Has the ECB's forward guidance helped (re-)anchor inflation expectations?

(1)

Has the ECB’s Forward Guidance

Helped (Re-)Anchor Inflation

Expectations?

MSc in Economics - Monetary Policy and Banking

Universiteit Van Amsterdam

Riccardo Rossi, 11376066

Supervisor: Dhr. Prof. Dr. Aerdt Houben

July 2017

(2)

Statement of Originality

This document is written by Riccardo Rossi who declares to take full responsibility for the con-tents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study investigates whether the ECB’s forward guidance has been effective in (re-)anchoring inflation expectations derived from Euro Area inflation-linked swaps (ILS). In order to fully grasp the impact of the forward guidance on interest rates, I focus on the overall expansionary stance taken on by the ECB. The results of the empirical analysis show that de-anchoring thrusts have emerged from second-round effects linked to developments in oil prices, news about the Euro Area unemployment rate, and from time-varying volatility affecting inflation swaps. Nonetheless, the signalling effects transmitted by the policy rate guidance and the expanded asset purchase program have steered inflation expectations at all horizons and have helped offset such de-anchoring pressures on longer-term maturities.

(4)

List of Figures

2.1 Euro Area Inflation dynamics 2004-2016 . . . 11

2.2 Contributions to inflation . . . 12

2.3 Movements in market-based inflation expectations . . . 13

2.4 Comparison between the 5y/5y forward rate and the 5-year ahead forecast from the SPF . . . 13

2.5 Path of ECB policy rates 2011-2016 . . . 15

2.6 Timeline of ECB measures . . . 16

3.1 Multicollinearity tests . . . 21

3.2 Correlation matrix of fitted macroeconomic surprises . . . 21

5.1 Volatility clustering in inflation swaps . . . 38

5.2 Conditional volatility in long-term forward rates . . . 40

5.3 Conditional volatility in long-term spot rates . . . 40

1 Macroeconomic news in GDP, HICP, changes in government expenditure, and trade balance with non-EU partners . . . 47

2 Macroeconomic news in industrial confidence, service confidence, consumer confi-dence, and PPI . . . 48

3 Macroeconomic news in industrial production, unemployment rate, and retail sales 48 4 Stationarity test on 5y/5y forward and 10-year spot . . . 49

5 Stationarity test on Brent price . . . 49

6 Stationarity test on VSTOXX index . . . 49

List of Tables

3.1 Dataset: variables and transformations . . . 22

4.1 Impact of macroeconomic surprises on forward inflation rates . . . 25

4.2 Impact of macroeconomic surprises on spot inflation rates . . . 25

4.3 Pass-through from short-term to long-term inflation expectations . . . 26

4.4 Reaction of forward inflation rates to the ECB’s forward guidance . . . 28

4.5 Reaction of spot inflation rates to the ECB’s forward guidance . . . 29

(7)

4.7 Reaction of spot inflation rates to the ECB’s forward guidance in 2013-2016 . . . . 30 4.8 Persistence of forward guidance on forward inflation rates . . . 32 4.9 Persistence of forward guidance on spot inflation rates . . . 33 4.10 Reaction of forward inflation rates to single forward guidance announcements . . . 34 4.11 Reaction of spot inflation rates to single forward guidance announcements . . . 35 4.12 Persistence of single forward guidance announcements on forward inflation rates . . 36 4.13 Persistence of single forward guidance announcements on spot inflation rates . . . 36 5.1 Forward guidance and time-varying volatility in forward inflation rates . . . 41 5.2 Forward guidance and time-varying volatility in spot inflation rates . . . 41 5.3 Forward guidance and time-varying volatility in forward inflation rates in 2013-2016 43 5.4 Forward guidance and time-varying volatility in spot inflation rates in 2013-2016 . 43 1 Evolution of explicit forward guidance announcements . . . 50 2 Important monetary policy announcements . . . 51

(8)

Introduction

Over recent years, many central banks around the world have stepped up the use of unconven-tional monetary policy instruments. Adjustments in central banks’ communication have typically preceded and then supported the adoption of different measures.

Regarding the Euro Area, the ECB introduced an explicit guidance on the expected future path of policy rates in July 2013. At that time, the sustained steepening of the term structure of money market rates and the increase in their volatility had caused a de facto tightening of previous monetary policy measures. By pre-committing to an accommodative stance, the ECB aimed at reducing uncertainty about future developments of monetary policy and enhancing its effectiveness in order to achieve the medium-term objective of an inflation rate below, but close to, 2%.

Against this background, the present work focuses on an aspect which has largely been over-looked by existing literature. That is, the impact of ECB forward guidance announcements on the anchoring of medium to long-term inflation expectations. The aim is to ascertain whether such a type of communication has contributed to (re-)anchor inflation expectations.

The main motivation underlying this inquiry rests on the heightened downside risks to both headline and core inflation experienced by the Euro Area since mid 2013 onwards. As stressed by several studies and by the ECB itself, such risks may have become somewhat entrenched into medium and long-term inflation expectations. As a result, the achievement of the ECB’s objective of price stability over the medium term may have been jeopardized and an unwanted tightening of monetary policy may have emerged due to an economic environment with negative nominal interest rates.

Although there is no general consensus on the best measure of inflation expectations, I rely on fi-nancial market-based measures extracted from inflation-linked swaps (ILS). Broadly speaking, they outperform survey-based measures in terms of frequency and the number of market participants involved. Compared to the main financial marked-based alternative, that is the breakeven rates computed as the difference between nominal and inflation-indexed bonds, they are less affected by distortions coming from liquidity and inflation risk premia.

In order to investigate empirically the impact of forward guidance on inflation anchoring, I adopt two main strategies. First, I run a number of OLS regressions. Then, I present a GARCH framework and compare the ensuing results with those obtained through the OLS estimations.

The contributions to existing literature are several. To our knowledge, the present work repre-sents a first attempt to study the effects of forward guidance on inflation expectations and inflation anchoring in the Euro Area. Similar studies have only been done for the US (e.g. Moessner (2015)). In this regard, I use an extensive sample size with daily observations until 30 December 2016.

I also expand the set of explanatory and control variables in the estimations. In particular, I consider news concerning the main macroeconomic variables for the Euro Area, but also the oil price, the Euro/Dollar exchange rate, the 3-month Euribor rate, and measures of market volatility. Furthermore, the approach to measure forward guidance does not only take account of the guidance on official interest rates, but the overall expansionary monetary policy stance adopted by the ECB. Among others, official announcements concerning other unconventional instruments - e.g. asset purchases, credit easing measures, and negative interest rates - arguably contain important explanatory information that cannot be neglected.

Finally, the use of a GARCH model improves existing empirical literature on inflation anchoring and forward guidance from a methodological point of view.

(9)

The thesis is structured as follows. Chapter 1 provides an overview of theoretical and empirical literature on forward guidance and inflation anchoring. Chapter 2 outlines the dynamics of Euro Area inflation and inflation expectations in recent years and links these to the introduction and development of the ECB’s forward guidance. Chapter 3 provides a detailed description of the dataset. Chapter 4 presents the first estimation methodology with the corresponding results. Chapter 5 presents the second estimation methodology and a comparison with the findings obtained through the OLS regressions. Chapter 6 draws conclusions.

