The Cost of Sovereign Default: The Impact of the Eurozone Debt Crisis on Domestic Stock Markets
An Event study
ROSECARMEN PIERRE LOUIS Student Nr: s1965646
February 2013
ABSTRACT
This paper documents the possible sovereign default costs associated with the Eurozone countries confronting sovereign debt crises. Stock market data of 219 companies from Portugal, Ireland, Italy, Greece and Spain (Acronym PIIGS) are analyzed to determine the impact of bail-out announcements on stock returns and liquidity measurements. The findings show that stock markets are mostly affected in the first 20 days following bail-out announcements. They indicate that Greece’s first bail-out announcement might have caused strong anticipation of default by the remaining four countries. Furthermore, the results confirm the Collateral view of market illiquidity during crises. Overall, this paper also documents that countries with relatively better fiscal or economic conditions are more negatively affected by the crisis than countries with relatively worse fiscal or economic conditions.
Keywords: sovereign default cost, abnormal returns, liquidity JEL-classification: G0, G1
1. Introduction
Throughout history many governments have defaulted on their domestic and external debts, leading their countries into severe debt crises. These types of crises, the consequence of sovereign default, have been associated with a variety of costs.
Macroeconomic studies have used a range of variables to identify the potential costs that defaulting governments might cause their countries to suffer. Some have used foreign direct investment (FDI) to look whether creditors’ countries shut their doors to defaulters’ investment abroad (Fuentes and Saravia, 2009). Other studies have analysed factors such as GDP, commodity prices, country short-and long-term interest rates, with the purpose to explore possible negative effects of sovereign default on these variables (Alfaro and Kanczuk, 2005; Boreinszten and Panizza, 2009; De Paoli et. al, 2006;
UNIVERSITY OF GRONINGEN, FACULTY OF ECONOMIC AND BUSINESS. DEPARTMENT OF FINANCE.The author would like to take
Reinhart and Roggof, 2008). From a microeconomic perspective, the impact on variables such as sovereign risk premium in credit default swaps, returns on asset prices, volume on assets traded, bid-ask spread on assets traded and the price reaction to trading, have been researched, to test the potential cost of sovereign default (Boreinsztren and Panizza, 2009; Chordia et. Al, 2005; Forbes, 2004; Hameed et. al 2010; Noy, 2008; Van Horen et. Al, 2008). However, despite all of the studies conducted, the cost of sovereign default on defaulters’ countries is still difficult to quantify.
Specifically, the cost of sovereign default on defaulters’ stock markets has not been given much attention.
Sovereign default can be defined as: “The failure of governments to meet a principal or interest payment on domestic or external debt on the due date (or within a specified grace period), or the rescheduling/restructuring of debt into terms less favorable to the lender than those in the original contract” (Reinhart and Rogoff, 2008).
In May 2010, the Greek government was officially bailed-out on its external and domestic debts with an 110bn euros rescue package. This bail-out acceptance followed an extended denial process of financial need by the Greek government. The high increase in the country’s debt level made it difficult for the government to service its loan. In March 2012, a second bailout package of 130bn euros was granted by the European partner states and the IMF to Greece. On top of these loans, the majority of Greece’s creditors agreed to write off more than half of the debts owed to them by Athens1. In addition, they agreed to replace existing loans with new loans at a lower interest rate. Following Greece’s first bail-out, other financially distressed countries in the Eurozone–Ireland, Italy, Portugal and Spain (acronym IIPS), have also been acknowledging their own need for monetary aid. In this paper, based on the given definition for sovereign default, a bail-out is also considered a debt default.
The abovementioned observations have motivated further investigation regarding the possible costs that sovereign default can impose on the secondary financial market of defaulting countries. Recently a number of studies have shared the view that sovereign default can be costly to defaulters’ financial systems. Some papers look at the impact of sovereign default on banks (Yeyati and Panizza, 2011). The premise is that sovereign default may weaken domestic banks, weakening their role as providers of liquidity and credit to the economy. Others have elaborated on contagion
1Eurozone crisis explained 2012, retrieved from http://www.bbc.co.uk/news/business-13798000 October 1, 2012
effects, emphasizing how default caused elsewhere might affect firms around the world (Forbes (2004). However, the majority focuses only on emerging markets. Thus it is still not completely apparent how sovereign default may be costly to developed countries’
financial systems. More importantly, the magnitude of the possible cost of sovereign default on the aforementioned Eurozone countries’ stock markets has not been investigated.
The aim of this study is to investigate the cost of sovereign default on the stock markets of Portugal, Ireland, Italy, Greece and Spain (acronym PIIGS). This is done by specifically analyzing investors’ reaction to sovereign bail-out announcements. Different scenarios are considered to analyze investors’ reaction. Firstly, I study whether Greece’s bail-out announcement might have caused possible default anticipation by IIPS investors as suggested in other studies (Boreinszten and Panizza, 2009; Levy-Yeyati and Panizza, 2011). This is because a Greek bail-out acceptance may have caused IIPS’
investors to anticipate a default by IIPS’ governments. Since a bail-out announcement is news, it is assumed to reflect symmetric information. This implies that investors’
reactions to such news might be related. It also suggests that due to the anticipation of default, IIPS’ investors might have taken speculative positions, causing them to start selling their assets, putting the stock market under severe stress. Throughout this paper, I refer to this anticipation process as“the anticipation effect”.
Two consecutive scenarios are considered to analyze investors’ reaction. A second scenario analyzes the specific public bail-out announcement dates for PIIGS separately. Here I look at the specific dates on which a bail-out package has been publicly announced to each country. A third approach consists of attempting to observe whether countries’ macroeconomic or fiscal conditions determine how these countries’
stock markets are affected by the crisis.
To carry out the analyses, the particular definition of a liquid market must be designated. From Amihud (2002), Bank for International Settlements (1999) and Chordia et. al (2005), a liquid market can be defined asa market where participants can rapidly execute large-volume transactions at low cost.
