Does the Greek redenomination risk affect the government
bond yield of other Eurozone countries?
Name: S. R. van Engen
Student number: S2017067
Student mail: s.r.van.engen@student.rug.nl Study Program: MSc IE&B
Supervisor: M. J. Gerritse Second Supervisor: T.M. Harchaoui
2
Abstract
Since the start of the sovereign debt crisis in the Eurozone, investors are uncertain about a Greek default or exit. This thesis has investigated the impact of redenomination risk stemming from a so-called grexit on the government bond yields of other Eurozone countries. A new proxy for redenomination risk is constructed with the help of Greek euro-denominated and yen-denominated short-term bonds, corrected for exchange rate risks. Unfortunately, none of the results regarding the redenomination risk are significant. However, the thesis confirms the effectiveness of the outright monetary transactions program of the ECB in reducing the government bond yield of Greece. Moreover, it shows robust results concerning the liquidity premium and risk aversion on government bond yields within the Eurozone.
3
I. Introduction
The introduction of the euro in 1999 has led investors to feel more safe as they believed a safety net was installed by the countries all agreeing to operate under a common currency (Bernoth, von Hagen and Schuknecht, 2004, & Manganelli and Wolswijk, 2009). Since the start of the financial crisis in 2008, government bond yields of the Eurozone have increased in spread vis-à-vis the German Bund (see graph 1, also consult De Santis, 2012; De Grauwe and Ji, 2014) and volatility in the government bond market increased.. Bank bailouts in Ireland and fraudulent Greek government budgets in 2009 have reduced the feeling of safety among investors. The market discipline of government bonds came into action which resulted in countries with higher risk having increased bond yields to compensate for this risk (Bernoth, von Hagen and Schuknecht, 2004). For a more detailed timeline of events since 2009, please consult table A1 in the appendix.
Graph 1. Spreads on 10-year government bond rates in selected Eurozone countries
Source: De Grauwe and Ji (2014)
4
The main difficulty in dealing with the sovereign debt crisis is the fact that all of these countries have the same currency. This poses a threat to the mission of the European Central Bank (ECB), which is maintaining price stability. Many countries in the Eurozone have different economic cycles (de Grauwe, 2012a). This complicates the policy making of the ECB as its monetary actions can lead to deflation when focusing on prosperous countries or to a high inflation (above 2 percent) in some countries when its policy is aimed at the countries in economic difficulty (de Grauwe, 2012a). This causes tension and proves a difficult job for the ECB to keep all parties happy under the current situation. As the lion’s share of the countries are doing relatively well, the ECB cannot do all in its power to adjust the situation in Greece as this would harm the economies of the relatively well performing countries.
Furthermore, political tensions rose during the height of the sovereign debt crisis as spillover risks and contagion increased. For example, Ang and Longstaff (2013) found that Greece had triple the systemic risk of countries like Portugal, Ireland and Spain, which in itself had twice the systemic risk of countries like Germany. They also indicate that the main systemic risk stems from the financial sector and not from macroeconomic factors. Most of the risk during the sovereign debt crisis stems from the fear of a collapsing currency union which poses multiple threats to investors. A possible default or redenomination of Greek debt leads to contagion effects as the demand of EMU government bonds of countries with weak fiscal fundamentals deteriorated, increasing their yields (de Santis, 2012). This could lead to a self-fulfilling prophecy of defaulting governments, forcing them to step out of the currency union.
5
Stepping out of the Eurozone would allow a country to print its own currency and thus regain full control over its monetary policy. Furthermore, the country would be able to determine its exchange rate vis-à-vis the world, increasing its competitiveness (de Grauwe, 2012a). However, stepping out of the currency union would possibly mean a redenomination of government debt. Investors could have less faith in the new currency and the defaulting government concerning its ability to service its debt. Furthermore, a redenomination increases exchange rate risks of investors (Manganelli and Wolswijk, 2009).
As indicated by Graph 1, and shown by the events that have taken place since the start of the sovereign debt crisis described in table A1 in the appendix, the likelihood of default was the highest for Greece. From this, one can also conclude that this country would have the most to gain from stepping out of the currency union and redenominate its debt into a new self-controlled currency (de Grauwe, 2012a). Hence, this research paper will focus on the effects of a possible redenomination of Greek government debt and the threats this poses to other Eurozone countries. The following research question will be investigated:
Does the possible redenomination of Greek bonds affect other Eurozone government bond yields?
In investigating this research question, the redenomination risk has proven hard to measure in previous literature. Data is either missing (Krishnamurthy et al., 2013), or measures are not precise regarding the country that will exit the currency union and thus redenominate its debt (Klose and Weigert, 2013). Therefore, a new proxy will be constructed with the usage of data on yen-denominated Greek government debt. The proxy will be explained in detail in the Methodology section.
6
(2013) and Al-Eyd and Berkmen (2013). Furthermore, as expected countries with a low risk profile benefit from increased risk aversion among investors as this results in lower government bond yields (which confirms findings of De Santis, 2012). The same holds for investors in need of liquid assets, they tend to move towards the bonds of low risk countries, as these are often also very liquid (same conclusion as Ejsing et al., 2012).
The rest of this thesis will be organized as follows. Section 2 discusses the literature review and derives the hypothesis. Section 3 explains the new proposed measure for redenomination risk and deals with the methodology and econometric model. Section 4 explains the dataset. Section 5 will give descriptive statistics and regression results together with robustness checks, and finally Section 6 will conclude and discuss the limitations of this research.
II. Literature review
i. Sovereign risk in a currency union
Once becoming part of a currency union, this poses several threats to sovereigns next to the many benefits it provides. These are highlighted to a great extent by De Grauwe (2012a, 2012b). One of the main problems is no longer having control over monetary policy. A member of a currency union issues debt in a currency over which it does not have any control, leading to increased default risk as the country is not able to summon its central bank to print excess money in order to repay debt. This has been one of the main reasons why countries like Spain and Italy have had severe difficulties during the financial crisis, while the UK, a country with an even higher government deficit during the time, was hardly impacted due to its influence on monetary policy and ability to let the central bank print money in order to repay debt (De Grauwe, 2012b). When entering a currency union, monetary tools like printing money to increase inflation or government spending are no longer available. The country is not able to decrease the real value of nominal debt or depreciate a currency to boost the economy and increase competitiveness (De Grauwe, 2012a).
7
and contagion risk between the financial sector (banks) and sovereigns. As large banks came into difficulty due to liquidity shocks in America, these banks had to be saved by national governments (as a banking union was lacking in the Eurozone). This caused deteriorating government budgets and weakening financial fundamentals. According to De Santis (2012), contagion “is a situation whereby instability in a specific market is transmitted to one or several other markets”. His research shows that a country in the Eurozone with weak fiscal fundamentals is more prone to contagion risk. Increased contagion risk thus leads to a lower demand for bonds, depressing market prices and increasing yields. The same concept is described by Allen and Gale (2000) for the banking system, which can be transformed to the country level. Once country A does not have sufficient funds and goes bankrupt due to a shock in liquidity (investors do no longer want to purchase their bonds, so the country becomes unable to (re)finance its debt), it is no longer able to service its obligations to country B. Country B needs to get new financing to make up for this loss, increasing the risk levels of the country as they suddenly run larger deficits which makes it harder to obtain liquidity. Overall, this reduces the value of country C’s assets in country B, leading to a vicious circle, increasing default risks and declining asset prices. This risk is more prone in a currency union as the countries are not able to create liquidity themselves through the monetary policy.
ii. Redenomination risk
These threats increase the likelihood of a country not being able to service its debt and function appropriately under the burdens of a currency union. When a country decides to step out, it will adopt its own currency and can allow its central bank to print currency, regaining control over monetary tools (De Grauwe, 2012a). However, when a country decides to step out of a currency union, it will need to redenominate its current debt into a currency it is now able to control in order to benefit to a maximum extend of the regained monetary control. This is only possible for the bonds under domestic-law, so euro-denominated bonds in the case of Greece, according to Krishnamurthy et al. (2013).
