Professional Documents
Culture Documents
Bond Futures and Orderflow Imbalance
Bond Futures and Orderflow Imbalance
Bond Futures and Orderflow Imbalance
Imbalance
An International Comparison
Lee A. Smales
School of Banking & Finance
University of New South Wales
Australia
Order imbalance methodology is utilized to examine the link between trading activity and returns
in the six most liquid international bond futures markets. Order imbalances are strongly related to
contemporaneous returns, in the expected direction (i.e. excess buy (sell) orders push down (up)
yields, even after controlling for aggregate market volume. There is evidence of contrarian
investor behaviour following an increase in yields, but continuation of order imbalances when
yields are falling (the prices of bond futures are rising). International bond futures markets are
strongly intertwined with the US market having a strong influence on the returns and order-flow
across all countries; this is likely an indication of the spill-over effect of US macroeconomic data.
Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1955810
http://ssrn.com/abstract=1955810
1. Introduction
Futures markets play a key role in the global financial system, allowing owners of physical
assets to hedge their price risk and providing speculators with a fast and inexpensive method of
positioning their capital. Ultimately, bond futures markets affect the cost of funding of both
corporate and governments. Previous literature has studied the relationship between trading
activity and the return of financial assets but has mainly focused on US equity markets and the
use of volume as a proxy for trading activity. Benston and Hagerman (1974), Foster and
Viswanthan (1990), Hiemstra and Jones (1994) and Lo and Wang (2000) find that volume is
positively related to price change, and closely linked to liquidity. Measuring trading activity by
volume may actually conceal information; trading volume can be high either due to a
preponderance of buyer-initiated or seller-initiated trades, or because there is generally a large
amount of trading interest that is evenly distributed between buyers and sellers; each possibility
having implications for prices and liquidity.
Chan et al (1999), Chan and Fong (2000), and Hasbrouck and Seppi (2001), study order
imbalance in US equity markets over relatively short periods and find that there is a strong
predictive ability for subsequent stock returns. Chordia et al (2002) conducted the first extended
study using order imbalance on NYSE stocks and found that order imbalances are strongly related
to contemporaneous absolute returns, as well as past market returns, and that investors exhibit
contrarian behaviour in aggregate.
Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1955810
http://ssrn.com/abstract=1955810
More recently, variants of the order imbalance methodology have been used to analyse
the foreign exchange and fixed income markets. Evans and Lyons (2002) and Evans (2002)
provide evidence that order flow contains information about economic fundamentals, and so
performs an important function that explains price movements. Pasquariello and Vega (2007) find
that unanticipated order flow has a significant and permanent impact on daily bond yield
changes, and correlation between order flow and price changes is higher when the dispersion of
beliefs among market participants is high. Underwood (2008) looks at the cross-market
relationship between equities and bonds and notes that aggregate order imbalances play a strong
role in explaining returns. Brandt and Kavajecz (2004) examine price discovery in the US
Treasury market, finding that order-flow drives price movements, accounting for up to 26% of
the variation in yields on days without macroeconomic announcements.
To date, non-US Fixed Income market and futures markets have largely been neglected in
this research. With the substantial size, scope and importance of the market for Government
bond futures there is need for a greater understanding of both the drivers of returns within
individual futures markets and the information flow between the markets of different countries.
The market for Government bond futures differs from equity markets in several crucial ways, and
thus provides numerous reasons for extending the existing work on order imbalance within
equity markets to the bond futures market. Firstly there is a different clientele with private or
individual investors constituting a smaller part of the futures market. Second, the bond futures
market is more frictionless than the equity market, with lower transaction costs, reduced liquidity
concerns, and no short-sale constraints. Third, equity order imbalance on any given day may be
driven by firm-specific factors, but this cannot be the case with bond futures which trade on the
basis of macroeconomic factors. Additionally, there is a different inventory adjustment process
within the bond futures market; while specific market makers exist within the equity market, this
is not the case with bond futures. However, the interest rate swaps (IRS) and physical bond
market does have market makers who use will use futures to hedge – it is therefore possible that a
different transmission process operates between order-flow and returns in such a market. Finally,
and perhaps most importantly, there is a lower level of fragmentation within the bond futures
market; equities trade across a number of exchanges and dark-pools, for example Madhaven
(2012) examines U.S. equities trading across 14 exchanges with just 40% of volume trading on the
Electronic
Electroniccopy
copyavailable
availableat:
at:https://ssrn.com/abstract=1955810
http://ssrn.com/abstract=1955810
two largest exchanges, while each bond future only trades on one primary exchange 1. As a result
of the difference in the level of fragmentation it may be possible that a more explicit, and directly
examinable, relationship exists between order imbalance and the returns on bond futures.
This paper has several key aims. Firstly, a variant of the order imbalance methodology
(developed by Chordia et al (2002) and continued by Chordia and Subrahmanyam (2004)) is
applied to futures markets for the first time in order to examine the principal causes of order
imbalance in the six largest Government Bond futures markets, and provides an international
comparison. Secondly, the relationship between order imbalance and market returns is
investigated, enabling a deeper understanding of aggregate trading behaviour in bond futures
markets. Finally, a Vector Auto Regression (VAR) framework is utilised to examine the
relationship between order flow and returns across the six international bond futures markets,
providing a greater understanding of the inter-market information flow and a demonstration of
the linkage of global capital flows.
The key findings are summarized as follows. Order imbalances are related to past market
returns. Market participants exhibit contrarian behaviour following market declines, but there is
continuation of order imbalances following market rallies. There is evidence of price reversals
following both positive and negative returns, with positive returns witnessing the larger reversal.
