We are committed to developing metrics that help improve our measurement and evaluation of market cleanliness. This helps us monitor and assess the effectiveness of our regulatory work on market integrity. Market cleanliness should not be gauged by a single metric. It requires numerous measures to reflect potential harm to relevant markets. As such, last year we committed to develop and publish additional measures that will further illustrate the state of UK markets.
Every new addition to our metrics’ portfolio needs to pass the following tests:
- it is a rigorous and meaningful measure
- it can be clearly explained and communicated
- it is clearly focused on a type of behaviour or asset class
- it can be consistently reproduced over time to ensure that trending is possible
This year we are introducing a new measure – the Potentially Anomalous Trading Ratio or PATR.
This new metric complements our measurements of potential insider dealing activity in UK markets, and will sit alongside our long running Market Cleanliness (MC) statistic[1] and the Abnormal Trading Volume ratio (ATV)[2] which we began publishing in 2019.
Additional Metric: The Potentially Anomalous Trading Ratio or PATR
The PATR for 2018 and 2019 encompasses 1917 potentially price sensitive news announcements (PPSNAs) and then examines trading behaviour around these announcements.
A PPSNA is defined in the Design Decisions table further down. The PATR captures the same products as those included in the ATV ratio, namely equities and CFDs and spread bets that are traded OTC where the underlying is a relevant equity. This gives us a wider pool of data to analyse when compared to the older MC statistic.
As with the ATV ratio, the PATR is also based on the premise that inside information should be properly controlled. It should only be disclosed to those who need to know it and should not be used to trade ahead of its disclosure to the wider market. Trading activity around a PPSNA, which is seen to be atypical for the trader in question and which leads to significant profit, could be an indicator of an unclean market.
Unlike the ATV ratio and MC statistic, the PATR is not looking for trading volume fluctuations or price changes ahead of unexpected price sensitive announcements; it focuses on the underlying trading behaviour around PPSNAs and whether the behaviour can be deemed anomalous.
Methodology
With the PATR we examine profitable trading around PPSNAs where there has been a significant price change for the relevant security following the announcement. We then look at trading activity prior to and immediately after a PPSNA. We call this timeframe the Observation Period.
Within the Observation Period, we identify those accounts that demonstrate anomalous behavioural patterns when compared to their historical trading behaviour. This helps us rule out profitable activity that is likely to be legitimate, such as well researched trading decisions or hedging activity. We split the Observation Period into two stages:
- Pre-Announcement Period: During which we assess whether the trading is anomalous
- Post-Announcement Period: Together with the pre-announcement period, we use this period to calculate profitability
To determine if an account is demonstrating anomalous behaviour, we look at certain characteristics of the account’s trading activity. These characteristics are:
- the participant does not typically trade in the instrument (we use a Benchmark Period to determine whether the instrument is typically traded)
- the participant traded significantly more in the direction of the announcement
- the participant made a significant profit from trading positions established in the period immediately prior to the announcement
The ratio is the value of total trading activity deemed to be potentially anomalous during the Pre-Announcement period over the total trading activity for the same period.
PATR is then calculated as:
PATR value ratio %= potentially abnormal trading value
total trading value
The following diagram shows a PPSNA (represented as the red line) and illustrates the time periods we consider in determining both profitability and the potentially anomalous nature of the trading. We provide further detail on the time periods in the Design Decisions table further down.
Time periods - measuring market cleanliness 2020
| Period (Yr-Quar) | Volume Ratio (%) | Volume Ratio (%) |
|---|---|---|
| 2018-1 | 4.5 | |
| 2018-2 | 7.3 | |
| 2018-3 | 5.1 | |
| 2018-4 | 6.2 | |
| 2018 | 6.1 | |
| 2019-1 | 9.8 | |
| 2019-2 | 8.3 | |
| 2019-3 | 4.4 | |
| 2019-4 | 5.8 | |
| 2019 | 6.7 |
Chart tips: hover over data series to view the data values and filter the data categories by clicking on the legend.
Chart
Data table
The PATR for 2019 encompasses 958 PPSNAs and the 2018 PATR encompasses 959 PPSNAs. The proximity of both numbers is deemed a coincidence and we expect the number of events to moderately vary year to year.