8
bank of the country is not able to execute its (possible) mandate of low inflation and economic stability (De Grauwe, 2012a). Furthermore, bond holders might now also be faced with an exchange rate risk that might not have been apparent with euro-denominated bonds (Manganelli and Wolswijk, 2009). Lastly, a redenomination of bonds is often seen as credit event (Thomson Reuters CDS – FAQ), as it gives off clear signals to investors that the country is no longer able to oblige to the demands of the currency union it is part of (Cantor and Packer, 1996). Default history will increase yields as bond holders will ask for a discounted bond price (Cantor and Packer, 1996) .
Existence of redenomination risk in the Eurozone is proven by Klose and Weigert (2013), who have researched the opinion of investors on the likelihood of a country leaving the Eurozone. They have defined redenomination risk as “the risk that a country will unilaterally exit the Eurozone and redenominate its public and private liabilities”. According to these authors, there is a systemic risk component that influences government bond yields, besides the effect of fundamental factors like budget deficits, credit ratings and the economic situation. This is confirmed by e.g. De Santis (2012), Ang and Longstaff (2013) and De Bruyckere et al. (2013). Klose and Weigert (2013) used the variable INTRADE as a proxy for redenomination risk. INTRADE is an online trading platform that allows its users to gamble on certain future events. In the research on redenomination risk, Klose and Weigert (2013) used a financial instrument available on this platform that allows investors to buy contracts which will pay out 10 euro’s in case at least one country leaves the Eurozone before the end of 2013. However, this data was only available for a limited time period and one is not able to link the investors opinion to a specific country. Furthermore, the online trading platform has been closed due to “pending investigation of financial irregularities” (BBC, 2013). Hence, a different proxy for redenomination risk is proposed in the methodology section of this paper.
euro-9
denominated Greek bonds and have compared this to the yields of foreign-law USD-denominated Greek bonds, which is closely linked to the proxy for redenomination risk proposed in this paper. Moreover, they have looked closely at the impact of ECB policies on government bond yields. They state that the Outright Monetary Transactions (OMT) program and the Securities Markets Program (SMP) were most effective, while the Long-Term Repurchase Operations (LTROs) hardly reduced government bond yields.
iii. Central Bank policies and effects on redenomination risks
The central bank in a currency union is able to influence the threats posed on countries through its monetary policy. In the Eurozone, the general objective of the ECB is to maintain price stability, besides supporting economic policies in the Eurozone (European Central Bank, 2011). Policies are implemented on a Eurozone wide basis and not specifically for one nation, this in order to not disrupt price stability elsewhere. There is a focus on price stability, and thus inflation, as this allows for the efficient allocation of savings and investments within the economy, prevents redistribution of income and wealth and overcomes risks related to deflation (European Central Bank, 2011). The main difficulty with which the ECB has had to deal during the sovereign debt crisis is the existence of asymmetric shocks. This paralyzes the ECB as it is unable to pick a side, as this would lead to increases in the shock elsewhere (either positive or negative). Sound fiscal policies are important in case of asymmetric shocks; they are an insurance mechanism to improve the sustainability of government debt and thus improve the capacity of countries to deal with asymmetric shocks (European Central Bank, 2011 and the European Commission).
10 Euro-Area Govern ments 62% Private sector 17% IMF 10% ECB 8% Others 3%
Graph 2. Owners of Greek Debt
Source: Bloomberg Brief, 7th of January 2015
and thus this program provides investors with the safety that they will be refunded for their bonds. So far, the OMT program has not been used, but has had the proposed effect on government bond yields (Krishnamurthy et al., 2013; Al-Eyd and Berkmen, 2013).
Next to becoming the lender of last resort, the ECB has given banks unlimited access to liquidity, either in euro’s or dollars (European Central Bank, 2011). Furthermore, the Quantitative easing program allows the ECB and national Central Banks to buy government bonds from Eurozone governments. The QE program has caused a fall in government bond yields and led to a depreciation of the euro vis-à-vis the US dollar (Financial Times, 2015). There is a strong opposition to the QE program; the program would disrupt the functioning of capital markets, as risks are artificially lowered. The distortion allows countries to follow unsound fiscal policies without being punished by the capital markets with high interest rates (Randow, 2015).
iv. Greece’s main debt holders
As this paper focusses on the redenomination risk stemming from Greece and the effect on other Eurozone government bond yields, it is important to see who the main bond holders of Greek debt are and how they will be effected by a possible redenomination. According to the Bloomberg Brief of the 7th of January 2015, the holders of Greek debt are distributed as shown in graph 2 (data is from the third quarter of 2014 and based on Bloomberg, the Greek Finance Ministry and the European Commission). The majority of Greek
debt is owned by Euro-area
11
for Eurozone countries, leading to increased default risk; a vicious circle that no Eurozone member would like to be part of. Furthermore, only 17% of Greek outstanding debt is hold by the private sector, which makes the secondary market fairly thin. This could affect 10-year bond yields of Greece, which is confirmed by the article of Whittall and Stubbington (2015) in which they state that only €2 million government bonds were traded in April 2015.
The risk a country runs by holding Greek government bonds is depicted by the height of these bonds volumes as a percentage of nominal GDP, which is shown in Graph 3 for the Eurozone (Bloomberg Brief, 7th of January 2015). The graph shows how much countries are indebted in % of nominal GDP (2013) and the total amounts. Interesting to see is that the relatively smaller countries like Slovenia and Malta have a higher exposure as a percentage of GDP than Germany, which on the other hand has the highest overall exposure in euros. One would thus expect these countries to be highly influenced by the redenomination risk stemming from Greece.
Graph 3: Indebtedness of Eurozone countries in Greece, showing both total billion euros and as a percentage of nominal GDP
v. Hypothesis
According to the literature, a country faces multiple threats when becoming part of a currency union. The financial crisis and the following sovereign debt crisis are an interesting case study
12
of these threats. Furthermore, part 1.4 showed that Eurozone governments owned the largest share of Greek debt (62%). When Greece would leave the Eurozone (with a redenomination of its government bonds), this will have large consequences for other Eurozone governments. De Santis (2012), Ang and Longstaff (2013) and De Bruyckere et al. (2013) have found proof of systemic risks within the Eurozone. Moreover, Klose and Weigert (2013) show that redenomination risk indeed has a negative effect on government bond yields of other Eurozone countries. Following from the literature review and the research question, the hypothesis that is tested is the following:
H1: The redenomination risk stemming from Greece has a significant negative impact on other Eurozone government bond yields.
The next section will describe the methodology used in this paper to investigate this hypothesis. Thereafter, the data to test the econometric model will be elaborated upon and results are given.
