Universiteit van Amsterdam
The Public Sector Purchase Programme and its Effectiveness
on Loan Provision within the Euro-Area
Aytac Avci
Master Thesis
14-Jul-17
Student Number: 11373075
Email: aytac.avci@student.uva.nl
Supervisor: Dr. Konstantinos Mavromatis
Program: MSc Economics – International Economics and Globalisation Word Count: 10669
Statement of Originality
This document is written by Student Aytac Avci who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
List of Abbreviations
ABSPP Asset-Backed Securities Purchase ProgrammesAPP Asset Purchase Programme
CBPP- Covered Bonds Purchase Programme
CSPP Corporate Sector Purchase Programme
DFR Deposit Facility Rate
ECB European Central Bank
EMU European Monetary Union
ESRB European Systemic Risk Board
FE Fixed Effects model
FSI Financial Soundness Indicator
GIIPS Greece, Italy, Ireland, Portugal and Spain)
HH Households
IMF International Monetary Fund
MFI Monetary Financial Institution
NCB National Central Banks
NFC Non-financial Corporations
PSPP Public Sector Purchase Programme
QE Quantitative Easing
RE Random Effects model
UMP Unconventional Monetary Policy
VAR Vector Autoregressive Models
Table of Contents
List of Abbreviations iii
I. Introduction 1
II. Literature Review on Measuring the Real Effects 4
II. I. Literature on Loan Provision 4
II. II. Literature on Unconventional Monetary Policy and its Real Effects 5
III. Monetary & Macro-prudential Policy after the European Sovereign Debt Crisis 8
III. I. Monetary Policy after the European Sovereign Debt Crisis 8
III. III. Macro-prudential Policy after the European Sovereign Debt Crisis 13
IV. Econometric Methodology 15
V. Data 18
V. I. Data on the Dependent Variable 18
V. II. Data on the Liquid Assets to Total Assets Ratio Variable 18
V. III. Data on other Control Variables 19
VI. Results 22
VI.I. Pre-estimation Tests 22
VI.II. Estimation Results 22
VII. Conclusion 29
I.
Introduction
There is a broad consensus that the European sovereign debt crisis was partly self-inflicted and partly imported from the US financial crisis. The impact coming from the US exacerbated the critical economic and financial situation of Iceland and Greece and led to contagion to other Eurozone countries. This was the start of the European sovereign debt crisis of 2010. In particular, international agents had started to doubt the creditworthiness of certain EU countries, which then highlighted previously mispriced risk and as such created uncertainty for economies and the monetary union itself. Eventually, the shocks dramatically hit the real economy of the distressed countries and negatively affected domestic provision of loan to the private sector, investment and employment, which has led to challenging times for the European Union. While Germany and France have absorbed the shocks relatively well, the GIIPS countries (Greece, Italy, Ireland, Portugal and Spain) have extraordinarily suffered for a prolonged time, as their economic exposure to asymmetric shocks was particularly high. These countries had either current account deficits, huge public debts, housing bubbles or weak financial institutions. Since there was the risk of bank failure, governments were concerned with banks that owned domestic and EU wide government bonds. Therefore, the monetary union was in turmoil and, hence, the European Central Bank (ECB) was compelled to act.
As a consequence of the following recession, the European Central Bank decided on lowering policy rates close to zero. This approach rapidly proved to have a major downside as the scope for conventional methods became limited, which later made unconventional methods necessary. One of these new measures is the Public Sector Purchase Programme (PSPP), which is part of the Asset Purchase Programme (APP).1 The open-ended PSPP buys sovereign and supranational bonds in order to bring back inflation to its target level of close to, but below 2 percent.2 However, the effects
1 Also known as quantitative easing (QE).
2 Supranational entities are defined as institutions that are formed by two or more governments. The function of these
institutions is primarily to support the economic development for the member states. In order to fund their activities supranational institutions issue debt securities, such as supranational bonds. Supranational bonds that are located in the
of the unconventional measures are rather ambiguous (Koijen et al., 2016 and Delivorias, 2015). With regards to the purchase programmes, researchers argue that various channels might have influenced the markets through positive as well as negative externalities.3 One important channel is the portfolio-rebalancing effect, which encourages investors to search for alternative more risky assets, preferably equities or corporate bonds and, thus, helps to improve the overall liquidity of financial markets (Koijen et al., 2016 and Delivorias, 2015). Monetary policymakers intend to ease financial constraints of financial institutions, by increasing the value of their assets, which in turn lowers the cost of borrowing for non-financial firms and households. Eventually, the outcome would raise the overall lending and so help economic growth in the euro-area to recover and restore the inflation rate (Koijen et al., 2016 and Delivorias, 2015).
According to the Economic Bulletin by the ECB (Issue 1, 2017), the European economic situation seems to recover, albeit slowly. Researches found that there is an overall trend to grant more loan applications to the private sector, indicating that monetary policy measures might have helped. In particular, since the implementation of unconventional monetary policy (UMP) measures, the aggregate cost of borrowing indicator for both non-financial firms and households decreased.4 Also, the yields on government bonds of the GIIPS gradually declined. Nevertheless, the purchase programmes, especially the PSPP, also raised some doubts. Many economists were concerned about the role of the ECB during the European Sovereign Debt Crisis and the potential financial instability that could have come along (Dabrowski, 2016 and Koijen et al., 2016).
In order to identify the impact and channels of the purchase programmes on the real sector, recent literature mainly used Vector Autoregressive Models (VAR) as well as counterfactual analyses. While some studies focused on the spillover on real economies coming from financial crises, others have concentrated on the effects of UMP measures or of announcements on real economies. In general, studies show that a financial crisis spillover of any kind has a major impact on the real euro-area and are eligible for the PPPS include, for instance, the European Investment Bank, Nordic Investment Bank and the European Union. A detailed list of eligible bond securities is shown in the appendix.
3 Further details will be discussed in section 3.
sector. On the other hand, the impact of UMP measures is more ambiguous within and across countries regarding its effect on financial markets and the real economy.
Since just few studies of the effect of quantitative easing on the response to loan provision have been undertaken so far, the intended contribution of this paper is therefore to extent the existing literature by investigating the links between monetary policy, loan supply and financial activity. Particularly, the thesis analyses the impact of the PSPP on bank lending through the liquidity channel effect.
The effect of the PSPP for loans to households and non-financial firms is heterogeneous mainly across countries. Some countries could significantly improve their loan situation through the PSPP, while others’ worsened. In the regression analysis positive and significant liquidity ratio coefficients indicate that the PSPP relaxes the financial constraints and increases the overall lending behaviour. For some countries loans to non-financial firms seem to be affected differently than loans to households by the PSPP implementation. Empirically this difference manifests itself in a sign reversal and a different magnitude of the effect. All in all, the coefficients are rather small. Therefore a higher amount of purchases are required in order to achieve a substantial result. The results suggest that the PSPP has been predominantly successful for most countries. Furthermore, policymakers should evaluate carefully if the continuation of the PSPP is desirable given the adverse effects and the complexity of the programme.
