Decomposing leverage in Quantitative Easing decisions : Evidence from the [618788]

1
Decomposing leverage in Quantitative Easing decisions : Evidence from the
UK

Dionisis Philippas1
Department of Finance, ESSCA School of Management, 55 Quai Alphonse Le Gallo, 92513, Boulogne, Paris,
France, [anonimizat]

Stephanos Papadamou
University of Thessaly, Department of Economics, 43 Korai str., 38333, Volos, Greece, [anonimizat]

Iuliana Tomuleasa
Department of Economics, University of Auvergne, 41 Blvd. F. Mitterand, 63000, Clermont -Ferrand, France,
ioana_iuliana.tomuleasa@etu.u damail.fr

Abstract
The paper analyses the implications arising from the responses of the financial sector in the
United Kingdom to the incentives determined by quantitative easing decisions. In a panel
vector autoregressive framework, we examine the effec ts of Bank of England asset purchases
on the profitability and disaggregated leverage components for different types of financial
institutions, which reflect differences in the sequencing of the quantitative easing strategy. We
find that quantitative easin g decisions are driven by economic activity, lending rates, and
financial institutions’ leverage. The transmission channel of QE on boosting economic growth
depends on the degree of financial institutions’ leverage and the securities holdings, but with a
diverging magnitude on the different types of the UK financial institutions.

Keywords : quantitative easing; financial institutions; leverage decomposition; panel VAR
JEL c lassification : G1, G21 , G28, E52

1 Corresponding author: Dionisis Philippas, Department of Finance, ESSCA School of Management, 55 Quai Alphonse Le
Gallo, 92513, Boulogne, Paris, France, [anonimizat], +33(0)241734747.

2
1. Introduction
The global financial crisis that started in 2008 and its aftermath, posed significant challenges
for monetary authorities. U nconventional monetary policies remain one of the few levers
available for policy makers to exercise, with the most common referred to the extension of their
balance sheets by large -scale asset purchases (LSAPs), known as quantitative easing (hereafter
QE). The QE strategy was initially applied by the Bank of Japan as it tried to handle the
Japanese real estate bubble and the deflationary pressures in the early 2000s. Afterwards, the
Federal Reserve System (Fed) and the Bank of England (BoE) followed suit in the late 2000s,
acting swiftly in order to evade a meltdown of their financial system.
Traditionally, QE is focusing on buying longer -term government bonds from b anks,
allowing the sovereign yields to serve as a benchmark for the pricing of riskier privately issued
securities (Krishnamurthy and Vissing -Jorgensen, 2011). In this context, the yields on privately
issued securities and consequently bank lending rates , are expected to decline in parallel with
those on government bonds, with the hope that this stimulates longer -term investments and
hence aggregate demand, thereby supporting price stability. However, recent studies
underlined the importance of banks on the effectiveness of QE strategy. Bowman et al. (2015)
argued that there was a positive effect of bank liquidity positions on lending. Moreover, Joyce
et al. (2012) note that banks may hold onto funds to improve their viability rather than on –
lending to the p rivate sector, driving the central banks to intervene with the direct provision of
credit, in order its policies to have an impact on the financial intermediation.
The paper analyses the interaction between leverage undertaken by different type of
financia l institutions and asset purchases by the BoE as part of its QE program and future QE
exit strategies , oriented to t he UK financial institutions , allowing them to enjoy vast financial

3
conditions .2 Addressing this issue is a challenge, because it is of grea t interest to disentangle
the implications of the effects of QE decisions for the UK financial sector. The setting of
monetary policy is done under several pressures that could force to abruptly change the policy
strategies being promoted, within a wide va riety of financial and macroeconomic signals.
Consequently, it is crucial question raised to what extent the critical role of the UK financial
sector’s leverage can ensure the success of quantitative easing. In periods with high deleverage,
QE is successful if it reduces the risks of a liquidity shortfall, encouraging the banks to extend
credit to higher interest -paying parties through the leverage decisions undertaken and thereby
boost economic growth, even though the banks are forced to undertake more risks.
Nevertheless, given the level of leverage the banking sector can experience, banks can stop
intermediating loans and may not pass on the additional liquidity to the real economy, thereby
making the QE policy ineffective.
Even though there is a considerable empirical literature concerning the broader
macroeconomic impact of QE via market rates ,3 few studies , to the best of our knowledge,
examined the impact of QE on the profitability and solidity of financial institutions focusing
mainly on US d ata (Lambert and Ueda, 2014; Montecino and Epstein, 2014; Mamatzakis et
al., 2015; Mamatzakis and Bermpei, 2016). These studies argue that unconventional monetary
policy reinforces bank solidity by allowing them to reduce leverage and extend the maturity o f
their debt. A handful recent studies attempt to highlight the role of financial institutions’
leverage decisions but for the case in the conduct of conventional monetary policy, business

2 During the first and second QE programs spanning from M arch 2009 to November 2012, the BoE purchased
£375 billion of medium – and long -term government bonds (representing approximately 24% of domestic GDP).
As a result, the balance sheet of the UK financial institutions has been significantly expanded due to th e liquidity
support. In 2013, the UK banking sector is 450% as a share of GDP in 2013 on a residency basis.
3 A strand of the literature has focused on the transmission channels through which asset purchases can affect
long-term interest rates by observing the policy signalling channel and portfolio balance channel. Contributors,
among others, are the studies of Meier, 2009; Joyce et al., 2011a; Joyce et al., 2011b; Christensen and Rudebusch,
2012; Hamilton and Wu, 2012; Joyce and Tong, 2012; D’Amico et al. , 2012; Gilchrist and Zakrajšek 2013;
Steeley, 2015; Neely, 2015. Fewer studies try to estimate the macroeconomic effects of unconventional monetary
policy measures via the linkages between interest rate spreads and the real economy (Lenza et al., 2010; Ch ung
et al, 2011; Chen et al., 2012).

4
cycles and real economic activity in USA (Geanakoplos, 2010; Serlet is et al., 2013; Istiak and
Serletis, 2016).
In light of the above discussion, it is important to go in further considerations when
discussing QE strategic policy interactions. We address these issues from a different angle that
innovates and contributes in filling some existing gaps in the literature in at least t wo
dimensions.
Firstly, we set up a panel vector autoregressive (panel VAR) framework, characterized by
cross‑sectional heterogeneity and dynamic interdependencies. We make two assumptions
within our modelling framework. In the first assumption, we employ different major types of
UK financial institutions and discuss to what extent QE has exerted different impacts on their
performance. This type of identification tries to shed light in a significant gap for the vital
importance of different types of UK financial institutions in studying the implications of QE
decisions, without been oriented narrowly on a macroecon omic perspective. In the second
assumption, we consider a decomposition of leverage into three main components , namely
gross loans to equity , the liquid assets to equity , and the securities to equity components . We
then analys e their discrete role on the Q E policies implemented and the ir interactions to real
economic activity for the different types of UK financial institutions . These types of
identification above differentiate our paper from other studies employing similar empirical
methodologies or addres sing related topics.
Secondly, we draw the policy implications based on both directions of impulse and
response functions between the QE strategies and the performance of UK financial institutions’
balance sheets, assessing the following main research ques tions. The first question is the
impulse analysis of QE on balance sheets and to what extent the financial variables of interest
can play a key role on the GDP growth. The second question investigates the QE policy
response to different shocks of leverage, profitability and real economic activity. At the last