III. Methodology
13
In order to test this hypothesis, the following econometric model is used:
∆𝑟𝑗,𝑡= 𝛽1𝑗 + 𝛽2𝑗𝑅𝑒𝑑𝑒𝑛𝑜𝑚𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑟𝑖𝑠𝑘𝑔𝑟𝑒𝑒𝑐𝑒,𝑡−1+ 𝛽3𝑗 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦 𝑝𝑟𝑒𝑚𝑖𝑢𝑚 𝑡−1 + 𝛽4𝑗 𝐺𝑙𝑜𝑏𝑎𝑙 𝑅𝑖𝑠𝑘 𝐴𝑣𝑒𝑟𝑠𝑖𝑜𝑛 𝑡−1+ 𝛽5 𝑂𝑀𝑇 + 𝛽6𝑗 𝐺𝑜𝑣𝑒𝑟𝑛𝑚𝑒𝑛𝑡 𝐷𝑒𝑓𝑖𝑐𝑖𝑡 + 𝜀𝑗,𝑡
In which ∆rj,t is the dependent variable: the daily difference in the government bond yield of country j at time t. The redenomination risk is the independent variable, see explanation below. The liquidity premium, global risk aversion, OMT and government deficit are control variables.
i. Dependent variable
The dependent variable is the daily difference in the 10-year government bond yield of a country at time t. As shown by the regression equation, there are several factors that are expected to influence the government bond yield. The main focus of this thesis is on the redenomination risk of Greek government bonds and what the effect of this will be on the government bond yields of other Eurozone countries. The first differences are taken as the data contains an unit root, which is often the case with time series data. For an explanation on this, see the test results of the Dickey-Fuller test in the descriptive statistics part of chapter 5.
ii. Independent variable
14
The Japanese yen denominated bond (with a coupon of 5%) used for this proxy matures on the 22nd of August 2016, while the Euro-denominated bond matures on the 11th of March 2019 (5% coupon). This causes a problem as bonds closer to maturity are expected to hold a value closer to their face value, due to lower default risks in the meantime or that there will be large shocks in inflation, interest rates or exchange rates. In order to cope with this difference in maturity, the yield to maturity (YTM) is calculated.
The YTM is mainly used to compare bonds, as bonds have different coupon rates and maturity. The YTM is calculated as:
𝑌𝑇𝑀 = 𝐴𝑛𝑛𝑢𝑎𝑙 𝐶𝑜𝑢𝑝𝑜𝑛 +
𝐹𝑎𝑐𝑒 𝑉𝑎𝑙𝑢𝑒 − 𝑀𝑎𝑟𝑘𝑒𝑡 𝑃𝑟𝑖𝑐𝑒 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑌𝑒𝑎𝑟𝑠 𝑡𝑜 𝑀𝑎𝑡𝑢𝑟𝑖𝑡𝑦 𝐹𝑎𝑐𝑒 𝑉𝑎𝑙𝑢𝑒 + 𝑀𝑎𝑟𝑘𝑒𝑡 𝑃𝑟𝑖𝑐𝑒
2
The annual coupon is the annual amount that the bond will pay out as interest. The face value of a bond is the amount which it will be worth when it has reached its maturity; normally this is €1,000 for euro-denominated bonds. Bonds that trade ‘at par’ are traded at their face value. The market price is the price at which the bond is currently traded. Number of years to maturity is the time period until the bond (when it carries a principal, which often is the face value) will be paid out to the bond holder. The coupon and face value are set over the duration of the bond, when the bond is not linked to inflation (non-indexed bonds). The YTM is thus mainly influenced by the bond’s market price. When the market price of a bond goes down, its YTM will go up, and vice-versa.
In order to compare the YTM of the two different bonds, the yen-denominated bond must be adjusted for fluctuations in the exchange rate. When investors expect large fluctuations in the exchange rate, they face more risk regarding their returns and thus want to be compensated for these risks, leading to a higher YTM. We adjust the yen-denominated bond by using the interest rate parity.
15
opportunity for investors. Investors will want to move towards the country that now offers the highest real return. In the end, this leads to an equilibrium in returns as this movement of investors changes the exchange rate (excess supply of currency with a low real return and excess demand of the currency with a high real return leads to changing exchange rates and thus also changing real returns, this takes place until the returns are equal in both currencies). Overall, this implies that the expected return of a domestic investment must equal the expected return on a foreign investment when this foreign investment is adjusted for exchange rate fluctuations. The investor can hedge its exchange rate risk exposure by the usage of a forward contract; a contract between two parties to buy/sell an asset at a specified price on a specified date. The forward contract takes away uncertainty and thus reduces the exchange rate risk exposure.
When the investor uses a forward contract, the interest rate parity is covered. This leads to the following equation:
(1 + 𝑖
𝑛) =
𝐹
𝑡𝑆
𝑡(1 + 𝑖
𝑓)
In which 𝑖𝑛 is the national nominal interest rate, 𝐹𝑡 is the forward rate at time t, 𝑆𝑡 is the spot rate at time t and 𝑖𝑓 is the nominal interest rate of the foreign currency. One could also read the 𝑖𝑛 and 𝑖𝑓 as returns on the national and foreign investments. In the case of this thesis, the 𝑖𝑛 and 𝑖𝑓 are the yields to maturity of the euro-denominated bond (𝑖𝑛) and the yen-denominated bond (𝑖𝑓).
The equation would then become:
(1 + 𝑖
€) =
𝐹
€ ¥,𝑡𝑆
€ ¥,𝑡(1 + 𝑖
¥)
16
variable is the YTM of the Yen-determined Greek bond, which is multiplied by the forward exchange rate divided by the spot exchange rate. A window of 30 is taken, as this is equal to approximately one month.
Therefore, the regression equation which will obtain the proxy for redenomination risk is the following: (1 + 𝑌𝑇𝑀 𝐸𝑢𝑟𝑜) = (1 + 𝑌𝑇𝑀 𝑌𝑒𝑛) 𝑥 ( 𝐹𝑜𝑟𝑤𝑎𝑟𝑑 𝐸𝑢𝑟𝑜 𝑦𝑒𝑛 𝑆𝑝𝑜𝑡 𝐸𝑢𝑟𝑜 𝑦𝑒𝑛 )
A high R-squared indicates that the Yen-denominated bond is able to explain the Euro-denominated bond very well. This is the case when the two bonds move in a similar fashion and are able to explain fluctuations of each other. A high R-squared is an indication that there is a low redenomination risk. This is due to the fact that the Euro-denominated bond does carry redenomination risk while the Yen-denominated bond does not. When the two bonds are able to explain each other, there must be a low redenomination risk in the Euro-denominated bond otherwise the two bonds would not be similar. The inverse is true for a low R-squared. The Yen-denominated bond has a low explanative power for the Euro-denominated bond, thus there is a large portion of the Euro-denominated bond that cannot be explained by the Yen-denominated bond. As the bonds are fairly similar (e.g. same coupon, same bond issuer and thus the same default risk), this unexplained factor could be indicated as the redenomination risk of the Euro-denominated bond.
17
Moreover, the research of Krishnamurthy et al. (2013) was focused on ECB policies and the channels through which these policies influence the government bond yields of other Eurozone countries. They have composed a method with the usage of USD-denominated bonds and state that government bonds issued in a foreign currency operate under foreign law. These bonds can thus not be redenominated by changing domestic law, hence, only bonds issued in domestic currency are eligible for redenomination and thus carry redenomination risk. This strengthens the argument made in this paper that yen-denominated bonds do not carry any redenomination risk and the difference between the two indicates redenomination risk (plus compensation for exchange rate risk as explained above with the interest rate parity). However, their research mostly focused on Spain, Italy and Portugal as they state to not have sufficient data on Greece. Luckily, with the usage of yen-denominated bonds as explained above, this research is able to conduct an investigation on redenomination risk stemming from Greece.