The remainder of this paper is structured as follows: the second section discusses various findings of the contemporary literature. The third section gives a brief overview of key problems starting shortly before the European Sovereign Debt Crisis as well as of policy decisions. The fourth section describes the structure of the panel regression on loan provision in the euro area. The fifth section illustrates the relevant data. The sixth section describes the results and the final section concludes.
II. Literature Review on Measuring the Real Effects
The European sovereign debt crisis was exceptional as it was partly homegrown as well as partly the aftermath of the US financial crisis. Nevertheless, there is relatively little academic literature on how UMP measures affect the real economic activities of countries, especially for the euro area. Possible explanations could be that many countries never experienced such situations until recently, while there is also a relatively short time period one can consider for the empirical part (Schenkenberg and Watzka, 2011). Recent analyses concentrated intensively on the relationship between financial crises and real effects supporting the assumption that there is indeed a causal relationship between financial crises and the subsequent real economic downturn (Dell'Ariccia et al., 2005, Gropp et al., 2014 and Dwenger et al., 2015). These results give interesting guidance and motivation for this paper. Especially, this motivates the hypothesis that QE measures are not neutral i.e. the programme might have real effects as well. The aim of this thesis is therefore to provide additional insights into the liquidity effect of the Public Sector Purchase Programme, as a consensus has not been yet reached.
II. I. Literature on Loan Provision
As Palenzuela and Dees (2016) point out, household demand seems to have been affected as precautionary savings increased during the post-crisis period, mainly due to higher uncertainty regarding the economic situation. In particular, high and increasing unemployment rates seem to have contributed to the adjustment of the savings decisions. However, Palenzuela and Dees (2016) argue that the adjustment process has not found its equilibrium yet. This is of paramount concern for policymakers, since weak investments and current account imbalances exacerbate the exposure to shocks. A further crisis affecting those countries in any way, could lead to severe consequences for euro-wide economies and the currency itself.
In the case of certain German and US banks, that were in critical situations in the aftermath of the financial crisis, Dwenger et al. (2015) and Gropp et al. (2014) observe a large credit rationing. According to some studies, it is important to differentiate between credit supply and demand for
households and non-financial firms. Gropp et al. (2014) argue that there are two views on the change of household behaviour. Households either suffered from credit rationing and, hence, were forced to alter their consumption behaviour or they were well aware of their new permanent budget constraint and, therefore, adjusted to a new optimal level. The latter would imply that a social planner is restricted in his ability to affect the consumer’s choice. However, their findings suggest that credit supply effects seem to have dominated the demand effects. Dwenger et al. (2015) prove that banks that are suffering losses decrease their credit supply to a much greater extent than healthy banks. Furthermore, they argue that credit rationing heavily affects firms’ behaviour in terms of borrowing, investment and employment. This is because many firms are unable to easily substitute credit from other banks, especially those with little collateral (Dwenger et al., 2015 and Bofondi et al., 2012). In countries where the crisis was particularly severe or access to foreign credit was limited, the private sectors seem to have suffered more from these effects (Dell'Ariccia et al., 2005).
II. II. Literature on Unconventional Monetary Policy and its Real Effects
Some authors have focused on the relationship between UMP shocks and the response of real macroeconomic variables. The results show that there is a significant stimulation of output growth (Schenkenberg and Watzka, 2011 and Baumeister and Benati, 2010), while the results for inflation growth are mixed. Schenkenberg and Watzka (2011) found significant real output growth and a decline in long-term interest rates after UMP shocks for Japan, while an increase in inflation could not be detected. They use a structural VAR model to examine the effects on inflation, production, reserves and Japanese government bond yields. Altavilla et al. (2014) carried out a multi-country VAR analysis with real GDP, consumer prices, money supply M3, retail credit, and government bond rates as variables. Interestingly, the authors found that Italian and Spanish 2-year government bond yields have decreased, namely by 2 percentage points, while German and French bonds were rather unaffected. One reason might be that these groups were sharing similar fundamentals during the European sovereign debt crisis. Baumeister and Benati (2010) use a counterfactual scenario analysis with the result that the asset purchase programmes significantly prevented a deeper recession and deflation in the US and the UK. Acharya et al. (2015) use a different approach, in particular, by looking at Outright Monetary Transaction programme announcement. The authors
find no significant correlation between the announcement and the real economic activity in GIIPS countries, even though UMP measures in general seem to have significantly improved the health of banks. The authors focused on the impact on loan provision, investment and unemployment for the GIIPS countries. Joyce and Spaltro (2014) suggest that the share of well-capitalised banks in the UK plays an important role for total bank lending during times of crises. In other words, macro-prudential policy, such as capital requirements or cap on leverage, might significantly influence the lending channel.
Bowman et al. (2011) study the transmission of monetary policy on banks through the liquidity channel. They investigate the impact of the Japanese quantitative easing programme on the loan provision, with the results that the stimulus had a positive and significant, albeit small effect on bank lending. Koijen et al. (2016) analyse the effect of asset purchases on portfolio holdings and asset prices of institutional investors. They introduce a measurement framework in order to calculate corporate and sovereign credit, equity, and duration risk exposures with regard to the purchase programmes and use an IV estimator to identify the impact of the PSPP on government bond yields. The results suggest that the PSPP reduced bond yields and significantly decreased the duration mismatch risk.5 Claeys et al. (2015) find similar results in response to asset purchases. The authors also argue that the PSPP pushed down the yields and, thus, expect the ECB’s returns of the PSPP to be small in response to the low yield environment. Due to the small loan-to-deposit spread, there were concerns that the PSPP and other purchases might have a negative impact on bank profitability, which eventually could affect the lending behaviour. However, Demertzis and Wolff (2016) see no clear sign that this is generally the case, but agree that there are indeed differences between countries. According to Martin and Milas (2012), it should be noted that QE implementations usually come along with other measures such as fiscal and regulatory policy reform, which complicates the identification of the QE impact.
This paper uses a similar approach to Bowman et al. (2011) for the panel regression analysis in order to study the effects of the PSPP on loan provision growth to the private sector. Furthermore,
5 The duration mismatch is a way for banks to generate profit. Banks use deposits (liability side) of customers to fund
their loans (asset side). While assets have a long duration, liabilities are usually of shorter term. This mismatch can be beneficial or harmful depending on the interest rate environment and the composition of the balance sheet.
the analysis should give an insight whether the findings of the recent literature apply also for the European Monetary Union.
III. Monetary & Macro-prudential Policy after the European Sovereign
Debt Crisis
III. I. Monetary Policy after the European Sovereign Debt Crisis
This section briefly summarizes the history of the European Sovereign Debt Crisis and technique behind the asset purchase programmes that the ECB has launched. The European sovereign debt crisis turned out to be the biggest challenge for the European Union until then. Unlike other countries, the euro members were in an exceptional position and thereby were, to some extent, restricted in their ability to react to a crisis, since a national currency devaluation is no longer feasible for the member states.