Order imbalances are strongly related to contemporaneous returns with excess buy (sell) orders
driving up (down) prices, even after controlling for aggregate market volume. International bond
futures markets are strongly intertwined, in terms of order-flow and returns. The US market has a
strong influence on returns and order flow across global markets; this is likely an indication of the
spill-over effect of US macroeconomic data.
The rest of this paper is organized as follows: Section 2 discusses the nature of the data
and methodology used in this paper. Section 3 provides analysis on the causes of order imbalance.
Section 4 provides discussion on the connection between market returns and order imbalance.
Section 5 examines the inter-market relationship between the six largest Government bond future
markets. Section 6 concludes the paper.
1
US Treasury Futures are an exception as whilst over 90% of the volume is transacted electronically
through CME Globex there is still the option of open outcry trading.
Government bond futures allow a wide variety of market participants to adjust their
interest rate exposure, and are among the most liquid financial products. The price of a bond
future is based on the price of a nominal underlying security, or basket of underlying securities, as
specified by the respective exchange, with a conversion factor adjusting for the delivery of eligible
securities that differ from the nominal specifications. Prices are therefore quoted in a similar
fashion to physical bonds whereby a bond future priced against a nominal bond with a yield to
maturity (YTM) higher (lower) than the nominal coupon rate will trade below (above) par.
Australian 10-Year Treasury Bonds are an exception to this pricing rule as their price is expressed
as 100 – YTM of an underlying basket of physical bonds.
Transaction level data on government bond futures is collected from Thomson Reuters
Tick History (TRTH) for the period 3rd January 2005 to 31st December 2010, a total of over 1,500
trading days. Data is gathered for the six most liquid and highly traded 10-Year Government Bond
Futures; US 10-Year Treasury Note (TY), German Bund (FGBL), Japanese Government Bonds
(JGB), UK Long Gilt (FLG), Australian 10-Year Treasury Bonds (YTC), and Canadian 10-Year
Government Bond (CGB) 2. In the interests of parsimony, and to ensure consistency across the
international comparison, only the results for 10-year bond futures are reported in this paper;
analysis of non-10-year bond futures (e.g. US 5-year notes and German Schatz 2-year notes) is
qualitatively similar. Consistent with prior work on fixed-income markets (e.g. Brandy and
Kavajecz (2004), Engsted and Tanggaard (2007) and Smales (2012)) returns are calculated on the
basis of implied yields rather than prices – this has the additional advantage of allowing
comparison of returns in bond futures priced in differing ways, for example the Australian 10-
Year Bond.
To ensure that the study concentrates on the most liquid contracts only the nearest
trading contract is considered. In-line with market convention, the contract under consideration
is rolled on the last day of the month preceding the delivery month. There are several motives for
using futures markets in this analysis; first, future prices and transaction data are readily available
on a tick-by-tick basis. Second, bond futures markets are open on an extended basis (virtually 24-
2
Equivalent Bloomberg tickers are: TY, RX, JB, G, XT and CN respectively.
In addition to transaction level data the daily closing level of the implied yield for each
contract is garnered from Bloomberg. Figure 1 illustrates the evolution of yields over the sample
period. Yields in Canada, Germany, UK and US have seemingly tracked closely, with the
relationship becoming stronger as yields fell quickly during the onset of the Global Financial
Crisis (GFC) in 2008, and then rose in unison from the start of 2009. The implied yield in the four
countries also fell together in mid-2010 before increasing towards the end of the year. Yields in
Australia have generally been higher over the sample period (averaging 5.36% v US average of
4.27%), although a clear association still exists and this intensifies during the crisis period of 2008-
09 as yields fall quickly before recovering. Japanese yields provide the biggest outlier during the
sample period, averaging just 1.65%, never rising above 2.22%, and suggesting a lower level of
correlation.
<Insert Figure 1>
2.3 Order imbalance variables
Futures market transactions are identified as a buy (sell) transaction when occurring at
the price equal to the offer (bid) prevailing immediately prior to the transaction. Prior work on
the equity markets has relied on the Lee and Ready (1991) algorithm to designate each transaction
as either buyer-initiated or seller-initiated using the established. The recognised difficulties faced
in applying the algorithm within US equity markets, namely trades inside the spread and
identification of the prevailing quote, are thus avoided when considering the electronically traded
Government bond futures market. Quotes and trades are excluded if recorded before the open or
after the closing time each day. For each trading day a number of variables are computed for each
of the identified contracts:
Rt: the return for the implied yield of the futures contract during interval t, defined as
Rt = log (Yt / Yt-1) × 100;
Panel B reports correlations among the measures of order imbalance, the concurrent daily
return, total volume, and the total number of transactions. Unsurprisingly, there is a strong
positive correlation between the various measures of order imbalance (OIBTRA and OIBVOL)
and market activity (NUMTRANS and VOLUME). Returns are negatively, and significantly,
correlated with order imbalance (OIBTRA); this makes intuitive sense since a positive order
imbalance, indicating a preponderance of buy orders, will likely push prices higher and yields
lower. There is generally no significance to the relationship between volume and returns,
suggesting that trading activity and returns are related through order imbalance, rather than
through trading volume. Given the high correlation between the two measures of order
imbalance across the six markets, and the strong relationship between OIBTRA and returns, the
findings reported in the remainder of the paper use only OIBTRA 3.
3
OIBVOL produces qualitatively similar results.