(10)

Chapter 1 A Review of Forward Guidance and Inflation Anchoring

1.1 Goals and channels of forward guidance

In broad terms, forward guidance can be described as ’explicit statements by a central bank about the likely path of future policy rates’1_.

Overall, forward guidance announcements aim at reducing monetary policy uncertainty and providing a higher degree of accommodation when conventional monetary policy instruments are limited or ineffective2_{. Indeed, by directly influencing expectations of future official interest rates,}

central banks are able to affect medium and long-term interest rates even if they are constrained by the effective lower bound (ELB)3_{. The impact on interest rates enacts two transmission channels.}

In the first place, the expectation of low interest rates for an extended period of time and the related flattening of the yield term structure reduce the cost of credit. As a result, consumption and investment spending rise.

Secondly, the fall in interest rates boosts asset prices, thereby increasing the value of stocks and financial wealth (wealth channel).

Finally, forward guidance can also work through the exchange rate channel. Again, it is impor-tant to stress that under forward guidance communications, it is not the reduction of current official interest rates that exerts downward pressures on the exchange rate. Rather, central banks can in-duce a depreciation by promoting an expectation that those rates will remain low, or even lower, for an extended period time. In an environment with negative interest rates, as the Euro Area has been experiencing since June 2014, such an indirect effect on the exchange rate is particularly relevant since it shifts demand from foreign to domestic goods and directly raises import prices. These changes, in turn, should help foster a sustained adjustment in overall inflation towards the 2% target.

1.1.1 Forms of forward guidance

Depending on its characteristics, forward guidance can be specified in different forms. A basic distinction put forward by Campbell et al. (2012) is between ’Odyssean’ and ’Delphic’ communi-cation. The former implies a commitment of the central bank to future monetary policy action. Conversely, Delphic forward guidance involves communicating a mere forecast of macroeconomic performance and a likely path of future monetary policy actions, without explicitly committing to it.

So far, no central bank around the world has resorted to pure Odyssean-type communications concerning the future path of policy rate4_{. Most of them have rather been referred to as of}

1_{Coeuré (2013), ’The Usefulness of Forward Guidance’, speech before the Money Marketers Club of New York} 2_{See, for instance, Filardo and Hofman (2014)}

3_{Existing literature typically uses the term ’zero lower bound’. Since the level of official rates in many Western}

economies have turned into negative territory since several years, we argue that it is more correct to refer to effective lower bound. In the remainder of the work, we will thus stick to this terminology

(11)

’Aesopian-type’. This is a form of Delphic guidance used by central banks on an irregular basis and mostly under special circumstances, such as the effective lower bound.

Within Aesopian communications, four different categories can in turn be distinguished 5_{: i)}

qualitative forward guidance, which has no explicit end-date and does not provide any numerical thresholds or specific economic conditions which would involve a change in the announced path of future policy rates; ii) qualitative forward guidance based on a narrative, which provides qualitative statements about the likely evolution of policy interest rates accompanied by a description of the underlying macroeconomic conditions upon which the central bank’s orientation is based; iii) calendar-based forward guidance, where the central bank refers to a clearly specified time frame; and iv) threshold-based forward guidance, where the central bank links the future path of policy rates to specific quantitative economic thresholds.

The ECB’s forward guidance that was first launched on 4 July 2013 is considered to be quali-tative based on a narrative. The expectation of the future path of policy rates is in fact based on an overall assessment of the macroeconomic performance, and most notably on the development of inflation over the medium term. A further element which has not been stressed by the literature is that this narrative has developed since 2013 to incorporate changes in macroeconomic conditions and an intensification of the ECB’s unconventional policy strategy.

1.1.2 Literature on forward guidance

Early theoretical contributions on the effectiveness of forward guidance refer to the first type of communication, that is, Odyssean forward guidance. In particular, Krugman (1998) and Eggerts-son and Woodford (2003) argue that in a lower bound environment, monetary policy can be still effective in affecting macroeconomic developments by committing to future values of the policy rate once the lower bound ceases to bind. The central bank would be able to reduce long-term interest rates today, which reflect expected future short-term interest rates, by promising to keep the policy interest rate ’lower for longer’, i.e. keep it below levels consistent with its reaction function once the lower bound is no longer binding.

Such a type of policy, however, is affected by a time-inconsistency problem, since the central bank may be tempted to renege ex-post on its past promise when balancing out the costs of higher inflation associated with protracted monetary stimulus. Therefore, credible commitment would turn out to be a necessary condition for forward guidance to be fully effective.

More recent theoretical studies such as Del Negro et al. (2012) and Carlstrom and Paustian (2012) find what has been referred to as the ’forward guidance puzzle’. In other words, standard medium-scale DSGE models tend to overestimate the impact of forward guidance on macroeco-nomic variables and bond yields. This has posed doubts on the way forward guidance is imple-mented in DSGE models and has led to the development of, inter alia, models with conditional forward guidance (De Graeve and Wouters (2014)), heterogeneous agents (Wiederholt (2014)), and imperfect credibility (Bodenstein and Ricardo (2012)).

On the empirical side, several studies have focused on the impact of forward guidance on financial market prices and private market expectations. For instance, Campbell et al. (2012) find that FOMC’s forward guidance has significantly affected US Treasury yields since 2007. Using an event study methodology and controlling for the effect of macroeconomic news, Moessner (2014) finds that FOMC’s forward guidance at the lower bound brought about an increase in US equity prices and a decrease in several risk indicators. As far as the ECB is concerned, Hubert and

5_{ECB (2014), The ECB’s Forward Guidance, in the April Monthly Bulletin}

(12)

Labondance (2016) find that forward guidance announcements decrease the full term structure of private short-term interest rates.

Notwithstanding this extensive literature, few attempts have been conducted so far to study the specific issue concerning the impact of forward guidance on inflation expectations. Among the few of them, Moessner (2015) finds that FOMC policy rate guidance announcements barely affected breakeven inflation rates, thus suggesting that inflation expectations in the US have re-mained well-anchored. For the Euro Area, Coenen and Warne (2014) show through model-based stochastic simulations that a time-based forward guidance may be successful in reducing the pre-vailing downside risks, but it may give rise to upside risks to price stability over the medium term if extended for too long.

1.2 Anchoring of inflation expectations: why central banks care about

it

Under the rational expectations assumption, the public knows the model structure of the econ-omy and the value of all model parameters, including the monetary authority’s reaction function. Accordingly, if the central bank announces an explicit inflation target over the medium/long-term (π∗_{) and is able to credibly commit to it, then medium/long-term inflation expectations should}

remain stable, or ’well-anchored’, at the announced objective (Dovern and Kenny (2017)). In particular, the conditional expectation of private agents based on information available at time t about inflation in the next period reduces to a weighted sum of current inflation and the central bank’s target, that is6

Etπt+1=

αθ 1 − φπ∗+

1 − φ − αθ

1 − φ πt (1.1)

where the parameters capture: (i) θ = weight on inflation in the central bank’s reaction function; (ii) 1-φ = degree of persistence in the inflation process, with 0≤ φ <17_{; (iii) α = sensitivity of}

inflation to excess demand.

