LiquidMetrix to Short Articles

07 Dec 2009

LM001 - What was the impact of the LSE outage on Thurs 26th Nov 2009?

Welcome to the LiquidMetrix short article series! Over the coming months, we will be publishing a series of short articles drilling down into more detail into our data sets and highlighting some of our findings. We are going to start with a recent and topical subject, namely what happened to MTF activity during the LSE outage on November 26th.

Background

First a little background. The LSE identified connectivity issues with some clients around 9:30 in the morning of the 26th. This ultimately led to the LSE placing all order books into auction mode at 10:33. The order books stayed in this state until 14:00 when trading was re-started with a standard uncrossing. (LSE also issued an update at 13:25 informing the market that trading will resume at 14:00). If you want to see a blow by blow description of the events and timeline, then the LSE incident report is at http://www.londonstockexchange.com/global/incident/26november2009-incidentcommunicationsoverview.htm. This report is helpful as it was what was disseminated to the market during the day and makes clear at what points in time the ‘market’ was aware of what was happening and what was going to happen next.

As this is the first article, we’ll briefly describe the data we have and the analysis we have performed. We have all on book trades and order by order events (entry/modification/deletion of orders) in the visible order books tick by tick. We use this data to replay order books over the day and then measure statistics like spreads, depths, best prices etc. To illustrate this, the two screenshots below show the re-built CHIX order books for Lloyds at 10:10:00.000 and 11:30:00.000 on the morning of November 26th.


CHIX order book at 10:10:00.000 (spread 6.6BPS)
(click the image to enlarge)


CHIX order book at 11:30:00.000 (spread 34.2BPS)
(click the image to enlarge)


Using these high frequency order book replays we can measure exactly what happens second by second during the trading day. The statistics we are showing in this article are based on 1 second samplings of order books that are then aggregated into 30 second intervals for graphing purposes. We are going to show results for a liquid and fairly competitive stock (Vodafone). We think most of the comments we make apply fairly generally to most of the more liquid LSE stocks (where there is significant competition with the MTFs).

Analysis

1. Order Book Messages

We’ll start with the basics, the graph below shows the number of order book messages (order entries, modification deletes) occurring on each of the competitive venues (LSE, CHIX, BATS, TRQX, NURO) during the day of the outage. The y-axis shows messages per 30 second period.


(click the image to enlarge)
The boundaries of the outage (10:33 to 14:00) are seen quite clearly. Order messages activity on all the MTFs is very quiet during this period although CHIX shows occasional spikes of higher activity.

2. Order Book Depths

We measure order book depth as being the total value (i.e. price * volume) of all bids/offers in the public visible order books within x basis points of mid price. The graphs below show this statistics for 25BPS (near ‘top’ of book volume). We have excluded LSE as the books were in auction i.e. volumes could not be ‘hit’.


(click the image to enlarge)
Points of interest:

  • Immediately following the LSE stocks going into auction at 10:33 the amount of liquidity in the MTF order books all dropped significantly.
  • Between around 11:30 to 13:30 CHIX and to a lesser extent BATS managed to recover some liquidity. For Vodafone there was approximately 500,000 EUR of visible orders within 25BPS of mid prices.
  • When LSE announced at 13:25 that trading would resume at 14:00, there was another collapse in book depths on the MTFs up until the point at which LSE started continuous trading again.

3. Spreads

The graph below shows the ‘At Touch’ spreads for the MTFs during the outage (again LSE is excluded as the spread during an auction period is meaningless).


(click the image to enlarge)
Spreads clearly became volatile on Neuro and Turquoise; however CHIX and BATS maintained more reasonable spreads over the outage period. The graph below shows CHIX and BATS spreads on their own.

(click the image to enlarge)
Although spreads on CHIX and BATS are clearly larger and more volatile than during normal market conditions, they are not ‘unreasonable’. In particular on CHIX there are periods of time when the spread is pretty much normal for this stock, although we know from the liquidity figures that the volumes available are smaller than normal.

As in the case for depths, there is a clear deterioration in spreads on both CHIX and BATS leading up to LSE restarting trading at 14:00.

4. Trading Volume

The graph below shows the value traded over the outage. Clearly the volume traded is extremely low compared to ‘normal’ on all MTF venues. This is despite there being some liquidity and reasonable spreads on CHIX and BATS.