Many factors may contribute to order imbalance: market returns, macroeconomic news,
daily return and weekly volatility regularities (e.g. Gibbons and Hess (1981) and Chordia et al
(2001)) and the reversal of temporary price pressures have all been cited as possible factors. This
section investigates whether order imbalance can be predicted using past market returns after
controlling for weekly regularities and past lagged values of order imbalance. Two regression
specifications are considered; first, for each bond the daily order imbalance in the number of
transactions (OIBTRA) is regressed on past values of order imbalance together with past market
returns and on day of the week dummies. Second, following Chordia et al (2002), the differential
impact of prior market declines, and market rallies, on order imbalance is examined.
𝑂𝐼𝐵𝑇𝑅𝐴𝑡 = 𝛼𝑐 + ∑4𝑖=1 𝛽𝑖 𝑂𝐼𝐵𝑇𝑅𝐴𝑡−𝑖 + ∑4𝑗=1 𝛽𝑗 𝑀𝑖𝑛(0, 𝑅𝑡−𝑗 ) + ∑4𝑘=1 𝛽𝑘 𝑀𝑎𝑥(0, 𝑅𝑡−𝑘 ) + ∑4𝑙=1 𝛽𝑙 𝐷𝑎𝑦𝑙 + 𝜀𝑡 (2)
Table 3 – Panel A reports the time-series regression for equation 1. A significant and
positive intercept indicates that, absent any other factors, the number of buy orders will outweigh
the number of sell orders, this is consistent with markets where yields have generally trended
lower (prices higher) – Australia is the exception with a negative intercept and yields trending
higher. There is also a significant and positive Monday irregularity for each of the markets,
indicative of risk-aversion among traders who seek to buy bond futures on a Friday to reduce risk,
or effectively undertake portfolio insurance against negative news, and then reverse this action on
a Monday. Such trading activity over weekends is common when markets are closed and there is
potential of asymmetric news events that could move yields sharply lower – for example, during
the Lehman bankruptcy in 2008 a number of the major actions by the US Treasury and the US
Consistent with the highly significant autocorrelation detailed in the descriptive statistics,
order imbalances are highly predictable based on lagged order imbalances for up to four days. This
finding is analogous to the US equity market results in Chordia et al. (2002) where the persistence
is found to last for at least five days. The persistence in bond future order imbalance is likely due
to two reasons; firstly, the macro-economy and hence yields tend to move in cycles, with a
recessionary (expansionary) environment often witnessing economic news that gradually gets
worse (better) with the result that order-flow into the market tends to trend too. Secondly,
investors in the physical market, particularly insurance and pension companies, are apt to
purchase bonds as assets in order to match their long-term liabilities, hence the market-makers
who deal with such investors will tend to hedge their interest rate exposure by purchasing
futures. Similarly, treasury departments of commercial banks will be inclined to buy bond futures,
rather than sell, when hedging their mortgage convexity risk.
A positive coefficient for lagged returns indicates that following an increase in yields
(decline in prices) the positive order imbalance increases. That is, participants in the bond futures
market are contrarian; buying following market declines and selling (or at least buying less)
following market advances. Again, this is consistent with Chordia et al. (2002) who identify US
equity market investors as contrarian. As expected given the insignificant level of autocorrelation
for returns past one-day there is no significance in lagged returns of more than one-day in this
regression.
<Insert Table 3 >
Equation 2 seeks to obtain a more detailed examination of the relationship between lagged
returns and order imbalance by disaggregating past returns into negative and positive
components, the results are reported in Table 3 – Panel B. A number of the prior results are
confirmed; a generally positive intercept, a significant Monday effect (Tuesday in Australia) and
high persistence in order imbalance across all six futures markets. Of more substance are the
coefficients for disaggregated returns; revealing that market participants are generally contrarian
In summary, consistent with prior studies on the US equity markets, order imbalances are
highly predictable based on lagged order imbalances for up to four days. A significant and positive
Monday irregularity exists for each of the markets, indicative of risk-aversion among market
participants seeking to protect against adverse market movements over a weekend. Lagged returns
tend to indicate that bond futures markets participants are contrarian in nature, buying following
market declines and selling following market advances. However, disaggregating lagged returns
reveals that market participants are generally contrarian in markets where yields are rising (prices
falling) but tend to follow continuation strategies when yields are falling (prices rising).
The relationship between market returns and order imbalances is examined by means of
three regression specifications. Firstly, returns are regressed against contemporaneous and lagged
order imbalance as well as lagged returns (equation 3). Second, following Chordia et al. (2002),
order imbalance is disaggregated into positive and negative parts to allow for the assessment of the
4 A negative return for the prior period multiplied by a negative coefficient results in a positive impact on
OIBTRA
10
Table 4 reports the results for the three specified regressions. For the base specification,
equation 3, contemporaneous order imbalance is negative and significant implying that positive
(negative) order imbalance induce lower (higher) yields as one would expect, and consistent with
Brandt and Kavajecz (2004). For Canadian, German and UK bond futures this relationship is
reversed for lagged order imbalances; this is consistent with inventory stabilization, or contrarian
behaviour, whereby the previous day’s imbalance is reversed. The intercept is indistinguishable
from zero in all cases, which is warranted given the average daily return close to zero. Finally, the
significant single period autocorrelation is confirmed by the positive and significant relationship
between returns and lagged returns.
<Insert Table 4>
Disaggregating order imbalance into excess buy and sell orders reveals additional
information. As expected, excess buy (sell) orders impel contemporaneous returns lower (higher)
across all markets. Importantly, excess sell orders appear to have an impact on returns of a much
greater magnitude than buy orders. This finding is likely a result of excess sell orders forming a
stronger signal than excess buy orders in a sample period where the majority of intervals witness a
positive order imbalance. Additionally, during the GFC, implied yields for the examined markets
reached the lowest level in decades and as a result market participants were likely wary of any
indication of a potential move lower in bond prices (yields moving higher).
