Are All Credit Default Swap Databases Equal?

Date01 September 2014
Published date01 September 2014
Are All Credit Default Swap Databases
Sergio Mayordomo
School of Economics and Business Administration, Universidad de Navarra, Spain
Juan Ignacio Peña
Department of Business Administration, Universidad Carlos III de Madrid, Spain
Eduardo S. Schwartz
Anderson Graduate School of Management, UCLA, CA, USA
We compare the ve major sources of corporate Credit Default Swap prices: GFI,
Fenics, Reuters, CMA, and Markit, using the most liquid single name 5year CDS in
the iTraxx and CDX indexes from 2004 to 2010. Deviations from the common trend
among prices in the different databases are not random but are explained by
idiosyncratic factors, nancing costs, global risk, and other trading factors. The
CMA quotes lead the price discovery process. Moreover, we nd that there is not a
full agreement among databases in the results of the price discovery analysis
between stock and CDS returns.
Keywords: credit default swap prices, databases, liquidity
JEL classification: F33, G12, H63
This paper was partially drafted during the visit of Sergio Mayordomo to the Anderson
School at UCLA. We acknowledge financial support from MCI grant ECO200912551. We
are very grateful to an anonymous referee for many helpful comments and to the editor John
Doukas for very useful suggestions. Moreover, the paper has benefitted from comments by
Rodolfo Campos, Teresa Corzo, Lars Norden, Yves Nosbusch, J. Pedro Nunes, Jose
Penalva, Maria RodriguezMoreno, Christina Wang, and other participants in the IX
INFINITI Conference, 2011 IFABS Conference, XIX Foro de Finanzas (Spanish Finance
Association), University of Valencia seminar, and 2012 EFMA Conference (Barcelona) for
useful comments. Correspondence: Sergio Mayordomo.
European Financial Management, Vol. 20, No. 4, 2014, 677713
doi: 10.1111/j.1468-036X.2013.12023.x
© 2013 John Wiley & Sons Ltd
1. Introduction
Over the last decade, the Credit Default Swap (CDS) market has grown rapidly.
the growth and the size of this market, quoted and transaction prices of CDS contracts are
widely thought to be a gauge of nancial marketsoverall situation, as suggested by the
GM/Ford credit episode in 2005, the US subprime asco in 20072009 or the Europes
debt crisis in 2010. Academic and policymakers alike have voiced concerns with respect
to the CDS markets role in the above mentioned episodes and its possible inuence in
other nancial markets, creditoriented or otherwise. However, to properly address
current concerns, careful empirical research is needed and therefore dependable CDS
price data is a key requirement. The CDS market is an Over the Counter (OTC) market
almost entirely populated by institutional investors and therefore, in contrast with an
organised exchange like the NYSE, there is no reliable information on prices. The
information on prices must be gathered from market participants on the basis of their
voluntary participation on periodic surveys, with all the potential shortcomings such a
situation may bring about. For instance, Leland (2009) reports that Bloombergs CDS
data is frequently revised weeks after and often disagrees substantially with other data
sources such as Datastream. Given that price data deserve special attention, as the validity
and power of the empirical results must be based on a dependable data source, in this study
we investigate the differences in the main data sources employed by researchers and
policymakers in this area. Specically, we compare the ve data sources for CDS prices
commonly used in almost all the extant research: GFI, Fenics, Reuters EOD, Credit
Market Analysis (CMA) DataVision (CMA hereafter), and Markit. We study the
consistency of these ve CDS data sources in the cross section and time series dimensions
using the most liquid single name 5year CDS of the components of the leading market
indexes, iTraxx (European rms) and CDX (US rms). First we look at their basic
statistical properties. Then we address two specic issues: (i) the factors explaining the
divergences from the common trend among different CDS quoted spreads, and (ii) the
relative informational advantage of the prices coming from different CDS databases.
Finally, we study the consistency among databases in the results of a price discovery
(causality) analysis between stock and CDS returns.
