The impact of the ECBs non-standard monetary policy measures on euro area stock and bond markets. Master Thesis MSc Economics

(1)

University of Groningen

Faculty of Economics & Business

Master Thesis

MSc Economics

The impact of the ECBs non-standard

monetary policy measures on euro area

stock and bond markets.

Supervisor: Mr. S. Pool

19

th

January, 2018

Jeroen Jansen

S2359421

(2)

1

The impact of the ECBs non-standard

monetary policy measures on euro area

stock and bond markets.

Jeroen Jansen

Master thesis MSc Economics

19

th

January, 2018

Abstract:

This paper investigates the effect of all non-standard monetary policy actions by the European Central Bank on equity prices and government bond yields. The non-standard policies investigated are the Securities Market Program, Outright Monetary Transactions, Asset Purchase Program and the (Targeted) Long Term Refinancing Operations. We use a panel data approach for six euro area countries. The measures for the policy actions consist of dummy variables for the announcement effects and ECB balance sheet data about outstanding loans and bond purchases. We find large positive effects of the policies on equity prices and negative effects on bond yields. The effects are larger for the periphery than the core countries.

JEL Codes: E52, E58, G15

Keywords: Monetary policy, equity prices, government bond yields, European Central

(3)

2

I. Introduction

From the end of 2007, central banks worldwide altered their policies as a response to the global financial crisis, started in the United States. The standard monetary policy tools were applied, however a wide range of non-standard policy measures were exercised as well to support the monetary transmission and fight low inflation. The cause of the crisis is debatable. The problems in the US subprime mortgage market triggered the crisis, however there are several other underlying contributions to the largest economic crisis since the Great Depression in the 1930s.

The substantial deregulation since 1980 and rapid financial innovation created enormous periods of growth in the financial sector, followed by several crises. The responses of the governments to these downturns were bailouts of distressed companies. Due to these bailouts, crises never completely pushed through and new expansions and financial booms were created (Crotty, 2009). The performance based bonus schemes across the executives of investment banks and other top traders in the financial system rose rapidly in the booming years pre-crisis. Even in the beginning of the crisis in 2009, bonusses paid were still excessively high despite the crisis has started (DiNapoli, 2009). The extremely high bonusses, even in times of crisis cause perverse incentives. During booming periods top traders and executives will still increase risk even if they know their decisions may cause a crises in the short run. This makes excessive risk taking to become rational risk taking which made the crisis as terrible as it was (Crotty, 2009). Credit rating agencies were also subject to agency problems. The loans and financial instruments were required to have high ratings in order to facilitate the high risk taking and to increase leverage.

(4)

3

together over $200 billion (Economist, 2009) These huge losses are the driving force of the crisis. Combined with the complex instruments, banks were highly levered. This high leverage was supported by the low short term interest rate set by the Fed. From 2002 to 2004 the short term funds rate was extremely low and supported the adoption of more debt. The falling housing prices caused a substantial amount of mortgage defaults triggering a deleveraging process by financial institutions. Off balance sheet packed assets needed to move back on the balance sheet and increasing losses on several assets required an increase in bank capital. To gain more capital, more assets were sold pressurizing prices even more in the falling market. The lower prices created margin calls and forced sell offs by investors. These effects together enforced an immense negative spiral (Crotty, 2009). The lack of trust and increasing costs of funding resulting from the price collapse slowed down the economy drastically and the worst financial crisis since the 1930s was born. This started in the United States, but considering the highly integrated financial market, the crisis spread out rapidly to Europe and the rest of the world.

The general implications of crises are widespread and the effects are visible all over the economy. The most important and well know effect is a substantial output loss. The median output loss of the crisis started with the banking crisis in 2007 is about 25 percent of GDP in four years for advanced countries affected by the crisis (Leaven and Valencia 2012). Other macroeconomic variables are affected as well. The recession after a crisis features sharply decreased consumption, investments, industrial production, employment and net exports; the whole economy slows down. Not only economic factors are largely affected, financial variables are hit by crises as well. The stock markets face huge losses and reasonable corrections. Credit supply falls and house prices drop dramatically (Claessens and Kose, 2013, working paper). As result of the decreased output, credit supply and slowing down of the economy, the inflation rate falls as well.

(5)

4

percent provides an adequate margin to avoid deflation. With interest rate cuts as monetary policy, in deflationary environments aggregate demand can be stimulated. Two percent inflation is low enough to obtain the benefits of price stability and it decreases the inflation differences within the euro area (ECB Europe). The ECB designed and implemented several non-standard policy measures to facilitate and accelerate the monetary transmission and improve the economic recovery after the crisis. The main non-standard policy measures are the Securities Market Program, Outright Monetary Transactions, (Targeted) Long Term Refinancing Operations and the Asset Purchase Program. In this research, we investigate the effect of those main non-standard policy measures on equity and bond markets, not taking into account the standard and remaining non-standard policies. The details about these programs are discussed in the next session.

The crisis started in 2007 had such a large scope and was followed by different stages evoking many different policy actions that now, 10 years later the recovery perseveres and policy actions can be repealed. The macroeconomic indicators adopt desirable values; the inflation rises towards 2 percent and the economic growth is picking up. However, stock markets show all time high prices and bond yields are historically low. A broad range of questions arise about the current economic situation and the large central bank intervention over the past decade. The non-standard policies have different impact on the stock and bond markets and these effects are widely studied, however there is a gap in the literature in the completeness of the different studies. Many studies focus on one monetary policy measure or one part of the global financial crisis and are written around 2013 and thus exclude the Asset Purchase Program started in March 2015 and Targeted Long Term Refinancing Operations. This research contributes by assessing the effects of all main non-standard policy measures on stock and bond markets. We use the approach similar as in Fratzscher, Lo Duca and Straub (2014). They use a panel approach to investigate the impact of the non-standard policies on stock and bond markets. The policies are measured with dummies, net outstanding loans and unexpected bond purchases, depending on the policy type.

(6)

5

purchase program, the increased demand pressurizes the yields. The low interest rates and supportive measures to increase bank lending lower the cost of capital for firms which stimulates equity returns. Italy and Spain are likely to be more sensitive to the policy actions since these countries are more risky and face sovereign tensions and the SMP and OMT focussed on these countries.

The remaining of this paper is structured as follows. In the next session we will briefly discuss the monetary policy actions of the ECB during and after the crisis. In Section III we link the policy actions to existing literature. We emphasize on the main non-standard policy measures and in Section IV we discuss the data and methodology used for investigating the effect of these non-standard policy measures on stock and bond markets. Consequently the results are presented in Section V, the robustness is discussed in Section VI, followed by the conclusions in Section VII.

