Ultimate SWR Portfolio - The Highest SWR Across Multiple Countries

[vand]

This follows on from my conviction of the lack of requirement for any exposure to bonds in the best asset allocation for retirement portfolios. 

In tinkering with the optimal blend with 3 assets or more I eventually came up with a mix of 45/30/25 in Stocks/REITs/Gold that supported a 6.4% SWR which I considered the optimal mix...

However, this is very much from a US point of view, and as I am not US based (rather UK) I'm more interested in what the portfolio does for me from a UK, and one of the great things about Portfoliocharts is that it lets you backtest any allocation mix from one of 12 countries.   The 45/30/25 mix drops to 5.8% from a UK perspective.. not bad, but I'd like to think it can still be improved from a UK perspective, and so in tinkering around I naturally wanted to refine it and then expand the question to determine:

What is the optimal portfolio that will work the best across all countries?

This is what I came up with, which I've rather immodestly called this the "Ultimate SWR Portfolio":
https://portfoliocharts.com/charts/portfolio-matrix/

US Stocks Large Cap Blend: 16%
US Small Cap Value: 10%
European Large Cap Blend: 4%
Emerging Markets Large Cap Blend: 6%
REITs: 23%
Gold: 23%
Domestic Stocks Large Cap Blend: 18%





I then used PortfolioCharts to backtest the SWR of this portfolio and compared it against 9 other portfolios that consistently rank highly in SWR, and I ran it across all 12 countries.

That is to say, I compared 10 portfolios across 12 countries to see how they fared in each.

Portfolios Tested:
Ultimate SWR Portfolio, Total Stock Market (Domestic), 60/40 (Domestic), Permanent Portfolio, Golden Butterfly, Ivy Portfolio, Pinwheel Portfolio, 7Twelve Portfolio, Weird Portfolio, Golden Ratio Portfolio.

Countries Tested are the 12 available in PortfolioCharts:
Australia, Canada France, Germany, Italy, Japan, Netherlands, Spain, Sweden, Switzerland, UK, US.

This makes it a truly all-compassing portfolio-agnostic country-agnostic test  -- Not only will it determine the "best" retirement portfolio, but also the best countries to retire to for the highest SWRs given the weight of evidence across all the portfolios tested.

How do they backtest? here's the raw data:

(update 2025-04-22)



This orders them the best Countries from top to bottom, ranking by the average SWR across all 10 portfolios for each country, and simultaneously the best Portfolios from left to right, ranking them by the average SWR across all 12 countries for each portfolio.

Some observations:


The spread from best to worst country across the average of all portfolios is 5.41% for the best country, Sweden, to 3.76% for the worst country with is Spain, so any given strategy with perform 44% better in the best country than it will in the worst country.

The spread from the best to worst portfolio across the average of every country is 5.73% for the best mix which is the Ultimate SWR Portfolio, compared to 3.08% for the worst strategy's average, which is 100% domestic stocks, so in any given country the best portfolio performs 86% better, on average, than the worst portfolio strategy. 

It should be pretty clear that a 100% portfolio in one's own domestic stock market is by far the riskiest and often worst performing portfolio you can hold.  Even in the best country, it performs worst than the best portfolio in the worst country - that is a 4.2% SWR for domestic TSM for the US vs 4.7% for the Ultimate SWR portfolio in Switzerland.

Even a Domestic 60/40 is terrible if you look at the average or worst case across all countries - reinforcing my belief that bonds are a near-useless diversifier for stocks and should not be considered as the automatic go-to balancing asset for multi-asset portfolios.

On the other end the portfolios that blend more uncorrelated asset classes tend to come out the best.

US Investors and US-centric SWR discussion, which tend to churn out all of the content in the asset allocation and withdrawals strategy space may "get away" with the worst or 2nd worst of these portfolios, but only because they are fortunate enough to be in the best country for the worst portfolios -- those same strategy adopted in some other countries will get you crushed.

I had not heard of the "Weird Portfolio" before, but looking at its composition, rather unsurprisingly it shares a lot of similarity with the Ultimate SWR Portfolio, with large allocation to gold, REITs, and SCV.  Personally, I didn't want to lean on the SCV too hard as, despite the history, there's just no way to know if its advantage will persist into the future, and with markets increasingly more efficient over time it is hard to imagine how they could outperform their large cap so much.

Anyway, Retire to Sweden, US or Canada with the Ultimate SWR Portfolio or Weird Portfolio if you want the best bang for your buck... or conversely retire to Spain or Italy with a domestic 60/40 or domestic stock-only portfolio if you fancy a challenge.   Them's the choices!

[MustacheAndaHalf]

Your other thread mentioned a start date, which you've conveniently left off here, so I'll repeat my criticism from your original thread: the U.S. cannot leave the gold standard twice, so you can't count on that performance in your portfolio.

Notice the start date: 1972.  People pushing gold ALWAYS include 1972-1974, which were 3 of the best 4 years of gold's performance.  What gets left out is U.S. leaving the gold standard in 1971, which ended the fixed price of gold.  The U.S. won't leave the gold standard twice, so performance from 1972-1974 will not repeat.

