University of Stockholm

University of Stockholm

School of Business

Master Thesis

Spring 1998

The Determinants of the Bid-Ask Spread on the Stockholm Stock Exchange:

A Market Microstructure Analysis

By:

Richard Silén

Erik Sundberg

Supervisor:

Adri de Ridder

Abstract:

This paper analyzes the determinants and the dynamics of the bid-ask spread on the Stockholm Stock Exchange. The study is conducted cross-sectionally on data between the 1st of September 1997 and the 27th of February 1998. During this period the stock exchange was open for trading 124 days. The data is divided and analyzed in the different market list the shares are listed in. Comparison reveals that the spread is widest at the OTC list and tightest at the most traded list. The main determinants of the spread are the share price, the risk associated with the share and the trading activity, with an inverse influence on the spread from an increasing share price and trading activity.

Table of Contents:

1. Introduction......

1.1 Problem......

1.2 Purpose......

1.3 Hypothesis......

1.4 Limitations......

2. Method......

2.1 Sample and data......

2.2 Biases......

2.3 Specifications of the regression model......

3. Determinants of the bid-ask spread......

3.1 Trading Activity......

3.2 Risk of the security......

3.3 Stock price......

3.3.1 Tick Size......

3.4 Trade Size......

3.5 Turnover......

4. Theories about the bid-ask spread......

4.1 Order-processing cost......

4.2 Adverse selection cost......

4.3 Inventory cost......

4.4 Previous studies......

5. Spread......

5.1 Quoted Bid-Ask Spread......

5.2 Effective spread......

6. Results......

7. Conclusions......

BIBLIOGRAPHY......

Appendix 1: Regression results......

Appendix 2: Descriptive data......

Appendix 3: Shares included in the study......

Appendix 4: Omitted sharesdue to incomplete data......

Appendix 5: Omitted shares due to crossing of tick size segment:......

Appendix 6: Listing requirements:......

Appendix 7: Market Descriptions:......

Appendix 8: Description of the Swedish equity market......

1. Introduction

In a market with perfect competition, no transaction costs, complete information and rational expectations, there would only be one price on the share. Hence the spread in an order book would not exist and all transactions would be executed immediately at the current market price. On dealer markets the spread would only consist of the dealers fixed and variable costs for performing the transaction. However if at least one of these assumptions does not hold in reality, a spread will exist. The cost of a trade has at least two components. Apart from a commission cost, there is an implicit cost, the bid ask spread. This cost arises as selling even the modest amount of share, one gets a lower price than one has to pay for it. By recognizing the market microstructure, it is possible to reduce this cost.

This study examines the determinants and the dynamics of the bid-ask spread on the Stockholm Stock Exchange. Considering the importance of the bid ask spread and the manifold of research of this subject made on the US exchanges, we were surprised not to find any single study examining the dynamics and the determinants of the bid ask spread on the Stockholm Stock Exchange. To understand and to be able to compare this study to studies on the US exchanges one has to be aware of the differences in market structure. The US markets are markets with less than full automation of the execution of the trades.

Moreover, studies conducted on markets with fully automated trading (LOB-markets) generally apply the existing theories, from dealer markets, without interpreting and analyzing the differences between the two different types of markets. We believe that in order to do this, a comparison of the different market systems is necessary. Essential differences in market structure most certainly influence market characteristics such as e.g. trading costs and liquidity[1]. This give rise to the theory that dealer spreads are not equivalent to spreads on LOB markets[2].

The outline of the paper is as follows. In the following section we specify the problem, purpose, hypothesis, limitations and the methodology of the study. Subsequently succeeds a presentation of existing theories in the field of market microstructure. We thereafter thoroughly describe the determinants of the spread. The two types of spread are thereafter modeled, succeeded by a presentation of the sample and data used in the study. The last section contains results from the regressions, analysis of the data and comparison with previous studies.

1.1 Problem

The questions we want to answer in this study are:

• Which factors influence the spread?
• Does an increase in the stock price automatically lead to an increase in the bid ask spread? An increase in share price ought to result in an identical relative enlargement of the spread. If not, there is relatively cheaper to trade in higher priced shares.
• Does the bid-ask spread vary with listing? We assume a higher liquidity on the main-list than on the complementary lists. If the spread is significantly related to liquidity we will see that the spread do vary with the list.
• What is the relationship between the quoted and the effective spread?

