Stock Exchange Automation

Jamal Munshi

• Publisher: Routledge
• Year: 2017
• Print ISBN: 9781138563759
• eBook ISBN: 9781351348560
• Topic: Business
• Subtopic: Business General
• Index: Business

Originally published in 1994, Stock Exchange Automation addresses the pivotal role played by capital markets in the market economics. Capital markets are an essential component of the free market system. The book argues that the capital markets function as an allocator of investable funds among competing uses. The movement toward automated markets requires that we understand how automation changes market behaviour. The book also examines the concept of market microstructure theory, and the implication that some forms of automation should affect prices. Theories of price formation in the specialist based trading system hypothesise that the trading mechanism induces short term price volatility.

Data Analysis and Results

Excess volatility data of 685 stocks at the NYSE that were converted from the manual book to the electronic book between April 1987 and December 1989 are studied. A cross sectional analysis of price volatility data of these stocks in ‘event time’ during this period reveals convincing evidence that the ‘excess volatility’ of stocks is lower after the electronic book is installed.

This effect is detectable only when the trading volume is high and when the market for the stock is ‘moving’. Additionally, the evidence does not suggest that the lower volatility has been achieved at the expense of liquidity. The observed effect may therefore be interpreted as an improvement in market quality.

Three sources of data were used in this study. These are: the implementation schedule of the Electronic Book (shown in Appendix 1) that was provided by SIAC (Securities Industry Automation Corporation); historical daily stock price and volume data purchased from CRSP (Center for Research on Securities Prices); and the Wall Street Journal Index.

The event window used is sixty trading days wide (there are approximately 22 trading days per month). This includes a 20-day ‘comparison period’ prior to EDB (Electronic Display Book) implementation during which the stock is assumed to be traded using the manual book. The comparison period is followed by a 20-day period, called a ‘learning period’ during which no data are sampled. The test period comprises the 20 days following the learning period when the stock is assumed to be completely traded on the EDB. The class variable ‘Trading’ is then assigned a value of ‘manual’ for trading days in the comparison period and a value of ‘EDB’ for trading days in the test period.

Ideally, we would expect there to be forty observations per stock; twenty on each side of the treatment However, the CRSP file does not contain all the required data values for all trading days. Intraday high and low prices are available only for trading days since April of 1987. Stocks with conversion dates prior to April 1987 were also dropped from the sample. Stocks with fewer than thirty observations were eliminated to ensure that at least ten observations will be made in each side of the EDB implementation. Observations of stock prices of less than five dollars and trading volumes of less than 10,000 shares per day and observations made during the market breaks of 1987 and 1989 were dropped. These elimination procedures left a sample of 795 stocks with 24,645 observations from a list of some 1,200 stocks in the implementation schedule.

A sample cleaning procedure was used to reduce the effect of uncontrolled external events that may change the stock volatility in the test or comparison period. For each stock in the implementation schedule, the WSJI (Wall Street Journal Index) was searched for unusual events in a 4-month window surrounding the implementation date. Data for stocks that were under the influence of external events were considered ‘noisy’ and removed from the sample since an observed difference in stock price volatility between the two periods may be due to the external event rather than the EDB. This process resulted in the elimination of 3876 observations (110 stocks) out of a total of 24,645 observations (795 stocks) leaving 20,769 observations (685 stocks). The external events that caused a stock to be classified as ‘noisy’ were taken from the synthesis of event studies by Copeland and Weston (1988) and include takeovers, mergers, new listings, bond rating changes, lawsuits, new stock offerings, and stock buybacks.

COMPUTATION OF EXCESS INTRADAY VOLATILITY

The hypothesis of this study concerns intraday volatility that is in excess of that explained by interday volatility and is defined in terms of a linear regression model as the residual when intraday volatility is regressed against the interday volatility (controlling for price). This quantity is the response variable of this quasi-experimental study and is referred to here as “excess volatility”. The regression model is described in Chapter 3.

Exhibit 4–1 shows the results of the regression analysis. The model explains over 42% of the sum of squared deviations from the mean intraday volatility of all 20,769 observations. The residuals of this model represent a portion of the intraday volatility that is not explained by the interday volatility and therefore, after Barnea (1974), may contain information on trading mechanisms and market microstructure. This regression serves as a way of separating mechanism information from economic information. The excess intraday volatility is then computed as: y = 2.411 + r, where 2.411 is the intercept term in the regression model and r represents the residuals of the regression model.

Check for Parallelism of Control Variables

The objective of the design is to make inferences about the treatment means of the response variable y=excess volatility. The ‘treatment’ here is the trading method which has two levels, (’manual’ or ‘EDB’). The level ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• I An Overview of Capital Markets and of the Nature of this Investigation
• II Market Microstructure and the Impact of Information Technology: Previous Research
• III The Empirical Test
• IV Data Analysis and Results
• V Summary and Conclusions
• Bibliography
• Index