Backtesting the PP all the way to 1971?

General Discussion on the Permanent Portfolio Strategy

Moderator: Global Moderator

Post Reply
User avatar
jason
Executive Member
Executive Member
Posts: 248
Joined: Mon Jun 10, 2013 4:10 pm

Backtesting the PP all the way to 1971?

Post by jason » Fri May 20, 2022 9:50 am

I used to use peaktotrough.com all the time for backtesting. Since that website shut down quite a while ago, I haven't been able to find a good replacement. Portfoliovisualizer and other websites I have found do not go all the way back to the Nixon shock, around August 15, 1971, the date that the gold price stopped being fixed. Are there any other solutions that go all the way back to around that date? I want to be able to play with the percentages such as testing with and without cash. And it would obviously be very helpful if the solution can do rebalancing bands.
User avatar
Tyler
Executive Member
Executive Member
Posts: 2066
Joined: Sat Nov 12, 2011 3:23 pm
Contact:

Re: Backtesting the PP all the way to 1971?

Post by Tyler » Fri May 20, 2022 10:39 am

While it uses simple annual rebalancing rather than bands, Portfolio Charts has the data coverage you're looking for (back to 1970).

Here's the dedicated Permanent Portfolio page: https://portfoliocharts.com/portfolio/p ... portfolio/
And here's a tool where you can adjust the percentages of the assets: https://portfoliocharts.com/portfolio/my-portfolio/

BTW, if you like Portfolio Visualizer but want more data, be sure to use "backtest asset allocation" instead of "backtest portfolio". The latter uses real fun data but has fewer years, while the former uses asset classes back to 1972.
User avatar
jason
Executive Member
Executive Member
Posts: 248
Joined: Mon Jun 10, 2013 4:10 pm

Re: Backtesting the PP all the way to 1971?

Post by jason » Fri May 20, 2022 10:54 am

Tyler wrote:
Fri May 20, 2022 10:39 am
While it uses simple annual rebalancing rather than bands, Portfolio Charts has the data coverage you're looking for (back to 1970).

Here's the dedicated Permanent Portfolio page: https://portfoliocharts.com/portfolio/p ... portfolio/
And here's a tool where you can adjust the percentages of the assets: https://portfoliocharts.com/portfolio/my-portfolio/

BTW, if you like Portfolio Visualizer but want more data, be sure to use "backtest asset allocation" instead of "backtest portfolio". The latter uses real fun data but has fewer years, while the former uses asset classes back to 1972.
I tried backtesting asset allocation on Portfolio Visualizer but it only goes back to 1978. It says they do not have LTT data older than that.
Regarding portfoliocharts, that goes all the way back to 1970, but my understanding is that PP data prior to the Nixon Shock on August 15, 1971 is not really valid because the price of gold was fixed before that date. Is there a way to set the start date to August 16, 1971? Thanks!
User avatar
Tyler
Executive Member
Executive Member
Posts: 2066
Joined: Sat Nov 12, 2011 3:23 pm
Contact:

Re: Backtesting the PP all the way to 1971?

Post by Tyler » Fri May 20, 2022 11:55 am

Hmm... It's interesting that PV doesn't have more LTT data than that. It's a terrific tool, but the way it automatically changes the start year is always kinda frustrating. You can rarely actually compare anything apples-to-apples.

For the record, I set the gold return for 1970 on PC to 0% since the price was fixed to the dollar. And while you can't directly crop data in the online tools, many of them allow you visually ignore the series starting in 1970 (first row, darkest line, etc). Because we're only talking about one year and the charts account for every start year after that, it shouldn't affect the visuals all that much.
User avatar
GT
Executive Member
Executive Member
Posts: 270
Joined: Sun Jan 20, 2013 7:54 pm

Re: Backtesting the PP all the way to 1971?

Post by GT » Fri May 20, 2022 1:46 pm

jason wrote:
Fri May 20, 2022 9:50 am
I used to use peaktotrough.com all the time for backtesting. Since that website shut down quite a while ago, I haven't been able to find a good replacement. Portfoliovisualizer and other websites I have found do not go all the way back to the Nixon shock, around August 15, 1971, the date that the gold price stopped being fixed. Are there any other solutions that go all the way back to around that date? I want to be able to play with the percentages such as testing with and without cash. And it would obviously be very helpful if the solution can do rebalancing bands.
Excel back testing spread sheet

Not sure if this has the info you are looking for....