While the PATR is marginally higher in 2019 when compared to 2018, both figures represent a small percentage of UK trading activity during those time periods. We do not yet have an ample historical comparison for the new statistic as the calculation uses MiFID II transaction reporting data (introduced only in January 2018). Over time, we will be able to gain a better understanding from year on year comparisons and quarter to quarter comparisons. For both 2018 and 2019, there are fluctuations in the quarterly PATR results, but both yearly averages are similar.
To fully understand the PATR results, it is important to view these within the wider context of UK trading activity (based on the number of MiFID II transaction reports reported to the FCA). As illustrated by the chart below, our analysis shows that for 2019, 99.2% of trading activity did not occur during a sensitive time period (i.e. preceding a PPSNA where the price of the relevant security moved significantly, please see the PATR design decisions for assumptions relating to price movements). For the 0.8% of trading activity that justified further review, the PATR shows that only 6.7% of that trading was considered potentially anomalous – a very small percentage of overall UK trading activity.
2019 summary pie chart - measuring market cleanliness 2020
It is important to note that even where activity meets the PATR characteristics to be deemed anomalous, it does not mean the trading was unlawful. Thorough investigation is always required to determine the exact nature of the activity and the reasons for trading before we can determine whether there are grounds to suspect misconduct has occurred. To do so, our market abuse case officers will dissect every trade pertaining to the activity and assess the behaviour against various intelligence sources. The PATR serves as an illustration of activity that may warrant further investigation.
Design Decisions
In all measures such as this there are many parameters that must be tuned to ensure that the statistical ‘noise’ is kept to a minimum while ensuring coverage is wide enough to produce a representative and valid measure. These parameters are based on our understanding of market participants’ behaviours. Each market event is unique, with different trading patterns, but parameters must be set to ensure that the measure can be tracked over time.
The following table details the scope of the metric and the key design elements that have been considered when creating it:
We determined the length of our periods by taking account of the following considerations:
- we are looking for ‘timely trading’ ahead of the announcement
- including a short period of trading after the announcement allows more of the profits to be realised.
- however, increasing the length of the periods reduces the number of announcements that are considered for the metric
We recognise that trading patterns around announcements vary for each announcement. Setting these periods is a balancing act which has been informed by our experience in investigating cases.
Design Decisions | Decision | Rationale |
|---|---|---|
| The time periods | Observation Period | The period immediately before and after a PPSNA, which is split into a Pre-Announcement and Post-Announcement Period. This period is used to calculate the profitability of the trading activity. |
| Pre-Announcement Period | The Announcement Period is the period during which potentially anomalous activity is likely to occur. For the PATR calculation, we assume this to be 10 trading days leading up to PPSNA. Based on anomalous trading activity observed, this time period allows us to capture a sufficiently representative sample of activity for our calculation. | |
| Post-Announcement Period | The Post-Announcement Period is a shorter period of 5 trading days after the PPSNA. This period is used in the profit calculation as it enables us to capture the profit being ‘realised’ when market participants trade out of positions taken prior to the PPSNA. | |
| Benchmark Period | The Benchmark Period is the period of 30 trading days immediately prior to the Pre-Announcement Period. This time period provides us with a sufficiently long observation window in which to ascertain 'normal' trading activity in the relevant instrument. We can then compare potentially anomalous trading behaviour against the period of normal trading activity. | |
| Potentially Price Sensitive News Announcements (PPSNAs) included in calculation | We consider all announcements which are potentially price sensitive and where no other potentially price-sensitive announcements occurred during the Observation Period. | The existence of other potentially price-sensitive announcements in any of the periods could harm the reliability of the metric. Where we have identified any other potentially price-sensitive announcements during either the Benchmark or the Announcement Periods, these announcements were removed from the sample. |
| Correcting for wider market movements | We take overall market movements into account when assessing the degree of price movement for an instrument. For example, if a financial instrument moves 2% and the overall market has also moved 2% in the same direction during the same period, we would not consider to be a price movement for the purposes of the PATR. | This seeks to correct for noise in the metric’s results. Where the price of a security moves in line with broader market direction, we can likely attribute the movement to general market momentum as opposed to any particular piece of information relating to that security. |
| The % price change required for a PPSNA to be deemed relevant to the PATR | With the PATR, we examine profitable trading around PPSNAs where there has been a significant price change for the relevant security following the announcement. A 3% change is required for the PPSNA to be included in the ratio. | See later sub-section on Announcement Selection explaining the Z-score and showing the trade-off between parameters and number of announcements considered. |
| Z-score for the Price Difference | The Z-score is a standard statistical measure to understand how typical the score (in this case the price change from T-1 to T) is relative to other scores. If the data is assumed to be normally distributed, then approximately 95% of the scores will have a Z-score between -2 and +2. | See later sub-section on Announcement Selection explaining the Z-score and showing the trade-off between parameters and number of announcements considered.