To identify the cost of sovereign default on the stock markets of PIIGS, I employ the insights from the overall literature discussed in this paper and fundamentals regarding negative shocks to the economy. In this paper, a public bail-out
This negative shock is likely to have a negative impact on asset prices (The first cost to be analyzed in this paper). This loss in asset value might be due to an asset recomposition effect. This effect originates from the foundational work of Kiyotaki and Moore (1997) and is referred to as theCollateral view. It is used by several authors to relate liquidity with market returns. According to the Collateral view theory, market participants obtain funding by pledging assets as collateral. During bad economic times, the liquidation value of the collateral can decrease, as potential buyers also confront difficult times. Because of the decrease in collateral value, debt capacity also decreases.
This in turn, reinforces the fall of the collateral price as potential buyers become even more cash constrained.
During crises periods, market participants can hit their margin constraints (Since asset prices decrease and investors are demanded to deposit more cash or securities to cover losses) and might be forced to liquidate. This, as stated by Van Horen et. al (2008), can induce a wider cost of trading. This cost will make it difficult to provide liquidity exactly when the market needs it. According to Van Horen et. al (2008), net withdrawals are a function of intermediaries’ performance and market liquidity is closely related to intermediaries’ funding needs. A drop in asset value will cause a decrease in short term inflow of funds. This effect forces financial intermediaries to sell, which adds to the price downturn and also generates a spiral fall in some liquidity measures. The liquidity impact is the second cost I disclose in this paper.
Liquidity and price impact are determined by using the famous model of Brown and Warner (1980), which allows for measuring the impact of events on securities. With this methodology, I try to empirically observe whether the sovereign debt crisis in the Eurozone caused investors to reassess their trading behavior. The observed abnormal values are used to conduct simple regressions to analyze the co-movements of these variables during the crisis period. Besides these analyses, the application of this methodology helps to answer key questions: Did the Greece’s official acceptance of bail- out cause anticipation effect on IIPS’ stock markets? Does this impact differ from specific public bail-out announcement of PIIGS? If so, what is the magnitude of this effect and how long does it last? Another important question addressed is: Does the economic or fiscal condition of a country determine how this country is affected by a sovereign debt crisis?
Based on the overall literature, which will follow in the second section, cost expectations are formed. First, an acceptance of bail-out package by Greece’s
government is expected to cause an anticipation of default by IIPS’ government, since these are also in financial distress. This anticipation effect will be stronger (as captured by the negative impact on the abnormal values of the proxies) than official public bailout announcement effect. Also, abnormal returns are expected to be negative, with positive abnormal increases in the liquidity proxies. In addition, volume traded and costs of transacting are expected to be positively related, suggesting illiquidity.
Furthermore, since crises mark departure from equilibrium, the expectation is that illiquidity is associated with lower returns.
Daily stock market data of the five above-mentioned Eurozone countries are gathered from the period of March 2010 to June 2012. From Datastream, information on 219 companies is gathered, comprised of data on returns, volumes traded, bid-ask spreads and prices. The data is used to construct four measurements: Abnormal returns, abnormal volume traded and abnormal cost of trading (Two proxies are used for cost). Volume traded is used as measurement for trading activity. Bid-ask spread and the Amihud ratio are used as measurements for trading cost. The Amihud ratio measures the price response to trading (Further explained in the methodology section).
Both trading activity and trading cost proxies are used as liquidity measurements as applied by Van Horen et. al (2008). In addition, macroeconomic data on total government debt, primary deficits, deficit ratios and growth rates are gathered from the period of 2007 to 2011. These are gathered from the World Bank database and the IMF database.
Results show that the Greek’s bail-out announcement might have caused IIPS’
investors to reassess their portfolio. This is reported in significant negative abnormal returns and significant abnormal increase in volume traded and cost of trading. More strikingly, the abnormal values for anticipation effect reveal stronger values and test statistics than for public bail-out announcements. In general the result related to the anticipation effect is consistent with the statement of Panizza and Yeyati (2011) who suggest that default marks the recovery of a crisis. This confirms the view that policy makers delay default, leading to anticipation of default. Another implication is that the market indeed prices the risk of default as suggested by Ardagna (2005) and Reinhart et. al (2003).
The findings also show that the impact of bail-out on the abnormal proxies is
spread and the Amihud ratio in the first 20 days following announcement. The impact decreases drastically in the middle of the event. At the end of the event, the values are still significant, but less pronounced than in the beginning. This suggests that the impact of bail-out is short-lived. Furthermore the findings indicate that countries with relatively better economic or fiscal conditions, suffer a higher cost (in terms of higher negative impact on the abnormal values) than countries with worse economic or fiscal conditions.
Results also indicate that markets are illiquid. An increase in volume traded is significantly related to an increase in cost of trading. This confirms the Collateral view and the expectation that investors, constrained by liquidity, liquidate their positions, thereby engaging in high volume of selling with high cost of transaction (Van Horen et.
al, 2008).
Furthermore, the findings report an ambiguous relationship between returns and illiquidity. For countries with relatively better fiscal or economic conditions, investors seem to be indifferent regarding the expectation of higher returns to compensate for illiquidity. This is however not true for countries with relatively poor economic conditions during the crisis. In these countries, the relationship between returns and illiquidity is significantly positive, confirming the theory of Amihud and Mendelson (1991b). This result is surprising, considering that Amihud and Mendelson predict this relationship to be present in equilibrium. Notwithstanding this contradictory finding, the overall results from this paper suggest that defaulters’ stock markets with relatively better economic conditions suffer a higher cost than countries with relatively worse economic conditions.
The findings contribute to the existing literature that attempts to determine the cost of sovereign default on domestic financial systems. Specifically, the paper gives new insight into how sovereign bail-out in developed countries transmits to the stock market through its impact on returns and liquidity. In addition, the findings confirm the Collateral view, that during crises, investors reorganize their portfolios, because of liquidity need. Furthermore, they show the degree of crises’ impact under different macroeconomic conditions of each country. This is important, because how a crisis is transmitted in a country, might depend on the fiscal economic characteristics of that country.
The remainder of this paper is structured as follows: Section 2 elaborates on the theory behind the cost of sovereign default. Section 3 follows with data, containing
preliminary descriptive statistics of the findings. In section 4, the methodology is explained, followed by section 5 reporting the main event study results and regression analyses results. Section 6 presents conclusions and discussions. Section 7 proposes areas for further research.
