Bad data and misinterpretation impede over-reliance on ETF flows data.
In yesterday’s flows article, I talked about one of the biggest problems with ETF flows data; that is, a lot of times, that data is buggy.
The second problem is thornier. Even if you believe the data, it’s important to understand how it can be misleading.
The Second Problem: Misinterpretation
Even assuming that the flows data was perfect, I frequently see flows stories with headlines like “Investors put $100 million into GLD yesterday,” which give me pause. The reality of why money flows into and out of an ETF is slightly more complex than that.
If you’re an investor—whether you’re a hedge fund or my mother—you get ETFs the same way. You buy them on the open market. Only APs actually “put money” into an ETF. And they only do it when it makes financial sense for them to do so. So while ETFs can go in and out of favor—and experience enormous swings in volume—they’ll only actually grow or shrink in size when the ETF becomes over- or underpriced.
Let’s look at a simple example. Here’s the data on the Market Vectors Indonesia ETF, IDX. It’s an annoying fund for a market maker, because to make new shares, they have to go buy a bunch of Indonesian stocks, and that market’s not open during U.S. trading hours. Consequently, they’ll let the fund trade to a bit of a premium or discount before they’ll step in.
So how do we look at a chart like this? Well, the first common mistake is to look at something like the activity on July 18 and think, “This ETF is broken; there must be something suspicious going on.” As reported on July 18, IDX closed at a discount of about 2 percent, but there was nearly $3 million in newly created shares!
If you think back to how creations/redemptions are supposed to work, that’s backward. You’d expect the authorized participant (AP) to show up and create new shares when she could sell them into the market at a premium, and making good on delivering those shares with fair-priced, newly minted shares from the issuer.
But remember from Part 1, this data is generally lagged. So that July 18 spike in new shares is likely related to the premium activity from two days earlier. It’s also worth pointing out that that spike is the creation of 100,000 shares, and with the ETF trading just over that on an average day (and no volume spikes in and around this creation), it’s entirely possible that the AP doing the creation was then stuck with inventory for a number of days.
Having that inventory means they were a better seller of the ETF than they’d otherwise be, so during the sell-off in Indonesian stocks that occurred mid-July, they were fine letting the price dip below “fair” value, as long as they could unload some of their position.
After all, the risk to the AP in unwinding their position through a redemption is that they could end up holding a bunch of Indonesian stocks they need to unload in a downmarket. Here’s what the ETF itself was doing, for context: