Central banks and climate change The impact of the ECB's willingness to tackle climate change on green bonds

(1)

Central banks and climate change

The impact of the ECB's willingness to tackle climate change on green bonds

University of Groningen

Faculty of Economics and Business Supervisor: Konstantin M. Wacker, Ph.D.

Master Thesis

in MSc Economic Development and Globalization and MSc International Economics (Double Degree)

(2)

I

Abstract

Green bonds play an essential role in financing the transition to a low-carbon economy in Europe. The European Central Bank will support this financing process legitimised by its mandate to maintain financial stability. The debate on the nature of the intervention is intense. Conducting an event study shows that the ECB's announcement on 29/05/2018 to take climate change somehow into account significantly benefits euro-denoted green bonds over conventional bonds with the same risk-maturity profile. However, this effect did not materialise when the new president publicly considered on 04/09/2019 to expand the range of instruments, e.g., by green quantitative easing.

Keywords:

(4)

III

List of Figures

Figure 1 Green bond supply by regions ... 6

Figure 2 Range of intervention instruments ... 8

Figure 3 Distribution of t-test statistics ... 25

(5)

IV

List of Tables

Table 1 The Green Bond sample ... 22

Table 2 Average daily Abnormal Green Bond Returns (unstandardized)... 22

Table 3 Standard deviations of Abnormal Green Bond Returns by Risk and Maturity ... 23

Table 4 Distribution of Raw and Abnormal Green Bond Returns ... 24

Table 5 Event Study Results ... 27

(6)

V

List of Abbreviations

AGBR Abnormal Green Bond Return

CSR Corporate Social Responsibility

ECB European Central Bank

ECON European Parliament's Committee on Economic and Monetary Affairs

EIB European Investment Bank

EP European Parliament

EU TEG European Union Technical Expert Group

ESG Environmental, Social and Governance

FSB Financial Stability Board

Green QE Green Quantitative Easing

IsDB Islamic Development Bank

KfW Kreditanstalt für Wiederaufbau

NGFS Network for Greening the Financial System

OTC Over-the-Counter

QE Quantitative Easing

TRACE Trade Reporting and Compliance Engine

UN United Nations

(7)

1

1. Introduction

Climate change is a concern not only for economic policymakers but also for financial regulators. While government intervention in response to the 'greatest market failure of all time' (according to Stern (2017)) is now widely accepted in the economic context, the debate on the role of central banks is relatively recent. Mark Carney (2015), Governor of the Bank of England, initiated this debate by linking climate change and the financial markets in his speech on the 'tragedy of the horizon'. Today, there is a degree of consensus that climate change should also be on the agenda of central banks of high-income economies since their mandate includes the preservation of financial stability, which is affected by climate change-related systemic risks. There are already 54 central banks and financial regulators in the Network for Greening the Financial System, the NGFS, which aims to help the financial system to make the transition to a sustainable economy a success.

First and foremost, the transition requires additional investment to ensure sustainable transport, energy, real estate, waste cycles, and others. Estimates of the annual global investment gap range from $650 to $900 billion until 2050 (Enkvist et al. (2010), IEA (2012)). This thesis argues that supporting green bonds, as a market-based debt instrument with precisely this financing intention, is essential for two reasons, especially in Europe. First, according to Campiglio (2016), the current macroeconomic situation and the related regulatory measures, prevent banks from lending for these purposes, as they continue to deleverage their balance sheets as a result of the 2007 financial crisis. The author thus claims that even in the presence of a carbon tax, the credit and capital markets do not react to the price signal to the socially desirable extent. Campiglio concludes that central banks are required to stimulate the credit creation mechanism to close the investment gap, for example, by lowering reserve requirements for green credit. However, this straight policy implication can only be applied by central banks, which mandate explicitly targets fostering a sustainable development or supporting economic policy measures made by governments. These kinds of mandates are often observed in central banks of developing countries. The People's Bank of China, for instance, can grant credit directly to sustainable projects.

Here the second argument for the urgency of an expanding green bond market in Europe comes into play. The European Central Bank (ECB), as a central bank for high-income economies, is unlikely to push lending because its mandate does not envisage active financing of projects intended by economic policy and because the mandate is difficult to amend. For this reason, the stakes are high with regard to green bonds, and the ECB's influence on them is of outstanding interest, although it is not yet being investigated. Hence, this thesis examines the overall research question: Does the presence of the ECB in the debate regarding financing the transition to a low-carbon economy positively affect euro-denoted green bonds?

(8)

2

thesis examines whether abnormal green bond returns, i.e. the difference between the daily green bond return and the normal return of conventional bonds, were statistically significantly positive. To take into account the heterogeneity of the bond market, 24 different risk and maturity specific matching portfolios were constructed to calculate the normal returns. Furthermore, following Ederington et al. (2015), returns were standardized by dividing them by their standard deviations, so that abnormal returns are less heteroscedastic and the parametric t-test for determining statistical significance becomes more reliable.

The impact of ECB statements on climate change was evaluated at two different points in time, which differ in terms of the level of the degree of intervention expressed. First, on 29/05/2018, the ECB's unexpected entry into the NGFS for the first time significantly signalled its willingness to reconsider its role in addressing climate change. Secondly, the new president Christine Lagarde announced on 04/09/2019 that she would give even more attention to climate change and further considered more active participation in the financing process towards the low-carbon economy, for example, by discussing green Quantitative Easing programmes (green QE). For the entry into the NGFS on 29/05/2018, significant positive abnormal green bond returns were observed. For the announcement on 04/09/2019 to consider extended measures, this only applies to parts of the euro-denoted green bond market. Furthermore, in the period from 23/03/2015 to 31/01/2020, events were identified that triggered a significantly greater positive valuation effect on green bonds relative to conventional bonds. Most events can be grouped into one of the following categories: The establishment of standards in the green finance segment, enormous new issues of green bonds, the pressure to disclose climate change-related risks and the signalling of international cooperation. Finally, the accumulation of abnormal green bond returns over the entire period shows that green bond premia have developed positively in terms of prices.

The thesis links to two different strands of literature. The work is thematically rooted in the literature on central banks and climate change. Campiglio et al. (2018) provide a summary of the academic literature. According to its early stage of development, the literature is mainly concerned with why central banks should be involved in the transition process to a low-carbon economy in the first place and what range of instruments is theoretically available to them. The literature review will discuss both aspects extensively to clarify that the tested dates differ fundamentally concerning the level of commitment of the ECB to tackle climate change. The second strand of the literature examines how central banks influence bond prices. Special attention will be paid to asset purchasing programmes and especially their announcement in an event study framework.

(9)

3

financing of the investment gap to a low-carbon economy in high-income countries is re-evaluated.

(10)

4

2. Literature Review

2.1 Why should central banks care about climate change?

To answer this issue, the first question that arises is: What is the purpose of central banks? Central banks are endowed with mandates from democratically legitimate legislators, but ideally, they are independent in their execution. Thus, the purpose depends on the form of the mandate and is, in principle individual to each central bank. The focus of this thesis is on central banks of developed industrial nations, in particular, the mandate of the ECB is subject of interest. Central banks endowed with a monopoly of creation for the most liquid form of money, namely cash and central bank reserves, can influence the quantity offered by monetary policy. Therefore, for those central banks, such as the ECB, maintaining price level stability is one of the primary tasks.