2. Literature Review
The cost of sovereign default has been investigated for decades by many researchers in order to come up with acceptable evidence of its impact on the economy. Macro- economic studies have reported several ways in which countries can be penalized when sovereigns stop servicing their debt obligations. De Paoli et. al (2006), Tomz and Wright (2007) investigate the relationship between default and output performance.
They find a negative relationship between the two. Specifically, De Paoli et. al (2006) shows that the fall in GDP is larger when combined with currency and banking crises.
On the other hand, Panizza and Yeyati (2011), using quarterly data, find that growth rate in the post-default period is never significantly lower than in normal times and that output contractions precede default episodes. According to them, the indication of an association between output losses and sovereign defaults is based on annual observations and suffers from measurement and identification problems. This result highlights the need for more in depth data analyses by using more micro data.
Default cost has also been related with the exclusion from future borrowing or with the exclusion from the financial market. Panizza and Yeyati (2011) argue that over the long run, default may exert its impact either through lower investments or reduced access to capital markets. In addition, Bulow and Rogoff (1989) reason that a country which defaults on its debt contract may no longer be able to borrow for domestic investment. According to De Paoli et. al (2006), exclusion from future borrowing could be considered as an incentive for debt repayment. However, she also argues that the empirical evidence suggests default is not necessarily related to a loss in market access, since other creditors are readily available to provide credit to defaulters.
The costs discussed above are a few of the penalties defaulters might suffer. Yet, these costs are not completely empirically proven. This evidence might be due to Panizza and Yeyati’s (2011) views who state that default is a turning point of a crisis,
economy. They conclude that this might be a result of the non- trivial costs of avoiding default. They also reason that most of consequences of default are normally reflected in the market before policymakers’ formal decision to default. Panizza and Yeyati’s (2011) reasoning is that policymakers might avoid default since the post-default cost might not outweigh fiscal effort to service debt. Thus the formal decision to default does not bring any additional cost and policymakers will then delay default; there is no tradeoff and default is then optimal. They conclude that a negative effect on the country’s economy is likely to be driven by the anticipation of default. This anticipation is independent of the country’s eventual decision to validate default.
As mentioned in the introduction of this paper, the anticipation of default might have an impact on the behavior of investors. An earlier research regarding the relationship between anticipation of default and GDP was conducted by Boreinszten and Panizza (2009). They split default into two dummy components to estimate proxies for anticipated and unanticipated default to measure their relationships with growth.
They find that the anticipation of default carry higher cost than unanticipated default.
Another paper by Bolton and Jane (2011) sheds insight on how anticipation of default might be manifested in other markets through investors’ perceptions. They analyze the co-movements of credit default swaps among the Eurozone countries in debt crisis. From their study, one can derive that the delaying of bail-out acceptance by Greece was already visible by drastic increase in Greek’s credit default swaps rate. This indicates that market participants might have anticipated the default of Greece. Later on when Greece posted a worse budget deficit, other Eurozone countries’ credit default swap rates also started to increase. When the official bail-out package was guaranteed to Greece, the equity market initially gave the impression that the effect on rates was abating. Nevertheless, the relief turned out to be temporary.
Consistent with the findings of Ardagna (2005), Bolton and Jane (2011) and Backe and Gardo (2012), one can derive that anticipation effect can transmit to a country’s financial system through its impact on investors’ expectations, confidence and risk preferences. Ardagna (2005) shows in her study that agents’ perception regarding fiscal adjustment can have a positive or a negative influence on a country’s economy.
Also Reinhart et al. (2003) suggest that citizens of financially distressed countries should be aware of the risk imposed on the market when their governments engage in heavy borrowings. Despite the fact that their study is based on emerging markets, it is this awareness, along with the anticipation effect, I attempt to disclose in this paper. A
Greece’s bail-out acceptance can be seen as a wake-up call, in the sense that it can potentially wake up IIPS’s investors to reassess the risk that default might impose on their stocks.
Besides the aforementioned research on credit default swaps, other recent microeconomic related studies have attempted to shed light on the impact of sovereign default on the financial system (Chordia et. al, 2005; Forbes, 2004; Panizza and Borensztein, 2009; Van Horen et. al, 2008). According to Panizza and Borensztein (2009), default episodes may cause a collapse in confidence in the domestic financial system and may lead to bank runs. The effect of bank runs may be banking crises or at least credit crunch. To test the presumed relationship between default and bank runs, they use an index for bank crises to investigate the possible effect of sovereign default on banks. Although they show that debt default seems to cause banking crises, they conclude that the cost of default is difficult to quantify. They attribute this to the mixed findings; default seems to have a small effect on credit ratings and bond spreads. They also find that government turnovers increase following debt crises. Also, the relationship between default and its effect on international trade credit and bank lending seems to be doubtful. The most striking and robust result is that the effect of default is short-term. Panizza and Borensztein (2009) suggest that this effect can never be detected beyond one or two years.
From a stock market perspective, Van Horen et. al (2008) apply event study methodology to analyze emerging market crises’ impact on liquidity. As liquidity measurements they apply volume traded, bid-ask spread and the price response to trading (as captured by the Amihud ratio for transaction cost) to test investors’
reactions during financial turmoil. They take different approaches to analyze liquidity impact. First, they typify crisis as sharp stock market downturns by using dummy variables to distinguish price decreases. These downturns typically include more than one event per country. Secondly, they regress each of the liquidity variables on a number of controls. A control variable includes a dummy variable that split prices into positive and negative changes to measure stock market downturns directly. Another control variable used is a separate dummy variable that split the data into early and late crisis dummy variables. This is to account for the fact that financial distress is often accompanied by intense but short-lived portfolio reallocation and fire sale. Their
The regression result of van Horen et. al (2008) with the control variables shows the following: Trading volume tends to remain stable at the beginning of crises. This however, declines later on, suggesting a drop in activity after reallocations have been completed. Price fluctuations are associated with an increase in volume traded. Thus, large price downturns seem to be associated with higher trading. The Amihud ratio remains stable in the beginning of the crisis, but decreases later on. However, they report an irregular pattern between price fluctuations and the Amihud ratio; price jumps are associated with lower Amihud ratio (greater liquidity) and price declines are associated with an increase in the ratio. This pattern is also confirmed in the behavior of the bid-ask spread. However the bid-ask spread barely reacts to price jumps.