II. ECB policies

In Europe, the crisis consist of three stages. First the liquidity crisis from summer 2007 till spring 2008. During this part of the crisis the interbank lending system collapsed. This was followed by the banking crisis form autumn 2007 till spring 2010 where the interbank market was dysfunctional and liquidity flew from peripheral countries to core countries due to distrust. Thereafter, the sovereign debt crisis from spring 2010 till July 2013 where the yields of highly levered European countries rose rapidly and created distrust and liquidity problems of several countries. After the sovereign debt crisis, the Euro area faced considerably low inflation and a slowed down economy. The European Central Bank responded to these stages of the crisis with several measures.

(7)

6

the broad based weakness of the economy and subdued monetary dynamics.” The Council repeated this in January, February and March 2014 as clear form of forward guidance. Figure 1: ECB policy interest rate developments

This figure presents the developments of the interest rate for the main refinancing operations and deposit facility in percentages.

A number of non-standard measures were employed to accelerate and improve the monetary policy transmission mechanism. We review the main monetary policy measures implemented by the European Central Bank as response to the global financial crisis and euro area sovereign debt crisis. We use the information given in the monthly bulletins of January 2010, October 2012 and January 2014 and the press releases about the monetary policy decisions from the ECB combined with Drudi et al (2012) and Thiman and Winkle (2013).

A. Main non-standard policy measures

The first main non-standard policy is the (i) Securities Market Program (SMP) as response to the euro area sovereign debt crisis. This program contains a bond purchase program of especially sovereign bonds in secondary markets to counteract disruptions on the monetary policy transmission. The program covers a total value of €210 billion and the bonds purchased will be held to maturity, however they will fully be sterilized. The SMP was announced on 10/5/2010 and exercised from May 2010 to September 2012 and reactivated on 8/8/2011 and focused on the sovereign unstable countries.

(8)

7

without prespecified limit. The prerequisites are that the relevant state complies with conditions specified in the EFSF/ESM program.

From March 2008, the ECB extends the maturities of the refinancing loans to (iii) Long Term Refinancing Operations (LTRO). In these refinancing operations maturities of liquidity provision of Main Refinancing Operations (MRO) are extended for providing longer term (6 months to 3 year maturity) refinancing to the financial sector, particularly the banking sector. Two large long term refinancing operations with maturity of 3 years were issued in December 2011. To further ease the private sector credit conditions and stimulate bank lending to the real economy, the ECB implemented From September 2014 to October 2017 (iv) Targeted Long Term Refinancing Operations (TLTRO). The TLTRO’s provide financing to credit institutions for periods up to 4 years. Figure 2 shows the total amount of outstanding loans under the LTRO and TLTRO program.

Figure 2: Balance of outstanding LTRO and TLTRO loans

This figure presents the amount of outstanding loans in billion euros on the ECB balance sheet.

(9)

8

government bonds from March 2015. From June 2016 under the Corporate sector purchase program (CSPP), corporate sector bonds were purchased to strengthen the pass-through of asset purchases to financing conditions as last part of the APP program. Figure 3 presents the APP holdings of every sub-program by the ECB. The Asset Purchase Program is a response to the ultralow inflation close to zero and the interest rates already at the zero lower bound.

Figure 3: Asset Purchase Program holdings

This figure presents the total ECB holdings of the different sub-programs of the Asset Purchase Program in thousands of billions euro.

B. Additional policy measures

(10)

9

US dollar funding. Combined with the LTROs, in addition the required reserve ratio decreased to 1% and the collateral availability increased by allowing national banks to accept additional credit claims on their own responsibility.

From June 2009 to June 2010 Covered Bond Purchase Program (CBPP) was applied and consists of the purchase of euro denominated covered bonds for a total value of €60 billion. This program was launched to revive the covered bond market since this market is the primary source of funding for banks. The Second Covered Bond Purchase Program (CBPP2) took place from November 2011 to October 2012 successive to the first covered bond purchase program. The APP is the large extension of the CBPP and CBPP2.

III Market impact ECB policies

A. Impact interest rate changes

The effects of the standard monetary policy mechanism of changes in the policy interest rate have been widely researched. Bohl et al. (2008) report that the short run impact of an unexpected change in the policy interest rates is negative. A hike of 25 basis points in the interest rate causes a fall in European stock markets by 1.42% to 2.30%. However, only 10% of all interest rate changes were unanticipated from 1999 to February 2007. This indicates the ECB is successful in communicating policy and managing expectations about it. Kholodilin et al (2009) complements the study of Bohl with a larger sample till January 2008 for ten sectoral stock market indexes. The findings are in line with the results in Bohl; a 25 basis point increase results in a decrease between 0.3% and 2.0% in stock market prices on the announcement day. A 25 basis point decrease in the interest rate aggregately results in a 1.0% increase in the stock market prices. Bernake and Kuttner (2005) analyse the impact of changes in monetary policy on equity prices. Their research find evidence a strong and positive reaction of unanticipated monetary policy and stock market returns. A 25 basis point unexpected decrease in the federal funds rate is associated with a 1 percent increase in broad stock indices. More permanent changes are likely to have larger market reactions.

(11)

10

forward guidance has improved the money market conditions, the control of money market rates and the expectations about the policy rates.

B. Impact main non-standard policies

The effect of the different non-standard policy measures on stock- and bond markets has been widely researched for the different measures. For instance Krishnamurthy et al (2017) find that the short term impact of the LTROs on the countries with the largest sovereign debt problems (Greece, Italy, Spain, Ireland, Portugal) on the corresponding stock markets are not significant. However the short effect of the Securities Market Program and Outright Monetary Transactions on the stock markets of these countries are clearly positive. The same effects hold for the core counties (Germany, France, The Netherlands, Belgium, Austria and Finland), however the effects are moderate.

Szczerbowicz (2015) find evidence that the sovereign bond purchases under the Securities Market Program and Outright Monetary Transactions significantly reduced the spreads of the sovereign bond yield. These measures were not only effective concerning the policy objective, but the positive spill over effect in reducing the bank covered bond spreads reinforces the effectiveness of the bond purchase programs. Furthermore, the three year LTROs and the low deposit rate of 0 percent were effective policy instruments to reduce the money market tensions. However, the shorter-term measures adopted by the ECB to provide credit were less effective in reducing those tensions.

Fratzscher, Lo Duca and Straub (2014) find that the OMT and SMP positively affected equity prices of the core and peripherical countries in Europe and had worldwide positive spill overs. Furthermore, those policies decreased the bond spreads in the peripherical countries and depreciated the euro slightly. The SMP program had an effect through 3 channels: Price channel, liquidity channel and balance sheet channel to ease the government bond markets. The research of Fratzscher includes several policy measures, however after 2014 new policies were introduced by the ECB which they do not cover. The APP and TLTROs are the missing policies in their research. In this paper we include the APP and TLTROs and investigate the effect of these programs as part of the total non-standard measures on equity and bond markets.