However, the fact that the central bank has a monopoly on issuing so-called 'high-powered' money does not mean that it can control the money supply of an economy deterministically and precisely. Greenham et al. (2012) describe how commercial banks are also able to create money by lending more than they have central bank money at their disposal, and thus generating fiat money. Minsky (1977) concludes that the financial system is inherently unstable following this feature due to a lack of incentives to restrict lending when the additional purchasing power of credit increases asset prices again. For this reason, central banks often have the task of regulating the financial markets, which, in addition to limiting lending, for example through minimum reserve requirements, also includes monitoring potential risks, the so-called macroprudential supervision.

The two purposes of central banks mentioned above can be extended in many ways. For some central banks, such as the Federal Reserve Board in the US, promoting economic growth is an equally important goal of monetary policy as inflation control. Other central banks actively pursue exchange rate policy in order to control the domestic exchange rate. To argue why central banks should not ignore climate change, however, a closer look at the objectives of financial stability is sufficient.

The financial stability argument

(11)

5

For instance, an rapid political exit from oil energy leads to the depreciation of undrilled oil reserves, which is known as stranded assets. Mercure et al. (2018) estimate that oil reserves worth 4$ trillion must not be extracted in order to meet the 2˚C target of the Paris Agreement and hence have to be written off.

Central banks are also often referred to as guardians of the currency because their actions are designed to preserve the functions of the money issued for transactions and as a store of value. If a sufficiently large volume of financial assets is spontaneously destroyed, the negative implications are directly threatening the viability of firms in the financial sector. It is then the responsibility of the central bank to prevent the negative positive-feedback-loops that could lead to the collapse of the financial system, which in turn would undermine the functions of money. Ideally, the situation in which banks, insurers and other asset managers are suddenly confronted with unmanageable losses does not even occur. This requires that physical risks and transition risks are correctly quantified, anticipated and priced. Although the assessment of such climate change-related risks is also in the self-interest of all actors with assets in question on the balance sheet, Lamperti et al. (2019) demonstrate that negative implications for the system as a whole are far more devasting. Therefore, the central bank, as a public institution, must identify these risks and point them out to the individual actors – known as climate change-related risk disclosure. Assessing climate-related risks for financial institutions is a rapidly evolving field in science. For instance, Battistion et al. (2017) attempt to assess this risk for European banks and conclude that precision in terms of both timing and raising expectations is the key when policies to mitigate climate-related risks are communicated in order to avoid a sharp devaluation of portfolios that consist to a significant amount of exposed assets.

The market failure in credit supply and green bonds

So far, there is a degree of consensus that central banks cannot completely ignore climate change and, in addition, that they must support the assessment of climate change-related risks and encourage financial market participants in their role as supervisors to disclose these risks. However, the debate on how far central banks should go is still intense. The range of intervention instruments is presented in the next subchapter, however, it is worthwhile anticipating why a central bank can go one step further than merely assessing and disclosing risks, i.e., taking a more active role in financing the transition.

Campiglio et al. (2018) argue that the physical risks can ultimately only be avoided if greenhouse gas emissions of the entire economy are reduced to such an extent that the rise in temperature is slowed down. While governments are the main actors responsible for implementing measures to achieve this, Campiglio (2016) shows that financial markets do not respond properly to the price signals of carbon taxes or traded carbon certificates since they underly market failures itself.

(12)

6

financing in Europe. Regulatory mechanisms, such as minimum capital and reserve requirements, play a large part in explaining this fact. The author concludes that financial regulators such as central banks have the responsibility to ensure a sufficient flow of capital in low-carbon investments, which under normal conditions, would be triggered by the price signals of carbon taxes. Differentiated reserve requirements for loans to polluting projects and green projects could make lending more attractive for desirable investments.

So, there is an argument for the central bank to promote investment as a second-best policy when government economic policy measures do not work as a first-best solution. The problem with most central banks in high-income countries, including the ECB, is that the pursuit of economic policy objectives is subordinate to the pursuit of price level stability. De Grauwe (2020) blames the ECB's strong independence in the execution of its mandate and its low accountability to the European Parliament for the fact that the ECB focuses almost exclusively on maintaining price stability. In the pursuit of price level stability, the ECB aims to avoid distorting markets in the direction of individual countries or sectors, making the use of an instrument such as differentiated reserves unlikely.

The combination of the prospect of limited bank lending in the euro area and the low priority given to supporting EU economic policy measures, coupled with low accountability to the European Parliament, dim the prospect of mobilizing large amounts of capital in the form of loans for the transition to a low-carbon economy. For this reason, it is particularly crucial for the euro area that an alternative source of financing develops rapidly, namely market-based debt, such as green bonds.

Green bonds are market-based debt instruments for investments with the intention of mitigating climate change, and in recent years they have rapidly developed into a source of huge investments. According to Deschryver and de Mariz (2020), in 2019 worldwide, $228.2 billion of green bonds were issued by governments, development banks, corporations and others predominantly in Europe. This represents an increase of 49.4% over the previous year

(13)

7

(Figure 1). Especially development banks, such as the European Investment Bank (EIB) or the German Kreditanstalt für Wiederaufbau (KfW) play a primary role in investments towards a sustainable economy and hence in the green bond market. However, unlike central banks, they cannot create reserves and therefore rely on market-based promissory notes. Therefore, it is vital to scale-up the green bond market and to examine the ECB's possible contribution to do so.

2.2 What can central banks do?

In principle, the central banks' activity consists of supporting the financial market in the financing phase in developing a low-carbon economy, the so-called transition. Dikau and Volz (2018) give an overview of the universe of potential intervention mechanisms and also provide examples of where the respective mechanisms are applied in the world. A more condensed overview of individual measures and a summary of the discussion on green QE is provided by Campiglio et al. (2018). The possibilities of individual policies are virtually infinitely diverse so that only groups of measures are discussed here and illustrated with a few examples.

The urgency of quantifying climate change-related financial risks and the implications for financial stability has already been highlighted, not least because this evidence ultimately serves as justification for the central bank to put climate change on its agenda. For this reason, research and the raising of awareness in this area is unavoidable. The Bank of England and the Nederlandsche Bank (2016) have already taken a leading role in this area outside the academic world. A second major area of research should also aim to incorporate the transition to a low-carbon economy into macroeconomic models, as attempted by Fontana and Sawyer (2016) with a growth model that also allows "potentially catastrophic environmental problems". Furthermore, due to their central position in the financial system, central banks have a coordinating role in ensuring an efficient diffusion of new insight and in coordinating individual measures. The establishment of the Network for Greening the Financial System (NGFS), a voluntary association of central banks and financial regulators around the world with a steadily growing membership, contributes in particular to the latter objective.