An earlier research using the same event study concept as Van Horen et. al (2008)and the impact of government default on the financial system is conducted by Forbes (2004). He uses microeconomic approaches to examine how different companies located around the world are affected by the Asian and Russian financial crises. Even though his research was not specifically related to default impact on domestic markets, it still illustrates how stock markets can be affected by sovereign crises. His findings indicate that in general affected firms have on average, significant lower abnormal returns during the crisis. Moreover he reports significantly lower returns for more liquid stocks during the latter part of the Asian crisis than the rest of the sample. He suggests that investors caused crises to spread through a forced- portfolio recomposition effect for global investors. He explains that due to large cash withdrawals and margin calls, investor might be forced to sell assets in markets not directly affected by the crisis.
The theory of asset recomposition also drew the attention of Chordia et. al (2005), who investigated cross-market liquidity dynamics during the Asian and Russian crisis of 1998 and the bond market crises of 1994. Motivated by the concept of portfolio rebalancing needs of investors during crises, they conduct several analyses to shed light on more primitive liquidity drivers and to explore liquidity movements.
Chordia et. al (2005) reasoned that systemic shocks that affect portfolio reallocation needs of investors and markets makers’ ability to provide liquidity, might be possible macro liquidity drivers. This is due to the idea that informational shocks could cause positive trading activity across securities, which in turn might cause co-movements across these markets. Regarding co-movement across markets, they imply that liquidity measurements can be driven by common influences and can exhibit co-movement
across asset classes. On that logic, they analyze the impact of expansionary monetary policy on stock and bond markets to investigate whether this policy would cause a change in these assets’ liquidities. As liquidity proxy, Chordia et. al (2005) also used bid- ask spread to account for transaction cost. Their study reports lower stocks’ liquidity in down-markets. They attribute this to the possible strained market making capacities during periods of market decline. In addition they find that a loosening monetary policy is contemporaneously associated with increased equity market liquidity during crises.
Motivated by the above-mentioned studies, the methodologies applied by several researchers are used to further analyze the cost of sovereign default. Van Horen et. al (2008) and Forbes (2004) identify default with dummy variables, implying that they only take special episodes during crises. This might cause ambiguity in the results, since not the entire default period is taken into consideration. This paper applies the definition used by Reinhart and Rogoff (2008) to account for default and takes the first two years as event window. In contrast to those who look at emerging markets under stress, this paper takes a new path by looking at the current Eurozone debt crisis; this has never been done before.
With regards to the research of Chordia et. al (2005), I use a different technique by analyzing the relationship between returns and liquidity measurements in countries with different economic conditions. This is important in order to determine whether the impact of bail-out on a country depends on the country’s fiscal or economic condition.
Also the statement that monetary policy can cause co-movements in liquidity sheds light on analyzing the impact of fiscal decision (in this paper bail-out acceptance) on the stock market. Most importantly, the view that the market is illiquid during crisis is adopted to analyze the cost of sovereign default on the stock market.
Based on the above, the following expectations are developed:
o There will be significant negative cumulative abnormal returns during bail- out episodes. This effect is expected to be stronger with the anticipation effect than with public bail-out announcement
o There will be significant cumulative abnormal increase in the liquidity variables. Specifically, trading activity and cost of trading are expected to increase
o Market is expected to be illiquid: The relationship between volume traded and transaction cost is expected to be positive. Thus high volumes are traded at high cost
o Abnormal returns is not expected to be compensated for abnormal illiquidity.
This means that an increase in abnormal volume traded is expected to be associated with a decrease in abnormal returns. Analogously, an increase in abnormal cost of trading is expected to be associated with a decrease in abnormal returns
3. Data
Financial data
The micro data: Company total returns, volume traded, numbers of shares on issue, stock price, and bid-ask spread are gathered from Datastream. The data ranges from March 2010 to June 2012. To calculate abnormal returns, the Total Return Index (RI) of each market is taken from Datastream. The (RI) tracks the capital gains of equities and assumes that any cash distributions, such as dividends, are reinvested in the underlying company. This gives a more precise illustration of the index’s performance, since it effectively accounts for stocks that do not issue dividends. The preliminary data consisted of: 75 companies for Greece, 70 for Ireland, 75 for Italy, 60 for Portugal and 72 for Spain. After evaluating and removing companies with too much missing data, the sample is reduced to 31 companies for Greece, 52 Ireland, 73 Italy, 10 for Portugal, and 53 for Spain. The macro data –taxes, GDP, public debt, deficits/surplus are gathered from the World Bank database and the IMF database.
News data
Information on relevant news is gathered from the BBC UK website. It hosts a specific timeline illustrating on the events related to the crisis since its inception. Table I.A shows specific public bail-out announcement dates for each country from May 2010 to March 2012. Table I.B groups each country according to certain macroeconomic criteria.
Descriptive statistics
Table II depicts the descriptive statistics of the cumulative average abnormal values of the variables in the event window D0through day D20. Panel A depicts the values since the first Greece bail-out announcement in May 2010. All countries are pooled to analyze a possible anticipation effect during the crisis period. The entire result of the descriptive statistic, which is included Appendix C of this paper shows that the impact was the highest during the first 20 days following the event and gradually decreased during the last days of the event window.
Panel B depicts the results of specific public bail-out announcement dates for the countries. As was expected, the preliminary results show that anticipation impact is stronger than public announcement impact. This is apparent in the mean values of the proxies. Table II reports a mean of cumulative average abnormal returns of -0.0722 in (D0, D20) reflecting anticipation effect. Comparing this with the result of specific public announcement, gives a much lower mean value in (D0, D20): -0.0077. Also the cumulative average abnormal result on D20is given. Deeper examination of the liquidity variables reveals similar results, except for the AMR. Overall, the result shows higher liquidity values for anticipation effect than for official public bail-out announcements.
