I was wondering how exactly ETF's are priced on a minute-by-minute basis... Here is an example. Suppose there is a fictional ETF that consists of all Auto Manufacturing Stocks, with 30% of the ETF being Toyota. And suppose this ETF is traded on the Amex, and that the total float of the ETF is an insignificant fraction of the total Auto Manufacturing stock float, like 0.01%. Now let's say that at noon, someone makes a very large buy of Toyota, causing the stock to go up 5% with a big spike. In theory, this means that the ETF should be valued at 1.5% more due to the Toyota spike. But, let's say that also at noon, someone makes a relatively large sale of the ETF, which requires all the existing buy limit orders within a few ticks of the price. But, remember that the total float of the ETF is only a tiny fraction of the Auto Manufacturing stocks, so the Toyota buy is a much bigger transaction than the ETF sell. SO, at 12:01, has the market maker for the ETF raised the price by 1.5% due to the big Toyota spike, or has he lowered the price due to the large sale of the actual ETF shares ?
Can't truly answer, but it is worth mentioning that if a price discrepancy did exist, it wouldn't be there for long because of arb opportunities.
One stock would never be 30% of an ETF... maybe max 3%... So the question is kinda pointless. Liquid ETFs are very efficiently traded. Many firms have very sophisticated, proprietary software that values ETFs in real-time... And 1 ms routing... Which requires a 6 or 7 figure infrastructure investment. It's not done on the back of an envelope with a pencil by some market maker. Also... The spread and orders for the ETFs... Are routinely manipulated at the AMEX, etc. IMO... One cannot profit by arbing ETFs at the retail level.
To address a spike DOWN in a stock specifically... What happens to the ETF? Let's say a stock is 3% of an ETF and it gaps down by 3% on the NYSE... To use an example. So a $100 ETF must go DOWN in price by $0.09 Within typical electronic latencies... Let's say roughly 10 ms or 1/100th of a second... The bid-ask will widen on the ETF... All manual buy orders in the $0.09 window will be sold into... And within several seconds trading will resume normally in an area $0.09 lower. So manual orders will tend to get the worst of it over and over throughout the day... Not even counting market manipulation and transaction costs. This would apply to the very liquid ETFs that trade > 1 million/day. Less liquid ETFs that trade in 50,000 shares/day range might be more tradeable by a quant with decent infrastructure. And it's never one stock and one ETF. To use gold as an example... Algo programs would track ALL the major gold ETFs, say 5 as an example ... And track perhaps the top 20 gold stocks that make up the ETFs... And crunch numbers all day long looking for specific opportunities to scalp a few pennies... Just classic program trading.
You can't call the question pointless if you don't know the answer. If you prefer, take the same example, and change the percentage of Toyota to 3%, but the question still holds - is the ETF priced by the underlying or priced by the trades in the ETF ?
Come on, do some research before you post. Individual stocks can be 15%+ of an ETF. Heck, I just chose RTH (retail ETF) as an example and there are two stocks in that ETF that are both over 16% of the ETF, Home Depot and Walmart. http://finance.yahoo.com/q/hl?s=RTH The question is not pointless at all. And if you don't think you can arb ETFs at the retail level, just look at the chart of EWS on February 27th. I made a bunch buying it below 11 and selling at $11.60 that day within minutes.
Single stocks routinely make up over 20% of an ETF. Just a few examples: QCOM is 38% of the Broadband Holdrr (BDH). DNA is 38% of the Biotech Holdr (BBH). XOM is 22% of the Energy Spider (XLE). Know what you're talking about before you call somebody's question pointless.
By the underlying. I thought that HoundDogOne did answer the question in his 2nd post. In a nutshell, the arbs will see to it that the ETF is priced at its fair value, derived from the underlying components. That process is not necessarily instantaneous at all times. HDO, the part about 3% max is not correct, though. Even for a broad market-based index like NDX (and so QQQQ), AAPL and MSFT are currently over 6% each, with another 3-5 names over 3% each at any given time. Moreover, the NASDAQ brain trust deliberately designed NDX to be modified cap-weighted, to keep the index from being dominated by a handful of stocks. The exact formula is proprietary. Without such shenanigans, MSFT alone would comprise as much as 25% or more of the Q's by market cap, for many years. There are also dozens of specialized or sector ETFs with double-digit-weighted names, even above 40%-50%. For example: BBH (Biotech), DNA=40.7%; BDH (Broadband), QCOM=41.2%; TTH (Telecom), T=52.4%; BHH (B2B Internet), CKFR=69.5%. If you take a look at, say, any HOLDRS ETF, you'll see that at least 15%-20%+ for the top component stock is the rule, not the exception. I would say the OP's question is neither pointless nor unrealistic.
Dan, do you watch the component stocks trading in Singapore to arb the EWS? They don't trade on US exchanges do they?
It didn't seem to answer the question, because he was just talking about the ETF price changing in proportion to the change in the underlying stock. However, in my example, I was proposing a gap up in the underlying stock at the same time as large transaction in the ETF's stock that was in the opposite direction. I had mentioned that the total cap/float of the ETF was tiny in proportion to the float of the underlying automobile stocks, so the large transaction in the ETF itself could not effect the price of the underlying stocks. So, are you saying that the entity responsible for making a market in the ETF soaks up the big sale of ETF shares, but nevetheless raises the listed price of the ETF to match the gap up in the underlying ?