Asset Correlation ... and diversifying

with help from the Diehards
Everybuddy talks about diversification and its efficacy when the assets have low (or negative) correlation and ...
>Yeah, so? Haven't you done this before?
Well, yes,
here and
here and
here and
here and ...
>Okay! So why yet another blurb on correlations?
I thought it'd be interesting to pick a few mutual funds which (hopefully) have low correlation and ...
>And see what happens, right?
Exactly. Anyway, I decided to use Vanguard funds ('cause there's lots of them) and you pick four and download the data and
you get the correlations between them and you decide on some allocations and ...
>I assume there's a spreadsheet, so why not just show it?
It looks like this:
To try the spreadsheet, you can Click on the picture
or to download a copy, RIGHTclick on the picture and Save Target.
>And how do you download the fund data?
There's a sheet which looks like so ... but there's gotta be eight years worth of monthly prices at Yahoo :
>Do they have to be Vanguard funds?
No. So long as you have a Yahoo symbol and there's eight years worth of data. **
>And there's a moneyback guarantee on the accuracy?
Of the spreadsheet? Of course. Always.
** I started off with ten years, but that eliminated too many funds
When you select allocations which you think are "optimal" ( in some selfdetermined sense ... and based upon past performance :^)
then there's a button which you can press and 96 monthly returns will be selected at random
(from the 96 months of historical data).
>96 months? That's 8 years.
Clever.
The final value of a $1.00 Portfolio is calculated from these 96 randomly selected monthly returns.
(A random month is selected and all four returns are selected from that month.)
This ritual is repeated a thousand times ( a la Monte Carlo :^)
and the Average final Portfolio is displayed, like so:
which shows an Average Portfolio of $1.83 (corresponding to an annualized return of 7.9%)
>What good is that?
Well, it shows you that, after you've selected what you think is the optimal allocation based upon the past eight years of monthly returns,
you might be surprised at what'd happen if random selections were made from those same set of returns ... for the next eight years
>Is that like: "Don't expect the future to be like the past" ?
Something like that.
Oh, I should also mention that you get the distribution of the 1000 Monte Carlo simulations.
It shows the percentage of those 1000 "final portfolio" simulations that end with less than $P
... as well as the Average Final Portfolio.
>The Average is the red dot?
Yes, the red dot.


>And a third of the final portfolios are less than $1.50?
$1.50? Yeah. That'd be about a 3.3% return.
>But that's just for a certain allocation, right?
Yes, so change the allocation and try again.
Drag your mouse from X to O
X
Diversification is a protection against ignorance
... Warren Buffett
O

motivated by stuff of Rick Ferri
When we look at the correlation of monthly returns over the past umpteen years, we don't realize that they can change from month to month.
>We don't realize? Speak for yourself!
Anyway, the above spreadsheet downloads eight years of monthly returns and calculates the correlation over all eight years.
I modified it so that you can see how these change as the months go by. The modified spreadsheet now calculates the correlation of
monthly returns over 36 months ... and shows how that changes as the years go by and ...
>Picture?
Yes. There's a new button
You click and get a picture, like so:
>And I get to choose the assets?
Well, any four (for which Yahoo has eight years of monthly prices).
Then you get six moving correlation graphs.
>When I click the button.
Yes.
>How about stocks?
Sure, you can do that:
>But that's only eight years worth, but what about ...?
Well, using the data here, looking at ten years worth of annual returns, we can follow the correlations and get, for example:
>Yeah, but what about ...?
We just wanted to illustrate the volatility of correlations (as Rick Ferri did). We've done that. It's enough. Now go back to sleep.
>zzzZZZ
Note: The spreadsheet (like the moving correlation) may change abruptly ... even if'n this tutorial don't ...