Another important paper on contagion risk is from De Santis (2012) who has mainly focused on credit rating changes of the country’s government bonds. Results show that a rating downgrade for Greece, Ireland and Portugal leads to increasing yield spreads between other Eurozone country pairs vis-à-vis Germany (e.g. the French government bond yield moves upward when the German bond yield remains the same or goes down). This especially holds for countries with weak fiscal fundamentals. The effect from Greece indicates severe contagion risk, especially for Italy, Portugal, Spain, Belgium and France. A disadvantage from using government bond credit ratings is that they do not portray the current situation but are decided upon with a delay. Other market instruments (like credit default swaps) react to new market information on a more timely manner and would therefore be better able to portray the events (Hull et al., 2004). In the last paragraph of the control variables section below is explained why credit default swaps data is not applicable to answer the research question and is therefore omitted.
iii. Control variables
18
bonds, depending on their perceived riskiness, and has an impact on the bond market as a whole.
Ejsing et al. (2012) have found evidence of liquidity and credit premiums in government bond yields. A liquidity premium is a higher price that has to be paid when the asset is more liquid than other assets; there are larger trading volumes in which the asset can be traded at lower costs (Ejsing et al., 2012). Liquidity can also be explained as the ease of selling or buying a bond at the price close to the mid-quote, which is the average of the bid-ask spread (de Haan et al., 2012). Liquidity becomes more important in uncertain markets as investors prefer to be able to sell off their assets quickly and possibly at a price close to their asking price, both are easier when the market of the asset is more liquid. The liquidity premium decreases the yield of a bond as it increases the market price. For the liquidity premium it was not possible to use bid-ask spreads of bonds as the data was not available on DataStream. Therefore, this research follows the liquidity premium method used by De Santis (2012) and takes the KfW-Bund spread as a proxy. The ‘Kreditanstalt fur Wiederaufbau’(KfW) is an agency which is backed by the German government. Both bonds therefore carry the same credit risk of Germany and the remaining difference should indicate the liquidity premium, according to De Santis (2012).
19
According to Cantor and Packer (1996), external debt affects government bond yields as they impact a country its ability to service its debt. Rising per capita income leads to higher revenue income for the government increasing the chance of debt repayment. A high level of external debt increases the chance that external parties could massively sell bonds, leading to a crash of bond prices and the self-fulfilling nature of a country defaulting. The government deficit as a percentage of GDP is included as a proxy for debt sustainability of a country. Once the government deficit becomes too high, investors lose trust that the country will be able to service its debts. Increasing deficits would thus lead to increasing yields as investors start to demand higher returns to make up for the risk.
There will be no usage of credit default swap (CDS) data, as the redenomination risk is possibly included within a CDS. A credit default swap gives an investor the opportunity to buy protection at a fixed premium per period until the contract matures or in case of default. When default occurs, the seller of the contract has the obligation to buy the defaulted bond at par value from the buyer (Longstaff et al., 2005). The CDS rate shows the perceived likelihood that a bond issuer will default on its obligations. The CDS rate will go up when more investors are willing to buy security against default than sellers are willing to sell this security, indicating that investors perceive increased risk. According to Thomson Reuters CDS - FAQ, a full restructuring means that in case of a restructuring event, this event will automatically be noted as a credit event. A redenomination of a bond is also a restructuring event, hence the overlap between the two measures. Furthermore, as Greece has had a restructuring event in March 2012, the CDS data of Greece is not complete (no trading activity between the 7th of March 2012 and 26th of November 2014).
IV. Data
i. Time period and countries
20
Data has been collected for the following Eurozone member countries: Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Malta, Netherlands, Portugal, Slovenia and Spain. Data collection has been determined by countries being an Eurozone member for the full period (August 2009 till August 2015) and the availability of data for 10-year government bond yields on a daily basis. Estonia, Latvia and Lithuania have adopted the Euro after August 2009 and are therefore not part of the dataset. For Cyprus, Luxembourg and Slovakia there is no (sufficient) government bond yield data available. Overall, data on 13 of the 19 Eurozone member countries is used.
ii. Data collection
For the long-term government bond yield, the 10-year yield is taken, following most research in this field. Daily data is collected from Thomson Reuters Datastream. Data is based on the secondary market yields of which the bond’s maturity is close to 10 years.
For the redenomination risk proxy, a euro-denominated Greek government bond trading in Italy and a Japanese Yen-denominated Greek government bond trading in Tokyo, Japan, are taken. The euro-denominated bond (ISIN code: IT0006527532) matures on the 11th of March, 2019 and has a coupon rate of 5%. The yen-denominated bond (ISIN code: JP530000CS83) matures on the 22nd of August 2016 and also has a coupon rate of 5%. The main differences between these bonds are thus the time to maturity and the currency in which they are denominated. For the forward rate, the most conservative option is used. The yen-denominated bond runs till the 22nd of August 2016. Therefore, the 1-year forward rate is used for the time period from the 24th of August 2015 till the end of August 2015. From the 25th of August 2014 till the 21st of August 2015 the 2-year forward rate is used. From August 2009 till the 25th of August 2014 the 5-year forward rate has been applied. Data on the bonds, spot rate and the forward rate is obtained from Datastream. The method used to obtain the proxy is explained in the methodology section above.
21
the maturing date and the same coupon value (in this case, zero). As bonds are stripped from coupons, their percentage of face value at which they are traded will move closer to 100% (trading at ‘par value’) when the bond is reaching maturity. Data is on a daily basis and obtained from Thomson Reuters Datastream. Expected is that the base points of the Bund are higher than the KfW bond due to the Bund being more liquid, and therefore having a higher value as a percentage of the face value.
Furthermore, De Santis (2012) indicates that risk aversion of investors and the overall global uncertainty play a role in determining government bond yields. Risk aversion negatively affects the yield of a bond. As bonds are seen as low-risk investments, demand of bonds will increase during risk adverse times, driving up bond prices and thus decreasing yields. The same holds for increasing global uncertainty. For global risk aversion, the spread between the US corporate BBB benchmark yield and U.S. government benchmark bond yield is used. Data is on a daily basis and from Thomson Reuters Datastream.
Government deficit data is obtained from the European Central Bank website. The data is on a quarterly basis and is presented as a percentage of annual GDP. The data is neither seasonally nor calendar adjusted.
V. Results
i. Type of data and regression method used
22
account the effect of control variable data of country A on the effect of Greek redenomination on country B (neither of which is Greece) is not the aim of this paper, therefore, the method chosen is not panel data with dummy variables for each specific country. The research specifically focusses on the contagion risk stemming from Greece via the redenomination risk channel, therefore, the effect of a non-Greek country on the contagion risk of country A is not investigated. When one would look at a single coefficient produced by panel data on redenomination risk stemming from Greece within the Eurozone, one could only be able to conclude that there is indeed an no effect or no effect on government bond yields of other countries, but not be precise in the effects on a specific country.
ii. Descriptive statistics
23 0 10 20 30 40 P e rc e n ta g e 7/1/2009 1/1/2011 7/1/2012 1/1/2014 7/1/2015
Greece Germany Spain Belgium
Ireland Italy France Malta
Netherlands Austria Portugal Slovenia
Finland
Government Bond Yield Spreads
Graph 4. The government bond yield spreads of the 13 Eurozone countries for the period 1/8/2009 – 31/8/2015 in percentages.
Table 1. Summary statistics of government bond yields
Government Mean Standard deviation Minimum Maximum
Austria 2.364 0.987 0.202 3.886 Belgium 2.784 1.134 0.344 5.865 Finland 2.151 0.918 0.193 3.784 France 2.442 0.894 0.352 3.780 Germany 1.901 0.883 0.077 3.500 Greece 12.936 7.609 4.423 39.850 Ireland 5.071 2.813 0.657 14.552 Italy 4.096 1.298 1.138 7.311 Malta 3.504 0.967 1.077 4.712 Netherlands 2.182 0.907 0.226 3.781 Portugal 6.517 3.293 1.570 17.355 Slovenia 4.557 1.507 0.776 7.569 Spain 4.220 1.418 1.138 7.586
24
periods of high yields; 14.55 and 17.36 percent respectively. Countries with low government bond yields are Austria, Finland, France, Germany and the Netherlands, all have maximum government bond yields of below 4 percent during August 2009 – August 2015 with minimums under half a percent. Germany overall has the lowest government bond yield with a mean of 1.9 percent.