5
one, we go in depth to examine the effects of leverage to profitability and the interactions across
the leverage components.
Our findings are of great importance to the existing literature by highlight ing both
directions of impulses and responses between the profitability and leverage of financial sector
and the central bank’s QE policies for real economy. The first finding is that asset purchases
by the BoE are not a determining factor that provide ban ks with the possibility to improve
profitability, a finding that is in line with the study of Mamatzakis and Bermpei (2016). A
significant reduction of profitability is identified for almost all the types of the UK financial
institutions, with a diverging magnitude between these types, mainly due to the securities which
are held and the diversification benefits of other institutions by their involvements in different
sectors of activity. Moreover, we observe an interdependency between profitability and
leverage and also, an indirect relationship between liquidity and lending relationship, which
depends on the type of financial institution. However, our paper recognizes that the significant
reduction of profitability for Real Estate banks present significant benefits for the economic
activity in UK.
The transmission channels of QE on GDP growth based on financial institutions’ leverage
have a significantly positive effect through securities holding , for Commercial banks and Bank
Holding companies. This second finding compliments previous studies about the positive effect
of conventional monetary policy on GDP via leverage in USA (see Geanakoplos, 2010; Adrian
and Shin, 2010; Serletis et al., 2013; Lambert and Ueda, 2014; Istiak and Serletis, 2016). The
contribution of commercial banks in liquidity and leverage responses to the QE shock is of
great importance and, consequently, the increase of leverage seems to be mainly attributed to
risk-taking behavior by commercial banks. The evidence show that a negative shock on
economic activity leads the majority of the UK financial institutions to increase their leverage
by undertaking significantly high risks, which indicating countercyclical effects. This result is

6
in contrast with Adrian and Shin, (2009 and 2010) arguing for procyclical behaviour of
leverage found in US over periods of convention al monetary policy .
The third finding is the evidence that the QE is also transmitted to real economy via the
significant reduction of the retail banking rates, in comparison with other studies focusing only
on the transmission via bond rates (i.e. Joyce et al., 2012; Pesaran and Smith, 2016; Weale and
Wieladek, 2016). We argue that the BoE reduces asset purchases when lending rates are
dropped, economic activity is augmented and leverage of commercial banks is increased. As
pointed out by Putnam (2013), exit strategies from QE by central banks could be awfully
challenging to implement and have the potential to suspend a return to the normal conduct of
monetary policy to the detriment of longer -term economic growth, rational leverage and
potential future i nflation.
The remainder of the paper is organized as follows. Section 2 provides a detailed description
of the data sources and draws some initial insights from a fundamental data analysis. Section
3 discusses the panel VAR framework, including the modelli ng assumptions. Section 4
illustrates the empirical findings, together with a discussion of the results and their policy
implications. Finally, Section 5 concludes.

2. Data selection
A part of our sample comes from the Bankscope database4 and covers the annual accounting
data of the financial institutions in the UK , for a period spanning from 2005 to 2013.5 However,
to ensure potential uniformity , which can be affected by the presence of missing data in

4 We should bear in mind that accounting data derived from Bankscope may suffer from a drawback, observing
that when inferences are drawn from the Bankscope database, there can be an implicit selectivity bias (Corvoisie r
and Gropp, 2001).
5 The quarterly frequency could, in principle, give better insight into the link between the accounting ratios and
the QE rounds. However, for most banks the quarterly data are not available. On the other hand, the bias in the
results o btained using annual data instead of quarterly data appears to be not significant (Gambacorta, 2005).

7
Bankscope, in some cases we use the annual reports of the financial institutions for the variables
of interest as data sources .
The timespan structure was chosen to segregate the impact of QE rounds and diminish the
likelihood of other puzzling factors (i.e. , purchases of other asset classes during successi ve QE
rounds) . Moreover, it can capture transformations observed in the UK financial s ector in recent
years. In the period preceding the crisis, UK financial institutions came to increasingly depend
on wholesale funding rather than their customer’s deposit s, an element that placed higher
pressure on their structure. At the brink of the financial crisis in the UK, financial institutions
ended up hav ing less capital and fewer liquid assets than they had had in the past, given the
fluctuations in the UK ’s financial environment. Thus, our timespan structure can evaluate the
overall impact of QE on the UK financial sector without segregating the impact of different
QE rounds.
We draw on two accounting quantities , which are associated with the present research study .
The first quantity straightforwardly derived from Bankscope is the returns on assets (ROA) ,
used as a key ratio for the evaluation of bank profitability and, as a measurement of the overall
performance of a financial institution regarding its effici ency in utilizing assets to generate
profits, given the structure of liabilities and equity (Athanasoglou et al., 2008; Garcia -Herrero
et al., 2009). The second one is the ratio of leverage , measuring the risk associated with non –
capital funding of overall balance sheets, and defined as total assets to total shareholders’
equity and subordinated debt . This definition is similar to the regulatory leverage ratio used by
the Office of the Superintendent o f Financial Institutions (OSFI), it is based on total regulatory
capital as defined in Basel II, including subordinated debt (Bordeleau et al., 2009) and it is not
subject to the model and measurement errors associated with asset -risk calculations. A high
leverage indicates a greater vulnerability to adverse shocks that can reduce the overall value of

8
assets. Similarly, it can decrease the long -term availability of funding and , in addition , increase
the reliance on volatile short -term sources of funding (i. e., higher funding liquidity risk).
Moreover, we drawn on three quantities derived from Bankscope namely the liquid assets ,
defined as the sum of cash and cash equivalents, public securities, and secured short -term loans,
the gross loans and the sum of securities , defined as the sum of investments of banks that
include bonds, equity derivatives and any other type of securities . We divide all the three
quantities over total shareholders’ equity to derive them as ratios. In this setting, leverage
defined above is decomposed in three components , denoted as liquid assets to equity , loans to
equity and securities to equity , which reflect to what extent the financial institutions are
(de)leveraging within the QE framework effect . This framework of deco mposition may expose
the financial sector's access to liquidate assets and its resilience to short -term liquidity stress,
whether it can provide loans to real economy and to withstand adverse non -performing loans'
shocks and, it can measure to what extent a financial institution should leverage in riskier
market securities and financing sources and can adverse market risks, respectively.
In standard quantitative easing framework, it is common to assume that the central bank sets
its policy interest rate t aking into account real -economy variables, e.g., the real GDP, the output
gap, the inflation deviation from target and so on, when deciding upon the amount of QE it will
engage in. In this context, we draw on the real GDP derived from BoE and examine to what
extent it may have an impact on bank performance due to the fact that the demand for lending
increases during cyclical upswings (Athanasoglou et al., 2008). Moreover, we derive the
lending rate, as the average long term rate from BoE, to examine the extent the lending between
banks is decreasing.6 This choice of lending rate relies on the hypothesis that certain bank –
specific characteristics (e .g., size, liquidity, short -term funding, cost -to-income proportion and

6 Gambacorta and Iannotti (2007) find that the interest rate adjustment in response to positive and negative
shocks is asymmetrical, in that banks adjust their lending rate faster during periods of monetary tightening.

