Alternative Permanent Portfolio Update

General Discussion on the Permanent Portfolio Strategy

Moderator: Global Moderator

Post Reply
EdwardjK
Executive Member
Executive Member
Posts: 151
Joined: Mon Oct 11, 2010 8:16 pm

Alternative Permanent Portfolio Update

Post by EdwardjK » Fri Apr 23, 2021 9:36 pm

Greetings all.

A while back I shared a modified version of the Permanent Portfolio using three (3) ETFs - VTI, TLT & GLD. I recently updated the results and want to share them.

For background, here are the rules I use:

The current closing price (CCP) of each ETF is divided by the closing price at the end of the prior quarter (PCP) divided by the PCP. Mathematically, the calculation is (CCP/PCP) / PCP. Make this calculation for all 3 ETFs and then sum the results for the 3 ETFs. Then calculate the percentage that each ETFs result is of the total. For example, if the results of (CCP/PCP) / PCP are 0.25, 0.40 and 0.30, then the percentage for the first ETF is 0.25/.95, or 26%. The second would be .40/.95, or 42% and the third would be 32%. (26% + 42% + 32% = 100%). The balance in the investment account would be allocated to the ETFs using these percentages on the next business day (the first day of the next quarter). The calculations and redistribution would occur again at the end of the next quarter.

Attached are the results in Excel.

Feedback welcome.

Ed
Attachments
Modified Permanent Porfolio.xlsx
(3.77 MiB) Downloaded 236 times
D1984
Executive Member
Executive Member
Posts: 730
Joined: Tue Aug 16, 2011 7:23 pm

Re: Alternative Permanent Portfolio Update

Post by D1984 » Fri Apr 23, 2021 11:44 pm

EdwardjK wrote:
Fri Apr 23, 2021 9:36 pm
Greetings all.

A while back I shared a modified version of the Permanent Portfolio using three (3) ETFs - VTI, TLT & GLD. I recently updated the results and want to share them.

For background, here are the rules I use:

The current closing price (CCP) of each ETF is divided by the closing price at the end of the prior quarter (PCP) divided by the PCP. Mathematically, the calculation is (CCP/PCP) / PCP. Make this calculation for all 3 ETFs and then sum the results for the 3 ETFs. Then calculate the percentage that each ETFs result is of the total. For example, if the results of (CCP/PCP) / PCP are 0.25, 0.40 and 0.30, then the percentage for the first ETF is 0.25/.95, or 26%. The second would be .40/.95, or 42% and the third would be 32%. (26% + 42% + 32% = 100%). The balance in the investment account would be allocated to the ETFs using these percentages on the next business day (the first day of the next quarter). The calculations and redistribution would occur again at the end of the next quarter.

Attached are the results in Excel.

Feedback welcome.

Ed

Quite interesting (and the returns look very juicy!). A couple of questions if you don't mind:

One, I take it you only went back to mid-2005 because that's as far back as GLD goes? Given that we have daily gold prices since early 1968, could you use those plus the expense ratio of GLD divided by 252 (since there are 252 trading days in a year) to simulate a quasi-GLD before 2005 or 2006?

Two, do you even need daily closing price data or would monthly or quarterly close price data work? I have daily data for VUSTX back to late 1986, the M*Star category returns for LTT back to 1-1-85 daily before that, and a daily LTT TR index back to mid-1973 for before that if needed but if we don't need daily data but only monthly/quarterly close then we can go back to 1968 (since that is when gold started and since we have monthly stock and LTT data back to the mid-1920s). Obviously if you did this you would need to norm the values to the earliest close data for the corresponding ETF (for instance, if VUSTX closed at a price of 85 on the day TLT was incepted and TLT closed that same day at a price of 40 then you would need to divide all previous VUSTX values by 2.125 to get a normed value in order to have a continuous unbroken concatenated chain of monthly closing values equivalent to the TLT ones.

Three, do you use split/dividend/cap gains distribution adjusted closing values for this or just closing values?

Four, any idea how this would do in a rising rate environment like 1964 to 1981 (since this portfolio has no cash in it and cash is the asset that actually benefits immediately from rising rates even in years like 2015, 1994, 1990, 1981, and 1969 when all the other assets get hurt by said rising rates)?

Fifth and finally, what is the MaxDD on this portfolio and how long was the recovery time? EDIT: Found this one.....scrolled over and there it was.
EdwardjK
Executive Member
Executive Member
Posts: 151
Joined: Mon Oct 11, 2010 8:16 pm

Re: Alternative Permanent Portfolio Update

Post by EdwardjK » Sat Apr 24, 2021 2:51 pm

D1984,

Thank you for your comments and questions. Here are my replies:

1. Yes, my model starts in 2006 because GLD data is not available prior to Nov. 2004. I wanted to calculate full year returns and the model looks back 63-days. So calculating the 2005 annual return is not possible.

I am not a fan of synthetic historical data. And even if I were, I prefer to use more recent history as being representative of current trends and relationships.

2. I use 63 trading days in the look-back period because some of the quantitative literature suggests that 63-days is representative of a 3-month trading period. It also has the benefit of avoiding the exact end of the quarter where investors and ETFs may be rebalancing their holdings and skewing results.

3. I am using Yahoo Finance "adjusted closing prices". These include dividends and interest. The assumption is that all dividends and interest are reinvested in the underlaying security, rather than holding a cash position. The annual returns will be different, likely a smidge less, if you did not reinvest the cash.

For full disclosure, the model does not take into account transaction costs (near zero nowadays anyway) and taxes (but if you trade in an IRA or 401(k) account, taxes are not an issue until you take a distribution). It also uses the closing price of the first day after quarter end when rebalancing.

Lastly, I am fully aware of data integrity issues with Yahoo Finance. However, I do not possess the skills to write any code for my model using more reliable data. If someone here can help me with that, I would welcome the opportunity.

My approach is more hands-on than the traditional HBPP. I rebalance four times a year whereas the traditional HBPP rebalances only when an asset value falls outside defined bands. However, I don't think rebalancing four times a year is excessive.

Again, I would welcome someone with coding skills to write the code and validate my results. Given the opportunity, I would also test a model where any of the assets are sold if its price declines more than "x %" (maybe 5%) from the last rebalance date and the funds are held in cash until the next rebalance date. I think there is some additional value there, but trading activity would further incease and you need to monitor the assets daily.

Thanks for your interest.

Ed
Post Reply