ETF with international constituents

I was just wondering how the fair price of ETF with international holdings are accurately priced. Say we have an ETF that trades on domestic exchange but the holdings are international and the market is closed i.e. we do not have the spot prices of the underlying holdings. How then can the ETF market makers accurately price the underlying holdings? From my understanding, one way to do so would be through the use of futures contracts for the stocks but I'm not sure how that is being done. Can someone shed light on this/provide an alternative? Thank you!

• A U.S. ETF that references a stock of a non-U.S. corporation is likely to use American Depositary Receipts (ADRs) that trade in U.S. exchanges in the same hours as U.S. stocks (or GDRs etc). Do you have an actual example of an ETF that uses underlying foreign assets? – Dimitri Vulis Jan 24 at 14:58
• Take a US traded ETF like FLKR for example (Franklin FTSE South Korea), it holds stocks which trade in South Korea on a timetable roughly opposite to US stock market hours (i.e. in the evening and night from a US perspectives). How then do the shares get priced during the US daytime hours? – noob2 Jan 24 at 15:18

Pricing an ETF always involves estimation. Most ETFs have a regular trading market that is only a few bps wide. Even in the US we often have ETFs with relatively illiquid underlyers but with an ETF that is priced tight.

Market makers look for all kinds of proxies to model the fair value of ETFs. For international ETFs it will be a combination of ADR's, foreign futures (Kospi, TOPIX, etc), and related local equities. As the clock cycle moves and some products become more - or less - liquid the market maker will adjust the way that the ETF price is projected from relevant underlying securities.

The goal of this constructive pricing is to come up with a fair value for the ETF. This is a precise number. We might determine that the fair value for a given security is \$100. That means that we will be willing to buy for \$99.9999999 or sell for \$100.00000001. But, there are frictions. It might cost another \$.02 to carry all of the underlying components. So now that means I will buy the ETF for \$99.979999999999 or sell it for \$100.0200000000001.

But, why would I want to do that? I'll try to buy the ETF for \$99 and sell it for \$101. OK, but then all of my competitors will cut that margin. Then I'll cut the margin. And pretty soon you are down to a market that might trade very close to fair value. And then put on top of that the possibility that you have people with real flow. For example, a customer puts in an order to a bank to buy 1mm sh. The bank might start working that order in the middle of the spread. So now, when you look at the market you will see the ETF bid over fair value!

Such is life.

• I understand that given the derivatives market trades roughly 23 hours a day, one could use the futures market to get a gauge but do you have a specific way of how that is being done? What could one infer from the futures price? Thank you! – user52091 Jan 25 at 1:56
• The futures market is certainly the go-to source for pricing non-US ETFs. That's what I used to do at my old shop. I beleive it's common with most competitors in this space. – JoshK Jan 25 at 1:59
• With your experience specialising in ETF, what would you say are the main factors that go into how narrow/wide the ETF bid/ask spread should be? I assume that with a simple vanilla ETF with domestic constituents, the fair price can be easily computed based on the underlying shares. But what factors go into the bid/ask spread? Thanks – user52091 Jan 25 at 2:16
• I will edit the answer to add that. – JoshK Jan 25 at 2:20
• @user52091: re first question. When Korea mkt closes you write down on piece of paper the NAV of the ETF and next to it the price of KOSPI futures. Hours later, with mkt closed but KOSPI open you project that closing NAV has changed by same percent as KOSPI fute. To get fancier you can have linear relationship with a "beta" other than one. If desired you can estimate a multivariate eqn with predictors other than KOSPI change on the right hand side. HTH. – noob2 Jan 25 at 9:22