9
capitalisation) only in fluence the loan supply. Finally, we derive the average annual asset
purchases made by the BoE over its total assets as an indicator of QE , which is commonly used
in the literature (Hancock and Passmore, 2011; Chen et al., 2012).
Using the Bankscope database, the types of financial institutions are not always mutually
exclusive (Bhattacharya, 2003) . Consequently, we have restricted our sample to five main types
of financial institutions in the UK , which are mutually exclusive. Even though the analysis is
implemented on a total sample of more than 300 financial institutions, the contribution of each
type of financial institution s to the QE responses is investigated further, given that each type
may reveal significant information. However, due to the data availab ility and and the low
relevance of some financial institutions to QE practices , the empirical analysis is focused on
five major types . Table 1 presents the types and the number of institutions included over the
period studied.

Table 1: UK financial institutions
Type of financial institution Number of
financial institutions Label
Commercial banks 76 ComB
Private Banking & Asset Management companies 32 PrivB
Real Estate & Mortgage banks 43 RealB
Investment banks 42 InvB
Bank Holding s companies 20 BkHo
Total 213
Note : The table presents the types and the number of UK financial institutions included over the period studied.
Bankscope divides financial institutions by specialisation, as follows: Commercial banks, Savings banks,
Investment banks, Real Estate and Mortgage banks, Cooperative banks, Credit banks, Islamic banks, Non –
Banking Credit institutions, Bank Holdings companies, Central Bank , Specialised Governmental Credit
institutions, and Multilateral Government banks. In terms of the distinctions between the five different types
presented in the table, Commercial banks are regarded as the financial institutions which are owned by
stockholders pursuing various lending activities to increase their profits. Real Estate an d Mortgage banks are
specialized on real estate lending. Investment banks are underwriters that serve as intermediary between issuer of
securities and the investing public. Private Banking and Asset Management companies are focused on the
management of a c lient's current investments. Finally, Bank Holding s companies own or control one or more
banks.

10
Next, we rely on some statistical analysis to provide insights that can further motivate our
analysis. The findings here are not decisive for the main conclusions of the paper, but they
offer a preliminary perspective of the data. Table 2 illustrates the mean and the standard
deviation for the variables of interest by type of the UK financial institutions. The idea behind
this is to examine whether the t ypes of UK financial institutions with comparable averages
have heterogeneous deviations from the mean .

Table 2: Summary statistics of accounting ratios for UK financial institutions

Commercial
banks Investment
banks Bank Holding
companies Private
Banking
companies Real Estate
banks
Ratio Mean SE Mean SE Mean SE Mean SE Mean SE
Leverage 13,508 9,271 10,865 8,012 15,034 11,073 19,470 10,492 17,792 5,525
Loans to
Equity 5,894 6,005 3,724 3,955 6,921 6,913 4,901 5,854 13,565 4,524
Liquid assets
to Equity 4,993 4,392 3,342 4,577 4,549 4,488 11,867 11,485 2,281 1,580
Securities to
Equity 2,621 3,410 3,799 4,793 3,564 3,834 2,702 3,386 1,946 1,581
ROA 0,679 1,992 1,133 5,156 1,940 3,666 0,827 2,166 0,387 0,370
Note : The table presents the summary statistics of the accounting ratios of interest for UK financial institutions,
namely the leverage and its components loans to equity , liquid assets to equity and securities to equity , and the
ROA. The panel illustrates the mean and the standard deviation of the UK financial institutions by type.

Comparing the results suggested in Table 2, we drawn on some interesting findings. Firstly,
there is a comparable (or close to) mean value between the types of financial institutions,
although their deviations are highly heterogeneous, suggesting that distinguishing financial
institutions by type and examining their partial contribution to the QE program can play a key
role because they all a re quite sensitive to unconventional shocks but differ in their degree of
sensitivity. Moreover, the heterogeneity of the l everage ’s components across the types of
financial institutions indicates short -term liquidity stress . To provide further insights ab out the
distribution of the leverage's components that can motivate the distinguishing of financial

11
institutions by type, we derive the histograms of the three components of leverage for the five
types of UK financial institutions, as shown in figure 1.

Figure 1: Histograms of leverage’s components by type of financial institutions

Note : The figure illustrates the histograms of the three components of leverage namely loans to equity , liquid
assets to equity and securities to equity , for all the types of UK financial institutions.

The findings indicate strong evidence of heterogeneity between the different types of
financial institutions across the components, indicating the handling of different processes for
each type of financial institutions , an element that robust our hypothesis not to consider all
financial institutions within the same modelling framework . This adjustment is in line with the
White Paper of Vickers Commission (2012) report, even though these measures are planned to
enter int o force in 2019 and, therefore, the effects will only become visible later on. When
010 20 30 40
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40Commercial Banks Investment Banks Real Estate Banks Bank Holding Companies Private Banking CompaniesPercent
Loans to Equity
050
0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30Commercial Banks Investment Banks Real Estate Banks Bank Holding Companies Private Banking CompaniesPercent
Securities to Equity

12
reviewing the loans to equity component, the majority of financial institutions have values
below 10%, while there are outliers in all the types with values that exceed 20%. This implies
that they promote a very aggressive growth strategy being accompanied by a correspondingly
increased insolvency risk. In the case of the liquid assets to equit y, there is evidence of a high
value crossing the 40% level for a few cases of Investment bank s, Real Estate banks and Bank
Holding s compan ies, indicating that they have high -quality liquid assets that can be converted
easily and immediately into cash. This fact can be confirmed by the results obtained for the
securities to equity compo nent, where it was registered for these institutions a high value of
this component , meaning that they deal with creditworthy securities with short -term maturities.
In 2010, the UK Financial Services Authority (FSA) addressed again the issue of liquidity,
adopting a tighter regulation with the purpose of withstanding new stress scenarios and to make
the financial system more resilient to major risks that placed pressure on the performance of
UK financial institutions , such as the economic downturn, borrowe r defaults, pressures in
funding markets, credit conditions and sovereign risk. At a minimum, the conditions for
achieving this objective are higher spreads on lending activities and reduced leverage.
Achieving these goals would imply a rebalancing of the financial institutions’ funding profiles
and a more focused approach on the activities that exploit their comparative advantage. In
reality, the transition determined a trade -off between deleveraging and revenue generation.
Though, as shown in fig ure 1, th is regulation framework had an impact, particularly on
Commercial banks and Bank Holding companies where a large part of the institutions ensure
a minimal level of liquidity.