Figure 1 shows the development of the yields of long-term government bonds for the GIIPS countries as well as for France and Germany. The yields are tied to the interest rate of the respective country. Since various risks such as political, inflation, interest rate, credit and economic risk are taken into account, the yield of a government bond tells investors how the economic situation of a country should be assessed, which may be either positive or negative. The path of the bond yields highlights three important points. First, since the beginning of the negotiations to form a monetary union, the respective bond yields showed a clear convergence, which was preferable especially for Italy, Spain, Portugal and Greece. The convergence indicated that these bonds were considered as almost risk-free. Second, after the collapse of Lehman Brothers in September 2008 and later the debt crisis in late 2009, the yield of the GIIPS rose again and showed massive spreads vis-à-vis the French and German yields. Finally, since the start of financial help packages, the austerity policies and the PSPP, the yields have been declining again (and for some members even reached negative rates), suggesting that the purchases and other measures were effective in reducing yields.
Source: Data from Long-Term Interest Rate Statistics of the Statistical Data Warehouse (ECB)
Also, financial and external imbalances played a significant role and were identified as main elements of the debt crisis. Financial imbalances often manifest in excessive levels of domestic credit in times of an economic upswing, as shown in table 1 (Lane, 2012). Except for Germany, the provision of credit rose from 1997 to 2007, to the highest level in Spain, and declined afterwards. Additionally, for financial institutions in peripheral countries the accession to the EU was beneficial due to several reasons: a common currency eliminated part of the exchange rate risk for banks when raising international funds. Furthermore, lower interest rates and easier access to credits boosted property demand and consumption (Fagan and Gaspar, 2007, cited in Lane, 2012). However, the excessive construction boom, for instance in Spain, and the related consequences caused a severe level of unemployment and a banking sector that had to cover the immense costs (Issing, 2011).
Germany Spain France Greece Ireland Italy Portugal -1 4 9 14 19 24 29 1993 Jan 1993 Jul 1994 Jan 1994 Jul 1995 Jan 1995 Jul 1996 Jan 1996 Jul 1997 Jan 1997 Jul 1998 Jan 1998 Jul 1999 Jan 1999 Jul 2000 Jan 2000J ul 2001J an 2001J ul 2002 Jan 2002J ul 2003 Jan 2003J ul 2004 Jan 2004J ul 2005 Jan 2005J ul 2006 Jan 2006J ul 2007 Jan 2007J ul 2008 Jan 2008J ul 2009 Jan 2009J ul 2010 Jan 2010 Jul 2011J an 2011J ul 2012 Jan 2012 Jul 2013 Jan 2013 Jul 2014 Jan 2014 Jul 2015 Jan 2015 Jul 2016 Jan 2016 Jul 2017 Jan
Dynamics of 10-Year Government Bond Yields, %
Table 1 Private Credit
Domestic Credit to Private Sector (% of GDP)
1997 2002 2007 2015 France 79.8 75.8 88.6 95.8 Germany 107.5 110.9 96.6 77.9 Greece 30.5 53.3 84.5 113.2 Ireland 80.6 75.4 157.9 54.4 Italy 53.2 62.3 82.1 88 Portugal 77.1 119 142.2 119.9 Spain 76.1 99.4 167.1 119.5
Source: Data from World Development Indicators database
Apart from financial assistance, the euro zone tried to influence the economy and to recover the lending behaviour through monetary help. Before a crisis, agents usually underestimate risk; while in post-crisis period they overestimate risk. This outcome is detrimental especially for households and non-financial firms that are able to maintain a sound credit standing, since in times of financial crises, banks tend to withhold additional money in order to strengthen their financial positions and to prevent an increase of the share of non-performing loans (see figure 2). Especially, Cyprus, Greece and Ireland have suffered from a steep incline in non-performing loans, which has further aggravated these crisis-afflicted economies. As a consequence of the turmoil, even healthy banks considered a “flight to safety” or raising lending spreads. The result of that shortfall was a disproportionally high credit rationing, which eventually affected households, enterprises and unemployment in the euro zone as a whole (Dell'Ariccia et al., 2005).
Source: Data from IMF Financial Soundness Indicator Database
Since the limitations of the conventional monetary policy created the necessity for other measures, the QE programme seemed to be a reasonable additional option. Additional to the Asset-Backed Securities and Covered Bonds Purchase Programmes (ABSPP and CBPP3), the ECB implemented the Public Sector Purchase Programme (PSPP) on March 2015, which comprises, with the recent Corporate Sector Purchase Programme (CSPP), the commonly known Asset Purchase Programme (APP). The APP targets various channels in order to influence the markets and achieve its objectives. Firstly, by lowering bond and credit interest rates, the ECB intends to influence prices, also known as the direct prices effect channel. Secondly, the portfolio-rebalancing channel describes the process of influencing demand behaviour. National central banks buy government or corporate bonds by creating new money. This helps to push up prices and so reduces the yield of these bonds. Investors will look for alternatives with higher yields such as corporate bonds. The increasing demand will decrease the borrowing costs for businesses, which will eventually result in increased consumption, investment and job creation. And finally, the signalling-effect channel is expected to
France Germany Greece Ireland Italy Portugal Spain Cyprus 0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00 40,00 45,00 50,00 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Non-performing Loans to Total Gross Loans Ratio, %
increase the confidence of agents in the shaken financial markets. Not surprisingly, this effect could be observed even before the implementation of the APP. In summary, the intention is to provide banks with additional money in order to grant further loans to the private sector, return inflation to target and boost the economy (Dell'Ariccia et al., 2005).
With regard to the PSPP, the ECB established certain eligibility requirements for the bonds in interest. In particular, the PSPP intends for national central banks (NCB) to buy sovereign and supranational bonds with a liquidity that amounted to € 60 billion per month until September 2016, which has been extended to € 80 billion until March 2017. Since April 2017, the purchases are back to the former level of € 60 billion per month. The purchases depend on the NCB’s share in the ECB capital key. The capital key is derived from the level of GDP and population share of each member. A large population and high GDP would therefore result in a relatively higher share (Koijen et al., 2016 and Dell'Ariccia et al., 2005).6
In addition to the capital key there are four eligibility criteria. Firstly, the bonds are required to have maturities between 2 and 30 years. Secondly, the ECB’s programs are restricted to securities that have an investment grade level, which corresponds to BBB or higher. Thirdly, the purchases are only carried out on the secondary market and have a 25 per cent issue and 33 percent issuer limit, which is imposed to prevent the ECB to be a monetary financing party of a member state as well as to preserve price formation and market functioning. (Dell'Ariccia et al., 2005 and Claeys et al., 2015). The ECB describes the issue limit as the highest share of a single eligible security that the euro-area is willing to hold. Similarly, the issuer limit refers to the highest share of a single eligible security that the euro-area is willing to buy. And finally, eligible securities are prioritised that have a yield higher than the deposit facility rate (currently -0.40%). 7
According to the ECB Governing Council, the programme will be open-ended as long as the medium-term inflation rate is under 2 percent (Claeys et al., 2015). Moreover, it should be noted that
6 Germany accounts for the highest share with 26.3%, followed by France and Italy with 20.7% and 18.0% respectively
(see table 7 in the appendix).
