Stochastic Volatility Structures and Intraday Asset Price Dynamics

Abstract

The behavior of financial asset price data when observed intraday is quite different from these same processes observed from day to day and longer sampling intervals. Volatility estimates obtained from intraday observed data can be badly distorted if anomalies and intraday trading patterns are not accounted for in the estimation process.

In this paper I consider conditional volatility estimators as special cases of a general stochastic volatility structure. The theoretical asymptotic distribution of the measurement error process for these estimators is considered for particular features observed in intraday financial asset price processes. Specifically, I consider the effects of (i) induced serial correlation in returns processes, (ii) excess kurtosis in the underlying unconditional distribution of returns, (iii) market anomalies such as market opening and closing effects, and (iv) failure to account for intraday trading patterns.

These issues are considered with applications in option pricing/trading strategies and the constant/dynamic hedging frameworks in mind. Empirical examples are provided from transactions data sampled into 5-, 15-, 30-, and 60-min intervals for heavily capitalized stock market, market index, and index futures price processes.

Appendix

At the time the ASX data were collected, the exchange had just previously moved from floor to screen trading with the six main capital city exchanges linked via satellite and trade data streamed to trading houses and brokers instantaneously via a signal G feed. The SFE maintained Pit trading for all futures and options on futures contracts at the time.

Legal restrictions on third party use and development of interfaces meant the ASX had a moratorium on such usage and development. The author was required to obtain special permission from the ASX to capture trade data from a live feed from broking house Burdett, Buckeridge, and Young (BBY). There was a further delay in reporting results of research following the legal agreement obtained from the ASX.

Trade data for stock prices and volume of trade were then sampled into 5-min files and subsequently into longer sampling interval files. The market index was refreshed at 1-min intervals and the above sampling scheme repeated. Futures price trades were supplied in two formats via feed: Pit (voice recorded) data and Chit data. Although the Pit data provides an instantaneous record of trade data during the trading day, some trades are lost during frantic periods of activity. The Chit records are of every trade (price, volume, buyer, seller, time stamped to the nearest second, etc.). The recorded Chits are placed in a wire basket on a carriageway and transferred up the catwalk where recorders on computers enter details via a set of simplified keystrokes. The average delay from trade to recording is around 30 s for the Chit trades. These are then fed online to trading houses and brokers. At the end of the trading day, these recorded trades are supplemented with a smaller set of records that were submitted to the catwalk late, e.g., morning trades that may have gone to lunch in a brokers pocket and submitted during the afternoon session and also some late submitted trades.

We created the intraday sampled files from both the Pit and Chit records. However, we employed the Chit trades for analysis in this paper so as to have the correct volume of trade details for each trading interval. All trades were reallocated using the time stamps to the relevant time of trade, including trades not submitted on time but supplied as an appendix to the trading day data. In this study the average number of late Chits were not a high proportion of daily trades. These futures price records were then sampled into relevant 5-min records and longer sampling frames generated in the same manner as was employed for the stock prices.

For all series the first price and last price closest to the opening and closing nodes for each sampling interval were recorded with volume of trade the accumulated volume of trade within the interval defined by first and last trade.