The negative coefficient found across all markets for lagged excess sell orders suggests that
lagged order imbalance exerts a mean-reversal impact on the current day’s return after controlling
for contemporaneous order imbalance. Although this result is only significant for US and
Australian markets it does add to the evidence of contrarian trading behaviour in bond futures
markets following intervals of excess selling (which would have eventuated into increases in
11
Excess buy orders appear to drive the identified difference in the relationship between
contemporaneous and lagged order imbalances for Canadian, German and UK futures; for excess
buy orders the contemporaneous relationship is significantly negative whilst the relationship with
the lagged variable is significantly positive. These three markets have similarities in the average
level of yield during the sample period and in the correlation of both order imbalance and
returns. For US bond futures the significance of the relationship between returns and order
imbalance appears to be driven by excess sell orders, consistent with the thought that a
preponderance of excess buy orders in the sample period places more emphasis on any days with
excess sell orders. In the market for Japanese bond futures, where yields are much lower than
other markets, order imbalance does not appear to have a significant impact on returns. Finally,
both excess buy orders and excess sell orders appear to have a significant impact on returns in the
Australian market, where yields have been higher on average. This could be a result of the
Australian market seeing a much more balanced number of negative and positive order
imbalances than the other markets. However, note that excess sell orders still have a greater
impact in Australia – where yields also reached multi-year lows during the GFC.
The final specification includes lagged negative and positive market returns, and order
imbalance is found to remain significant even in the presence of lagged returns. Lagged return
coefficients are significant for Canadian, and US markets, where changes in yields are closely
linked owing to the strong economic and trade links between the two countries. In both cases,
lagged positive returns (rising yields) are followed by positive returns in the current period, while
lagged negative returns (falling yields) are reversed. The magnitude of the coefficient for negative
lagged returns is much greater than that for positive lagged returns; since negative returns (falling
yields) in bond futures markets are often driven by negative economic news this is consistent
with work on the equity markets where Cox and Peterson (1994) find reversals following large
stock price declines.
Previous literature has focused extensively on the relation between volume and volatility.
However, this paper argues that daily imbalances could provide information about movements in
the prices of bond futures in addition to that provided by daily volume. For example, if aggregate
12
|𝑅𝑡 | = 𝛼𝑐 +𝛽1,𝑡 Max[0, 𝑂𝐼𝐵𝑇𝑅𝐴𝑡 ] + 𝛽2,𝑡 |Min[0, 𝑂𝐼𝐵𝑇𝑅𝐴𝑡 ]�+𝛽3,𝑡 Log(Volume𝑡 � + 𝛽4,𝑡 |𝑅𝑡−1 | + 𝜀𝑡 (7)
Table 5 provides evidence on the relationship between the daily market contemporaneous
return (a proxy for market volatility) order imbalance, aggregate volume, and the lagged absolute
market return. Considering the result for aggregated order imbalance (equation 6) first; the
coefficient for OIBTRA is generally positive implying that order imbalance is significant and
positively related to volatility. Surprisingly, the volume coefficient is generally significant and
negative, suggesting that once order imbalance is controlled for an increase (decrease) in volume
results in falling (rising) yield volatility. For German Bunds the overall level of trade volume has
no significant effect on market returns. Lagged volatility has a positive and significant relationship
with contemporaneous volatility, confirming the persistence of returns identified earlier. That is,
order imbalance is significant in explaining contemporaneous market returns even after
controlling for volume.
<Insert Table 5>
The disaggregation of order imbalance into positive and negative parts confirms the prior
findings, and exposes additional information. Again an increase (decrease) in order imbalance
induces an increase (decrease) in the magnitude of market movements, while aggregate volume
has the opposite effect. The positive relationship with lagged returns remains. Importantly,
disaggregation of order imbalance reveals that excess sell orders have a much greater impact on
market volatility than excess buy orders. This finding is consistent with the notion that sell orders
reveal a stronger signal than excess buy orders in an era where buy orders predominate, and
corroborates evidence presented in Table 4. The result also aligns with previous work on US
equity markets (e.g. Chordia et al. (2002)) which notes an asymmetric impact of excess buy and
sell orders.
13
The analysis of the previous sections reveals strong evidence to support the existence of a
significant relationship between order imbalance and market returns at the level of individual
markets. Order imbalance appears to exhibit a high degree of persistence in the individual
process, whilst there is a significant first-order autocorrelation in bond returns. In addition, Table
6 – Panel A reveals a high degree of cross-correlation across markets. In order to correctly model
the dynamics of global bond futures markets the dependencies and interdependencies between
markets have to be explicitly taken into account. A twelve-dimensional model for the endogenous
variables (returns and order imbalances in each of the six countries) is specified in a Vector Auto-
Regression (VAR) framework.
14
Linkages involving returns and order-flow between the six countries seem to be driven by
economic fundamentals, consistent with Froot et al. (2001) and Engsted and Tanggaard (2007).
The US – home to the largest bond futures market as well as the world’s largest economy - has a
strong influence on the returns and order flow across all countries; lagged US returns have a
positive (negative) relationship with contemporaneous returns (order imbalance) in other
countries. A positive return (rising yields) in the US futures market will induce sell orders and
rising yields in other markets. This is also consistent with the positive relationship between
lagged US order imbalance and current period order imbalance in other markets. Furthermore,
the bond futures markets of countries with close economic linkages, such as Germany with the
UK, and Canada with the US, appear to have close relationships in terms of a significant
association between lagged returns and current returns and order imbalance.