Simba's backtesting spreadsheet

https://www.bogleheads.org/wiki/Simba%2 ... preadsheet
Kbg
Executive Member
Executive Member
Posts: 2815
Joined: Fri May 23, 2014 4:18 pm

Re: Backtesting the PP all the way to 1971?

Post by Kbg » Mon May 23, 2022 12:12 pm

GT wrote:
Fri May 20, 2022 1:46 pm
jason wrote:
Fri May 20, 2022 9:50 am
I used to use peaktotrough.com all the time for backtesting. Since that website shut down quite a while ago, I haven't been able to find a good replacement. Portfoliovisualizer and other websites I have found do not go all the way back to the Nixon shock, around August 15, 1971, the date that the gold price stopped being fixed. Are there any other solutions that go all the way back to around that date? I want to be able to play with the percentages such as testing with and without cash. And it would obviously be very helpful if the solution can do rebalancing bands.
Excel back testing spread sheet

Not sure if this has the info you are looking for....

Simba's backtesting spreadsheet

https://www.bogleheads.org/wiki/Simba%2 ... preadsheet
Yes but...

1. Annual info only which provides a butt load of smoothing that wallpapers over the real day to day pain one might feel

2. Don't forget, the USD and gold were the same thing for a good chunk of the Simba spreadsheet.

Personally, I don't think a backtest of the permanent portfolio is even valid before gold was allowed to float. I suppose one could find a dataset from a country where gold was not fixed...but then is a backtest even useful when evaluating something that is fundamentally not the same anymore. My answer is no.
User avatar
Tyler
Executive Member
Executive Member
Posts: 2066
Joined: Sat Nov 12, 2011 3:23 pm
Contact:

Re: Backtesting the PP all the way to 1971?

Post by Tyler » Mon May 23, 2022 1:02 pm

I admit this qualifies as a shameless plug (and I promise not to make a habit of it), but another option is to try one of my new standalone backtesting tools. You'll have to supply your own source data, but the individual PP asset histories aren't too hard to dig up and you also have control over the (post-1970) start year. With just a little work you'll also be able to see how varying the start month and rebalancing bands affects the results.
User avatar
Xan
Administrator
Administrator
Posts: 4392
Joined: Tue Mar 13, 2012 1:51 pm

Re: Backtesting the PP all the way to 1971?

Post by Xan » Mon May 23, 2022 1:34 pm

Tyler wrote:
Mon May 23, 2022 1:02 pm
I admit this qualifies as a shameless plug (and I promise not to make a habit of it), but another option is to try one of my new standalone backtesting tools. You'll have to supply your own source data, but the individual PP asset histories aren't too hard to dig up and you also have control over the (post-1970) start year. With just a little work you'll also be able to see how varying the start month and rebalancing bands affects the results.
Tyler, no need to ever be shy about plugging yourself! Please keep us all informed of the tremendously useful tools you build.
D1984
Executive Member
Executive Member
Posts: 730
Joined: Tue Aug 16, 2011 7:23 pm

Re: Backtesting the PP all the way to 1971?

Post by D1984 » Mon May 23, 2022 4:32 pm

Tyler wrote:
Mon May 23, 2022 1:02 pm
I admit this qualifies as a shameless plug (and I promise not to make a habit of it), but another option is to try one of my new standalone backtesting tools. You'll have to supply your own source data, but the individual PP asset histories aren't too hard to dig up and you also have control over the (post-1970) start year. With just a little work you'll also be able to see how varying the start month and rebalancing bands affects the results.
Hi Tyler,

I was wondering....is there any chance of any of the PortfolioCharts backtesting tools available for purchase (either as downloadable Excel sheets or on the site itself) being extended back to, say, 1964 or 1965 or even 1955 or so? As it is the site is already very useful and a wonderful tool....however, bonds got creamed from 1955 to 1959 and then again from 1965 or so to April or May 1970 as rates rose sharply and starting in 1970 misses almost all of this period of awful fixed-income performance. As a result, portfolios heavy in fixed-income (Wellesley, the PP, the 80/20, etc....maybe including the GB as well) end up with backtests not going through their harshest (so far) worst case SWR scenario...the one that started in 1965 or 1966 (although to be fair I think starting in 1956 would have been almost as bad if not slightly worse for the PP and for Wellesley than the 30 or 40 year sequence starting in 1965 or 1966); as such, people might have an unreasonably high expectation of SWR for these types of portfolios. IMO anyone expecting, say, a three or four percent real (not nominal) return from the PP for the withdrawal period starting from 1-1-2022 on is going to be painfully disappointed and had better learn to appreciate the finer points of the taste of Alpo....especially if they have 35 or 40 years (or more!) of retired lifespan because they did FIRE/early retirement/ERE!