1.2 Purpose

The purpose of this paper is to examine the bid ask spread on the Stockholm Stock Exchange. It is important as the spread contains a part of the total trading cost. We will investigate if the spread vary with the price of the security and the different lists. The relative importance of security characteristics such as trading activity, the stock price, risk and the average trade size of the security will be identified. By performing regression analysis we will determine which variables explain the size of the bid-ask spread and to what extent. We intend to discuss the fundamental differences between dealer markets and limit order book markets.

1.3 Hypothesis

We assume that the spread is inversely related to liquidity. With this in mind, we expect a higher spread on stocks listed on the O-list and the OTC-list, and the lowest spread on the list for the most traded securities – the A1-list. Previous studies indicate that the spread should be positively related to the risk of the security and the average trade size. A U-shaped form of the total spread in relation to trade size is expected.

1.4 Limitations

Many studies dealing with the bid-ask spread are conducted on an intraday basis when analyzing the spread. We assume that the study by Niemeyer and Sandås (1995), who investigated the intraday behavior of the spread, holds. We therefore restrict our study to comprise only daily data.

Such variables as foreign competition, firm size, turnover and ownership concentration are omitted in our analysis. Some of these variables are implicitly included. E.g. the variable firm size is included through the determinant volume as larger companies normally demonstrate higher trading volume.

The spread is also influenced by such factors as earning announcements and dividend policy. Our paper will not estimate the effects of new information on the bid-ask spread.

2. Method

We will perform a cross-sectional study on daily closing data from the Stockholm Stock Exchange. This is in line with previous studies and will simplify comparisons.

The minimum price variation rule for equity[3] may influence the relationship between the determinants of the spread and the quoted/realized spread. We therefore split the data in different price segments, i.e. in each tick size segment, before conducting the regression analysis. Arithmetic means of both the quoted and the effective spread will be calculated on all segments.

2.1 Sample and data

202 shares were included in the study. The shares on Stockholm Stock Exchange not included in the study were omitted due to either incomplete data or crossing of tick size rules. The sample period is six month, starting the 1st of September 1997 until the 27th of February 1998. During the period the stock exchange was open for trading 124 days, which also leaves us 124 observations of closing data. In appendix three to five the included and omitted shares are listed.

When conducting the regressions, we sort the stocks by price and also by the lists. We define the A1 list as the most traded shares on the A-list. A2 is the rest of the stocks on the A-list. The OTC-list contains smaller companies and the O-list embodies the unlisted securities.

We identify to types of spreads, the quoted and the effective spread. We define the quoted spread as the ask quotation minus the bid quotation at any given time in the limit order book. The effective spread on the contrary is based on the actual transaction prices. The average difference between the price at which a dealer sells at one point in time and the price at which a dealer buys at an earlier point in time. In the literature one can find definitions such as effective and realized spread. Effective and realized bid-ask spread are however identical by definition. In this study, the term “effective spread” will be used.

2.2 Biases

The effective spread should be a measurement of the actual difference between the price at which an broker buys at one time and sells at an earlier or later point in time, i.e. the actual inferred transaction cost. When approximating the effective spread, the recorded transaction price and the quoted bid-ask spread are supposed to be at the same time. In our study though, there might be a brief time difference between the both.

2.3 Specifications of the regression model

We will use the following regression when estimating the determinants of the bid-ask spread.

(1)

where:

spread = quoted or effective spread

 = a constant

activity = the number of transactions for each trading day

risk = the intraday stock price volatility

price = the closing stock price

size = the average trade size for each trading day

3. Determinants of the bid-ask spread

3.1 Trading Activity

An inverse relationship between spreads and trading activity is expected. This is partly explained by the fact that every trade adds new information to the market, a large number of trades reduces informational asymmetries and lowers informed trader advantages[4]. The tradeoff between traders placing market orders and traders placing limit orders amplify this inversely relationship. By this we mean that traders that places limit orders gain on the expense of traders demanding immediacy. According to basic economic theory, an inverse relationship between spreads and trading activity holds if economies of scale predominate.

We measure the trading activity with the amount of trades during a trading day.

3.2 Risk of the security

We estimate the risk of the security as the stock price volatility. The models dealing with price volatility in information asymmetry states that in periods with high volatility, the informed traders are likely to gain at the expense of the less informed traders. As a consequence, less informed traders will compensate themselves from expected loss, by widening the spread.