(14)

8

investment is assessed in terms of environmental, social and governance aspects (ESG) in addition to its pure financial performance. If, for example, a corporation is characterized by high returns and low risk, but causes a disproportionately high level of greenhouse gases and toxic waste compared to the industry average, best-in-class ESG screening can help to ensure that fewer shares in this company are held. Here too, Nederlandsche Bank can be mentioned as a pioneer in this practice since they already apply ESG guidelines to their own portfolio. ESG screening can even be applied in the next step in terms of accepting collateral when the central bank lends reserves to commercial banks. As a result, commercial banks can borrow liquidity at better conditions for the same financial performance of a security if it has a high ESG score. The attractiveness of suitable ESG securities from the perspective of the commercial banks means that the incentive is passed on to the companies, as they can, ceteris paribus, finance themselves more advantageously from the commercial bank. Besides, Goss and Roberts (2011) detect a lower cost of capital in terms of bank loan interest payments for companies with a high corporate social responsibility (CSR) rating and infer that banks assess the performance of responsible companies as superior. At this point at the latest, a threshold is crossed, from a mere monitoring and support function to active participation in financing the transition with direct consequences for the real economy; indicated by the bold line in Figure 2. The Norwegian Central Bank, for example, takes the ESG score of an asset into account when it lends against central bank reserves.

Within the framework of QE already mentioned, a central bank mainly purchases bonds from states, various agencies such as development banks, and companies in order to provide banks with sufficient liquidity with newly created reserves. This procedure is intended to stimulate lending in order to create inflation and support the economy. The instrument is a non-conventional monetary policy instrument that is used when the non-conventional range of action has been exhausted, i.e. when the refinancing rate has reached the lower boundary. It is usually a temporary purchase that is reversed if there are signs of excess liquidity. The ECB, for example, applies the principle of market neutrality when intervening in the capital market in this way. Accordingly, bonds are purchased in relation to their market volume in order to keep market distortions as low as possible. Theoretically, it is possible to impose additional criteria on a QE programme, such as a disproportionate purchase of green bonds. Bonds issued by governments as well as institutions and companies can be classified as green if their financing goal is to transform them into low-carbon issues. Two main arguments against this are obviously the violation of market neutrality and the fact that central banks, as non-democratically legitimized institutions, pursue an apparently political goal.

(15)

9

However, the suggestion of a green QE programme cannot be invalidated that easily. A rather odd measure to support the financing process of transition is to monitor carefully that standard measures do not impede this process. Matikainen et al. (2017) argue that the ECB's QE programme in the aftermath of the euro crisis has ensured that traditional industries have benefited from improved financing conditions due to low risk and liquidity premiums. Green companies, on the other hand, were often still too small and had a higher risk due to their disruptive technology character, so that these bonds were not classified as ECB-eligible. As there is still no prospect of a relaxation of liquidity-promotion policies, including the anti-cyclical QE programme, the financing advantage of conventional technologies seems to be permanent. According to this argumentation, the establishment of a green QE programme would therefore not be a relative improvement of the position of green bonds. However, it would be necessary in order not to distort the market in such a way as to inhibit innovation. Beyond green QE, central banks with the goal to promote sustainable development, such as the Reserve Bank of India and Bangladesh Bank, have the option of requiring commercial banks to provide a minimum proportion of loans to be granted for sustainable development or even to grant loans directly. Alternatively, differentiated reserve requirements could be imposed on loans for unsustainable and sustainable investments, thereby providing an incentive for banks to provide more loans for the latter option. The instruments above, including green QE, make use of the money creation potential of central banks, which is why vast sums of financial resources can be mobilized suddenly. On the other hand, this further increases leverage, which, according to Brunnermeier (2009), is a decisive factor for financial instability.

In conclusion, there is a variety of instruments that central banks can use to support the financing process towards a low-carbon economy. The order of the presented measures is by no means arbitrary. Figure 2 shows that on a scale from left to right, the measures presented show an increasing degree of intervention. The sequence in this form makes it clear that with an increasing degree of intervention, the direct influence on the financing of the transition increases, but also that the measures are more difficult to reconcile with the mandate and are hence viewed more critically. Therefore, jumps on the scale from left to right seem unlikely for central banks without a sustainability component and without an equally important goal of supporting the general fiscal policy of the national state, for instance, the ECB. It has already been mentioned that from the ESG screening of the collaterals for the lending of reserves, it is possible to speak of a more active role of the central bank in the financing process of transition. This break is vital for the terminology of the specific research questions that will be presented in Chapter 2.4.

2.3 Central banks and the financial market

(16)

10

Bonds belong to the class of fixed-income financial instruments. The issuer promises regular coupon payments, usually in a fixed amount, to the bondholder and repayment of the face value after a fixed agreed term. In the bond pricing theory (e.g. Duffie (2010)), the price is derived from the present value of all future outstanding payments. This means that future payments are discounted at a reference interest rate of a theoretically risk-free security. In practice, however, bonds are by no means homogeneous, apart from the size of their coupon. Firstly, some bonds can be sold more quickly because the market is more liquid than before. Secondly, the risk that payments may be partially or wholly cancelled due to the issuer's payment difficulties differs for all bonds. The lower the default risk (see, e.g. Rodriguez (1988)) and the higher the liquidity (see, e.g. Archaya and Pedersen (2005), Chen et al. (2007), Favero et al. (2010)), the more investors are satisfied with lower yields. As a consequence, the lower the yield of fixed cash flows, the higher the price of the bond. If the central bank announces that it will become an additional buyer of a group of bonds, the market becomes more liquid, and investors consider the risk of default to be lower, which may reflect in a higher price. Issuers also benefit from lower yields, as investors may be able to offer a lower coupon when issuing new bonds. Mésonnier et al. (2017) speak of an 'interest of being eligible'.

As explained above, maintaining price stability is the primary objective, and the instruments are, therefore, strongly focused on this objective. For this analysis, it is in principle sufficient to assume the following relationship: If the central bank anticipates inflationary tendencies, it tries to let capital market interest rates rise in order to dampen consumption and investment. The opposite is true for deflationary tendencies. The most important conventional instrument is the determination of the main refinancing operations rate. At this rate, commercial banks can borrow from the ECB in exchange for eligible collateral, central bank reserves. Forward guidance refers to the announcement of future main refinancing rate determinations, which also depresses the yield of bonds with longer maturities as the interest-rate risk is reduced. The idea behind determining this actually short-term interest rate is to reduce the opportunity cost of bonds in order to strengthen demand for them. When the lower bound, i.e. an interest rate of 0%, was reached in spring 2016, the scope for this instrument was exhausted.

(17)

11

The portfolio rebalancing channel shows that quantitative easing programmes not only reduce the yield of directly purchased bonds. The portfolio rebalancing channel predicts that these investors will move into riskier markets in order to maintain their returns after the central bank depressed returns of low-risk bonds. Albertazzi et al. (2018) were able to empirically prove this channel as a result of the ECB's large purchasing programmes. Zaghini (2019) also demonstrates this effect explicitly for corporate bond purchase programmes.