Table I.A: Timeline of Specific Bail-out Announcement Dates for the 5 Countries From 2010 to 2011 Countries Default dates Specification announcement
Greece May 2nd 2010 110bn euro bailout package
March 13th 2012 Second bailout package of 130bn euro Ireland November 22nd2010 85bn euros bailout package
Portugal May 16th2011 78bn euros approved for bailout
Italy August 7th 2011 ECB announce purchase of Italian bonds (this is considered a bail-out since a purchase of a bond by the ECB implied that the ECB injects money in the system to prevent further distress)
Spain August 7th 2011 ECB announce purchase of Spanish bonds
Table I.B: Several Fiscal or Economic Conditions For the 5 Countries in Sovereign Debt Crisis Debt ratio
Preceding crisis Growth rate
during crisis Primary deficit
Preceding crisis Deficit ratio preceding crisis
Criteria >100%GDP <100%GDP <0 >0 ≥14%GDP <14%GDP ≥14%GDP <14%GDP Country Greece
Italy Ireland
Portugal Spain
Greece Ireland Portugal
Italy
Spain Greece
Ireland Italy Portugal Spain
Greece
Ireland Italy Portugal Spain
Figure A plots the time series for anticipation effect and specific bail-out announcements. As can be seen from figure A. 1, CAAR was never positive in the first 20 days following announcement, indicating a loss in asset value during crises. It also shows that all liquidity measures are constantly positive during the first 20 days.
Looking at the specific public bail-out announcements, gives the impression that CAAR only decreases in the first days following announcement and starts increasing from day 2 to day 6. It also shows increasing positive values for the liquidity proxies from day 2.
Afterward the liquidity proxies fluctuate throughout the event window, though they remain positive. On the other hand, figure A.2 shows that the impact is less severe in the first 20 days for specific bail-out than for anticipation effect.
Table III depicts descriptive macroeconomic statistics for the 5 countries. At first glance the statistics demonstrate an increase in debt ratio for all countries prior to Greece’s bail-out in 2010. Focusing on each country individually, the table shows that out of the 5 countries, Greece and Italy have on average the highest debt levels. From 2007 to 2011, Greece and Italy have average debt ratios of 140% and 114% of GDP respectively. On average, Spain has the lowest average debt level. Ireland and Portugal Table II Descriptive Statistics Using Event Window (D0,D20)
The results provided account to the cumulative average abnormal variables. Anticipation Effectillustrates default expectation by the countries.Specific Bail-outillustrates actual bail-out reaction. P is the expected abnormal value in the estimation window.
Statistics Return Amihud
Ratio Volume Bid-Ask
Spread Panel A
Anticipation Effect
(D0,D20) Mean -0,0722 0,0362 0,0152 0,0327
Sample Variance 0,0060 0,0058 0,0017 0,0141
Minimum -0,3547 -0,2324 -0,0387 -0,3160
Maximum 0,2904 0,5005 0,3184 0,9718
D20 Mean -0.0682 0.0553 0.0220 0.0583
P 0.4729 0.3379 0.3077 0.3783
count 219
Panel B
Specific Bail-out
(D0,D20) Mean -0,0077 0,0383 -0,0019 0.0037
Sample Variance 0,0130 0,0315 0,0023 0,0933
Minimum -0,5232 -0,4638 -0,2375 -1,6715
Maximum 0,6033 1,5326 0,4221 0,7215
D20 Mean -0.0160 0.0580 -0.0031 0.0024
P 0.4656 0.3417 0.3052 0.3793
count 249
followed with slight increase in the ratio, though both are below 100% of GDP. The results also show that since 2008 up to 2011, Greece, Ireland, and Portugal’s growth rates have each exhibited a decreasing pace. Italy and Spain have decreasing growth from 2008 to 2009. However for Italy, growth rate shows a slight increase in 2010 and 2011. Spain’s growth rate shows only a slight increase in 2011. The results also show that Ireland, Portugal and Spain have the highest deficit ratios. Regarding primary deficit level, on average Greece and Ireland have the highest deficits. Overall, the macro statistics show that Greece and Ireland are the most distressed countries based on their fiscal or economic conditions.
Figure A: Trend of the Cumulative Average Abnormal Variables (CAAX) Over Event Days D0Through D20.
Here the CAAR, CAAAMR, CAASVOL and CAAPBAS reflect Cumulative Average Abnormal values for Returns, Amihud Ratio, Volume Traded, and Bid-Ask Spread respectively.Anticipation Effectillustrates default expectation by the countries.Specific Bail-outillustrates actual bail-out reaction
A. 1
A.2
-0.15 -0.1 -0.05 0 0.05 0.1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Trend of the CAAX: Anticipation Effect
CAAR CAAAMR CAASVOL CAAPBAS
-0.15 -0.1 -0.05 -1.7E-16 0.05 0.1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Trend of the CAAX: Specific Bail-out Dates
CAAR CAAAMR CAASVOL CAAPBAS
2The primary information was retrieved on March 2012 from the IMF database and on May 2012 from the world bank database
Table III Fiscal or Economic Conditions Prior and During Bail-out Episodes as Percentage of GDP2
Greece 2007 2008 2009 2010 2011 Averages
Sovereign debt(b) 125.2820 126.7240 141.9699 142.757 163.343 140.01518
Growth rate(∆y) 2.9962 -0.1569 -3.2506 -3.5167 -6.9072 -2.1670
Primary Deficit(g-t) 6.7450 9.6806 15.5529 10.6595
Deficit ratio(∆b) -2.7808 1.4421 15.2459 0.7871 20.5860 7.0560
Ireland
Sovereign debt(b) 28.3722 49.0498 70.4825 92.4850 104.95 69.0679
Growth rate(∆y) 5.1823 -2.9721 -6.9945 -0.4294 -0.7000 -1.1827
Primary Deficit(g-t) -0.2425 7.0332 14.1464 6.97033
Deficit ratio(∆ b) -0.2852 20.6777 21.4326 22.0025 12.4660 15.2588
Italy
Sovereign debt(b) 104.2962 107.1373 118.4371 118.6520 120.1060 113.7257
Growth rate(∆y) 1.6830 -1.15623 -5.4944 1.8044 0.4311 -0.5464
Primary Deficit(g-t) 1.3239 2.3397 4.8620 2.8419
Deficit ratio(∆ b) -4.7903 2.8411 11.3000 0.2149 1.4540 2.2039
Portugal
Sovereign debt(b) 67.5100 72.4622 83.9667 93.4430 106.7850 84.8334
Growth rate(∆y) 2.3653 -0.0085 -2.9084 1.4013 -1.6096 -0.7599
Primary Deficit(g-t) 2.5669 2.6291 8.6743 4.6234
Deficit ratio(∆b) -1.8487 4.9522 11.5045 9.4763 13.3420 7.4853
Spain
Sovereign debt(b) 30.0803 34.1938 46.4292 47.8129 68.4710 45.3974
Growth rate(∆y) 3.4792 0.8887 -3.7408 -0.0695 0.7080 -0.0301
Primary Deficit(g-t) -2.4510 2.3272 8.5296 5.1699 3.5256
Deficit ratio(∆ b) -3.9690 4.1135 12.2354 1.3838 20.6581 6.8843
4. Methodology
In this paper the three proxies used by van Horen et. al (2008) as measurement for liquidity will be applied – the Amihud ratio, bid-ask spread and volume traded. The Amihud ratio and bid-ask spread are proxies for trading cost and the volume represents trading activity. The Amihud ratio represents a measure of price impact as it can be interpreted as the daily price response associated with one dollar (in this paper euro) of trading volume. Amihud (2002) refers to the proxy as illiquidity, since he perceives liquidity as an elusive concept. He reasons that it is a measurement that cannot be directly observed but rather has a number of aspects that cannot be captured in a single measure. He further clarifies that illiquidity reflects the discount that a seller accepts or the premium that a buyer pays when executing a market order. However, Amihud acknowledges that the bid-ask spread is a finer measurement for illiquidity. Also according to Amihud (2002), the Amihud ratio is coarser and less accurate.
Nevertheless, unlike the bid-ask spread, it is a measurement that is calculated from data that are readily available for long periods of times for most markets.
Amihud’s (2002) findings show that across stocks and over time, expected stock returns are an increasing function of expected illiquidity. This implies that in equilibrium, an increase in the cost of transacting and volume traded should be associated with an increase in expected returns. This is analogous to the capital asset pricing theory, which states that investors should be promised higher returns for more risky stocks. In that sense, Amihud argues that in equilibrium, stocks’ excess returns should also reflect compensation for market illiquidity. In a previous paper, Amihud and Mendelson (1991b) imply that since investors prefer to promise capital to liquid investments, in equilibrium, the expected returns on capital assets are an increasing function of illiquidity.
The proxies are determined as follows
= − (1)
= (2)
( ) = (3)
(4)
= |∆ |( ) (5)
In equation (1),BASitrepresents bid-ask spread in Euro unit for company i on dayt. The MIIDQUOTE is calculated by taking the average of the bid-ask (BIDit+ASKit)/2. PBASit represents the ratio for company i, on day t. SVOLit in (3) is measured as the daily volume traded in share divided by share on issue(SOIit).Equations (1)-(3) are adopted from Frino et. al (2006). Equation (4) is a standard equation for returns. Equation (5) represents the Amihud ratio (AMR), with |∆Pit| as the absolute value of the daily price change during the day.
Abnormal Values
To determine the abnormal behavior of each proxy (AX) –(Abnormal PBAS, abnormal SVOL, Abnormal R and Abnormal AMR), the mean-adjusted method of Brown and Warner (1980) is applied.
The general abnormal value for each proxy is determined as follows:
= , (6)
where AXit, Xit and i are the abnormal, actual, and expected values respectively for company in time periodt.
In this study a 2 years period event window is applied leading to 540 days. Since a sovereign debt crisis can last up to 2 years, the paper takes this into consideration to analyze how the abnormal values fluctuate during this event window. This is due to the reasoning that a bail-out is more a macroeconomic event and might have a longer impact on the market. However, in this paper more focus is put on the first 20 days following an event. This is because generally, the impact of an event is more pronounced in the first days following an event. According to the BBC UK online news, it
was not until May 2nd, 2010 that the first tranche of an initial bail-out package of 110bn euro was allowed to Greece. In March 2012, a second bail-out package of 130bn euro was provided. For the other Eurozone members with similar problems, bail-out packages were provided during that same crisis period. From the first bail-out date on, different events took place which would affect investors’ confidence (see timeline in Table I.A). The timeline is used to track the specific dates of bail-out announcements during the event window.
Two approaches are used to analyze the impact of a bail-out announcement on investors’ reactions. Firstly, Greece’s bail-out announcement on May 2010 is taken as the inception date. From May 2010 to June 2012, PIIGS stocks’ data are pooled to analyze the anticipation effect mentioned in the literature review. Secondly, I separate the data for specific bail-out dates. These are then pooled to analyze the magnitude of the impact. Note that Greece’s two bail-outs dates are also included in this second analysis. The event date is indicated as day 0.
Following the methodology of Cowan (1992), an estimation window of 100 days before the event date is applied. The estimation window is 100 days before May 2nd 2010. This same estimation window is also used to analyze specific bail-out announcement to avoid ambiguous results, since the 540 days are considered as a period of financial turmoil. The event date is the indicated day on the BBC timeline for the month of bail-out or rescheduling announcement. The event window starts from May 2nd 2010 up to June 30th2012.
Cumulative Abnormal Values
To determine the cumulative abnormal value for each proxy (CAX) over an event window, days D0through Dd, the Cowan (1992) method is used:
CAX= ∑ (7)
The cumulative average abnormal value for each proxy (CAAX) for a sample ofnstocks over the event window D0through Dd, is calculated as follows:
(8)
Secondly, the Jarque Bera test for normality shows that the data is not normality distributed. Hence a non-parametric test –the generalized sign test applied by Cowan (1992), is conducted to test the significance of the proxies.
Following the Cowan notation, the number expected for each proxy is determined by:
= ∑ ∑ , (9)
where for the abnormal returns (AR)
= 1 < 0
0 ℎ
The objective is to test whether the abnormal are significantly negative compared to the number expected. This expectation is due to the result of several studies that report negative abnormal returns during crises (Chordia et. al 2005; Forbes 2004; Hameed et.
al 2010; van Horen et al, 2008). For the liquidity proxies the opposite notation will be applied, since the expectation is that during crises the liquidity proxies increase (Amihud 2002, Chordia et. al, 2005; Hameed et. al 2010; van Horen et. al 2008).
Appendix Adepicts results for equation (9).
Based on the abovementioned considerations, the following notation for the liquidity proxies is applied:
= 1 > 0
0 ℎ
The test statistic for the abnormal values is as follows:
= ( ( )) , (10)
where w, is the number of stocks in the event window for which the cumulative abnormal returns (CAR) is negative or the cumulative abnormal liquidity (CAL) measures are positive. The values forw in equation (10) are given inAppendix B.