Theories about the risk impact on the spread are not uniform. Stoll (1978) argues that the spread is affected by the total risk, i.e. both systematic and unsystematic risk, of the security and others e.g. Tripathy and Peterson (1991) dispute whether or not systematic- and/or unsystematic risk are correlated with the bid-ask spread. A relative simple model for measuring the volatility of the stock, on an intraday basis, has been used by e.g. Aitken and Frino (1996). This model suggest that the volatility may be proxy as the standard deviation of the midpoints of the bid-ask spread occurring within each measured time interval.

We believe that a better measurement of the risk of the security, on a daily basis, are the difference between the highest and the lowest price, for a specific trading day, divided by the average share price. Where the average share price is defined as the total value of the traded shares divided by the total number of traded shares. This is also consistent with Tinic and West (1972).

(2)

where:

ASP = the average share price

TVT = the total value of the traded shares for each trading day

NTS = the number of shares traded for each trading day

High = the highest traded price during a trading day

Low = the lowest traded price during a trading day

3.3 Stock price

The absolute bid-ask spread is suggested to be direct proportional to the stock price. The relative spread, however, proves to be inversely related to the stock price. This was first shown empirically by Stoll (1978).

For higher priced stocks, the fixed order processing costs will be spread over a greater trade value. If variable costs, such as labor and communication cost increase with the number of transactions, and if transactions are made in standard trade sizes it will be relative cheaper to execute larger trades and this in turn will lead to lower relative spreads and therefore an inverse relationship is expected. Too big differences in the relative spread also stimulate arbitrage transactions[5].

Benston and Hagerman (1974) also showed that the relative spread is less than proportional with the stock price. Therefore investors prefer to trade in high-priced stocks so that total transaction cost are reduced[6]. They also found that the price per share was the most important explanatory variable determining the spread.

If trades are executed in standard sizes (in number of shares) and expenses increase with the number of transactions but are invariant to transaction value, then stocks with higher price will tend to have lower relative spreads.

The above-presented theory however, is not applicable on the Stockholm Stock Exchange where standard trade size is not measured as a fixed number of shares[7]. Nevertheless Silén and Sundberg (1997) found share price to be a significant determinant of the bid-ask spread. We have not found any theories and are unable to explain why this inversely relationship also exists on the Stockholm Stock Exchange.

3.3.1 Tick Size

Many studies suggest that the tick size is a significant determinant of the spread. The Tick Size on the Stockholm Stock Exchange minimizes the spread to between 0.2 and one percent for normally priced share and a spread at up until five percent for shares priced above SEK 10.00[8].

The intraday average spread is on the Stockholm Stock Exchange considerable larger than one tick according to Niemeyer and Sandås (1995). This corresponds with Silén and Sundberg (1997) whose findings suggest that the tick size is an insignificant determinant of the spread. However to avoid the tick size from not influencing the regressions we decide to divide the shares in different tick segments when conducting the regression analysis.

The tick size rule determines the minimum movement in the share price. One consequence of using a tick size rule is that a particular share with a large tick size relative to the price will have a minimum quoted spread that is greater than a share with a lesser relative tick size. This implies a greater spread for a share with a larger relative tick size than for a share with a lesser tick size. A wider tick size enhances liquidity by reducing bargaining and processing costs and by providing more incentives for limit orders and market makers to provide liquidity. A wider relative tick size, however, also increases the minimum quoted bid-ask spread. A Company may split its stocks to move its share price into the range where the tick size is optimal relative to the share price[9].

A nonzero tick puts a floor on the quoted bid-ask spread, which provides incentives for dealers to make markets and thus increase liquidity. On a LOB market a great tick size increases the chance that an investor will place a limit order, and hence provide liquidity and depth in the market.

Table 1. Tick size rules on Stockholm Stock Exchange[10]

Share price, SEK / Tick size
< 5 / 0.01
5  / 0.05
10  / 0.10
50  / 0.50
500  / 1.00

3.4 Trade Size

As described in the section about the components of the spread, trade size measure all three components of the spread. Therefore the expected relation between the spread and the average trade size is not so clear-cut. Still, earlier studies have indicated that in most cases a positive connection between trade size and the spread exists. Consequently, we anticipate a positive relation between trade size and the spread.

Degryse (1996) finds a U-shape form in the bid-ask spread varying with the trade size on the Belgian Stock Exchange[11]. The same pattern is observed on the Paris Bourse[12]. His results on the London SEAQI[13] market, showed an independent relationship between trade size and the bid-ask spread.

Figure 1. Relation between bid-ask spread and trade size[14]