3. Model setup
The panel VAR framework is a coherent approach to estimating interdependencies by treating
all variables as endogenous and allowing time lags across variables . Recent relevant studies
have used empirical panel VAR modelling frameworks with different structural identification

13
approaches to address a variety of issues such as the transmission of shocks across units,
countries , and time .7 In a panel VAR framework, a cross -sectional dimension is added to the
common VAR representation that may reveal additional information about interdependencies .
Within a panel VAR approach, we obtain financial institutions’ dynamic responses to shocks
because of the model’s ability to approximate complicated, interdependent adjustment paths
with the time -series information . On the other hand, we can control for individual heterogene ity
and can specify the time varying relationships between dependent and independent variables .
Without loss of generality, we illustrate the specification of our panel VAR framework,
assuming one lag. Let 𝑦𝑖,𝑡 be the 𝑘𝑖×1 vector of endogenous variables for each unit 𝑖, 𝑖=
1,…,𝑁. The 𝑘𝑖×1 vector of endogenous variables takes the form 𝑌𝑖,𝑡=[𝑦′1,𝑡… 𝑦′𝑁,𝑡]′.
The panel VAR is written as:
𝑌𝑖,𝑡=𝐴𝑖,0+𝐴𝑖(𝑙)𝑌𝑖,𝑡−1+𝑢𝑖,𝑡 (1)
where 𝐴𝑖,0 is the vector of all the deterministic common components (e.g., constants, seasonal
dummies, and deterministic polynomial in time ) of the data for all units 𝑖, 𝑡 denotes the time
parameter where 𝑡=1,…,𝑇, coefficients 𝐴𝑖(𝑙), and 𝑢𝑖,𝑡 is the 𝐺×1 vector of
contemporaneously correlated random disturbances with zero mean and the non-singular
variance -covariance matrix Σ𝑢.
Assuming t hat the data generating process features dynamic homogeneity, the pooled
estimation approach with fixed effects can be used to estimate the parameters of the model by
potentially capturing idiosyncratic but constant heterogeneities across variables and units.
However, if different assumptions are imposed i n the model specification (e.g., for 𝑁 and 𝑇),
the pooled e stimation approach is biased. One way to overcome this difficulty is to employ the
generalized method of moments ( GMM ) approach initially proposed by Arellano and Bo nd

7 Contributors, among other, are Canova and Ciccarelli (2009), Beetsma and Giuliodori (2011), Canova et al.
(2012), Ciccarelli et al. (2013), De Graeve and Karas (2014).

14
(1991 ). According to them, when the cross -sectional size (number of units, denoted as 𝑁) is
large, 𝑇 is fixed and small and, given the fact that lagged regressors are used as instruments,
the first assumption is derived by estimating the model parameters with the GMM procedure ,
which is consistent when 𝑇 is small. Nevertheless, the GMM app roach also requires
differencing model specifications.
In this paper, we impose two assumptions to obtain plausible results . The first assumption
of the panel VAR framework derived herein is that cross -sectional heterogeneity and dynamic
interdependencies are assumed by introducing fixed effects , thus allowing for time -variant
individual characteristics.8 Therefore, t he panel VAR is characterized by dynamic
interdependencies where the lags of all endogenous variables of all units enter the model for
every unit 𝑖, cross -sectional heterogeneity where innovations are correlated
contemporaneously, where intercept, the slope and the variance of the shocks 𝑢𝑖,𝑡 may be unit-
specific. In th is setting, we impose a block structure on the matrix of contemporaneous
coefficients (i.e., short -run restrictions) to compute structural parameters prior to generating
impulse -response functions, based on the stud y of Frame et al. (2012).
Under the firs t assumption and a common set of 𝐿≥𝑘+𝑙 instruments, recall equation (1)
in a compact form:
𝑌𝑡=𝑍𝑡𝐴+𝑈𝑡 (2)
where 𝑌𝑡 is the vector of the endogenous variables, 𝑍𝑡=𝐼𝑁𝐺×(𝐴0 𝑦′𝑖,𝑡1), which contains all
the remain ing deterministic common components of the data for all units 𝑖, 𝐴=(𝐴𝑖(𝑙))′=
(𝑎′𝑖)′ with 𝐺𝑘×1 vectors, and 𝑈𝑡 is the 𝐺𝑁×1 vector of innovations serially correlated
contemporaneously with zero mean and variance -covariance matrix Σ𝑢. The individual

8 One way to address implicit selectivity bias of our accounting data is to use fixed effects in order to ensure
robustness in the empirical analysis in relation to non -random selectivity, rather than the random effects estimator.

15
heterogeneity is endorsed in the levels of the variables .9 Subtracting the means of each variable
calculated for each firm -year and by introducing fixed effects, eliminates any bank -specific
time dummies that ca pture aggregate and global shocks which may a ffect all firms in the same
way and preserves the orthogonality between transformed va riables . Since 𝐴 varies with cross –
sectional units, it depends on a lower dimension vector that prevents any meaningful
unconstrained estimation. For a structural interpretation, we use the following standard linear
accounting identity, as:
𝑌𝑡=∑𝑍𝑡𝛾𝑗𝜗𝑗+𝑈𝑡+𝑍𝑡𝑒𝑡 𝑗 (3)
where 𝑍𝑡𝛾𝑗 can capture any potential common, unit-specific, variable -specific , and lag-specific
information in the regressors, 𝜗𝑗 are factors that capture the determinants of 𝐴, and 𝑒𝑡 is the
error term of the linearization. The decomposition allow s us to measure the common and unit –
specific influences for endogenous 𝑌𝑡. Finally, the equation -by-equation GMM estimation
yields consistent estimates of panel VAR , where j oint estimation of the system of equations
makes cross -equation hypothesis testing straightforward (Holtz -Eakin et al. , 1988 ). To robust
the GMM estimator, we test the optimal lag order in both panel VAR specification and moment
condition using the moment and model selection criteria (MMSC) for GMM models based on
the J statistic of over -identifying restrictions proposed by Andrews and Lu (2001).
The dynamics of the model can be investigated by impulse res ponse analysis (IRF). The
IRFs are informative for the shocks and interactions arising between the endogenous variables
of the system. The standard errors of the impulse response functions and confidence intervals
are generated using Monte Carlo simulation s. The impulse response function is derived to one
standard deviation shock to equation 𝑗 corresponding to variable 𝑘 at time 𝑡 on expected values
of 𝑌 at time horizon 𝑡+ℎ.

9 Within this context, if the data generating process features dynamic heterogeneity, both a within – and a between –
estimator will give inconsistent estimates of the parameters, even when N and T are large, since the error term is
also likely to be correlated with the endogenous regressors.