Also, starting in 1964 or 1965 would allow backtesting of stock/bond combos during the (so far) worst -case SWR scenario (the 1965 or early 1966 start that had worse 30 and 40 year SWRs than even 1929 or 1937 during the withdrawal sequences that included much of the Great Depression).

Finally, starting in those years would show how non-US stocks could actually help a portfolio greatly; whether it was simulated EAFE or just simulated EASEA (the latter in order to get around the hypothetical complaint "well, of COURSE portfolios that included foreign stock that started their SWRs in the mid-1960s did well; Japan kicked butt from 1965 to 1989 and that was the cause of almost all their outperformance"; the fact of the matter is that even if you deliberately exclude Japan by using only EASEA--or EASEA+Canada--as your foreign stock allocation, every start year--except for maybe one or two in the early or mid 1960s--with starting dates from 1955 to the early or mid-1980s outperformed in SWR terms similar portfolios whose stocks were only US TSM or S&P 500) they in fact increased SWR rates for return sequences starting in these years. I mention this because you had previously stated that adding foreign stocks was of no more help (and generally less so) than simply adding gold or US SCV. For the years starting before the early 1970s this was most certainly not always the case.
User avatar
Tyler
Executive Member
Executive Member
Posts: 2066
Joined: Sat Nov 12, 2011 3:23 pm
Contact:

Re: Backtesting the PP all the way to 1971?

Post by Tyler » Mon May 23, 2022 5:00 pm

Thanks Xan! I'm always a bit self-conscious about pushing paid products on forums, as I dislike spam just as much as the next guy. But I really appreciate your support, and I hope you guys find the new stuff useful.

D1984 wrote:
Mon May 23, 2022 4:32 pm
Hi Tyler,

I was wondering....is there any chance of any of the PortfolioCharts backtesting tools available for purchase (either as downloadable Excel sheets or on the site itself) being extended back to, say, 1964 or 1965 or even 1955 or so? As it is the site is already very useful and a wonderful tool....however, bonds got creamed from 1955 to 1959 and then again from 1965 or so to April or May 1970 as rates rose sharply and starting in 1970 misses almost all of this period of awful fixed-income performance. As a result, portfolios heavy in fixed-income (Wellesley, the PP, the 80/20, etc....maybe including the GB as well) end up with backtests not going through their harshest (so far) worst case SWR scenario.
I probably won't be extending the online tools any time soon unless I can find new data sources, as more years isn't very useful if there are only a small handful of assets to model. But I have an open mind, especially with the downloads that have more flexibility. And in the meantime, even if you can't change the year labels there's nothing stopping you from using the standalone Withdrawal Rates tool and entering the inflation-adjusted returns data from 1962-2021. There are 60 years of open slots, after all. ;)

BTW, I do cover the SWR issue you're talking about in the Withdrawal Rates FAQ. At least for traditional stock/bond portfolios where you can compare apples to apples, the difference in SWRs calculated since 1970 and since 1870 is only about 0.3%. I hear ya, though.
D1984
Executive Member
Executive Member
Posts: 730
Joined: Tue Aug 16, 2011 7:23 pm

Re: Backtesting the PP all the way to 1971?

Post by D1984 » Mon May 23, 2022 7:12 pm

Tyler wrote:
Mon May 23, 2022 5:00 pm
Thanks Xan! I'm always a bit self-conscious about pushing paid products on forums, as I dislike spam just as much as the next guy. But I really appreciate your support, and I hope you guys find the new stuff useful.

D1984 wrote:
Mon May 23, 2022 4:32 pm
Hi Tyler,

I was wondering....is there any chance of any of the PortfolioCharts backtesting tools available for purchase (either as downloadable Excel sheets or on the site itself) being extended back to, say, 1964 or 1965 or even 1955 or so? As it is the site is already very useful and a wonderful tool....however, bonds got creamed from 1955 to 1959 and then again from 1965 or so to April or May 1970 as rates rose sharply and starting in 1970 misses almost all of this period of awful fixed-income performance. As a result, portfolios heavy in fixed-income (Wellesley, the PP, the 80/20, etc....maybe including the GB as well) end up with backtests not going through their harshest (so far) worst case SWR scenario.
I probably won't be extending the online tools any time soon unless I can find new data sources, as more years isn't very useful if there are only a small handful of assets to model. But I have an open mind, especially with the downloads that have more flexibility. And in the meantime, even if you can't change the year labels there's nothing stopping you from using the standalone Withdrawal Rates tool and entering the inflation-adjusted returns data from 1962-2021. There are 60 years of open slots, after all. ;)

BTW, I do cover the SWR issue you're talking about in the Withdrawal Rates FAQ. At least for traditional stock/bond portfolios where you can compare apples to apples, the difference in SWRs calculated since 1970 and since 1870 is only about 0.3%. I hear ya, though.
Tyler,

Thanks for the info on this. I wasn't aware of the ability to start before 1970; from what it appeared on the website that was the earliest possible start date (the grey blocks on the Excel spreadsheet always were shown as starting at 1970 or later).

