September 5/6, 2019 – Ljubljana, Slovenia
Today I’m providing my thoughts on the Forum’s discussions and debate around “Market Surveillance”.
As mentioned in my previous articles, what follows is my personal assessment of, and reflection on, the two days of discussion and is not intended to reflect any official pronouncements by ACER or any of the stakeholder interventions by the industry associations present.
The keynote address was made by Vinay Tripathi (Office of Enforcement) of the US Federal Energy Regulatory Commission (FERC). Vinay provided an overview of FERC’s electricity market surveillance programme, addressing the following areas:
1. Description of the North American power grid and wholesale power markets, including US and Canadian interconnectors, balancing authorities and RTOs/ISOs
2. Market manipulation
3. Data overview and the complexities that it brings to the surveillance programme
4. Review of some scenarios and how they are presented in the FERC’s surveillance tools
5. The broader surveillance process
FERC continues to view market manipulation as a complex and evolving area, where an exhaustive classification of manipulative schemes would be extremely difficult or impossible to define. FERC consider there to be two broadly applicable types of market manipulation:
1. Cross-market manipulation
2. Gaming of applicable market rules i.e. taking advantage of the rules of a market in a way that provides a market participant, or market participants, with an advantage that others do not take
FERC described how it set about identifying manipulative behaviour with reference to the concepts of “tool”, “target” and “benefiting position”:
1. Tool (mechanism)
a. Usually a physical tradeable instrument or a physical asset used by a potential market manipulator
b. Often characterised by uneconomic or sub-optimal trading, although assessing this can often be extremely challenging e.g. in the case of physical assets, it requires evaluating a complex matrix of economic parameters
2. Target (objective). FERC referenced the following examples:
a. Sending a false or misleading signal to the market or software used by that market
b. Affecting an index or benchmark price
c. Targeting an out-of-market payment outside its intended purpose e.g. out-of-market make whole payments to generators. “Make whole payments are made to market participants (also known as resource owners or generators) when they face a shortfall between their resource’s offer price and the revenue earned through market clearing prices. Ideally, market prices would fully compensate all resources for the variable cost of providing energy. In reality, market outcomes are impacted by a number of technical and operational limitations that may not be foreseen or modeled sufficiently in the software used to clear the market” (http://www.ieso.ca/-/media/Files/IESO/Document-Library/market-renewal/Fact-Sheet-18-Make-Whole-Payments.pdf?la=en)
3. Benefiting position. Positions that could be benefited include
a. Swaps, futures, swap-futures and options
b. Financial transmission rights (FTR) that settle based on locational price spreads of nodal electricity prices
c. Generation portfolios
d. Out-of-market make whole payments
FERC then described the buildout and screening for market manipulative strategies:
1. A mixture of public and non-public data is used
2. The Division of Analytics and Surveillance (DAS) builds alert scenarios based on market rules and known manipulative strategies prioritised using the risk framework described above (tool, target, benefiting position)
3. The alerts scenarios (FERC calls “screens”) are run on a monthly basis and the results are posted to a DAS intranet
4. DAS analysts then review the results and present the screening findings to the broader DAS team
FERC then demonstrated some of the interactive tools that it has available to it on its electric surveillance intranet and how these can be used, including the collaborative work it does with the US Commodity Futures Trading Commission (CFTC) with respect to cross-market manipulation and financial contracts.