(12)

11

domestic firms has a significant portfolio balance effect indicating persistent higher stock prices. The ECB did not adapt as aggressive policies as the bank of Japan, however literature find supportive evidence for the positive effect of purchase programs on financial markets. De Santis (2016) investigates the effect of the APP program on euro area bond yields using an expectation based approach by evaluating euro area news on Bloomberg. They find evidence for a 63 basis point decrease of the bond yields and that the expectations about the APP program were formed before announcement.. Vulnerable countries are more sensitive for this effect. Brunnermeier and Schnabel (2015) state that expansionary monetary policy and historically low interest rates drive asset prices and would give rise to asset price bubbles, indicating a strong positive relationship between the APP and stock prices. Also Demertzis and Wolff (2016) find positive effects of the Asset Purchase Program. Due to the APP, key macro variables as investments, employment and growth move in a positive direction. Together with the substantial lower exchange rate the APP is effective in achieving its goals. Markets believe in the ability of managing the low inflation rate by the ECB.

In the next section we discuss our methodology and data. We explain how we can investigate the effects of the above discussed non-standard policies with our empirical design. We update the model used in Fratzscher and expand it with the Asset Purchase Program to assess the total impact on stock prices and bond yields.

IV. Methodology

A. Data

(13)

12

The policy effects are likely to differ for both groups, for instance, the yield spread between the core and periphery countries increased substantial during the crisis (Figure 4) indicating that bond yields per group may react different to monetary policy. The SMP and OMT were aimed at the risky countries with sovereign tensions. According to De Santis (2012) The increased risk aversion for Germany Bunds and bonds of other safe core countries lowered the yields. This safe haven motive together with together the negative spill over effects from Greece to other periphery countries supports the splitting of both groups. The SMP and OMT mainly focussed on the periphery countries, targeting the sovereign credit risk so we expect the policies to have different effects for both groups and clustering is necessary.

Figure 4: 10 year government bond yield developments

This graph presents the yield in percentages of the 10 year government bond yield for all countries.

Also the stock market returns seem to be clustered, however less evident than the bond yields. The stock market index of the core countries seem to outperform those of the periphery countries (Figure 5). The German DAX outperforms all other indices considerably. According Reuters (2010) this is among others due to the low share of banks in the index, bank stocks performed worst during the crisis. Furthermore, the DAX consists of large global companies, having little euro zone exposure. The depreciation of the Euro and low interest rate boost performance of these global oriented companies.

(14)

13

market is the daily difference of the 10 year government bond yield. See appendix Table 1 for the summary statistics for all variables. The sample is perfectly balanced and we see a higher standard deviation for the periphery countries for both the stock prices and the bond yields. This implies the stock prices and bond yields are more volatile than those of the core countries.

Figure 5: Stock market index developments

This graph presents the developments of the main stock market index for all countries. Theindexes are rescaled with 1/5/2007 equals 100 for all indices.

We use daily data, this is required in order to assume no reversed causality. The daily stock and bond market changes do not shape the monetary policy decisions, but the changes over a longer period can do so. Monetary policy initially does not respond to stock and bond market changes, but the primary objective of monetary policy is to maintain price stability. After the price stability, the ECB supports policies aiming at full employment and balanced economic growth. The stock and bond markets do not influence the policy decisions directly, however, the stock and bond prices are influenced by these macroeconomic variables. This all holds in the long term, meaning that the daily changes in the market cannot have any influence. The impact of policy decisions and implementations in return do have daily impact. The decisions probably change the long term developments and the long term market developments affect the policy measures, but for the daily changes, we can assume there is no reversed causality. The broad picture forms the monetary policy actions and can impact the market in one day, but one day market developments cannot change the monetary policy decisions.

(15)

14

Monetary Transaction policy, Securities Market Program and Asset Purchase Program. See appendix table 1 for the details about the selection of announcement dates. These dummies are equal to 1 at the date an important policy announcement occurs to capture the announcement effect of the corresponding policy. Since the OMT and APP policies are prespecified in magnitude and timing, at time the actual purchase is done, the purchase is already fully priced in due to the efficient market hypothesis.

Second, the Long Term Refinancing Operations and Targeted Long Term Refinancing Operations are included as explanatory variable as average weekly change in the outstanding loans on the ECB balance sheet. The net change in (T)LTRO loans per week divided by 5 for every weekday of that week is the way the LTRO and TLTRO variables are defined. Spreading the net change around the exact auction (repayment) date captures the market dynamics. Some days prior to the auctions, banks may change their demand for bonds to provide sufficient collateral for the (T)LTRO auction, to which investors can anticipate. The money received from the (T)LTRO auction is reinvested by the bank a couple of days after the auction. These bond auctions have dynamic effects on stock and bond markets prior and after the auction, so it is reasonable to define the explanatory variable in a larger time span around the auction. However, the effects might be underestimated due to spreading the auction (repayment) over 5 days. The summary statistics are presented in appendix Table 1.

The SMP is not only covered in the impulse dummy for the announcement effect, but also the net unexpected SMP purchases are important in order to investigate the effect of this program. Since neither the duration or magnitude of the SMP purchases are prespecified, the markets form expectations about the purchases to price them in. If there are unexpected purchases, the market reacts since the unexpected purchases are not priced in the markets already. We use the data about the unexpected SMP purchases as calculated by Fratzscher, Lo Duca and Straub (2014) who estimate the expected purchased based on market expectations and conditions. Comparing these estimations with the actual purchases provides a measure for unexpected SMP purchases which may have an effect on the financial markets. The summary statistics for the unexpected SMP purchases are presented in appendix Table 1.

(16)

15

controls for movements in stock market prices and bond yields due to macroeconomic surprises.

(ii) Effects of key monetary policy announcements in The US are captured by dummy variables for the days an important announcement about the unconventional monetary policy of the Fed occurs. We expand the table of Fratzscher with the dates and details about these dummy variables presented in appendix Table 2.

(iii) To control for sovereign shocks induced by the countries with sovereign tensions, i.e. the periphery countries, we include a set of dummy variables. The dummy variable equals 1 if one of the three largest credit rating agencies, Moody’s, Fitch, or Standard & Poor’s has changed the credit rating of Spain or Italy and otherwise zero. We make distinction between the countries and between a up or down grade of the rating.

