Trading into Hong Kong Close
In October 2012, the Beijing-based China Asset Management Company (China AMC) and the Guangzhou-based E-fund launched the first two cross-border ETFs that have Hong Kong stocks as underlying instruments but are listed on Shanghai and Shen Zhen stock exchanges. The China AMC Hang Seng Index ETF is currently being traded under the ticker of 159920 on Shen Zhen Stock Exchange. The E-fund Hang Seng China Enterprises Index ETF is currently being traded under the ticker of 510900 on Shanghai Stock Exchange. As the ETFs’ net asset values (NAVs) are marked to market using the Hong Kong Stock Exchange (HKEx) official closing price for the brokers who make market for these two ETFs, capturing the close prices of the underlying stocks is a major task..
These two ETFs are just one example of many reasons why Market on Close (MOC) strategies are often popular among brokers and dealers who trade into Hong Kong. Other than demand from buy-side firms such as China AMC and E-fund for close-end fund mark-to-market valuation and index rebalancing transactions, the unique nature of how HKEx determines stock closing price is another reason. Comparing to other major exchanges in the world, the closing mechanism of HKEx has gone through several changes in recent years. The exchange introduced a closing mechanism in May of 2008 that was modelled after similar mechanisms of other markets such as those of the London Stock Exchange and the Korea Stock Exchange. However, some extreme volatilities of several large cap stocks near close on several trading days after May of 2008 forced HKEx to suspend the closing auction mechanism in March of 2009 and roll back to the “median of 5 snapshots” closing price determination mechanism.
The “median of 5 snapshots” closing price determination takes the median of the last 5 nominal prices in the last minute of the continuous trading session. The system will take 5 snapshots on the nominal prices at 15-second interval starting from 3:59:00 p.m. Under normal circumstances and in particular for relatively liquid stocks, the nominal price is the last recorded trade price as long as this trade price is within bid-ask spread. For other cases, the nominal price has to be determined by comparing the current bid price, the current ask price and the last recorded price in accordance with Rule 101 of the Rules of the Exchange.
This unique rule for closing price determination, when combined with technology aspects such as the throttle (1) arrangement for message submissions and influences from other markets such as China close and Europe open near 4pm of Hong Kong local time, leads to specific market microstructure characteristics near Hong Kong close. As a result, the design and development of an effective MOC strategy require quantitative models that can capture such microstructure details and specific technology treatment that utilises throttles efficiently.
Key market microstructure features of Hong Kong close
The “median of 5 snapshots” of Hong Kong closing price determination has led to concentration of liquidity around each of the 5 15-second snapshots, which are 3:59:00, 3:59:15, 3:59:30, 3:59:45 and 4:00:00. Figure 1 shows a typical intra-minute volume profile of a Hong Kong stock. In general, volume in the last minute spikes around the 5 15-second snapshots. Because of the throttle limit that many market participants face during the last minute trading, some orders submitted exactly at the 15-second snapshots may not get filled immediately. Instead, they may be filled in the second right after the snapshot. This is why people often see relatively large volume at 3:59:01, 3:59:16, 3:59:31 and 3:59:46. For relatively liquid stocks, volume profile can exhibit minor spikes in between two consecutive 15-second snapshots, as people often combine these “in-between” executions with 15-second volume spikes to ensure execution certainty, especially for relatively large orders.
For most of the cases, intra-minute volumes tend to concentrate more on the first 3 15-second snapshots than the last 2. This is mostly due to the fact that people tend to ’front-load‘ their executions at the beginning of the last minute to ensure high fill rate and avoid under-fill risks. As a result of this, liquidity starts to dry up after 3:59:45, leading to wider bid-ask spread (see Figure 1), which further reduces people’s willingness to trade in the last 15 seconds due to elevated transaction cost.
Figure 1: A typical intra-minute volume profile and bid-ask spread change of a Hong Kong stock in the last one minute of continuous trading
The anatomy of a MOC algorithm for Hong Kong market
a. Key strategy set-up requirements: A client asked to buy 8,000 shares of 1299 HK (AIA Group) that targets the official close price on May 16, 2012. The 8,000 shares order was less than 1% of average daily volume in 1 minute interval. The round lot size of 1299 HK was 200 shares. Given the relatively small order size, the trader adjusted the MOC strategy using two key parameters: first, the trader allowed the strategy to place the order passively in the market 3 minutes before market close; second, the trader required that the fill rate must be 100%; in order to achieve that, the trader allowed the strategy to cross the spread whenever necessary while ignoring the potential impact on the market, which was likely to be small. Based on these parameter set-up, the MOC strategy decided to divide the 8,000 share parent order to two groups: the “passive” group that consisted of just one order that was 60% of the total parent order, and the “aggressive” group that consisted of the remaining 40% of the total parent order, which was further sliced into three smaller orders of 1,600 shares, 1,200 shares and 400 shares in size, respectively. The three orders in the aggressive group would target the 3:59:15 pm, 3:59:30 pm and 3:59:45 pm snapshots, respectively.