There then followed, from my perspective at least, an interesting section on market surveillance data sources. The principal data components, on which DAS analysis is based, are:
1. ICE transactional data for physical power and gas products
2. RTO/ISO data – non-public data from organised electricity markets
3. Large trader reports – essentially daily open financial positions of large traders of financial natural gas and electricity contracts
4. E-Tags (North American Reliability Corporation (NERC) Tag) which represent a transaction on the North American bulk electricity market scheduled to flow within, between or across electric utility company territories
5. Publicly available data such as prices, volumes and other subscription data
6. Electric quarterly report detailing transactional electric sales data submitted by electric utility companies
Given the differences in the ISO data schemas, and in order to make such data useable and comparable, some 80% of DAS effort is spent on rationalising and optimising the data prior to its analysis. This insight then led into a broader discussion of the surveillance challenge faced by FERC, including:
1. Extraction, transformation and loading of data – this is not just a problem of volume but also multi-dimensional complexity and, often, the extreme rawness of the data provided. This is a predicament that I have extreme sympathy with
2. Data cleansing – essentially the process of detecting and correcting corrupt or inaccurate records from the source data provided. FERC considers this to be a significant part of the surveillance effort and, with my eight years’ experience of working with transaction surveillance systems, I would concur
3. Creation of dynamic and flexible analysis methods often with the use of open source programming languages and visualisation techniques, with tools like Power BI or Tableau
4. FERC emphasised the fact that market surveillance is an iterative process as market design changes and market participant behaviour continues to evolve. A supplemental area of complexity, which also requires an iterative approach, is the evolution of data (and data schemas) and its sources. FERC noted that data schemas can change as often as four times a year – this, of itself, imposes a significant burden on surveillance resources and maintaining the reliability and scalability of the market surveillance programme
FERC then provided an overview of the surveillance process, including:
1. Screening of surveillance alerts
2. Determining whether further investigation is warranted, having regard to:
a. all data available to FERC, including market fundamentals, physical transactional data, financial positions and relevant contextual data e.g. outages, generating asset economic and dispatch data
b. Contacting the relevant market participant(s) asking for details (Request for Information (RFI)) about the observed behaviour, additional relevant data and, as necessary, interviewing key personnel such as traders and trading supervisors
c. Follow-up analysis of data provided by market participants as part of the RFI
3. If DAS continue to have suspicions in respect of the market participant’s behaviour or conduct, then it will consult with FERC’s Division of Investigations as to whether an investigation should be opened
I thought that this was one of the most useful and insightful presentations of the EMIT forum. It dealt with the complex and often “dirty” world of data, understanding the data, data quality, data processing, summarising the data, data analysis and data visualisation. Even with all the resources that FERC has available to it, or can call upon, it is clear that c. 80% of the market surveillance effort is below the surface of this iceberg and that what the market sees, in terms of FERC enforcement or FERC white papers, is but a small part of the effort required.
Clearly, the amount of market surveillance effort required is dependent upon the context of what the market is, ranging from an individual trading venue to transnational infrastructure, and whether the surveillance teams have access to, and control of, the data that is required to inform its surveillance efforts. Nonetheless, even if only in microcosm, all market surveillance efforts, on the part of participants, venues, exchanges and trading facilities, will face many of the challenges described by FERC on an ongoing basis.
I would consider that FERC operates a reasonably sophisticated market surveillance programme and, therefore, it should be of interest to all EU regulators and market participants the amount of continuing effort and diligent maintenance that is required, often just to keep pace with “data change” below the waterline of the iceberg, let alone continuing to build a proactive, resilient and scalable surveillance programme as market design and participant behaviour changes.
I would also refer my readers to two FERC papers that should be of interest in the context of surveillance and how such efforts are resourced from a market participant perspective:
1. Staff white paper on anti-market manipulation enforcement efforts ten years after Energy Policy Act 2005 (FERC November 2016)- https://www.ferc.gov/legal/staff-reports/2016/marketmanipulationwhitepaper.pdf
2. Staff white paper on effective energy trading compliance practices (FERC November 2016) - https://www.ferc.gov/legal/staff-reports/2016/tradecompliancewhitepaper.pdf
That’s it from me on the EMIT forum. My next article will compare, and contrast, the world of regulator/venue market surveillance programmes and the surveillance programmes operated by individual market participants. I will attempt to answer two market participant’s questions, which have been posed of me as a Head of Compliance:
1. Why does the regulator’s RFI about our behaviour not match up with the output from our own surveillance processes?
2. Should we have been expected to see this RFI coming and been better prepared to respond to it?
Comentários