(vi) Three dummy variables control for special days with extraordinary equity returns and yield changes. The first is 14/5/2010 when equity market losses worldwide were excessive as a result of the rumours about the French president threatening to leave the euro area and the possible downgrading of the sovereign credit rating. Second on 10/8/2011 again a possible downgrade of France’s credit rating feared the worldwide equity markets. Furthermore the severe situation of European banks, especially French and Italian banks contributed to the worldwide plunges of stock markets. These special days are also included in Fratzscher, complementary to this we include one additional special day dummy. The last special day is 24/8/2015, this Monday, better known as ‘Black Monday’ financial markets got in a free fall when a large drop in Asian equities triggered falling future prices of European and American indices followed by panic, lack of liquidity in the markets and forced sell offs. See appendix table 1 for details about the control variables.

B. Empirical approach

We extend the model used in Fratzscher, Lo Duca and Straub (2014) by including the Asset Purchase Program, TLTRO’s and LTRO’s from 1/10/2012 to 30/11/2017. By including these measures, our model captures the full set of main non-standard monetary policies, also the policies after the end of the sample used in Fratszcher (1/10/2012). The model to evaluate the impact of the policy measures on the stock- and bond markets is as follows:

𝑦𝑖,𝑡= 𝛽𝑀𝑃𝑡+ 𝛾𝐹𝑡+ 𝜀𝑖,𝑡 Equation (1)

(17)

16

The dependent variable yi,t is the log return on the main stock market index or the first

difference of the 10 year government bond yield for country i at time t. The matrix MPt

reflects the explanatory variables involving the non-standard policy measures of the ECB at time t. ANOMT, ANSMP and ANQE are the dummy variables reflecting the announcements of respectively the Outright Monetary Transactions, Securities Market Program and Asset Purchase Program. SMP is the unexpected Securities Market Program purchases. LTRO and TLTRO are the net change in outstanding loans of the Long Term Refinancing Operations and respectively the Targeted Long Term Refinancing Operations.

Ft contains the following set of control variables at time t: (i) The first difference of the

Citibank Economic Surprises Index for the EU, (ii) the first difference of the Citibank Economic Surprises Index for the US, (iii) dummy variables for a change in the credit rating of countries experiencing sovereign tensions, distinguished between Italy and Spain and an up- or downgrade, (iv) the set of dummies for US expansionary monetary policy announcements as listed in appendix Table 2 and (v) dummy variables for the special days 14/5/2010, 10/8/2011 and 4/8/2015. Details and summary statistics for all the variables is presented in appendix Table 1.

We run a panel regression with country fixed effects and bootstrapped standard errors with 1000 repetitions clustered by country. As showed in Fratzscher, Lo Duca and Straub (2014), this method is appropriate as benchmark model because in practice the inclusion of other control variables or estimation methods does not change the sign nor significance of the estimates for most of the results. There might be some potential issues we need to address. The LTRO and TLTRO auctions and repayments might be asymmetric, i.e. the auctions and repayments might have different impacts on the stock market and the bond yields. In our benchmark model the auctions and repayments are bundled to the net change in outstanding loans. We will discuss this in more detail in the robustness section. In that section, we will also discuss the impulse dummies for credit rating changes reflecting sovereign tensions more extensive since sovereign tensions might be longer lasting.

V. Results

(18)

17

A. Estimation results

We start with the model without control variables and stepwise add control variables until we end up in the benchmark model. Table 1 summarizes the estimation results about the impact of the ECB policies on stock markets and bond markets for the benchmark model for the different groups clustered by country. For the detailed results of the different models for all countries see appendix Table 3, for the periphery countries Italy and Spain see appendix Table 4 and for the core countries Austria, Finland, Germany and The Netherlands see appendix Table 5.

Table 1: Benchmark model results

Equity prices 10 year government bond yield

All countries Periphery Core All countries Periphery Core

ANSMP 1.451* 4.281*** 0.036 -0.200* -0.614*** 0.007 ANOMT 3.294*** 4.952*** 2.465*** -0.094 -0.370*** 0.044** ANQE 0.403* 1.100*** 0.114 -0.030*** -0.056*** -0.019*** SMP 0.188*** 0.114*** 0.225*** -0.005* -0.013*** -0.001 LTRO 0.338*** 0.470*** 0.272*** -0.003 -0.011 0.001 TLTRO 0.402*** 0.610*** 0.297** -0.036*** -0.048*** -0.030*** Observations 16,578 5,526 11,052 16,578 5,526 11,052 R-squared 0.037 0.048 0.034 0.037 0.103 0.035

This table reports the estimated coefficients for the benchmark model. *, ** and *** corresponds with a 10, 5 and 1 percent significance level.

(19)

18

thus decrease the yields. However, the assumed symmetric auctions and repayments might be inappropriate resulting in incorrect effects.

The policy effects on the core countries are much less evident. It is in line with our expectations that the countries without sovereign tensions are less sensitive to ECB policies, since some policies are mainly designed for the countries with sovereign tensions. Although all coefficients for the effect on equity prices are positive, they are smaller of magnitude than the coefficients for the Periphery and not all the coefficients are significant. The Securities Market Program announcements as well as the Asset Purchase Program announcements do not have a significant effect on stock prices for Austria, Finland, Germany and The Netherlands. The direction of the government bond yields differ between the policies. The effect of the total package monetary policies on the core bond yields is ambiguous since the OMT announcement has a significant positive effect and the QE announcements and the TLTROs have a significant negative effect. The SMP and TLTRO seem not to have any effect for the core countries. The interest rate approaching the zero lower bound might explain this. Rates are not able to decrease far below the zero lower bound, so the policies cannot have a severe pressurizing effect once the yields are already historically low. Furthermore, the safe haven motive can explain some of the upward effects. During the crisis the government bonds of the core countries are safe. Due to the policies aimed at the periphery the risk of those countries decrease, resulting in less demand for government bonds of safe countries.

(20)

19

announcement on 10/5/2010 was in the same week as a large unexpected SMP purchase and the 14/5/2010 special day dummy. The second SMP announcement on 8/8/2011 was again in the same week as a large unexpected SMP purchase and the second special day dummy of 10/8/2011. Due to the distribution of the unexpected SMP purchases over the whole week, the purchases interact with the announcements and the special days which explains the changes after controlling for the special days. Controlling for the large impact of these special days created by an exogenous event is reasonable. The events do not directly influence the monetary policy measures. Not controlling for the events create a bias on the policy effects, in that case the event should be the result of the policy measures which they are clearly not as explained in the data section and appendix Table 1. In the broad picture evaluating all policies for all countries the results are reasonable, the effect on equity prices of both the announcement and purchases are positive and significant and the effect on bond yields are negative and significant. The expansionary monetary policy boosts equity prices and drives down government bond yields.

