This dissertation investigates the effects of earnings announcements on the bid-ask spread in the stock market. The purpose of this investigation is to see how earnings announcements affect the dead-weight loss of trading in the stock market. To perform this investigation we employed data from Oslo Stock Exchange (OSE), provided by Oslo Børs Informasjon (OBI).
We define the bid-ask spread as the difference between the best ask and bid-price quoted in the Order Book at OSE. The bid-ask spread is set to cover the dealers cost of trading and is considered as a friction in the stock market. Earnings announcements are regular and considered as highly anticipated events. In the period before year 2000 the earnings announcements were published between 2 and 4 times a year depending on the company. From year 2000 and onwards all companies listed at OSE present their results 4 times a year.
Assume a stock market with two kinds of market participants: investors and market makers. Investors are traders who buy and sell shares. A market maker is a middleman who matches buyers and sellers and takes the opposite position of each transaction. The market maker is a supplier of immediacy.
The bid-ask spread consists of three main parts. These parts are the order processing cost, the holding cost and the informational cost. Hence, we can express the spread as a function:
Bid-Ask Spread = f (order-processing cost, holding cost, informational cost)
Demsetz (1968) introduces the order-processing cost. The order-processing cost is the cost of standing ready to trade and to process the actual transaction. Stoll (1978) model the inventory holding cost, which is the cost of holding a less than fully diversified portfolio. Finally, Copeland and Galai (1983), Glosten and Milgrom (1985) and Kyle (1985) present different theories describing the impact of asymmetric information on the bid-ask spread.
Kim and Verrecchia (1991) model how anticipated earnings announcements affect the bid-ask spread prior to the announcements. They find that the spread will increase prior to the announcements due to an increase in the informational asymmetry. Kim and Verrecchia (1994) models how earnings announcements affect the bid-ask spread after the announcements. They find that the spread widens during and after earnings announcements due to increased informational asymmetry. Lee, Mucklow and Ready (1993), Krinsky and Lee (1996), Brooks (1996) and Yohn (1998) support these results.
Investigating the impact of earnings announcements on the spread we employed real time data containing information about all orders and informational events registered at OSE in the period 02.01.1995 to 23.10.2000. As a result of the introduction of a new trading system at OSE we divided our sample into two periods. The new system was introduced 05.02.1999. The period prior to 05.02.1999 is called period 1 and the period from 05.02.1999 and onwards is called period 2.
Employing the method of Ordinary Least Squares we find that the bid-ask spread is significantly wider 24 hours prior to earnings announcements for less frequent traded shares in period 1. In addition, we find that the bid-ask spread is significantly wider 24 hours after the earnings announcements. Our results support the theory and previous findings. The trading activity seems to increase in the 24-hours period after the announcements.
In addition, we find that the bid-ask spread is negatively related to the volume in Period 1, but positively related in period 2. This may indicate that the spread is more sensitive to large volumes in period 2. We find a positive relationship between the inter-transaction time and the spread in both periods, which may be interpreted as reduced holding costs. The variability of returns and the spread exhibits a positive relationship for less frequent traded firms in period 1, which supports the theory of a positive relationship between the risk and holding cost. The rest of the companies show a negative relationship, which may be a result of cyclical changes in the variance and use of historical data to measure the variable. Finally, we find that the spread exhibit intra-day patterns. The spread is wider during the first-half hour of trading in the morning as a result of a higher informational asymmetry.
The introduction of a new computerised trading system at OSE seemed to reduce the bid-ask spread. We find evidence of reduced order-processing costs as a result of replacement of the traders and less paperwork associated with each transaction. Finally we find evidence of reduced informational costs as the market became more transparent and efficient. Hence, the new system resulted in smaller frictions in the stock market at OSE.
We experienced problems with heteroscedasticity and autocorrelation employing the OLS method. In order to account for autocorrelation we employed the AR(1) correction and applied the method of Maximum Likelihood to estimate our parameters on three randomly selected firms. We did not attempt to control for heteroscedasticity as we consider this beyond the scope of this dissertation. In short, the spread seemed significant wider for the small firms in the 24-hours period prior to the announcements in period 1. The spread seemed significant wider during the 24-hours period after the announcements. These results support the results from earlier investigations. Most of the parameters explaining the non-event spread were highly significant.
We suggest that more sophisticated econometric methods are applied in further investigations. It would be interesting in further investigations include a variable for the variability of the companies results to make the investigation more interesting for managers in companies.