7 The deposit facility rate (DFR) is one of the main interest rate tools for conventional monetary policy used by the
ECB. It is set every six weeks and is applied to overnight monetary deposits by banks. After June 2014 the DFR has been consistently below zero.
the purchase of Greek bonds is not (yet) part of the PSPP. The profits that the ECB generates from the APP is either repatriated to each member’s treasury or serves the purpose of helping members with financial problems. However, given the low yields on sovereign and supranational debts, the returns are not expected to be high (Claeys et al., 2015).
The new mandate of the ECB was not welcomed universally, fearing the ambiguous effects of new monetary policy instruments, potentially leading to negative externalities and to a conflict with the price stability target. Especially, there was the concern of rising financial instability, since the resulting lower yields of safe assets and lower borrowing cost of loans may lead to excessive levels of leverage, liquidity or credit risk. (Stein, 2012, Woodford, 2011, Coimbra and Rey, 2016 cited in Koijen et al., 2016). It may seem paradoxical, since exactly this is the objective of the ECB and, hence, the way out of the crisis. However, according to Koijen et al. (2016) there is the threat that these risks may get concentrated in certain sectors and eventually lead to unexpected financial instability. The authors also argue that such developments can be addressed through prudential policy measures, but the effectiveness of these policies would critically depend on the timely and efficient design. Furthermore, critics were concerned about the independence of central banks as the new instruments, especially the APP, could even be considered as fiscal assistance for banks and euro members, which could eventually lead to a national lethargy in pursuing reforms (Dabrowski, 2016).8
III. III. Macro-prudential Policy after the European Sovereign Debt Crisis
The EU and the ECB agreed that the Asset Purchase Programme would be fruitful in the long-term, only if the members comply with their national responsibility of fiscal discipline and prudential supervision. Besides micro-prudential policy, macro-prudential policy has become particular important for the European Monetary Union (EMU), as membership implies the aforementioned restrictions on the interest and exchange rate channel that non-members usually would not face (Margerit et al., 2017). The EMU hopes to achieve positive externalities with the prudential
8 According to Dabrowski (2016), it was even seen as a probable infringement of “Article 123 of the Treaty of the
supervisory such as lowering financial stability risks and fostering the integration of European financial markets and the euro area itself (Dabrowski, 2016).
Macro-prudential policy is a rather new policy measure and should be seen as complementary to monetary policy. It intends to prevent and mitigate systemic risk by introducing, if necessary, adequate capital buffers/requirements and dynamic provisions on financial institutions in order to influence the loan supply. In particular, prudential policy is set up to respond in a countercyclical manner; raising capital requirements when systemic risk seems to be dangerously high and reducing it, if needed, when risk is overestimated and the economy is in dire need of a push out of a recession. Supervisors monitor the evolution of the financial cycle, which capture systemic patterns in the financial environment. The financial cycle is represented by credit, leverage, debt levels, asset and house prices (Margerit et al., 2017 and Buch et al., 2017).
In the EU, the macro-prudential policy framework is decided on the national, ECB and European level, with the main responsibilities lying on the national level as financial and business cycles are rather a national phenomenon (Margerit et al., 2017 and Buch et al., 2017). The debt levels of the EU members are a good example of the heterogeneity of financial cycles (see Appendix). The European Systemic Risk Board (ESRB) is supervising, coordinating and advising members on threats of systemic risk and potential spillovers. Moreover, the ECB is strongly linked with the ESRB by sharing information and advising the institution on an optimal mitigation of systemic risk (Margerit et al., 2017).
Just like unconventional monetary policy, macro-prudential policy has also not been spared from criticism. Since the ECB plays an important role on the ESRB, concerns were raised about the possible contradictory objectives of financial stability and price stability at a given point in time. Furthermore, as mentioned above, many were concerned that the purchase programme could fuel asset bubbles and thereby increase the threat of systemic risk (Dabrowski, 2016).
Since macro-prudential policy is an important measure for financial regulation and partly implemented by the ECB, it is reasonable to control for potential measures taken after the financial and sovereign debt crisis. Hence, for the analysis the variable Regulatory Capital to Risk Weighted Assets ratio is included, which is discussed in the next two sections.
IV. Econometric Methodology
The hypothesis of this thesis is that the Public Sector Purchase Programme has positively affected the provision of loans through the liquidity channel. By using a fixed effects (FE) panel regression, I study the effect of the interaction between the PSPP and the liquidity ratio of the respective euro member on the loan provision on households and non-financial corporations. In particular, the panel regression follows a similar approach as Bowman et al. (2011), who studied the effect of the Japanese quantitative easing programme on the growth of loans. The FE model estimates the relationship of dependent and independent variables within a country (or entity in general) taking into account the country-specific characteristics that may be permanent and could, thus, influence the independent variables. For instance, the demographic situation or the common attitude towards risk are country specific and, thus, could differ across countries. The FE model controls for these time-invariant characteristics that could bias the estimations. Furthermore, the FE model assumes that these characteristics are unique for each entity and, hence, not correlated with other entities. In particular, an entity differs in its constant and error term and should not be correlated across other entities. If, however, the error terms are correlated, then the random effects (RE) model is more efficient and therefore, more suitable. The estimation period goes from 2008 Q1 to 2017 Q1. The dataset includes 19 individual euro-member listed in the appendix.
The estimations of the models are the following: Δ ln 𝐿𝑜𝑎𝑛!!
!,! = 𝛼 + 𝛽! + 𝛿𝑋!,!!!𝐷!!"!!+ 𝜃𝑋!,!!!+ 𝜀!,! (1)
Δ ln 𝐿𝑜𝑎𝑛!"#
!,! = 𝛼 + 𝛽!+ 𝛿𝑋!,!!!𝐷!!"!!+ 𝜃𝑋!,!!!+ 𝜀!,! (2)
Where Δln Loani,t is the loan growth and is measured as the first difference of the natural logarithm
of country i at time t. Moreover, LoanNFC
i,t and LoanHHi,t represents the loans to non-financial
corporations and households respectively. In order to filter the relevant effect for the topic, the regression uses Xi,t-1 as a vector of control variables, which might affect the prospect of lending. The
growth, bad loan ratio, capital outflow, government bond yields, government expenditure, short-term interest rates.
Some variables are, however, removed, because of multicollinearity. Section 6 elaborates the details of the problem. DiPSPP is the PSPP time dummy variable for each country i, which takes either the
value 1 when the purchase programme is in effect or 0 otherwise. The programme started in March 2015 and is still operating. From this date onwards the dummy will take the value 1 continuously. The IMF considers the liquidity ratio a core financial soundness indicator (FSI). According to the FSI Compilation Guide and Amendment, a higher ratio indicates that financial institutions hold more liquid assets, but also eventually forego profitable investments. 9 Hence, financial institutions may tend to hold the ratio low in order to increase the performance of the asset side. However, the level of liquidity is crucial in order to absorb shocks. A sufficiently large shock can damage the health of financial institutions with a low liquidity ratio and lead to a liquidity crisis.10 Especially depository institutions must consider the possibility of unexpected withdrawals of depositors and, hence, allow for higher liquidity ratios than non-depository institutions. The Guide defines liquid assets as the sum of “currency and deposits and other financial assets that are available either on demand or within three months or less, but deposit takers’ deposits (and other non-traded claims) with other deposit takers in the reporting population are excluded.”
