1
On the announcement effect of ECB unconventional measures on
inflation expectations.
Bachelor Thesis June, 2016
By: Sander Louis Mulders Student number: 10203877 Supervisor: Mr Gabriele Ciminelli MPhil
Since the global financial turmoil the European Union has faced periods of dysfunctional financial markets, low growth prospects and declining expected inflation. Jeopardizing the ECB’s mandate to maintain stable (expected) inflation, close but below two percent. The ECB announced various unconventional and Quantitative Easing (QE) programs. This thesis quantifies the announcement effect of the unconventional and quantitative easing policies of the ECB on euro-zone expected inflation. Using an OLS regression we find evidence that in particular the more recent QE policies did have a significant announcement effect on expected inflation over the medium term.
Keywords: Unconventional Monetary Policy – Quantitative easing - (expected) Inflation – Monetary transmission mechanism - Eurozone
2
Content.
1. Introduction………..3
2. (Un)conventional monetary policy, GDP and inflation expectation: theory and
implementation………5
2.1 Does money matter; Monetarists versus Keynesians……….5
2.2 Recent theory and evidence from Money, GDP growth and (expected) inflation…..7
2.3 Defining monetary transmission channels……….9
2.3.1 Monetary policy: a structural model………9
2.3.2 Quantitative Easing: a structural model……….11
2.4 (Un)conventional monetary policy after the financial crisis of 2008………13
2.4.1 Conventional instruments of the European Central Bank……….13
2.4.2 Unconventional programs implemented by the ECB since 2008…………..14
3 Data……….19
3.1 Dependent variable……….19
3.1.1 Inflation linked swap rates………19
3.1.2 Inflation swaps, a financial market based measure for expected inflation..20
3.2 Independent Variable………..21
3.3 Control Variables……….22
4 Hypotheses and method………..24
4.1 Hypotheses………24
4.2 Method………25
5 Results and discussion……….26
6 Conclusion………28
7 Reference list……….29
3
1. Introduction
With the advent of the subprime financial crisis in the summer of 2007, central banks around the world began a very aggressive easing of monetary policy. As a result central banks lowered their official interest rates to close to zero percent. However these measures did not provide enough stimulus to the real economy. The euro area experienced a strong economic decline after the U.S. had gone into recession. With interest rates close to zero and risks of deflation as a result of the severe recession, the European Central Bank (ECB) decided to implement unconventional policies in order to maintain price stability over time and protect the economy from further recession (Mishkin F., K. Matthews and M. Giuliodori, 2013).
The ECB wishes to maintain price stability over time. In order to reach this objective the ECB uses conventional and unconventional monetary tools. It sets monetary stance and tries to control the expected inflation over the medium term. Financial agents, households, institutions and companies throughout the world change their decisions as a consequence of unexpected changes in monetary policy. Besides that a stable and positive expected inflation is relevant for consumer spending. Positive (expected) inflation stimulates consumer spending today rather than tomorrow. It implicates that goods and commodities will have a higher price in the future. The opposite holds if there is expected deflation.
The structural model of Keynesian economists offer monetary transmission channels through which unconventional monetary policy directly influences inflation expectations. Joyce M., Tong M. and Woods R.(2011) elaborate on the confidence channel: Unconventional monetary policy may directly boost consumer confidence as a consequence of an improved economic outlook. This will lead to more willingness to spend and therefore future inflationary pressures are expected. This structural model is one of many channels that provides clear insight in ‘how’ the announcement of unconventional monetary policy directly affects inflation expectations. But their remains uncertainty to what extend the direct impact is on inflation expectations for different time horizons.
The contribution of this thesis is to quantify the effect of the announcements of unconventional and QE-policies on the expected inflation proxied by swap rates. Data from January 2008 till mid-june 2016 is used to measure these effects. Within this time period the ECB decided to implement several unconventional monetary programs. These measures were implemented to restore the normal functioning in different markets in the euro zone, and provide stimulus to the real economy. I want to measure the effect of the announcement of those unconventional programs over the short to medium term inflation expectation. For this event study I opt for a methodology that has been
4
used by Moessner (2013) to quantify these effects. The methodology of Moesner (2013) is a linear OLS-regression using robust standard errors and binary independent variables.
The study done in this thesis shows that the most recent QE-program announcements from the ECB had a positive (inflation rise) and significant effect on inflation expectation over the medium term (3 to 5 year). The same QE announcements didn’t show significance on a shorter horizon (1 year). I also found that the covered bond purchase program, securities market program and covered bond purchase program 2 on average didn’t have a significant impact over the short to medium term expected inflation. Despite this outcome, these programs were a response to the global financial crisis and the euro sovereign debt crisis.
The outline of this thesis is as follows: Section 2 discusses the academic debate and a theoretical framework that provides a background for a better understanding of the analysis. Section three presents the data. Sections four states the hypotheses and presents the method. Section five presents and discusses the results. In Section six I will draw a conclusion of this thesis. The appendix discusses the assumptions of the regression model.
5
2. (Un)conventional monetary policy, GDP and inflation expectation:
theory and implementation.
In this section I will discuss the necessary theoretical background for an adequate understanding of the analysis that is performed in this thesis. The first Subsection discusses the academic debate and evidence provided on the effects of monetary policy on output and inflation. The second Subsection describes the relation between money GDP growth and inflation and provides more recent evidence. The third Subsection defines a theoretical framework that is used throughout the theses. Subsection 4 describes in more detail (Un)conventional monetary policy instruments and programs decided by the ECB.
2.1 Does money matter; Monetarists versus Keynesians
Early Keynesians focused on structural model evidence1 to explain the relation between money supply growth and output. The general consensus for Keynesian post war theory stated that monetary policy does not matter when it comes to movements in aggregate output. This was supported by mainly three arguments. First, during the great depression low nominal interest rate didn’t seem to affect investment spending. Monetary policy could not explain why an easing policy didn’t affect aggregate demand. So Keynesians concluded that money does not matter. Second, early empirical studies didn’t find a relation between a change in nominal interest rates and investment spending. Third, surveys of business people at that time showed that their decision on new investments were not influenced by the interest rate. If there was any relation between monetary policy and output then this was mainly explained through a structural model that defined the channels (often interest rate) through which it affects output.
Monetarists led by Milton Friedman opposed to this view. Friedman &Schwartz’s (1963) showed in their research: ‘a monetary history of the United states’ that monetary policy was rather contractionary during the great depression. Besides that they indicated that investment does not necessarily rely on nominal interest rates, but rather on real interest rates. Real interest rates were high during the great depression because of the high (expected) inflation. Monetarists stated that structural evidence based model is as good as the model it is based on. Besides that, they suggested
1
Structuralmodel evidence examines the effect of changes in the money supply on economic activity by building a structural model, a description of how the economy operates using a collection of equations that describe the behaviour of firms and consumers in many sectors of the economy. These equations then show how monetary policy affects aggregate output and spending. (Mishkin F., K. Matthews and M. Giuliodori. 2013).