B. Interpretation and total effects

The broad picture is in line with our expectations, however the interpretation of the coefficients differ for the policies. The coefficient of the announcements reflects the percentage change of the stock market and the change in percentage points of the bond yields due to the policy announcement. Note, the SMP and OMT are announced twice and the APP three times. The coefficients in Table 1 are the changes per announcement meaning the total effect of the policy announcements is larger; for the SMP and OMT two times and for the APP three times.

(21)

20

effect of the TLTRO on the stock markets of the periphery countries was at the time the outstanding loans reached the maximum was 2.4 percent increase. Now at the end of the sample period where the total net outstanding loans equals 27 billion the total effect is almost zero.

As described above, the estimated coefficients have different implications for the total effects. The results in Table 1 do not directly present the total effects of the monetary policies over the total sample time but only the estimation results. Table 2 presents the total effects for the different groups up to 30/11/2017. For the stock market returns the total effects are in percentage change and the total effects for the bond yields are in percentage points change. The total effects are obtained transforming the estimation coefficients using the number of announcements, net allotted amounts of loans and total unexpected purchases. In this way the real effect of the different policies becomes visible.

Table 2: Total effects ECB monetary policy on equity prices and bond yields over full sample period from 1/5/2007 to 30/11/2017.

Change in percentages Change in percentage points

ANSMP 2.9* 8.6*** 0.1 -0.40* -1.23*** 0.01 ANOMT 6.6*** 9.9*** 4.9*** -0.19 -0.74*** 0.09** ANQE 1.2* 3.3*** 0.3 -0.09*** -0.17*** -0.06*** SMP 9.1*** 5.5*** 10.9*** -0.24* -0.63*** -0.05 LTRO 2.1*** 2.9*** 1.7*** -0.02 -0.07 0.01 TLTRO 0.0*** 0.0*** 0.0** -0.00*** -0.00*** -0.00*** Observations 16,578 5,526 11,052 16,578 5,526 11,052 R-squared 0.037 0.048 0.034 0.037 0.103 0.035

The table reports the total effects of the policies for the benchmark model over the full sample period. * ** and *** corresponds with a 10, 5 and 1 percent significance level.

(22)

21

Figure 6: Actual and counterfactual stock market developments periphery

This figure presents the actual and counterfactual developments of the main stock market index for Italy and Spain. The indexes are rescaled with 1/5/2007 equals 100 for all indices.

Figure 7: Actual and counterfactual bond yield developments periphery

This figure presents the actual and counterfactual 10 year government bond yield development for Italy and Spain.

(23)

22

policy measures had an overall positive effect on the bond markets of the Euro area and the spread between the core and periphery decreased. This is in line with the results of Szczerbowicz (2015) and Fratzscher et al. (2014). The average increase of stock prices by 21.9 percent might be a good development as well, if it is a recovery of the large losses, however, several stock markets are a at record levels, especially the Germany DAX. If the increase due to the policies is less fundamentally grounded, the stock price rises may create an overpricing environment, however investigating this is beyond the scope of this paper. Figure 8: Actual and counterfactual stock market developments core

This figure presents the actual and counterfactual developments of the main stock market index for Austria, Finland, Germany and The Netherlands. The indexes are rescaled with 1/5/2007 equals 100 for all indices.

Figure 9: : Actual and counterfactual bond yield developments core

(24)

23

Remarkable is the small effect of the QE announcements on the stock markets but even more on the bond markets. The largest purchase program in history should drive bond yields down because of the massive scope of the total purchases. The increased demand for bonds and investors anticipating on those purchases should induce severe announcement effects. According to the efficient market hypothesis, the effect of the purchase program should be fully priced in by announcement, i.e. the actual purchases do not have any effect on the bond prices because the amounts and timing are prespecified. However, the markets may already have formed their expectations about the program meaning the expected program is already priced in and the actual announcement does not provoke large movements of the market if the announcements are in line with the expectations. This corresponds with the evidence presented in De Santis (2016). The sovereign bond yields have decreased dramatically since the adoption of the QE program, but apparently not during the announcements, supporting the theory of the pricing in effect of expectations. Furthermore, the forward guidance as mentioned before effectively formed expectations about the very low interest rates. The APP is the last policy and the interest rates already approached the zero lower bound. Around this bound, the interest rates have very little space to move further downwards.

VI. Robustness

A. Sovereign tensions

The choice of controlling for sovereign tensions in Italy and Spain by means of impulse dummies for a downgrade in the sovereign credit rating by Standard & Poor’s, Moody’s or Fitch might not be the optimal way of controlling for sovereign tensions. Sovereign tensions are likely to evolve gradually over time, so the effects of the tensions might not be fully captured by the impulse dummies only. Now we control for the announcement effects of one of the largest credit rating agencies which is possibly a part of the total sovereign tensions leading to an omitted variable bias. We can also control for the actual rating assigned to Italy and Spain by the credit rating agencies to see what effect the level of creditworthiness has. However, our model is a first differences model so it is reasonable to look at the first differences of sovereign tensions measured as the change in rating.

B. Symmetricity LTRO and TLTRO

(25)

24

repayment. Within the sample period, the balance of outstanding loans fluctuates with at its maximum a total effect of 2.4 percent equity price increase for the periphery and 1.2 percent increase for the core countries. For the effect on the bond yields the same idea holds and the effect at 1/7/2016 was a 19 basis point decrease for the Periphery and 12 basis points for the core countries. The total effects for the LTRO net outstanding loans are not zero since the net outstanding loans at the end of the sample period is still 6,137 billion euro. However if all the loans mature and the balance approaches zero, the total effect also vanishes for the LTROs. The assumed symmetry imposes a temporary effect of the policies. It is debatable whether the effect indeed is temporary or that auctions have different effects than repayments. The auction provides the bank of cheap liquidity, improves the balance sheet and lending activities. The auction should have a positive effect. The repayment removes the liquidity and counteracts the balance sheet improvements and lending activities. However, timing may be crucial. At the moment of the auction, additional liquidity might be of higher value for the bank than 3 years later at repayment. Repayment can also be a signal of strength. If this holds, the auction has a larger effect than the repayment. To check this, we distinct between auctions and repayments and the full estimation results are presented in appendix Table 6.

The auctions and repayments are not symmetric. In our first results the effects of the net outstanding LTRO loans were moderate but significant for the equity prices and very small and insignificant for the government bond yield. We make a distinction between the auction and the repayment of the loans and it appears that for equity prices the effect of the auction is larger than the effect of the repayment and both are significant. For the bond yields the effects are even more asymmetric. The auction as well as the repayment both lower the bond yields significantly for all countries. This explains the former small and insignificant effect of the LTRO loans, the auction and the repayments cancel out. For the periphery and core countries the auctions are slightly insignificant, but the repayments are significant at a 1 percent level.

