Minimising correlation

[Newroad]

Hi All.

Does anybody know of a convenient way to calculate and rank correlation between listed financial instruments? This could be done at a fundamental level (common holdings and their weightings) but I'm guessing would more likely be at a proxy level, e.g. based on prior price movements. Though I'm sure there are other applications one might apply this to, my specific reason is that I want to compare VWRL to a selection of Investment Trusts - with a possible view to picking the one with the LEAST correlation.

To expand slightly, there are six relevant portfolios in my family, namely

My ISA
My SIPP
Spouse's ISA
Spouse's SIPP
Child 1 JISA
Child 2 JISA

The first four are constructed in a A/B/C/D 30%/30%/20%/20% structure, the last two A/B/C/D/E 35%/35%/15%/15%/?% where

A is a Global Investment Trust
B is VWRL
C is a Global Debt (Loans and Bonds) Investment Trust
D is VAGP
E is MYI (the exact percentage varies as it is a grand-parental contribution, but for purposes of discussion, you can assume it is around 10% in addition to the 100% total of the rest)

For the SIPP's (ATST) and the JISA's (FCIT), I'm happy enough with the A's. The others for historic reasons are WTAN (the other two A's) and MYI (the E's) - I am aware the latter is Global Equity Income. For these, I am looking to switch when the time is right, and am considering a pool of high rated (by Morningstar) Global Investment Trusts, ideally with weak(er) correlation to VWRL. This pool is currently MNP, BNKR, MNKS.

You might reasonably ask why not LTI and/or SMT as well - in short, they don't "feel right" for me - but I may change my mind on these - they might come under consideration for the E's in particular. You might also reasonably ask why not some other *** rated trusts if I'm OK with ATST. The short answer is, I may look at these as well if one of the initial three don't seem to fit. However, my intent is not to make this post a voting thread on any of the above - just trying to put some system and impartiality into an impending decision. Finally, I may just keep WTAN and/or MYI for a while until the situation becomes clear in my mind.

So, anyone able to help on the correlation bit at the top?

Regards, Newroad

[Adamski]

Morning, you could try Morningstar portfolio tool. I find nearly everything moves in tandem, except China, Japan stocks, gold etf, and higher risk such as SMT less correlated, so if want diversified pf need some of thess.

[Newroad]

Thanks, Adamski.

I am happy to use the Morningstar tool, but I didn't realise it had this sort of functionality?

On your broader points, I agree that my three proposals (and a host of other possibles) will be correlated with VWRL, probably quite strongly, i.e. have a correlation coefficient of greater than 0 and probably somewhere between 0.75 and 1 - the idea is to identify the least so (if I'm lucky, perhaps 0.5 to 0.75). I do not wish to explore other asset classes and/or go sector specific myself - though I am happy for a Global Trust manager to do so on my behalf.

Finally, I would expect SMT to be more rather than less correlated with VWRL than, say BNKR, but will happily stand corrected if I figure out a way to calculate the numbers and that proves not to be the case. This is not to say SMT and VWRL have or will perform the same, but rather, in a simplistic sense, that they are more likely to move in the same direction at the same time.

Regards, Newroad

[jonesa1]

This would be ideal, if you could find some way of getting it to analyse the funds you're interested in. I've not been able to find a way to change them.

https://www2.trustnet.com/Tools/Compari ... ACDT,FGMTX

Trustnet provide quite a lot of data for a portfolio, but doesn't appear to include correlation. A decent X-ray tool is what you need, but I'm not aware of one that can be used stand-alone. I hold BNKR & MNKS, the AJ Bell X-ray shows them as having a correlation of 0.90, I don't hold the other funds you're interested in. Maybe people who do hold them could tell you how they correlate based on X-rays of their portfolios?

