Miguel and Mykola,
Thanks for the nice writeup. After checking out the spreadsheet that was provided I have a question about the correlation calculation.
Since my version of the spreadsheet does not display the formulas (only the results) of the “Correlations” tab I was unable to see how the comparison was made for ETFs that had differing start points. For example: SIL only goes back to 2010, but GLD goes back to 2006. Mean returns will vary – it seems – if:
1) the empty rows in SIL are included back to GLD’s start point,
2) GLD is averaged back to 2006, but SIL is averaged back to 2010
3) GLD is only averaged as far back as there is data for SIL.
Could you please comment more on how these differences should be handled?
Hi tuckandturn, thats actually a very relevant question . When you compute correlation in Excel, the two assets need to have the same number of values so if there are no values for one of the assets, it will just ignore those cells and the corresponding cells for the other assets. For the GLD vs. SIL example, the correlation will be computed only up until 2010 and all GLD data prior to 2010 will be ignored. In other words, the most recent date of the two is used.
This also means that some correlations on the table are based on 10 years of data while others, involving more recent assets such as the junior gold minors or the Israel ETFs, are based on much less data.
Also, the reason it does not display the formulas is that I used the correlation tool in the data analysis add-in on Excel. It is a much quicker way to build a correlation matrix. The data analysis tool also helps you build a variance covariance matrix which is helpful in portfolio optimization.
I hope this answers your question, feel free to ask if you have other questions.
Financial Analyst Warrior
- You must be logged in to reply to this topic.