(26)

25

decrease of 1.65 percentage point for the periphery and 1.07 percentage point for the core countries.

Table 3: Total effects asymmetric LTRO and TLTRO

Change in percentages Change in basis points

LTRO auctions 10.4*** 14.5*** 8.4*** -0.37** -0.68 -0.25 LTRO repayments -7.2** -8.5*** 6.6 -0.87*** -0.97*** -0.82*** LTRO total 3.2 6.0 1.8 -1.24 -1.65 -1.07 TLTRO auctions 3.3 *** 6.0*** 2.0*** -0.31*** -0.48*** -0.22*** TLTRO repayments -1.3** -1.0*** 1.5* 0.10*** 0.06 0.11*** TLTRO total 2.0 5.0 0.5 -0.21 -0.42 -0.11

The table reports the total effects of the asymmetric LTRO and TLTRO for the benchmark model over the full sample period. * ** and *** corresponds with a 10, 5 and 1 percent significance level.

The total value for the auctions of the TLTRO policy is 5,875 billion euro, and nearly all these allotted amounts are already repaid, namely 5,848 billion euro. The auctions and repayments have contrary to the LTRO opposite effects: the auctions increase the equity prices and decrease the bond yields, the repayments cancel this partly out with a decrease of equity prices and increase of bond yield. Note, the auction effects are larger than the repayment effects resulting in a net increase in equity prices and net decrease in bond yields. The former symmetric total effects were zero for both the equity prices and the bond yields because of the net outstanding loans approached zero. With the asymmetric approach, the total effects of the TLTRO on equity prices for all counties is an increase of 2.0 percent, for the periphery and core is the percentage 6.0 and 1.8 respectively. This new evidence supports the method of asymmetric LTRO auctions and repayments. Table 4 compares the total effects of the symmetric and asymmetric approach.

Table 4: Comparison total effects symmetric and asymmetric LTRO and TLTRO

Change in percentages Change in basis points

LTRO symmetric 2.1 2.9 1.7 -0.02 -0.07 0.01

LTRO asymmetric 3.2 6.0 1.8 -1.24 -1.65 -1.07

TLTRO symmetric 0.0 0.0 0.0 0.00 0.00 0.00

TLTRO asymmetric 2.0 5.0 0.5 -0.21 -0.42 -0.11

(27)

26

C. Persistence effects and expectations

The assumption that all policy effects are fully persistent is probably incorrect. If the effects are symmetric, at time of repayment or sterilizing the effects will vanish. As showed with the LTRO and TLTRO, the assumed symmetry does not certainly hold. Assessing the total effects make the effect of the total package of policy measures visible, however markets price in all new information and new developments about policies. Also investors’ expectations about the policies change over time. The total effect of the policies may change over time after the policy announcement. The announcements have impact and provoke a shock, only when the announcement is unexpected. After the shock markets gradually move to the new equilibrium since new expectations are made. In our approach with announcement dummies, we only capture the announcement effects. The convergence after the announcements can be of greater scope than the announcements. For instance, the QE program had three different announcements, our results show week evidence of a small effect of the announcements of this program. This program is the largest asset purchase program in the history of the European Central Bank and we expect that this program has severe influences on the equity and bond markets. If the expectations about the program were already formed, the effects were already priced in and if the announcements are conform the expectations, the markets do not heavily react on the announcements. In this case, the announcement effect is not persistent and does not reflect the total effect of the policy. The unexpected announcements have a large impact effect, the SMP for instance. However, we don’t know whether the announcement effect is still present after 1 day, 10 days, or even after some years. It would be a great extension of this research to investigate the persistency of all policies. Our data is constructed by using the day end market values. Looking closer at specific moments after the announcement effect, for instance after 1 hour, 2 hours, 6 hours 12 hours give more insight about the dynamics of the announcement effects.

D. Omitted variable bias

(28)

27

VII. Conclusions

We find large effects of the European Central Banks non-standard monetary policy measures on euro area equity and bond markets. Our approach evaluates the effect of the Securities Market Program, Outright Monetary Transactions, Long Term Refinancing Operations, Targeted Long Term Refinancing Operations and the Asset Purchase Program. We distinguish between announcement effects, actual outstanding loans and unexpected purchases. Overall we find the non-standard measures positively affect stock market prices. The effect of the full set over the total sample period is a 30.2 percent increase in equity prices for Italy and Spain, and a17.5 percent increase for Austria, Finland, Germany and The Netherlands. The effect on the 10 year government yield is a considerable decrease of the yield with 2.77 percentage points for Italy and Spain. For the core countries contrary, the total effects are negligible with a total decrease of 3 basis points of the 10 year government bond yield. This ambiguous effect is explained by opposite effects of the different policy measures involving the safe haven motive and the zero lower bound restriction the interest rates of the core countries face. Taking into account asymmetry between (T)LTRO auctions and repayments, the total effect is more evident since auction effects seem not to be cancelled out by repayments.

The effects are, in line with our expectations and the research of Fratzscher, Szczerbowics and De Santis more severe for the periphery because the SMP and OMT mainly aimed at the sovereign unstable countries. This indicates the ECB measures were successfully targeted. The bond yields of the periphery decreased considerably indicating the policy makers were able to restore trust in the financial market and support the recovery of the global financial crisis.

(29)

28

(30)

29

VIII. References

Barbon, A., and Gianinazzi, V., 2017, “Quantitative easing and equity prices: Evidence from the ETF program of the Bank of Japan”, Working paper

Bernake, B.S. and Kuttner, K.N. (2005) “What Explains the Stock Market’s Reaction to Federal Reserve Policy?” The Journal of Finance, Vol. 60 I3.

Bohl, T., Siklos, P.L., and Sondermann, D., 2008, “European stock markets and the ECB’s monetary policy surprises” International Finance, 7 May 2008

Brunnermeier, M.K. and Schnabel, I., 2015, “Bubbles and Central Banks: Historical Perspectives”, Cambridge University Press, Web

Chacko, G., Sjoman, A., Motohashi, H., and Dessain, V., 2006, “Credit derivatives”, Philadelphia Wharton School.

Claesens, S., Ayhan Kose, M., 2013, “Financial Crises: Explanations, Types, and Implications”, IMF Working papers 13/28, International Monetary Fund.