[Itsallaguess]

This would be ideal, if you could find some way of getting it to analyse the funds you're interested in. I've not been able to find a way to change them.

https://www2.trustnet.com/Tools/ComparisonReport.aspx?typeCodesCF=,FMGAM,FACDV,FACDT,FGMTX

The compared funds can be changed by altering the final four codes in the above URL, but please note that we need to keep the initial additional 'F' before each code...

Using the spreadsheet linked below, we can see where the current URL code details can be found for your four initially-compared funds -

https://www.oldmutualwealth.co.uk/globalassets/documents/literature-library/funds-list-for-investments-in-selfselect1.xlsx

FMGAM,FACDV,FACDT,FGMTX -





I've just tried it with four new random funds to check compliance with the spreadsheet data -

FCCD6,FACDQ,FLSX2,FSPSEA -





URL for the above four fund comparison - https://www2.trustnet.com/Tools/ComparisonReport.aspx?typeCodesCF=,FCCD6,FACDQ,FLSX2,FSPSEA

Again, note the requirement for an additional 'F' at the front of each fund code...

I'm not quite sure on why some funds need the MexID code and others need the FE Citicode, but given that it's a quick swap to check the alternative if something doesn't work, then hopefully there's enough in there to have a useful play with...

Cheers,

Itsallaguess

[Newroad]

Hi ItsAllAGuess.

Please correct me if wrong, but the spreadsheet contains "funds" (i.e. OEICS and similar) rather than ETF's and Investment Trusts? Assuming so, do you know if and where it is possible to source similar information for the latter two types?

{JonesA1} Thanks for the reply. With II (rather than AJ Bell) I also get access to (Morningstar's ?) X-Ray tool. There are two things which prevent me from using it to answer my question: {1} I can't seem to create a "test" portfolio to X-Ray, and {2} With my existing portfolios, I can get an "Overlap" report, where it states the common holdings and their percentages in each, e.g. in VWRL and WTAN, but not a correlation figure.

It's perhaps a question for fellow II users rather than you, but if anyone thinks the functionality is there and I am simply overlooking it, please advise how to see it as specifically as practical?

Regards, Newroad

[Newroad]

Hmm.

I did it the hard way. Downloaded monthly prices for VWRL, BNKR, MNKS, MNP and for good measure, SMT from Yahoo Finance. I did it for June 2012 (first month after VWRL became available in London) to Feb 2020 (for some reason, BNKR data wasn't complete after that).

The results are in, correlation of VWRL vs ...

BNKR: 0.246
MNKS: 0.224
MNP: -0.140 (yes, negative)
SMT: 0.225

We appear to have a winner! I'm also surprised about the lack of correlation in general - as you know, I was expecting it to be quite high.

I have a follow up question, unrelated to the maths above. Martin Currie manages both MNP and STS, but is losing the mandate of the latter to Troy Asset Management I understand, despite it being quite well rated (I was considering it as a replacement for MYI). Anyone know why - and whether it is relevant to a decision about investing in MNP?

Regards, Newroad

[Itsallaguess]

Please correct me if wrong, but the spreadsheet contains "funds" (i.e. OEICS and similar) rather than ETF's and Investment Trusts?

Assuming so, do you know if and where it is possible to source similar information for the latter two types?

I think the nearest thing I've found to the earlier Trustnet tool that can compare against those different types is this one from Hargreaves Lansdown, which allows you to stack up to 7 different elements and then plots 1, 3, or 5-year price or total-return performance comparisons against the selections.

Here's an example 5-year total-return chart from the tool, showing your BNKR, MNKS, MNP, SMT selection -



Source - https://www.hl.co.uk/funds/fund-discounts,-prices--and--factsheets/search-results/f/fundsmith-equity-class-i-accumulation/charts

I don't think it's as useful to you for correlation comparisons, but if you've not seen it before then I hope it's handy for an additional bookmark for future reference at least..