The investigation in this thesis is not based on an explicit announcement of a purchase programme. However, investors may anticipate a possible customized purchase programme only for green bonds. Cieslak and Schrimpf (2019) show in an event study that statements made by central bank representatives in press conferences also have an instantaneous effect on the bond market. Green (2004) also shows that the trading volume of government bonds increases significantly when news is published. The higher the surprise component of the news, the higher the simultaneous trading volume.

2.4 Development of the research questions

So far, the consensus has been discussed that central banks cannot ignore climate change because of the financial risks it implies. Furthermore, the market failure argument was presented, which envisages supporting economic policy with far more active interventions than promoting risk-disclosure if they are not effective in the financial market. Subsequently, it was argued that such interventions by the ECB are unlikely because support for the general economic policies of the EU is deficient in the ECB's own interpretation of the mandate and because the ECB has little accountability to third parties. From this, it has been concluded that bank credit will only help to a very limited extent to close the investment gap to achieve climate targets, which is why green bonds remain a kind of hope in Europe and their development is essential.

Chapter 2.3 gave an overview of the potential set of instruments available to the ECB, or to central banks in general, to support the financing process towards a low-carbon economy. The further to the right the measure is on the scale in Figure 2, the more controversial its implementation is likely to be with regard to the ECB mandate. If measures behind the line in Figure 2 are applied, this study speaks of a more active role in financing the transition. In this way, the general research question of whether the presence of the ECB in the debate regarding financing the transition positively affects euro-denoted green bonds can be divided into two specific research questions:

Research Question 1: Does the expression of the ECB to take climate change somehow into

consideration positively affect euro-denominated green bond prices?

Research Question 2: Does the expression of the ECB to consider a more active role of the

ECB to support the financing phase towards a low-carbon economy positively affect euro-denominated green bond prices?

(18)

12

(19)

13

3. Methodology and Data

An event study approach is used to examine the reaction of the market for euro-noted green bonds after two crucial announcements of the ECB concerning its role in the challenge of tackling climate change. In event studies, the basic idea is to test the deviation of bond prices during the event period from a return expected without the event, the abnormal return, for statistical significance. In principle, the implementation of an event study can be divided into three steps. First, the events and their timing relevant to the scientific question are determined. Second, the abnormal return on the securities is determined, which is done by subtracting an expected return from the actual return. The expected return represents a hypothetical return, in which the new information from the event is not yet taken into account by the market. Finally, thirdly, a statistical test is performed to determine whether the abnormal returns are significantly different from zero.

In this chapter, the choice of the approach is extensively discussed and justified. This includes the determination of the events to investigate the research questions, an overview of the possibilities to design a bond event study, a discussion of the extent to which an event study can be used to answer the research questions and a section on the data. Finally, for clarity, the methodology is presented compactly and formally, which makes it easier to look up the terminology quickly when reading the results.

3.1 Determination of the events

The existence of abnormal returns requires an event through which unexpected information that directly or indirectly affects the price of the bonds reaches market participants. Appropriate central bank events which, firstly, have an unexpected effect and, secondly, are relevant to the decisions of financial market actors, may, for example, take the form of measures announced in press releases or considerations in speeches by key ECB officials. In order to find these events, the first step was an extensive internet search on the topic ECB and

climate change. By working through discovered newspaper reports, online articles, press

releases, and homepage entries, it was possible to make an initial assessment of which information or which steps the ECB was considering at what point in time.

(20)

14

challenge of the transition process towards a low-carbon economy. The event was not publicly announced before the NGFS press release, which is why it is suitable to be used to answer the research question 1, whether the presence of the ECB in the transition-debate has a positive impact on the green bond market.

The second research question, on the other hand, relates to the expressed willingness of the new ECB President Christine Lagarde to take a more active role in the financing process. On 02/07/2019 it was announced that Christine Lagarde was nominated as successor to Mario Draghi. At that time, according to the published reports, it could not yet be foreseen that the ECB would in future expand its role in the challenge against climate change, which is why this date is also not considered suitable to examine research question 2, although this date should also be checked to rule out a possible impact.

In order to find the crucial event that triggered the increased reporting on the ECB's future role under Lagarde in climate change, all articles published in the Financial Times Europe Edition (E-Print Edition) from 29/02/2020 back to 01/05/2019 were manually checked for new information in this context. This step is to ensure that no significant events are missed in addition to the keyword search in the very relevant period. This procedure has mostly ensured that all relevant information was assembled. However, the exact timing and source of the event could not yet be identified.

As it turned out, the Financial Times published a series of articles on the role of the ECB under Lagarde in climate change not until four weeks after the actual event. The actual event was not reported. Therefore, the E-Print search was again extended by a more precise keyword search on the internet, which finally pointed to an article (Lagarde promises to paint

the ECB green - Politico 29.08.2019, updated 04.09.2019), which refers to two events that

suggest that the ECB is taking a green course under Lagarde. The European Parliament's Committee on Economic and Monetary Affairs (ECON) presented Christine Lagarde with a questionnaire in which she was asked to take a position on individual issues at the ECB and to outline her future course. The answered questionnaire (Q&A paper) was submitted to ECON, which published it on 29/08/2019.

The contents of the 48-page document already indicated the green course, although she did not explicitly target an active financing role, such as the use of green QE programmes. The topic was not given media attention until after the nomination hearing on 04/09/2019 in front of ECON, in which Lagarde discussed topics from the Q&A documents and furthermore, the application of green QE in more detail. There she said, among other striking statements, that she declared climate change to be "mission crucial" for the ECB. After this, no further new events could be identified that would trigger new information on the ECB's course on climate change. Thus, the nomination speech in front of the ECON on 04/09/2019 was determined, which is used in an event study to examine whether the ECB's announcement to consider a

more active role in the financing process of the transition actions has a positive price effect on

(21)

15

In the event study literature for stock markets, the cumulative return of the event date, as well as one day before and one day after the event, is often calculated. However, the procedure of Bessembinder (2009) was followed and instead only daily abnormal returns were calculated, but for control, also separately on the days before and after the event.

3.2 The Event Study approach in bond markets

The event study literature to investigate reactions of stock market prices to various events is very extensive. Between 1974 and 2000, Kothari and Warner (2007) counted 565 articles that include an event study in the top five finance journals. Maul and Schiereck (2017), on the other hand, 12 years later, could only identify 118 bond event studies in the 15 top finance journals. The stock market event study approach, in principle, can also be applied to other securities, such as bonds. However, because the bond market has different characteristics than the stock market, the actual implementation differs considerably. For instance, Goyenko and Ukhov (2009) found a higher heterogeneity in terms of characteristics such as liquidity and risk. Besides, according to Bessembinder et al. (2009), different classifications of the various bonds in terms of maturity, interest payment conditions and other special characteristics such as putability and convertibility result in heterogeneous returns that cannot be adequately explained by stock market pricing models. This circumstance, in turn, has implications for the modelling of expected returns, which are necessary to determine the abnormal returns.

Furthermore, particular care should be taken in the choice of test statistics, since an unbiased test result of a parametric test, such as the t-test, requires normally distributed abnormal returns with a uniform variance. These conditions are often met in stock market event studies, which use market models to calculate abnormal returns, but not for abnormal returns in bond prices. For this reason, the use of non-parametric tests needs to be discussed. Ederington et al. (2015) propose standardization of abnormal returns in terms of maturity and risk to reduce the problem of heteroscedasticity and thus make parametric as well as non-parametric tests more reliable.

3.2.1 Calculating Abnormal Returns

Since daily closing bond prices are provided from the Eikon database, daily returns must first be calculated. This is done by the following formula, where 𝐵𝑃_𝑖,𝑡 is the closing price for bond 𝑖 at day 𝑡:

𝐵𝑅_𝑖,𝑡 = 𝐵𝑃𝑖,𝑡− 𝐵𝑃𝑖,𝑡−1 𝐵𝑃_{𝑖,𝑡−1}

Both the selection of the model for determining abnormal returns and the selection of the statistical tests are based on the study by Bessembinder et al. (2009). This study tests the reliability of the different approaches to correctly identify simulated events from previously randomly compiled firm-event observations by detecting abnormal returns. Events were simulated by modifying the daily return ranging from +/-10 to +/- 20 basis points.

(22)

16

available from the Eikon database. Bessembinder et al. (2009) summarise the three most common methods for determining the expected return. Firstly, Fama and French (1993) developed a five-factor bond model based on their three-factor stock return model. In this model, the daily excess return of a bond is explained by the risk-free return in a linear regression with a constant, the variables market risk premium, small minus big factor, high minus low, the slope of Treasury yield curve, a default premium and an error component. The abnormal return of the day is then captured by the regression constant. Compared to the other models for determining abnormal returns, however, the five-factor model in the study by Bessembinder et al. (2009) proved to be inferior, since abnormal returns from simulated events were less often correctly identified with all test statistics. Consequently, this approach will not be explained in detail here.

Another, technically less sophisticated, method for determining abnormal returns is the mean-adjusted model. The idea is to determine the expected return at time 𝑡 based on an average daily return from a past estimation window, as Brown and Warner (1980) developed for stock returns. Handjinicolaou and Kalay (1984) modified this approach to account for the change in term structure that applies equally to all bonds. First, a premium bond return 𝑃𝐵𝑅𝑖,𝑡 is

calculated for each bond 𝑖 at time 𝑡

𝑃𝐵𝑅𝑖,𝑡 = 𝐵𝑅𝑖,𝑡 – 𝑇𝑅𝑖,𝑡

where 𝐵𝑅_𝑖,𝑡 is the observed return of the bond 𝑖 and 𝑇𝑅_𝑖,𝑡 is the return of a matched treasury security 𝑖 for the same day 𝑡. The matched treasury security should have the same coupon payments and time to maturity. The next step is to calculate the mean expected excess return 𝐸𝐵𝑅_𝑖 during the defined estimation window (−𝑦, −1)

𝐸𝐵𝑅_𝑖 = 1

𝑦 ∑ 𝑃𝐵𝑅𝑖,𝑡 −𝑦

−1 .

Finally, to calculate the abnormal return 𝐴𝐵𝑅_𝑖,𝑡 for bond 𝑖 at event 𝑡, the expected excess return 𝐸𝐵𝑅𝑖 is subtracted from the premium bond return 𝑃𝑅𝑖,𝑡

𝐴𝐵𝑅𝑖,𝑡 = 𝑃𝐵𝑅𝑖,𝑡 − 𝐸𝐵𝑅𝑖.

(23)

17

calculated in this analysis by subtracting the return of the matching portfolio, consisting of conventional bonds with the same risk and maturity 𝑀𝑃𝑅_𝑖,𝑡

𝐴𝐺𝐵𝑅𝑖,𝑡 = 𝐺𝐵𝑅𝑖,𝑡 – 𝑀𝑃𝑅𝑖,𝑡

In principle, this equation will be used to examine the research questions. The extent to which this procedure captures the pure effect of the treatment Information regarding supporting

green finance needs to be discussed. The return of a reference portfolio consisting of

conventional bonds with the same rating and maturity is subtracted from each green bond. The effects that are independent of the greenness characteristic are hence filtered out in this procedure. That means that the expected return already contains abnormal return movements induced by the information that does not only influence green bonds. If the abnormal green bond returns are then still significantly positive, it is concluded that the information that only affects green bonds, namely the ECB's commitment to tackle climate change, will have a positive impact on green bonds.

3.2.2 Statistical tests on the significance of abnormal returns

The overview of conducted bond market event studies from Maul and Schiereck (2017) shows that most of the earlier studies check the abnormal returns for significance using simple t-statistics. Under the assumption of a normal distribution, the respective abnormal returns are checked whether they differ significantly from zero on average. The problem is that abnormal returns, regardless of the model chosen, are rarely normally distributed and usually heteroscedastic, as Bessembinder et al. (2009) show. While the property of symmetry is still present, distributions of abnormal returns exhibit significant excess kurtosis due to the abundance of observations close to zero. Even though the assumption of symmetry is more important for an unbiased test result, the simulations from Ederington et al. (2015) and Bessembinder et al. (2009) show that the non-parametric tests sign-test and Wilcoxon Signed-rank test are more reliable, since these tests do not require normally distributed abnormal returns.

Ederington et al. (2015) suspect varying variances in abnormal returns, i.e. heteroscedasticity, as another reason why the simple t-test is less reliable. The weaker the rating, and the longer the maturity, the higher the variance in returns. While Kothari and Warner (2007) consider heteroscedasticity in equity returns to be negligible, especially for short event periods, the variance of returns in the reference portfolios varies considerably. For this reason, Ederington et al. propose an additional standardization of abnormal returns in terms of their maturity and ratings. First of all, for all bonds, i.e. for green bonds as well as for conventional bonds used to create the reference portfolios, standard deviations of the daily returns in the combined period from t-55 to t-6 and t+6 and t+55 are calculated. All returns are then divided by their (time-varying) standard deviation. After standardizing the returns, the determination of abnormal standardized green bond returns by the matching portfolio method follows.

(24)

18

intrinsically different from conventional bond returns, which violates the assumption of independence of the intra-day sample of AGBRs. As a consequence, daily AGBRs are rarely zero on average, instead green bond returns systematically deviate positively or negatively from the normal return so that abnormal returns would too often be considered as significant using this method. The fact that the majority of green bond returns on any given day deviate either positively or negatively from the returns of conventional bonds is an unexpected feature of the green bond market. One possible explanation could be that due to their characteristic greenness, bonds react very sensitively to information that is only relevant to this characteristic. Due to that unique feature, the test procedure must be modified in such a way that particularly strong deviations are screened out of the fundamental deviation of the green bond market. This is achieved by aggregating all bond-specific AGBRs for each day into one day-specific AGBR, which then forms the basis for the distribution of the t-test statistics. This procedure ensures that the test statistics are time-dependent, and then statistically significant daily AGBR can be identified.

