Working Papers, Data and Programs

Low-Latency Trading (with Gideon Saar, October, 2010)

The Best Bid and Offer: A Short Note on Programs and Practices (October, 2010)

Trading Costs and Returns for US Equities: Estimating Effective Costs from Daily Data (August, 2006).
This supercedes the February, 2005 draft.

Technology and Liquidity Provision: The Blurring of Traditional Definitions (with Gideon Saar)

Limit Orders and Volatility in a Hybrid Market: The Island ECN (with Gideon Saar)

Price Discovery Analysis in SAS (programs and data related to "Intraday Price formation in US Equity Index Markets, J. Finance)

Liquidity in the Futures Pits: Inferring Market Dynamics from Incomplete Data

Also see: Markov Chain Monte Carlo Methods for Bayesian Estimation of Microstructure Models

SAS programs used to generate the results in Stalking the “Efficient Price” in Market Microstructure Specifications: An Overview (Journal of Financial Markets, 2002)

SAS routines to do random-walk decompositions with arbitrary number of securities (innovations)

Trading Fast and Slow: Security Market Events in Real Time (February, 1999)

New York Stock Exchange Systems and Trading Procedures (with George Sofianos and Deborah Sosebee, 1993)

TORQ Database (50 MB, zipped). Using the TORQ Database (1993)


Trading costs and returns for US equities: Estimating Effective Costs from Daily Data (August, 2006)

Joel Hasbrouck

The effective cost of trading is usually estimated from transaction-level trade and
quote data. This study proposes a Gibbs estimate that is based on daily closing prices. In
a broad sample of US firms over a period when both high-frequency TAQ and daily
CRSP data are available (1993-2005), an annual Gibbs estimate based on daily data
achieves a correlation of 0.965 with the TAQ value. The approach is extended to a panel
specification in which effective costs for individual stocks are driven by a latent common
factor. In the comparison sample, the estimated series for the common factor based on
daily data achieves a correlation of 0.447 with the corresponding TAQ value at a daily
frequency (0.670 at a monthly frequency). The firm-specific factor loadings estimated
from daily data are also positively correlated with the loadings estimated from
transactions data.

The Gibbs estimates are employed in asset pricing specifications over a longer
historical sample (1927-2005). The results offer only weak support for the view that
effective cost (as a characteristic) affects expected stock returns, except when interacted
with a January seasonal dummy variable. An asset’s return covariance with the common
factor of effective cost is not found to be a determinant of expected returns. The
difference between these results and those of analyses based on other liquidity proxies
indirectly suggests the importance of trading volume. The latter quantity is used in most
daily liquidity proxies, but does not enter the effective cost estimates constructed here.

Download pdf

Datasets (including historical Gibbs estimates of effective cost)


Technology and liquidity provision: the blurring of traditional defintions

Joel Hasbrouck and Gideon Saar

This paper presents a cross-sectional empirical investigation of the relations between volatility and various

The usual economic perspective on a limit order emphasizes its role in supplying liquidity. We investigate the trading of 300 Nasdaq-listed stocks on the Island ECN, an electronic communication network organized as a limit order book. We find that a substantial portion of the limit orders are cancelled within an extremely brief time. We term “fleeting orders” those limit order that are cancelled within two seconds of submission, and explore the role they play in trading strategies. Our principal finding is that fleeting limit orders are closer substitutes for market orders than for traditional limit orders. Our results suggest that the aim of a trader who submits a fleeting order is to demand immediacy. This contrasts with the traditional view of limit order traders as patient providers of liquidity. We hypothesize that a “new equilibrium” has arisen, driven by improved technology, the emergence of an active trading culture, and increased market fragmentation. The new environment transforms the market from one in which prices are posted (visible limit orders) into one where searches (for hidden liquidity) are needed in order to achieve better terms of trade.

Download pdf


Limit Orders and Volatility in a Hybrid Market: The Island ECN

Joel Hasbrouck and Gideon Saar

This paper presents a cross-sectional empirical investigation of the relations between volatility and various measures of activity on the Island ECN, an Alternative Trading System for US equities that is organized as an electronic limit order book. We find that higher volatility is generally associated with

We find weaker evidence that higher volatility is associated with lower depth in the book. In addition, we find that Island's market share for a given firm is positively related to the overall level of Nasdaq trading in the firm, and document substantial use of hidden limit orders (for which the submitter has opted to forgo display of the order). Finally, over one quarter of the limit orders submitted to Island are canceled (unexecuted) within two seconds or less. The extensive use of these "fleeting" orders is at odds with the view that limit order traders (like dealers) are patient providers of liquidity.