(click the image to enlarge)

EDSP Auctions

As a side note there is a regular mini ‘outage’ similar to November 26th that occurs on the 3rd Friday of each Trading Month. On that day between 10:10 and 10:15, LSE puts stocks that have option/future expiries into an auction state for around 5 minutes. The graphs below show the effect of the EDSP auction on spreads, market depths and value traded for Barclays on the MTFs for Friday 20th November zoomed in between 10:00 -> 11:00.

1 Contract MTF Spreads during EDSP auction at LSE 25BPS

MTF Liquidity during EDSP auction

MTF value traded during EDSP auction
(click the images to enlarge)

Summary:
  • Spreads on all the MTFs widen markedly at 10:10 at the beginning of the LSE EDSP auction, although CHIX maintains a reasonable spread.
  • Depths on BATS, Turquoise and Neuro drop to virtually zero. Again CHIX maintains a degree of liquidity, albeit about a third of normal liquidity.
  • During the 5 minute LSE EDSP auction there is virtually no value traded on all MTFs.

Obviously this is a very different type of ‘outage’ event in that it is known in advance and by definition prices can be volatile and volumes large due to the fact that there are derivative expiries happening. However, it’s interesting to see the basic patterns (CHIX maintain spreads and some liquidity but has minimal or no trading activity) are similar to what happened on November 26.

Conclusions

Following the previous more minor LSE outage on the afternoon of 9th November that affected a small number of stocks, BATS published a short article http://www.batstrading.co.uk/resources/participant_resources/LSE_outage_whitepaper.pdf noting that significant liquidity and matched trading volumes continued on the MTFs (as a whole) during the outage.

However, on the prolonged 26 November LSE outage, it’s clear from our analysis that matched trade volumes during the outage were a small fraction when compared with ‘normal’ volumes on the MTFs. This is in spite of the fact that some of the MTFs seem to have at least some liquidity providers willing to quote reasonable spreads during the outage albeit with smaller volumes and less ‘depth’ in the visible books as would be usual. This implies that although prices and some liqiudity was available on the MTF venues during the outage, aggressive orders were not being routed to hit these available prices. Some think this is due to the LSE being in ‘Auction Mode’ rather than suspension, auction mode meaning that many Smart Order Routers would not route orders away from the primary (or at all). BATS provided an update in their article to largely express this opinion.

The behaviour of the markets during regular EDSP auctions seems to have given a good forewarning of what would happen on November 26th. It will be interesting to see if any trading begins to occur on MTFs during these EDSP periods in the coming months. If SORs were to be adapted to route orders even if a primary venue is in auction mode this might start to happen.

Further Thoughts…

Something we haven’t looked at in detail here but may come back to is to what extent ‘price formation’ was continuing on the MTFs. In other words were prices really ‘moving’ in the usual way during the outage or were the market is a kind of limbo awaiting the re-opening of the primary market.

If you have comments / feedback on this article please email liquidmetrix@if5.com.


The above analysis was done using a LiquidMetrix WorkStation. Please click here to find out more.

Disclaimer

The information contained within this website is provided for information purposes only. IFS will use reasonable care to ensure the accuracy of the information within this site. However, IFS will not be held liable for any errors in the information provided within this website or for accuracy or completeness of the information, or for delays, interruptions or omissions therein, any difficulties in receiving or accessing the website and/or for any loss direct or indirect (including without limitation, loss of profits or consequential loss and indirect, special or consequential damages) howsoever arising and whether or not caused by the negligence of IFS, its employees or agents. The information contained within this site may be changed by IFS at any time.

The information available within this website may include ‘Evaluations’ which are not reflections of the transaction prices at which any securities can be purchased or sold in the market but are mathematically derived approximations of estimated values. Nevertheless, reference may sometimes be made to Evaluations as pricing information, solely for convenience or reference. Evaluations are based upon certain market assumptions and evaluation methodologies reflected in proprietary algorithms and may not conform to trading prices or information available from third parties. No liability or responsibility is accepted (and all such liability is hereby excluded) for any information or ‘Evaluations’.

The copyright of this website and all its content belongs to IFS. All other intellectual property rights are reserved. Redistribution or reproduction of the information and data contained within this website is prohibited without the prior written permission by IFS.

www.liquidmetrix.com is an Intelligent Financial Systems Service: www.if5.com ©Copyright IFS 2009