Information transmission between stock and bond markets during the Eurozone debt crisis: Evidence from industry returns Nuno Silva *,+ * University of Coimbra, Faculty of Economics + CeBER - Centre for Business and Economics Research Abstract I analyze the Granger causality in distribution between sovereign bonds and industry indexes in the five European countries most affected by the debt crisis: Greece, Ireland, Italy, Portugal, and Spain. Prior research assessed the impact of the debt crisis on the financial firms, but its effect on other industries was broadly neglected. My results reveal that, at the height of the crisis, delayed shocks transmission from the sovereign bond to the stock market occurred mainly in Greece. At the industry level, there is no evidence of lagged response of the financial industry to negative sovereign debt shocks, but sovereign debt leads other industries in, at least, one country. These findings are consistent with the investor inattention hypothesis, which states that investors tend to specialize in specific markets, due to their limited availability of time and resources and the cost of information gathering, and information flows slowly across markets. Electronic copy available at: https://ssrn.com/abstract=3697288
22
Embed
Information transmission between stock and bond markets ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Information transmission between stock and bond markets during the
Eurozone debt crisis: Evidence from industry returns
Nuno Silva*,+
*University of Coimbra, Faculty of Economics +CeBER - Centre for Business and Economics Research
Abstract
I analyze the Granger causality in distribution between sovereign bonds and industry indexes
in the five European countries most affected by the debt crisis: Greece, Ireland, Italy, Portugal, and
Spain. Prior research assessed the impact of the debt crisis on the financial firms, but its effect on
other industries was broadly neglected.
My results reveal that, at the height of the crisis, delayed shocks transmission from the
sovereign bond to the stock market occurred mainly in Greece. At the industry level, there is no
evidence of lagged response of the financial industry to negative sovereign debt shocks, but
sovereign debt leads other industries in, at least, one country. These findings are consistent with the
investor inattention hypothesis, which states that investors tend to specialize in specific markets, due
to their limited availability of time and resources and the cost of information gathering, and
information flows slowly across markets.
Electronic copy available at: https://ssrn.com/abstract=3697288
1. Introduction
The European debt crisis began at the end of 2009 when the new Greek government revised
upwards the budget deficit projection to 12.7% of GDP. This revision led to downgrades of the Greek
sovereign debt by several rating agencies, and to a surge in bond yields that effectively barred Greece
from international credit markets and culminated in a request by Greece of an initial loan of 45 billion
euros from the EU and the IMF.
The Greek debt crisis quickly spread to the rest of the Eurozone, as investors become
increasingly aware of the fragility of other peripheral countries. In 2010 sovereign bond yields rose
sharply in Portugal and Italy, as investors began questioning the sustainability of their public debt,
and in Ireland and Spain who had to rescue several banks that were plagued by a large stock of non-
performing loans, after the world financial crisis.
The European debt crisis left its mark, not only in the sovereign bond market but also in other
asset classes, such as stocks. Several authors studied the impact of this crisis on banking stocks, which
were particularly affected by it, due to their high exposure to the sovereign credit risk. A study group
established by the Committee on the Global Financial System of the Bank for International
Settlements (BIS, 2011) identified four main transmission channels through which sovereign risk can
affect banks:
i) Asset holdings- Banks may suffer losses due to their holdings of sovereign debt;
ii) Collateral/liquidity- Increases in sovereign credit risk reduce the value and/or eligibility of
sovereign bonds as collateral in banks’ funding operations;
iii) Sovereign ratings and banking ratings- Sovereign ratings usually represent a ceiling for
domestic banks’ ratings. Thus, a sovereign downgrade tends to be followed by downgrades in
domestic banks’ ratings;
Electronic copy available at: https://ssrn.com/abstract=3697288
iv) Government guarantees- The deterioration of the sovereign fiscal position leads to a
decrease of both explicit and implicit guarantees on bank funding.
Even though banking was the most severely affected and thoroughly studied sector during
the European debt crisis, its impact was felt across the whole economy. The linkages between other
industries and sovereign bonds are deep and often neglected. First, they are both affected by a
deterioration of investors’ expectations regarding the future growth of the economy. Second, a
higher sovereign yield may lead to contagion in the form of higher financing costs for firms, especially
in highly leveraged industries, such as telecommunications and utilities. Third, an increase in the
perceived likelihood of sovereign default increases the prospects of further fiscal consolidation,
which decreases internal demand and affects, particularly, industries most exposed to the domestic
market. Finally, a deterioration in firms’ business conditions may generate lower profits and
employment, which translates into lower tax revenues, and may compromise the sovereign debt
sustainability.
