A Leveraged ETF Trading Flow Case Study: Gold Miners – $GDX $NUGT $DUST


Today I came home from soccer practice (coaching, not playing) and was catching up on the FOMC-minutes-induced madness that had occurred in the previous few hours.   What caught my eye was a large MOC (market on close) print in the Market Vectors Gold Miners ETF ($GDX – no positions).   $GDX had a monster move on the day – up 6.7%, almost all of that in the last 2 hours of the trading day after the Fed minutes – but what I found odd was the direction of the market on close trade.  21.8MM shares traded in the official closing print on ARCA, but what I noticed was the price action on the print:

GDX's MOC action, 10/8/14

GDX’s MOC action, 10/8/14

GDX was crazy to buy all day, so who the hell was selling this massive order on the close?   Now, don’t get confused and say “Hey Kid Dynamite, don’t be an idiot, every seller has a buyer and every buyer has a seller – maybe this was a buy order on the close.”   Well, sure – there WAS a buy order – that’s how an auction works: the buyers and sellers get matched at the clearing price.   My point is that sellers overwhelmed buyers on the close.  That’s clear from looking at the chart.  But we’ll come back to that in a minute.

So @benjaminburke replied to me on twitter that this was a $NUGT rebalance.   $NUGT (no positions) is a triple-leveraged fund that is indexed to 3 times the daily return of the Arca Gold Miners Index, which is the underlying index for $GDX.   This explanation didn’t sit well with me, because you see, the leveraged ETF’s rebalance should be in the same direction as the price move in the underlying: on big up days, the triple leveraged $NUGT will be a buyer of $GDX exposure, and on down days, $NUGT will be a seller of $GDX exposure.   Let’s go through the math real quick, shall we – because leveraged and inverse leveraged ETFs constantly confused new readers in the way they work.

Yahoo Finance tells us that, as of 9/30/14, $NUGT had assets under management of $ 697MM.    Since they need to get triple the daily performance of their index, they need triple that amount in exposure.  So their exposure is long $2.091 Billion.   Now today, the underlying index (we can use $GDX as a proxy)  rises 6.7%.  We can calculate that $NUGT’s exposure is now worth $2.231 Billion.   They’ve gained $140 MM.   That’s a 20% gain on their $697 in AUM, which is exactly what they want (well, that’s no accident, it’s just math based on their exposure).

Here’s the important part:  now $NUGT has $697 MM + $140 MM = $837  MM in assets under management.   So they need triple that in exposure for the next day: $2.511 Billion.   They only have (calculated just above) $2.231B, so they need to buy $280 MM in exposure.

Summarizing:  $697 MM in assets –> $2.091 Billion in exposure, which yields $2.231 Billion at the end of the day, which results in $837  MM in assets for the next day, which requires $2.511 Billion in exposure to maintain triple leverage, which requires $NUGT to increase exposure by $280 MM.

Makes sense, right?  So $NUGT needs to buy $GDX exposure at the end of the day on days like today.  Notice that I wrote “GDX exposure” and not just “$GDX”.   $NUGT holds a bunch of swaps with counterparties instead of just holding $GDX itself.   Of course, the counterparties will hedge these swaps – so someone is going to be a net buyer of GDX a result of this rebalancing, even if it’s not $NUGT itself.  You can view the exact details of $NUGT’s holdings of GDX and swaps (including actual up to date numbers) at the Direxion website.

But wait – there’s more.  There’s a  triple leveraged INVERSE ETF  – $DUST (no positions) – which seeks to deliver negative 3 times the daily return of the same index. Here’s another confusing part for some people – its rebalance flows are in the same direction, even though it’s leveraged short.  Let’s walk through it, shall we?

$DUST had $209 MM in assets as of 9/30/14.  That means they’d need -3*209 = -$627 MM in (short) exposure to the GDX.   Today, GDX was up 6.7%, so their short hedge portfolio is now worth $42 MM more (a loss of $42 MM for $DUST), or -$669 MM (their short went up in gross notional value).   Their assets are now $209 MM  – $42 MM  = $167 MM.   For the new (tomorrow) assets number of $167  MM, they’d need -3*167 = -$501 MM in exposure – so they need to COVER $168 MM in short exposure.  In other words, the leveraged short ETF ends up short too much exposure when the underlying index goes higher, so they need to cover some of their short.

Take our back of the envelope numbers of $280 MM in exposure to buy from $NUGT, and $168 MM in exposure to buy from $DUST and you get $448 MM in exposure.   The MOC print in $GDX was 21.8 MM shares at $21.94, which equals $478 MM in notional.   Not bad for some quick math with slightly outdated Yahoo Finance data.

Now, let’s get back to the main point of my initial question:  you can see that the rebalancing of the leveraged ETFs leads to size buy orders in $GDX market on close.   Why, then, did we see GDX get marked down into the MOC print?  Where did all the sellers come from?

Well, I have to admit, I shouldn’t have been confused about this – after all, this is one of the reasons I’m retired now:  my trading desk made a living off of MOC-related flows in my prior life.   Index changes were our specialty – we figured out the index-related flows before everyone else did (they’re common knowledge now, and very hard to trade), and were able to effectively manage the risk that our clients didn’t want to take.   The basic formula was to buy 15-20% of the order early, buy 65-70% of the order on the benchmark – MOC – and then short 15-20% of the order to the client on the print as a “cleanup” trade.    This is the point where some idiot will pipe in “Hey KD – you were frontrunning your clients and giving them worse executions as a result.”  On the contrary – by taking risk in early executions (and guaranteeing our clients their benchmark, while working for better prices for them), we MINIMIZED their MOC impact.  Similarly, every share we shorted to our client on the close was one less share of buy-impact they had.

Anyway, my point here is that these days this game is very very hard – EVERYONE knows the flows, so everyone tries to profit from “pre-trading” them.   Guess what happens when everyone knows that XYZ is to buy on the close and they go and buy XYZ before the close and sell it on the close?   The trade goes the wrong way:  sellers overwhelm buyers because the Market has over-played the trade.

Markets may not be efficient, but they are efficient ENOUGH, and anyone who trades index-related flows can attest to the mammoth shift in this business over the last 15 years.  What was once basically free money is now a multi-level poker game of trying to guess how much The Market has played the trade that everyone knows is coming, and if this specific trade is one that’s ripe to blow up and go the wrong way on the close.    Did guys get blown up on the last trade, and might that make them less likely to play this trade?   This is the mult-level “what is the other guy thinking?” game that is The Market.

Which finally brings us back to today’s $GDX MOC action.  Guess what – The Market knows the math I did in this post above.  The Market knows that when $GDX rallies, $NUGT and $DUST have to rebalance their exposures by buying more GDX.   So what do traders do?  They buy $GDX early, and sell it back to the rebalancers on the close.   In this case, in fact, they got it mostly right:  GDX traded down a little over 1% on the close, but only guys who bought in the last 20 minutes lost money on the pre-trade.   Of course, The Market didn’t get it exactly right: it was a slightly over-played trade, as evidenced by the net selling on the close to cancel out the massive buy rebalance (see 1, below).


Leveraged ETFs and High School Level Mathematical Truths

My ETF Archive



1 Similarly, if no one had pre-traded the rebalance, and thus no one had accumulated supply of $GDX to sell on the close into the rebalance buy order, we would have expected to see an uptick in $GDX on the close from the rebalance buy order, which would need to find sellers at some higher level to reach a clearing price.

postscript – I haven’t addressed the “feedback” loop of leveraged ETFs in this post, although I did the math to show that it exists.  In other words, one may very well argue that the trading mechanics we saw today are very different from the trading examples from my own career I gave where The Market was trading index changes.   Here, you see, the pre-trading by the market affects the trade itself!  The more the market buys in anticipation of the rebalance trade, the more the rebalancers have to buy as a result.



Kid Dynamite is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to Amazon.com. If you click on my Amazon.com links and buy anything, even something other than the product advertised, I earn a small commission, yet you don't pay any extra. Thank you for your support.

The information in this blog post represents my own opinions and does not contain a recommendation for any particular security or investment. I or my affiliates may hold positions or other interests in securities mentioned in the Blog, please see my Disclaimer page for my full disclaimer.

blog comments powered by Disqus
Kiddynamitesworld Blog