This paper addresses the issue of the optimal trading system for less actively traded (i.e., ‘thinly-traded’) stocks. We compare the performance of a pure order driven market with limit order book (POD) with that of a hybrid order driven market with specialist and limit order book (HOD). We find that the HOD system offers superior performance along several dimensions of market quality. In particular, the specialist-based system offers lower execution costs, greater depth, a significant increase in the depth-to-spread ratio, and lower adverse selection costs. Very ‘thinly-traded’ stocks benefit more than less inactive stocks from the adoption of a hybrid trading system both in terms of greater liquidity and in terms of lower adverse selection costs.
The structure and design of exchanges is an important issue that has received considerable attention in market microstructure studies.1 The research in this field has mainly focused on the implications of alternative market structures for metrics of market quality, such as explicit and implicit trading costs, liquidity, price discovery, and informational efficiency. One of the most important developments in financial markets over the past decade is the proliferation of new markets and automated trading systems. Many of these automated markets use an order driven system, with a high degree of transparency where current and away public limit orders are continuously displayed. Even though there has been a tremendous growth in order driven systems, there is little empirical research into the relative performance of order driven and quote driven systems, especially for less frequently traded (i.e., ‘thinly-traded’) stocks.
Recent reorganization of the Italian Stock Exchange (ISE) offers a unique institutional setting to study the relative performance of pure and hybrid order driven systems for trading illiquid stocks. In May 1997, the ISE implemented a plan, named Thin Stocks Plan, to improve market quality of ‘thinly-traded’ stocks. 2 Under this plan, stocks defined as ‘thinly-traded’ by ISE were given the choice to trade under one of two alternative regimes: a pure order driven system with a limit order book (POD), or a hybrid order driven system with a specialist and a limit order book (HOD). In this paper we examine the quality of the two market structures for trading illiquid stocks and test two hypotheses, the first related to trading costs and the second to the adverse selection component of the spread.
Examining the effects of trading architecture on market quality for less actively traded stocks is important for several reasons. First, several empirical studies document a statistically significant relation between thinness and poor market quality. ‘Thinly-traded’ stocks are usually associated with poor market quality indicators – such as large spreads, high transitory volatility, and low price efficiency (Pagano, 1997), and high adverse selection costs (Easley et al., 1996). Second, negative effects associated with “thinness” may not only be detrimental for the market behavior of firms that are already listed, but they may also deter new companies from listing for fear of falling into a vicious cycle of low trading activity leading to low attractiveness for investors which in turn lead to even lower trading activity.3 Thus, improving the market quality for these firms could also help the listing of new companies. The third motive for this study is the widespread presence of ‘thinly-traded’ stocks on the main stock exchanges. As documented by Easley et al. (1996), on the London Stock Exchange the bottom 50% of stocks by capitalization account for only 1.2% of the overall market trading volume, and over 1000 stocks average less than one trade a day. When considering all the major European exchanges, the evidence is similar: stocks with a turnover ratio lower than 25% of the average market turnover account for only 1.4% of the overall trading volume and for almost 26% of the total number of listed stocks (Garcia Coto, 1996).
There are several important differences between previous empirical studies of market quality and the present study. First, previous empirical studies that have compared the quality of alternative market structures have generally focused on actively traded stocks (Petersen and Fialkowski, 1994; Huang and Stoll, 1996; Keim and Madhavan, 1997), whereas in this study we focus on ‘thinly-traded’ stocks. Second, the present study compares different trading structures for companies listed on the same market. This feature reduces structural differences across stocks in the sample, and highlights differential effects induced by different trading mechanisms. A number of studies have a much more limited ability to control for differences in structural characteristics of the underlying securities. For example, existing studies that compare execution costs on NASDAQ and NYSE (e.g., Huang and Stoll, 1996; Bessembinder and Kaufman, 1997) may suffer from differences in the listing requirements and risk characteristics of the securities under examination.4 Third, the data sample differs from existing studies. This study examines the ISE, an electronic screen-based market whose dimension and importance are constantly increasing in the context of organized exchanges. According to the Federation of European Stock Exchange statistics (FESE, 2002), as of September 2002, ISE is the fifth (fourth) largest exchange in Europe and the third (third) largest exchange in the Euro area in terms of total market capitalization (monetary trading volume). Finally, this study provides specific insights on the operation of order driven markets. Interest in limit order trading has grown rapidly in recent years as it plays a vital role in the liquidity provision for both pure (e.g., Paris, Tokyo, Stockholm) and hybrid (e.g., NYSE) order driven structures. In this paper we examine a comprehensive set of quality measures, and focus on specific features displayed by spread behavior in an electronic order driven system. In particular, the difference between quoted and effective spread, that has already been studied in quote driven environments (Petersen and Fialkowski, 1994), has a totally different interpretation in an order driven system.
