Liquidity and price discovery in the European corporate bond ...
Liquidity and price discovery in the European corporate bond market
Research proposal submitted by Bruno Biais and Fany Declerck
Toulouse University – IDEI
Bruno Biais is Professor of Finance at IDEI, Toulouse University, Place Anatole France,
31000 Toulouse. email@example.com. 33 (0)5 61 12 85 98
Fany Declerck is Professor of Finance at IDEI, Toulouse University, Place Anatole France,
31000 Toulouse. Fany.Declerck@univ-tlse1.fr. 33 (0)5 61 12 85 73
Liquidity, transparency and price discovery in the European corporate bond market
1) Research objectives
The European corporate bond market is quite large. In 2004, in the Eurozone, the amount
outstanding was close to € 600 billion for bonds issued by non financial corporations, and
close to € 4000 billion for bonds issued by banks and financial intermediaries.i The sum of the
two amounted to about 70% of the Eurozone GDP (Source ECB statistics). Its counterpart in
the US was close to 60% of GDP.
While this market is large and important, for the financing of the economy and in terms of
investment opportunities, it is somewhat under researched. In particular, while we know a lot
about price formation and liquidity of European stock markets, we still have very little
empirical knowledge about the microstructure of the European corporate bond market.
The purpose of this research project is to investigate empirically the workings of the
secondary market for corporate bonds in Europe. How are trades conducted? How are prices
formed? How is liquidity supplied? Who are the major players in the market (dealers, fund
managers, platform organizers, etc…) and what is their role? What information is available to
the different market participants? How efficient and liquid is the market? What is the bid-ask
spread? How do these features vary across segments of the market (industries, maturity,
For stock markets, a rich set of empirical microstructure studies is available. For corporate
bond markets, in contrast, there are only very few empirical microstructure studies.ii Until
recently, high frequency datasets were not available for corporate bonds. A few papers have
recently analyzed US data, and, as far as we know, the present paper will be the first one to
analyze microstructure data from the European corporate bond market.
Results obtained for the US Municipal Bond Market
Because they are tax exempt, munis are held in large measure by individual investors (35% in
2001), insurance companies (10%) or personal trusts. Such investors could differ from the
typical investors in stock markets, either because they are less sophisticated (individuals) or
because they are unlikely to desire to trade frequently (insurance companies). Market makers
play a key role to provide intertemporal intermediation and price discovery services.
Harris and Piwowar (2004) and Green, Hollifield and Schurhoff (2004) obtained data from
the Municipal Securities Rulemaking Board (MSRB). The MSRB requires that dealers report
all trades by midnight of the trade date. Most bonds in their samples average less than one
transaction per week. The two studies find that Municipal Bonds are more expensive to trade
than equity. Effective spreads average 2% for retail size trades of $ 20,000 and about 1% for
institutional trade size of $ 200,000. In contrast with stock markets, small trades incur greater
transactions costs than large ones.
Harris and Piwowar (2004) and Green, Hollifield and Schurhoff (2004) argue that large
spreads on small trades stem from lack of transparency. It’s difficult for retail traders to figure
out an appropriate estimate of the market valuation of the security. This puts them in a weak
bargaining position. To the extent, however, that the level of transparency did not vary in this
market, during the data collection period of these 2 studies, it’s impossible to be sure that the
results are driven by the opaqueness of the market. The case of the corporate bond market,
which we discuss below, offer an interesting contrast in this respect.
Results obtained for the US Corporate Bond Market
In contrast with munis, corporate bonds are not tax exempt. Correspondingly, they have a
very different investor basis. As mentioned by Schultz (2001), in 1997 individual investors
owned only 9% of the total, while institutions held the remaining 91%. Among the latter, life
insurance companies were the biggest holders (32%), while private pension funds owned 10%
of the bonds, and foreign investors held 16%. Like the muni market, the secondary market for
bonds has traditionally been opaque, but its transparency has been raised recently.
In 2002, the National Association of Securities Dealers began a program of increased
transparency for corporate bonds, known as the Trade Reporting and Compliance Engine.
Initially trades were to be reported within one hour and fifteen minutes. At the July 2002 start
of TRACE, post trade information was publicly disseminated only for very large and high-
quality issues. Post-trade information dissemination for the other bonds was phased in later.
Edwards, Harris and Piwowar (2004) obtained reports on every bond trade reported to
TRACE in 2003. At that time, for bonds with dissemination requirements, the TRACE system
disseminated transaction prices no later than 45 minutes after the trade. Edwards et al estimate
the bid-ask spread for a retail order size of $ 20,000 to 1.38% of price (138 basis points). This
is much larger than the average spread for retail trades in the equity market, which
approximately amounts to .4%. This ranking is surprising, as one would expect spreads to
increase in risk and adverse selection, which are likely to be greater for stocks than for bonds.