|
| Profitability Thresholds | We set minimum profitability thresholds which individual trading accounts must hit for them to be deemed as potentially anomalous: £5000 for Individuals and £50,000 for Legal Entities | This filter aims to reduce noise in the PATR results by removing those cases where relatively low profits are secured. Insider dealers seek to maximise their profit from insider dealing, therefore where we see relatively low profits, there is a higher likelihood that the profit is a result of chance, not unlawful trading. For example, a firm trading in a security ahead of a PPSNA, where the trading is driven by the need to hedge their position in line with their risk management needs, may secure a small profit, entirely lawfully, during an Observation Period. As with other parameters, every case is different and would be investigated on its merits, but for the purpose of a repeatable measure, we have set values that we feel are appropriate based on our experience. |
| Trading Direction Filter | 3 | Must have traded 3 times more volume in the direction that benefits from the price move than in the opposite direction. |
| Benchmark Comparison | 3 | Must have traded 3 times more volume in the Pre-Announcement Period than the Benchmark Period, in the direction that benefits from the price movement. |
| Financial instruments included | All equities that are traded on a market with the operating Market Identifier Code (MIC) ‘XLON’. All CFDs and spread bets that are traded OTC where the underlying is an equity that is traded on a market with the operating ‘XLON’ MIC. | We consider this to be a reasonable representation of the UK equity market because the XLON MIC identifies most of the instruments on the primary UK listing venues. We also include key derivative contracts for those instruments, namely CFDs and spread bets, to capture the broader trading activity in those instruments. |
Price movement graph - measuring market cleanliness 2020
It is crucial to factor in the volatility of an instrument when assessing whether a price movement is significant enough to be captured in the PATR.
The figure above illustrates this by showing two instruments with an identical 3% price movement in one day.
The first instrument has a volatile price history (with regular price changes). The second instrument has a very stable pricing history. Therefore, in the example below, the 3% change caused by a PPSNA is only significant for the second instrument due to the volatile nature of the first instrument.
It is harder to attribute the price movement that may be generated by a PPSNA to an instrument with a volatile price history. We will therefore not include such events in the PATR calculation.
Price movement: % change and Z-score selection
Choosing the % change and z-score of the price change for the metric is a trade-off between statistical noise and the number of anomalous events captured by the metric. To ensure the metric is representative of market activity, it is important to set these parameters ‘low’ enough to ensure we capture a sufficient number of quantifiable anomalous events. It is also important however that the metric is not overly distorted by false positives.
The following table shows how the number of events captured by the metric and the resulting PATR of varying the % change and z-score parameters over the 2-year period (2018 and 2019).
| Price Change (%) | Z-score of the Price Change | PATR | Event Count |
|---|---|---|---|
| 2 | 2 | 6.2 | 2080 |
| 2 | 3 | 6.5 | 1321 |
| 2 | 4 | 6.8 | 850 |
| 2.5 | 2 | 6.2 | 2013 |
| 2.5 | 3 | 6.5 | 1302 |
| 2.5 | 4 | 6.8 | 841 |
| 3 | 2 | 6.4 | 1917 |
| 3 | 3 | 6.6 | 1272 |
| 3 | 4 | 6.8 | 829 |
| 3.5 | 2 | 6.5 | 1822 |
| 3.5 | 3 | 6.8 | 1237 |
| 3.5 | 4 | 7.2 | 813 |