The use of the non-parametric sign-test is subject to the same weakness. In the sign test, the null hypothesis states that on a given day, the individual AGBRs have a median of the average of all individual AGBRs over the entire period (𝐴𝐺𝐵𝑅̅̅̅̅̅̅̅̅). This is in contrast to the alternative hypothesis that the median is greater than 𝐴𝐺𝐵𝑅̅̅̅̅̅̅̅̅. For the one-sided hypothesis test, those bonds are counted first whose AGBRs are greater than 𝐴𝐺𝐵𝑅̅̅̅̅̅̅̅̅ (S). Assuming the binomial distribution with the probability of success of ½ and the number of green bonds on a given day 𝑛_𝐺𝐵, the test statistic Z is calculated. Since a sufficiently large 𝑛_𝐺𝐵 is assumed, the normal distribution is approximated with the estimators for mean and variance from the binomial property. For reasons of continuity correction, ½ is deducted from S:

𝑍 = (𝑆 − 1 2) − 𝑛 2 √𝑛𝐺𝐵∗1₂

However, due to the dependency among green bond returns, a deviation from zero median within one day is more the rule than the exception. For this reason, the time-dependent test statistics are used as the primary test decision due to their strict test discipline. The sign-test serves merely as an additional robustness test. What the described problem means in concrete numbers, i.e. rejection rates, is evaluated in the result part.

3.3 A Summary of the Methodology

Firstly, the daily returns of green and conventional bonds are calculated from dirty closing bond prices:

𝐺𝐵𝑅_𝑖,𝑡 =𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1− 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡 𝑃𝑟𝑖𝑐𝑒𝑖,𝑡−1

(25)

19

𝑆𝐺𝐵𝑅_𝑖,𝑡 = 𝐺𝐵𝑅𝑖,𝑡 𝜎𝑖,𝑡

where 𝜎_𝑖,𝑡 is the standard deviation of the combined period t-55 to t-6 and t+6 to t+55. The same holds for standardized returns of conventional bonds 𝑆𝐵𝑅_𝑗,𝑡.

Thirdly, Abnormal Standardized Green Bond returns 𝐴𝐺𝐵𝑅_𝑖,𝑡are calculated by applying: 𝐴𝐺𝐵𝑅_𝑖,𝑡 = 𝑆𝐺𝐵𝑅_𝑖,𝑡− 𝑆𝑀𝑃𝑅_𝑡(𝑖)

where SMPR𝑡 = ∑𝑛_𝐵𝑗=1𝑡𝑤𝑗∗ 𝑆𝐵𝑅𝑗,𝑡 and 𝑆𝑀𝑃𝑅(𝑖) is a function that assigns 1 out of 24

matching portfolio return to green bond 𝑖 and 𝑛_𝐵𝑡 is the number of bonds per portfolio per

day.

Finally, the aggregation of all green bond-specific 𝐴𝐺𝐵𝑅_𝑖,𝑡 each day delivers the basis for the t-test statistic:

𝐴𝐺𝐵𝑅𝑡= ∑ 𝑤𝑘∗ 𝐴𝐺𝐵𝑅𝑘,𝑡 𝑛_𝐺𝐵𝑡

𝑘=1

where 𝑤_𝑘 is the volume-based weight for bond 𝑘.

Correcting for the mean µ_{𝐴𝐺𝐵𝑅} of all 𝐴𝐺𝐵𝑅_𝑡 for all 𝑡 and division by a standard deviation estimated from a 100-day estimation period around 𝑡 gives the final test statistic:

𝑡_𝑣𝑎𝑙𝑢𝑒_𝑡 = 𝐴𝐺𝐵𝑅𝑡− µ𝐴𝐺𝐵𝑅 𝜎_𝑡

which is assumed to be standard normally distributed. Hence, for testing one-sided statistical significance the 90%, 95% and 99% quantiles of the standard normal distribution were used, i.e., 1.282, 1.645 and 2.326, respectively.

3.4 Data

(26)

20

Due to a limited selection of databases for licensing reasons, I refrain from a detailed discussion about the selection of the appropriate database. In this thesis, the Eikon database of Thomson Reuters is used, which already contains calculated daily bond prices. The Eikon database is the successor of the better-known Datastream database by Thomson Reuters. The upgrade from Datastream to Eikon brings a decisive advantage for this investigation. For each bond, information is available on whether it is classified as a Green Bond, according to the Climate Bond Initiative (2020). A further advantage is that maturity and rating information is also available and does not need to be imported from another database.

For both the daily prices of green bonds and the conventional bonds that are used to construct the reference portfolios, the same restrictions were made. Convertible, puttable and zero bonds were excluded, as these instruments, in general, are expected to have a higher probability of detecting abnormal returns due to their unique characteristics (Bessembinder et al., 2009). Furthermore, in order to calculate a maturity-rating specific normal return, bonds with at least one missing information regarding maturity, rating and bonds without return data during the event window were excluded. First, the availability of the rating by Moody's is checked. If no rating is available from Moody's, Standard & Poor's ratings are used, a common procedure to avoid a loss of observations. As there is primarily an effect on euro-listed bonds expected, in a final restriction, all bonds in another currency are excluded from the analysis.

(27)

21

4. Results

The presentation of results comes in three steps. In the first step, after a brief presentation of the composition of the green bond sample, descriptive statistics for returns and abnormal green bond returns (AGBR) are presented. The objective of the descriptive analysis is, on the one hand, to demonstrate the advantages of standardizing returns and, on the other hand, to motivate the construction of the test statistics.

In the second step, the event study methodology is applied to the selected events in order to test the two hypotheses. Positive AGBR on 29/05/2018 (NGFS entry) are tested in terms of significance to give insights in research question 1, and positive AGBR on 04/09/2019 (Lagarde's ECON hearing) for research question 2, respectively. As a reminder, while research question 1 refers to the impact on the euro-denoted green bond market after the ECB announced to take climate change somehow into consideration, question 2 aims for the impact after the announcement to consider a more active role by expanding the set of instruments.

In the last step, AGBR are examined over the entire period between 2015 and 2020 to identify additional events that caused abnormal green bond returns. The evolution of the difference between green bond prices and conventional bond prices is also used to draw conclusions about a possible green bond premium over time.

4.1 Descriptive Statistics

Daily normal returns were calculated for each rating-maturity portfolio from a total of 13,916 conventional bonds. All individual returns or standardized returns are weighted with a bond-specific weighting based on its issue volume. The weighted daily return of each matching portfolio forms the normal return of the matching green bond. Due to the aggregation of multiple bonds per day in a class-specific normal return, I forego to present summary statistics on individual conventional bonds.

The sample of green bonds comprises 309 bonds whose distribution of rating, maturity and ECB eligibility is shown in Table 1. It is noticeable that only a few bonds have a rating below Baa, so that a category 'below Baa' would have also been conceivable. However, the rating classes Ba and B differ in terms of average abnormal returns (Table 2) and the variation in abnormal returns (Table 3), which is why a separation was made despite the small sample size. Throughout the whole thesis, returns and abnormal returns are consistently given in basis points, i.e. in hundredths of a percent.