1972 - 2024: gold returned 8%/year
1975 - 2024: gold returned 5%/year

(According to Portfolio Visualizer, using portfolio of 100% gold)
https://www.portfoliovisualizer.com/backtest-asset-class-allocation

[vand]

My REITs are a sea of GREEN today lol

[Fru-Gal]

Why are REITs green?

[ChpBstrd]

Interesting work @vand . Maybe it's a pedantic quibble, but you should always use past tense when talking about backtesting. I.e. this is "what worked best" not "what works best" or "what will work best". This clarifies that we are using inferential reasoning, and not making the much bolder claim that the future WILL look like the past.

Accordingly, Spain might be a great opportunity today because valuations are low, whereas it was a poor opportunity in the past because valuations didn't go up. Japan might be a great opportunity today for the same reasons it was tough sledding for the past 35 years. Or maybe there's a chance that a developing market like India or Brazil will have the highest SWR in the future, so any claims arising from this list of past performers cannot account for this probability and is therefore invalid? 

What was the backtesting timeframe? This may explain certain results. E.g. Japan's difficulty's or where TF did Sweden come from?

[vand]

Interesting work @vand . Maybe it's a pedantic quibble, but you should always use past tense when talking about backtesting. I.e. this is "what worked best" not "what works best" or "what will work best". This clarifies that we are using inferential reasoning, and not making the much bolder claim that the future WILL look like the past.

Accordingly, Spain might be a great opportunity today because valuations are low, whereas it was a poor opportunity in the past because valuations didn't go up. Japan might be a great opportunity today for the same reasons it was tough sledding for the past 35 years. Or maybe there's a chance that a developing market like India or Brazil will have the highest SWR in the future, so any claims arising from this list of past performers cannot account for this probability and is therefore invalid? 

What was the backtesting timeframe? This may explain certain results. E.g. Japan's difficulty's or where TF did Sweden come from?

The point of the exercise isn't to try to take a punt on what markets might provide the best outcomes in the future due to their current valuations when the data isn't there to support the thesis from a historical point of view...

It's to show what mix produced the least worst outcome and was most insensitive to external factors, so can reasonably claim to work well under the widest range of scenarios - that's how you build confidence in your portfolio when the future is unknowable.

From a fundamental pov Japan, Spain etc may be cheap, but as the saying goes they are cheap for good reason (growth, demographics, political instability, national finances etc)

[swashbucklinstache]

It appears that your table is missing Australia data though I can't confirm as I'm not a member of that site.

{Interlude: nicer to read the above than 'you're a propagandist!' when someone thinks your model could be better}

With deference to the idea that any backtesting is always of limited usefulness, I'll challenge you to go one step deeper on this one. Checking against other portfolios and other countries is indeed somewhat similar to building any portfolio for theoretical reasons and then seeing how it performed in backtesting. Your results are clearly not incorrect, but the next level to this is to assess how much confidence should we have that such a portfolio will outperform in the future given these results? That is, let's quantify the word "reasonable" below.

Please note that I'm assuming you went from your original 3 assets to the observed mix without observing the other (non-UK) backtesting. If that is incorrect and this is instead simply the result of "what asset mix gives the best average across all countries" i.e. purely identifying a portfolio based on backtesting with any intent to extrapolate from the results , including the below, I would advise extreme caution on any extrapolating and advise you to stop until you learn more.

It's to show what mix produced the least worst outcome and was most insensitive to external factors, so can reasonably claim to work well under the widest range of scenarios - that's how you build confidence in your portfolio when the future is unknowable.

I'll get you started on either of the above with a few optional prompts. Assume reasonable timeframes for the below. We have more data available, so we should use it. We also know this is not all data, because there's more historical and future data. You got us started by not only looking at average SWR, which is good.

• This portfolio was not the highest for every country. Can you tell me the likelihood that we would observe this frequency of outperformance or more extreme if, in fact, your model was actually exactly as good as or worse than the next best model? The one after that?
• We also have data here about magnitude of difference. Can you calculate a confidence interval for outperformance?
• 11 or 12 countries is better than 1, but it's far from all countries one might reasonably retire in. What does a power analysis say?
• To what extent are these selected countries' large cap performances correlated to US performance outside of the sample window? This is optional and hard but could be a very good exercise in seeking extra data instead of extrapolating. It's a common problem in a globalized economy. For example, a nation having a trade agreement with the US and then not.

[vand]

It appears that your table is missing Australia data though I can't confirm as I'm not a member of that site.

{Interlude: nicer to read the above than 'you're a propagandist!' when someone thinks your model could be better}

With deference to the idea that any backtesting is always of limited usefulness, I'll challenge you to go one step deeper on this one. Checking against other portfolios and other countries is indeed somewhat similar to building any portfolio for theoretical reasons and then seeing how it performed in backtesting. Your results are clearly not incorrect, but the next level to this is to assess how much confidence should we have that such a portfolio will outperform in the future given these results? That is, let's quantify the word "reasonable" below.