Download pdf

Return to top of page


Price Discovery Analysis in SAS (Version 1.0)

Joel Hasbrouck

This paper describes a suite of SAS programs to analyze price discovery high-frequency market microstructure price data. The programs implement the price discovery analysis for the S&P 500 reported in "Intraday Price Formation in US Equity Index Markets"

Download pdf documentation

Download programs and data (about 5MB, zipped)

Return to top of page


Liquidity in the Futures Pits: Inferring Market Dynamics from Incomplete Data
forthcoming JFQA

Joel Hasbrouck

Motivated by economic models of sequential trade, empirical analyses of market dynamics frequently estimate liquidity from regressions of price changes on transaction volumes, where the latter are signed (positive for buyer-initiated trades; negative for seller-initiated trades). This paper implements these models for futures transaction data from pit trading. To deal with the absence of timely bid and ask quotes (which are used to sign trades in most equity-market studies), this paper proposes new techniques based on Markov chain Monte Carlo estimation. The model is estimated for four representative Chicago Mercantile Exchange contracts.
The model structure implies a decomposition for long-run price volatility into trade- and non-trade-related components. For the pork belly, Euro and UK £ contracts, trades explain over half of the (long-term) price volatility. Trades in the S&P 500 Index contract, however, explain only about eight percent of the volatility.

Download pdf

Programs and data

Return to top of page


Markov Chain Monte Carlo Methods for Bayesian Estimation of Microstructure Models

Joel Hasbrouck

This computational appendix to "Liquidity in the Futures Pits" describes the Markov Chain Monte Carlo (MCMC) estimation of microstructure models with bid/ask spreads, discreteness, clustering and trade impacts. In all cases, the data are presumed to consist solely of trade prices and (optionally) trade volumes. The exposition discusses models of increasing complexity.

The appendix is distributed in two forms: a pdf text document, and a Mathematica notebook in which derivations are interspersed with the text. The notebook requires the Mathematica reader (freely available through Mathematica).

Download pdf version. Download Mathematica notebook. The notebook requires the Mathematica reader, a free download from the publisher (Wolfram).

Return to top of page


Trading Fast and Slow: Security Market Events in Real Time

Joel Hasbrouck

February, 1999

Continuous security markets evolve as a sequence of timed events. This study is a descriptive analysis of NYSE market data in which trades, quote revisions and orders are considered to constitute a stationary multivariate point process, which can be analyzed by standard time- and frequency-domain techniques. There are three principal findings. (1) Although occurrence intensities for different types of events are positively correlated, they are not characterized by the uniform proportionality that a strict sense of time deformation would require. (2) The frequencies and durations of informational epochs (periods of uncertainty and informational asymmetry) are highly variable. (3) The correlation in arrivals of market orders and opposing limit orders is zero or negative over periods of thirty minutes or less.

Download pdf

Return to top of page


New York Stock Exchange Systems and Trading Procedures

Joel Hasbrouck, George Sofianos and Deborah Sosebee

April 27, 1993

This paper provides a selective description of New York Stock Exchange systems, trading rules and procedures. The paper's primary objective is to provide researchers with a detailed institutional framework for studying quote and transaction data generated by U.S. securities trading. It is also meant to serve as a guide to the New York Stock Exchange system, for economics, business and legal scholars needing a reference aid for their research. Among the topics examined are: order entry and execution, trade and quote reporting, the audit trail, SuperDot, the Intermarket Trading System, crossing orders and the upstairs positioning of large block trades. The paper provides descriptions of New York Stock Exchange systems, rules and procedures that are constantly changing, as they were at the beginning of 1993.

Download pdf

Return to top of page


Using the TORQ Database

Joel Hasbrouck

1992

The TORQ database contains transactions, quotes, order processing data and audit trail data for a sample of 144 NYSE stocks for the three months November, 1990 through January 1991. This document covers installation, formatting and use of the data. (The data are distributed by the NYSE.)

Download pdf

Return to top of page