Two price time series for the same single name CDS reported by different data sources
should, in principle, be very close in the sense that both share a common trend, the
underlying true value of the asset. Even if there are deviations from the common trend
between the price series reported by the different datasets, one should expect that these
deviations are random errors and therefore unrelated both to idiosyncratic factors such as
The global notional value of CDSs outstanding at the end of 2004, 2005 and 2006 was $8.42,
$17.1 and $34.4 trillion, respectively. The CDS market exploded over the past decade to more
than $45 trillion in mid2007 and more than $62 trillion in the second half of the same year,
according to the ISDA. The size of the (notional) CDS market in mid2007 is roughly twice
the size of the US stock market (which is valued at about $22 trillion) and far exceeds the $7.1
trillion mortgage market and $4.4 trillion US treasuries market. However, the notional amount
outstanding decreased signicantly during 2008 to $54.6 trillion in mid2008 and $38.6
trillion at the end of 2008. This declining trend followed in 2009 (31.2 in mid2009 and 30.4 at
the end of 2009), but seems to have stabilised in 2010 and 2011.
© 2013 John Wiley & Sons Ltd
678 Sergio Mayordomo, Juan Ignacio Peña and Eduardo S. Schwartz
rm size, the volatility of the rm equity prices, or the disagreement/agreement of
analystsforecasts on the rms earnings, and to systematic factors such as the nancing
costs of the participants in the CDS market, global risk, or trading activity. If all the data
sources are consistent among them, the use of a given data source should not affect
research results and their nancial and policy implications. But, if there are signicant
deviations among them, the research implications may be sensitive to the specic data
base employed. Any inconsistency in prices from private providers would also have
implications for market transparency which would affect all nancial agents such as
investors, risk managers, and regulators.
We nd that there are systematic departures from the common CDS spreadstrend
across databases. Our analysis suggests that, although the different CDS quotes move
broadly together, there are very noticeable divergences for some entities in some days.
Also, the discrepancies among databases appear to be more marked in specic time
periods, probably reecting market turbulences. No single database, however, provides
quotes that are consistently above or below the quotes from other databases. We also nd
evidence suggesting that on average the days without trade price information have higher
quote dispersion than the days with trade price information.
Most importantly, deviations (in absolute value) from the common trend among the
different CDS quoted spreads are not purely random, but are related to idiosyncratic
factors such as the disagreement of analystsforecasts of the companys earnings per
share (EPS) or the rm size, and also to specic CDS liquidity, global risk, global
nancing costs, and trading factors. We also nd that the different data sources do not
reect credit risk information equally efciently. Our results suggest that, for the sample
period considered, CMA quoted CDS spreads led the credit risk price discovery process
with respect to the quotes provided by other databases.
The discrepancies among databases might have a material impact on the results of
price discovery tests between stock and CDS returns given that we do not nd a full
consistency among databases in the results of the stockCDS price discovery (causality)
Our results have a number of important implications for empirical research using
CDS prices. First, for singlename CDS with low trade frequency, our results cast
doubts on the reliability of the existing quoted price information. Second, the larger
the disagreement in the analystsforecast on the rms EPS and the smaller the
rm, the larger are the deviations among databases. Third the higher nancing costs of
nancial institutions and, the higher the global risk (measured by implied volatility
indexes), the larger are the deviations from the common trend in prices across the different
databases. Fourth, there is a CDS database that leads the price discovery process when
they are compared to each other. Fifth, the price discovery analysis between stock and
CDS returns reveals that there is not a full consistency among databases in terms of the
causality relations found between stocks and CDS. Extensive robustness tests would be
needed before a causality relationship between stock and CDS returns could be
This paper is structured as follows. Section 2 presents a literature review. Section 3
describes the data employed in the analysis. Section 4 motivates the research hypotheses
and introduces the methodology. Section 5 shows the empirical results. Section 6
addresses the extent to which the differences among databases impact the price discovery
tests between stock and CDS returns. Section 7 conrms the robustness of the results and
presents some extensions. Section 8 concludes.
© 2013 John Wiley & Sons Ltd
Are All Credit Default Swap Databases Equal? 679

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT