Technicals of ETF structure (and why they matter more in volatile times)

I’ve seen a number of guides around ETFs and how the mechanics work, but I’ve never been fully satisfied with the explanation. They feel like a glossy car brochure that explain why this car is awesome, have the price laid out, and so on. You get super excited in the dealership and pull the trigger, then they pull out the contract with the $100 admin fee, $300 documentation fee, etc. Even if you walk out of the dealership with the keys, the glossy brochure has lost its shine.

Though the benefits of ETFs are well known, I do feel like there are dynamics that aren’t very well spelt out. I’ll let you google for the basic mechanics to avoid rewriting everything. Below spells out the basics.

You log onto your account on your trading platform (your broker) and you buy a share of QQQ. The broker sends the money to an authorized participant who understands the contents of the ETF, buys the same underlying and gives them to the ETF manager and in return creates a share of the ETF which gets passed back. Simple.

The complexities arise because generally ETF managers have minimum blocks of creation units, e.g. 50,000. So if you buy one share of the ETF the AP has to figure something out. Though the share will show up in your account, the share takes 2 days to be settled.

Sounds simple so far, right?

Many ETFs have very simple baskets. For example QQQ aims to replicate the Nasdaq 100. We could manage this creation/redemption process quite quickly. When you get to Fixed Income, there are more idiosyncrasies.

Firstly, many of the large bond ETFs (e.g. BND, BOND, etc) do not necessarily replicate an underlying index exactly, but have similar or better performance. Their prospectus leave a little bit of wiggle room. This is understandable because there is a liquidity aspect to thing about.

At the point an AP wants an ETF unit, they may not be able to buy everything to exactly replicate a benchmark. Sometimes, there simply isn’t someone wanting to buy/sell specific bonds. It is not reasonable to expect the AP to exactly replicate a benchmark as it may pay a liquidity premium to secure the bonds.

Instead there is a negotiation process between the AP and the ETF manager.

The aspect of negotiation may come as a surprise to you as you’ve been trained that these ETF units are like shares. They are referred to as ‘custom baskets’. Basically the AP will suggest a list of instruments and say ‘these are equivalent in price’; the ETF manager will review their pricing of those instruments vs. their view of the NAV of their basket and if equivalent or better, they will accept the basket.

There is a slight challenge with the process. This is a nefarious and oversimplified example, but the purpose is to show the potential ‘fees’ that weren’t advertised in the brochure.

  • The ETF basket has 1 security in it worth $95. There is 1 ETF share which represents 100 units of that bond, so total value $9500.
  • The AP gets an order for 1 ETF share, so the AP goes and buys 100 units of that same bond. They buy 50 at $95, which pushes up the price so the next 50 are bought at $96. The average price they paid as $95.50.
  • The AP turns to the ETF manager to negotiate the basket.

There is room for maneuver:

  • The ETF share value may still be $9500
  • The AP may say that they will deliver 99 units of that bond (because 99*96 ~ $9500)

In this case the client gets their 1 ETF share with underlying valued at $9500. Yet you should be annoyed! The AP has kept one of the shares to themselves!

Hopefully the ETF manager you have chosen is full of smart people who can tell the AP to go do something to themselves. They should revalue the ETF upon this event and increase the value of the ETF share to $9600 to reflect the increase in value. This protects the existing unit holders.

However, by protecting the existing unit holders, the AP needs to charge a higher price to the person looking to buy the unit. Let’s replay that in a different way, with a smart altruistic AP:

  • The investor initiates a buy order when the ETF share is at $9500
  • The AP sees that this will cost more than $9500 so they charge $9550, which equals the average price paid for the security.
  • The investor will be happy because they executed at $9550 and the NAV of the ETF moves up to $9600.

Great. What about if the AP is smart and seeking profit?

