The third calendar week of SEF trading is completed. So, how are the first few weeks of SEF’s trading going? Regulators have been quick to announce that SEF’s have seen a successful start, while industry critics have quickly countered that more time is required to integrate the SEFs and onboard clients.
Thankfully, in the interest of transparency, the CFTC requires SEF’s to publish daily market data reports. This has allowed interested parties – like me – to mine the data and perform analysis, such as the ones we regularly post on the Clarus website.
However, this analysis has proved difficult. The difficulty stems from the fact that the different SEF reports are in various file formats, published at different times of day, and contain varying degrees of information, among other nuances. As my own frustration with the analysis grew, I began taking notes of every grief I came across; after all, this is supposed to be transparent.
When finished, I had a list of 11 attributes that I felt all SEF’s should strive to adhere to. The table below summarizes the best practices and my assessment of how each of the SEF’s currently conform to them.
click to enlarge
Below follows a brief description of the best practices I would prescribe:
1) Data available on website. It should go without saying that SEF’s need to make this easily accessible. I found general conformity, with only those inactive ones yet to make the data available.
2) Publish daily data. There were many occasions when I would check for the latest reports at 8pm, then at 6am, and still not see data for the previous day. In some cases data a report is not made available if there was no activity, and in some other cases a report is delayed and reported for multiple transaction dates. SEF’s should make the report available on a daily basis regardless of activity, with the leniency to interpret holidays and non-working days as appropriate.
3) History Available. There had been a couple cases where the SEF did not publish data for a particular day, or I was late in checking for the activity report. The archival of public reports should be done and made publicly available.
4) Machine manipulable table of data. My research purposes require me to manipulate the data in Excel. Some firms deliver the data in CSV/XLS formats, some are in HTML, and some in PDF. PDF’s are notoriously bad for extracting data; some may claim that this is in fact one of the intents of the format. I am often able to cut and paste into a text editor and import as space delimited, however spaces in complex data like this mean that data is often offset and not readily usable. Consumers of SEF data should be able to extract and manipulate the data in spreadsheets or more sophisticated data mining tools.
5) Clear Asset Class Designation. When the data is extracted, I want to readily interpret what I am looking at. There is a wide spectrum of conformity here, but I’ve noticed some SEF’s making clear the asset class (IRD=IR, FXD=CU, CRD=CD, CMD=CO, EQD=EQ). The other end of the spectrum requires one to interpret some strings of product description, culminating in my personal favorite product I’ve seen: anyone care to make me a market in “8Y 6M_null”?
6) Readily identifiable CCY Notional. The volumes reported need to be understood. Common complexities are with FX. With my FX background, I am used to interpreting notional data in terms of the primary currency, so that for example a EUR/JPY notional of 1mm means it is 1 million EUR. Just when you think you understand how the data is being reported, you get some USD/IDR trades (market rate is approx. 15,000 IDR to 1 USD) reported as 80 billion. While possible to trade 80 bn USD worth of IDR, it does represent 10% of the country’s GDP! The most favorable example I have seen on this is a few firms reporting USD equivalent volumes, which makes the data easy to understand. Cross ccy swaps, equities and commodities also lend themselves to difficulty here.
7) Readily identifiable price range. While my analysis does not mine the price data, the spirit of SEF reporting includes transparency into prices. The best examples I have seen are the reporting of open, high, low, and close.
8) Readily identifiable date of report. Particularly when you are not sure if the SEF has updated their daily activity, it would be nice to have an indication of what days’ data you are looking at. File names, tab names, columns or headers are generally used, but in some cases this had been neglected entirely. This has generally been cleaned up.
9) Readily identifiable cutoff times. The FX market has a fairly well agreed end of “business day”. However the SEF reports are currently made available at widely disparate times of day. One SEF clearly denotes on the report the time window of their reporting – for example from 5pm EST Monday through 5pm EST Tuesday.
10) Trade counts. Is that 100 million notional one trade or 100 small trades? Particularly when more clients onboard and use SEF’s, there is the expectation that average trade sizes will decrease. We will require trade counts in order to understand this.
11) Timely reporting. As a best practice, effort should be made to have reports available at a particular time after the close of the reporting period (as discussed in point #9). I often found myself in the late afternoon waiting for a report to be made available for the previous business day. Particularly for those firms that do not make history available, you need to grab the data before it gets overwritten by the new data. My analysis above is admittedly qualitative on this practice.
Ultimately, firms such as Clarus would like to mine this SEF data in an automated fashion and provide tools for the industry to consume the information in a digestible and useful manner, similar to how we already do this for the public SDR data with our SDRView product. But we’re not the only ones. This data can be used for academic studies, decision support system development, trading strategy analysis and many other purposes. In the spirit of transparency with which the SEF rules were created, we should expect at least some standardization of data reporting.
It is important to note that the CFTC do not publish a standard for SEF reporting. Given the nascent nature of SEF’s and the relative complexity of some of this data, this is understandable. Don’t forget as well that even SDR reporting lacks standards. There is, however, an initiative by consumers of SDR data (e.g. the CFTC) to standardize the data. A similar initiative is required for SEF reporting.