The main objective of this study is to analyze if information flowed swiftly between the
sovereign bond and eleven industry equity indices, during the height of the sovereign debt crisis
(2010-12), in the five most affected countries (Greece, Ireland, Italy, Portugal, and Spain). To obtain
a clearer picture of the information transmission in these countries during the crisis, I benchmark it
against the two largest Eurozone economies. I also compare the speed of information flow during
the crisis and in the period afterward (2013-19). To achieve this goal, I use the nonparametric test for
Granger causality in distribution proposed by Candelon and Topkavi (2016). This method can test
causality over several quantiles of the distribution and offers a more complete picture of the
information transmission across markets than the traditional Granger causality in mean.
Furthermore, it extends the method proposed by Hong et al. (2009) by allowing the researcher to
test causality in several quantiles simultaneously, effectively making it a causality test in the
Electronic copy available at: https://ssrn.com/abstract=3697288
distribution. My results reveal there is evidence of contagion, in both directions, for several
industries, in the left and right tails of the distribution.
My contribution to the literature is twofold. First, I assess the impact of the sovereign debt
crisis in other industries beyond finance, across different parts of the distributions. Second, I show
that the European debt crisis had a broad impact on several industries and that new information is
not incorporated in the sovereign bond and stock markets at the same time. These results are
consistent with Hong et al. (2007) and Menzly and Ozbas (2010), among others, who report that
industry returns exhibit positive cross-momentum because Information gathering is costly, and
investors tend to specialize in specific sectors. Thus, news flows slowly across industries.
The remainder of this paper is organized as follows. Section 2 presents the main related
literature. Section 3 describes the dataset. Section 4 presents the econometric methodology. Section
5 displays and analyzes my main results. Finally, section 6 presents the concluding remarks.
2. Related literature
This study is related to a vast strand of literature on the cross-asset spillover of shocks, that
experienced rapid growth following the 2008 Global financial crisis and the subsequent European
debt crisis.
The strong linkages between sovereign bonds and banking stocks led several authors to study
the existence of contagion and spillover effects between them. Allegret et al. (2017), using a
multifactor model of equity returns with a sovereign risk premium, conclude that the negative impact
of the European Debt Crisis is confined to European banks. Bhanot et al. (2014) find that an increase
in Greek sovereign bond yields generates negative abnormal returns in financial stocks of Greece,
Portugal Italy, and Spain, and this effect is reinforced when there are negative news announcements
about Greece. Tamakoshi and Hamori (2012) analyze the relation between Greek sovereign bond
Electronic copy available at: https://ssrn.com/abstract=3697288
yields and Southern European banking stock indices and conclude that there is unidirectional
causality-in-mean from banking stocks to Greek sovereign bond yields and bidirectional causality-in-
variance. Using a database of 33 systemically important banking stocks and 36 sovereign bond yields,
Corsi et al. (2018) develop a measure of connectedness between these markets in times of financial
distress. They show that this indicator peaks at the beginning of the European, which implies that the
“flight-to-quality” phenomenon was especially prevalent during this period. The credit default swap
(CDS) spreads is used to test the transmission of shocks between sovereign bonds and banks by Alter
and Beyer (2014), who show that the interconnectedness between banks and sovereign CDS’s
increased from 2010 to 2012, and De Bruyckere et al. (2013) who report evidence of increased
correlation (contagion) between sovereign and bank CDS spreads during the European debt crisis,
especially in the GIIPS countries. Grammatikos and Vermeulen (2012) show that an increase in Greek
CDS spread causes a decrease in both financial and non-financial stock returns, not only in the fragile
Southern European countries but also in the more robust Northern ones. Using a vector
autoregressive model, Coronado et al. (2012) analyze the transmission of shocks between the CDS
and stock markets, in several European countries, between 2007 and 2010. They find that the stock
markets lead CDS markets throughout most of the period considered, but the CDS markets played a
key role in shock transmission at the heyday of the debt crisis. Ballester et al. (2016) assess return
spillovers between bank CDS markets in different countries. After decomposing CDS returns into
systematic and idiosyncratic factors, using principal component analysis, they apply a generalized
VAR model to measure contagion. The authors report that global contagion is always greater than
idiosyncratic contagion, but the role of idiosyncratic risk in information transmission increased during
the European debt crisis.