16
The second model assumption is identifying as a restricted version of the pan el VAR
framework, and examines dynamic heterogeneity in the responses to shocks that m ay arise for
different consistent formulations of the cross -sectional panel. Suppose we run the model for
one type of financial institution denoted as 𝑑, from the full panel sample. Comparing the
impulse response functions obtained for the 𝑑-type financial institutions each time, allows us
to roughly assess the contribution of the 𝑑-type institutions. Therefore, the restricted vector to
be estimated in equation (3) is now specified as:
𝑌𝑡∗=[𝑦′1,𝑑,𝑡… 𝑦′𝑁,𝑑,𝑡] (4)
where 𝑌𝑡∗ is the 𝑘𝑖,𝑑×1 vector of endogenous variables for unit 𝑖, 𝑖=1,…,𝑁 and 𝑑 denotes
the type of financial institutions examined for the restricted model setup. In addition, suppose
we run the model excluding one of the variables in the full endogenous vector, denoted as
(𝑘𝑖−1)×1. This form of the restricted model is obtained by the exclusion of the 𝑘-variable
and it can reveal the contribution of the omitted variable in the impulse response functions of
the 𝑑-type restricted model. The restricted vector to be now estimated in equation (4) is given
as:
𝑌𝑘,𝑡∗=[𝑦′1,𝑑,𝑡(𝑘𝑖−1)… 𝑦′𝑁,𝑑,𝑡(𝑘𝑖−1)] (5)
where 𝑌𝑘,𝑡∗ now is the (𝑘𝑖−1)×1 vector of endogenous variables included in the restricted
model setup for unit 𝑖, 𝑖=1,…,𝑁 and (𝑘𝑖−1).
We estimate the panel VAR model repetitively, for all the five major categories of financial
institutions mentioned in Section 2, under the second model assumption. The cross -sectional
interactions within the different types of financial institutions can each time reflect the extent
to which the institutions are subject to QE imposed by the central bank. Finally, we expect that
central banks pay part icular attention to the performance of the components in the endogenous
vector, compared to all the other type of banks in conducting monetary easing policies, given
their size, number and importance as traditional financial intermediaries.

17

4. Empirical findings under the model setup
In this section, we present the empirical results from the panel VAR model framework
illustrated and discuss the implications associated with the present research questions. We start
by select ing the optimal lag length for panel VAR framework, using MMSC for the GMM
models based on the J statistic of over -identifying restrictions (Andrews and Lu , 2001). The
first-order lag specification is chosen to insure no serial correlation of residuals in the VARX
models after estimatin g the model. Finally, we bear in mind that when computing the
bootstrapped error bands by simulating the model, we use the sample covariance matrix, since
the number of endogenous variables in our model is lower than the dimension of the time -series
includ ed. Under the model assumption s, our panel VAR framework is repetitively estimated
for all types and the 𝑑-type of UK financial institutions with the analysis focus ing on IFRs (one
standard deviation) .10

4.1 Quantitative Easing impulses and transmission to GDP growth
We start the empirical analysis by setting the QE effect impulses and the transmission to the
UK real GDP growth, as shown in Table 2 (panels A and B). The first important finding is the
evidence that during the period of the positive shock o f QE, profitability ( indicating by ROA )
of Commercial banks and Real Estate banks is reduced significantly , highlighting their role,
compared to the others. This finding is in line with Mamatzakis and Bermpei (2016) who
identifies a reduction of profitability of US banks during quantitative easing implementation
by the Fed.

10 Analysing the response of the financial sector to shocks resulting from the QE policy, it is implicitly assumed
that the variables of interest respond within the period to the BoE QE policy. We simulate the model 5,000 times
to obtain confidence intervals and median estimates for the impulse responses. In addition, we perform forecast –
error variance decomposition (FEVD) analysis on the dynamics of the model setup under the model assumptions,
derived after 5,000 runs. The FEVD is interpreted as the impact ac counted for by innovations in each variable in
proportion to the total impact of all innovations reported over the horizon ahead selected. The results are not
reported but are available to the reader upon request.

18
However, the finding above also contributes to the ongoing debate (i.e., separate ban king
system reported in White Paper , Vickers Commission , 2012) , by highlights the significant
difference across different types of UK financial institutions. This effect can be beneficial for
the real economy when looking on the effect of a positive shock of ROA on real GDP growth
after one period for Real Estate banks, Investment banks and Commercial banks. In the case of
Real Estate banks, there is a significant relationship between ROA and GDP growth with one –
year time lag. The reduction of their net int erest margin may be attributed to the reduction of
lending rate implying significant benefits for the real economy. Therefore, the above finding
adds to the transmission channel of QE to real economy via ROA for Commercial banks and
the Real Estate banks.
Another significant finding derived from table 2, is the evidence that the positive shock of
the QE coexists with the significant increase on the securities to equity for Commercial banks
and Bank Holdings companies. Real GDP growth responds positively and significant after one
period to QE positive shock. Therefore, these two types of banks may contribute to the UK real
GDP growth, because of their significant activity in terms of asset leverage. Moreover, the drop
of liquid assets to equity for Private Ba nking companies and Bank Holdings companies may
contribute to the increase of real GDP growth , given the response of the later to a positive shock
on the liquid assets to equity , for these types of banks. Finally, the results of table 2 provide
evidence that the positive shock on QE leads to a significant reduction of lending rate with
beneficial effects on real GDP growth for all cases of financial institutions, amplifying the
investors’ mood , in line with the study of Lutz (2015).

Table 2: Impulses of QE and transmission to GDP
Note : The figure presents the responses of all the financial variables of interest to a quantitative easing shock. The
thin black line represents the median estimate of the response. The shadow area around the median estimate li ne
of response represents the 95% confidence bands generated from 5,000 Monte Carlo bootstraps resamplings . To
avoid any misunderstanding, in the table we denote the leverage components, namely securities to equity , loans
to equity and liquid assets to equity as "Securities to Equity", "Loans to Equity" and "liquid to Equity",
respectively.

19

QE Impulses for
Private Banking
companies
QE impulses
for Private
Banking
companies
GDP responses
in case of Bank
Holding
companies
QE impulses
for Bank
Holding
companies
GDP
responses in
case of Real
Estate banks
QE impulses
for Real
Estate banks
GDP responses in
case of
Investment banks
QE impulses
for Investment
banks
GDP
responses in
case of
Commercial
banks
QE impulses
for Commercial
banks

20
4.2 The role of financial institutions’ variables on the GDP growth response and QE shock
We further explore the role of financial institutions’ variables on the effects of quantitative
easing to the real GDP growth and on profitability. To do so, we repetitively estimate our panel
VAR framework , by excluding each time one relevant variable of interest and compare the
responses with the ones from the full baseline framework. Table 3 presents the results associate
with this q uestion. The main finding of our analysis herein is that the leverage component
securities to equity amplifies the effect of QE shock on real economic activity for Bank
Holdings companies, and to a less degree for Private Banking companies and Commercial
banks. In the case of Real Estate banks, the effect of QE shock on GDP growth is much less
positive with the exclusion of profitability ( ROA ). Therefore, the transmission of quantitative
easing decisions on real activity pass through the effect on ROA of Real Estate banks.
The monetary policy makers keep the net interest margin low for the case of Real Estate
banks which may add to the efficiency of the transmission of monetary policy to the real
economy. In the majority of the cases, ROA responses to a positive QE shock are negative with
the exception of the Private Banking companies. However, when the securities to equity is
omitted, ROA also response in the same manner for this type of financial institutions, following
the others’ ROA response to QE . Therefore, the leverage component securities to equity is of
great importance, providing a tool to the Private Banking companies not to experience a
significant reduction on their profitability. In the case of the Bank Holdings companies, the
same levera ge component has beneficial effect by reducing the negative effect of QE on ROA .
These results have significant implications for bank managers when facing a significant easing
monetary policy. A well -diversified bank strategy to interest and non -interest income activities
may reduce the negative effect of QE strategy on bank profitability.

