I am always trying to tear down the PP. I like to think of myself as a pretty smart guy, and that damn four split just looks so stupid. When I was messing around with ETF replay I found something that could have indicated a serious problem.
When you look at the last 3 years the portfolio has a 0.84 correlation with gold, 0.24 correlation with LTT, and 0.33 correlation with stocks. This indicates that the portfolios performance is largely driven by gold. In a perfectly balanced portfolio, the correlation to each of the underlying assets would be the same, meaning that each asset was playing an equal role in determining performance. Maybe the 25% gold allocation is seriously overdoing it.
However, that test was for the last three years. Has gold always explained the Permanent Portfolios returns? I did a new test using some daily data that I synthesized (which distributes inflation, dividends, down to the daily level) which I talk about here: http://www.stableinvesting.com/2012/10/ ... raphs.html
Here are the underlying correlations to the PP using trailing 300 days...
As you can see, lately gold has been what is driving the PP. But you can also see that this has been anything but consistent over time. If gold was consistently the highest correlation asset I would consider under-weighting it, but that isn't the case. The Permanent Portfolio has different asset classes dominate at different times, depending on complex changes in expectations regarding changes in macro regimes.
This damn portfolio has beat me again, so I guess I will keep the four way split.
everything comes from somewhere and everything goes somewhere
melveyr wrote:
I like to think of myself as a pretty smart guy, and that damn four split just looks so stupid.
I run into this a lot.
It bothers people that something could be so coloring-book-simple, seem so counterintuitive and yet apparently work so well.
When talking with people about the PP, a common response I get is something to the effect of: "Well, I know this guy, he's REALLY smart, and he works with a lot of other really smart guys/quants/analysts/gurus/etc., and they pretty much know what the market is going to do before anyone else. I may run this by him and see what he thinks. I'll let you know what he says."
So far, I've never heard any reports about what these smart guys think of the portfolio.
Another thing that I've run into (and this honestly surprised me) is that sophisticated money managers sometimes don't want to talk about the PP because it has performed so much better than whatever they have been using and they don't want to look like they weren't able to outperform such a simple-looking allocation. It hadn't occurred to me that money managers would simply ignore performance because it was too challenging to their worldview of what is supposed to work.
There is more dogma in the minds of many investment advisors than I think most investors realize. Many of these people are like gamblers whose minds are only open to new gambling strategies, but whose minds are completely closed to a mental framework that might allow them to move to the other side of the gaming table and begin enjoying an actual statistical house advantage.
Q: “Do you have funny shaped balloons?”
A: “Not unless round is funny.”
Slotine wrote:
Actually, melveyr, I think it does tell you something. Look at the 79-87 range for gold. The correlation graph hides the sign and magnitude of the actual delta.
Maybe use the indicators corresponding the two quadrant axes as your factors? You'll get the non-varying factor coefficients separated from the time-varying ones. If the four quadrant model holds well then you should be able to see it in the stable coefficients.
Edit: I had forgotten that the GestaltU guys did do a simple factor analysis over a three year range for PRPFX. Shakedown Article They went directly for the underlying components, which works just as well it looks like.
Perhaps I am interpreting this wrong, but doing a multiple regression analysis in the PP would simply lead to 0.25 performance attribution for each asset class, correct? That is what I got when I just did it. I think they are trying to isolate alpha for PRPFX in that article, but we already know alpha is 0 for the PP because we are using indices.
everything comes from somewhere and everything goes somewhere
I think we can take into account the magnitude by using covariance. Is this more instructive?
(daily numbers so the units tiny!)
You can really see the gold bubble, the 1987 crash, uptick in turbulence of equities during the tech bubble and our most recent crash!
Kind of looks like LTT have never really had their time to shine compared to the others (justifying a higher weighting). Also makes me wonder if gold allocations should be reduced if a speculative fervor makes its volatility enormous. Volatility is somewhat sticky; Mandelbrot correctly identified that this is one of the many flaws with EMH.
Last edited by melveyr on Wed Nov 14, 2012 9:46 pm, edited 1 time in total.
everything comes from somewhere and everything goes somewhere
Perhaps I am interpreting this wrong, but doing a multiple regression analysis in the PP would simply lead to 0.25 performance attribution for each asset class, correct? That is what I got when I just did it. I think they are trying to isolate alpha for PRPFX in that article, but we already know alpha is 0 for the PP because we are using indices.
Ya, that's right. My bad. We really need to use factors that are independent of construction.
Regarding the covariance graph, is volatility that bad really? We're typically measuring portfolio performance against our currency, not something real. Hence volatility in the currency itself would simply manifest itself as additional volatility in a portfolio like the PP that has some 'outside' peg.
This probably makes the 80's gold bubble look worse than it really is. I still don't like it, but I'd be cautious to do anything to upset the balance.
The daily returns underneath these charts are adjusted for inflation. I took the monthly rates and applied them on a daily level (taking into account TVM so month end values tie out).
everything comes from somewhere and everything goes somewhere
melveyr wrote:
This damn portfolio has beat me again, so I guess I will keep the four way split.