As seen in graph 5, there are differences between the yields of Greek government bonds. The yield of the 10-year government bond lies below the yields of bonds with shorter maturities. This might be due to the fact that investors perceive more risk on the short-term than long-term. Besides, holders of long-term government bonds are often financial institutions, and in the case of Greece, Euro-Area governments (see part 2.2). This fact might also explain the difference between the euro and yen denominated short-term bonds during 2012 and the beginning of 2015. Normally, one would expect a bond with a shorter maturity to have a market price closer to the face value of the bond, as the risk of default, sudden inflation or increasing interest rates is smaller and easier to foresee in the nearby future then when needing to make assumptions over a 10 year period. However, as shown in the graph, the YTM of the yen-denominated bond is higher than the YTM of the euro-denominated bond, which could indicate that investors perceive the nearby future to be more risky than in three years. This high peak could also be due to changes in the monetary policy of the Eurozone, for example the introduction of quantitative easing has led to a large depreciation of the euro versus the dollar.
25 0 50 1 0 0 1 5 0 Pe rce n ta g e 7/1/2009 1/1/2011 7/1/2012 1/1/2014 7/1/2015 Date
10-year government bond Euro-denominated bond Yen-denominated bond
Greek bond yields
Graph 5: The development of Greek government bond yields for different bonds over the time period 1/8/2009 – 31/8/2015.
Table 2. Summary statistics regarding the yield of Greek government bonds
Bond type Mean Standard Deviation Minimum Maximum
10-year Government bond 12.936 7.609 4.423 39.850 Yen-denominated Short-term bond 22.782 19.012 3.248 154.893 Euro-denominated Short-term bond 15.793 12.294 2.549 47.256
Table 2, which covers the summary statistics of the Greek government bond yields, shows that the 10-year government bond has a lower mean, less fluctuation and a closer range for the yield than a corresponding shorter-term euro-denominated bond. Furthermore, it indicates that the fluctuations in the yield of the yen-denominated short-term bond are larger (a higher standard error and broader range) than for the short-term euro-denominated bond.
26
a case of unit root within the data. Therefore, the first difference is taken of all variables used within the regression before computing the rolling R-squared in order to take out the unit root.
Table 3. Test results for the Dickey-Fuller unit root test on variables of redenomination risk equation
Country Z(t) Statistic MacKinnon approx. P-value
Euro-bond yield -2.109 0.2409
Yen-bond yield 2.912 1.0000
Forward Rate -1.416 0.5747
Spot Rate -2.039 0.2695
Forward / Spot -0.829 0.8106
Note: the corresponding Z(t) values hold for all countries; 1% critical value = 3.430, 5% critical value = -2.860, 10% critical value = -2.570. Absolute numbers hold for this test. All variables have 1268 observations.
The proxy for redenomination risk, the rolling R-squared of the regression between the euro and yen bond (as explained in methodology section 3.2), is shown in graph 6. The redenomination risk proxy fluctuates over the time period, however, certain areas can be distinguished in which the R-squared was close to zero, indicating a high redenomination risk. One such time period runs from the end of 2012 till the half of 2013. The high fluctuations of the proxy can be due to different reactions to certain events, however, it remains hard to determine what these are. For details on the spot and forward exchange rates used, please consult the appendix, graph A14 and table A3.
27
Graph 6. The rolling R-squared statistic between the yen and euro short-term bonds (proxy for redenomination risk) with a window of 30 data points which is approximately one month, from 1/8/2009 – 31/8/2015
Table 4. Summary statistics of the redenomination risk proxy
Mean Standard deviation Minimum Maximum
R-squared 0.089 0.150 0.000 0.979
Concerning the government deficits of the specific countries, there are large differences as depicted by graph 7. The country with the highest government deficit is Greece, which should not come as a surprise, as its government bond yields are significantly higher than the bond yields of the other countries. Furthermore, a tendency towards increasing government deficits is visible for most countries. The Maastricht treaty, in which the Stability and Growth Pact was established, limits government deficits as a percentage of GDP to 60%. As shown by graph 7, none of the countries were able to adhere to this rule in 2015, while about 5 countries did so in 2009. The table below (table 5) with summary statistics of the government deficit data confirms that 5 countries have had a government deficit below 60 percent during this time period. For the maximum, none of the countries are below 60 percent.
0 .2 .4 .6 .8 1 R -sq u a re d 7/1/2009 1/1/2011 7/1/2012 1/1/2014 7/1/2015 Date
28
Graph 7. The Government deficit of the 13 Eurozone countries as a percentage of GDP over 1/8/2009 – 31/8/2015 on a quarterly basis.
Regarding the liquidity premium, there are large fluctuations over time, with very few periods that are clearly distinguishable (graph 8). This might be due to the fact that ever since the financial crisis of 2007-2008 investors have diversified towards bonds, increasing the liquidity in the bond market in general since that time. A high liquidity premium indicates that there is a large difference between the KfW and the Bund, the Bund being the most liquid of them all generally and thus having the highest market value as investors are willing to pay more for this high liquidity. The graph shows that on numerous occasions, the KfW bond traded at a higher market price than the Bund (indicated by a negative spread), this is against expectations. The bonds are fairly similar in maturity and both have no coupon, thus those two factors could not cause this negative difference.
Table 5. Summary statistics of government deficit (% of GDP)
Government Mean Standard Deviation Minimum Maximum
Austria 82.178 2.086 73.484 86.404 Belgium 105.398 3.351 99.462 110.932 Finland 51.346 6.176 37.953 62.360 France 88.641 5.888 77.444 97.659 Germany 76.825 2.853 71.876 81.007 0 50 1 00 1 50 2 00 Pe rce n ta ge 7/1/2009 1/1/2011 7/1/2012 1/1/2014 7/1/2015 Date
Austria Belgium Germany Spain
Finland France Greece Ireland
Italy Malta Netherlands Portugal
Slovenia
29 -5 0 5 10 Sp re a d Kf W - Bu n d 01/07/2009 01/01/2011 01/07/2012 01/01/2014 01/07/2015 Date
Liquidity Premium
Greece 157.281 16.316 124.833 177.446 Ireland 103.332 18.895 59.240 125.331 Italy 123.255 7.781 112.462 135.961 Malta 69.598 2.000 66.933 74.752 Netherlands 63.793 4.287 56.493 69.247 Portugal 115.132 17.241 80.170 132.761 Slovenia 55.335 16.528 34.280 81.792 Spain 78.302 17.160 49.041 99.848Graph 8. Movement of the proxy for liquidity premium (spread between KfW bond and German Bund) from 1/8/2009 – 31/8/2015
Table 6. Summary statistics of the Liquidity premium and Risk aversion
Mean Standard deviation Minimum Maximum
Liquidity Premium 2.158 2.416 -4.829 9.56
Risk Aversion 2.523 0.297 1.857 3.601
30
Graph 9 shows the risk aversion in the market; a clear pattern is seen over time in which the periods of investors becoming more risk averse are clearly shown. For example, from the beginning of the dataset we see that the proxy for risk aversion, which is the spread between the yield of US Corporate BBB-rated bonds and the US government bond, decreased but fluctuated heavily, indicating tension in the market. When investors become more risk averse, they will want to sell their corporate BBB-rated bonds and buy government bonds as the latter are presumably less risky. This decreases the corporate BBB-rated bonds prices and increases the yield while it increases the government bond price and decreases its yield. For example, an increase in risk aversion is seen between January 2011 and around January 2012, then a small dip occurs in which investors become less risk averse, followed by an increase and then from about June/July 2012 a decrease in risk aversion is visible, which indicates that investors have regained trust in the market. Corresponding summary statistics are shown in table 6.