6
that the interest rate effect on investment is just one of a variety of channels that affects aggregate demand and there is need for a different model.
Therefore the monetarists (Friedman) came up with a reduced-form model2 that suggested a strong relation between money and economic output for the short run. This evidence was based on three important pillars. Timing evidence, statistical evidence and historical evidence. The first revealed that money supply growth fluctuates relative to the economic cycle. However, this happens with long and variable lags that might lead to reverse causation. Does money supply growth affect the bussines cycle or is it the opposite way? Statistical evidence examined the correlation between money growth and output. The friedman-Meiselman(1963) evidence proved that money is stronger correlated with output than investment and government spending of the Keynesian components framework3. Although critics of this study found that the results were the opposites if the ‘way’ of determining the autonomous expenditure were constructed more carefully. Later study on the aptness of the different ways of determination of the autonomous expenditure was indifferent about a higher correlation for either Money or Consumption and Investment with output. The last pillar is mainly based on the historical evidence found in ‘a monetary history of the United states’ of Fieldman and Schwartz (1963). Several episodes in this book occur, in which changes in the money supply seem to be exogenous events. For these events the post hoc, ergo propter hoc4 condition is more likely to be valid and historical evidence provides strong reason that money does matter for output.
The evidence provided by the monetarists made many Keynesian economists think again about the effect of money on output. Economic science couldn’t give a clear cut about the situation and therefore many economist switched to an intermediate position toward this debate. Since then research has diverged into two directions. One that focusses on structural form evidence and one that focusses on the reduced form evidence. (Mishkin F., K. Matthews and M. Giuliodori. 2013).
This theses applies mainly the structural model theories for explanation of effect of monetary policy towards GDP. The following subsection discusses the link between growth in GDP towards (expected) inflation and more recent evidence.
2
Reduced-form evidence analyses the effect of changes in M on Y in general terms. It does not describe in specific ways how or through which channels money effects output. It rather examines whether changes in money are highly correlated with changes in output
3
Kenysian components framework: Y= C + I + G + NX
4 After event Y, event X happens, therefore, because of Y another event X happens. This is a classic fallacy
reasoning as Y could have been a consequence of another event Z. Then Y is endogeneous and Y and X are both consequences of event Z. But if Y is exogeneous it is more likely that this principle holds.
7
2.2 Recent theory and evidence from money, GDP growth and (expected) inflation.
This sections discusses more recent theory and empirical evidence of money, GDP and Inflation. It concludes that there is recent empirical evidence that a change in monetary policy affects GDP and inflation with certain time lags.
According to simple Keynesian economic theories as the DAD-SAS model (under rational expectations5) a monetary expansion will lead, over the short term, to a lower interest rate. Which causes more lending and increases investments and output
(Point A1). In the long-run this growth in GDP will causes the prices to rise (inflation) and the economy will fall back in equilibrium at the same level of GDP but at a higher inflation equilibrium (point A3).
This long-term mechanism also corresponds with the quantity theory of money QTM6. This reduced form evidence model states that if the CB prints money this will result in GDP growth and inflation in the long-run if the velocity of the money remains constant.
The monetary policy committee (2005) of the bank of England complements on the
theoretical relations in the DAS-SAS model. Total domestic expenditure plus the balance of trade in goods and services reflects aggregate demand in the economy, and is equal to GDP at market prices. The difference between actual GDP and potential GDP is known as the ‘output gap’ when there is a positive output gap, a high level of aggregate demand has taken actual output to a level above its sustainable levels. Booms in the economy that take the level of output significantly above its potential level are usually followed by a pick-up of inflation.
The ECB(2000) states in their monthly bulletin of July that the speed in which changes in spending (increase in output) translate into price pressures depends on the degree of nominal price rigidities and the flexibility of the economy more generally. In normal circumstances, increases in aggregate demand beyond the potential output tend to create bottlenecks in the economy, which fuel inflationary pressures.
5 The theory of rational expectations (i.e. = π) assumes that expected inflation predicted by the model will
be equivalent to the future inflation.
6
GDP*P = V*M
8
Graph 2
Graph 3 Graph 4
However the OECD (2000)provided existing econometric evidence for the euro area that while nominal price rigidities have been comparable to
the U.S. On average the speed of adjustment in real wages to unemployment and productivity changes in the EU has been much slower. The ECB concluded that the formation of inflation expectation by households, firms, wage setters and financial market participation form an essential element in the transmission and effectiveness of monetary policy
To conclude the ECB(2000) and the BOE(2005) also provide their forecast models in which a
change in the interest rate is visible in a graph . An interst rate rise first effects GDP and second (expected) inflation. The ECB state that an unexpected temporary rise in the short term interest rate of 25 basis points tends to be followed by a temporary fall in output after two quarters. Prices are more slow-moving and start to fall significantly below zero after six quarters (graph 2).
The BOE shows similar effects for an increase of 100 basis points of the interest rate (graph 3 & 4). These graphs
indicates that the empirical test of this theses is likely to find no significance on expected inflation within a
9
2.3 Defining monetary transmission channels
This subsection provides two structural frameworks to define more specifically the variety of channels through which monetary policy and QE affects GDP and so expected inflation. The provided frameworks are partly used for section 2.4 on elaborating the possible effects of the monetary programs on inflation expectations.
2.3.1 Monetary policy (MP) transmission: a structural model
We opt for a structural model evidence framework for defining the transmission channels of monetary policy (see figure below) that was part of the course of Monetary & Fiscal Policy at the University of Amsterdam and can also be found in the book Economics of Money, Banking, and Financial Markets from Mishkin F., K. Matthews and M. Giuliodori. (2013). They define three main categories for 9 channels through which monetary policy affects GDP. Figure 1 provides a equation overview of the different channels.
According to new Keynesian and monetarist theory, the traditional interest rate channel can affect not only investment by firms. Also housing and consumer durable expenditure are affected. Real interest rates (rather than nominal) affect agent’s decisions. Lower real interest rate cause more investment. Which in turn leads to more output.
There are three asset price channels. First, the exchange rate effect. In open economies with flexible exchange rate, expansionary MP becomes more effective due to exchange rate changes. The exchange rate depreciation leads to a net export increase shifting the IS curve to the right. Which causes output (Y) to rise. A second asset price channel is tobin’s q effect. Tobin’s q is the market value of the firm divided by the replacements of cost of capital. If MP affects
stock price (PS), new plants and capital costs changes relative to market value. A rise in the market value of the firm relative to the replacement cost of capital makes it possible to issue new equity for the firm to finance their investment. Besides stock prices, tobin’s q can also be applied to housing prices. MP affects stock prices through roughly two ways. First, expansionary MP leads to more spending as the public finds that it has more money than it wants. The extra spending takes also place at the stock market, increasing the demand for stocks and raising their prices. Second, a lower nominal interest rate raises the present value of the expected return of the stocks. So stock prices
Graph 5 Expansionary monetary policy in IS-LM-BP Model
10
rise. The third asset price channel is the wealth effect. If net financial wealth is affected by MP, consumption changes. Based on Modigliani’s life cycle hypothesis 7of consumption: people smooth consumption based on lifetime resources. So not only today’s income, but also net wealth affects consumption spending. Large components of wealth are based on equities and bonds but also housing or land. More spending increases output.
