How Effective Are Trading Pauses? – Evidence From NASDAQ
University of Vienna – Department of Statistics and Operations Research
Vienna Graduate School of Finance; University of Vienna
June 25, 2016
In response to the Flash Crash of May 2010, a new control mechanism called trading pause was introduced across all U.S. stock exchanges in order to keep extraordinary volatility at bay. In this paper we investigate whether this new regulatory measure achieves its goal by looking twenty-level deep into the NASDAQ limit order book around trading pauses, as well as exploring intra-pause order book dynamics for the first time in the literature. We identify the effect of trading pauses on return volatility, market liquidity and order frequency on a high-frequency basis by making comparison to a control group of matched extreme price movement events from the pre-regulation period. Implementing a difference-in-differences framework, we find that order frequency and price discovery are enhanced, already during trading pauses, however, this beneficial effect comes at the cost of greater volatility and bid-ask spread after the resumption of trading.
How Effective Are Trading Pauses? – Evidence From NASDAQ – Introduction
The statutory objectives of financial market regulation are to maintain market confidence and stability, protect consumers and reduce financial crime (U.S. Financial Services and Markets Act, 2000). In order to achieve this orderly operation of financial markets, certain control measures are imposed on market participants and trading activity. For example, price limits and trading halts, which are collectively also referred to as circuit breakers, are designed to prevent abrupt price changes, and thus protect market participants from sudden capital losses. However, the Flash Crash in May 2010 revealed that the control mechanisms at the time had not been sufficient to avoid rapid price movements and maintain an orderly market.
Hence, in June 2010, trading pauses (also known as `volatility circuit breakers’) were introduced on U.S. stock exchanges. These are five-minute-long call auctions, which are automatically triggered for any listed stock, should its price rise or fall too rapidly. A distinctive feature of this new control mechanism is that only order execution is suspended, whereas submitting or canceling limit orders is still possible. The purpose of this temporary suspension of trading activity is to provide market participants enough time to process information and revise their open positions. This in turn should curb excessive volatility, thereby creating an additional layer of investor protection on the stock market (Mary L. Schapiro, SEC Chairman).
Market microstructure theory is undecided about the effects of circuit breakers in general. While Subrahmanyam (1994) argues that artificial constraints on price movements may have the undesirable effect of exacerbating the situation and self-fulfilling a market failure, Kyle (1988) takes the view that trading interruptions have a pacifying effect and help resolve market tensions. At the same time, empirical research has been focused on older control mechanisms, such as trading halts1 (Lee et al., 1994; Corwin and Lipson, 2000; Christie et al., 2002) and price limits (Harris, 1998; Cho et al., 2003), while the effects of trading pauses have not been fully explored. Yet, the understanding of this recently introduced control mechanism is of crucial importance for both regulators and market participants to assess its effectiveness and impact on today’s high-frequency market environment.
In this paper, we focus on this particular topic, and contribute to the literature by providing novel empirical evidence on the market effects of trading pauses. We measure changes in relative bid-ask spreads and relative limit order book depth to analyze the supply-side of liquidity, as well as changes in spot volatility and trading activity to analyze the demand-side of liquidity around trading pauses. In addition, we look into the connection of the paused stock to other stocks on the market that are not paused. Furthermore, we analyze the informativeness of the mid-quote by estimating how much of the long-run (i.e. stable) stock price is revealed during trading pauses, and thus to what extent the trading suspension facilitates information processing and price discovery.
The key to answering these questions lies in isolating the effect of the trading pause from the effect of the related price movement which triggered the trading pause in the first place. To account for this factor, we implement a difference-in-differences analysis where the post-regulation state of the market is compared to the pre-regulation (control) state of the market along the cross section of trading pauses. The control group is a collection of price movements which are matched by paused stock and trigger direction (upward or downward) for each trading pause, and which are large enough that they would have certainly triggered a trading pause, had they been in force at the time.
For this analysis we collect the exact date and time of all trading pauses on NASDAQ from their implementation in June 2010 to June 2014.We use twenty-level deep limit order book data, updated at a message-level frequency, which enables us to follow every market change (i.e. order submission, cancellation or execution) on a microsecond basis. From this high-frequency limit order book data we calculate the market measures mentioned before: spot volatility, relative bid-ask spread, relative market depth, the size and frequency of trades as well as market imbalance. This is done at regular time intervals in a one-hour window around each trading pause, creating a cross section of time series for each market measure, which we use to perform the above described difference-in-differences analysis. To address the question of price discovery, we also look into the order book during trading pauses. Since order execution during this five-minute period is suspended, the buy and sell side of the market may cross over. Hence, in order to make an equitable comparison to the around-trading-pause market possible, we bring the during-trading-pause market to the same level by matching limit orders and virtually clearing the market. Using this virtually cleared order book, we calculate market depth and price discovery statistics, and explore intra-pause order book dynamics for the first time in the literature.
See full PDF below.