Overall, the dynamic analysis confirms the effects identified in Section 4. Additionally,
there is evidence that the bond futures markets of countries with significant economic ties are
closely intertwined. Importantly, the US futures market has a strong influence on the returns and
order flow across all countries considered in this analysis.
5 For conciseness only one lag for each of the dependent variables is reported, this is deemed appropriate
since additional lags are not statistically significant.
15
The relationship between trading activity and market returns have been explored
extensively, particularly with regards the US equity markets. Trading activity has typically been
measured by volume, but more recent work (e.g. Spiegel and Subrahmanyam (1995) and Chordia
et al (2002)) suggests that the imbalance between buyer and seller initiated orders could be a
powerful determinant of price movements beyond trading volume. Analysis of the determinants
of order imbalances, and of the relation among order imbalance and returns, on an international
basis produces clear evidence of similarities across markets during the sample period.
Order imbalances are related to past market returns. Examining lagged market returns
there is evidence of contrarian investor behaviour among market participants. However,
disaggregating returns into positive and negative components reveals that the contrarian
behaviour is only prevalent following an increase in yields (falling futures prices), whilst investors
tend to follow a continuation strategy (continue to buy) when yields are falling and prices rising.
Order imbalances are strongly related to contemporaneous returns after controlling for
aggregate market volume. Contemporaneous order imbalance exerts an impact on market returns
in the expected direction i.e. excess buy (sell) orders drive yields down (up). Order imbalance also
has an impact on contemporaneous volatility (absolute daily returns) even after controlling for
trading volume. These results are consistent across all countries apart from Japan where the
market differs owing to the larger notional size of futures contracts, and significantly lower
average yields.
International bond futures markets are strongly intertwined. Returns and order-
imbalance are strongly correlated between markets. The US bond futures market has a strong
influence on the returns and order-flow across all countries; this is consistent with the spill-over
effects of US data.
16
Bjursell, J., A. Frino, Y. Tse, and G.H.K. Wang, 2010, Volatility and trading activity following
changes in the size of futures contracts, Journal of Empirical Finance, 17, 967-980
Bonfim, A., 2003, Pre-announcement effects, news effects, and volatility: Monetary policy and
the stock market, Journal of Banking and Finance, 27, 133-151
Brandt, M., and K. Kavajecz, 2004, Price discovery in the US Treasury market: The impact of
orderflow and liquidity on the yield curve, Journal of Finance, 59, 2623-2654
Chan, K., Y. Chung, and W-M. Fong, 1999, The informational role of stock and option volume,
Working Paper
Chan, K., and W. Fong, 2000, Trade size, order imbalance, and the volatility-volume relation,
Journal of Financial Economics, 57, 247-273
Chordia, T., R. Roll, and A. Subrahmanyam, 2001, Market Liquidity and trading activity, Journal
of Finance, 56, 501-530
Chordia, T., R. Roll, and A. Subrahmanyam, 2002, Order imbalance, liquidity, and market
returns, Journal of Financial Economics, 65, 111-130
Chordia, T., and A. Subrahmanyam, 2004, Order imbalance and individual stock returns: Theory
and evidence, Journal of Financial Economics, 72, 485-518
Cohen, Benjamin H., and Hyun S. Shin, 2002, Positive feedback trading under stress: Evidence
from the U.S. Treasury Securities Market, Working paper, International Monetary Fund
Cox, D., and D. Peterson, 1994, Stock returns following large one-day declines; evidence on short-
term reversals and longer-term performance, Journal of Finance, 49, 255-267
Engsted, T., and C. Tanggaard, 2007, The co-movement of US and German bond markets,
International Review of Financial Analysis, 16, 172-182
Evans, M., 2002, FX trading and exchange rate dynamics, Journal of Finance, 57, 2405-2447
Evans, M., and R. Lyons, 2002, Order flow and exchange rate dynamics, Journal of Political
Economy, 110, 170-180
Faust, J., J. Rogers, E. Swanson, and J. Wright, 2003, Identifying the effects of monetary policy
shocks on exchange rates using high-frequency data, Journal of the European Economic
Association, 1, 1031-1057
Foster, F., and S. Viswanathan, 1990, A theory of the interday variations in volume, variance, and
trading costs in securities markets, Review of Financial Studies, 3, 593-624
17
Gibbons, M., and P. Hess, 1981, Day of the week effects and asset returns, Journal of Business, 54,
579-596
Glosten, L., and P. Milgrom, 1985, Bid, ask and transaction prices in a specialist market with
heterogeneously informed traders, Journal of Financial Economics, 14, 71-100
Hasbrouck, J., and D. Seppi, 2001, Common factors in prices, order flows and liquidity, Journal of
Financial Economics, 59, 383-411
Hiemstra, C., and J. Jones, 1994, Testing for linear and non-linear Granger causality in the stock
price-volume relation, Journal of Finance, 49, 1639-1664
Ho, T., and H. Stoll, 1983, The dynamics of dealer markets under competition, Journal of Finance,
38, 1053-1074
Kim, S-J., and D. Nguyen, 2008, The spill-over effects of target interest rate news from the U.S.