There are five credit view channels. First, the bank lending effect. Expansionary MP increases bank reserves and bank deposits. This will increase the quantity of the available loans at the bank. This increase in loans will cause investment and (probably consumer) spending to rise. Which in turn leads to more output. Another credit view channel is the balance sheet effect. Expansionary MP decreases adverse selection (AS) and moral hazard (MH) problems for lending institutions. The potential losses from AS and MH are smaller as firms do have more collateral value on the balance sheet for their loans. More lending will occur which leads to more investment and output. A third credit view channel is the cash flow effect. Expansionary MP lowers the short term interest rate. This causes firms to become more liquid. The potential losses from AS and MH are smaller when firms do have more liquidity. It ensures lenders that the firm is able to pay their bill. As a consequence more lending will occur which leads to more investment and output. A fourth credit view channel is the unanticipated price level effect. The real value of liabilities will become lower if there is an unanticipated rise in the price level. Potential losses from AS and MH are lower as firm do have a larger net worth value. This increases lending which in turn increases investment and output. The last credit view channel is the household liquidity effect. The other credit view channels should also be applied to households. Through these channels the likelihood of financial distress is lower. What causes more durable and housing expenditure. This increases in turn the output.
7
The life-cycle hypothesis is an economic theory that concerns the spending and saving habits of households over the course of a lifetime. The concept was developed by Franco Modigliani. It assumes that people base consumption on a constant percentage of their anticipated life income. An example that supports this concept is that people save money for their retirement.
11
2.2.2 QE transmission mechanism; a structural model
Besides the more general described transmission framework for monetary policy we opt for another framework to define the transmission channels through which the QE-programs have effected GDP and inflation. The concept of this framework is formed by Joyce M., Tong M. and Woods R.,(2011) for explaining the possible effects of QE for the Bank of England (BoE). It defines three main categories and 5 channels.
First channel is the Confidence channel. QE may directly boost consumer confidence as a consequence of an improved economic outlook. This will lead to more willingness to spend and therefore future inflationary pressures are expected. This channel is also likely responsible for the majority of a possible announcement effect. As it causes a direct effect.
There are three asset price channels. The firs is called Policy signaling. The CB may signal low future policy rates (lowering long-term rates) and determination to avoid deflation (e ) leading to lower real rates. This in turn stimulates investments and boosts GDP and creates inflationary pressures. Another asset price channel of QE is portfolio rebalancing. By buying assets from non-bank in exchange of money, the CB stimulates rebalancing of portfolio from money to better
substitutes. These firms may want to change their acquired money for corporate bonds and equities
Figure 1
Channel Structural model explanation of Monetary policy (MP) on output (Y) Traditional interest rate
Exchange rate Tobin's Q Wealth Bank lending Balance sheet Cash flow
unanticipated price level Household liquidity effects
MP e i r I Y MP i r E NX Y MP i r PS q I Y MP i r (PS , PH , PL WC Y
MP reserves deposits loan supply I and C Y MP P
S Net- worth of firms (collateral value) AS & MH lending to firms Y MP i cash flow (liquidity) AS & MH lending I Y
MP i unanticipated P real value of liabilities AS & MH lendingI Y MP i P
S value of financial assets likelihood of financial distress consumer durable and housing expenditureY
12
at firms. Henceforth a chain of firms buying new assets will get into equilibrium at a higher asset price. Higher asset prices mean lower yields, which stimulates spending. Besides that, there is a wealth effect of raising asset prices (see former subsection for wealth effect) The last asset price channel is the market liquidity channel. By targeting QE on specific financial debt or security
markets. The ECB can increase liquidity in those markets and as such increase asset prices through a reduction in the liquidity premium. The outcome of this channel may only persevere if the CB is conducting a QE policy in the market.
To conclude with the bank lending channel. QE-programs lead to a higher level of reserves and expansion of new deposits which ‘may’ lead to new loans (similar to bank lending channel in other framework).
13
2.4 (Un)conventional monetary policy measures after the financial crisis of 2008
This sections unveils the conventional and unconventional instruments and programs that the ECB uses. This subsection tries to give an insight into the current possibilities of the ECB to achieve its goal to maintain price stability.
2.4.1 Conventional instruments of the ECB
As mandated in article 2 of the statute of the ECB. The ECB strives for price stability in the Euro-zone. Three conventional instruments are used to achieve that goal.
i) Standing facilities ii) Open market operations
iii) Minimum reserve requirements.
The ECB provides two standing facilities, i.e. the lending facility and the deposit rate. Lending facilities are operated to provide (against eligible collateral) overnight reserves to banks at a lending rate, which is typically set above the official rate. Deposit facilities are offered for absorbing
overnight reserves of banks wishing to deposit at the central bank at a deposit rate, which is set below the official rate. (Mishkin F., K. Matthews and M. Giuliodori, 2013)
Besides the standing facilities. The ECB affects money market interest rates by providing more (or less) liquidity to banks if it wants to decrease (increase) interest rates. It allocates an amount of liquidity that allows banks to fulfil their liquidity needs at a price that is in line with the ECB policy intentions. To manage liquidity in the money market and steer short-term interest rates, it uses open market operations, i.e. it buys (or sells) financial assets. If assets are bought from (sold to) a bank, the reserves of that bank at the central bank increase (decrease). These operations are carried out by the National Central Banks (NCBs) in the euro area. The most important open market operations of the ECB are the main refinancing operations (MROs) and longer-term refinancing operations (LTROs) (Pattipeilohy, Van den End, Tabbae, Frost and J. de Haan, 2013). LTROs and MROs can be categorized in two different operations. Outright purchases/sales by the ECB in a secondary market, which permanently changes the amount of reserves held by the market. The second category refers to temporary open market operations. In which the ECB engages in (reverse) repurchase agreement (repo’s). The ECB buys (sells) securities from the certain sector with the agreement that the seller (buyer) will repurchase (resell) the securities at the date of maturity. In such a way the ECB is able to create loans for a specific market.
14
The last instrument of the central bank is to define the amount of required reserves that credit institutions should deposit at the NCB. An increase (decrease) of the required reserve ratio by the ECB means that there is a rise (decline) in demand by the market for reserves and leads to an increase (decrease) in the euro overnight index average (EONIA).
In special economic circumstances; an economic crisis or the MRR reaching the zero lower bound. The ECB might face the problem that its conventional instruments are no longer effective. In these circumstances the ECB can decide to use unconventional monetary instruments. So far this has only been done by the ECB if the economic circumstances also jeopardize its objective of price stability.
2.4.2 Unconventional instruments of, and programs implemented by the ECB since 2008
Pattipeilohy, Van den End, Tabbae, Frost and J. de Haan (2013) describe a framework that comprises three elements for unconventional monetary policy of the ECB
i) Large-scale liquidity support to banks;
ii) Forward guidance of ultra-low policy rates over extended policy horizons; iii) Large-scale financial market interventions, in particular huge asset purchases. All the measures taken by the central bank are to be categorized by these three elements.