(28)

22

appropriate to separate the two rating groups Aaa and Aa, as 12.9% of all green bonds listed in euro have the best rating Aaa. The low average risk of the entire green bond sample is also reflected in a share of 68% that meets the ECB's eligibility criteria, which includes rating standards.

Overall, in Table 2, the heterogeneity in abnormal green bond returns highlights why it is necessary to create rating/maturity-specific portfolios to determine normal returns. Consider that the heterogeneity of abnormal returns between the groups was not taken into account, for instance, by subtracting a return from a uniform index from each green bond return. Then, the value-weighted AGBR of all green bonds per day would depend on the composition of the sample in terms of risk and maturity. The approach chosen in this thesis, however, is robust to this problem, since a bond-specific normal return is subtracted before a value-weighted aggregation. Therefore, it is not necessary to mimic the composition of the uniform index, which would have potentially results in a smaller sample size.

Let us turn to a second issue, which refers to the power of the t-test to detect significant AGBR. Table 3 Panel A shows the standard deviations of AGBRs, depending on their rating/maturity classification. It is noticeable that abnormal returns vary to different degrees depending on the group, which violates the assumption of homoscedasticity and makes the parametric t-test unreliable if this circumstance is not taken into account. It becomes evident that the variance increases both with higher risk and with longer time to maturity. A decreasing variance in the Ba-rated portfolio with increasing maturity is the exception in this sample. However, it should not be of particular importance due to the small sample size of Ba-rated bonds of 30. The standardization method proposed by Ederington et al. (2015) was used to address the problem of heteroscedasticity. Before subtracting normal returns from green bond returns, each return, both of green bonds and conventional bonds, was divided by its standard deviation of a 100-day estimation interval around the particular day. Panel B of Table 1: The Green Bond sample

Rating (%) Maturity (%) ECB-eligible (%)

Aaa 12.9 1-3 years 12.9 Yes 68.0

Aa 25.6 3-5 years 19.4 No 32.0

A 35.3 5-10 years 45.0

Baa 23.0 over 10 years 22.7

Ba 1.0

B 2.3

Note: The total sample size of euro-denoted green bonds is 309.

Table 2: Average daily Abnormal Green Bond Returns (unstandardized)

Bond term-to-maturity

Rating 1 to 3 years 3+ to 5 years 5+ to 10 years Over 10 years

Aaa _-0.0866 _0.1109 _0.0132 _0.4255 Aa 0.2263 0.3074 0.1874 -0.3321 A -0.1044 0.6978 0.0697 0.0316 Baa 0.0925 -0.2419 -0.3356 -1.4928 Ba 1.2045 1.3487 NA NA B NA 2.0325 -2.3813 NA

(29)

23

Table 3 shows that the variance of AGBRs has converged significantly after standardization. While for panel A the standard deviation of the most volatile portfolio is more than seven times higher than the standard deviation of the least volatile portfolio, the factor after standardization was reduced to less than two.

Following the suggestions of the current state of research on bond market event studies, firstly, the matching portfolio method was applied to determine normal returns and, secondly, returns were standardized to tackle heteroscedasticity. Table 4 shows, these procedures result in a distribution of AGBRs that is, in principle, satisfactory in terms of the appropriateness to perform the t-test. AGBRs are only slightly negatively skewed (-0.0135) and show a low excess kurtosis of 0.6100 (the kurtosis above the value of the standard normal distribution, which is three). Both characteristics suggest an adequate t-test, which tests whether the distribution of AGBRs within a day is different from zero on average; let us say the intra-day t-test. The fact that half of the AGBRs are positive, and half are negative also suggests the applicability of the sign-test, which is also limited to the sample of green bonds on the event day.

However, these standard test procedures are not suitable for this particular study, as the price movements of the entire green bond market differ systematically from the price movements of conventional bonds, which means that the intra-day sample of green bonds is not independent. This, in turn, means that AGBRs in the daily samples are rarely zero on average and statistically significant AGBRs are the rule rather than the exception. According to the intra-day t-test at a 5%-significance-level, on 28.80% of all intra-days, the AGBRs are statistically significant and 29.58% negative - not a suitable acceptance rate to test the importance of the events.

Therefore, this study uses a time-dependent t-test statistic, which means that intra-day AGBRs are aggregated to a daily observation. Finally, the distribution of all daily aggregated AGBR forms the basis for the following test statistics, which is the same as already shown in in the Table 3: Standard deviations of Abnormal Green Bond Returns by Risk and Maturity

Panel A: Bond term-to-maturity

Unstandardized 1 to 3 years 3+ to 5 years 5+ to 10 years Over 10 years

Aaa _5.9762 _7.8745 _10.4572 _24.8277 Aa 6.0245 6.7394 11.8373 22.8692 A 6.2458 8.3437 13.1704 22.6543 Baa 12.2597 16.3123 22.8300 45.2708 Ba 20.6699 13.6524 NA NA B NA 28.4371 30.2726 NA Panel B: Standardized Aaa 0.6975 0.5376 0.4747 0.4609 Aa 0.6512 0.5277 0.4694 0.4462 A 0.6769 0.5691 0.5512 0.5469 Baa 0.8289 0.7859 0.8142 0.9020 Ba 0.6585 0.6742 NA NA B NA 0.8284 0.7662 NA

Note: NA indicates that there are no green bonds in the sample with the respective risk/maturity profile. A

(30)

24

Chapter Methodology and Data, where AGBR are aggregated abnormal standardized green bond returns for day t with a day-specific standard deviation:

𝑡_𝑣𝑎𝑙𝑢𝑒_𝑡 = 𝐴𝐺𝐵𝑅𝑡− µ𝐴𝐺𝐵𝑅 𝜎_𝑡

The distribution of all test statistics, ranging from 20/03/2015 to 31/01/2020, is presented in Figure 3. Test-statistics are plotted as a histogram, and a standard normal distribution is drawn as a line and serves as a target value. The bold red line indicates the critical 95%-percentile. Due to construction, the t-test statistic has a mean close to zero (0.0039), and the standard deviation is approximately one (0.9978). The t-statistics are slightly positively skewed (0.0047) and show an excess kurtosis (0.7384). The excess kurtosis indicates an accumulation of observations around zero, which makes the distribution higher and steeper and more pronounced at the ends than the normal distribution. Although the P-value of the Jaque-Bera test statistic, a formal test for normal distribution, is 0.000 and thus rejects the hypothesis of the normal distribution, the test outcome relies on this property. The main reason to do so is that the test procedure detects on 5.27% of days significantly positive AGBR, which is found to be a desirable proportion to call an event essential. However, because of the non-normal distribution, the non-parametric sign-test is used in robustness tests.