31 2 2 .5 3 3 .5 Sp re a d U S co rp o ra te BBB a n d g o ve rn me n t b o n d s 7/1/2009 1/1/2011 7/1/2012 1/1/2014 7/1/2015 Date
Risk Aversion
Graph 9. Movement of the proxy for risk aversion (spread between the US Corporate BBB-rated bonds and the US government bond) from 1/8/2009 – 31/8/2015
Table 7. Test results for the Dickey-Fuller unit root test on country government bond yields
Country Z(t) Statistic MacKinnon approx. P-value
Austria 0.239 0.9744 Belgium -0.408 0.9088 Finland 0.235 0.9741 France 0.105 0.9664 Germany -0.254 0.9317 Greece -0.270 0.9297 Ireland -1.009 0.7498 Italy -1.333 0.6141 Malta 0.204 0.9725 Netherlands -0.023 0.9565 Portugal -1.219 0.6655 Slovenia -0.049 0.9543 Spain -0.820 0.8130
32 iii. Empirical results
In table 8 the OLS regression results are shown. What immediately can be noticed are the extremely low R-squared; this is unexpected as most independent variables are commonly used in the literature to explain changes in government bond yields.
The redenomination risk proxy is not significant for any country, therefore, we can conclude that either this measure is not adequate or there is no redenomination risk present within the Eurozone. We are thus unable to give a clear answer to the research question as the results of the data cannot be guaranteed with certainty. It is remarkable that the results are not even significant for Greece, which could indicate that Greece does not plan on redenominating its debt, even while its new government (the Syriza party, see Table A1 in the appendix for more information) advocated to step out of the Eurozone.
33
Table 8. Results of OLS regression. Dependent variable is the first differences government bond yield. Time period from 1/8/2009 – 31/8/2015.
Austria Belgium Finland France Germany Greece Ireland Italy Malta Netherlands Portugal Slovenia Spain Redenomination Risk 0.007 (0.010) 0.008 (0.011) 0.003 (0.010) 0.003 (0.010) 0.001 (0.010) 0.039 (0.103) -0.022 (0.025) 0.010 (0.016) -0.005 (0.008) -0.003 (0.010) -0.019 (0.038) -0.021 (0.025) 0.011 (0.018) Liquidity Premium -0.000 (0.001) -0.001* (0.001) -0.000 (0.001) -0.001 (0.001) -0.000 (0.001) -0.006 (0.006) -0.003 (0.002) -0.001 (0.001) -0.000 (0.001) -0.000 (0.001) -0.001 (0.002) -0.001 (0.002) -0.002 (0.001) Risk Aversion -0.004 (0.005) -0.010* (0.005) -0.003 (0.005) -0.008* (0.004) 0.001 (0.005) 0.032 (0.048) -0.030*** (0.012) -0.016** (0.008) -0.003 (0.004) -0.003 (0.004) -0.038** (0.017) -0.005 (0.011) -0.009 (0.008) Government Deficit 0.000 (0.001) -0.001 (0.001) -0.000 (0.000) -0.001 (0.000) 0.001* (0.000) 0.003** (0.001) -0.001** (0.000) 0.000 (0.001) -0.000 (0.001) -0.001 (0.001) -0.001** (0.001) -0.000 (0.000) 0.000 (0.000) OMT -0.002 (0.003) -0.002 (0.005) 0.001 (0.005) 0.002 (0.005) -0.000 (0.003) -0.149*** (0.041) -0.002 (0.010) -0.014 (0.010) -0.002 (0.003) 0.004 (0.006) 0.021 (0.020) -0.007 (0.012) -0.019* (0.011) Constant -0.001 (0.060) 0.103 (0.082) 0.012 (0.026) 0.068* (0.039) -0.074* (0.042) -0.379 (0.237) 0.144*** (0.038) 0.013 (0.086) 0.021 (0.048) 0.046 (0.046) 0.236*** (0.072) 0.022 (0.036) 0.020 (0.031) Observations 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 R-squared 0.0017 0.0063 0.0005 0.0047 0.0030 0.0114 0.0152 0.0075 0.0017 0.0013 0.0099 0.0021 0.0075 Adj. R-squared -0.0024 0.0023 -0.0036 0.0007 -0.0011 0.0074 0.0112 0.0035 -0.0024 -0.0028 0.0059 -0.0020 0.0035
Note: standard errors in parentheses. * indicates significance at the 10% level, ** is significance at the 5% level, *** is significance at the 1% level.
Greece, Ireland and Portugal are the countries with the most significant results. The government deficit is significant at the 5 percent level for all three countries. In case of Greece, an increase in the government deficit (as a % of GDP) increases government bond yields by 0.003 percentage points. An increase in government deficit has a positive effect on the bond yields of Ireland and Portugal as these will go down with 0.001 percentage points.
34
Bond yields of Italy are also negatively affected by an increase of risk aversion, leasing to a decrease of 0.016 percentage points (significant at the 5 percent level). This seems odd as these three countries have all been indicated as having weak fiscal fundamentals. According to De Santis (2012) this would result in increased contagion risk, hence, it would be expected that their bond yields would move upward when risk aversion increases among investors. Increased contagion risk leads to higher risk of default, and in times of increased risk aversion, investors disinvest in high risk bonds, hence prices move down and yields of risky bonds move upward.
One would expect countries to benefit from increased risk aversion by decreasing yields as the government bonds of these countries are often perceived as low risk, and countries with weak fundamentals and thus higher risk to increase yields. One argument that could explain this phenomenon is that investors diversified away from other investment instruments to bonds, but still demanding a high return on their investments, increasing the demand on bonds of countries with weaker fundamentals that in the first place gave higher yields, but are still a relatively safer investment than other non-government bond instruments. It is important to note that the risk aversion variable is not significant for Germany, this contradicts findings of De Santis (2012) who found that increased risk aversion led to increased demand in the German Bund. Furthermore, we are able to reject the null hypothesis that the constant is zero at the 1 percent level for both Ireland and Portugal.
The OMT program of the ECB has been highly effective for Greece, as government bond yields have reduced by 0.149 percentage points after its introduction. The OMT program has also been effective for Spain (significant at the 10 percent level) by reducing bond yields with 0.019 percentage points. This finding corresponds to the findings of Krishnamurthy et al. (2013).
35
obtained funds from the IMF and the ECB. The government bond yields of both Ireland and Portugal will fall by 0.001 percentage point when the country’s government deficit increases with 1 percentage point (both significant at the 5 percent level). Apparently the market mechanism is not at work here; investors do not punish these countries for worsening government deficits (de Grauwe, 2012a).
The liquidity premium is only significant for Belgium at the 10 percent level. This can be due to the fact that De Santis (2012) has never specified which KfW bond and German Bund he has taken as a proxy for the liquidity premium. In this research, one of the most similar pairs for which data was available has been taken. However, these bonds can differ from the ones used by De Santis (2012), explaining that only for Belgium the liquidity premium variable is significant. A 1 point increase in the liquidity premium decreases the Belgian government bond yield by 0.001 percentage point. In other words, when investors start to value liquidity more, they are willing to pay more for Belgian government bonds, hence an increasing price and decreasing yield. Overall, there is a negative relationship between the liquidity premium and government bond yields, as expected. When investors start to request higher liquidity (and hence are willing to pay a premium for it), they move towards the government bond market, as this is known for being highly liquid. More demand in the same market increases prices and reduces the yields.