The remainder of the paper is organized as follows. In Section 2 we discuss theoretical and empirical studies related to the comparative analysis of alternative trading systems. In Section 3 we describe the microstructure of the ISE, and summarize the reforms adopted by ISE to improve the quality of the market for less actively traded stocks. In Section 4 we describe the hypotheses to be tested and the research design. In Section 5 we describe data and sample selection criteria. In Section 6 we present quality measures and methodologies used to investigate the relative quality of the two trading systems. In Section 7 we present empirical results, and conclusions are provided in Section 8.
In this paper we examine the impact of specialist intervention for trading illiquid
(i.e., thinly-traded) stocks. Empirical evidence from the ISE shows that a hybrid
order driven system – with a specialist and a limit order book providing liquidity –
offers better market quality than a pure order driven structure. In particular, we find
significant improvements under the hybrid system compared to a pure order driven
system across several quality metrics – bid–ask spread, depth, depth-to-spread ratio,
and adverse selection costs. Very thinly-traded stocks benefit more than less inactive
stocks from the adoption of a hybrid trading system with specialist intervention both
in terms of greater liquidity and in terms of lower adverse selection costs. Partitioning
the sample by trading volume quintiles we also find that liquidity improvements
are most important for the central quintile.
Results are robust to different variables specification (quoted and effective
spreads), and different comparative approaches (simple event study and combined
approaches). Our results are also consistent with theoretical analysis (Grossman
and Miller, 1988; Glosten, 1989; Benveniste et al., 1992) and related empirical evidence
(Christie and Huang, 1994; Madhavan and Sofianos, 1998; Kavajecz, 1999;
Chung et al., 1999; Heidle and Huang, 2002). Christie and Huang (1994) study firms
moving from NASDAQ to NYSE and find that less liquid stocks prior to listing on
the Exchange realize the most liquidity gains from listing in an auction market.
Madhavan and Sofianos (1998) analyze NYSE specialists market making activity
and provide evidence that specialists services are most valued for thinly-traded
stocks. Kavajecz (1999) compares the spread on a limit order book with the specialists
quoted spread; he argues that specialists play an important role in narrowing the
spread, especially for smaller and less frequently traded stocks. Chung et al. (1999)
examine the interaction between limit orders, specialist, and bid–ask spread; they
find that specialists provide greater liquidity to low volume stocks. Heidle and
Huang (2002) study a sample of firms switching exchange listing, and analyze the
probability of information-based trading across different exchanges; our findings
are consistent with their in terms of lower adverse selection when moving to a
specialist-based market.
Our results are also consistent with the behavior of the ISE. In fact, Italian market
authorities have progressively reduced the number of stocks included in the Inactive
Stocks segment of ISE that are eligible for the hybrid order driven system. Hence,
less thinly-traded stocks have been progressively excluded from the Inactive Stocks
segment, and from the possibility to opt for the hybrid order driven system. 27 This behavior is consistent with our findings that specialist intervention is more valuable
for very thinly-traded stocks.
On April 2001 the ISE launched a new market segment, named STAR, in substitution
of the Inactive Stocks segment. The STAR market segment is built upon the
same principles as those of the Thin Stocks Plan; in fact, for each stock that chooses
to be listed on the STAR segment, specialist intervention is required. 28 Moreover,
twice a year, the ISE issues a list of stocks to be traded under a double (opening
and closing) auction system. Those are typically very low trading volume stocks,
and this is also consistent with our results on liquidity improvements partitioned
by trading activity quintiles.
However, our research design suffers from a potential source of bias. The decision
to switch to the hybrid system is a voluntarily, endogenously-determined choice, reflecting
some underlying optimization process. Firms that choose to switch do so
because they expect to benefit from the new trading system. Thus, the switching
event reflects an inherent self-selection: we only observe those firms for which the decision
to opt for the hybrid system was expected to be beneficial. This suggests that
our sample could be biased in favor of finding better performance for the specialistbased
system. This might be the case, and in fact not all the firms eligible to switch
did so. However, several other factors also affect this decision. Given the endogeneity
of the switching choice, our paper thus suggests that a hybrid order driven system
with a specialist represents an optimal trading systems for some thinly-traded
stocks, and not for all thinly-traded stocks.
Why does a firm choose the hybrid system? Benefits for the firm from the switching
can arise in terms of greater stock liquidity and positive signals to the market,
both factor especially valuable for firms planning to raise additional capital on the
market. On the other hand, there are both tangible and intangible costs attached
to the switching decision. Thus, the choice to opt for the hybrid order driven system
depends on several firms-specific characteristics. Exploring this issue is left for future
research.
Acknowledgements
