عملکرد نسبی توقف تجاری و محدودیت های قیمت: شواهدی از بورس اوراق بهادار اسپانیا

We study the relative performance of trading halts and price limits using data from the Spanish Stock Exchange where both mechanisms have coexisted. According to our evidence, trading activity increases after either mechanism is triggered. Volatility stays the same after trading halts but increases after price limit hits. Our evidence also shows that the bid–ask spread is narrower after trading halts but wider after price limit hits. Information is efficiently reflected in stock prices once trading resumes after trading halts, but there is evidence of market overreaction for upper price limits. Our overall result may have important policy implications for financial markets in the world.

In semi-strong efficient markets, asset prices reflect all publicly available information, and prices change only in response to relevant new information (Fama, 1970). Therefore, any artificial interruption imposed on the market should have little impact on price movements. However, organized exchanges generally have special rules or procedures that come into play in connection with events that result in, or are likely to result in, large changes in asset prices. Following the 1987 market crash, the level of interest in procedures to limit large or sudden changes in prices has increased. The Brady Report (1988) suggests that circuit breaker mechanisms, such as trading halts and price limits, should be imposed to protect the market system. It appears that stock exchanges in the United States prefer trading halts to price limits.1 The NYSE has imposed both market-wide trading halts and individual news or order-imbalance trading halts. News-related trading halts also exist at the NASDAQ. Unlike those stock exchanges, the U.S. futures markets seem to favor price limits. Many countries in Europe and Asia also impose price limits on their stock markets.2 Although trading halts and price limits are both circuit breakers, they differ in several ways.3 First, by definition, trading halts represent a temporary interruption in the trading of an individual asset on an exchange to disseminate information, whereas price limits are boundaries set by market regulators to confine the daily movements of security prices within a predetermined range to mitigate excessive price volatility. Therefore, trading halts indicate a complete cessation of trading activity, whereas in the case of price limits, trading is still permissible as long as it remains within the preset trading range. Second, trading halts do not include limitations on price movements. Once trading resumes after a trading halt, the price is determined solely by the market. Third, trading halts are not mechanically or predictably imposed but rather are subjectively imposed in certain circumstances by exchange officials or supervising authorities. That is, trading halts are called at the discretion of officials. In contrast, the activation of price limits depends solely on the price movements. Price limits are therefore easier for investors to observe and predict than are trading halts. Despite the differences between trading halts and price limits, we believe that they could be viewed as trying to achieve either directly or indirectly the same objective, which is to reduce information asymmetry. Problems associated with asymmetric information include excessive price volatility (Spiegel & Subrahmanyam, 2000), unwarranted trading uncertainty (Greenwald & Stein, 1991), and transactional risk (Kodres & O'Brien, 1994). The activation of circuit breakers attempts to provide investors with more time to evaluate new information and make rational decisions. In the case of trading halts, trading is suspended so investors are forced to cool off and obtain and digest new information. Although price limits allow investors to continue trading at the limit price, they also give investors the option of choosing not to trade. In the latter case, the effect is similar to trading halts because during the cooling-off period, investors can reevaluate the market information. On the basis of this cooling-off argument, regulators expect that trading halts and price limits cause stock prices to become more informative, reduce uncertainty, and protect uninformed investors from excessive price movements. In previous literature, trading halts and price limits have been either treated equally (e.g., Kyle, 1988) or studied separately (e.g., see Christie et al., 2002, Corwin and Lipson, 2000, Edelen and Gervais, 2003, Lee et al., 1994 and Schwartz, 1982 for trading halts and Bildik and Gülay, 2006, Brennan, 1986, Kim and Rhee, 1997 and Kodres and O'Brien, 1994 for price limits).4 Although Telser (1981) informally argues that rule-based price limits are superior to discretionary trading halts because they are more predictable, no theoretical model has been developed, nor any empirical test performed, to support that argument. To the best of our knowledge, only two studies formally compare trading halts with price limits. Subrahmanyam (1995) theoretically analyzes the relative desirability of discretionary and rule-based procedures and argues that discretionary closures enable exchange regulators to consider more information (e.g., market liquidity, volatility) in the closure decision than just the size of the price movement, which makes them more effective than price-triggered closures. Therefore, according to Subrahmanyam (1995), discretionary trading halts should be more effective than rule-based price limits. Coursey and Dyl (1990) conduct an experimental study to compare the market's adjustment to significant new information in the presence of price limits or trading halts. Their findings seem to indicate that the adjustment of asset prices to new information is more effective in markets with price limits than in those with trading halts, in contrast with Subrahmanyam's (1995) argument. We attempt, for the first time to our knowledge, to provide empirical evidence for the relative performance of trading halts and price limits. The Continuous Spanish Stock Market, or SIBE,5 provides a natural setting to study the performance of trading halts and price limits because both mechanisms had been used in this market prior to May 2001. The supervisory committee of the SIBE has the authority to halt trading on any individual stock, similar to NASDAQ trading halts, in that they are mainly imposed to force information disclosure to eliminate asymmetric information or await a pending announcement by a listed firm. During our study period from January 1998 to April 2001, the SIBE also set a daily maximum price fluctuation of 15% from the previous day's closing price. Although reasons for hitting price limits are not officially determined, they often relate to news or announcements about the listed firms. Therefore, both discretionary trading halts and rule-based price limits are associated with essentially the same market conditions and have the same objectives: to eliminate unwarranted asymmetric information and uncertainty about an individual stock value. In the spirit of Subrahmanyam (1995), we hypothesize that trading halts are more effective than price limits because exchange officials incorporate relevant information into the trading halt decision and can ask companies to provide relevant information. For the purpose of this paper, we call this the “discretion hypothesis”. We investigate the pattern of trading activity, liquidity, and volatility, as well as the speed of price discovery, in the period surrounding trading halts and price limit hits and then compare the performance of these two circuit-breaking mechanisms. Because trade-to-trade movements are essential for conducting microstructure studies, we examine transaction data from the SIBE. Our results show that trading activity increases after trading halts and price limit hits. Volatility stays at the same level after trading halts but increases after price limit hits. We also show that the bid–ask spread is narrower after trading halts but wider after price limit hits. Our spread decomposition analysis suggests that the change in spreads is likely attributed to the changes in the fixed component of trading costs. For price discovery, information is efficiently reflected in stock prices when trading resumes after trading halts, but market overreaction may occur after upper price limit hits. Overall, the results support our discretion hypothesis. Although our results are based on a market where both trading halts and price limits are present, we believe that our results are relevant to markets with either trading halts or price limits. First, in an effort to remove any potential contamination to our sample, we exclude observations whose price limits are reached and trading is halted concurrently. Second, our daily analysis of market quality shows that the individual results of trading halts and price limits are consistent with the findings from the existing trading halt and price limit literature. That is, the coexistence of trading halts and price limits does not seem to generate different behaviors of market participants than what have been observed in other markets. Third, the fact that firms with trading halts are larger on average than those with price limits is consistent with Kim and Limpaphayom's (2000) finding that small market capitalization stocks hit price limits more often and Bhattacharya and Spiegel's (1998) finding that larger capitalization stocks are suspended more often on the NYSE. Thus, the coexistence of trading halts and price limits does not seem to alter firm characteristics of trading halt and price limit firms. However, the interpretation of our results needs to be cautious. The support of the discretion hypothesis seems to indicate that trading halts are more desirable than price limits, but we have no intention, nor are we able, to make that argument because of the following reasons. First, trading halts and price limits are two different mechanisms with possibly different immediate objectives. Exchanges may use trading halts to disseminate new information while using price limits to restrict daily price movements. In this sense, both mechanisms have achieved their goals. Second, the occurrence of a trading halt or the hitting of a price limit is an endogenous event that itself might have caused our results. Although the reasons for trading halts are available, we do not always know the cause of a price limit hit and thus unable to adjust for the potential endogeneity problem. Lastly, with price limits already imposed, the officials in the Spanish market could use trading halts more efficiently at their discretion. For example, they can treat price limits as a safety net and halt trading only if price limits are not useful or trading halts will resolve issues quickly. This opportunity does not exist in markets with only trading halts. Nevertheless, our results represent the first empirical evidence for the relative performance of trading halts and price limits. The remainder of the paper is organized as follows: the next section describes the institutional background of the SIBE. Section 3 outlines the hypothesis and presents our research methodology, followed by data selection and sample descriptions in Section 4. Section 5 provides empirical results of various tests of trading halts and price limits, as well as comparisons of them, and Section 6 provides a conclusion.

The performance of trading halts and price limits has been studied extensively and separately, particularly after the 1987 market crash. However, the relative performance of trading halts and price limits has not been examined using market data. In this paper, we study these two mechanisms using data from the Spanish stock market, where both trading halts and price limits have been imposed. We make no attempt to test the effectiveness of either trading halts or price limits but focus instead on their relative performance. Using Subrahmanyam's (1995) findings, we hypothesize that trading halts may be more effective than price limits because exchange officials can incorporate related information into their trading halt decisions and ask companies to provide relevant information. Specifically, we investigate the pattern of trading activity, liquidity, and volatility, as well as the speed of price discovery in the periods surrounding trading halts and price limit hits. Our results show that (1) trading activity increases after trading halts and price limit hits; (2) liquidity increases after trading halts but decreases after price limit hits; (3) volatility stays at the same level after trading halts but increases after price limit hits; and (4) information is efficiently reflected in stock prices when trading resumes after trading halts, but market overreaction may occur for upper price limits. Overall, our results are consistent with Subrahmanyam (1995). Although our results may have important policy implications, three potential issues do remain. First, in Spain, firms are required to release information during trading halts but not price limit hits. Therefore, a question may arise: “Would price limits perform equally well if they require information release?” (or “Would trading halts perform equally badly if no information is released?”) We are not addressing this important but hypothetical question in this paper. Second, this study is subject to a form of “joint hypotheses” problem. That is, we made an implicit assumption that current trading halt and price limit mechanisms are optimally imposed. However, trading halts have performed better than price limits perhaps because the 15% price limit is sub-optimal. Third, the Spanish market is completely order-driven and has no market makers, so we do not know if trading halts are still preferable in markets where market makers exist. We leave these three issues to future research.