Fed and the European Central Bank on the Asia-Pacific stock markets, International Research
Journal of Finance and Economics, 20, 27-45
Krauss, A., and H. Stoll, 1972, Price impacts of block trading on the New York Stock Exchange,
Journal of Finance, 27, 569-588
Kyle, A., 1985, Continuous auctions and insider trading, Econometrica, 53, 1315-1335
Lee, C., and M. Ready, 1991, Inferring trade direction from intraday data, Journal of Finance, 46,
733- 746
Lo, A., and J. Wang, 2000, Trading volume: Definitions, data analysis, and implications of
portfolio theory, Review of Financial Studies, 13, 257-300
Madhaven, A., 2012, Exchange-traded funds, market structure, and the flash crash, Financial
Analysts Journal, 68, 20-35
Pasquariello, P., and C. Vega, 2007, Informed and strategic order flow in the bond markets,
Review of Financial Studies, 20, 1975-2019
Smales, L.A., 2012, RBA monetary policy communication: The response of Australian interest rate
futures to changes in RBA monetary policy, Pacific-Basin Finance Journal, 20, 793-808
Stoll, H., 1978, The supply of dealer services in securities markets, Journal of Finance, 33, 1133-
1151
Underwood, S., 2009, The cross-market information content of stock and bond order flow,
Journal of Financial Markets, 12, 268-289
18
Ave. Daily
Reuters Remaining Term Value of 1 Ave. Daily Volume
Contract Exchange Contract Size Min Tick Size Turnover
Code of Bond (Years) tick (Contracts)
(US$ Million)
10‐Year US Treasury Note Futures TY CME/CBOT 6.5 to 10.0 $100,000 1/32 $31.25 648,044 64,804
German‐Bund Futures FGBL EUREX 8.5 to 10.5 €100,000 0.01 €10 207,705 27,801
Japanese Government Bonds JGB TSE 7.0 to 11.0 ¥100,000,000 0.01 ¥10,000 35,502 43,877
UK‐Long Gilt Futures FLG LIFFE 8.75 to 13.0 £100,000 0.01 £10 71,081 11,098
Australian 10‐Year Treasury Bond
YTC ASX 10.0 A$100,000 0.005 A$38 28,043 2,869
Futures
Canadian 10‐Year Government Bond
CGB MX 8.0 to 10.5 C$100,000 0.01 C$10 1,329
Futures 13,382
Swiss‐CONF Futures CONF EUREX 8.0 to 13.0 CHF100,000 0.01 CHF 10 8,533 916
Australian
li 10YR Bonds d (YTC)
( ) Canadian
di Government Bonds d (CGB)
( ) German Bund d (FGBL)
( )
Panel A: Summary Statistics Panel A: Summary Statistics Panel A: Summary Statistics
Mean Std. Dev. Min Max Mean Std. Dev. Min Max Mean Std. Dev. Min Max
OIBTRA 12 166 ‐ 1,421 1,568 OIBTRA 166 212 ‐ 455 1,435 OIBTRA 1,654 1,069 ‐ 226 7,004
Electronic copy available at: https://ssrn.com/abstract=1955810
OIBVOL 298 3,371 ‐ 12,826 15,628 OIBVOL 2,010 2,462 ‐ 5,188 15,052 OIBVOL 88,008 81,576 ‐ 429,478 1,260,574
NUMTRANS 1,029 429 16 3,282 NUMTRANS 1,015 683 1 3,659 NUMTRANS 6,329 1,518 1,021 13,087
VOLUME 28,043 13,938 63 107,500 VOLUME 13,382 8,209 1 43,611 VOLUME 207,705 291,343 84,232 2,295,806
Implied Yield 5.536 0.511 3.848 6.778 Implied Yield 3.918 0.516 2.667 4.874 Implied Yield 3.649 0.550 2.139 4.746
Returns 0.002 1.118 ‐ 4.882 6.762 Returns ‐ 0.024 1.367 ‐ 7.544 8.560 Returns ‐ 0.013 1.309 ‐ 5.730 10.778
Panel B: Correlations Panel B: Correlations Panel B: Correlations
OIBTRA OIBVOL NUMTRANS VOLUME OIBTRA OIBVOL NUMTRANS VOLUME OIBTRA OIBVOL NUMTRANS VOLUME
OIBVOL 0.438 OIBVOL 0.804 OIBVOL 0.444
NUMTRANS 0.007 0.060 NUMTRANS 0.727 0.611 NUMTRANS 0.840 0.037
VOLUME 0.004 0.098 0.541 VOLUME 0.476 0.572 0.773 VOLUME 0.088 0.428 0.092
Returns ‐0.207 ‐0.381 0.000 0.010 Returns ‐0.079 ‐0.059 0.016 0.028 Returns ‐0.062 ‐0.030 0.018 ‐0.019
Panel C: Autocorrelations Panel C: Autocorrelations Panel C: Autocorrelations