Table 4: Distribution of Raw and Abnormal Green Bond Returns

Mean Median deviation Standard Skewness kurtosis Excess positive % n Raw returns 1.0143 0.6829 23.3788 -0.2663 3.3889 54.37 151,094 Standardized raw

returns 0.0609 0.0525 0.8830 -0.1104 0.4032 54.73 134,545 Abnormal green bond

returns 0.0564 0.0574 14.1550 -0.0097 1.8522 50.31 128,537 Abnormal standardized

green bond returns 0.0032 0.0014 0.6146 -0.0135 0.6100 50.13 128,537

Note: Only to demonstrate the difference between AGBR before and after standardization the term 'Abnormal

(31)

25

4.2 Event Study Results

Test results of the hypotheses for the two specific research questions are presented in Table 5. Research question 1, whether the ECB's announcement to take climate change into consideration positively affects the euro-denoted green bond market, is examined by testing the significance of positive AGBR on the corresponding event, the entry of the ECB into the Network for Greening the Financial System (NGFS) at 29/05/2018. On that day, the test statistic is with 2.4898 in the rejection range (1.645,∞), which is why the hypothesis that the AGBR was smaller or equal to average can be rejected. In other words, the ECB's entry into the NGFS triggered significant positive abnormal green bond returns. The result of the intra-day sign test confirms the result at a 1%-level. The test result is also robust to a variety of modifications of the test specification. The test statistic remains significantly positive for, ceteris paribus, the use of 2-day returns (t=2.9290), the use of clean prices instead of dirty prices (t=2.4466), and the use of unstandardized returns (t=4.0368). The sign-test also indicates significance if several bonds per firm are aggregated to a firm-specific bond in order to exclude distortions caused by firm-specific events of large firms with several bonds (P-value is 0.0018).

Concerning the research question, it can thus be concluded that the ECB's mere presence in the financing-the-transition debate has a positive impact on the market for euro-denominated green bonds. It is also worth noting that mainly bonds with weaker rating support the test result. It is not reasonable to assume that the described mechanism of the reduced risk and liquidity premium relates exclusively to bonds that can actually be bought by the ECB. The

(32)

26

event could just as well trigger the expectation that the ECB could extend eligibility for green bonds to Baa in the future, with the result that this group of 'marginal' bonds would become more attractive at the time of the announcement in anticipation of becoming eligible, manifesting in high AGBR. Furthermore, for Aaa-bonds, there may no longer be a risk-specific spread between green bonds and conventional bonds, and market participant will not rate the additional criterion 'greenness' high enough to reverse the spread.

The second research question asks whether the announcement of the new ECB president, Christine Lagarde, to reassess the ECB's role in the process of financing-the-transition by considering an expanded set of instruments, like green QE, positively affected the green bond market. On 04/09/2019, the date of the ECON hearing, positive AGBR are examined in terms of significance, respectively. Additionally, testing AGBR at the dates of the nomination of Lagarde 02/07/2019, and of the publication of the Q&A paper on 29/08/2019 ensures to capture this effect when it has arisen before.

The primary event on the 04/09/2019 could not produce significant AGBR; the test statistic of 0.5438 is lower than the critical rejection value of 1.645. Furthermore, the p-value of the sign-test is 0.3215. Looking at different sub-samples, green bonds issued by either governments, agencies or central banks (no-corporation), with the best rating (Aaa) or with a Ba rating, showed significant results. Most of these are green bonds with a low remaining maturity (1-3 years).

(33)

27

Overall, there is little evidence for a significant positive impact on the green bond market due to the new ECB president's expressed commitment to participate more actively in financing the transition process. The commitment of the new ECB president to pay special attention to the challenges of climate change and even to question traditional principles of monetary policy, such as market neutrality, for this purpose did not result in any excess performance of green bonds at the time. Before the analysis, it was assumed that Christine Lagarde's clear commitment to declare climate change 'mission crucial' for the ECB, combined with the questioning of the market neutrality of QE measures, would be surprising and relevant enough to reassess green bonds relative to conventional bonds. On the one hand, a reduced risk premium and on the other hand, a lower liquidity premium due to the central bank's greater willingness to buy green bonds would have been conceivable. The new pricing would have satisfied investors with a ceteris paribus lower yield, which is why the prices of green bonds should rise relatively more than those of conventional bonds. Limited credibility of realistic implementation as an interpretive approach can play a central role. Besides, Lagarde's outlined course immediately met with fierce resistance from some of her new colleagues. The president of the German Bundesbank, Jens Weidmann, has publicly stated that he is 'very critical' of any commitment by, for example, 'green QE', precisely because it violates the Table 5: Event Study Results

Research Question 1 Research Question 2

Entry NGFS Lagarde Nomination Lagarde Q&A Lagarde ECON

29.05.2018 02.07.2019 29.08.2019 04.09.2019

t-value sign-test t-value sign-test t-value sign-test t-value sign-test

Specification: All 2.4898*** 0.0085 -0.6973 1.0000 -0.5431 1.0000 0.5483 0.3215 Subsets: ECB-eligible 1.7490** 0.1524 -0.5876 0.9996 -0.3965 0.9975 0.5268 0.0754 No corporations 0.5067 0.7388 -0.6365 1.0000 -0.5308 0.9888 2.2321** 0.0000 Sector-specific:

Fin & Insurance -0.2095 0.2752 -0.1273 0.9058 -0.4637 0.9699 -0.9846 0.9641 Utilities 3.0458*** 0.0173 -0.8150 0.5679 -0.3204 0.9931 0.5051 0.1279 Public 1.7036** 0.3770 -0.6384 0.9682 0.4044 0.1796 1.4310* 0.0318 Other 0.1448 0.9539 2.1549** 0.0539 -1.0143 0.9904 1.8582** 0.0096 Rating-specific: Aaa -2.3586 0.9979 -0.1929 0.9822 -0.4055 0.9739 1.8771** 0.0000 Aa 1.7534** 0.2498 -0.1177 0.9476 -0.5332 0.7825 0.0648 0.6506 A 0.6107 0.1110 -0.2175 0.6778 0.2938 0.8668 0.3892 0.4555 Baa 2.8991*** 0.0001 -0.8298 1.0000 -0.4589 1.0000 -0.2690 0.9799 Ba 2.1320** 0.2500 1.1653 0.1250 -0.2557 0.8750 2.4993*** 0.2500 B 2.1744** 0.5000 -1.7316 1.0000 -2.3568 1.0000 -0.9810 1.0000 Maturity-specific: 1-3 years -2.3586 0.9979 -0.1929 0.9822 -0.4055 0.9739 1.8771** 0.0000 3-5 years 1.7534** 0.2498 -0.1177 0.9476 -0.5332 0.7825 0.0648 0.6506 5-10 years 0.6107 0.1110 -0.2175 0.6778 0.2938 0.8668 0.3892 0.4555 Over 10 years 2.8991*** 0.0001 -0.8298 1.0000 -0.4589 1.0000 -0.2690 0.9799

Note: The table reports test statistics regarding the relevant events for the research question 1 and 2, respectively. The t-value is the

Central banks and climate change The impact of the ECB's willingness to tackle climate change on green bonds