Finally, the adjusted R-squared of the regression is very small, indicating a low explanative power of the regression. There have to be other variables outside of the regression that explain the increasing bond yields. For Austria, Finland, Malta, the Netherlands and Slovenia none of the variables are significant. Therefore, the results of these countries have not been touched upon specifically.
iv. Robustness checks
36
correct beta’s, the White's heteroscedasticity consistent standard error estimates have to be taken, as taking logs of the data is not an option as some of the data turns negative from time to time. Table 10 shows both test results for the Breusch-Godfrey LM test and the Durbin’s alternative test for autocorrelation within the data. The Breusch-Godfrey LM test has 5 time lags and the Durbin’s alternative test has 1 time lag, this is the main difference between the two tests. When looking at the test results, the prob>chi2 for most of the countries is below 0.05, making it possible to reject the null hypothesis of no serial correlation between the errors. In most cases we have to deal with autocorrelation.
As the Breusch-Pagan / Cook-Weisberg test for heteroscedasticity (table 9) indicated heteroscedasticity is present in the data and the latest two tests for autocorrelation (table 10) show positive test results, the regressions should be run with a Newey-West procedure. This procedure gives heteroscedasticity and autocorrelation consistent (HAC) standard errors, thus correcting for the heteroscedasticity and autocorrelation. This makes the outcome of the regression reliable. When not dealing with possible autocorrelation and heteroscedasticity, the standard errors obtained from the regression could be inappropriate. Interpretation of the data could thus be untrue (Brooks, 2008).
Table 9. Test results of the Breusch-Pagan / Cook-Weisberg test for heteroscedasticity for data on government bond yields (first differences).
Country: Chi2(1) Prob > chi2
Austria 4.11 0.0425 Belgium 6.06 0.0138 Finland 7.61 0.0058 France 28.74 0.0000 Germany 48.03 0.0000 Greece 635.18 0.0000 Ireland 45.53 0.0000 Italy 50.89 0.0000 Malta 2.54 0.0110 Netherlands 10.42 0.0012 Portugal 187.24 0.0000 Slovenia 7.45 0.0063 Spain 15.05 0.0001
37
Table 10. Durbin’s alternative test for autocorrelation with 1 time lag and the Breusch-Godfrey LM test for autocorrelation with 5 time lags.
Country: Chi2 (1) Prob>chi2 (1) Chi2 (5) Prob>chi2 (5)
Austria 15.908 0.0001 19.197 0.0018 Belgium 84.726 0.0000 88.804 0.0000 Finland 2.931 0.0869 5.976 0.3086 France 14.497 0.0001 22.582 0.0004 Germany 1.161 0.2813 7.466 0.1882 Greece 0.568 0.4509 19.340 0.0017 Ireland 34.248 0.0000 54.553 0.0000 Italy 13.543 0.0002 41.797 0.0000 Malta 35.864 0.0000 40.063 0.0000 Netherlands 2.946 0.0861 7.511 0.1853 Portugal 38.911 0.0000 54.287 0.0000 Slovenia 20.821 0.0000 25.910 0.0001 Spain 41.931 0.0000 57.777 0.0000
38
Table 11. Results of the Newey-West OLS regression. Dependent variable is the first differences government bond yield. Time period from 1/8/2009 – 31/8/2015.
Austria Belgium Finland France Germany Greece Ireland Italy Malta Netherlands Portugal Slovenia Spain Redenomination Risk 0.007 (0.008) 0.008 (0.011) 0.003 (0.008) 0.003 (0.009) 0.001 (0.008) 0.039 (0.066) -0.022 (0.018) 0.010 (0.011) -0.005 (0.006) -0.003 (0.007) -0.020 (0.028) -0.021 (0.022) 0.011 (0.013) Liquidity Premium -0.000 (0.001) -0.001 (0.001) -0.000 (0.001) -0.001 (0.001) -0.000 (0.001) -0.006 (0.006) -0.003 (0.002) -0.001 (0.001) -0.000 (0.000) -0.000 (0.001) -0.001 (0.002) -0.001 (0.001) -0.002 (0.001) Risk Aversion -0.004 (0.006) -0.010 (0.007) -0.003 (0.005) -0.008 (0.005) 0.001 (0.005) 0.032 (0.052) -0.030** (0.013) -0.016** (0.008) -0.003 (0.003) -0.003 (0.005) -0.038* (0.021) -0.005 (0.010) -0.009 (0.011) Government Deficit 0.000 (0.001) -0.001 (0.001) -0.000 (0.000) -0.001 (0.001) 0.001* (0.001) 0.003 (0.002) -0.001** (0.000) 0.000 (0.001) -0.000 (0.001) -0.001 (0.001) -0.001 (0.001) -0.000 (0.000) 0.000 (0.001) OMT -0.002 (0.003) -0.002 (0.005) 0.001 (0.006) 0.002 (0.006) -0.000 (0.003) -0.149** (0.058) -0.002 (0.014) -0.015 (0.012) -0.002 (0.002) 0.004 (0.008) 0.021 (0.036) -0.007 (0.013) -0.019 (0.021) Constant -0.001 (0.065) 0.103 (0.086) 0.012 (0.028) 0.068 (0.050) -0.074 (0.045) -0.379 (0.372) 0.144*** (0.042) 0.013 (0.089) 0.021 (0.044) 0.046 (0.054) 0.236** (0.116) 0.022 (0.030) 0.020 (0.039) Observations 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 1233 R-squared 0.0017 0.0063 0.0005 0.0047 0.0030 0.0114 0.0152 0.0075 0.0017 0.0013 0.0099 0.0021 0.0075 Adj. R-squared -0.0024 0.0023 -0.0036 0.0007 -0.0011 0.0074 0.0112 0.0035 -0.0024 -0.0028 0.0059 -0.0020 0.0035
Note: standard errors in parentheses. * indicates significance at the 10% level, ** is significance at the 5% level, *** is significance at the 1% level. Newey-west lag taken is 5.
39
The Spanish government bond yield is no longer influenced by the introduction of the Outright Monetary Transactions program. The significance of Greece’s reaction to the OMT program has decreased from the 1 percent level to the 5 percent level. Austria, Belgium, Finland, France, Malta, the Netherlands, Slovenia and Spain do not have any significant variables and therefore their results are not described in the text.
In table 12, the empirical results of another robustness check are presented. Graph 5 showed the yields to maturity of the different Greek government bond yields. The Yen-denominated short-term government bond showed a large increase in yield to maturity since the start of 2015. It would therefore be wise to test whether this large increase has had a significant impact upon the empirical results presented in table 8. By taking out all data after the 31st of December 2014, this large increase in yield to maturity is omitted.
40
Table 12. Results of the Newey-West OLS regression. Dependent variable is the first differences government bond yield. Time period from 1/8/2009 till 31/12/2014.