The main focus is on the interaction of the liquidity ratio and the PSPP dummy. The coefficient of the interaction term indicates to what extent the interactions affect the level of loan provision. If the coefficient is significant and positive, the liquidity ratio positively affects loan growth by means of the PSPP. If, however, the coefficient is zero, then the effect of the liquidity ratio on loan growth is independent of the PSPP. Furthermore, if the policy dummy takes the value zero, only the variables that do not interact with the policy dummy become relevant for the analysis.
Additionally, total assets are calculated as the sum of non-financial and financial assets. Moreover, by using lagged values, the problem of simultaneity is mitigated. To check for differences, I also run
9 Source: http://data.imf.org/?sk=51B096FA-2CD2-40C2-8D09-0699CC1764DA
10 In particular, a financial crisis could influence the confidence of depositors and investors and eventually harm the
access to funding. Financial institutions are then forced to sell their assets at lower prices in order to re-establish a healthy liquidity buffer. This outcome could even affect solvent, but illiquid, banks, forcing these banks to declare bankruptcy.
the random effects regression and compare both models. Furthermore, the Hausman test tests whether the fixed effects or random effects model is more appropriate for the estimation.
V. Data
V. I. Data on the Dependent Variable
In order to measure, the impact of the PSPP on loan provision in the euro area, this thesis uses the available data from the Statistical Data Warehouse of the European Central Bank. In particular the ECB provides a rich database of monetary financial institution balance sheet statistics, which is a useful basis for financial analysis in the euro area. The aggregated loans are offered in various ways such as growth rates, flows and stocks. Moreover, it is possible to choose from loans with different maturities. In case for households the database offers different forms like mortgage, consumption and other loans.
As for the raw data, I concentrate on the monthly stock of total loans to euro area non-financial corporations and households for each euro member state, since the dataset is ample and most relevant for the panel regression analysis. The explanatory note further describes that the data is “adjusted for loan sales, securitisation and notional cash pooling (all currencies combined, all maturities, denominated in euro, not seasonally adjusted, outstanding amounts at end of period). Loans adjusted for sales and securitisation include loans, which have been sold or securitised and are no longer reported on banks’ balance sheets.”11 As the regression analysis would incorporate variables of different frequencies, the dependent variables are translated to quarterly frequency.
V. II. Data on the Liquid Assets to Total Assets Ratio Variable
The Liquid Assets to Total Assets ratio is an important indicator for regulators, policymakers and bank managers to measure if the liquid assets are able to improve its asset base. According to Bowman et al. (2011) the asset purchases had a positive influence through the credit channel and so increased the liquidity of Japanese banks. The results suggest that the higher liquidity level has led to an expansion of loan supply. Therefore, the liquidity ratio will be a crucial variable for the thesis. The data for the liquid assets to total assets ratio are taken from the IMF database for core financial
soundness indicators, which is compiled quarterly and on country level. The database provides the most consistent data from 2008 Q1 onwards. Therefore, I will focus on the period from 2008 Q1 to 2017Q1. For some time periods, the database provides data only semi-annually. I, therefore, make use of the linear interpolation approach to circumvent the problem of varying time frequencies. Furthermore, the thesis concentrates on the PSPP as the dummy variable. The Statistical Data Warehouse provides PSPP data for the exact date of the monthly net purchases of every euro area country.
V. III. Data on other Control Variables
To control for other effects that could have an impact on loan provision, it will be useful to include a vector of control variables. For instance, Bowman et al. (2011) suggest using data on bank size, bad
loan ratio and lags of loan growth in order to control for banks’ soundness and other characteristics,
which could have affected the lending behaviour. The next page will give an overview of these and additional control variables. Additionally, as for the estimation I use lagged variables to mitigate potential endogeneity issues.
Table 2
12 Also known as non-performing loans ratio.
Overview of Variables
Variables Description
Bank Size
The Bank Size is measured by total assets that are owned by financial institutions. The statistical data warehouse provides aggregated monthly data for total assets of monetary financial institutions of each euro area country. As I deal with country level data, total assets of financial institutions are therefore chosen for the country as a whole. As for the regression, I include the logarithmic growth of this variable.
Bad Loan Ratio
The Bad Loan Ratio is provided as net non-performing debt instruments [% of total own funds for solvency purposes] by the consolidated banking database of the statistical data warehouse.12 During financial unrest the ratio is normally increasing as the growth prospects decline and banks withhold new loans, which then could further aggravate the performance of loans.
Government Bond Yields
Financial institutions tend to hold government bonds as a low-risk investment. However, if the prices of these bonds are falling and financial institutions are forced to liquidate their assets, this eventually could cause a negative effect on the balance sheets. The statistical data warehouse provides ample data for the countries in question. The most consistent data can be found on the 10-year government bond yields.
Government Expenditure
In times of crises, the governments could raise expenditures in order to push consumption or to bail out financial and non-financial firms. As there is a variety to measure government expenditure, I focus on total government expenditure and on government assistance to the financial sector from the statistical data warehouse.
Variables Description
Capital Outflow
The Capital Outflow is measured as net portfolio investment (assets minus liabilities) provided by the balance of payment database of the statistical data warehouse. A financial account deficit would imply that the respective country faces a net capital outflow during national financial unrest. For the regression, the variable is expressed as the share of GDP. In particular, Eurostat provides the country data for GDP at market prices.
Short-Term Interest Rate
The Short-Term Interest Rates are taken from the MFI Interest Rate Statistics database of the statistical data warehouse. I consider the interest rate for mortgage and consumption loans to households, as well as large and small loans to non-financial corporations. In particular, by taking the average rate the two interest rates are merged for each group. All interest rates have initial rate fixation periods of one year.
Lagged Loan Growth To prevent potential problems with autocorrelation, it would be important to
include the lagged loan growth, i.e. the respective lagged independent variables.
Regulatory Capital to Risk Weighted Assets
The Regulatory Capital to Risk Weighted Assets ratio controls for potential prudential measures by regulators. The IMF provides the data, where the sector-wide Tier 1 capital is the numerator and sector-wide risk-weighted assets is the denominator.
Cost of Borrowing
Cost of borrowing is defined as interest and other potential incurred costs to a
debtor for taking a debt. Usually, borrowing costs tend to increase during economic upswings with rising inflation and market interest rates. This can be the case even if the creditworthiness of households and non-financial firms remains strong. The data for the variable is taken from the MFI Interest Rate Statistics database of the statistical data warehouse. The cost of borrowing is in percentage terms.
VI. Results
VI.I. Pre-estimation Tests
The Hausman test tests whether there is a significant difference between the fixed and random effects estimators. The result supports to use the fixed effects over random effects estimation, since this estimator seems to be more consistent. Additionally, the modified Wald test for groupwise heteroscedasticity in fixed effect models supports the rejection of the null hypothesis of homoscedasticity and leads to the conclusion of heteroscedasticity. I, therefore, use the fixed effects estimation with heteroskedasticity-robust standard errors.