1) Weight them according to their percentage of historical occurences. I did a simple real returns optimization mentioned in another thread and the weights came very close to that.
2) Use tactical allocation.
I do #2, but haven't had enough proof-satisfaction that #1 is also worth implementing.
"All generous minds have a horror of what are commonly called 'Facts'. They are the brute beasts of the intellectual domain." -- Thomas Hobbes
Disclaimer: I am not a broker, dealer, investment advisor, physician, theologian or prophet. I should not be considered as legally permitted to render such advice!
Interestingly enough, I've had this exact thought earlier this morning with a concern that I might want to raise my VP relative to my PP so that I can underweight gold in my overall portfolio.
Then I thought about hyperinflation. LTTs will drop like a brick. Stocks will fail because the economy will be failing. Cash may only lose a little to hyperinflation, after-tax, assuming interest rates keep up.
However gold should soar. And if so, I'll need a 25% allocation to gold to cover the 25% of my assets (LTT) that are taking heavy losses, and the 25% of my assets that are taking moderate losses (equities) and the 25% of my assets that are taking light losses (cash).
Additionally, suppose I wasn't 25% gold. What would that money be in? Either more cash, more equities, or more LTTs. All 3 which would fail during hyperinflation to various degrees as described above. So if I go below 25% gold then not only am I taking away from gold's ability to protect the other 75% of the portfolio, but I make the rest of the portfolio (the non-gold part) bigger than 75% which means it needs more protecting from hyperinflation but I'll have less gold to do it with.
Thus, I think 25% gold is prudent. I think the original post is more about finding a more "correct" and efficient way to split assets rather than 25x4 because how can that possibly be the "best" way that just happens to be random, especially since everyone will choose slightly different ways to invest in the underlying assets that shift returns by 1% to 5% depending on what they picked (i.e. a zero coupon bond versus a 30 year LTT versus a 20 year LTT may have a several percent spread in returns).
And the answer to this is that you won't be 4x25 for long. Things will move around and you only rebalance when it hits 15 or 35. So maybe 20, 30, 28,22 is the most efficient split. Well, you may be there at some point through random movements.
And let's say that we could lock down 21, 29, 28, 22 as the most efficient split looking backwards. That doesn't mean it will work the best going forwards.
And let's say that we could lock down 21, 29, 28, 22 as the most efficient split looking FORWARDS... that doesn't mean the 4 assets wont move around through daily movements... and will you rebalance back to 21,29,28,22 on a daily basis? The rebalancing trading costs and taxes would be cost prohibitive so your perfect 21,29,28,22 portfolio will look very similar to what a 25,25,25,25 portfolio will look like after a while.
There's also an added benefit of simplicity. While I personally enjoy tweaking and measuring, most people have no idea what the term "metric" means or how to calculate an IRR or even what an IRR is. By saying 25,25,25,25 with 15, 35 rebalance bands, you make it easier to do which saves the average person time. It won't save you time if you have a degree in Finance and Excel is always open in the background, but it might save an average person hours of headaches trying to figure things out every couple of months.
And again, we can't know any deviation from that 25x4 split will really be the best going forward, so making it more complicated for no benefit doesn't make sense.
Certainly HB started off with a significantly more complex PP. Look at the PP mutual fund to see what he started with. Later on he realized such complications were unnecessary and similar returns could be achieved through 4x25.
It is really amazing how much complication you can introduce and have the model still spit out something very close to a PP. I find this with lots of different models that I guinea pig.
everything comes from somewhere and everything goes somewhere
TripleB wrote:
Interestingly enough, I've had this exact thought earlier this morning with a concern that I might want to raise my VP relative to my PP so that I can underweight gold in my overall portfolio.
Then I thought about hyperinflation. LTTs will drop like a brick. Stocks will fail because the economy will be failing. Cash may only lose a little to hyperinflation, after-tax, assuming interest rates keep up.
You make a compelling argument, but I don't think there would be an issue of using the 10% VP to invest in that single asset outperforming relative to the other three while staying in all four in the 90% PP, whether 25%x4 or historical probabilities. In case of hyperinflation, everything goes to pot so having any gold, even if it is less than 25%, would be like comparing a golden delicious apple to a fuji apple. But both are apples and you came out ahead even if the golden delicious is a bit tart. I believe historically the "optimal" allocation to gold was 17% so its not a huge reduction that is going to cause any amount of worry.
At first thought, I like the idea of the banal simplicity of doing 25%4 buy-and-pray and leaving the tactical allocation to the VP alone, but that puts a lot of pressure on the VP to outperform when the PP is taking heavy losses. And what if both the VP asset and the PP asset go down together? Its not easy to deal with having worse loses from being active than if you were passive.
"All generous minds have a horror of what are commonly called 'Facts'. They are the brute beasts of the intellectual domain." -- Thomas Hobbes
Disclaimer: I am not a broker, dealer, investment advisor, physician, theologian or prophet. I should not be considered as legally permitted to render such advice!