Crotty, J., 2009, “Structural causes of the global financial crisis: a critical assessment of the ‘new financial architecture’”, Cambridge Journal of Economics, 33, pp 563-580. Demertzis and Wolff, 2016, “The effectiveness of the European central bank’s asset purchase program” Bruegel Policy Contribution

DiNapoli, T., 2009, “Wall Street Bonuses Fell 44% in 2008”, Office of the New York State

Comptroller, January 28

Drudi, Durré and Mongelli, 2012, “The interplay of economic reforms and monetary policy: The case of the eurozone”, ECB working paper series, No 1467

ECB, Monthly bulletin January 2010 ECB, Monthly bulletin October 2012 ECB, Monthly bulletin January 2014 ECB, Monthly bulletin April 2014

(31)

30

Filardo, A.J., and Hofmann, B., 2014, “Forward guidance at the zero lower bound”, BIS

Quarterly Review, March 2014

Fratzscher, M., Lo Duca, M., Straub, R., 2014, “ECB unconventional monetary policy actions: market impact, international spillovers and transmission channels”, 15th_Jacques

Polak Annual Research Conference, November 13-14, 2014

Kholodilin, K., Montagnoli, A., Napolitano, O., Siliverstovs, B., 2009, “Assessing the impact of the ECB’s monetary policy on the stock markets: A sectoral view”, Economic

Letters, Elsevier, December 2009, pp. 211-213

Krishnamurthy, A., Nagel, S., and Vissing-Jorgensen, A., 2017, “ECB Policies Involving Government Bond Purchases: Impacts and Channels”, working paper

Leaven, L. Valencia, F. 2008. “Systemic Banking Crises: A New Database”, IMF Working Papers 08/224, International Monetary Fund.

Reuters, 2010, “ANALYSIS-German DAX to outperform in sovereign debt crisis” Santis, De, R. A., 2012, “The Euro area sovereign debt crisis, safe haven, credit rating agencies and the spread of the fever from Greece, Ireland and Portugal”, ECB working paper series no 1419

Santis, De, R. A., 2016, “Impact of the asset purchase program on euro area government bond yields using market news”, ECB working paper series no 1939

Szczerbowicz, U., 2015, “The ECB unconventional monetary policies: Have they lowered market borrowing costs for banks and governments?”, International Journal of Central

Banking, December 2015

Tett, G., 2009, “Lost through destructive creation”, Financial Times, 10 March

(32)

31

(33)

32

Table A2: Dummy variables for important US expansionary monetary policy announcements

Event Date Description Source:

EV1 25/11/2008 LSAP announced Fratzscher (2014)

EV2 01/12/2008 First suggestion Bernake of extending QE to treasuries Fratzscher (2014) EV3 06/12/2008 First suggestion of extending QE to treasuries by FOMC Fratzscher (2014) EV4 28/01/2009 Fed stands ready to expand QE and buy Treasuries Fratzscher (2014)

EV5 18/03/2009 QEs expanded Fratzscher (2014)

EV6 27/08/2010 Bernake suggest additional QE Fratzscher (2014)

EV7 12/10/2010 FOMC says additional QE may be appropriate Fratzscher (2014) EV8 15/10/2010 Bernake says Fed stands ready for action Fratzscher (2014)

EV9 03/11/2010 QE2 announced Fratzscher (2014)

EV10 21/09/2011 Maturity Extension Program announced Fratzscher (2014) EV11 20/06/2012 Maturity Extension Program extended Fratzscher (2014) EV12 22/08/2012 FOMC says additional monetary accommodation is likely Fratzscher (2014)

EV13 13/09/2012 QE3 announced Fratzscher (2014)

EV14 12/12/2012 QE3 expanded Fratzscher (2014)

EV15 19-06-2013 Hint of first tapering QE Author, FOMC

EV16 18-12-2013 Tapering from 85 to 75 billion per month Author, FOMC EV17 29-01-2014 Monthly purchases lowered to 65 billion Author, FOMC EV18 19-03-2014 Monthly purchases lowered to 55 billion Author, FOMC EV19 30-04-2014 Monthly purchases lowered to 45 billion Author, FOMC EV20 18-06-2014 Monthly purchases lowered to 35 billion Author, FOMC EV21 30-07-2014 Monthly purchases lowered to 25 billion Author, FOMC EV22 17-09-2014 Monthly purchases lowered to 15 billion Author, FOMC

EV23 29-10-2014 Announcement final QE purchase Author, FOMC

(34)

33 Table A3: Estimation results for all countries

Model 1 Model 2 Model 3 Model 4 Model 5

No controls Surprises US QE Rating Dummies

Dependent variable: Equity prices

(35)

34 RATEITD - - - 0.423*** 0.431*** - - - (0.039) (0.038) RATEITU - - - -0.323*** -0.320*** - - - (0.077) (0.077) RATESPD - - - -0.492*** -0.497*** - - - (0.057) (0.052) RATESPU - - - -0.057** -0.062*** - - - (0.022) (0.023) DU24082015 - - - - -5.238*** - - - - (0.188) DU10082011 - - - - -5.203*** - - - - (0.591) DU14052010 - - - - -4.844*** - - - - (0.521) Constant -0.010 -0.011* -0.010 -0.008 -0.006 (0.006) (0.006) (0.006) (0.006) (0.006)

Fixed effects yes yes yes yes yes

Surprises no yes yes yes yes

US QE no no yes yes yes

Credit rating no no no yes yes

Special days no no no no yes

Observations 16,578 16,578 16,578 16,578 16,578

Groups 6 6 6 6 6

R-squared 0.006 0.013 0.025 0.026 0.037

Dependent variable: 10 year government bond yield

(36)

35 - - (0.021) (0.021) (0.020) EV14 - - 0.054*** 0.054*** 0.054*** - - (0.008) (0.008) (0.008) EV15 - - 0.186*** 0.186*** 0.186*** - - (0.032) (0.033) (0.032) EV16 - - 0.021*** 0.021*** 0.021*** - - (0.008) (0.008) (0.007) EV17 - - -0.007 -0.007 -0.007 - - (0.016) (0.016) (0.015) EV18 - - 0.049*** 0.049*** 0.049*** - - (0.002) (0.002) (0.003) EV19 - - -0.039*** -0.039*** -0.039*** - - (0.006) (0.006) (0.005) EV20 - - -0.051*** -0.052*** -0.052*** - - (0.004) (0.004) (0.004) EV21 - - 0.020*** 0.020** 0.020*** - - (0.007) (0.008) (0.007) EV22 - - 0.029*** 0.029*** 0.029*** - - (0.007) (0.007) (0.007) EV23 - - -0.037*** -0.037*** -0.037*** - - (0.012) (0.012) (0.011) RATEITD - - - -0.014*** -0.014*** - - - (0.002) (0.002) RATEIDU - - - -0.018*** -0.018*** - - - (0.003) (0.002) RATESPD - - - -0.009* -0.009* - - - (0.005) (0.005) RATESPU - - - -0.004 -0.004 - - - (0.004) (0.004) DU24082015 - - - - 0.013* - - - - (0.007) DU10082011 - - - - -0.114*** - - - - (0.033) DU14052010 - - - - -0.023 - - - - (0.029) Constant 0.001*** -0.001*** -0.001*** -0.001** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000)