The first regression results show low coefficients and large standard errors, which makes it worth exploring if there are multicollinearity issues. Multicollinearity occurs when the independent variables are highly correlated, leading to inaccurate estimated coefficients. However, there is no formal test for this problem. I, thus, use the variance inflation factor (VIF) output in order to check informally for the potential presence of multicollinearity. Analysts usually determine multicollinearity by examining the variables’ VIF and applying it on an agreed tolerance. This approach suggests removing variables if the highest VIF is greater than 10 and if the mean of all the VIFs is substantially higher than 1. Here, this yields an indication of multicollinearity between two or more independent variables: namely, for the Cost of Borrowing, Short-term Interest Rates, Government Expenditure and Regulatory Capital to Risk Weighted Assets. These variables are removed from the final regression. The final results show a VIF with 2.64 as the highest output and an acceptable mean VIF of 1.79, making a further investigation not necessary. 13
VI.II. Estimation Results
This section presents the main estimation results. The interaction effect of the PSPP and the liquidity channel is estimated by using equation (1) and (2). The estimation results are summarised in table 2 and 3, where heteroscedasticity-robust standard errors are in parenthesis. For the sake of brevity the interaction term for variables other than the liquidity ratio are not shown in the
estimation results, but are listed in the appendix. Firstly, by controlling for other factors, the coefficients on the dummy interactions for households appear to be heterogeneous in terms of sign and magnitude, but mainly significant for most of the countries. A positive and significant outcome suggests that the PSPP was effective in strengthening the liquidity channel by increasing the effect liquidity had on loan provision, relaxing the financial constraints of financial institutions and pushing the household loan supply up. The implied increase without PSPP is given by the coefficient on loan provision.
Secondly, the R-squared overall indicate that the fit is high with 0.5844 for households and 0.4217 for non-financial corporations. Rho is the proportion of variation that explains the individual specific term, while the rest is due to idiosyncratic error. In case for households the proportion explained by the individual specific term is moderate (with 41.12%). The proportion for non-financials corporations turned out to be smaller (with 32.33%).
VI.III. Discussion
VI.III.I. Control Variables
This section discusses the coefficients on the variables without interactions. Structurally interpreted, these coefficients capture the general effect of the variables on loan provision, regardless of the PSPP. Since Bank Size and Lagged Loans enter the regression in log growth rates, their estimated coefficients represent elasticities, i.e. the percentage point change in loan provision growth as a consequence of a one percentage point change in the variable growth rates. All other variables enter the estimation in levels and their coefficients thus represent semi-elasticities, measuring the percentage point change in the loan provision growth rate to a one-unit change in the variables. Since Bad Loan Ratio, Government Bond Yields, Capital Outflow and Liquidity Ratio are expressed in percentage terms, the interpretation of their estimated semi-elasticities coincides with the interpretation of elasticities.
The liquidity ratio coefficient is insignificant for both types of loans (0.0123 for HH, -0.0728 for NFC). On the other hand, the liquidity interaction term is predominantly significant. This suggests that the PSPP was successful in influencing the liquidity channel.
Furthermore, government bond yields have a highly significant and negative effect on the provision of the two loan types. Non-financial corporations’ loan growth seems to be more affected by bond
yields (-0.1230) than households’ (-0.0742). This implies that a one percent point increase in the government bond decreases loan provision by ca. 0.12 percentage points for non-financial corporations and by around 0.07 percentage points for households.
Similarly, the bad loan ratio shows a weak, but slightly higher, magnitude for firms than for households (-0.0749 and -0.0324 respectively). Generally, an increase of non-performing loans seems to decrease the loan growth in the next quarter.
The bank size of euro members is positively and significantly correlated with dependent variables (0.0543 for HH and 0.1805 for NFC). If total assets of financial institutions grow, loan growth will improve. One interpretation could be that banks benefit from a more diversified asset portfolio and are thus able to reduce their risks.
Additionally, the household loan growth depends on the previous level of loans with 0.3572 at the 1 percent significance level, while for non-financial corporations the lagged loan growth is slightly lower with 0.2792.
Finally, the capital outflow shows a weak negative relationship with firm loan growth (-0.0746 at the 1 percent significance level) and weak positive for households (0.0292 at the 5 percent significance level). An increase of capital outflow would imply that investors and financial institutions prefer to invest abroad, which then pushes down the growth of non-financial corporation loans for the respective country. On the other hand, an increase of capital outflow would increase the growth of household loans.
The interaction between control variables and the PSPP dummy is significant for most countries. This implies that the PSPP significantly altered the effect the control variables had on both types of loan provision growth.
VI.III.II. Effects of PSPP on the liquidity channel
This section describes the effects the PSPP had on the strength of the liquidity channel in the individual countries which is captured by the interaction terms of the country-specific PSPP dummy variable and the liquidity ratio. It should be noted that some interaction terms are omitted because
of collinearity or because the liquidity ratio data for the respective country is insufficient.14 In general, the coefficients on the interaction terms indicate how the PSPP has changed the effect of liquidity on loan provision: a significant and positive coefficient implies that the implementation of the PSPP has strengthened the liquidity channel and liquidity has, hence, a larger effect on loan provision. A significantly negative coefficient implies the opposite. An insignificant coefficient would imply that the PSPP has had no effect on the strength of the liquidity channel.
Firstly, the coefficients for some countries are rather small. For instance, with regard to Austria when the PSPP is in effect, if the liquidity ratio increases by 1 percentage point, the non-financial firms loan growth increases by 0.2752 percentage points the next quarter; in excess of the implied increase when the PSPP is absent. It would therefore require a considerable amount of asset purchases to boost lending noticeably. The country with the largest coefficient is the Netherlands with 2.0947; Slovakia, with -1.0012, has the lowest. This outcome may illustrate the dynamics of the high household debt in the Netherlands, even during the PSPP.
Secondly, the effects are heterogeneous. This may be the outcome of differing country characteristics. On the one hand members such as Slovenia and Belgium show a positive and significant increase for both loan types (respectively: 0.1117 and 0.0189 for HH; 0.3412 and 0.0829 for NFC). On the other, there are members like Luxembourg and Portugal, where both loan types have worsened (respectively: -0.13 and -0.4005 for HH; -0.1065 and -1.0012 for NFC). The results seem to be in line with the findings of Demertzis and Wolff (2016), where the authors find indeed different impacts on countries after an APP. According to the authors, one reason could be due to the small loan-to-deposit spread. The APP might have a negative impact on bank profitability, which could eventually affect the lending behaviour. However, the authors cannot find a strong indication that these developments occurred.
Thirdly, Austria has benefited from the purchase programme for loans to non-financial firms, but has seen a worsening loan situation for the other (0.2752 and -0.2286 respectively). Consequently,
14It is also important to note that the purchase of Greek sovereign bonds is not (yet) part of the PSPP. Therefore, the effects on the loan growth with regard to Greece should be treated with caution. One could say that the effects also involve Greek financial institutions, since their financial constraints depend on European financial markets. Also, supranational agencies or institutions are partly owned by Greece.
loans to non-financial firms may be affected differently than loans to households by the PSPP implementation. For Italy and France the signs reverse. However, the coefficients are insignificant for households.