Regression Analyses
Subsequently time series (eq 11) and cross-sectional regressions (eq 12) are performed with the abnormal values. Since the results for anticipation effect are more pronounced than the results for specific bail-out announcement, this analysis focusses only on the anticipation effect. The time series regression is related to the Collateral view and the liquidity theory. The objective is to test market liquidity during crises by regressing volume traded with cost of trading using the definition of Amihud (2002), Bank for International Settlements (1999) and Chordia et. al (2005). In the event window, cumulative average abnormal volume traded is regressed against cumulative average abnormal bid-ask spread and cumulative average abnormal Amihud ratio. The relationship between volume and cost is given as:
= + . + . + (11)
where is the cumulative average abnormal volume, , the coefficient proxies for the bid-ask spread and the Amihud ratio respectively in the event window. The and the represent the constant and the error term respectively. Two regressions are employed. The first one takes the entire event period of 540 days into account. The second one takes only the first 20 days following announcement into account.
The last regression is a cross-sectional analysis to test the relationship of abnormal returns with liquidity. Amihud and Mendelson (1991b) state that in equilibrium investors should be promised higher returns with less liquid stocks. The relationship is given by the equation
= + . + . + . + , (12)
whereCAR is the average of the cumulative abnormal returns for theith company in the event window days D0through Dd. The same analogy applies to the other variables.
On both time series and cross-sectional regressions different OLS checks are performed. First, the White test for heteroskedasticity is performed to check whether the standard errors have constant variances. In the event that the variances are not constant, the heteroskedasticity consistent coefficient test is used to correct for this.
Also the Jarque Bera test of normality is conducted to test whether the error terms are normally distributed. In addition, the Ramsey test is conducted to see whether the correct functional form has been used, to account for possible misspecification of the model.
Furthermore, to detect the possibility of simultaneity between the variables, the Hausman test for exogeneity is conducted. This is done because there might be a two way relationship between the exogenous and the endogenous variables. Some studies indicate that expected volume is a function of trading cost. However, expected cost might also be a function of volume traded. Regarding the cross-sectional analysis, according to Chordia et. al (2005), liquidity might influence returns. Nevertheless, Amihud and Mendelson (1986) argue that returns might influence liquidity as well.
Thus a test of exogeneity helps to clarify the relationships between the variables.
Nonetheless, in the event that this exogeneity is present, this effect is not further analysed in this paper.
Macroeconomic robustness checks
Additional robustness checks are conducted by analyzing groups of countries with different macroeconomic conditions as categorized in Table I.B and Table III. In his recent paper, Mun (2012) argues that macroeconomic developments can affect the market’s expectation about the economic conditions of a country. He argues that these expectations are revealed in the financial system by their influence in the expected returns on all asset classes. According to Butt et. al (2009) financial information and macroeconomic variables could predict a notable portion of stock returns. In their research regarding the stock return variation to specific economic conditions, they find that stock returns of different industries behave differently to different macroeconomic conditions. Other researchers have also investigated stock returns variation caused by various economic factors (Chen et. al., 1986; Errunza and Hogan, 1998; Getler and
Grinols, 1982; Hu and Mei, 2000). Their findings also suggest that stock returns are sensitive to changes in macroeconomic factors.
Conventional macroeconomic theory suggests that government running a primary deficit, with income growth falling short of interest rate can only stabilize by appropriate adjustment of the primary deficit ratio. Economic theory also implies that a country that grows and accumulates wealth may find a growing debt acceptable, as long as the debt ratio does not grow (Gartner, 2009, p 394). From Table III, one can notice that Greece’s growth rate has decreased extremely. Ireland and Portugal follows, though with less severe rates than Greece.
Hosono and Sacuragawa (2011) use historical joint distributions of interest rates and growth rates to test the fiscal sustainability of Japan. They evaluate fiscal sustainability by looking whether the debt to GDP stabilizes or increases without bound.
Their result indicates that Japan’s fiscal sustainability depends on its projected growth rate and fiscal policy rules. Their finding suggests that a government should react to their fiscal crisis in order to avoid an increase in the debt-to GDP ratio.
Fiscal policy matters because a large deficit and high debt might force the central bank to inject money in the market. In the EU, the Maastricht Treaty specifies a debt to income ratio not exceeding 60% and a deficit ratio below 3%. These constraints are to ensure price stability. However as can be seen from Table III, only Spain complied with the debt ratio requirement and Italy with the deficit ratio. According to the European Central bank (2005), economies with high public debt ratios and slow fiscal adjustment, experience larger deviations of the price level from its trend after fiscal expansion. They suggest that different fiscal adjustment rules could be taken into account to analyze the response of the economy to fiscal expansion. Furthermore they imply that fiscal policies affect the equilibrium condition of the money market. In order to restore equilibrium, price level should be adjusted, which reflects the mechanism through which fiscal policy affects the price level. In a simulation analysis, they find that temporary fiscal expansion negatively affects price stability.
Anncchiaricco and Marini (2003) show that high levels of debts and high deficits affect the money market equilibrium through their effect on wealth. This effect can be revealed through its impact on investors’ uncertainty. According to Acosta and Loza (2004) external debt level can represent the evolution of external credit and
may signal over-indebtness and may negatively impact investors’ expectations. Their study confirms the results of previously made studies that external debt level together with other fiscal conditions of a country, are the main variables that guide investment decisions in the long run.
Ardagna (2005) also provides empirical evidence on the impact of fiscal policy shock on financial variables. She argues that financial markets are forward looking and that agents’ perception can explain the effect of fiscal shocks on financial variables. Her findings suggest that financial markets react in anticipation of the future path of government debt-to-GDP. The finding predicts that a decline in public debt will lead to an increase in stock market index and its growth rate. Additionally she tested whether the effect of changes of the government budget on financial variables depends on the decrease of public debt in the future. She finds that stock markets react positively only when governments are successful in reducing the debt-to-GDP ratio.
This sensitivy analysis is important because it sheds light on what type of macroeconomic variables would display a stronger association between asset returns and liquidity. Four economic conditions are used: debt ratio (b), deficit ratio (∆b), primary deficit/surplus (g-t) and growth rate (y).
