6
7
8
9
10
11
12
Impulse Responses of Inflation
Impulse Responses of Industrial Production
.7
4
.5
.30
DFEDFUNDS
DBANKPROFITSPREAD
DYIELDSPREAD
INDPRODGR
INFL
DUNEMP
.25
.20
.4
.15
.3
.10
.2
.05
.1
.00
.0
-.1
-.05
1
2
3
4
5
6
7
8
9
10
11
12
46
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3. Impulse Responses for UK
Impulse Responses of Inflation
Impulse Responses of Sonia rate
.4
.28
DSONIA
DBANKPROFITSPREAD
DYIELDSPREAD
INFL
INPRODGR
DUNEMP
.24
DSONIA
DBANKPROFITSPREAD
DYIELDSPREAD
INFL
INPRODGR
DUNEMP
.3
.20
.16
.2
.12
.1
.08
.04
.0
.00
-.04
-.1
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
.16
DSONIA
DBANKPROFITSPREAD
DYIELDSPREAD
INFL
INPRODGR
DUNEMP
0.8
12
Impulse Responses of Bank Spread
Impulse Responses of Industrial Production
1.0
11
DSONIA
DBANKPROFITSPREAD
DYIELDSPREAD
INFL
INPRODGR
DUNEMP
.12
0.6
.08
0.4
.04
0.2
.00
0.0
-0.2
-.04
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
Figure 4. Impulse Responses for Germany
Impulse Responses of Industrial Production
Impulse Responses of Yield Spread
1.6
.30
DYIELDSPREAD
DBANKPROFITSPREAD
DUMMY
INPRODGR
INFL
DUNEMP
.25
DYIELDSPREAD
DBANKPROFITSPREAD
DUMMY
INPRODGR
INFL
DUNEMP
1.2
.20
0.8
.15
.10
0.4
.05
0.0
.00
-0.4
-.05
1
2
3
4
5
6
7
8
9
10
11
1
12
2
3
Impulse Responses of Inflation
5
6
7
8
9
10
11
12
Impulse Responses of Unemployment
.4
DYIELDSPREAD
DBANKPROFITSPREAD
DUMMY
INPRODGR
INFL
DUNEMP
.3
4
.06
DYIELDSPREAD
DBANKPROFITSPREAD
DUMMY
INPRODGR
INFL
DUNEMP
.05
.04
.03
.2
.02
.1
.01
.00
.0
-.01
-.1
-.02
1
2
3
4
5
6
7
8
9
10
11
12
47
1
2
3
4
5
6
7
8
9
10
11
12
Figure 5. Impulse Responses for France
Impulse Responses of Yield Spread
Impulse Responses of Inflation
.4
INFL
INPRODGR
DBANKPROFITSPREAD
DYIELDSPREAD
DUNEMP
.30
INFL
INPRODGR
DBANKPROFITSPREAD
DYIELDSPREAD
DUNEMP
.25
.3
.20
.15
.2
.10
.1
.05
.00
.0
-.05
-.1
-.10
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
.08
INFL
INPRODGR
DBANKPROFITSPREAD
DYIELDSPREAD
DUNEMP
1.2
12
Impulse Responses of Unemployment
Impulse Responses of Industrial Production
1.4
11
1.0
INFL
INPRODGR
DBANKPROFITSPREAD
DYIELDSPREAD
DUNEMP
.06
0.8
.04
0.6
0.4
.02
0.2
0.0
.00
-0.2
-.02
-0.4
1
2
3
4
5
6
7
8
9
10
11
1
12
2
3
4
5
6
7
8
9
10
11
12
Figure 6. Impulse Responses for Italy
Impulse Responses of Yield Spread
.6
Impulse Responses of Inflation
INFL
INPRODGR
DBANKPROFIT SPREAD
DYIELDSPREAD
DUNEMP
.5
.4
.20
INFL
INPRODGR
DBANKPROFIT SPREAD
DYIELDSPREAD
DUNEMP
.16
.12
.3
.08
.2
.1
.04
.0
.00
-.1
-.04
-.2
1
2
3
4
5
6
7
8
9
10
11
1
12
2
3
Impulse Responses of Industrial Production
1.6
5
6
7
8
9
10
11
12
Impulse Responses of Unemployment
INFL
INPRODGR
DBANKPROFIT SPREAD
DYIELDSPREAD
DUNEMP
1.2
4
.20
INFL
INPRODGR
DBANKPROFITSPREAD
DYIELDSPREAD
DUNEMP
.16
.12
0.8
.08
0.4
.04
0.0
.00
-0.4
-.04
1
2
3
4
5
6
7
8
9
10
11
12
48
1
2
3
4
5
6
7
8
9
10
11
12