Iterating Eq. (1.1) forward up to time t+h we have

Etπt+h= ( αθ 1 − φ)[1 + 1 − φ − αθ 1 − φ + ... + ( 1 − φ − αθ 1 − φ ) h−1_]π∗_{+ (}1 − φ − αθ 1 − φ ) h_π t (1.2)

Then, for h large enough it can be shown that long-term inflation expectations will converge to π∗8:

lim

h→∞Etπt+h= π

∗ _(1.3)

Interpreting equations (1.1) and (1.3) together, short-term inflation expectations will deviate from the inflation target as a result of temporary shocks driving inflation away from that target. However, the longer the forecast horizon, the smaller should be the reaction of expectations, so that long-term expected inflation should not react at all to economic shocks.

6_{Cf. Eq. (2.1) in Dovern and Kenny (2017)}

7_{Specifically, the quantity 1-φ measures the degree of correlation between inflation in periods t and t+1. The}

lower φ, the more inflation at time t+1 is correlated with current inflation

(13)

Accordingly, long-term inflation expectations can be defined as well-anchored if they do not respond to short-term news about macroeconomic variables and remain consistent with the central bank’s inflation objective (Dräger and Lamla (2013)). Crucial for such a relation to hold is that the central bank is fully credible, that is, it is believed by the public to be capable of delivering its inflation target under any circumstance.

Furthermore, as suggested by Dovern and Kenny (2017) this model also implies that long-term inflation uncertainty, which is measured by the long-term forecast error variance (V ar[et+h|It]),

will be equal to the unconditional variance of inflation (V ar[πt+h]). In particular, we will have

that9

V ar[et+h|It] = V ar[πt+h− π∗] = V ar[πt+h] (1.4)

As long as the unconditional variance of actual inflation is stable, we should expect to observe a stable variance in the long-term distribution of expected inflation. Thus, changes in the variance of such a distribution may signal movements in the degree of anchoring of inflation expectations.

In reality, however, inflation expectations may become de-anchored for various reasons. For instance, private agents may not understand well the structure of the economy or its evolution over time. Likewise, they may not understand correctly the central bank’s reaction function or may doubt the central bank’s ability to deliver on its objective (Bernanke (2007)).

As such, central banks are primarily concerned about the anchoring of inflation expectations. On the one hand, well-anchored inflation expectations signal that the central bank’s objective is considered to be credible by the public and thus medium and long-term expectations will not react to incoming data10_{. On the other, since inflation expectations are one of the main drivers of}

current inflation, a better anchoring enhances the central bank’s ability to deliver price stability over longer-term horizons. Therefore, well-anchored inflation expectations ultimately imply that inflation will be less sensitive to the level of economic activity - i.e. the Phillips curve will become flatter - and less responsive to supply shocks, e.g. oil shocks (Mishkin (2007)).

Monitoring the degree of anchoring becomes even more important when the interest rates are in negative territory or close to the effective lower bound. In these circumstances, a decrease of medium and long-term inflation expectations amounts to an increase in the real interest rate. To the extent that the central bank is providing an accommodative stance, this would result in an unwanted monetary tightening. This is exactly the environment that has been prevailing in the Euro Area since 2013 and that highlighted the need for the ECB to resort to unconventional instruments to counter unfavourable developments in inflation and inflation expectations.

9_{Cf. Eq. (2.2) in Dovern and Kenny (2017)}

10_{ECB (2011), Inflation Expectations in the Euro Area: a Review of Recent Developments, in the February}

Monthly Bulletin

(14)

Chapter 2 The ECB’s Forward Guidance: catching up with inflation

dynamics

2.1 Developments of Inflation and Inflation Expectations in the Euro

Area

Over the period 2012-2016, both headline and core inflation have declined, and most worryingly, inflation persistence beneath 2% has increased. The latter effect has raised concerns about the anchoring of inflation expectations.

2.1.1 Causes of low inflation

Understanding the causes underlying inflation dynamics is pivotal for central banks to undertake the proper policy response. Typically, central banks react to sustained demand shocks, whereas they are less concerned about one-off supply shocks, unless these get embedded into agents’ expec-tations (so-called ’second-round’ effects) (Ciccarelli and Osbat (2017)).

As depicted in Fig. (2.1), both measures of Euro Area inflation - headline and core - have embarked on a steadily declining path since the last quarter of 2012.

Figure 2.1: Euro Area Inflation dynamics 2004-2016 Source: Eurostat

The unprecedented and prolonged fall in oil prices in more recent years has placed strong down-ward pressure on headline inflation (Fig. 2.2), while the negative contribution of demand factors has increasingly added to the downward trend of both inflation measures. By contrast, spread shocks1_{have turned from an initial disinflationary impact to a consistent reflationary contribution}

1_{Using a structural BVAR model and imposing restrictions on the spread between long and short-term rate,}

Ciccarelli and Osbat (2017) identify the spread shock as a decrease in the long-term rate relative to the short-term rate. Such a decrease in the spread is mainly associated with ECB expansionary monetary policies, but it is also affected by other factors (e.g. repercussions from the sovereign debt crisis)

(15)

in 2015-2016, mostly as a result of the ECB’s accommodative stance.

The overall picture suggests that the subdued development of core inflation has proved to be quite persistent and second-round effects stemming from the drop in oil prices may have in fact emerged.

Figure 2.2: Contributions to inflation Source: Ciccarelli and Osbat (2017)

2.1.2 Decline in inflation expectations

Since mid-2013, the prolonged period of low inflation has been accompanied by a worrisome decline in inflation expectations. By focusing on financial market-based measures of inflation expectations extracted from inflation-linked swaps (ILS), we see that this decline has characterized both the short-end and the long-end of the term curve. For instance, the 2-year spot rate reached a trough below 0% in early 2015. The 3-year spot and the 2-year forward rate followed a similar decreasing path (Fig. (2.3), panel (a)).

The high volatility of energy and commodities prices, especially the large drop in the oil price, have seemingly played a major role in the dynamics of these short-term market-based measures. However, longer-term inflation expectations have also displayed a declining path, which became particularly pronounced in years 2014-2015. Three common measures of medium and long-term inflation expectations from ILS i.e. the 7year, the 10year, and the 5year 5year forward rates -steadily fell throughout 2014 and troughed at levels below (or far below) the ECB’s medium term objective in 2015 (Fig. (2.3), panel (b)).

