I got an email recently from a reader (and relative) that I thought I’d answer through posts. Here is the first one:

Q: Does an ETF have to go up or down in general proportion to the underlying stocks that make it up?  Like if ARKK was made up of just 10 stocks, and one day all 10 went down 5% each, would ARKK itself drop 5% in aggregate?  Or could people decide they liked ARKK as a long term investment and buy it and ARKK goes up 5% despite each and every stock that makes it up dropping 5% the same day?

A: An ETF is like a mutual fund in that both are special purpose corporations (details). One key difference is that mutual funds handle recordkeeping and sales/redemptions themselves and deduct prospectus-specified administrative charges from overall fund assets for doing so. Mutual funds sell/redeem once daily, after the US market close, and at net asset value. So the question of transactions away from NAV never comes up.

In contrast, brokers that an ETF designates as authorized agents keep the fund’s records of trades. They also make a market in the shares, buying and selling them during normal market hours, just like they do for individual stocks, and taking a bid-asked spread as compensation.

An aside: the major stock exchanges update prices of US-traded stocks every 15 seconds; brokers use these feeds to calculate the NAV of any ETF precisely (a practical impossibility for you and me). So we can’t know exactly how our buy/sell price differs from NAV at the time we transact. This is a big edge for the broker. But in a normal market, the broker markup will likely be less than 0.5% of NAV.

In the real world, brokerage house traders routinely keep open positions, both long and short, that they plan to close at a more profitable point later on. Let’s assume, though, that the trader’s goal is not to have any risk on his books and to close positions as fast as possible.

Let’s look at the case of common stock XYZ trading at $10 right now.

In the simplest case, one client wants to sell 100 shares of XYZ and at the same time a second customer wants to buy the same number. The trader simultaneously agrees to pay the seller $9.95/sh and charge the buyer $10.05/sh, and pocket the $.10/sh spread. The trade will settle in two days.

If the sells are 300 shares and the buys 100, the broker will execute the buy/sell of the first 100 shares in line and lower the price he is willing to pay for the rest to, say $9.80 or $9.75. This process both draws in new buyers and encourages price-sensitive sellers to withdraw their orders.

If, in contrast, the buys are 300 and the sells 100, the broker will act in analogous fashion by raising the price he is willing to pay for the rest.

In either case, the right-now hopes and fears of the market can move XYZ up and down, without apparent regard for any intrinsic value of the underlying company.

An ETF is different, however. The underlying assets are themselves publicly traded stocks, whose asset value the broker calculates every 15 seconds. Also, the agreement between ETF and authorized agent allows the agent to both exchange ETF share for the underlying assets and to exchange packages of the underlying assets for ETF shares.

So if the ETF begins to trade above NAV, the broker will sell the ETF short in the market, buy the underlying assets, convert them into new ETF shares and use them to close out the original short. If the ETF is trading below NAV, the broker buys the ETF, shorts the underlying stocks and converts the ETF into stocks to close the short.

In simpler terms, if the ETF is trading at $100 and NAV is $105, the broker buys the ETF (spending $100) while shorting the underlying stocks (taking in $105), and then exchanges the ETF share for the underlying stocks and closes the trade. This arbitrage is simple, can be done automatically, is virtually costless–and nets $5.

If the ETF is trading at $105 and the NAV is $100, the opposite happens. Short the ETF, buy the stocks, present them to the ETF for a new ETF share, close out the short. Net $5.

So, although in theory the situation you describe could happen–the ETF trading at a substantial premium to NAV–it’s highly unlikely, both because the ETF can create new shares to meet stepped-up demand and because the arbitrage opportunity, were a premium to develop, is so lucrative.

One possible exception: at the bottom of the market during the 2008–09 financial crisis, shares of ETFs containing only foreign stocks closed one day at discounts of as much as 12% to NAV. The main reason, I think, is that foreign markets weren’t open at that time of day for arbitrage to happen.