Table 3: Responses of the real GDP and ROA to a positive QE shock – Identifying the
role of omitted variables for the types of financial institutions

21
Real GDP Response to a QE shock ROA response to a QE shock
Commercial Banks

Investment Banks

Real Estate Banks

Bank Holding Companies

Private Banking companies
-0.0500.050.10.150.2
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.12-0.1-0.08-0.06-0.04-0.0200.02
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.0200.020.040.060.080.10.120.14
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.25-0.2-0.15-0.1-0.0500.050.1
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.0500.050.10.150.2
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.08-0.07-0.06-0.05-0.04-0.03-0.02-0.010
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.08-0.07-0.06-0.05-0.04-0.03-0.02-0.010
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.14-0.12-0.1-0.08-0.06-0.04-0.020
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA

22

Note : The figure presents the responses of the real GDP and ROA to a positive QE shock for 10 period -horizons
ahead. The blue line with rhombuses represents the sample with all financial institutions included, the red line
with the squares represents the sample when the securities to equity (denoted as Sec/Equity) is excluded, the green
line with the triangles represents the sample when the loans to equity (denoted as Loans/Equity) is excluded, the
purple line with 2 -ray asterisks represents the sample when the liquid assets to equity (denoted as Liquid to Equity)
is excluded and the light blue line with 3 -ray asterisks represents the sample when the ROA is excluded . Statistical
significance is obtained from 5,000 Monte Carlo bootstrap resamplings .

4.3 Does QE policy respond to shocks of leverage and profitability?
Table 4 show s the responses of the BoE QE policy to leverage and profitability. The findings
illustrate that the BoE reduces asset purchases when a positive growth shock occurs and
increases asset purchases when a positive le nding rate shock exists. Looking into the financial
institutions’ variables, we observe a significant reduction of asset purchases as evidence after
a positive shock on the leverage component securities to equity for Commercial banks. The
same finding holds for Bank Holdings and Private Banking companies but with the absence of
the statistical significance. Our findings also provide evidence that the BoE seems to be
interested for the increased profitability of Re al Estate banks given their importance on the
lending activity and its effect on the real economy. The response of the QE variable is positive
after a positive shock in profitability for the Real Estate banks, in order to reduce the lending
rates and to he lp the economy boost, given the significant role of this type of financial
institutions on housing lending.

Table 4: The BoE QE policy response on leverage and profitability
Commercial banks Investment banks Real Estate banks Bank Holdings
companies Private Banking
companies
00.020.040.060.080.10.120.140.16
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA
-0.15-0.1-0.0500.050.1
0 1 2 3 4 5 6 7 8 9 10
All Without Sec/Equity
Without Loans/Equity Without Liquid/Equity
Without ROA

23

Note : The figure presents the response functions of QE to all type of macroeconomic and financial shocks. The
thin black line represents the median estimate of the response. The shadow area around the median estimate line
of response represents the 95% confidence bands generated from 5,000 Monte Carlo bootstraps resamplings . To
avoid any misunderstanding, in t he table we denote the leverage components, namely securities to equity , loans
to equity and liquid assets to equity as "Securities to Equity", "Loans to Equity" and "liquid to Equity",
respectively .

4.4 Does economic activity affect leverage and profitab ility?
We address this question by testing the impulses of real GDP growth to the financial
institutions’ v ariables of interest . Table 5 illustrates the results of IFRs for all the types of
financial institutions. The findings are of great interest, indica ting a number of aspects. Real
GDP growth has a major positive effect on Real Estate banks’ and Bank Holdings companies’
profitability and to lesser extend on the profitability of Commercial banks and Private Banking
companies (first row of table 5). A sec ond main finding is that a negative shock on real GDP
growth may increase the securities to equity for three out of five types of financial institutions
namely Commercial Banks, Real Estate Banks and Bank Holdings companies (second row of
table 5). Moreove r, the leverage component loans to equity increases after a negative GDP
growth shock for the Real Estate and Commercial banks adding to their leverage. The liquid

24
assets to equity is reduced in case of a negative GDP growth shock for Commercial banks and
Bank Holdings companies adding more to their risk profile, while for Real Estate banks is
increased lowering their risk profile. Our results imply that risks are undertaken, when
economic conditions are worse. This is especially apparent for Commercial banks and Bank
Holding companies. By increasing their leverage, these institutions hope to resist on a potential
reduction in their profitability, due to low economic activity. Howeve r, this may increase
significantly their risk, given that bad conditions in the economic environment leading them to
loses. Even though the monetary authorities are afraid of deleverage over weak economic
growth, they should take measures for bank capital adequacy due to a possible worsening of
economic conditions.

Table 5: Effect of e conomic activity impulses to leverage and profitability
Commercial banks Investment
banks Real Estate
banks Bank Holdings
companies Private Banking
companies

Note : The figure presents the profitability ( ROA ) responses to a shock from the three components of leverage,
across the different types of UK financial institutions, for 10 period -horizons ahead. The thin black line represents
the median estimate of the r esponse. The shadow area around the median estimate line of response represents the

25
95% confidence bands generated from 5,000 Monte Carlo bootstraps resamplings . To avoid any
misunderstanding, in the table we denote the leverage components, namely securiti es to equity , loans to equity
and liquid assets to equity as "Securities to Equity", "Loans to Equity" and "liquid to Equity", respectively .

4.5 Does profitability responds significantly to leverage components’ shocks?
Next in our analysis, we notice some interesting aspects by comparing the magnitude across
the financial institutions’ variables (i.e. profitability and leverage components). We start by
examin ing if leverage undertaken increases profitability. Table 6 ill ustrates our findings of
ROA responses to leverage shocks, for the different type of financial institutions. The majority
of our results indicate that there is no evidence of increased profitability due to a leverage
shock. A positive shock on the leverage components reduces significantly ROA of Real Estate
banks. This finding implies that increased leverage leads to non -profitable risky activity. A
positive shock on loans to equity has a positive but statistically not significant effect on ROA ,
only for th e cases of Investment banks and Private Banking companies. Based on this finding
managers may have additional information to what extent an increase in loans to equity
contribut es to bank profitability.