It is useful for the purpose of understanding the different actions to distinguish certain phases and responses of the ECB’s monetary policy. These phases and actions are listed below. The design and possible transmission channels through which the APP have effected expected inflation are discussed in the next subsection. A detailed description of the other ECB programs is not provided as the focus of this thesis is on the APP.
15
A. Extended Asset Purchase Program (APP)
This program comprises several subprograms that have been implemented and announced on different dates. Nowadays these programs are bundled together and called the extended Asset Purchase Program (APP). It’s consists of the following programs:
A1. Covered Bond Purchase Program 3 (CBPP 3) A2. Asset Backed Securities Purchase Program (ABSPP) A3. Public Sector Purchase Program (PSPP)
A4. Corporate Sector Purchase Program (CSPP) Tabel 1
Overview implemented monetary programs
Phase Global financial crisis september 2008
Response i Enhanced Credit Support program 10-2008* **** ii coverd bond purchase program (CBPP) 7-5-2009 60 Billion
Phase Euro sovereign debt crisis may 2010
response i Securities Market Program (SMP) 10-5-2010 220 Billion***
Phase Sovereign debt crisis and banking sector strains intensified in summer of 2011
response i Activily re-implementation of (SMP) 4-8-2011 220 Billion*** ii Covered bond purchase program 2 (CBPP2) 6-10-2011 40 billion iii Long term refinance operation (LTRO) 20-12-2011 489 Billion iv Long term refinance operation (LTRO) 28-2-2012 529 Billion
phase Forecast of low growth and decling inflation in 2014 for 2015 by ECB
Response Extended asset purchase program** 1084 Billion i Covered bond purchase program 3 (CBPP3) 4-9-2014 183 Billion ii Asset backed securities program (ABSP) 4-9-2014 20 Billion iii Public sector purchase program (PSPP) 22-1-2015 875 Billion iv Corporate sector purchase program (CSPP) 10-3-2016 6 Billion v targeted long term refinance operation (TLTRO) 5-6-2014
approx 400 billion
*Not a specific announcement date, for dat reason left out of research ** Name for collection of CBPP3, ABSP, PSPP and CSPP
*** Amount is shared by first announcement and re-implementation **** Exact amount unknown, rather credit-easing than monetary expansion
Announcement
16
Previous non-standard measures were mainly aimed at redressing impairments in the monetary policy transmission mechanism and fostering a regular pass-through of the monetary policy stance. Their implications for the ECB’s balance sheet were accommodated in a merely passive way to satisfy the liquidity demand created by banks. In contrast, with the new measures
implemented since June 2014, the Governing Council is more actively steering the size of the ECB’s balance sheet towards much higher levels in order to avoid the risks of too prolonged a period of low inflation in a situation where policy rates have reached their effective lower bound. (Vitorio Constâncio, 2015).
From March 2015 until March 2016 the average monthly puchases done by the ECB under the APP set around €60 billion euro. Since April 2016 this has increased on average to an amount of 80 billion euro8
A.1 Covered Bond Purchase Program 3 & Asset Backed Securities Purchase Program
On the 4th of September 2014 the ECB (Governing Council of the ECB, 2015) decided to implement the CBPP3 and the ABSPP. These programs were implemented as a consequence of the decline of the inflation rate since mid-2013 with fragile growth prospects. With economic growth rates stabilizing at and below 1% in 2014 the prospects seemed to be deteriorate again for 2015. Forecasts indicated a declining and weak economic growth accompanied by declining inflation of 3% at the end of 2011 to -0.2% in December 2014. This combination of low growth and declining inflation was particularly suggestive for a shortfall in aggregate demand and called for a further easing of monetary policy (Vitor Constâncio,2015)
Asset backed Securities is a financial security that is backed by a loan, receivables or other assets than real estate. The ECB made a list of all securities that were eligible for the ABSPP by buying. According to the ECB the ABSPP further enhances the transmission of monetary policy, facilitates credit provision to the euro area economy and generates positive spillovers to other markets. As a result, it eases the ECB’s monetary policy stance, contributing to a return of inflation rates to levels closer to 2%. The ABSPP also helps banks to diversify funding sources and stimulates the issuance of new securities. Asset-backed securities can help banks to fulfil their main role: providing credit to the real economy. For instance, securitising loans and selling them can provide banks with the necessary funds to provide new lending to the real economy.
8
17
A.2 Public Sector Purchase Program (PSPP)
On 22nd of January 2015, the Governing Council of the ECB decided that the CBPP3 and ABSPP should be expanded to include a secondary markets public sector asset purchase program. Under the PSPP the NCBs, in proportions reflecting their respective shares in the ECB's capital key9, and the ECB may purchase outright eligible marketable debt securities from eligible counterparties on the secondary markets. This decision was taken as part of the single monetary policy in view of a number of factors that have materially increased the downside risk to the medium-term outlook on price developments, thus jeopardizing the achievement of the ECB's primary objective of
maintaining price stability. (Governing Council ECB, 2015)
A.3 Corporate Sector Purchase Program (CSPP) (Amount of Euro still needed to define) On 10 March 2016 the Governing Council of the ECB(2016) decided to further expand the asset purchase programs and initiate the CSPP as part of the single monetary policy and in pursuit of its price stability objective. This decision was taken in order to further strengthen the pass-through of the Eurosystem’s asset purchases to the financing conditions of the real economy, and in order to provide, in conjunction with the other non-standard monetary policy measures in place, further monetary policy accommodation and contribute to a return of inflation rates to levels below, but close to, 2 % over the medium term.
Extended APP transmission to expected inflation
The extended APP concerns mostly QE-programs and for that reason all the transmission channels of Joyce M., Tong M. and Woods R.,(2011) are mainly relevant for this program. I want to mention the confidence channel in particular. The amount of money that is spent to the extended APP is big. For that reason there is a higher chance that this program will cause a larger direct impact on expected inflation. Besides that the program will also have larger effect during the adjustment phase because of it’s enormous amount of money.
9
The capital of the ECB comes from NCBs of all EU Member States and amounts to €10,825,007,069.61. The NCBs’ shares in this capital are calculated using a key which reflects the respective country’s share in the total population and gross domestic product of the EU. These two determinants have equal weighting. The ECB adjusts the shares every five years and whenever a new country joins the EU. The adjustment is made on the basis of data provided by the European Commission. For more information see
18
By the time that the first APP program was launched the interest rate could not lower much further. For that reason the general framework, which oftens operates through the interest rate, is not a very solid framework to deduct effects on expected inflation. There is only one remark on this assumption. As it is likely that the confidence channel will directly increase inflation expectation. The real interest rate might become lower during the adjustment phase. According to monetarists theory the QE-programs might nevertheless leak through several transmission channels of the general framework through that way.
19
3. Data
In this Section I will elaborate on the data that I gathered for the regression analysis that is performed in this thesis.