(16)

(a) Short-term inflation rates from inflation swaps (b) Medium and long-term forward inflation rates

Figure 2.3: Movements in market-based inflation expectations Source: Datastream, Author’s calculations

A comparison between the quarterly average of the 5-year 5-year forward rate extracted from ILS and the average 5-year ahead forecast provided by the ECB’s Survey of Professional Forecasters (SPF) shows that the latter slightly declined between 2011 and 2015 but remained nonetheless more stable than the former, and did not diverge from the target of an inflation rate below, but close to, 2% over the medium term (Fig. 2.4).

Figure 2.4: Comparison between the 5y/5y forward rate and the 5-year ahead forecast from the SPF

Source: Datastream, ECB

All these patterns raise questions about what drivers have been at play. In particular, the question arises whether the fall in market-based measures of inflation expectations over longer horizons only reflected the impact of liquidity and other risk premia, or rather signalled a threat of de-anchoring resulting from disinflationary, or even deflationary, risks and second-round effects. In fact, the ECB has looked through these developments by taking on an increasingly watchful stance. It first introduced forward guidance and then refined this guidance by stepping up its unconventional monetary policy orientation.

(17)

2.2 Introduction and developments of forward guidance

On 4 July 2013, for the very first time in its brief history, the ECB adopted an explicit forward guidance on policy rates. In the official introductory statement, President Draghi stated:

”The Governing Council expects the key ECB interest rates to remain at present or lower levels for an extended period of time”

In the next sentence he specified the reasons underlying this decision:

”This expectation is based on the overall subdued outlook for inflation extending into the medium term, given the broad-based weakness in the real economy and subdued monetary dynamics”

Three elements clearly emerge. First, the qualitative nature of the policy communication, which does not involve any exact end-date or specific quantitative thresholds. Second, the narrative ac-companying the statement, which strongly emphasizes the link to the weak outlook for inflation over the medium-term as well as the underlying weaknesses in real economic activity and in mon-etary dynamics. Third, the easing bias incorporated in the statement, through which the ECB explicitly takes account of the possibility that policy rates could be lowered further in the future. Nonetheless, President Draghi stressed at that time that the risks to the outlook for price developments were expected to be broadly balanced over the medium term and this justified the consideration of inflation expectations as still being well-anchored.

The wording of the forward guidance was then reinforced and reiterated in January and Febru-ary 2014, when the ECB began to point out that a protracted period of low inflation may be underway:

” We may experience a prolonged period of low inflation. [...] Accordingly, we firmly reiterate our forward guidance that we continue to expect the key ECB interest rates to remain at present or lower levels for an extended period of time”

On 5 June 2014, all policy rates were lowered to a historical minimum, with the deposit facility turning to negative territory (-0.10%) (Fig. 2.5). Maintaining well-anchored inflation expectations became crucial because a potential downside de-anchoring would amount to an unwanted monetary tightening resulting from an increase in the real interest rate. Accordingly, the key policy interest rates were expected to remain ’at present levels’ for an extended period of time.

In addition, the ECB stood prepared, ’[...] if required, to act swiftly with further monetary policy easing’. Such a wording, though not strictly related to the forward guidance on policy rates, featured an enhancement of the Odyssean orientation in the ECB’s narrative and paved the way to the unprecedented step taken some months later. On 22 January 2015, an expanded asset purchase programme (APP), which also included purchases of sovereign bonds, was launched. At this time, the threat of inflation expectations de-anchoring stemming from the ongoing prolonged period of low inflation and second-round effects linked to the fall in oil prices was admittedly taken into account by the ECB.

Importantly, it was stressed that the impact on, inter alia, inflation expectations was expected to come through the so-called ’signalling channel’ of the quantitative easing (QE) programme. Specifically, the forward guidance on the APP contained a Delphic component concerning the horizon within which the program was intended - initially, until end-September 2016 - and a state-contingency clause justifying its continuation, which has not yet changed since then - that is, ’[...]

(18)

until the Governing Council sees a sustained adjustment in the path of inflation consistent with its inflation target’2. Accordingly, the programme was intended to complement and reinforce the forward guidance on interest rates.

Figure 2.5: Path of ECB policy rates 2011-2016 Source: ECB

Notwithstanding the launch of QE and a first positive assessment by the ECB on the effective-ness of the programme in stabilizing inflation expectations, downside risks to inflation projections extending over the medium-term heightened again in late 2015 and early 2016. In January 2016, for instance, Draghi explicitly admitted that ’all inflation expectations measures have declined [...] and their correlation with current inflation has increased’. This higher correlation indicated that inflation expectations had likely experienced a departure from the 2% objective over the medium-term, which signalled a de-anchoring of longer-term maturities.

This led to a reaffirmation, in January, and then an enhancement of the forward guidance in March 2016. The announcement was also accompanied by an extension of the APP up to €80 billion per month, a further reduction down to -0.40% of the deposit facility rate (Fig. 2.5), and the launch of a new series of targeted longer-term refinancing operations (TLTRO II) with a four-year maturity starting in June.

The guidance on asset purchases specified that the latter were intended to run ’until the end of March 2017, or beyond, if necessary’, whereas the guidance on policy rates re-introduced the downward bias and extended the expectation of present or lower levels of interest rates ’[...]well past the horizon of (our) net asset purchases’.

Finally, in December 2016 the horizon of the APP has been further extended ’until December 2017, or beyond, if necessary’ and an additional Odyssean clarification as to the preparedness of the ECB to increase the size and/or the duration of the program, under certain conditions3_{, has}

been added4_.

As we have hinted in the above description, important announcements concerning other aspects of the ECB’s expansionary stance - such as the ones concerning the reduction of official rates, the continuation of TLTROs, and the initially more limited asset purchase program encompassing covered bonds and asset-backed securities - have accompanied the development of forward guidance.

2_{The detailed wording of the announcements encompassing the forward guidance on interest rates and those}

associated with the APP is provided in Appendix C, Table (1)

3_{’[...] If the outlook becomes less favourable, or if financial conditions become inconsistent with further progress}

towards a sustained adjustment of the path of inflation’

4_{Coeuré (2017), ’Central Bank Communication in a Low Interest Rate Environment’, Speech at an Event}

(19)

All of them served to strengthen the message that the ECB was willing, if needed, to take all the necessary steps to achieve its inflation target over the medium term.

Similarly, unconventional monetary policy instruments had already been introduced since the outbreak of the European sovereign debt crisis and during the global financial crisis. Since many of those measures may as well have produced a movement in financial-market based measures of expected inflation, we investigate their potential impact when performing our empirical analysis.

2.2.1 The impact of expansionary monetary policy announcements

The ECB began adopting unconventional instruments in the wake of the financial crisis when it en-acted important packages to enhance liquidity provision and reduce financial market segmentation5

(see Fig. 2.6).

Then, in the midst of the Euro Area sovereign debt crisis (2011-2012) it introduced a series of 3-year VLTROs (very long-term refinancing operations) and launched the Outright Monetary Transactions (OMTs) plan with the aim of getting rid of the redenomination risk entrenched in unsustainably high sovereign yield spreads and breaking the so-called ’bank-sovereign nexus’6_.