Lag (days) OIBTRA OIBVOL VOLUME Returns Lag (days) OIBTRA OIBVOL VOLUME Returns Lag (days) OIBTRA OIBVOL VOLUME Returns
1 1.576 1.723 0.925 2.136 1 1.209 1.404 0.594 2.109 1 0.681 2.002 2.013 2.126
2 1.664 1.856 1.160 2.070 2 1.420 1.539 0.848 2.045 2 0.900 2.002 2.013 1.853
3 1.787 1.889 1.208 1.954 3 1.416 1.522 0.953 1.960 3 1.060 1.716 1.725 2.023
4 1.959 2.036 1.118 2.038 4 1.637 1.671 1.060 2.033 4 1.012 2.001 2.013 1.890
Lagged Order Imbalance 0.034 * 0.051 ** 0.02 *** ‐0.046 0.013 ** 0.011
(OIBTRAt‐1) (1.92) (2.49) (3.43) (‐1.49) (2.27) (1.45)
Excess Buy Orders ‐0.092 *** ‐0.089 *** ‐0.074 *** ‐0.075 *** ‐0.018 *** ‐0.019 *** ‐0.083 ‐0.079 ‐0.028 *** ‐0.029 *** ‐0.008 ‐0.058
Max(0,OIBTRAt) (‐3.47) (‐3.33) (‐3.43) (‐3.48) (‐3.86) (‐4.04) (‐1.40) (‐1.21) (‐2.77) (‐2.89) (‐0.91) (‐0.66)
Excess Sell Orders 0.172 *** 0.170 *** 0.159 0.143 0.160 0.173 0.344 0.300 0.539 ** 0.527 ** 0.153 *** 0.152 ***
|Min(0,OIBTRAt)| (5.78) (5.71) (1.26) (1.14) (0.43) (0.47) (0.91) (0.80) (2.14) (2.09) (4.15) (4.12)
Lag(Excess Buy Orders) 0.013 0.006 ‐0.052 ** ‐0.030 ** ‐0.014 *** ‐0.016 *** ‐0.048 ‐0.079 * ‐0.011 ** ‐0.012 ** ‐0.004 ‐0.008
Max(0,OIBTRAt‐1) (0.49) (0.21) (‐2.42) (‐2.36) (‐3.08) (‐3.27) (‐1.52) (‐1.83) (‐2.04) (‐2.20) (‐0.48) (‐0.85)
Lag(Excess Sell Orders) ‐0.075 ** ‐0.068 ** ‐0.075 ‐0.069 ‐0.160 ‐0.172 ‐0.437 ‐0.442 ‐0.128 ‐0.097 ‐0.074 ** ‐0.081 **
|Min(0,OIBTRAt‐1)| (‐2.52) (‐2.25) (‐0.59) (‐0.55) (‐0.43) (‐0.47) (‐1.17) (‐1.19) (‐0.51) (‐0.39) (‐2.01) (‐2.19)
Lagged Return 0.049 ** 0.045 ** 0.08 *** 0.081 *** 0.061 ** 0.043 **
(Rt‐1) (1.95) (1.98) (3.13) (3.08) (2.37) (2.17)
Lag(Positive return) ‐0.021 0.098 ** 0.089 * 0.030 0.055 0.102 **
Max(0,Rt‐1) (‐0.48) (2.12) (1.91) (1.28) (1.17) (2.24)
Lag(Negative return) ‐0.078 * ‐0.208 *** 0.071 ‐0.027 0.063 ‐0.192 **
Min(0,Rt‐1) (‐1.71) (‐4.12) (1.44) (‐1.25) (1.43) (‐2.04)
No Observations
No. 1535 1536 1535 1337 1340 1340 1513 1514 1514 1451 1458 1458 1506 1508 1508 1506 1508 1508
Note: ***, ** & * indicates statistical significance at the 1%, 5%, and 10% level respectively
Table 5 Does order imbalance have explanatory power for market returns in addition to aggregate volume?
This table reports the regression results for the models specified in equations 6 and 7. In each case the dependent variable is the absolute value of the daily return on the respective bond
future (|Rt| ), a proxy for return volatility. For equation 6, the explanatory variables include contemporaneous daily order imbalances measured in number of trades (OIBTRAt )x100 together.
For equation 7 the order imbalance measure is disaggregated into excess buy (Max(0,OIBTRAt) ) and excess sell orders (|Min(0,OIBTRAt)| ). Both specifications include the log of aggregate
volume (log(Volume t ) ) and lagged absolute returns (|R t‐1 | ) as explanatory variables. (t‐statistics are in parentheses). Data: Jan 2005 ‐ Dec 2010
Adjusted R2 0 074
0.074 0 079
0.079 0 113
0.113 0 119
0.119 0 153
0.153 0 153
0.153 0 258
0.258 0 260
0.260 0 144
0.144 0 149
0.149 0 135
0.135 0 156
0.156
Note: ***, ** & * indicates statistical significance at the 1%, 5%, and 10% level respectively
Panel A: Correlations
OIBTRAt AU CA GE JA UK Returns AU CA GE JA UK
CA 0.072 CA 0.347
GE 0.061 0.385 GE 0.473 0.551
JA 0.022 0.060 0.027 JA 0.233 0.129 0.176
UK 0.071 0.537 0.525 0.094 UK 0.424 0.500 0.757 0.126
US 0.042 0.549 0.608 0.272 0.611 US 0.341 0.738 0.513 0.290 0.465
Panel B: Returns, intermarket order imbalance and volume.