3.1 Dependent variable
As I want to measure the impact of the announcements of the ECB programs over the short till medium-term expected inflation proxied by inflations swaps. I will use a two day interval10 change in 1-year, 3- year and 5-year ahead zero coupon inflation swap rates from the second of January 2008 till the fourteenth of June 2016 as the dependent variable. Evidence of a 2 day interval change is more relevant for our research question. One day interval might assume that there is a transitory effect while a two day interval measures a durable effect on inflation expectations. Within the described timeframe I am able to measure the effects of the monetary programs after the financial crisis. The swap rates are obtained from ICAP11 via DataStream.
3.1.1 inflation linked swap rates
An inflation swap is a contract in which an exchange of cash flows that allows investors to reduce or increase their exposure to the risk of a decline in the Harmonised Index of Consumer Prices (HICP), also known as inflation. In this thesis, data of zero coupon inflation swaps are used. The zero coupon inflation swap is an inflation derivative; a cash flow that is linked to the rate of inflation is traded for a cash flow with a fixed interest rate. The cash flows are paid on the date of maturity. Hence, the risk of the inflation is hedged by a person who buys (inflation seller) the inflation swap on the market from the seller of the inflation swap (inflation buyer), who is willing to take the risk of inflation versus paying a fixed interest rate to the buyer of the swap. The income stream for the inflation buyer is the nominal amount multiplied by increase/decrease in inflation from start date till maturity of the contract.
(1)
10
with with on the 2nd of January.
11
ICAP is a large market operator and provider of post-trade risk mitigation and information services. For more information visit ICAP.com
N= Nominal value of the contract Mt = time to maturity in years T = Start date of contract.
20
The income stream for the inflation seller is the nominal amount of the contract multiplied by fixed interest rate till maturity.
(2)
Standard no-arbitrage pricing theory implies that the following identity holds
) (3)
So if I solve for R I find:
-1 (4)
In the market, ‘R’ is quoted such that the time to maturity zero coupon inflation swap has zero value today. It’s equivalent to the rates quoted in the nominal swap market.
From the fourth identity I can derive the relation of the Swap market rates to the expected inflation. The is in this sense an estimation of the expected inflation by the market. To be
precise, equilibrium is provided through the anticipation of by the agents who sell inflation
rate swaps (inflation buyer). According to the fourth identity a rise in will lead to a higher
fixed rate return. This matches with the theory that an inflation buyer (seller of an inflation swap contract) wants to be compensated for the anticipated rise in inflation by a rise in the swap rate ‘R’.
3.1.2 Inflation swaps, a financial market based measure for expected inflation
The reliability of the equilibrium rate as an indicator for inflation swaps depends also on the functioning of the market for swap rates in the Euro-area. The amount of trading and liquidity are important indicators in making a judgement if swap rates are a valuable proxy for expected inflation.
Deacon, Derry, and Mirfendereski (2004) and JP Morgan (2008) state that in a number of countries, bonds or interest rate swaps that are linked to some measure of domestic inflation are
N = Nominal value of the contract Mt = time to maturity in years T = Start date of contract. R = interest rate per year
21
actively traded. These instruments can be combined with nominal bonds or nominal interest rate swaps to back out financial markets inflation expectations. The main advantage of this type of measure is that, given its high frequency, it allows examining more formally changes in the behavior of expectations over a relatively short horizon. Galati, Poelhekke, Zhou , 2009 continue on that part and found that inflation swaps are actively traded—the most liquid ones among inflation-indexed products in the euro-area. Market commentary indicated that the monthly trading volume of ten-year euro-area inflation swaps averaged around six billion in 2007 (JP Morgan 2008). To conclude the governing council of the ECB (February 2011) 12holds the inflation linked swaps in high esteem when it comes to measuring the expected inflation, especially on the short run. They complement: In the specific case of the euro area, the inflation-linked swap market has grown rapidly since 2003, reflecting the increasing demand for inflation-linked instruments and the relatively limited supply of index-linked bonds in the euro area. The euro area is currently likely to be the most developed market for inflation-linked swaps in the world, which makes its information particularly suitable for monitoring developments in inflation expectations, most notably for short and medium-term horizons. Given this empirical evidence by several institutions and researches; inflation swaps are a useful proxy for expected inflation in the Eurozone.
3.2 independent variables.
The announcement of the different monetary programs of the ECB are proxied by two independent dummy variables. Refers to the announcement dates of all ECB
programs except the APP. These programs were implemented as a response on the global financial crisis and the sovereign debt crisis. It concerns the CBPP, SMP and CBPP2. Refers to the
announcement dates of the extended Asset Purchase Program. These programs were implemented as a response to a forecast in low growth and declining inflation for 2015. It concerns the CBPP3, ABSP, PSPP, CSPP. On the dates that the ECB announced for the first time that they would
implement a new program the dummy variable takes a value of ‘1’ on the other dates the variable takes a value of ‘0’. I refer to the appendix for more information about the use of dummy-variables in a regression analysis. The dates of all programs are given in table 1 of section 2.4.
The distinction between these two dummy-variables has one reason that is supported by two arguments. I want to measure the effect of the single APP program. Henceforth, I expect the APP
12
There are two papers of the governing council of the ECB from 2011. So this one is written down with ‘february’ to see the difference in the reference list.
22
program to have a stronger announcement effect than the earlier announced ECB programs. First, the ECB programs were less severe (in amount of billions) than the recently announced APP. Second, they were mainly implemented to fix impairments in the market and restore normal functioning of the MTM. While the APP is addressed to create inflationary pressures over the medium term.
3.2 Control variables.
An event study measures the effect of an event on the value of a dependent variable. In our case the zero coupon inflation swaps. In reality there are many other unexpected news variables (Macro- economic news, credit raiting news, news political instability, natural disasters etc.) that also influence the rates of zero coupon inflation swaps for the euro. In order to find valuable results for the effect of independent variable. I need to incorporate control variables in such a way that my sample regression model resembles the true population model (i.e. reality ). In practice I can’t build such an exact model. But I can create a regression model that approximates this population model. For that reason I incorporate 12 Macro-economic surprise variables that are thought to be the most important announcements influencing the European financial markets. As such I try to control for the potential effect of other Macro-economic news that explains the rate of zero coupon inflation swaps besides my independent variable.
Five Eurozone, four German, two United States (US) and one Chinese surprise indicator are included in the regression. A detailed description of the variables is given in the table on the next page. Only the surprise effect is relevant as surprises are not expected by the market. Hence, creating a change in the swap rates. They are founded on the calendar of forexfactory.com13. By subtracting forecasted data from the actual data I can calculate the surprise effect of the variable. To conclude a dummy variable is included to control for the effect of a meeting of the governing council of the ECB on changes in the dependent variable.
13 The Forexfactory.com is designed to provide high-quality information that traders can apply in their pursuit of profits.
On the financial calendar people can find economic surprise releases of relevant financial data. Forex Factory is operated with complete independence, and is not owned by or affiliated with any forex broker or company.