For anyone interested in digging a little deeper in that HL tool, the chart URL's themselves can be manipulated in a similar way to my earlier Trustnet post, but this time benefiting from using much simpler EPIC entries in the URL text, as can be seen using the following chart-URL from the above page, but where here we can clearly see the four BNKR, MNKS, MNP, SMT EPIC's being used (each with a leading 'E'...) to build the specific 5-year total-return chart -

https://webfund6.financialexpress.net/clients/Hargreaves/chartbuilder.aspx?codes=EBNKR,EMNKS,EMNP,ESMT&color=1a83f6,efd715,53e166,8aa3d8&hide=&span=60&totalReturn=true

Enjoy!

Cheers,

Itsallaguess

[Newroad]

Thanks, ItsAllAGuess.

Based on what I've observed, I'm now considering replacing WTAN with MNP in the two A's (i.e. the ISA's) and MYI with SMT in the two E's (the extra grand-parental 10% in the JISA's). They are two of the three ***** rated options from Morningstar, MNP is nicely uncorrelated with VWRL it seems, and the kids can take some more risk with SMT.

Will mull over it for a short while though still.

Regards, Newroad

[monabri]

I don't know what you mean by "correlation " ...is it not the case you have 2 assets classes (shares in VWRL, MYI, Global IT) and bonds.

[Newroad]

Hi Monabri.

If you look above, I explained what I did - a quite literal version of correlation (coefficient). To reiterate and elaborate slightly ...

Downloaded June 2012 to Feb 2020 inclusive start of month adjusted closing prices for VWRL, BNKR, MNKS, MNP and SMT
Imported them into separate tabs in Excel
For the latter four, ran the Excel "CORREL" function against VWRL for each

Bonds were not part of this anaylsis - though I may do the same for VAGP with respect to CMHY, IPE and HDIV (not that I'm currently thinking of changing any of these). It would purely be for interest's sake.

Regards, Newroad

PS Not sure about the location of that apostrophe in "interest's"

[Newroad]

STOP PRESS

Just had a closer look at the data - it appears the Yahoo Finance data is dodgy - not just the BNKR's stuff this year.

Please ignore the alleged correlations stated earlier - they are likely rubbish

[Newroad]

Hi All.

The good news, if there is any, is that my original thoughts proved accurate - VWRL is highly correlated with all the Global Investment Trusts.

BNKR: 0.840
MNKS: 0.825
MNP: 0.842
SMT: 0.843

The bad news, of course, is that I now have no clear "winner".

I had to manually correct the data, in particular for VWRL and MNP (mainly, some was out by two decimal points) and may have made the odd typo, but I believe it's pretty close to right now, if not exactly so.

Regards, Newroad

[Newroad]

Hi All.

I've now done some more correlations for bond investment trusts with both VWRL and VAGP.

With VWRL (Jun 2019 to Oct 2020 inclusive, VAGP only started then in London)

VAGP: - 0.111

With VWRL (Jun 2012 to Oct 2020 inclusive)

CMHY: 0.797
HDIV: 0.659
IPE: 0.788

With VAGP (Jun 2019 to Oct 2020 inclusive)

CMHY: 0.002
HDIV: 0.253
IPE: -0.238

Once again, some data anomalies have been hand corrected.

Regards, Newroad

[monabri]

Have a shufty at the pensioncraft videos....they are very informative ( well, I found them to be so).


https://youtu.be/tHvlU1WGjpU

[Newroad]

Thanks, Monabri.

I had a look - informative enough as you say.

From his examples, he "picked" VWRL and VIGBBD as his two ETF portfolio (ratios adjusted to suit the investor's risk tolerance). I can't find VIGBBD now - I assume that pre-dates VAGP. If so, what he describes is exactly what I've done for 50% of my portfolio - the passive side.

As you may be able to recall, I mirror that with the equivalent active side - so mine is a 4 product portfolio. However, it seems to me to be very in keeping with his basic premise of simplicity - a Pareto Rule like approach.

Regards, Newroad