I have PM'd you regarding a good potential source for ex-US annual return info that isn't copyrighted (or at least effectively isn't; some of the data may be copyrighted but if the copyright holder/s provide it free to anyone who asks they have a pretty weak case against allowing the data to be shared....esp since it would be for non-commercial use and thus likely would come under "fair use" ).
User avatar
vnatale
Executive Member
Executive Member
Posts: 9422
Joined: Fri Apr 12, 2019 8:56 pm
Location: Massachusetts
Contact:

Re: Backtesting the PP all the way to 1971?

Post by vnatale » Mon May 23, 2022 7:24 pm

Xan wrote:
Mon May 23, 2022 1:34 pm

Tyler wrote:
Mon May 23, 2022 1:02 pm

I admit this qualifies as a shameless plug (and I promise not to make a habit of it), but another option is to try one of my new standalone backtesting tools. You'll have to supply your own source data, but the individual PP asset histories aren't too hard to dig up and you also have control over the (post-1970) start year. With just a little work you'll also be able to see how varying the start month and rebalancing bands affects the results.


Tyler, no need to ever be shy about plugging yourself! Please keep us all informed of the tremendously useful tools you build.


EXACTLY!!!!
Above provided by: Vinny, who always says: "I only regret that I have but one lap to give to my cats." AND "I'm a more-is-more person."
User avatar
Tyler
Executive Member
Executive Member
Posts: 2066
Joined: Sat Nov 12, 2011 3:23 pm
Contact:

Re: Backtesting the PP all the way to 1971?

Post by Tyler » Mon May 23, 2022 9:19 pm

D1984 wrote:
Mon May 23, 2022 7:12 pm
Thanks for the info on this. I wasn't aware of the ability to start before 1970; from what it appeared on the website that was the earliest possible start date (the grey blocks on the Excel spreadsheet always were shown as starting at 1970 or later).

I have PM'd you regarding a good potential source for ex-US annual return info that isn't copyrighted (or at least effectively isn't; some of the data may be copyrighted but if the copyright holder/s provide it free to anyone who asks they have a pretty weak case against allowing the data to be shared....esp since it would be for non-commercial use and thus likely would come under "fair use" ).
Well, all of the data on the site does start in 1970. And the new downloadable spreadsheets also say 1970. But really those are just labels on placeholders for an annual return (just like the placeholders for future years through 2029), so if you want to use it off-label to study different timeframes it won't affect the actual calculations. BTW, here's a free sample to tinker with if you like to see how it all works.

And thanks for the data suggestions!
User avatar
jason
Executive Member
Executive Member
Posts: 248
Joined: Mon Jun 10, 2013 4:10 pm

Re: Backtesting the PP all the way to 1971?

Post by jason » Mon May 23, 2022 10:18 pm

GT wrote:
Fri May 20, 2022 1:46 pm
jason wrote:
Fri May 20, 2022 9:50 am
I used to use peaktotrough.com all the time for backtesting. Since that website shut down quite a while ago, I haven't been able to find a good replacement. Portfoliovisualizer and other websites I have found do not go all the way back to the Nixon shock, around August 15, 1971, the date that the gold price stopped being fixed. Are there any other solutions that go all the way back to around that date? I want to be able to play with the percentages such as testing with and without cash. And it would obviously be very helpful if the solution can do rebalancing bands.
Excel back testing spread sheet

Not sure if this has the info you are looking for....

Simba's backtesting spreadsheet

https://www.bogleheads.org/wiki/Simba%2 ... preadsheet
Thanks! I was able to use this to test from the beginning of 1972 through the end of 2022 using the Simba spreadsheet. I was curious what the performance would be with a PP without cash, so one third each of stocks, LTTs, and gold. I was surprised that it came out to a little over 10% CAGR. That's really strong. But I'm not totally sure I am using the spreadsheet correctly. Can anyone else confirm this?
Post Reply