23
Overview control variables (Surprise)
Coëfficient Variable description
Meeting of the governing Council of the ECB.
Nearly every month the Governing Council of the ECB gathers to make decisions with regard to monetary policy. Afterwards they read up a statement of the decisions that have been made during the meeting. This variable is the only one that does not act as a surprise variable.
Change in the Main Refinancing Rate.
The governing council of the ECB decides whether the main refinance rate changes. Surprise changes in the MRR are incorporated in the regression. Unemployment rates Eurozone Percentage of the total work force that is unemployed in the eurozone and
actively seeking employment during the previous month. Monthly data. German Information und
forschungstelle Bussines Climate Survey
Level of a composite index based on surveyed manufacturers, builders, wholesalers, and retailers; This survey is respected due to its large sample size and historic correlation with German and wider Eurozone economic conditions. It tends to create a hefty market impact upon release.
CPI Flash estimate Eurozone
Change in the price of goods and services purchased by consumers. Eurostat bases this estimate on energy prices and 13 euro area member states that report early CPI data. There are 2 versions of this report released about two weeks apart – Flash and Final. The Flash report is extremely early and tends to have a significant impact. The data is monthly
Flash manufacturing Purchase Manager Index Eurozone.
Level of a diffusion index based on surveyed purchasing managers in the manufacturing industry. Survey of about 3000 purchasing managers which asks respondents to rate the relative level of business conditions including employment, production, new orders, prices, supplier deliveries, and inventories. Monthly data announcement is incorporated in the regression. There are 2 versions of this report released about a week apart – Flash and Final. The Flash release, is the earliest and thus tends to have the most impact.
Flash services Purchase Manager Index Eurozone
Level of a diffusion index based on surveyed purchasing managers in the services industry. Survey of about 600 purchasing managers which asks respondents to rate the relative level of business conditions including employment, production, new orders, prices, supplier deliveries, and inventories. Monthly data announcement is incorporated in the regression. There are 2 versions of this report released about a week apart – Flash and Final. The Flash release, is the earliest and thus tends to have the most impact.
Preliminary GDP Germany
Change in the price of goods and services purchased by consumers. Eurostat bases this estimate on energy prices and 13 euro area member states that report early CPI data. There are 2 versions of this report released about two weeks apart – Flash and Final. The Flash report is extremely early and tends to have a significant impact. The data is monthly
German ZEW Economic Sentiment.
Survey of about 275 German institutional investors and analysts which asks respondents to rate the relative 6-month economic outlook for Germany. Quarterly available data.
Non-Farming Employment Change US
Change in the number of employed people during the previous month, excluding the farming industry. This economic data is released shortly after the month ends. The combination of importance and earliness makes for hefty market impacts. Monthly data
Advance Gross Domestic Product (GDP) US
Annualized change in the inflation-adjusted value of all goods and services produced by the economy. While this is quarterly data, it's reported in an annualized format (quarterly change x4). There are 3 versions of GDP released a month apart – Advance, Preliminary, and Final. The Advance release is the earliest and thus tends to have the most impact.
GDP China
Data represents the quarterly value compared to the same quarter a year earlier. Chinese data can have a broad impact on the markets due to China's influence on the global economy and investor sentiment.
German unemployment change Change in the number of unemployed people during the previous month. Monthly data
24
4. Hypothesis and method.
This Sections explains several hypotheses that I want to test by doing a regression analysis. Explanation on how to perform this regression is also part of this Section and is elaborated in paragraph 4.2.
4.1 Hypothesis
As explained in Section 3.2 I made on purpose a distinction between the severe QE-programs and ECB program independent dummy variable. I want to test for both group of programs whether they had a significant impact on expected inflation or not. For that reason there are two hypotheses that I would like verify. Both hypotheses are tested for the short and medium term.
4.1.1 Hypothesis 1
“The monetary programs, CBPP, SMP and CBPP2, announced by the ECB did have a positive effect on the expected inflation (inflation rise)”
I will test this hypotheses for the one, three and five year ahead expected inflation. If there is sufficient evidence to reject H0 and accept H1, there is a positive effect of the announcement of the indicated programs on expected inflation. As economic theory indicates that a monetary expansion lead to a rise in inflation, I expect to find a positive relation.
4.1.2. Hypothesis 2
“The extended APP program implemented by the ECB did have a positive effect on the expected inflation (inflation rise)”
I will test this hypotheses for one, three and five year ahead expected inflation. If there is sufficient evidence to reject H0 and accept H1, there is a positive effect of the announcement of
25
extended APP on expected inflation. As economic theory indicates that a monetary expansion lead to a rise in inflation I expect to find a positive relation.
4.2 Method
I opt for a methodology that has been suggested and used by Moessner (2013) for quantifying the effects of explicit Federal Open Market Committee policy rate guidance on interest rate
expectations. The regression is ran by using white heteroscedasticity-robust standard errors. I refer to the appendix for the explanation on the use of heteroscadastic-robust standard errors. Similar regression have been widely used in the literature on the effect of central bank communication and news on financial asset prices (see, e.g. the survey of knütter et al. 2011).
For this thesis I will regress two daily changes in euro inflation swap rates for K-year ahead (K = 1,3 or 5 years ahead)on:
1 The independent dummy variable and .
2 The control dummy variable that represents all the meeting and public statements
made after the governing council meetings from the ECB
3 The surprise components of 5 Euro-zone, 4 German, 2 US and 1 Chinese Macro-economic variables.
The regression model is as follows:
∑ ( )
(5)
The dummy variables. , take the value of ‘1’ when there has been an
announcement of a program that falls in their category (for specific dates, see table 1 in Section 2.4). Zero otherwise. The takes the value of 1 in case there has been a meeting by the ECB,
zero otherwise. The takes the value of the surprise variable on dates that there has been a surprise announcement within our range of surprise variables, zero otherwise.
26
5. Results & discussion
I can see from the results presented in table 2 that the announcement effect of the APP did have a significant effect at the 5% level for a three and five year time horizon on expected inflation. That is an announcement of one of the APP programs led to a rise in expected inflation of 4,711 (bèta + constant) basis points on average for a time horizon of three years. The same program had a larger average effect on the 5 year time horizon. A rise of 5,485 basis points in expected inflation for an APP announcement. I therefore reject and accept for the medium term horizon, respectively on the 5% significance level. The APP didn’t show significance14 on the short term. For that reason I accept as the effect of the announcement of APP on expected inflation for the short term. The APP does not show an announcement effect on 1 year ahead inflation
expectations.
An announcement of one of the other ECB programs also had a positive impact on expected inflation on average in this model. However the ECB programs do not provide significance at a 10% level or lower for any time horizon. This means that this outcome has no meaning and I accept To clarify, there is high uncertainty an no statistical proof for the announcement effects of the CBPP, SMP and CBPP2 on expected inflation.