Observations 16,578 16,578 16,578 16,578 16,578

Groups 6 6 6 6 6

R-squared 0.015 0.024 0.032 0.033 0.034

(37)

36 Table A4: Estimation results for Italy and Spain

ANSMP 5.551*** 5.447*** 5.478*** 5.471*** 4.281*** (0.540) (0.564) (0.565) (0.565) (0.588) ANOMT 5.120*** 4.958*** 4.955*** 4.953*** 4.952*** (0.165) (0.178) (0.177) (0.178) (0.177) ANQE 1.122*** 1.100*** 1.100*** 1.100*** 1.100*** (0.071) (0.081) (0.082) (0.082) (0.081) SMP -0.190*** -0.185*** -0.194*** -0.192*** 0.114*** (0.021) (0.022) (0.023) (0.023) (0.029) LTRO 0.343*** 0.382*** 0.410*** 0.383*** 0.470*** (0.026) (0.023) (0.024) (0.026) (0.024) TLTRO 0.0619*** 0.620*** 0.618*** 0.623*** 0.610*** (0.110) (0.115) (0.115) (0.116) (0.116) Constant -0.025*** -0.025*** -0.025*** -0.023*** -0.021*** (0.004) (0.004) (0.004) (0.004) (0.004)

Observations 5,526 5,526 5,526 5,526 5,526

Groups 2 2 2 2 2

R-squared 0.015 0.021 0.033 0.034 0.048

ANSMP -0.617*** -0.619*** -0.619*** -0.619*** -0.614*** (0.086) (0.087) (0.084) (0.085) (0.084) AOMT -0.363*** -0.370*** -0.370*** -0.370*** -0.370*** (0.043) (0.044) (0.043) (0.044) (0.042) ANQE -0.054*** -0.055*** -0.055*** -0.056*** -0.056*** (0.006) (0.005) (0.005) (0.005) (0.005) SMP -0.012*** -0.012*** -0.012*** -0.012*** -0.013*** (0.000) (0.000) (0.000) (0.000) (0.000) LTRO -0.010 -0.009 -0.009 -0.010 -0.011 (0.010) (0.010) (0.010) (0.010) (0.009) TLTRO 0.051*** -0.049*** -0.049*** -0.048*** -0.048*** (0.005) (0.005) (0.005) (0.005) (0.005) Constant 0.000 0.000 -0.000*** 0.000* 0.000* (0.000) (0.000) (0.000) (0.000) (0.000)

Observations 5,526 5,526 5,526 5,526 5,526

Groups 2 2 2 2 2

R-squared 0.083 0.088 0.102 0.103 0.103

(38)

37

Table A5: Estimation results for Austria, Finland, Germany, The Netherlands

ANSMP 0.867*** 0.769*** 0.793*** 0.786*** 0.036 (0.222) (0.222) (0.208) (0.210) (0.248) ANOMT 2.633*** 2.471*** 2.468*** 2.466*** 2.465*** (0.238) (0.241) (0.240) (0.237) (0.237) ANQE 0.132 0.119 0.117 0.115 0.114 (0.160) (0.160) (0.157) (0.162) (0.157) SMP 0.033 0.037 0.030 0.032 0.225*** (0.036) (0.035) (0.036) (0.037) (0.032) LTRO 0.179* 0.215** 0.245** 0.219** 0.272*** (0.100) (0.100) (0.097) (0.103) (0.106) TLTRO 0.293** 0.299** 0.299** 0.305** 0.297** (0.124) (0.123) (0.121) (0.121) (0.120) Constant -0.003 -0.003 -0.002 -0.000 0.002 (0.007) (0.007) (0.007) (0.006) (0.006)

Observations 11,052 11,052 11,052 11,052 11,052

Groups 4 4 4 4 4

R-squared 0.003 0.010 0.024 0.025 0.034

ANSMP 0.031* 0.029* 0.029* 0.030* 0.007 (0.017) (0.017) (0.017) (0.017) (0.020) ANOMT 0.052** 0.044** 0.044** 0.044** 0.044** (0.021) (0.021) (0.022) (0.021) (0.021) ANQE -0.017*** -0.019*** -0.019*** -0.019*** -0.019*** (0.001) (0.001) (0.001) (0.001) (0.002) SMP -0.006** -0.006*** -0.007*** -0.007*** -0.001 (0.002) (0.002) (0.002) (0.002) (0.002) LTRO -0.001 0.000 -0.000 -0.001 0.001 (0.004) (0.004) (0.004) (0.005) (0.004) TLTRO -0.032*** -0.029*** -0.029*** -0.029*** -0.030*** (0.003) (0.003) (0.003) (0.003) (0.004) Constant 0.001*** -0.001*** -0.001*** -0.001*** -0.001*** (0.000) (0.000) (0.000) (0.000) (0.000)

Observations 11,052 11,052 11,052 11,052 11,052

Groups 4 4 4 4 4

R-squared 0.003 0.018 0.030 0.031 0.035

(39)

38

Table A6: Symmetric and asymmetric LTRO and TLTRO for the benchmark model

LTRO 0.338*** 0.470*** 0.271*** -0.003 -0.011 0.001 (0.078) (0.024) (0.106) (0.005) (0.009) (0.004) LTROAUC 0.336*** 0.467*** 0.271*** -0.012** -0.022 -0.008 (0.069) (0.017) (0.089) (0.007) (0.014) (0.006) LTROREP 0.292** 0.343*** 0.266 0.035*** 0.039*** 0.033*** (0.116) (0.031) (0.177) (0.004) (0.002) (0.005) TLTRO 0.402*** 0.610*** 0.297** -0.036*** -0.048*** -0.030*** (0.102) (0.116) (0.120) (0.004) (0.005) (0.004) TLTROAUC 0.565*** 1.026*** 0.334*** -0.052*** -0.082*** -0.037*** (0.169) (0.167) (0.128) (0.012) (0.020) (0.005) TLTROREP 0.230** 0.170*** 0.259* -0.016*** -0.010 -0.019*** (0.103) (0.065) (0.152) (0.004) (0.010) (0.002) Observations 16,578 5,526 11,052 16,578 5,526 11,052 Groups 6 2 4 6 2 4