  • The AP charges the investor $9600
  • The AP executes at $9550
  • The investor is happy because after the transaction their $9600 equals a $9600 value

The AP is happy because they’ve pocketed $50 by making this happy. The AP is providing a valid service in doing this, but I would understand why you might be surprised and a bit upset. This cost is difficult to explain, prove and audit. So it’s never on any transaction report. Next time you buy an ETF unit, check the market price shown and the price you executed at. The price you pay will almost always be a few cents higher than you were shown. This whole dynamic is why.

Why are APs stealing my money?!!?

They’re not. I hope.

The APs are often banks and large trading firms. They are taking on lots of potential risk in these transactions. So in a sense they need to build up these profits to pay for an occasional loss.

Instead of a creation, let’s look at a redemption. In this case the AP will be delivered a basket of securities which they have to then in turn sell. In the case of QQQ this may be simple, they sell a future contract of Nasdaq to hedge the price. For a fixed income basket with many different securities, they cannot do this. Some instruments do not tick in small fractions, but can jump a lot.

Some large players use ETFs. Some of the oil producing nations will hold billions of dollars in these ETFs. If they come in and sell $1 billion of ETF shares of something like BND, the selling will move the market. And that selling is mostly the APs problem. Don’t get me wrong, the ETF manager will work with the AP to try to limit the potential damage to their securities. Any drop in price will impact the NAV of their other shares so their interests are aligned. In these cases the ETF structure is stressed, though no-one will want the general public to know about it.

If the price of the ETF unit is $100, and oil-producing nation wants to sell $1 billion, that’s 10 million ETF shares. If the AP executes this with the investor at $97, the instant drop will cause consternation among other investors. And if it’s a large ETF it may impact the wider market. Instead, these transactions are managed behind the scenes.

The ETF manager may choose to sell off its most liquid securities to limit the price movement. But in that case the rest of the ETF investor base is left with a less liquid underlying basket. In the background the ETF manager would then gradually sell off its less liquid positions and recycle into higher liquid transactions. There are two problems with this approach. 1) they are in effect doing two transactions in the liquid instruments and paying slightly more in transaction fees which you won’t notice in the NAV as its share across a large investor base. 2) The ETF is holding less liquid securities and may be more vulnerable to a shock if they cannot liquidate the other securities at a reasonable price.

I say all that not to scaremonger, but to simply explain that there are a few fine print details to mention with these structures.

And what’s all this semi-transparent ETF stuff?

Semi-transparent ETFs (ETMFs) will become a bigger part of the market in the future. The rationale is that there are many trading strategies where the performance is protected when you cannot see the details of their strategies. These ETFs will publish holdings once per quarter. This is counterintuitive to some of the states goals of ETFs relating to transparency.

In the semi-transparent world the AP doesn’t know the contents of the basket explicitly. The nature of the basket will be known to the APR, not the AP. It’s not really clear to me why you need both an AP and an APR in this model. It seems duplicative, but given these structures are not in place and being heavily traded yet, my understanding may not be perfect.

To get approved there has to be some mechanism to provide transparency. ActiveShares is looking to use a “Verified Intraday Indicative Value” that gets published once per second. But it’s not clear whether the APR or the ETF manager will produce that price. You can also drive a truck through that definition. What is an ‘indicative’ value? It certainly isn’t a ‘firm’ value that you are guaranteed to trade at.

Given the lesson you learnt about how the AP can pick up some $$$ in the process of greasing the wheels of the ETF structure its not clear who picks up that money in the process of creations and redemptions.

The SEC can approve other approaches. Another approach includes providing a ‘proxy portfolio’. This involves the ETF manager/APR producing a portfolio that doesn’t specifically exist but is expected to behave like the underlying. Whenever there are these kind of gaps/omissions/reduced transparency, there is the introduction of mispricing risk. Who will wear that cost is not clear.

Again, I don’t want to be downbeat about ETMFs. Though they will come with risks, they do offer the benefit that sophisticated fund managers may choose to expose their higher return strategies via ETMFs, safe that their secret sauce will not be discovered. Though there are pros and cons, I’m sure the brochure will only contain the positive!

Spent career at various financial institutions across continents and across the buy/sell side. Content/thoughts are my own, not my employers’