Although the dynamics of inflation were not yet a major concern for the ECB in this period, the above measures may have contributed to underpin a firm anchoring of inflation expectations. As such, they should be taken into account when estimating the regressions.

In addition, the gradual progression towards a negative territory by the deposit facility rate and the launch of the two series of TLTROs may convey significant explanatory information over the period 2013-2016.

Both sets of measures were explicitly directed to enhance the transmission mechanism of mon-etary policy by encouraging banks to increase lending to the real economy, rather than holding excess reserves at the ECB. Their adoption ultimately signalled the willingness and readiness by the ECB to act, thereby improving monetary policy effectiveness and underpinning the message conveyed through the explicit forward guidance on interest rates.

Figure 2.6: Timeline of ECB measures

5_{For instance, it expanded the assets that could be used as collateral in credit operations with the Eurosystem}

and introduced the fixed-rate allotment policy in all refinancing operations

6_{Incidentally, the decision on the OMTs plan followed the speech by Draghi in London in July 2012 announcing}

the well-known ’whatever it takes’

(20)

Chapter 3 Data Description

3.1 Measuring inflation expectations

This section discusses advantages and disadvantages of the two broad categories conventionally used to gauge inflation expectations - i.e. financial market-based and survey-based measures - and explain why we focus on inflation-linked swaps (ILS) for our empirical analysis.

In the final section we turn to a detailed description of the dataset used for the model regressions.

3.1.1 Market-based versus survey indicators

The most common source of survey-based inflation forecasting in the Euro Area is the Survey of Professional Forecasters (SPF), which is directly conducted by the ECB on a quarterly basis

1_{. This specific survey is based on responses from financial and other institutions settled in the}

European Union, and has had an average actual participation of about 60 respondents - out of an active panel of approximately 75 professional forecasters - since the start of the Euro. The forecasting covers the short-term - 1-year and 2-years ahead - and stretches up to 5 years ahead, thus capturing a medium-term horizon.

The SPF also includes a quantitative assessment of uncertainty surrounding the point forecasts, which is reflected in the reported probability distributions of future inflation outcomes falling within given ranges. Furthermore, the standard deviation of the point forecasts of all respondents, referred to as ’disagreement’, is provided2_.

Survey-based measures like the ECB’s SPF contain valuable information and can be used as a benchmark when assessing developments of inflation expectations at different horizons. For instance, the provision of individual density functions allows to build up the whole probability distribution of inflation forecasts at a specified horizon, thereby enabling the detection of per-ceived tail risks. Likewise, the broad set of respondents implied by the survey reflects an array of expectations which goes beyond those provided by financial markets.

However, these measures have a number of drawbacks that make them unsuitable to gauge the impact of monetary policy decisions, which presumably happens in a very short timespan. In the first place, information from surveys is only available at low-frequency and with a lag. In addition, the panel of forecasters is generally limited. Another problem is ’inertia’3_{. In other}

words, the timing of survey-based forecasts may not be aligned with the forecast schedule of the respondents, so that their reported forecast may not be revised and may not reflect the latest available macroeconomic data. Finally, the incentives governing their answers may be biased and skewed towards the ECB’s objective, particularly for longer-term forecasts.

1_{Other surveys of inflation expectations for the Euro Area, such as for instance the European Commission’s}

consumer survey or that of the Euro Zone Barometer, mostly focus on short or very short forecasting horizons and/or are qualitative in nature. As such, they are not particularly suitable to gauge circumstances relevant to the monetary policy stance, which is rather geared towards medium and long-term time spans

2_{See also ECB (2011), Inflation Expectations in the Euro Area: a Review of Recent Developments, in the February}

Monthly Bulletin, for a general discussion

(21)

Given these limitations, financial market-based measures of inflation expectations are often used as an alternative in empirical research. The two most common indicators for this purpose are the break-even inflation rates (BEI) and inflation rates extracted from ILS. Applying the Fisher equation, the former are calculated as the yield spread between nominal and inflation-linked bonds of equal maturity4_.

Both types of measures have significant advantages compared to survey indicators. First of all, they are very timely and are available at high-frequency (daily or intraday). Secondly, they are determined on the basis of financial transactions among numerous market participants that have clear incentives for accurate pricing. However, market prices also include two main sources of distortions: the liquidity premium and inflation risk premium.

3.1.2 Inflation-linked swaps and break-even inflation rates

Measures of inflation expectations based on actively traded financial instruments may be contam-inated by liquidity and inflation risk premia and as such, they may not reflect a pure indicator of expected inflation (Hördahl (2009); Gürkaynak et al. (2010); Galati et al. (2009)). Although it is difficult to quantify the impact of both premia, the use of ILS seems to produce better estimates of inflation expectations than BEI rates.

Broadly speaking, investors may demand a liquidity premium to be compensated for the diffi-culty with which a security can be sold or exchanged. Such a premium is different across maturities - longer tenors are typically more affected than shorter ones - and is deemed to lower the extracted inflation expectations. This is particularly relevant for BEI rates, since inflation-linked bonds are typically traded less frequently than nominal bonds (Nautz et al. (2015)).

On the other hand, ILS are less prone than BEI rates to liquidity distortions. First, only one market is involved, so that possible asymmetries in liquidity between nominal bonds and inflation-linked bonds are not a major concern. In addition, liquidity effects are limited by the structure of the swap contract itself, which implies that only residual cash flows - rather than the whole notional value of the contract - are exchanged at maturity. Among all ILS rates, the 5-year 5-year ahead forward is considered to include the smallest liquidity premium5_.

However, when financial markets experience periods of turbulence and high volatility, even inflation swaps can be distorted to some extent by illiquidity premia. Therefore, in our regressions we include liquidity control variables - specifically, the VSTOXX index - to take account of residual liquidity effects and volatility.

The other major distortion that may affect financial instruments linked to inflation is the infla-tion risk premium. Such a premium compensates investors for the risk of unexpected changes in inflation over the period of security holding. In particular, it is argued that the premium increases with maturity and pushes the market-based measure of expectations upward6_{. Nonetheless,}

infla-tion risk premia that are included in ILS rates are considered to be limited in size and variability and are most likely to affect short and very long horizons7_.

Beyond being less distorted by both types of premia (liquidity and inflation risk), ILS rates present other advantages compared to BEI rates. In fact, inflation swaps are available on a daily basis for a wide range of maturities. In the Euro Area, maturities go from 1-year to 10-year with an annual spacing, and from 10-year up to 30-year with a five-year spacing. This wide availability

4_{Euro Area BEI rates are typically extracted from French nominal bonds and HICP ex-tobacco linked bonds} 5_{See for instance ECB (2011)}

6_{See e.g. Gürkaynak et al. (2010) and Garcia and Werner (2010)} 7_{See Garcia and Werner (2010) and ECB (2011)}

(22)

avoids the problem of estimating nominal and real term structures that emerges when using break-even rates. As a result, the fixed rates from ILS allow to monitor movements in both short-term and long-term market-based expectations. Likewise, they are less affected by HICP seasonality8_.

