Market Lens #4
We further explore Market Impact in this Market Lens mini-series by looking at the extremely volatile market condition in March last year. In the previous edition ‘Understanding Market Impact’, we introduced the Market Impact framework, and set out expected patterns of Market Impact in correlated stocks. Does the same market impact pattern apply during the volatile period?
The Question: How does investor behaviour change in volatile markets, and how can the BMLL Market Impact Framework provide insight into these changes?
The Answer: During periods of increased volatility, such as during the Covid sell-off of March 2020, traders reduced posted volumes by 90% despite trading twice as much as they did at the beginning of 2020. In the meantime average Market Impact increased significantly to 0.91bps compared to 0.19bps in post-pandemic markets. In steady market conditions, traders tend to react a third more vigorously to sell trades than buy trades. However, during the highly volatile period they responded in equal measure to both sides during the heightened market movements.
The Context: The plots below highlight this divergent behaviour. The first plot shows the asymmetric market impact, in May this year, between buy and sell trades on the CME’s E-mini and the NYSE listing of the SPY S&P500 ETF. The average market impact of buy orders is nearly a third of the size of corresponding sell orders. Note how sell orders have a larger impact (dotted blue and red lines) and the secondary “reverse impact” of the SPY trade on the CME market (the small bump at 5.5ms after the SPY).
Exhibit 1: May 2021, Asymmetric Market Impact
Key:
- ES: E-mini futures contract on S&P500 trading on the CME Exchange
- SPY: SPDR ETF that indexes the S&P500 and trades fungibly across US equity venues. The graph uses the data for the New York Stock Exchange listing.
The second plot, derived during the earlier period of March 2020 at the height of the pandemic threat, highlights a symmetrical market impact, i.e., traders are reacting in equal measure to both buy and sell orders, albeit at much higher levels, 0.91bps vs. 0.19bps. This behaviour characterises a market in free fall where traders are equally reactive to buyers covering shorts or pre-emptying a bounce as to sellers increasing shorts or cutting losses. In such markets, participants are looking for any positive signs of mean reversion, therefore have become more reactive to such trades.
Exhibit 2: March 2020, the first Covid-instigated sell off
Exhibit 2 indicates no asymmetry between buy & sell orders at this time reflecting the unprecedented pressure in the market due to the threat of the pandemic. Market participants care as much about buying as they do selling.
The So What: Understanding Market Impact is a central component of tuning smart order routing algorithms and optimising execution more broadly. Understanding Market Impact is a contributing factor to gaining competitive edge, and failing to understand impact asymmetry places a limit on how effectively execution algorithms can be tuned. In volatile markets with Market Impact responses nearly 5 times larger, this is especially true. Market participants who understand state changes that occur around volatility can leverage this added insight to optimise their performance.
Footnote: How Did We Do This? The BMLL Data Lab replays every order book message to get a true picture of market impact. For an event, BMLL computes metrics in each considered order book at the event time and offsets around it, considering each order book message processed between them. By scaling the compute up over the 18 months of the Covid crisis, we are able to see the evolution of the true impact of the considered events on the metric. Here, we considered how the 10,000 largest daily executions move the midpoint price. Note, these are the executions as reported in the public feed and correspond many-to-one to a large aggressive. We use the largest executions as smaller executions carry additional noise. This analysis utilised over 2 billion order book messages to determine these curves with maximum reliability. The BMLL Data Lab has multi-asset class nanosecond granularity in a scalable research environment as well as predefined market impact frameworks, giving quants the ultimate suite of tools to harness market impact and make more informed trading decisions.
More Market Lens articles
Insight #3 : Understanding Market Impact - can understanding correlated Market Impact improve execution?
Insight #5 : BMLL Market Lens. The View From The Feed