The economic notations that are used in the paper are as follows:
1) Primary deficit:
2) Deficit to income ratio/budget deficit: ∆B/Y=∆b 3) Debt ratio:B/Y=b
4) Growth rate in GDP:∆y=
In 1), G represents all public spending on consumption and investment, plus transfer payment in the year that bail-out is announced. T is the total taxes collected. In 2), B and Y stand for total government debt and output as measured by GDP preceding bail- out year respectively. A country can run a deficit (g>t) or surplus (g<t), with g as (G/Y) and t as (T/Y).
B=debt; ∆B=change in public debt, ∆B/Y=∆b= deficit to income ratio or budget deficit.
Table I and Table III represent the several economic conditions considered to analyze the impact of sovereign default.
Hypothesis formation
The first four hypotheses are derived from the abnormal performance of each proxy.
For equations (7) and(8),the statement that returns are negative during crises leads to the expectation that cumulative abnormal returns will be negative during bail-out episodes. Following the Cowan (1992) method, the null and alternative hypotheses are
, : , ≥ 0
, : , < 0
The cumulative average abnormal illiquidity proxies are expected to increase. This implies that volume traded and costs of trading are expected to increase. These are given as
Trading activity (volume traded):
, : , ≤ 0
, : , > 0
Trading cost (Amihud ratio):
, : , ≤ 0
, : , > 0
Trading cost (Bid-Ask spread):
, : , ≤ 0
, : , > 0
The second group of hypotheses is related to market liquidity from equation (11). The definition of a liquid market is applied here to account for the expected relationship between volume traded and cost of trading. From the mentioned studies, it can be
derived that increased volume traded is expected to be associated with increased cost of transaction during crisis episodes.
The supposition concerning the relationship between abnormal volume traded and abnormal bid-ask spread is given as
, : = 0
, : > 0
Analogously, the supposition regarding the relationship between abnormal volumes traded and abnormal Amihud ratio is given as
, : = 0
, : > 0
The last groups of hypotheses test the relationship between returns and liquidity from equation (12). Taken from the analogy of Amihud and Mendelson (1991b), the following can be derived: Amihud uses the CAPM theory to develop an equilibrium expectation for stocks. He states that in equilibrium, an illiquid stock should be expected to gain higher returns. As already derived, a stock that trades at a high cost and high volume is considered to be illiquid. Since crises mark departures from equilibrium, the expectation is that abnormal returns are negatively associated with illiquidity. The hypotheses regarding this relation for the different proxies are as follows:
Abnormal returns and abnormal bid-ask spread:
, : = 0
, : < 0
Abnormal returns and abnormal Amihud ratio:
, : = 0
, : < 0
Abnormal returns and abnormal volume traded:
, : = 0
, : < 0
5. Results
Cumulative abnormal results
Figure A.1, Table II panel A and Table V display statistical results for the Generalized Sign Test classifying for possible anticipation effect by IIPS. From Table V, it can be noticed that CAAX show strong significant Z-values from D1through D20. The entire result, which is includedAppendix C, shows that from D20to D540 the Z-values fluctuate between positive and negative Z-values. However the values are significant at D540. The latter are smaller, suggesting that the impact is stronger only during the first days following announcement and decreases at later stages of the crisis. This is in line with the statement of Panizza and Borensztein (2009) who argue that default is short lived.
Combining Table II.A with Table V, one can notice that on D20, anticipation effect is significantly associated with CAAX of -0.0682, 0.0553, 0.0220 and 0.0583 for Returns, Amihud Ratio, Volume TradedandBid-Ask Spreadrespectively. In view of these results, H1,0-H4,1 are rejected and H1,1-H4,4 come into effect. This conclusion is in line with the results of Chordia et. al (2005), Hameed et. al (2010) and van Horen et. al (2008), who found a decrease in returns and a decrease in liquidity during crises.
I then explore the effect of specific public bail-out announcements dates. Figure A.2, Table II panel B and Table V present these results. The proxies fluctuate from D0to D20 with positive and negative values and show that the impact is stronger in the first three days following bail-out. On D20, public bail-out announcement is significantly associated with CAAX of -0.0160, 0.0580, -0.0031 and 0.0024 forReturns, Amihud Ratio, Volume TradedandBid-Ask Spread respectively. Except for the Volume Traded, the Z- values are still significant on D20. Notably, the impact on all the variables is less strong for specific announcement. Particularly volume traded decreases sharply. The effect can be seen in the degree of impact on the cumulative average abnormal values and the test statistics. These values are less strong for public bail-out than for anticipation effect. This striking result leads to the conclusion that investors do price the risk of
result also implies that bail-out announcement might be perceived as a turning point and has a positive impact on investors’ expectation. This is consistent with the findings of Panizza and Yeyati (2011) who suggest that default marks the recovery of a crisis.
Sensitivity Analysis
Table VI shows results as a check of how companies in different countries with different fiscal or economic conditions are affected by bail-out episodes. This analysis takes only the anticipation effect into account. Countries are grouped according to economic conditions as reported in Table I. For all economic conditions, the result shows significant abnormal Z-values over the first three days following announcement. The Table also shows the first dates where CAAR started to be insignificant. This helps to have a clearer understanding of how the crisis developed throughout the 540 days event window. The table also depicts the dates where CAAR has the highest insignificant Z-value. As can be observed, the crisis’ impact on CAAR starts to be less pronounced from the third month following announcement. On the 9th month CAAR reaches the highest insignificant Z-value. Yet and overall, the majority of the variables are significant over the entire 2 year event window.
Further, I conduct additional robustness analyses on the groups as listed in Table I. First I look at companies listed in Greece and Ireland, which are countries with debt levels >100% of GDP preceding bail-out. These are compared with Italy, Portugal and Spain, which are countries with debt levels <100% of GDP. The results show that Z- values for abnormal returns and liquidity proxies are higher for Greece and Ireland than for Italy, Portugal and Spain. Thus the result shows that Greece and Ireland have relatively high decrease in abnormal returns. The result also suggests that these two countries have relatively high increase in transaction cost and volume traded in comparison to the remaining three countries.
I further look at the results presented in panels C and D for other groups with
∆y<0 and ∆y>0
