We realize we've written a number of blogs now about 'nuances' in financial analysis. This is the nature of it --- you must learn some of the details involved in assumptions. It is a part of what being a financial analyst is about.
The nuance we will discuss here concerns 'tracking error.' Tracking error just means how closely the ETF price reflects the underlying portfolio value.
The way an ETF provider measures ETF tracking error and posts on their website is to calculate the midpoint of the closing bid-ask spread vs the underlying per share value of the benchmark it is tracking. The nuance here is that this is not how regular closing price is determined in your brokerage or quote service.
Closing price defaults to the last trading price of the day. So we have a mismatch. On the one hand, if you went to trade a given ETF, you obviously would get it near the bid or ask. But if you are looking at the official exchange closing price on any given day, it will be the last price -- not the closing bid-ask midpoint.
This actually has important implications in backtesting as you could have large tracking error on market price -- even if there is low tracking error as calculated by the ETF providers.
Happily, there is an easy way to adjust for this --- only use data where the ETF is liquid (and WAS liquid for the backtest period under analysis). It the ETF is liquid, it will actually be trading at the end of day and the closing price and the midpoint of the closing bid-ask will be very close -- true tracking error will be small.
We will point out that in this case, it is not that the ETF providers are doing something wrong --- they are showing a good reflection of an actual trade if done at the close. It is clearly up to the financial analyst to think about this and do their own diligence. You are the one responsible for your decisions in the end -- nobody else.
Let's say an ETF only trades a few times a day -- and let's say that on a particular day the last price was 10am. Then let's say the underlying index moved materially up or down into the close -- the closing price would reflect the 10am price -- despite clearly not being a true price you could trade -- and note that this difference would NOT be considered 'tracking error' because that is determined by the midpt of the closing bid-ask. But there clearly will be substantial tracking error using closing price data when the security is illiquid.
Again, there is a simple solution to this --- only use data where there there is substantial liquidity. We have hundreds of choices of liquid ETFs that have been liquid for many years now -- you don't have to bother with the 8th US Consumer Discretionary ETF that is slightly different than the first 7 but trades no volume. The data won't be any good. If its a valid investment -- 9 times out of 10 it will get to liquidity relatively quickly. If you want to use less liquid ETFs, that is fine so long as you understand the difference. Take pride in understanding these nuances -- it's part of being a good analyst.
Related to this topic, take note that we've listened to user requests and added a page to show New Additions to ETFreplay.com. We will continue to add any ETF that trades at least some level of volume. For what its worth, ETFreplay already covers >98% of overall ETF/ETN assets and >98% of the volume traded. The bottom 400 ETFs (some of which we do include if they are interesting new products) make up about 1% of overall ETF/ETN industry assets. There are plenty of choices already. Other than the handful of innovative new ETFs that come out every 6-12 months, focus your efforts on the existing liquid ones.