Clarus Financial Technology

FX NDF Package Transactions

Today, I looked at packages of FX NDF trades on the SDR, expecting to be able to find some basic packages, or even just logical groupings of trades such as:

However, what we find however is not so obvious.

Background

We at Clarus do a good job of normalizing trades reported to the SDR.  Taking Interest Rate swaps for example, we are able to assign our own product subtypes (Spot start, MAC, IMM, etc) as well as packages (Curve, Butterfly, Compression, etc).  Are there any such package trades in the world of FX NDF?

Headline Statistics

I grabbed data for the month of August 2017, and took the 4 most active USD ccy pairs on the SDR (BRL, KRW, INR, and TWD).

After processing all of the cancel corrects, and only looking at NEW trades, we see over 100,000 trades for August, so roughly 4,100 per day.

I defined a “Package” loosely for this exercise; a package FX NDF is any group of trades having the same time-stamp (down to the second) and the same currency pair.  Using this logic, I found 14,627 package trades, accounting for 48,330 trade legs.  Hence nearly half of the trade legs we see are part of a package of some sort.  In numbers, for our month of data across these 4 currency pairs:

NDF Trades & Package. August 2017. KRW, INR, BRL, TWD.

So packages of NDF’s seem prevalent.  I had anticipated most packages to be 2-legged swaps.  So I then broke down these packages by the number of trades contained within them:

NDF Package Count. August 2017. KRW, INR, BRL, TWD.

So here I am thinking that the 2-legged packages are FX NDF Swaps, and the packages containing 3 or more would be Par Forwards.

I should clarify that yes, there were 119 distinct packages of trades that each had 10 legs.  There were even some packages of 130 and 135 trades!

So I dug into some of these structures a bit more.

2-Legged Packages

The oddest finding here is that 60% (5,900) of the 2-legged packages had the same value dates!  So they are not Swaps.  And slightly concerning here is that the 90% of these (5,200) had the same dates and price.  Prime Brokerage?  And of these, 60% (3,000) have the same dates, price, and notional.  Duplicate trades?  Maybe, possibly, auto-hedging, but at the same price and with no reported fee?  And what are the trades with the same dates and prices, but different notionals?  Maybe:

However on the bright side, this does leave us with roughly 4,000 packages that we could call real NDF swaps.  And probably 3,000 that can be called Prime Brokerage (back to back, same price and dates).

Last comment here, the vast majority of the swaps we see were fixed USD notional.  I would have expected some FX NDF swaps to roll out the foreign currency amount from one date to another.  For example if the near leg is 1mm USD vs 64mm INR, I would expect the INR amount to stay the same (64mm) and the USD notional change.  However this accounts for precisely zero trades in our sample set.

EDIT:  Some readers have told us that FX NDF swaps are often traded as a single contract, but booked as separate legs, which might cause different execution times.

3-Legged Packages

Of the 1,809 packages here, 1,450 of them had the same delivery dates.  I could only describe these as allocations.  Although a surprising number have the same rate and notional.  Could these be duplicates instead?

The other 350 trades had just two delivery dates.  So for example, 30-Aug, 30-Aug, & 29-Sep.  Very odd.

N-Legged Packages!

So what were the massive trade packages?

The 135-leg package was a strip of forwards, with 29 different value dates.  Very similar was the 130-leg package.  Those do not fit the Par Forward model.  Nor do they resemble a straight allocation.

A real mystery.

Summary

I have to leave it there and open this up to our readers to chime in with their thoughts.  Either public comments below, or privately to us.

In summary, we attempt to assess package trades in FX NDF and find:

Is it possible that some firms are reporting NDF’s in batches throughout the day with the same timestamp?

If anyone has any better ideas, we are all ears.

Stay informed with our FREE newsletter, subscribe here.

Exit mobile version