Introductory Study of Ultra-High Frequency Forex Spot Rate Data
Quant Finance Wednesday Wisdom: You need high-frequency data for high-frequency trading. In 1991, the first academic study to look at high-frequency data was published.
High-Frequency Trading System Needs
High-frequency trading (HFT) systems depend on technology to execute and gain a competitive edge.
Hand-wavily, the system needs to encompass
Co-location services (locate servers in the same data centers as the exchanges, etc.)
Low-latency networks with direct market access (minimal delay between a trade being executed and when it’s confirmed - FPGAs, ASICs, high-speed fiber-optic cables, etc.)
Trading software and platforms (order management system [OMS], execution management system [EMS], smart order routing [SOR], liquidity manager [LM], risk management system [RMS], etc.)
Algorithms (scrapper, analysis, selection, execution, risk, back-testing, confirmation, EDA, etc.)
Data (historical, current, proprietary, etc.)
Data Management (scrapper/aggregator, data storage, data processing, data visualization, data security, data monitoring, data reporting, etc.)
Development and Research Tools (systems to do the daily work of managing all of the above)
All are important and need to work together to deliver a high-frequency trading system that works. At the heart of developing an HFT system is market data for the particular market you’ll be working in.
In today’s Quant Finance Wednesday Wisdom post, we look at the first academic study to look at high-frequency data. This study was published in 1991 and looked at ultra-high frequency, minute-by-minute data (!!!) for forex spot rates (bid-ask Reuters quotes) on three days, Autumn 1987. This is amazing, given that HFT traders are now looking at “5 nano-second wire-to-wire reactions”.1
The paper
The paper is called “Every minute counts in financial markets”.2
The abstract of the paper reads:
This paper represents an introductory study of ultra high frequency, minute-by-minute data, for forex spot rates (bid-ask Reuters quotes) on three days, Autumn 1987. The frequency of price revision, size of spread, and statistical characteristics are measured. The series exhibit (time varying) leptokurtosis, unit roots, and first-order negative correlation, the latter especially in disturbed ‘jumpy’ markets. The effect of time aggregation on these characteristics is examined, and variance ratios are analyzed. Multivariate analysis revealed significant relationships between lagged exchange rates, both the own rate and the key Deutsche mark/US dollar rate, and the current spot rate.3
Note, if you don’t have a Science Direct account, you can find my copy here (just for friends, dropbox link, don’t distribute)4 . You can also ask your local economist for a copy as well.
Paper Summary
This paper, published in the "Journal of International Money and Finance" by C.A.E. Goodhart and L. Figliuoli, studies the minute-by-minute movement in foreign exchange (forex) rates. It analyzes data from three specific days in Autumn 1987, focusing on the frequency of price changes, bid-ask spreads, and other statistical features of forex spot rates.
The five key findings include:
Price Revisions and Spreads: The paper observes high frequency of price revisions. Bid-ask spreads are found to mostly take a few standard values and don’t show significant widening even during turbulent market conditions.
Statistical Characteristics: The data exhibit certain statistical traits like leptokurtosis (higher likelihood of extreme values than a normal distribution) and unit roots (indicating a random walk behavior). The paper notes negative autocorrelation, particularly during volatile market periods, implying that a rise or fall in prices tends to be followed by a move in the opposite direction.
Impact of Time Aggregation: The study also examines how the aggregation of data over time affects these characteristics. It finds that high-frequency data show less leptokurtosis and heteroskedasticity (variability of variance) compared to lower frequency data.
Inter-Relationships Between Currencies: A noteworthy aspect of the research is its analysis of how changes in one currency's value could influence others. The study suggests that some currencies, especially the Deutsche Mark [editor’s note: RIP], have a significant influence on others.
Market Efficiency and Reaction to News: The paper delves into the efficiency of the forex market and its reaction to new information. It finds that market reactions to news are not always immediate or clear, indicating potential inefficiencies.
Overall, this 1991 paper provides valuable insights into the high-frequency dynamics of the forex market, exploring how currencies interact and respond to changes within minutes, and the implications of these movements for market efficiency.
It’s an interesting historical read and gives you a sense of how much things have changed and how much they have stayed the same.
Enjoy!
That’s all for today :) For more Quant Finance treats, check out our archives.
Stay quanty!
All the best,
Sebastian
https://www.linkedin.com/posts/stefanschlamp_eurex-latency-marketdata-activity-7014356023566028800-V1mI
Goodhart, C. A. E. and Figliuoli, L., (1991), Every minute counts in financial markets, Journal of International Money and Finance, 10, issue 1, p. 23-52.
https://www.sciencedirect.com/science/article/abs/pii/026156069190025F
https://www.dropbox.com/scl/fi/ujnx63eu3gf9x52fn40l5/goodhart1991.pdf?rlkey=rvbochtamdlz2pd0n0spsbpxz&dl=0