Please note that I'm assuming you went from your original 3 assets to the observed mix without observing the other (non-UK) backtesting. If that is incorrect and this is instead simply the result of "what asset mix gives the best average across all countries" i.e. purely identifying a portfolio based on backtesting with any intent to extrapolate from the results , including the below, I would advise extreme caution on any extrapolating and advise you to stop until you learn more.

It's to show what mix produced the least worst outcome and was most insensitive to external factors, so can reasonably claim to work well under the widest range of scenarios - that's how you build confidence in your portfolio when the future is unknowable.

I'll get you started on either of the above with a few optional prompts. Assume reasonable timeframes for the below. We have more data available, so we should use it. We also know this is not all data, because there's more historical and future data. You got us started by not only looking at average SWR, which is good.

• This portfolio was not the highest for every country. Can you tell me the likelihood that we would observe this frequency of outperformance or more extreme if, in fact, your model was actually exactly as good as or worse than the next best model? The one after that?
• We also have data here about magnitude of difference. Can you calculate a confidence interval for outperformance?
• 11 or 12 countries is better than 1, but it's far from all countries one might reasonably retire in. What does a power analysis say?
• To what extent are these selected countries' large cap performances correlated to US performance outside of the sample window? This is optional and hard but could be a very good exercise in seeking extra data instead of extrapolating. It's a common problem in a globalized economy. For example, a nation having a trade agreement with the US and then not.

A thousand apologies for missing off Australia - It was in there but I had a hidden row in the sheet so wasn't showing throw.

Here it is (I've also updated the top post):



Unfortunately it looks like since I started this thread Tyler has paywalled the portfolio matrix page that I used to run the tests
https://portfoliocharts.com/charts/portfolio-matrix/
:(

[vand]

This portfolio was not the highest for every country. Can you tell me the likelihood that we would observe this frequency of outperformance or more extreme if, in fact, your model was actually exactly as good as or worse than the next best model? The one after that?
We also have data here about magnitude of difference. Can you calculate a confidence interval for outperformance?
11 or 12 countries is better than 1, but it's far from all countries one might reasonably retire in. What does a power analysis say?
To what extent are these selected countries' large cap performances correlated to US performance outside of the sample window? This is optional and hard but could be a very good exercise in seeking extra data instead of extrapolating. It's a common problem in a globalized economy. For example, a nation having a trade agreement with the US and then not.

all very good questions - from a statistical perspective it may be useful to quantify the answer to those and I honestly won't say that I can put numbers to them.

However, the theory is that the 12 countries tested are just what PortfolioCharts had data for, so I'm just testing the widest dataset available.  Hopefully we're in agreement that you use as much data as possible and not exclude anything that might produce outlier results as those results are the ones we should question our theories the most.

PortfolioCharts' own reasoning is that they are major 12 developed economies for which history is available - of course we just don't have the data to be able to say if you are retiring in Iceland or Egypt or China that these assumptions will hold up as things like rule of law, taxation, inflation and exchanges rates can be very country dependent.

[vand]

While not specifically related to the sub-topic ofSWRs, Portfoliocharts' latest delve into the outright most resilient portfolios to survive downturns basically confirms that from their list of model portfolios the Weird Portfolio wins.

https://portfoliocharts.com/2025/04/09/how-to-succeed-in-the-worst-stock-markets/

note how it very effectively deals with the whole "gold in the 70s" question


Also, if people want to backtest their own portfolios the SWR page is still available for free:

https://portfoliocharts.com/charts/withdrawal-rates/

Also to quote one of the closing lines from the article...

"So if you’ve ever felt pressured into investing in an asset you just don’t personally care about, take some time to evaluate your options. There’s a portfolio for all types of people."

I will freely admit that one of the personal drivers of the Ultimate SWR Portfolio was to challenge the orthodoxy that bonds should be the go-to diversifying asset. So if you like the idea of diversification but are prejudiced against repackaged federal reserve notes bonds as much as I am then go with something like the Ultimate SWR Portfolio. 

Also, at 52%, the USWRP actually has a considerably higher total stock allocation than Weird, Permanent, Golden Ratio or any of the other model portfolios that emerge as the "best" from the article. So if you have a bias towards stocks then again USWRP should be something worth considering.

[vand]

I have added 100% US Stocks and 60/40 US also



While for some of the worst outcome countries sticking to the US stock/bond market over their domestic equivalents would have helped by avoid the trainwreck of the domestic stock market, for other countries the currency swings that it introduces actually worsens the portfolio.

This highlights the point I made in the other thread that if you are a non-US investor it is a bad idea to hold all your wealth in the assets of another country -- even if that country is the US.  When market risk and FX risk both go against you it will more than offset any gain you get from exposure to the best (long term) equity market.


This is pertinent right now because.. look at how the weak USD has amplified the losses for foreign investors:  Here's S&P Total Return in GBP -- the current decline is almost at the same level as Covid. Another bad day or two and we'll be right there.