Choosing ETF Histories for Backtests Without Survivorship Bias
Summary
The document considers how to assemble an ETF universe for historical portfolio-allocation backtests. One response suggests using a finance screener to sort ETFs by inception date, but cautions that requiring funds to exist by a chosen early date can sharply restrict the universe and exclude relevant exposures. It recommends defining the indices or asset classes the strategy needs before selecting matching ETFs.
The discussion distinguishes ETF inception history from the history of the underlying index. It argues that index ETFs may avoid some forms of survivorship bias seen in single-stock tests, while emphasizing that researchers must judge whether an index would plausibly have been considered among the popular choices at the historical date. Examples contrast established US and international benchmarks with a country exposure that became prominent later. Other replies disagree about whether a comprehensive free database exists and mention screening services. A final reply reports a separate study using a weekly unemployment-claims moving average as a trading signal, but the document gives no methodology or detail sufficient to evaluate that result.
Key ideas
- Define the asset classes and indices for a backtest before selecting ETFs by inception date.
- A strict ETF-age cutoff can exclude useful exposures and reduce the historical universe.
- ETF inception dates do not by themselves establish whether an underlying index was historically plausible.
- Index ETF tests still require careful consideration of survivorship and historical selection bias.
- The database suggestions vary, and the document does not establish that a complete free source is available.
Tags
Full text
# Where can I find a database of ALL ETFs, sorted by age? # Where can I find a database of ALL ETFs, sorted by age? I have a portfolio allocation strategy I want to backtest, but I need a large "universe" of ETFs for it to choose from at each time period. I was thinking of starting with a criteria such as "all ETFs available in 2000" or something like that to ensure sufficient data for a backtest. Any suggestions as to where I can find a databse of all the ETFs out there, complete with their inception date? ## Answer by Tal Fishman (score 1, accepted) https://quant.stackexchange.com/a/1818 Yahoo! Finance allows you to list all ETFs and sort by inception date or size. Considering your follow-up question, though, I think you are asking the wrong question here. ETFs are a relatively recent innovation. Only 32 of these ETFs were around in 2000. For example, the Russell 2000 ETF (IWM), today's third most actively traded ETF overall, was only introduced in May 2000, thus missing your cutoff date. The CBOE only started listing options on SPY in 2006, and now it is one of the most liquid and popular options series. If your real goal is to backtest an ETF strategy, then think first about what indices/asset classes belong in your backtest then see what ETFs match those choices best. Alternatively, start with what is most actively traded today. Unlike single stocks, a backtest involving index ETFs is not subject to backfill bias (a.k.a. survivorship bias) in quite the same way. Instead, you'll have to think hard about whether the index on which the ETF is based could reasonably have been considered one of a handful of popular indices as of your backtest date. Russell 2000 (EF: IWM) and MSCI EAFE (ETF: EFA) can reasonably be considered survivor-bias free index choices as of at least 1990, but Brazil (ETF: EWZ) wasn't on most people's radar until some time in the 2000s. This is despite the fact that their ETFs were all created around the same time. ## Answer by SRKX (score 0) https://quant.stackexchange.com/a/1809 I seriously don't think such a thing exists. I recently saw an ETF specialist who was working along with his team to standardize their own database in order to be able to sell its access to the public. It's a question I've heard several times and I seriously don't think you'll find that for free anywhere. ## Answer by user6546 (score 0) https://quant.stackexchange.com/a/9489 http://funds.ft.com/us/Screener/PreScreen can screen by inception date, i.e. <1 yr, 1-3, 3-5, 5-10, >10 years. ## Answer by AZman (score 0) https://quant.stackexchange.com/a/21116 See www.nanlee.net for a book doing what you are trying to do. It scientifically tested 26 older ATFs and found that the average annual return doubled using a moving average of weekly unemployment claims as a buy and sell indicator. A summary of the results is in the Fall, 2015 issue of the Journal of Index Investing.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.