Finally, with regard to the regressions, it is not possible to examine the overall effect for the euro-area due to the omission of certain countries. Germany, Spain and Ireland, for instance, would be essential to examine, since these countries attribute to a crucial proportion of the economic soundness within the euro-area. These countries either have missing independent variable data or are omitted because of collinearity. The predominantly positive effects suggest, however, that the PSPP was rather beneficial.
Table 3
Interaction:
Liquidity Ratio*DPSPP Household (e/se) Non-Financial Corporations (e/se)
Austria -0.2286 *** (0.0065) 0.2752 *** (0.0089) Belgium 0.0189 *** (0.0037) 0.0829 ** (0.0277) Cyprus -0.0897 *** (0.0221) -0.6099 *** (0.0411) France -0.0088 (0.00812) 0.0920 *** (0.0152) Greece 0.8328 *** (0.0331) 0.3307 *** (0.0800) Italy -0.5305 (0.02670) 0.0599* (0.0324) Luxembourg -0.1300 *** (0.0038) -0.1065 *** (0.0023) Latvia 0.0667 *** (0.0086) 0.2966 *** (0.0243) Malta 0.1408 *** (0.0074) 0.0565 *** (0.0133) Netherlands 0.1054 *** (0.0107) 2.0947 *** (0.0121) Portugal -0.0317 ** (0.0169) -0.1425 *** (0.0410) Slovenia 0.1117 *** (0.0063) 0.3412 *** (0.0111) Slovakia -0.4005 *** (0.0155) -1.0012 *** (0.02621) ***, **, * denote significance at 1, 5, and 10 percent significance levels respectively.
Countries that are not listed in table 3 are omitted because of collinearity or because the liquidity ratio data for the respective country is insufficient.
Table 4
Variables Household (e/se) Non-Financial Corporations (e/se)
Constant 0.5424 (0.7160) 2.9753 (1.7351) Liquidity Ratio 0.0123 (0.0199) -0.0728 (0.0471)
Government Bond Yields -0.0742 ***
(0.0180)
-0.1230 *** (0.0442)
Bad Loan Ratio -0.0324 *
(0.0160) -0.0749 ** (0.0270) Bank Size 0.0543 ** (0.0236) 0.1805 ** (0.0617)
Lagged Loan Growth 0.3572 ***
(0.0652) 0.2792 ** (0.0943) Capital Outflow 0.0292 ** (0.0125) -0.0746 *** (0.0127) R2 Overall 0.5844 0.4217 Rho 0.4112 0.3233
VII. Conclusion
The ECB has launched a series of unconventional methods in order to mitigate the effects of the financial and European sovereign debt crisis. The asset purchase programmes is one of them and perhaps the most controversial and most discussed measure. Section 2 shows that various researchers focused on the impact of the QE programmes with ambiguous results. Furthermore, in section 3 the obstacles of the euro-area are summarized, where the focus is mainly on monetary and macro-prudential policy.
This thesis examines the relationship between bank lending and monetary policy of the ECB. Particularly, the thesis analyses the impact of the PSPP on bank lending in the euro-area through the liquidity channel effect. By undertaking a fixed effects estimation, the regression analysis controls for time-invariant country characteristics that could bias the outcome. Dummies are used in order to investigate the impact of the PSPP through the liquidity channel.
The key results are as follows: Firstly, the impact of the PSPP for loans to households is mainly significant, yet heterogeneous across countries. While some countries have benefited from the purchase programme, others’ loan situation has worsened. Clear evidence that would indicate a positive and homogeneous result for the whole euro-area cannot be found. As for most members, the PSPP appears to be effective in loosening the financial constraints and increasing the overall lending, albeit the coefficients are small. To have a stronger effect through the lending channel, it would require a considerable amount of asset purchases. However, policymakers will have to consider carefully whether this is feasible, as the purchases come with consequences as well, e.g. the deterioration of loan growth for some countries. There seems to be no pattern regarding the effect across countries. Even within a country households and non-financial corporations need not be affected in the same way, sometimes displaying considerable differences in the magnitude of the effect.
At times of financial unrest and a zero interest rate environment there was not much scope for the ECB’s conventional policies and therefore this unconventional attempt was worth considering. The predominantly heterogeneous effect of the PSPP, however, could not support an overall
positive effect. The ambiguousness of the asset purchases illustrates the complexity that comes along. Member states appear to react differently to the PSPP. However, a more targeted measure is not easy to implement, since there might be unexpected externalities that have yet to be discovered. All these unexpected problems could raise the urge for a termination of the APP. As soon as the inflation target is reached and the national economies have recovered, it would therefore be necessary to think about an appropriate exit strategy for the PSPP in order to prevent an adverse impact on the economies. However, it is likely that the ECB will follow a tapering strategy, since a sudden stop might cause unforeseeable externalities or might even lead to a recession.15
The multitude of significant interaction terms between the PSPP dummy variables and the other control variables might indicate that not just the liquidity, but also other channels are affected by the asset purchasing program; an area that might warrant closer inspection in further research. Nevertheless, potential weaknesses could arise from the small sample size. For instance, the data for the liquidity ratio, which is provided by the IMF database, is incomplete, leading to omission of important countries. Additionally, it should be noted that the PSPP is usually accompanied by other policy measures such as CBPP3 and CSPP, but also fiscal and regulatory policy reforms. Some of these variables controlling for these issues, are omitted, because of multicollinearity issues. These problems might complicate the identification of the effect of the PSPP measure.
The results open new questions for further research. This analysis has shown that the PSPP had had heterogeneous effects across countries but makes no indication of what drives this heterogeneity. A better understanding of which factors determine the success or failure of asset purchasing programs in the individual countries could be crucial in designing better policy responses in the future.