Another line of research focuses on the international transmission of shocks across several
European countries during the debt crisis. Tola and Walti (2015), using a narrative approach, find
Electronic copy available at: https://ssrn.com/abstract=3697288
evidence of contagion in the European sovereign debt market, and Tamakoshi and Hamori (2011)
show that significant causal relationships between European stock markets disappeared during the
Greek sovereign debt crisis.
The literature on cross-asset interdependence is a rich one and covers a wide range of classes.
Chang and Cheng (2016) study the cross-asset contagion between REIT, stock, money, bond and
currency markets in the US, Norden and Weber (2009) analyze the transmission of shocks in stocks,
CDS spreads and bonds, Beirne and Gieck (2014) focuses on global bond, equity and exchange rate
markets, Longstaff (2010) shows that a subprime asset-backed collateralized debt obligations index
leads stocks, and corporate and treasury bonds by as much as three weeks, and Chulia and Torro
(2008) test the volatility transmission between European stock and bond markets using futures
contracts.
The research on contagion at the industry level is much scarcer. Bekaert et al. (2014) analyze
the information transmission for 415 equity industry portfolios from 55 countries during the global
financial crisis, using a three-factor model. They find that domestic contagion dominates
international contagion, and it is particularly severe in countries that present poor economic
fundamentals. Phylatkis and Xia (2009) also use a factor model to test for contagion across several
country-industry indexes, from U.S., Latin America, Europe, and Asia, between 1990 and 2004. They
conclude that the transmission of information is heterogeneous across industries: some industries
are plagued with contagion, while others seem almost immune to it. Using an asymmetric dynamic
conditional correlation GARCH model, Alexakis and Pappas (2018) test for international contagion at
the sector level in 15 European countries, during the global financial crisis and the European
sovereign debt crisis. They conclude that contagion exists in all the business sectors, and it is
especially prevalent in financials and telecommunications. The information flow between 11 US
industry stock indexes and their corresponding CDS spreads, is studied by Shazad et al. (2017). They
Electronic copy available at: https://ssrn.com/abstract=3697288
report that all the stock market indexes Granger-cause the CDS markets, but there is also some
evidence of bidirectional causality for some industries.
3. Data
The database consists of daily sovereign bond total return and equity indexes, covering the
years 2010 to 2019, for five of the Eurozone countries most affected by the European sovereign debt
crisis- Greece, Ireland, Italy, Portugal, and Spain- and their two largest economies (France and
Germany), which are used as benchmarks. For each country, I extracted, from Datastream, the 10-
year sovereign bond total return index, a broad equity market total return index, and eleven industry
, where ����! = max ����, 0�. ����� = −min����, 0�, ����! = max ����, 0�. ����� = −min����, 0�,
and s = 1,…,m indicates the risk level. As in Engle and Manganelli (2004), I estimate the parameters
in equations (10) and (11) by minimizing the regression quantile loss function, and I assess the
quantile adequacy using their dynamic quantile test.
Using the estimated values-at-risk, I compute 0B�� ≡ 0��DE�� , … , E � F and 0B�� ≡0��DE�� , … , E � F, the empirical counterparts of 0����� and 0�����. Then, I obtain the test statistic
proposed in Candelon and Tokpavi (2016) by following the steps below.
1 – Compute the sample cross-correlation matrix between 0B�� and 0B��
ΛBH� ≡⎩⎪⎪⎨⎪⎪⎧M�� N D0B�� − ΠB�FD0B��P� − ΠB�F33
�Q�!P 0 ≤ H ≤ M − 1M�� N D0B�!P� − ΠB�FD0B�� − ΠB�F33
�Q��P 1 − M ≤ H ≤ 0
(12)
, where ΠB� and ΠB� are the sample means of 0B�� and 0B��, respectively.
2- Calculate the corresponding sample cross-correlation matrix
REH� = SDΣE�F��/;ΛBH�SDΣE�F��/; (13)
, where D represents the diagonal form of a matrix, and ΣE� and ΣE� are the sample covariance matrices
of 0B��and 0B��, respectively.
Electronic copy available at: https://ssrn.com/abstract=3697288
3- Compute the quadratic form
VE = N W; X HYZ [EH�3��PQ�
(14)
, where W is a kernel function (Bartlett, Daniel, and Parzen kernels are popular choices), M is a