AU CA GE JA UK US
Explanatory Variable Rt OIBTRAt Rt OIBTRAt Rt OIBTRAt Rt OIBTRAt Rt OIBTRAt Rt OIBTRAt
Intercept ‐0.013 ‐21.83 ** ‐0.165 ** 67.83 *** ‐0.093 352.6 *** ‐0.079 97.30 *** 0.027 152.4 *** ‐0.243 *** 39.50 **
(‐0.20) (‐1.98) (‐2.01) (5.11) (‐1.19) (6.89) (‐0.80) (5.44) (0.33) (6.13) (‐2.60) (2.16)
Rt‐1, AU ‐0.222 *** 25.48 *** 0.009 ‐11.93 ‐0.018 25.67 0.122 ** 15.79 ‐0.012 ‐3.353 ‐0.041 11.58
(‐6.69) (3.68) (0.20) (‐1.63) (‐0.43) (0.91) (2.35) (1.60) (‐0.27) (‐0.244) (‐0.83) (0.62)
Rt‐1, CA 0.035 ‐0.585 ‐0.121 *** 6.492 *** ‐0.068 17.64 0.071 22.11 * ‐0.024 ‐3.264 0.133 ** ‐46.01 **
(0.97) (‐0.07) (‐2.58) (2.77) (‐1.54) (1.19) (1.27) (1.93) (‐0.51) (‐0.21) (2.48) (‐2.11)
Rt‐1, GE ‐0.021 ‐2.453 ‐0.007 ‐1.87 ‐0.094 * 26.19 ** 0.022 ‐5.499 0.126 ** ‐35.89 ** ‐0.090 ‐12.74 *
(‐0 51)
(‐0.51) (‐0 31)
(‐0.31) (‐0 12)
(‐0.12) (‐1 20)
(‐1.20) (‐1 84)
(‐1.84) (2 15)
(2.15) (1 42)
(1.42) (‐0 48)
(‐0.48) (2 34)
(2.34) (‐2 25)
(‐2.25) (‐1 45)
(‐1.45) (‐1 87)
(‐1.87)
Rt‐1, JA ‐0.024 2.514 0.028 3.040 0.031 6.966 ‐0.207 *** 11.99 ** 0.033 9.982 ‐0.016 ‐4.631
(‐1.20) (0.63) (1.08) (0.72) (1.28) (0.43) (‐6.74) (2.12) (1.30) (1.27) (‐0.55) (‐0.43)
Rt‐1, UK 0.047 * ‐12.51 * ‐0.034 7.618 0.014 9.635 ‐0.021 ‐12.14 ‐0.170 ** 18.16 ** 0.004 21.31
(1.69) (‐1.69) (‐0.70) (0.97) (0.30) (0.32) (‐0.36) (‐1.15) (‐2.36) (2.24) (0.08) (1.06)
Rt‐1, US 0.335 *** ‐4.042 ** 0.112 ** ‐10.33 *** 0.311 *** ‐57.30 ** 0.240 *** ‐9.721 ** 0.281 *** ‐2.459 * ‐0.166 *** 10.27 **
(9.98) (‐2.57) (2.54) (‐2.82) (7.40) (‐1.99) (4.52) (‐2.34) (6.39) (‐1.80) (‐3.28) (2.54)
OIBTRAt‐1,AU ‐0.015 0.224 *** 0.040 * 0.004 0.042 0.139 ‐0.003 0.048 0.050 ‐0.064 0.030 0.062
(‐0.76) (6.65) (1.79) (0.11) (1.01) (1.01) (‐0.10) (1.00) (1.26) (‐0.96) (1.19) (0.68)
OIBTRAt‐1,CA ‐0.026 0.064 * 0.004 0.309 *** ‐0.018 ‐0.030 0.005 0.033 0.017 0.270 * 0.028 0.122
(‐1.27) (1.81) (0.18) (8.22) (‐0.81) (‐0.21) (0.19) (0.66) (0.74) (1.85) (1.06) (1.26)
OIBTRAt‐1,GE ‐0.007 ‐0.014 ‐0.004 ‐0.004 ‐0.005 0.552 *** ‐0.001 0.023 * ‐0.009 * 0.034 * 0.001 0.080
(‐1 17)
(‐1.17) (‐1 43)
(‐1.43) (‐0 73)
(‐0.73) (‐0 41)
(‐0.41) (‐1 11)
(‐1.11) (13 99)
(13.99) (‐0 12)
(‐0.12) (1 69)
(1.69) (‐1 85)
(‐1.85) (1 77)
(1.77) (0 07)
(0.07) (1 05)
(1.05)
OIBTRAt‐1,JA 0.007 0.019 0.008 ‐0.058 0.016 0.065 ‐0.003 0.397 *** 0.005 ‐0.077 0.043 ‐0.045
(0.48) (0.77) (0.56) (‐1.25) (1.17) (0.65) (‐0.20) (11.34) (0.36) (‐1.57) (1.56) (‐0.67)
OIBTRAt‐1,UK ‐0.001 0.002 0.023 0.035 * 0.020 ** 0.086 0.028 ‐0.034 0.003 0.408 *** 0.013 ‐0.006
(‐0.05) (0.11) (1.20) (1.75) (1.99) (1.13) (1.18) (‐1.28) (0.31) (11.03) (1.07) (‐0.11)
OIBTRAt‐1,US ‐0.004 0.029 ** 0.006 0.048 *** 0.004 0.150 *** ‐0.010 0.046 ** ‐0.010 0.081 *** ‐0.005 0.432 ***
(‐0.52) (2.25) (0.76) (3.48) (0.55) (2.80) (‐1.07) (2.45) (1.24) (3.12) (‐0.61) (12.16)
2
Adjusted R 0.228 0.135 0.026 0.376 0.088 0.628 0.18446 0.423 0.07192 0.589 0.031 0.654
F‐Statistic 24.52 5.45 2.26 21.03 8.06 58.69 18.81 25.51 6.44 49.97 2.68 65.75
Log‐likelihood ‐1367.2 ‐5582.3 ‐1638.5 ‐5631.5 ‐1584.8 ‐6793.1 ‐1820.0 ‐5888.2 ‐1631.1 ‐6170.5 ‐1772.2 ‐6442.3
Note: ***, ** & * indicates statistical significance at the 1%, 5%, and 10% level respectively
5
Australia
Canada
4
Yield (%)
Germany
Japan
3
UK
US
2
0 Source:
Jan‐05 Jan‐06 Jan‐07 Jan‐08 Jan‐09 Jan‐10 Date Bloomberg