Furthermore, Euro Area BEI rates also incorporate an element of credit risk which distorts the signal on the extracted inflation expectations measure. By contrast, inflation swaps are traded on central-clearing platforms, so that the impact of both credit and counterparty risks is limited.

Overall, ILS seem to reflect a better measure of expected inflation. Therefore, in our empirical analysis we are going to use the fixed rates from inflation swaps at different horizons to extract a measure of inflation expectations and study the impact of the ECB’s forward guidance.

3.2 Description of the dataset

Most of the data for our estimations comes from two sources: Datastream and Bloomberg. Specif-ically, data for ILS rates, the oil price, and the VSTOXX index are collected from Datastream, whereas data concerning macroeconomic variables, economic indicators, and the Euribor rate are drawn from Bloomberg. Table (3.1) presents a summary of the dataset along with a description of the transformation of each variable.

3.2.1 Inflation swap data: forward and spot rates

We use daily data for mid-quotes ILS rates from Datastream. The sample period goes from 3 January 2005 until 30 December 2016. For the estimation, we focus on both spot rates and forward rates. The implied forward inflation rate between maturity n and m, with m>n, is computed as

fn,m= m−ns (1 + s m)m

(1 + sn)n

−1 (3.1)

where smand sn are the ILS spot rates with m and n time to maturity, respectively.

In economic terms, the difference between spot and forward rates is subtle, yet significant. On the one hand, spot rates reflect today’s expectation of what inflation will be in the future. On the other hand, forward rates capture expectations about future inflation at a future date. Thus, for instance, the 10-year spot rate measures today’s expectation of what inflation will be in ten years. By contrast, the 10-year forward rate reflects the rate of inflation which is expected to prevail in ten years nine years from today.

Although both types of rates ultimately measure future expected inflation, forward rates are more stable than spot rates and filter out some noise that may influence the latter over the whole spectrum of maturities9_{. Since the ECB’s price stability objective should be delivered over the}

medium and long-term, we assess the impact of the ECB’s forward guidance at different horizons but we gauge its effectiveness in steering inflation expectations by focusing on the reaction of longer tenors.

In particular, the most relevant measure for the transmission channel of monetary policy is the 5-year forward rate five years ahead. Such a forward is preferable for several reasons. First of all, it is based on liquid inflation swaps and is difficult to manipulate because of the size of the market. Secondly, it is strongly correlated with other measures of inflation expectations. Finally,

8_{ECB (2011)}

(23)

it is highly recognizable, since both the ECB and the FED as well as empirical literature use it to investigate the degree of anchoring stemming from financial markets.

3.2.2 Macroeconomic news variables: calculation, selection, and

multi-collinearity tests

In line with existing empirical literature, we control for the effect of macroeconomic news on our measures of inflation expectations to determine the degree of anchoring. We collect data on macroeconomic announcements from Bloomberg. Realizations of macroeconomic variables are released on a monthly or quarterly basis. For each announcement, Bloomberg provides the median survey forecast Mt, the actual realized value At, the prior value Pt, and the revised value Rt10.

Since the predicted component of the news variable should already be priced in on the release date, it should not affect inflation expectations (Gürkaynak et al. (2010)). On the contrary, if inflation expectations are not well-anchored, we should expect them to react to the news compo-nentof a macroeconomic announcement. A reaction of short-term expected inflation may not be surprising to the extent that it reflects the impact of transitory economic dynamics.

Therefore, we calculate the surprises of the macroeconomic announcement data as the difference between the actual release value and the median forecast and we normalize them by their standard deviation:

St=

At− Mt

σt (3.2)

Surprises are realized on the date the value of the underlying variable is published by official sources. On non-publication dates, surprises are set to zero11_{. The sample period goes from 3}

January 2005 to 30 December 2016.

The initial set of macroeconomic data includes the following thirteen variables for the Euro Area: year-on-year HICP inflation, quarterly GDP, industrial production (ind_prod), unemploy-ment rate (unemrate), retail sales (retsales), changes in governunemploy-ment expenditure (gvmtexp), trade balance with non-Euro Area partners (trade_bal), producer price index (PPI), industrial confidence index (ind_conf ), business confidence index (bus_conf ), service confidence index (service_conf ), economic confidence index (ec_conf ), and consumer confidence index (cons_conf ).

However, since the inclusion of too many variables with similar characteristics may increase the degree of multicollinearity, which in turn would inflate standard errors and make the estimated coefficients unstable, we perform multiple variance inflation factor (VIF) tests and check which variables may create such a problem. The first VIF test on the set of all of the thirteen variables is reported in Fig. (3.1), panel (a). The rule of thumb is to avoid using variables that have a VIF factor exceeding four. As we can see, all variables are in fact below four, but some confidence indicators - specifically, industrial confidence, economic confidence, and business confidence - are above three. Although no serious problem of multicollinearity is detected, we follow the European Commission’s standards and we exclude the economic and business confidence indicators to improve the test and remove redundant variables. Hence, we perform a new VIF test on the resulting set of macroeconomic variables. The results of this second test, which are reported in Fig. (3.1), panel (b), show an improvement of both the mean VIF factor and the single VIF factors of the three remaining confidence indicators. Furthermore, a cross-checking with the correlation matrix from the model regression confirms that all correlations between the variables in the second set are lower

10_{The prior value refers to the previous month’s indicator, whereas the revised value is added after the publication}

of the actual variable to incorporate a revision of the previous value

11_{The realizations of the news variables are shown in Appendix A}

(24)

or significantly lower than |0.5|, which is the threshold above which highly collinear relationships are detected (see Fig. (3.2)). Therefore, we keep all three remaining confidence indicators, that is: the industrial and service confidence indexes for the business sector, and the consumer confidence index for the consumer sector.

The final set of macroeconomic news consists of eleven variables for the Euro Area and is reported in Table (3.1).

(a) Set including economic and business confidence (b) Set excluding economic and business confidence

Figure 3.1: Multicollinearity tests

(25)

3.2.3 Oil price, exchange rate, Euribor rate and volatility indicators

The literature on event studies regarding inflation anchoring typically focuses on the impact of macroeconomic news only. We improve on this strand of research by adding further explanatory and control variables in order to reduce the risk of omitted variable bias.

First of all, we consider the oil price to be a pivotal variable. On the one hand, we expect short-term inflation expectations to be affected by oil price developments, especially over the recent years of high volatility in oil prices. On the other, we would interpret a reaction of longer-term expectations as a clear sign of de-anchoring. Accordingly, we use daily changes in the price of Brent oil, which is traded in Europe. The series is taken from Datastream.

Similarly, fluctuations in the exchange rate may have played a significant role in the dynamics of actual and expected inflation in recent years12_{. Therefore, we include daily changes in the}

Euro/Dollar exchange rate. Data are collected from Bloomberg.

Furthermore, we include the 3-month Euribor rate as a proxy for conventional monetary policy surprises. The Euribor should not respond to the anticipated component of central bank announce-ments. However, the change in the Euribor rate on the day after a monetary policy announcement should capture reactions to (conventional) monetary surprises. Data are taken from Bloomberg.