b. Step-by-step action of the MOC strategy:
[3:56:30 – 3:57:30 pm] The MOC strategy started to examine the price trend at 3:56:30; the sampling of past trade/quote date was finished at 3:57:30, and an analysis based on market data was done as well. The strategy’s short-term alpha signal model indicated that price was under downward pressure and might tick down in the next 1-2 minutes;
[3:57:45 pm] The strategy sliced 60% of the order (in total 4,800 shares) and placed a limit order at the current bid (26.35 HKD), trading passively at first as the short-term alpha indicator showed that price might move downwards;
[3:58:07 pm] 4,200 shares of the passive order was hit on the bid at 26.35 HKD, with 600 shares remaining at the top of the queue; the strategy decided to stay put as the downward price pressure was still present;
[3:58:08] 200 more shares was hit on the bid at 26.35 HKD; the strategy decided to keep waiting for another hit;
[3:58:10] the remaining 400 shares of the first 4,800 share child order were filled; as 60% of the order had been filled, the strategy decided to stay out of the market until the last minute trading commenced; behind this decision was that the short-term alpha signal indicated that the price might tick downwards eventually;
[3:58:19] both bid and offer of the stock ticked down; still significantly ahead of schedule, the strategy calculated the potential duration of current tick and was not convinced that the current bid at 26.30 HKD might last long;
[3:58:39] both bid and offer ticked up; the strategy was waiting to enter the last one minute trading to capture official close price;
[3:58:52] the strategy sliced a child order of 1,600 shares as the first child order in the “aggressive” group; however, the MOC strategy instructed to place this order “passively” on the bid and wait to be hit instead of crossing the spread immediately;
[3:59:00] the final minute of continuous trading started;
[3:59:02] to capture the first snapshot, the strategy cancelled the 1,600 share order waiting on the bid and sent it across the spread to be filled at 26.40 HKD;
[3:59:07] it was 8 seconds before the second snapshot, and the strategy sliced another child order of 1,200 shares as the second child order in the “aggressive” group; again, the MOC strategy instructed to place this order “passively” on the bid (now at 26.35 HKD) to wait to be hit instead of crossing the spread immediately; at this moment, the strategy found that the “queue size” on the bid was rather long;
[3:59:18] to capture the second snapshot, the strategy cancelled the 1,200 share order waiting on the bid and sent it across the spread to be filled at 26.45 HKD;
[3:59:22] to prepare itself to capture the third snapshot at 3:59:30, the MOC strategy decided to send the remaining 400 shares to the market; this is the third slice of the “aggressive” group; again, the strategy instructed to place the order on the bid side (at 26.40 HKD as of now) so as to be passive;
[3:59:45] the third snapshot had passed and the fourth snapshot arrived; as the price ticked up significantly for the final child order slice, the MOC strategy decided to get aggressive to ensure 100% fill rate;
[3:59:47] the strategy sent a one lot order of 200 shares first to cross the spread and got filled at 26.45 HKD; the strategy put the remaining 200 shares into reserve and hoped that the price may come down;
[3:59:50] 3 seconds later, the strategy decided that the risk of under fill was so high at that moment; it immediately crossed the spread and got the remaining 200 shares filled at 26.45 HKD; as of now, all of the 8,000 shares were filled. So far, the execution VWAP of the MOC strategy reached 26.38 HKD;
[4:00:00] the continuous trading finished, and the official close price of 1299 HK on that day was announced by the exchange to be 26.40 HKD. As a result, the MOC strategy beat the official close price by 7.5758 bps.
1 “Throttle” refers to the standard order input capacity within which orders can be submitted to the trading host system through the Open Gateway (OG) within a specific period of time. For most of the cases, “one throttle rate” refers to one order submission per second through one OG. An individual OG can accommodate more than 1 throttle rate but the number of throttle rate needs to be in integral multiples of one order per second.