Table 6: ROA responses to leverage component s’ shoc ks
Commercial banks Investment banks Real Estate banks Bank Holdings
companies Private Banking
companies

26

Note : The figure presents the profitability ( ROA ) responses to a shock from the three components of leverage,
across the different types of UK financial institutions, for 10 period -horizons ahead. The thin black line represents
the median estimate of the response. The shadow area around the median estim ate line of response represents the
95% confidence bands generated from 5,000 Monte Carlo bootstraps resamplings . To avoid any
misunderstanding, in the table we denote the leverage components, namely securities to equity , loans to equity
and liquid assets to equity as "Securities to Equity", "Loans to Equity" and "liquid to Equity", respectively .

Going our analysis in depth in this last step , we examine the interaction of leverage
components and the effect of ROA on these components. The findings are pres ented in Table
7 (panels A, B and C). The results of panel A in table 7 shows some interesting aspects. First,
the higher the profitability for Commercial and Real Estate banks the higher their leverage
component securities to equity . A significant decreas e in liquidity leads to higher securities to
equity for all the type of financial institution, implying a substitution effect between liquidity
and securities. Another interesting finding is the positive significant response of securities to
equity on loans to equity for three out of four types of financial institution types. Among them,
the highest response presented on Investment banks , followed by Commercial banks, Bank
Holdings companies and Real Estate banks. Consequently, when significant amount of lo ans

27
are given over equity then a significant amount of securities are also bought in terms of equity.
Therefore, these two leverage components are complimentary for these types of financial
institutions. An increased lending to real economy may be used as a signal indicator of the
trend in security markets driven mainly via main types of financial institutions.
We present the response of loans to equity to the rest of banking variables shocks at panel
B in table 7. There is evidence of a unidirectional eff ect from loans to equity to securities to
equity shock for all the type of financial institutions. This finding implies that the leverage in
securities is complementary to leverage in l oans. Considering profitability effects, higher
returns on asset leads to higher loans to equity with the exception of Bank Holdings companies.
A positive shock on liquidity leads to higher loans after three to four periods ahead for Real
Estate banks and lowe r loans for Investment banks. Th e implications arising from this finding
are of great importance, because it indicates the different behaviour of how the different types
of financial institutions manage their liquidity usage. Real Estate banks in contrast to
Investments banks has a higher contribution on economic growth, leaving space for
discretionary policy by the BoE.
We finally turn our analysis to the liquidity impulses and responses. The results are shown
at panel C in table 7. There are two main fin dings emerging from this panel. We note that a
positive shock on loans to equity leads all the type of financial institutions to increase their
cash holdings. However, in the case of Real Estate banks, the response of liquid assets to equity
to loans to eq uity fades out smoothly and slowly, without being statistically significant after
the third period ahead. The second finding is the positive response of liquid assets to equity to
a positive shock of securities to equity for Investment banks. This finding implies a higher level
of conservatism compared to other types of financial institutions, a finding also presented in a
lower degree for Commercial banks. When the leverage component of securities is increased,

28
it is followed by a higher level of cash hold ings, while profitability shocks do not statistically
affect profitability.

Table 7: Leverage impulse responses and profitability shocks
Panel A
Securities to Equity responses to leverage component s and profitability shocks
Commercial banks Investment banks Real Estate banks Bank Holdings
companies Private Banking
companies

Panel B
Loans to Equity responses to leverage component s and profitability shocks
Commercial banks Investment banks Real Estate banks Bank Holdings
companies Private Banking
companies

29

Panel C
Liquid Assets to Equity responses to leverage component s and profitability shocks
Commercial banks Investment banks Real Estate banks Bank Holdings
companies Private Banking
companies

30

Note : The table (panels A, B and C) presents the interaction between profitability ( ROA ) and the three components
of leverage, across the different types of UK financial institutions, for 10 period -horizons ahead. The thin black
line represents the median estimate o f the response. The shadow area around the median estimate line of response
represents the 95% confidence bands generated from 5,000 Monte Carlo bootstraps resamplings . To avoid any
misunderstanding, in the table we denote the leverage components, namely securities to equity , loans to equity
and liquid assets to equity as "Securities to Equity", "Loans to Equity" and "liquid to Equity", respectively .

5. Conclusion
Considerable efforts have been made by the central banks in recent years to effectively provi de
a sufficient monetary stimulus to their economy during recent global and domestic downturns
and ensur e the sound functioning of financial sectors. In the UK, the financial institutions are
the main collectors of funds and suppliers to the non -financial and households’ sectors;
therefore, a strong understanding of the UK financial institutions' role during the
implementation of BoE QE strategy is vital because it raises a series of concerns regarding the
economic spin -off that could be triggered through these monetary policy decisions. The paper
gauges how the different types of UK financial institutions’ le verage responded to the

31
incentives determined by the QE decisions realized in BoE asset purchases, using a panel VAR
framework.
We find that QE decisions are driven mainly by real economic activity , lending rates, and
to a diverging degree of the leverage components with differen t effects on the five main types
of UK financial institutions. The findings highlight the crucial role played by Commercial
banks in explaining these interrelationships. When the BoE proceeds to positive shock on asset
purchases, t he financial institutions’ profitability is significantly reduced . Turning to the
relationship between unconventional monetary policy and financial institutions’ leverage, we
find that QE rounds seem to have a positive effect on the leverage components, im plying riskier
behavior during QE rounds for busting the real economy .
The quantitative easing policies aim to increase the money supply by inundating financial
institutions with capital in a struggle to encourage lending and implicitly liquidity . Our study
presents that during the implementation of QE strategy, the leverage of the banking sector is
increased. This implies a signal of credit easing conditions that were disappeared during the
involvement of the financial crisis. The decrease of ban ks’ profitability implied negative
signals from the financial sector to the monetary authorities in order to reduce unconventional
easing strategies and assess financial stability, which is the main goal derived from these
policies. Moreover, given the hig h uncertainty and low interest rates, it can be observed the
heightened risk -taking behaviour of financial institutions, as a response to a possible restraint
on their policy choices. This attitude pro -risk has a high potential to influence the market pric e
of risk in the economic system. Likewise, a higher level of risk affects financial sector's stability
and soundness, particularly if the additional risk is condensed in systemically important
financial institutions. As a result, these issues accentuate t he policymakers' concerns related to
the limitation of financial institution's risk taking behaviour.

References

32
Adrian, T., and Shin, H.S., 2009. Money, Liquidity, and Monetary Policy. American
Economic Review 99, 600 –605.
Adrian, T., and Shin, H.S., 20 10. Liquidity and leverage. Journal of Financial
Intermediation 19, 418 –437.
Andrews, D.W.K., Lu, B., 2001. Consistent model and moment selection procedures for
GMM estimation with application to dynamic panel data models. Journal of Econometrics 101,
123–164.
Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo
evidence and an application to employment equations. Review of Economic Studies 58, 277 –
297.
Athanasoglou, P.P., Brissimis, S.N., Delis, M.D., 2008. Bank -specific, i ndustry -specific
and macroeconomic determinants of bank profitability. Journal of International Financial
Markets, Institutions and Money 18, 121 –136.
Beetsma, R., Giuliodori, M., 2011. The effects of government purchase shocks: Review
and estimates for th e EU. Economic Journal 121, F4 –F32.
Bhattacharya, K., 2003. How good is the Bankscope database? A cross -validation exercise
with correction factors for market concentration measures. BIS Working Paper No.133.
Bordeleau, E., Crawford, A., Graham, C., 2009. Regulatory constraints on bank leverage:
issues and lessons from the Canadian experience. Bank of Canada Discussion Paper No. 2009 –
15.
Bowman, D., Cai, F., Davies, S., Kamin, S., 2015. Quantitative easing and bank lending:
Evidence from Japan. Journal of I nternational Money and Finance 57, 15 –30.
Canova, F., Ciccarelli, M., 2009. Estimating multi -country VAR models. International
Economic Review 50, 929 –959.