In Section 3.2 we discussed how financial market-based measures of inflation expectations may be distorted by liquidity and inflation risk premia. Although we argued that inflation-linked swaps are allegedly less affected than break-even rates by these premia, we control for financial market uncertainty by including daily changes in the VSTOXX indicator for the Euro Area13_{. This index}

is commonly used by empirical studies as a proxy for swings in financial volatility and market liquidity in European markets14_{. The series is collected from Datastream.}

The sample period for all the above-mentioned variables goes from 3 January 2005 until 30 December 201615_.

Sample period: 03/01/2005-30/12/2016

Variable Source Transformations

1. ILS rates Datastream daily difference

2. y/y HICP Bloomberg Normalized At-Mt

3. Quarterly GDP Bloomberg Normalized At-Mt

4. Industrial production Bloomberg Normalized At-Mt

5. Unemployment rate Bloomberg Normalized At-Mt

6. Retail sales Bloomberg Normalized At-Mt

7. Changes in government expenditure Bloomberg Normalized At-Mt

8. Trade balance with non-EU partners Bloomberg Normalized At-Mt

9. PPI Bloomberg Normalized At-Mt

10. Industrial confidence index Bloomberg Normalized At-Mt

11. Service confidence index Bloomberg Normalized At-Mt

12. Consumer confidence index Bloomberg Normalized At-Mt

13. Brent oil Datastream daily difference

14. Eur/Dol Exchange rate Bloomberg daily difference

15. 3-month Euribor rate Bloomerg daily difference

16. VSTOXX index Datastream daily difference

Table 3.1: Dataset: variables and transformations

12_{See Section 2.1}

13_{The VSTOXX is a measure of implied volatility for options on the Eurostoxx equity index} 14_{See e.g. Nautz et al. (2015) and Galati et al. (2016)}

15_{All variables refer to the Euro Area}

(26)

Chapter 4 Estimation technique I: Multiple OLS

4.1 Testing for de-anchoring

In this chapter, we are going to run a series of OLS regressions to investigate the effects that the ECB’s forward guidance may have had on inflation expectations and, specifically, on the degree of anchoring1_.

We divide the analysis in two parts. We first look at possible signs of de-anchoring stemming from the impact of macroeconomic surprises, changes in the oil price, and changes in the (nominal) exchange rate on inflation expectations.

In the second part, we begin by running two sets of regressions on forward rates and spot rates on the whole sample to check whether forward guidance has significantly impacted inflation expectations, and if yes, at which horizons. We then repeat the analysis on a subsample of more recent years (2013-2016) where the ECB’s unconventional stance has been stepped up. Finally, we carry out a robustness check and we present results for a longer 5-day window event study on the same regressions.

4.1.1 Impact of macroeconomic surprises on forward and spot ILS rates

We regress daily changes in ILS forward rates with maturities m= 2, 5, 7, and 10 years ahead, daily changes in the forward rate 5-year 5 years ahead, and daily changes in spot rates with maturities m= 2, 3, 5, 7, and 10 years ahead2 _{on the set of macroeconomic surprises listed in}

Table (3.1), daily changes in the price of Brent (∆Brentt), and daily changes in the Euro/Dollar

exchange rate (∆Et). Daily changes in the 3-month Euribor rate (∆Euribort) and VSTOXX index

(∆V ST OXXt) are included as control variables for conventional monetary policy surprises and

market volatility, respectively. Thus, the full regression is the following:

∆zm t= α +

n

X

i=1

βisuri,t+ γ1∆Brentt+ γ2∆Et+ γ3∆Euribort+ γ4∆V ST OXXt+ t (4.1)

where ∆zm

t = zmt− zmt−1 represents the daily change in the zero-coupon or forward inflation

rate m years ahead, and n=11 is the number of macroeconomic news included3_.

In all regressions we use Newey-West standard errors to take account of heteroskedasticity and serial correlation in the errors terms, which are likely to affect financial market-based instruments such as ILS rates4_{. The results are shown in Tables (4.1) and (4.2) for forward and spot rates,}

re-1_{Before running the OLS estimations, we check whether our dependent variables - namely spot and forward rates}

at different maturities - and some of our explanatory and control variables - specifically, the oil price, the euro/dollar exchange rate, the 3-month Euribor rate, and the VSTOXX index - are stationary. Since both ILS rates and the covariates are found to be non-stationary, we take first differences in all variables to remove the unit root. Some results are presented in Appendix B

2_{We use a subset of ILS forward and spot rates that capture maturities at short, medium, and long-term horizons} 3_{See Table (3.1) in Chapter 3}

(27)

spectively. In both tables, we only present the macroeconomic news variables showing a significant coefficient and those that are useful for a comparison with the results of the regressions in next sections. If inflation expectations have remained firmly anchored throughout the sample period - 3 January 2005 - 30 December 2016 -, medium to long-term spot and forward rates should be relatively unresponsive to short-term shocks as captured by our macroeconomic news variables, changes in the oil price, and changes in the exchange rate.

At first glance, the results seem to suggest a somewhat sound degree of anchoring over the full sample period. Indeed, we can see that both forward and spot rates are only slightly affected by some macroeconomic surprises. Most notably, macroeconomic news concerning HICP inflation and changes in the Euro/Dollar exchange rate do not influence either types of rates. Furthermore, the fact the the R2 _{is low in all regressions does not necessarily mean a lack of explanatory power}

of the model and, above all, does not reflect a multicollinearity problem, as we already highlighted in Chapter 3. Conversely, as suggested by Galati et al. (2009), the fact that most of the estimated coefficients on the set of macroeconomic news variables are insignificant can be a sign of well-anchored inflation expectations.

On the other hand, however, daily changes in the oil price have a statistically significant impact on every spot rate, which may signal the emergence of second-round effects. Likewise, the signif-icance of the coefficient on the VSTOXX index at all maturities and the consistent negative sign it takes on, suggests that market uncertainty and liquidity concerns might have been transmitted also to longer-term inflation expectations, thus potentially weakening the anchor to the ECB’s target.

Therefore, following a standard approach used in empirical literature5_{, we investigate this issue}

further and we check for pass-through effects from short-term to long-term expectations6_{. We}

use changes in the 10-year forward and the 5-year 5 years ahead forward as measures of long-term inflation expectations, while we pick up changes in the 2-year (∆2y) and 3-year (∆3y) spot inflation rates as indicators of short-term expected inflation. Again, we take account of market volatility and liquidity risk by including daily changes in the VSTOXX as control variable.

The equation is the following:

∆ftLT = a + b1∆stST + b2∆V ST OXXt+ t (4.2)

where ∆ftLT denotes changes in the two long-term forward rates, whereas ∆stST represents

changes in the two measures of short-term expected inflation.

If inflation expectations are well-anchored, short-term developments in expected inflation should not have a significant impact on the two measures of long-run inflation expectations and the corresponding coefficients should be statistically insignificant.

