A Framework For Choosing ETFs

The Ideal ETF

Most investors choose an ETF to act on a market view — gold will rise, U.S. equities will fall, and so on. ETFs are great instruments for implementing these market views because they allow for exposure to a group of securities with the convenience of being able to trade these as a single stock. But at the time of this writing, there are over 1,900 ETFs with around 200 new ETFs that are created and 100 existing ETFs that close each year.

Given the number of ETFs and its changing landscape, it wasn’t straightforward to me how I should choose ETFs to implement my market views, so I started looking for a systematic approach.

Drawing from first principles, the ideal ETF is an ETF with a return stream that is identical to the return stream of the market segment it represents. In the real world, these ideal ETFs do not exist due to two reasons: costs and tracking error. Therefore, investors should choose ETFs that minimize costs and minimize tracking error.

Minimize Costs

Costs come in the form of fund management expenses, trading costs, and taxes.

Expense Ratio

The expense ratio represents the annual percentage that the fund management company charges investors. Most ETFs have an expense ratio between 0% and 1%, but there is considerable variation. Larger funds generally charge lower expense ratios because fees can be spread over a larger base. Some funds over fee reimbursements for a period of time while they are accumulating assets — this is temporary and the fund management company will end this after an introductory period. Lower expense ratios are better.


Trading Costs

Trading costs refer to the costs faced when an investor buys or sells the ETF. The most direct trading cost is the bid-ask spread. Think of this as the cost for one roundtrip trade (buying and then selling). Most ETFs are easily traded with a bid-ask spread between 0% and 0.25%, but again there is considerable variation. It’s my opinion that a lot of investors place a lot of emphasis on the expense ratio, but trading costs can be as or more important. For example, two roundtrip trades on a ETF with a 0.25% bid-ask spread equal 0.50% which is equivalent to a moderately-high expense ratio. If you plan on trading often, look at trading costs carefully. Lower bid-ask spreads are better. 



Taxes are another commonly overlooked cost, but they can matter for retail investors holding ETFs in a taxable account. I won’t spend too much time on taxes since everyone’s situation is different, but certain funds are better than others at not distributing capital gains. You don’t want an ETF that distributes capital gains because you have to pay taxes on them. You don’t want an ETF that distributes its cash flows as ordinary income. There are also five different ETF fund structures and certain fund structures combined with certain assets are tax inefficient.

Minimize Tracking Error

ETFs that track an index can be managed in one of two ways: full replication or optimization. Sometimes ETFs will just buy every single security in the index that it tracks — this is full replication. Sometimes ETFs will instead buy a representative sample of the index or use derivatives in an attempt to track the index — this is optimization.

Both approaches require a good amount of skill. Some people think that the full replication approach is really simple and that a computer just takes care of everything. The reality is more nuanced. There are always special situations and edge cases that come up when it comes to index construction — dealing with minor changes in methodology or index constituents, mergers and acquisitions, trying to enter and exit positions while minimizing costs, and skillfully lending out securities for a small extra return.

For optimization, skill plays a larger role. Fund managers go the optimization route in the first place because its not feasible to do full replication for certain securities (like fixed income). Investors want to minimize tracking error because this allows the return stream of the ETF to be more predictable. An ETF that has tracked its index closely in the past is indicative of a well-managed fund. 

There are a variety of ways to measure tracking error. You could calculate the correlation between the returns of the index and the returns of the ETF. You could also calculate the mean difference in daily returns over some time period.

How to Choose ETFs

Generally, choose an ETF with a low expense ratio, low trading costs, low tracking error, and a tax-advantaged structure. Depending on what kind of investor you are, you can weigh each of these elements differently. A long-term investor should heavily weight a low expense ratio and perhaps not care too much about trading costs. On the other hand, an investor that trades often should heavily weight trading costs. At the extreme, an investor with no long or short bias should only care about trading costs and not the expense ratio because a high expense ratio is good if you are shorting the ETF.

In my next post, I will provide what I think are the optimal ETFs for each market segment. If you enjoyed reading this post, please consider following Signal Plot via email.


I am interested in building investing systems, and this blog contains my research and analysis on this topic. I previously worked as an analyst at Bridgewater Associates, a hedge fund that utilizes a systematic, global macro investing style.

  1. so helpful! do you work at wall street?

    • Hey Xiao Bao. I used to work at an investment management firm. Now I work at home on various projects like this blog. Glad you found it helpful.

    • I agree with Xiao Bao, this has been very helpful and informative. Keep them coming!

Leave a Reply

Your email address will not be published. Required fields are marked *