Note also that the average retail spread for municipal bonds is greater than its counterpart for
corporate bonds. Again, this is surprising, as default risk is on average much lower for
municipal bonds than for corporate bonds. For trades of $ 200,000 the average spread is
estimated to only .54%. Spreads in the corporate bond market decrease with trade size. This
contrasts strikingly with the results obtained in the equity market.
This analysis is complemented by a fascinating real scale experiment conducted by Goldstein,
Hotchkiss and Sirri (2005). The authors designed an experiment, in collaboration with NASD.
They formed a sample of 90 BBB rated bonds, for which transparency was introduced. They
also collected data for a matched sample of 90 bonds, for which transparency was not
introduced. By comparing liquidity in the treatment and control samples, the authors are able
to identify the effect of increased (ex--post) transparency. Goldstein, Hotchkiss and Sirri
(2005) find that, for all but the smallest trade size group, spreads decrease for bonds whose
prices become transparent more than for the control set of non disseminated bonds. This effect
is strongest for intermediate trade sizes: for trades between 51 and 100 bonds, relative to their
controls, spreads on the 90 disseminated bonds fall by approximately 30 basis points more
(per $100 dollars face value). On the other hand, Goldstein et al find that transparency does
not affect trading volume. Furthermore, they find no significant effects of increased
transparency for very infrequently traded bonds.
Bessembinder, Maxwell and Venkataraman (2005) complement the two papers discussed
above by focusing on institutional trades. They also find a decrease in transactions costs for
bonds that are not subject to TRACE reporting.
3) Method, expected results and practical value of the project
We have already obtained two extremely interesting data sets for more than 1000 bonds,
denominated in € or in £. IIC, the company which computes the IBOX index, has provided us,
for 2004 and 2005, with the daily closing bid and ask quotes for all the bonds in its index.
ICMA, the self regulatory organization based in London, has provided us with data on all the
trades for these bonds in 2004 and 2005. For each trade we know the price, the quantity, the
time and if the customer was buying or selling to the dealer. We also observe an anonymized
dealer code, enabling us to compute the market share of the different dealers. To complement
this data, we will retrieve, from data stream, information on the characteristics of the bond,
such as issue size, yield, maturity, duration, rating etc…
We will compute measures of liquidity and transactions costs in this market: number of trades
and volume per day, quoted spread, effective spread. We will study how these measures vary
across bonds as a function of the characteristics of the bonds: issue size, yield, maturity,
duration, rating, and currency of denomination. We will compare these results to those
obtained in the US corporate bond market. This will shed light on the relative liquidity of the
two markets. This is particularly interesting as the US market is now post-trade transparent,
due to the TRACE regulation, while the European market is not.
We will also study how the supply of liquidity varies across bonds. How many dealers quote
prices for a given bond? What is the market share of the most active dealer in this bond? How
do these features vary across bonds? How do they relate to the liquidity of the bond, and its
Further, we will study if trades have information content. In stock markets and in the foreign
exchange market, trades have been found to convey information. In the bond market this issue
has not been studied yet. The information content of trades can be expected to be lower than
for stocks, but it can’t be ruled out that trades convey information about market perceptions of
default risk. To test for this, we will run regressions of changes in the midquote onto signed
trades. To study the progressive adjustment of prices to information in this opaque market, we
will estimate how long it takes for the information content of the trade to be fully impounded
in the price. We will study how these measures vary with the characteristics of the bonds. In
particular, we expect the informational content of trades to be larger for low ratings bonds.
The set of results to be generated by this study should be extremely valuable in practice. Fund
managers need to be able to estimate and forecast transactions costs when making investment
and trading decisions. The bid-ask spread estimates we will obtain will be useful in that
respect. Hedge funds need to anticipate how the market will respond to their trades. The
informational content measure we will estimate should be useful in that respect. Regulators
and professional bodies need to evaluate how liquid and efficient the market is, in order to
inform their policy recommendations. Our empirical study will provide the first elements of
information on the microstructure of the European corporate bond market.
February 2006 – June 2006: Data collection, organizing the data, first summary statistics and
June 2006 – November 2006: First draft of the paper, seminar presentations, further
December 2006 – May 2006: Finalized draft of the paper, available for presentation at an
Inquire seminar, revision of the paper and submission to an academic journal.
In this project, we will consider jointly the non financial corporate bond market and the market for bonds issued by
financial institutions. In terms of secondary market microstructure the two work similarly and are closely integrated.
Some interesting studies have focused on Treasury bond markets, where data was released somewhat earlier than for
corporate and municipal bond markets. The Treasury market is markedly different from the corporate market and is
beyond the scope of the present project.