The shows are very low fit15. So the variation of our dependent variable is poorly
explained by the regression model that I used. Although a higher is preferred as it indicates that
14 No significance means that there is ‘not a low’ or high P-value. This p-value is the chance that the outcome
of our model is established by coincidence. If this chance is too large, the outcome is not significant. There is insufficient statistical evidence to conclude that the outcome is not established by coincidence. Henceforth, it has no meaning and I accept
15
is the percentual amount of the response variable variation that is explained by the linear model Table 2
Announcement effect of ECB programs on expected inflation
Variable 1 year 3 year 5 year
Constant -0,115 -0,141 -0,130
3,415 4,852** 5,615*
1,471 3,107 3,032
0,0231 0,021 0,022
No. Variables 2203 2203 2203
Use of white heteroskedasticity-consistent standard errors *significant at 5 % (p-value = 0,047)
** significant at 5 % (p-value = 0,041) No asterisk = no significance below 10%
Dependent variable: changes on a two day interval in zero coupon swap rates for k-years ahead.
27
I have a better model, the statiscal inference, w.r.t. the bètas, that I obtained from this regression is still valuable.
Comparing my findings of the time lag to the literature review in this thesis. I can state that the results are conform the recent research done by Central Banks with regard to the APP program. Just as the papers of monetary policy committee (2005) and of the ECB(2000), I find evidence that expected inflation is not likely to change within a term of one and a half year. Although, these researches were more focused on conventional monetary policy, my findings suggest that the APP program did not affect expected inflation for a horizon shorter than 1 year.
To review the quantification effect on expected inflation, I make a comparison to metrics made by several studies on the effect of QE programs in the UK. The table below is provided by Joyce M., Tong M. and Woods R.,(2011) who made a comparison of estimates of the macro-economic peak impact of QE on output and inflation. Interesting to see is that the range made among the studies show a change of inflation from 75 till 150 basis points. Although these numbers are larger, few conclusion can be made. As this research is focused on another QE program, another country and is using different models to measure.
Finally, the research of the ECB (2000) shows that a rise of the interest by 25 basis points follows an average decline of 10 basis points in inflation after three years (see graph Section 2.2). Assuming the opposites holds. The announcement effect of an APP program was on average equal to a drop in the interest rate of 12 basis points. Again there are large uncertainties but it provides a rough order of the magnitude on the announcement effect of the APP program.
28
6. Conclusion
The aim of this research was to quantify the effects of the unconventional monetary policies on inflation expectations proxied by zero coupon inflation swaps rates from January 2008 till mid-june 2016. I quantified the announcement effect of two categories of unconventional policies from the ECB. One category consists of non-standard monetary policies. These were mainly aimed at redressing impairments in financial markets or to restore the normal functioning of the monetary transmission mechanism (MTM). These programs were implemented as a response to the global financial crisis and the euro sovereign debt crisis. The second category consist of QE programs. These were mainly aimed at preventing the economy from a downfall and future deflationary pressures. These QE programs are a response on a low growth prospect with a decline in the expected inflation over the medium term. By making use of an OLS-estimation to quantify the effects, I can draw the following conclusion from this research.
First the announcement of QE-programs did have a significant effect on expected inflation on a three to five year horizon. The average rise in three year ahead expected inflation is equal to 4.711 basis points when there was an announcement of a QE-program. The average rise in five year ahead expected inflation was equal to 5.485 per announcement of a QE-program. These findings are in line with recent studies on the effect of QE programs on expected inflation.
Second, the ECB programs (CBBP, SMP and CBBP2) didn’t have a significant announcement effect on inflation rate expectations over the short to medium term. This has the following possible explanations. First the seize of the programs were relatively small. Second, the programs were mainly aimed at restoring normal functioning of the markets and MTM than stimulating economic output and creating inflationary pressures over the medium term.
Third, QE programs do not have a significant announcement effect on one year ahead inflation expectations. This conclusion is also in line with other literature on the effect of conventional, expansionary monetary policy on inflation rates. Although, this conclusion provides relevant evidence. As this thesis measures the announcement effect of QE rather than conventional expansionary monetary policy.
Lastly, monetary policy affects economic output and inflation through a variety of
intertwined channels. Although an effort has been made in this thesis to elaborate on the historical evidence, explain on the MTM and provide evidence through use of a financial marked based measure. There remains high uncertainty about the exact impact and time lags that come along with expansionary monetary policy on output and expected inflation.
29
7. Reference list.
ECB(2000). Monetary Transmission Machenism, ECB monthly bulletin. July 2000.58 pages. Obtained on 7th of july on
https://www.ecb.europa.eu/pub/pdf/other/mbjul2000_article07.pdf.
Cour-Thirman, P., Winkler B., (2013). The ECB’s non-standard monetary policy measures; the role of institutional factors and financial structure. ECB working paper series. No 1528/April2013. 44 pages. Obtained on 1st of May 2016 via https://www.ecb.europa.eu/pub/research/working-papers/html/papers-2013.en.html.
Constâncio V. (2015). Assessing the new phase of unconventional monetary policy at the European Central Bank. Panel remarks by Mr Vítor Constâncio, Vice-President of the European Central Bank, at the Annual Congress of the European Economic Association. Obtained on 14th of June via: https://www.ecb.europa.eu/press/key/date/2014/html/sp141006.en.html Deacon, M., A. Derry, and D. Mirfendereski. 2004. Inflation-Indexed Securities: Bonds, Swaps, and
Other Derivatives. 2nd ed. Wiley Finance.
Friedman M., Meiselman D., (1963) ‘The relative stability of monetary volicity and the investment multiplier’ in stabilization policies, ed. Commission on money and Credit (Upper Saddle River, NJ: Prentice-Hallo, 1963, pp. 165-268
Friedman M., and Schwartz A.J.(1963), A Monetary History of the United States, 1867-
1960. Princeton: Princeton University Press (for the National Bureau of Economic Research). Gabriela G., Poelhekke S., Zhou C., (2009). Did the crisis affect inflation expectations. DNB Working
paper. No. 222 September 2009. 30 pages.
Governing Council of the ECB, (2009). Decision of the European Central bank on the implementation of the coverd bond purchase program. Official journal of the European Union. 4/7/2009. 2 pages.
Governing Council of the ECB, (2010). Decision of the European Central Bank on establishing a securities markets programme. 2 pages. Found on ECB.eu
Governing Council of the ECB. (2011). Decision of the European Central Bank on the implementation of the second covered bond purchase programme. Official journal of the European Union. 16/11/2011. 2 pages.
Governing Council of the ECB. (2011, February). Inflation expectations in the euro area: A review of recent developments. Monthly ECB bulletin. Found on 1st of july on:
30
Governing Council of the ECB. (2015). Decision of the European Central Bank on a secondary markets public sector asset purchase programme. Official journal of the European Union. 14/5/2015. 5 pages.
Governing Council of the ECB. (2016). Decision of the European Central Bank on the Implementation of the corporate sector purchase programme.
Joyce M., Tong M., Woods R. (2011). The United Kingdom’s Quantitative Easing Policy: Design, Operation and impact. Bank of England’s Quarterly Bulletin. Quarter 3, 2011. 13 pages. Obtained on the 7th of july on:
http://www.bankofengland.co.uk/publications/Documents/quarterlybulletin/qb110301.pdf JP Morgan. 2008. Inflation-Linked Markets. April 2009. Global Data Watch. 31 July
Knütter, R., Mohr, B. and Wagner, H., 2011. The effects of central bank communication on financial stability: a systematization of the empirical evidence. Fernuniversität Hagen Discussion Paper No. 463.
Mishkin F., K. Matthews and M. Giuliodori. (2013). The Economics of Money, Banking, and Financial Markets (European Edition), Harlow: Pearson.
Moesner R. (2013). Effects of explicit FOMC policy rate guidance on interest rate expectations. DNB Working paper. No. 384/july 2013. 20 pages.
Monetary policy committee BOE (2005). The transmission mechanism of monetary policy, Report for treasury commission of the house of commons UK. May 1999. 10 pages. Obtained on
http://www.bankofengland.co.uk/publications/Documents/other/monetary/montrans.pdf Montgomery, D. C., Peck, E. A. and Vining, G. G. (2001). Introduction to Linear Regression Analysis.
3rd Edition, New York, New York: John Wiley & Sons OECD (2000) EMU One Year. Paris. Obtained on 9th of july via
http://www.oecd.org/economy/outlook/1889423.pdf
Pattipeilohy, Van den End, Tabbae, Frost, Haan, J., (2013).Unconventional monetary policy of the ECB during the financial crisis: An assessment and new evidence. DNB Working paper. No 381/ may 2013. 41 pages. Obtained on 8th of june 2016 via
http://www.dnb.nl/publicatie/publicaties-dnb/dnb-working-papers-reeks/dnb-working-papers/working-papers-2013/dnb291732.jsp.
President of ECB (2011). Statement by President of ECB. Press release ECB. Pages 1 obtained on 14th of june via https://www.ecb.europa.eu/press/pr/date/2011/html/pr110807.en.html
31
Trichet J.C., (2009). The ECB’s Enhanced Credit Support. CESIFO WORKING PAPER NO. 2833. 18 pages. Obtained on 10th of june via
http://econpapers.repec.org/paper/cesceswps/_5f2833.htm
Important websites for general background information used:
n
https://www.ecb.europa.eu/mopo/intro/objective/html/index.en.html
ghttp://www.euromoney.com/Article/2985829/LTROs.html.
https://www.ecb.europa.eu/mopo/implement/omt/html/index.en.html
https://www.ecb.europa.eu/mopo/implement/omo/html/index.en.html
32
8. Appendix
A1. Elaboration on assumptions and aspects of the Method:
In this part I extend on several technical aspects an assumption of the method that is used within this thesis.
i. Regression with binary variables
Binary variable is another word for dummy or indicator variable. When a variable is binary it can only take two values, 0 or 1. In the case of this thesis; the binary variable meeting and statement made by the governing council of the ECB, , is ‘1’ for every date that there was a meeting
of the governing council. The same logic counts for the variable & .
Hence:
{
The mechanics of a regression with a binary regressor are the same as if it is continuous. The interpretation of , however, is different and it turns out that regression with a binary variable is equivalent to performing a difference of means analysis. This means that the difference in the sample averages of the dependent variable is equivalent to the . To clarify I provide a simple regression model with zero coupon inflation swap rates for k-year ahead as dependent variable and the as dummy variable for announcement dates of a program of the ECB.
The is then the difference in the average of the sample of zero coupon inflation swap
rates on all the dates that the was equivalent to 0 (i.e. there was no announcement of
a ECBprogram, hence almost all the data) and all the dates that was equivalent to 1 (i.e.
there was an announcement of a ECBprogram, hence just few numbers of data).
ii. Auto correlation in time series data
33
next period. The correlation of a series with its own lagged values is called autocorrelation or serial correlation. Stock and Watson (2012) state that The first autocorrelation (or autocorrelation
coefficient) is the correlation between , that is, the correlation between values of Y at two
ρadjacent dates. The autocorrelation is the correlation between - . In this theses I do use
time series data for our regression. For that reason it is relevant to test whether the residuals of the regressions are auto correlated. I can perform this test by doing a Durbin-Watson Test.
∑
∑ where = )
And Ŷ and Y are the predicted an observed values of the regression.
I will test the following hypotheses with regard to the correlation:
D-stat becomes smaller as the serial correlation increases. In table three we see the results on the Durbin-Watson statistic. The statistic is smaller than the lower bound and for that reason I reject and accept . I can assume that our regression model contains serial autocorrelation.
The fundamental assumptions in linear regression are that the error terms have mean zero and constant variance and uncorrelated = 0 , = , and . For
purposes of testing hypotheses and constructing confidence intervals I often add the assumption of
Table 3
Durban watson-statistic.
k=16 and n=2203
1 year 3 year 5 year
D-stat: 0,745 0,609 0,569
Upperlimit* 1,942 1,942 1,942
Downlimit* 1,911 1,911 1,911
*5% significance on n=2000. Critical values for larger observations are not directly available. Inference is still usefull as D-stat is low and difference between critical values of N=2000 and n=2203 is very smal.
34
normality, so that the are ). Some applications of regression involve regressor and response variables that have a natural sequential order over time. Such data are time series data. The assumption of uncorrelated or independent errors for time series data is often not appropriate. Usually the errors in time series data exhibit serial correlation, that is, . Such error
terms are said to be auto correlated (Montgomery, Peck and Vining, 2012)
iii. Heteroscedasticity
An error term is homoscedastic if the variance of the conditional distribution of given
the independent variable | ), is constant for and in particular does
not depend on x. Otherwise , the error term is heteroskedastic (Stock and Watson, 2012) . In many financial/economic data the conditional variance of the error depends on past values of the error term. This phenomenon is also known as autoregressive conditional heteroscedasticity (ARCH). Thus, the conditional variance is not heteroscedastic with respect to the independent variable. But it is heteroscedastic with respect to it’s lagged value (former errors in the time-series) . As I proved in the former paragraph that the error term of our regression is auto correlated. I now want to test whether the residual of our regression show ARCH effects to prove for heteroscedasticity16 Most regressions use a white test to see if there is heteroscedasticity. As our regression has independent binary variables it is not useful to perform this test.17 The regression of the ARCH(p) model for the residual of our normal regression is as follows:
+ + …… + (7)
I want to test the following hypotheses: or or…or
I perform an autoregressive conditional heteroscedastic lagrange multiplier (ARCHLM) test to test for arch effects. An arch effect of any order would be enough evidence to conclude that our
regression equation (5) shows heteroscadatic residuals. For that reason the amount of lags I choose
16
If there is a positive relation between a recent squared d errors and the current squared error. 17 As dummy variable take an amount of 1 or zero it is difficult to regress the residuals with the indepent
variables using a white test. This will lead to perfect multicollinearity and the statiscal inference that I have obtain from the regression is then not usefull.
p = the amount of lags a= unknown coefficients σ =unsquared residual of the primary regression