In this regard, the results reported in Table (4.3) show that changes in the 2-year spot rate significantly affect at the 0.1% level both long-term ILS forward rates alike, whereas changes in the 3-year spot only affects the 10-year forward, albeit at the 0.1% level. These results indicate that some pass-through effects may indeed have been at play. Analysis of the full regression including monetary policy announcements should help pinpoint the source of these effects.

5_{See e.g. Jochmann et al. (2010)}

6_{Notice that in the present study we cannot detect pass-through effects from actual inflation to inflation}

expec-tations. In fact, we do not have access to daily data on HICP inflation, which would be needed for this type of analysis. Hence, we only investigate pass-though effects from short-term to long-term inflation expectations

(28)

Table 4.1: Impact of macroeconomic surprises on forward inflation rates (1) (2) (3) (4) (5) ∆5y5yf wd ∆10f wd ∆7f wd ∆5f wd ∆2f wd ind confidence -0.00267 -0.00484 -0.00624 0.0157∗ -0.00483 (0.238) (0.200) (0.191) (0.018) (0.410) trade balance -0.00146 -0.00606 -0.00193 0.00797∗ -0.00367 (0.401) (0.184) (0.544) (0.016) (0.151) HICP 0.00000468 -0.00106 0.00102 0.000940 -0.000165 (0.998) (0.756) (0.774) (0.797) (0.953) GDP 0.00374 0.00899∗∗ 0.00871 -0.00271 0.000603 (0.086) (0.004) (0.078) (0.717) (0.902) ∆E 0.0727 0.222 0.0707 -0.123 0.143 (0.374) (0.070) (0.636) (0.550) (0.429) ∆Euribor -0.0566 0.0508 -0.168 0.0520 0.0814 (0.440) (0.590) (0.094) (0.699) (0.521) ∆Brent 0.000803 0.000426 0.00108 0.000536 0.00273∗∗∗ (0.051) (0.504) (0.143) (0.555) (0.001) ∆VSTOXX -0.00176∗∗∗ -0.00139∗ -0.00252∗∗∗ -0.00272∗∗∗ -0.00378∗∗∗ (0.000) (0.012) (0.000) (0.001) (0.000) Observations 2428 2428 2428 2428 2428 R2 0.032 0.014 0.018 0.010 0.042

p-values in parentheses. Significance: ∗_{p < 0.05,}∗∗ _{p < 0.01,}∗∗∗ _{p < 0.001.}

Sample period: 03/01/2005-30/12/2016.

Table 4.2: Impact of macroeconomic surprises on spot inflation rates

(1) (2) (3) (4) (5)

∆10y ∆7y ∆5y ∆3y ∆2y

cons confidence -0.00118 -0.00205 -0.00377 -0.00600 -0.00620∗ (0.518) (0.330) (0.160) (0.064) (0.038) HICP -0.000294 -0.000620 -0.000306 -0.000317 0.000196 (0.808) (0.633) (0.820) (0.826) (0.919) PPI -0.00295 -0.00345 -0.00546∗ -0.00489 -0.00310 (0.093) (0.124) (0.039) (0.086) (0.330) ∆E 0.0748 0.0619 0.0902 0.144 0.196 (0.322) (0.454) (0.346) (0.195) (0.097) ∆Euribor 0.0460 0.0654 0.138 0.198 0.222∗ (0.464) (0.394) (0.143) (0.066) (0.045) ∆Brent 0.00193∗∗∗ 0.00238∗∗∗ 0.00281∗∗∗ 0.00407∗∗∗ 0.00473∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) ∆VSTOXX -0.00222∗∗∗ -0.00253∗∗∗ -0.00257∗∗∗ -0.00232∗∗∗ -0.00258∗∗∗ (0.000) (0.000) (0.000) (0.000) (0.000) Observations 2428 2428 2428 2428 2428 R2 0.088 0.096 0.084 0.079 0.096

p-values in parentheses. Significance: ∗p < 0.05,∗∗ p < 0.01,∗∗∗p < 0.001. Sample period: 03/01/2005-30/12/2016.

(29)

Table 4.3: Pass-through from short-term to long-term inflation expectations (1) (2) (3) (4) ∆5y5yfwd ∆5y5yfwd ∆10fwd ∆10fwd ∆2y 0.134∗∗∗ _0.131∗∗∗ (0.000) (0.000) ∆3y 0.0338 0.163∗∗∗ (0.381) (0.000) ∆VSTOXX -0.00143∗∗∗ _-0.00185∗∗∗ _-0.00116∗ _-0.00110∗ (0.001) (0.000) (0.040) (0.049) Observations 2504 2504 2504 2504 R2 0.062 0.024 0.016 0.021

p-values in parentheses. Significance: ∗_{p < 0.05,}∗∗_{p < 0.01,}∗∗∗_{p < 0.001.}

Sample period: 03/01/2005-30/12/2016.

4.2 Augmented OLS specifications: gauging the impact of

forward guidance

To determine the reactions of inflation expectations based on ILS to the ECB’s forward guidance, we develop two sets of event study regressions. In the first set, we stick to the 1-day window followed in Section 4.1 and we present results for two samples: the whole sample - from 3 January 2005 to 30 December 2016 - and a subsample from 1 January 2013 to 30 December 2016. That specific subsample is chosen to look at the dynamics of ILS rates over the more recent period where deflation risks were detected and the ECB stepped up its unconventional monetary policy stance. In the second set, we present results for a longer 5-day window event study on the subsample period 2013-2016. This serves as robustness test to check whether the reactions of ILS rates to the ECB’s monetary policy and the other explanatory variables show persistence or not.

4.2.1 Whole sample regressions

Our approach to measure forward guidance is the following. First, we capture the surprise com-ponent of ECB policy rate guidance announcements. Following Moessner (2015), we do so by considering only those announcements where a new word was introduced or those signalling a change in the policy guidance. We point out two dates: (i) 4 July 2013; and (ii) 9 January 2014. On the first date, the ECB introduced the explicit forward guidance for the first time. On the second one, it ’firmly reiterated’ it. Accordingly, we construct a single dummy variable (dumf g,t)

that takes a value of 1 on those dates, and zero otherwise7_.

At the same time, we construct another dummy (dumf grep,t) that takes on a value of 1 since

the date on which a previous wording was repeated until the next announcement with new explicit forward guidance on interest rates, and zero otherwise8_{. The choice of making this dummy more}

continuous controls for possible lags in the transmission of policy rate guidance communications. Furthermore, the introduction and development of the expanded asset purchase program (APP) incorporate additional forward guidance components. In this regard, we acknowledge that

move-7_{In the last section of this chapter we split the forward guidance announcements in two separate dummies to}

look at the explanatory power of each single announcement

8_{The periods where the dummy variable on repeated forward guidance takes a value of 1 are the following: 1}

August 2013 - 13 January 2014, 6 February 2014 - 4 June 2014, 3 July 2014 - 3 September 2014, and 21 April 2016 - 7 December 2016

Has the ECB's forward guidance helped (re-)anchor inflation expectations?