Your browser is unsupported

Please visit this URL to review a list of supported browsers.

Closing Auction: Immediate market impact, price drift and transaction cost of trading

Part 1

Author
Choey Li

Quantitative Research Lead, NYSE

Published
August 22, 2023

Market volume at and near the closing auction has regained momentum recently, after the retail trading boom in 2020 and 2021 shifted some flows earlier in the day. This liquidity concentration at the end of the day creates opportunities for all market participants to trade in large size with less market impact relative to other time periods. In this analysis, we leverage NYSE’s rich auction imbalance data to better understand market impact in and around the closing auction and provide insights on relationships among order sizes, market impact, and trading costs.

Market impact and price drift

This analysis measures market impact of orders entered into the closing auction, based on order size. We limit our measurement to the 3:55-4:00 pm period, when all eligible order types are included in the auction imbalance calculation, and focus on the largest sized orders relative to the CADV of each individual stock. To assess price impact, we measure the reference price change within the next 10 imbalance updates (given roughly each second), giving the market enough time to react to newly added orders.

Each imbalance dissemination message is a snapshot of current auction liquidity, and the change in imbalance quantity represents the net value of auction eligible interest. We believe it is necessary to aggregate these snapshots together to derive a full understanding of the imbalances. Without the flexibility of joining the imbalances, the increasing buy imbalance from the previous update to the current one should drive the reference price higher, and the increasing sell imbalance should drive it lower.

Examining all immediate changes in imbalance quantities relative to CADV, we put all imbalance changes into 100 bins ranked from largest 1% to the smallest 1% and measured both the reference price (i.e., the current market price) change in next 10 imbalance dissemination messages and the price drift from the order arrival reference price to the auction price.

A median is chosen as a descriptive statistic for the price drift because we care more about how the reference price deviated from the auction for most of the stocks, to help calibrate for less-active names. For the immediate market impact measurement, we chose a weighted average to reflect the fact that even when filtering to the largest orders the immediate price impact can often be 0. Results are summarized below. For example, between 15:55-15:56, for Russell 1000 names, the largest auction orders were on average 2.4% of CADV and had an immediate price impact of 0.34X the daily average spread with 52% of imbalance updates within this minute showing non-zero price impact.

Chart 1: Imbalance Change/CADV in % - Russell 1000 Names

2023 YTD Imbalance Change / CADV in 100 Buckets - Russell 1000

Chart 2: Imbalance Change/CADV in % - Other Names

2023 YTD Imbalance Change / CADV in 100 Buckets - Rest

Chart 3: Reference Price Change vs Imbalance Change Bins
Russell 1000Others
Weighted Average Immediate Reference Price Move
(in Multiple X relative to Daily Average Spread)
0.2X-0.39X0.2X-0.34X
Price Drift to the Close
(in Multiple X relative to Daily Average Spread)
0.6X-1.7X0.5X-0.9X
Average Imbalance Change
(in %CADV)
0.01% - 2.7%0.01% - 5.3%

Thresholds in imbalance quantity change bins

The next step is to find the threshold, if any, in sizing an order that could be used as an upper bound to prevent large immediate market impact; we want to find the largest order size that will not produce undue price impact. First, we find all “inflection points” in imbalance change bins where the immediate change in reference price from the current bin jumped to a significantly higher level and remained elevated compared to rest of smaller imbalance change bins.

The process to find the inflection point is iterative. We start with the largest % of CADV bin - Bin#1 - and calculate the average price impact for the remaining smaller CADV bins in a given minute. We then compare the price impact from the current bin to the one with the multiples of one to two times the average of the remaining bins with 0.01 added to the multiple in each iteration. For example, if the first bin is greater than 1.01 times the average of remaining bins, we store it, move to the second bin, and perform the same calculation until all bins have been considered. Then, we go back to the first bin and check if it is greater than 1.02 times the average of the remaining bins etc.

If we find too many inflection points across the bins or if the first bin is smaller than the average of the remaining bins, we do not find a valid inflection point. We find that significant inflection points with a price impact of 1.08X - 1.23X the average of remaining bins successfully differentiated the level change in immediate price impact. Thus, the threshold is the CADV bin showing the most significant inflection points.

The threshold represents the smallest order size relative to CADV where the reference price is less likely to revert to its norm compared to an order sized smaller than the current bin. We use it as a proxy for how large an average order could be sized relative to the immediate imbalance quantity change before triggering large and lasting market impact.

For Russell 1000 names, the first two minutes into the imbalance period did not show a consistently elevated level in immediate reference price changes. However, in the last three minutes, we found a threshold for each minute, above which larger market impact remained elevated. For example, between 15:57-15:58, for Russell 1000 names, an order could be sized up to the top 0.47% of CADV (7th largest bin) before triggering a large lasting market impact. For other stocks, we saw high variability in the immediate changes in reference price across all bins, making it difficult to find the thresholds.

Time of day effect

As we move later in the trading day, the immediate price impact increases while price drift decreases (for a given order size). We believe the opposite direction of the two trends reflects the ability for the market to better react to information when it has more time to do so; when meaningful orders arrive later the market has a larger immediate reaction, and the auction tends to occur at a price reflecting this late impact. This is demonstrated in the below charts; we see 0.3X daily spread price impact for orders of ~2.4% CADV between 15:55 - 15:56, while the same size orders have impact of 0.4X daily spread price impact at 15:59-16:00.

Chart 4: Immediate Reference Price Change and Imbalance Quantity Change Bins – Russell 1000 names

2023 YTD Market Impact with respect to Imbalance Change / CADV in 100 Buckets - Russell 1000

Chart 5: Immediate Reference Price Change and Imbalance Quantity Change Bins – Other names

2023 YTD Market Impact with respect to Imbalance Change / CADV in 100 Buckets - Rest

Summary

The above research indicates that closing auction orders as large as 2.5% of CADV for Russell 1000 stocks results in an immediate price move of around 0.3X the daily average spread. In stocks not in the Russell 1000, we actually see less immediate price move (below 0.3X the daily average spread), suggesting the market takes more time to digest and react to information in auctions of less-active stocks. Across all stocks, impact is likely to increase when order entry gets very close to the end of trading. In the last three minutes during the imbalance period between 15:57 - 16:00, auction orders could be sized up to approximately 0.47%, 0.86% and 1.18% of CADV before triggering the large and persistent market impact.

We will soon publish a second related analysis, examining the trading cost impact of entering orders on the opposite side and the same side of an auction imbalance.

Recent articles


2023: The Year in U.S. ETPs - Part 3

February 1, 2024

2023 was an exceptional year for global equity ETPs. Until the fourth quarter, profits were limited generally to a few tech companies, but gains broadened in Q4. We review fund flows, returns and activity in multiple equity ETP categories.

2023: The Year in U.S. ETPs - Part 2

January 17, 2024

Actively managed ETPs now account for 1/3 of all U.S. ETP listings, accounting for 6.5% of AUM and 22.5% of fund flows in 2023. Futures-based Crypto ETP activity grew as the year progressed, likely in anticipation of the 11-Jan-2024 launch of cash bitcoin ETPs.

Read More

2023: The Year in U.S. ETPs - Part 1

January 11, 2024

U.S.- listed ETP AUM continued to grow, with NYSE Group markets leading the way. Equities remain the largest asset class, but fixed income products gained ground.

Read More