Austria Belgium Finland France Germany Greece Ireland Italy Malta Netherlands Portugal Slovenia Spain Redenomination Risk 0.007 (0.009) 0.010 (0.091) 0.004 (0.009) 0.005 (0.009) 0.004 (0.008) 0.029 (0.068) -0.020 (0.019) 0.013 (0.012) -0.007 (0.006) -0.001 (0.008) -0.019 (0.031) -0.021 (0.023) 0.014 (0.014) Liquidity Premium -0.000 (0.001) -0.001 (0.001) 0.000 (0.001) -0.000 (0.001) -0.000 (0.001) -0.006 (0.007) -0.003 (0.002) -0.001 (0.001) 0.000 (0.001) -0.000 (0.001) -0.001 (0.002) -0.001 (0.001) -0.001 (0.001) Risk Aversion -0.004 (0.006) -0.008 (0.007) -0.002 (0.005) -0.007 (0.005) 0.002 (0.005) 0.030 (0.052) -0.029** (0.014) -0.016** (0.008) -0.002 (0.003) -0.002 (0.005) -0.039* (0.022) -0.004 (0.010) -0.009 (0.011) Government Deficit -0.000 (0.001) -0.001 (0.001) -0.000 (0.001) -0.001 (0.001) 0.001** (0.001) 0.003 (0.002) -0.001* (0.000) 0.000 (0.001) 0.000 (0.001) -0.000 (0.001) -0.001 (0.001) -0.000 (0.000) 0.000 (0.001) OMT -0.002 (0.003) -0.002 (0.005) 0.003 (0.007) 0.003 (0.007) -0.001 (0.003) -0.147** (0.058) -0.002 (0.015) -0.013 (0.013) -0.004* (0.002) 0.002 (0.008) 0.020 (0.036) -0.006 (0.014) -0.018 (0.021) Constant 0.042 (0.078) 0.094 (0.091) 0.022 (0.032) 0.078 (0.054) -0.097** (0.046) -0.370 (0.374) 0.144*** (0.044) 0.025 (0.100) -0.019 (0.044) 0.029 (0.056) 0.236** (0.118) 0.020 (0.031) 0.020 (0.041) Observations 1130 1130 1130 1130 1130 1130 1130 1130 1130 1130 1130 1130 1130 R-squared 0.0015 0.0058 0.0011 0.0051 0.0046 0.0118 0.0150 0.0079 0.0024 0.0006 0.0101 0.0020 0.0080 Adj. R-squared -0.0030 0.0014 -0.0033 0.0007 0.0002 0.0074 0.01056 0.0035 -0.0020 -0.0039 0.0057 -0.0025 0.0036
Note: standard errors in parentheses. * indicates significance at the 10% level, ** is significance at the 5% level, *** is significance at the 1% level. Newey-west lag taken is 5.
41
program. The significant result for Malta could thus be coincidence. For Austria, Belgium, Finland, the Netherlands and Spain none of the variables are significant.
Table 13. Results of the Newey-West OLS regression. Dependent variable is the first differences government bond yield. Time period from 1/8/2009 – 31/8/2015 on a weekly basis.
Austria Belgium Finland France Germany Greece Ireland Italy Malta Netherlands Portugal Slovenia Spain Redenomination Risk 0.020 (0.036) 0.035 (0.048) 0.014 (0.030) 0.016 (0.039) 0.008 (0.030) 0.496 (0.439) -0.069 (0.083) 0.028 (0.066) -0.020 (0.031) -0.014 (0.028) -0.117 (0.102) -0.044 (0.098) -0.012 (0.060) Liquidity Premium -0.007** (0.003) -0.007** (0.003) -0.009*** (0.002) -0.007*** (0.003) -0.010*** (0.003) 0.020 (0.033) -0.003 (0.005) -0.000 (0.004) -0.006*** (0.002) -0.009*** (0.002) 0.004 (0.010) -0.000 (0.005) -0.003 (0.005) Risk Aversion -0.040** (0.020) -0.053** (0.021) -0.042** (0.017) -0.035* (0.018) -0.029* (0.016) 0.084 (0.186) -0.143*** (0.047) -0.041 (0.034) -0.022* (0.013) -0.038** (0.016) -0.119* (0.061) 0.002 (0.037) -0.035 (0.036) Government Deficit 0.001 (0.003) -0.003 (0.003) -0.002 (0.002) -0.001 (0.002) 0.003 (0.002) 0.018*** (0.006) -0.002 (0.001) 0.001 (0.003) -0.002 (0.002) -0.003 (0.003) -0.001 (0.002) -0.001 (0.001) 0.002 (0.001) OMT -0.012 (0.011) -0.009 (0.017) 0.007 (0.026) -0.003 (0.021) -0.010 (0.011) -0.590*** (0.198) -0.010 (0.062) -0.055 (0.047) -0.021** (0.008) 0.012 (0.030) -0.059 (0.105) -0.007 (0.052) -0.111** (0.049) Constant 0.052 (0.244) 0.454 (0.298) 0.197* (0.103) 0.217 (0.157) -0.153 (0.162) -2.801*** (1.043) 0.588*** (0.167) -0.036 (0.352) 0.205 (0.176) 0.298 (0.195) 0.452 (0.279) 0.047 (0.111) 0.007 (0.142) Observations 308 308 308 308 308 308 308 308 308 308 308 308 308 R-squared 0.0409 0.0359 0.0560 0.0414 0.0653 0.0294 0.0451 0.0211 0.0636 0.0627 0.0241 0.0108 0.0317 Adj. R-squared 0.0250 0.0199 0.0404 0.0256 0.0499 0.0133 0.0293 0.0049 0.0480 0.0472 0.0080 -0.0056 0.0156
42
The empirical results of the last robustness check are presented in table 13. Here, the data is transformed into weekly data, by taking the average of the five observations of that week. When testing for heteroscedasticity and autocorrelation, the tests for autocorrelation were positive thus the Newey-West standard errors are still applicable. Firstly, the R-squared and adjusted R-squared have increased hugely in comparison to the first empirical results of table 8. Independent variables converted into weekly data thus seem to be better in explaining government bond yields.
The redenomination risk proxy unfortunately still remains insignificant for all countries. However, the liquidity premium has become highly significant for a number of countries, namely, for Finland, France, Germany, Malta, the Netherlands at the 1 percent level and for Austria and Belgium at the 5 percent level. For all of these countries, the government bond yield falls when the liquidity premium increases. The decrease in government bond yield is between 0.006 and 0.010 percentage points. This also holds for all other non-significant countries except for Greece and Portugal. There has been a large shift in this variable, as in the first OLS regression, only Belgium was significant at the 10 percent level.
43
its government bond yields when the risk averseness of investors increases. The explanation for Slovenia remains unclear.
When running the regression with weekly data, the government deficit is only significant for Greece; a 1 percentage point increase in government deficit (as a percentage of GDP) leads to a 0.018 percentage point increase in the Greek government bond yield. Hence, Greece is punished by investors for its unsound fiscal policies by investors. The government deficits of Germany, Ireland and Portugal are no longer significant.
Lastly, the OMT variable remains significant for Greece at the 1 percent level. The program has led to a 0.590 percentage point decrease in Greek government bond yields. Furthermore, the variable also remains significant at the 5 percent level for Malta and Spain, with a reduction of 0.021 percentage points and 0.111 percentage points respectively. The influence of the OMT on Malta has not been significant in the initial regression, but the link has also been indicated in the robustness regression with a shorter time period (table 12).
Overall, the redenomination risk proxy remains insignificant throughout the robustness checks. We are thus not able to answer the research question nor reject/fail to reject our hypothesis. This measure for redenomination risk might either not be adequate or there are other flaws within the research. The adjusted R-squared, and thus the explanative power of the regression, has been low throughout the research, while it increased when bundling the data into weekly points. It might be true that this dataset would lend itself better for regressions on a weekly or monthly basis, however, it seems wasteful to reduce such a rich daily dataset. There seem to be other variables outside of the regression that explain the increasing bond yields.
