How To Calculate Correlation for Stocks, Bonds and Funds

[@NitinSaiJ]

Awesome Explanation!

[@Pensioncraft]

Thank you!

[@RamonBalthazar]

Man, thanks for this video. So good! 🙏

[@Pensioncraft]

Glad you enjoyed it Ramon Balthazar

[@simeon-1383]

Top notch video, thank you!

[@Pensioncraft]

Glad you liked it SimOneYouMayKnow

[@valeriy07]

Great video, thanks a lot for sharing all these materials! If possible, I'd love to see more correlations rollovers for very commons ETFs and mutual funds, eg vanguard's funds

[@FxBrawler]

Very useful, thank you.

[@Pensioncraft]

I'm glad it was helpful! Thank you for watching. Ramin

[@sierraecho884]

Good video, I was on the fence about the correlation between Gold and Stocks. I always thought it is a good idea to own gold for diversification purposes but this shows it is not as great I thought. However gold has a better performance than Bonds especially the ultra safe gov. bonds. So I might keep my gold after all but diversify into Bonds as well.

[@vladk9152]

shouldn't we use log returns for stocks?

[@Pensioncraft]

Hi Jorge, unless returns are really large there's not much difference between log returns and geometric returns. But you're right log returns are commonly used and given the choice I prefer them but geometric is more common. Thanks, Ramin.

[@alessandroanderlucci]

It's possible to let the sheet update automatically? Or how can I do it?

[@x2x3456]

Do you ever look at correlations using data that's over a longer period of time? I.e annual returns?

[@Pensioncraft]

Hi Nathanial, I don't because that requires much, much longer time series i.e. to get a reasonable number of data points you would need decades of data which simply isn't available for most funds. Thanks, Ramin.

[@joseph354]

How many is the minimum data points do we need to calculate the correlation?

[@cliffkwok]

Great video. I have one question not sure anyone can answer, I am a bit confuse on rolling concept. First, we calculate the 12 month correlation and we rolling month by month to get the curve. I am confuse what we are comparing is the last year to this year or what?

[@Pensioncraft]

Hi Kwok, the rolling correlation "window" doesn't do any comparison. It just calculates correlation over, say, 2000-01-01 to 2001-01-01, then the window moves on and calculates form 2000-01-02 to 2001-01-02, then 2000-01-03 to 2001-01-03... That would be a one-year rolling window. Thanks, Ramin.

[@cliffkwok]

PensionCraft totally clear now! Very professional !!

[@Rafaelkenjinagao]

Very nice video! I liked specially how you presented an interesting idea in such a simple and coherent way. 7:59 which software or library did you use for this graph?

[@MRAROCKERDUDE]

Is there a way to view correlation change over time. In other words, you have used historical data to determine that the movement of the pound and the FTSE 100 aren't correlated. Would be interesting to see if this correlation has altered over time. How would you view that?

[@MRAROCKERDUDE]

In other words, is it possible that the correlation has become more positive over time? Wondering if this might have occurred as if companies on the FTSE 100 might have been earning more overseas income as time goes on. That's not a fact, just a theory.

[@Pensioncraft]

Hi MRAROCKERDUDE, the spreadsheet has "rolling correlation" which is precisely what you describe. The correlation between the FTSE 100 and sterling is very unstable and inconsistent. The German DAX index and oil are much more reliably positively correlated with the FTSE 100. This is 13:07 into the video: kzbin.info/www/bejne/jYOqnGuGd6eZaNE Thanks, Ramin.

[@conduit242]

Great video again! Quick question: - Are you doing an adjustment to daily percentage change returns for the price time series and then normalizing to an incremental z-score?

[@conduit242]

One additional comment, I note that the market has a very Laplacian behavior when you look at those crazy tail returns. It’s likely reasonable to regularize various estimators along those lines.

[@Pensioncraft]

Hi Rob, you mean use shrinkage? If I was worried about the covariance being positive definite I would do that, but to introduce people to correlation I wanted to keep it simple. But equity returns are very leptokurtic as you say.

[@Pensioncraft]

Hi Rob, it's not incremental in the spreadsheet and there's no return adjustment. I do include rolling correlation just to demonstrate that correlation depends on market conditions.

[@conduit242]

PensionCraft oh I wasn’t judging! I’ve just found that an incremental z-score tends to scrub out any remaining testy root dynamics driving a daily percentage change time series off a zero mean. I thought you were doubling up too 😁.

[@conduit242]

PensionCraft I tend to be a Bayesian (eek) but yep, same thing. There’s an interesting article on Quantopia about using the OAS shrinkage estimator for hierarchical risk parity allocation estimation that gives some lovely improvements out of sample. Understand this is not what you’re targeting here, was literally up experimenting with distance correlation vs linear correlation after noting how limply it correlates leveraged versions of two etfs (eg spy vs upro).

[@usamashoaib3696]

Where can I find such data for other funds?

[@Pensioncraft]

Hu usama, there's a function in Google Sheets called GOOGLEFINANCE() which allows you to get time series for many stocks and funds, documentation here support.google.com/docs/answer/3093281?hl=en It doesn't have everything but it's pretty good. I pay for ShareScope (my affiliate link is sharescope.co.uk/pensioncraft ) which gives me stock, bond, fund and currency data which I use for my videos and discussions about markets with PensionCraft clients. Thanks, Ramin.

[@usamashoaib3696]

@@Pensioncraft Thank you

[@MrZweene]

I would be more interested in a correlation between central bank interest and inflation. I fell in love with Ray Dalio all weather portfolio that takes this into account.

[@iandrake4683]

This calculation of returns seems to ignore total returns. Am I missing something?

[@Pensioncraft]

Hi Ian in the UK if you use Accumulation funds their price is the total return so that might be an easy way to do the calculation. But I think you'll find that the correlation of total return and capital return are _very_ similar if you're using daily returns. That's because income varies slowly and so doesn't show up much in daily return. Thanks, Ramin.

[@tc9634]

This is why it's so stupid when people don't hold stocks because of a fear of a crash. "Oh, so you only hold bonds then?" "Yeah, it's safer." "If you put 20-30% of your portfolio in stocks, your would get a more stable portfolio with a higher long-term return." "..."

[@Pensioncraft]

Hi Tim, I wrote a little Monte Carlo simulator for a client to explore the probability of hitting a return target assuming future returns resemble those in the past. For higher return goals, the probability of reaching the goal drops to zero for bond funds because of their limited upside no matter how long the investment horizon. However, speaking to "real" investors I can see how fear can drive people to take too little risk even when they have a very long lifetime of investment ahead of them. Thanks, Ramin.

[@iamonline3221]

there are many type of bonds, I'm one of those stupit people who only posses bonds nowadays, how you doin' lately ?