15 Tapering is the gradual unwinding of asset purchases by a central bank in order to ensure a smooth transition for the
VIII. References
Acharya, V.V., Eisert, T., Eufinger, C. and Hirsch C. (2015), “Whatever it takes: The Real Effects of Unconventional Monetary Policy”, Paper presented at the 16th Jacques Polak Annual Research Conference Hosted by the International Monetary Fund, Washington, DC, November 5–6
Altavilla, C., Giannone D. and Lenza, M. (2014), “The Financial And Macroeconomic Effects Of OMT Announcements”, Working Paper Series No 1707, August 2014
Baumeister, C. and Benati, L (2010), “Unconventional Monetary Policy and the Great Recession. Estimating the Impact of a Compression in the Yield Spread at the Zero Lower Bound”, Working Paper Series, No 1258, October 2010
Bowman, D., Cai, F., Davies, S. and Kamin, S. (2011), “Quantitative Easing and Bank Lending: Evidence from Japan”, Board of Governors of the Federal Reserve System, International Finance Discussion Papers Number 1018, June 2011
Buch, C. M., Krause T. and Tonzer L. (2017), “Drivers of systemic risk: Do national and European perspectives differ?”, Deutsche Bundesbank, Discussion Paper No. 09/2017
Claeys, G., Leandro, Á. and Mandra A. (2015), “European Central Bank Quantitative Easing: The Detailed Manual”, Bruegel Policy Contribution, Issue 2015/02, March 2015
Dabrowski, M (2016), “Interaction between monetary policy and bank regulation: lessons for the ECB”, CASE Networks Studies & Analyses No. 480
Delivorias, A. (2015), “The ECB’s Expanded Asset Purchase Programme. Will Quantitative Easing revive the euro area economy?”,European Parliamentary Research Service, Briefing February 2015
Dell'Ariccia, G., Detragiache, E. and Rajan, R.G. (2005), “The Real Effect of Banking Crises”, Discussion Paper No. 5088 May 2005, Centre for Economic Policy Research
Demertzis, M. and Wolff, G. B. (2016), “What impact does the ECB’s quantitative easing policy have on bank profitability?”, Policy Contribution Issue No. 20, 2016
Dwenger, N., Fossen, F. M. and Simmler, M. (2015), “From financial to real economic crisis: Evidence from individual firm-bank relationships in Germany”, DIW Discussion Papers, No. 1510 Edison Y. (2016), “Did Quantitative Easing Work? Did QE lower yields and stimulate the economy? What about risks? Weighing the evidence requires a bit of theory”, Economic Insights, First Quarter 2016, Federal Reserve Bank of Philadelphia Research Department
European Central Bank (2017), “ECB Economic Bulletin”, Issue 1/2017
https://www.ecb.europa.eu/pub/economic-bulletin/html/index.en.html
European Central Bank (2016), “Macroprudential Bulletin”, Issue 1/2016
https://www.ecb.europa.eu/pub/pdf/other/ecbmpbu201603.en.pdf
Gropp, R., Krainer, J. and Laderman, E. (2014), “Did consumers want less debt? Consumer credit demand versus supply in the wake of the 2008-2009 financial crisis”, SAFE Working Paper Series, No. 42
Issing, O. (2011), “The crisis of European Monetary Union - Lessons to be drawn”, Journal of Policy Modeling 33, pp. 737-749
Joyce M. A. S. and Spaltro M. (2014), “Quantitative easing and bank lending: a panel data approach”, Working Paper No. 504, August 2014
Koijen R. S. J., Koulischer F., Nguyen B. and Yogo M. (2016) “Quantitative easing in the euro area: The dynamics of risk exposures and the impact on asset prices”, Banque de France, Working paper September 2016
Lane, P. R. (2012), “The European Sovereign Debt Crisis”, Journal of Economic Perspectives, Volume 26, Number 3, pp. 49–68
Margerit, A., Magnus, M. and Mesnard, B. (2017), “The EU macro-prudential policy framework”, Economic Governance Support Unit (EGOV), Briefing, March 2017
Martin, C. and Milas, C. (2012), “Quantitative Easing: a Sceptical Survey”, Working Paper, Department of Economics, University of Bath, Bath, U. K.
Mercier T. (2016), “Eurozone ECB: The PSPP parameters” BNP Parisbas Research Paper, September 2016
http://economic-research.bnpparibas.com/Views/DisplayPublication.aspx?type=document&IdPdf=29112
Palenzuela, D. R. and Dees S. (2016), “Savings and investment behaviour in the euro area”, Occasional Paper Series No. 167, January 2016
Peek, J. and Rosengren E. S. (2000), “Collateral Damage: Effects of the Japanese Bank Crisis on Real Activity in the United States”, The American Economic Review, Vol. 90, No. 1, March 2000, pp. 30-45
Schenkelberg, H. and Watzka, S. (2011), “Real Effects of Quantitative Easing at the Zero Lower Bound: Structural VAR-Based Evidence from Japan”, University of Munich, August 2011
Appendix
Source: Data from ECB Statistical Data Warehouse
Figure 3
Table 5
Modified Wald Test
H0: Sigma(i)2 = Sigma2 for all i Chi2 (13) = 897.90 Prob > chi2 = 0.0000
Testing for Heteroskedasticity
Source: Data from ECB Statistical Data Warehouse
Households Non-Financial Firms 0 1 2 3 4 5 6 7 2017-04 2016-12 2016-08 2016-04 2015-12 2015-08 2015-04 2014-12 2014-08 2014-04 2013-12 2013-08 2013-04 2012-12 2012-08 2012-04 2011-12 2011-08 2011-04 2010-12 2010-08 2010-04 2009-12 2009-08 2009-04 2008-12 2008-08 2008-04 2007-12 2007-08 2007-04 2006-12 2006-08 2006-04 2005-12 2005-08 2005-04 2004-12 2004-08 2004-04 2003-12 2003-08 2003-04
Cost of Borrowing in the Euro Area, %
Table 6
Hausman Test
Fixed Random Difference Sqrt (diag(V_b-V_B)) S.E.
Liquidity Ratio .0118565 .0126071 -.0007505 .0117932
Bond Yields -.0694835 -.0448164 -.0246671 .0128129
Bad Loan Ratio -.0285786 -.0049149 -.0236637 .0060075
Bank Size .050079 .0328997 .0171792 .
Lag Loan Growth .3765265 .6815203 -.3049938 .0306657
Capital Outflow .0306036 .0289296 .001674 .0104499
b = consistent under H0 and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under H0; obtained from xtreg Test: H0: difference in coefficients not systematic
Chi2 (6) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 123.83
Prob>chi2 = 0.0000
(V_b-V_B is not positive definite)
Testing for Fixed or Random Effects
Table 7
Variance Inflation Factors
Variable VIF 1/VIF
Liquidity Ratio 2.64 0.379113
Bad Loans Ratio 2.37 0.422816
Bond Yields 2.19 0.457330
Lagged Loan Growth 1.44 0.696040 Bank Size 1.08 0.925626
Capital Outflow 1.01 0.990146
Mean VIF 1.79
Table 8
Country Member of the
Euro-Zone since: Capital Key With Respect to PSPP Monthly Purchases Germany 1 January 1999 26,3% 16,6 France 1 January 1999 20,7% 13,1 Italy 1 January 1999 18,0% 11,3 Spain 1 January 1999 12,9% 8,1 Netherlands 1 January 1999 5,9% 3,7 Belgium 1 January 1999 3,6% 2,3 Austria 1 January 1999 2,9% 1,8 Portugal 1 January 1999 2,6% 1,6 Finland 1 January 1999 1,8% 1,2 Ireland 1 January 1999 1,7% 1,1 Slovakia 1 January 2009 1,1% 0,7 Lithuania 1 January 2015 0,6% 0,4 Slovenia 1 January 2007 0,5% 0,3 Latvia 1 January 2014 0,4% 0,3 Estonia 1 January 2011 0,3% 0,2 Luxembourg 1 January 1999 0,3% 0,2 Cyprus 1 January 2008 0,2% 0,1 Malta 1 January 2008 0,1% 0,1 Greece 1 January 2001