33
Canova, F., Ciccarelli, M., Ortega, E., 2012. Do institutional changes affect business
cycles? Evidence from Europe. Journal of Economic Dynamics and Control 36, 1520 –1533.
Chen, H., Cúrdia, V., Ferrero, A., 2012. The macroeconomic effects of large -scale a sset
purchase programmes. Economic Journal 122, F289 –F315.
Christensen, J.H.E., Rudebusch, G.D., 2012. The response of interest rates to US and UK
quantitative easing. Economic Journal 122, F385 –F414.
Chung, H., Laforte, J.P., Reifschneider, D., Williams, J.C., 2010. Have we underestimated
the likelihood and severity of zero lower bound events? Journal of Money, Credit and Banking
44, 47 –82.
Ciccarelli, M., Maddaloni, A., Peydró, J.L., 2013. Heterogeneous transmission mechanism:
Monetary policy and financia l fragility in the eurozone. Economic Policy 28, 459 –512.
Corvoisier, S., Gropp, R., 2001. Bank concentration and retail interest rates. ECB Working
Paper No. 72.
D’Amico, S., English, W., López -Salido, D., Nelson, E., 2012. The Federal Reserve’s
large -scale asset purchase programmes: Rationale and effects. Economic Journal 122, F415 –
F446.
De Graeve, F., Karas, A., 2014. Evaluating theories of bank runs with heterogeneity
restrictions. Journal of European Economic Association 12, 969 –996.
Frame, W.S., Hanco ck, D., Passmore, W., 2012. Federal Home Loan Bank Advances and
Commercial Bank Portfolio Composition. Journal of Money, Credit and Banking 44, 661 –684.
Gambacorta, L., 2005. Inside the bank lending channel. European Economic Review 49,
1737 –1759.
Gambacor ta, L., Iannotti, S., 2007. Are there asymmetries in the response of bank interest
rates to monetary shocks?. Applied Economics 39, 2503 –2517.

34
Garcia -Herrero, A., Gavila, S., Santabarbara, D., 2009. What explains the low profitability
of Chinese banks? Jou rnal of Banking and Finance 33, 2080 –2092.
Geanakoplos, J., 2010. Solving the present crisis and managing the leverage cycle. Federal
Reserve Bank of New York Economic Policy Review, 101 –131.
Gilchrist, S., Zakrajšek, E., 2013. The impact of the Federal Reserve’s large -scale asset
purchase programs on corporate credit risk. Journal of Money, Credit and Banking 45, 29 –57.
Hamilton, J.D., Wu, J.C., 2012. The effectiveness of alternative monetary policy tools in a
zero lower bound environment. Journal of Mon ey, Credit and Banking 44, 3 –46.
Hancock, D., Passmore, W., 2011. Did the Federal Reserve’s MBS purchase program lower
mortgage rates? Journal of Monetary Economics 58, 498 –514.
Holtz -Eakin, D., Newey, W., Rosen, H.S., 1988. Estimating vector autoregressions with
panel data. Econometrica 56, 1371 –1395.
Istiak, K., Serletis, A., 2016. A note on leverage and the Macroeconomy. Macroeconomic
Dynamics 20, 429 –445.
Joyce, M., Miles, D., Scott, A., Vayanos, D., 2012. Quantitative easing and unconventi onal
monetary policy – An introduction. Economic Journal 122, F271 –F288.
Joyce, M., Tong, M., Woods, R., 2011b. The United Kingdom’s quantitative easing policy:
Design, operation and impact. Bank of England, Quarterly Bulletin 51, 200 –212.
Joyce, M.A.S., L asaosa, A., Stevens, I., Tong, M., 2011a. The financial market impact of
quantitative easing in the United Kingdom. International Journal of Central Banking 7, 113 –
161.
Joyce, M.A.S., Tong, M., 2012. QE and the gilt market: A disaggregated analysis.
Econom ic Journal 122, F348 –F384.
Kapetanios, G., Mumtaz, H., Stevens, I., and Theodoridis, K., 2012. Assessing the economy
wide effects of quantitative easing. The Economic Journal 122, F316 –F347.

35
Krishnamurthy, A., Vissing -Jorgensen, A., 2011. The effects of qu antitative easing on
interest rates: Channels and implications for policy. Brookings Papers of Economic Activity
43, 215 –287.
Lambert, F., Ueda, K., 2014. The effects of unconventional monetary policies on bank
soundness. IMF Working Paper No. 14/152.
Lenz a, M., Pill, H., Reichlin, L., 2010. Monetary policy in exceptional times. Economic
Policy 62, 295 –339.
Lutz, C., 2015. The impact of conventional and unconventional monetary policy on
investor sentiment. Journal of Banking and Finance 61, 89 –105.
Mamatzak is, E., Bermpei, T., 2016. What is the effect of unconventional monetary policy
on bank performance?. Journal of International Money and Finance 67, 239 –263.
Mamatzakis, E., Matousek, R., Vu, A.N., 2015. What is the impact of bankrupt and
restructured loan s on Japanese bank efficiency?. Journal of Banking and Finance,
doi:10.1016/j.jbankfin.2015.04.010.
Meier, A., 2009. Panacea, curse, or non -event? Unconventional monetary policy in the
United Kingdom. IMF Working Paper No. 09/163.
Neely, C.J., 2015. Unconventional monetary policy had large international effects. Journal
of Banking and Finance 52, 101 –111.
Pesaran, M.H., and Smith, R.P., 2016. Counterfactual analysis in macroeconometrics: An
empirical investigation into the effects of quantitative easi ng. Research in Economics 70, 262 –
280.
Putnam, B.H., 2013. Essential concepts necessary to consider when evaluating the
efficiency of quantitative easing. Review of Financial Economics 22, 1 –7.
Serletis, A., Istiak, K., Gogas, P., 2013. Interest rates, lev erage, and money. Open
Economies Review 24, 51 –78.

36
Steeley, J.M., 2015. The side effects of quantitative easing: Evidence from the UK bond
market. Journal of International Money and Finance 51, 303 –336.
Weale, M., and Wieladek, T., 2016. What are the macro economic effects of asset
purchases?. Journal of Monetary Economics 79, 81 –93.
WHITE PAPER, 2012. Banking reform: Delivering stability and supporting a sustainable
economy. HM Treasury and the Department for Business, Innovation and Skills, London.