100% defensive stocks vs 60/40

[rule of law guy]

in addition to "Consumer Staples (which I understand to be things you buy in a supermarket), Healthcare and Utilities" for defensive equity positioning now, I think you have to add energy/oil. especially royalty trusts and pipeline companies, which have the current income stream aspect but also should preserve their yields even as we go into recession (yes commodity prices generally will come down as we move from inflation to deflation, but the demand for energy should not decrease substantially because there is no substitution effect available). it is an interesting paradox, which Buffett clearly understands, that as the developing nations move to more green energy as a policy goal, hydrocarbon energy becomes more enticing as an investing option.

[Logan Roy]

Can this defensive stocks idea be translated into a simple, Bogleheads-style move ...?

A big chunk of my portfolio is in index fund-related stuff (for example, my 401k is in Vanguard Target 2025 fund); might SCHD (Schwab U.S. Dividend Index ETF, in my brokerage account) be seen as defensive? If not, I'm considering trading that for the Vanguard Total Stock Market fund. I've already got bond-y things like Wellesley and Dodge & Cox Balanced in IRA and Roth accounts (balanced funds have always been my comfort zone).

Edit: I see that Vanguard has a Consumer Staples ETF, VDC. I guess that's my answer.

I'd say what's fairly consistent so far is overweighting Consumer Staples (Consumer NonDurables here) and Healthcare. It's generally leaving Utilities out. Lower long-term returns and relatively high volatility don't favour the criteria I'm using at the moment. And what's a bit surprising, although as rule of law guy says, is it tends to overweight Energy. So this portfolio optimises for a high Sharpe ratio and return (Sharpe x Arithmetic Mean Return). There's something somewhat Harry Browne about this. A combination of market risk, earnings stability and inflation protection, in similar sized chunks: (this test leaves the volatile period for Utilities out – the next one uses all the data.)

25% Energy
24% S&P 500
24% Consumer NonDur
22% Healthcare
5% Tech/Other
(34 month rebalance)



I'm using Evolution Strategies now. So unlike random portfolio weightings, it can now optimise and investigate effective designs. This portfolio has gold as an option. It's quite consistently using between 10 and 23%, and disregarding T bonds.

33% Healthcare
27% S&P 500
12% Gold
11% Consumer NonDur
8% Energy
4% Tech
(34 month rebalance)



So I'd say a Boglehead-friendly approach could be adding moderate Consumer Staples and Healthcare overweights, and a smaller Energy overweight. And, it seems, rebalancing less often.

[Lawrence of Suburbia]

Can this defensive stocks idea be translated into a simple, Bogleheads-style move ...?

A big chunk of my portfolio is in index fund-related stuff (for example, my 401k is in Vanguard Target 2025 fund); might SCHD (Schwab U.S. Dividend Index ETF, in my brokerage account) be seen as defensive? If not, I'm considering trading that for the Vanguard Total Stock Market fund. I've already got bond-y things like Wellesley and Dodge & Cox Balanced in IRA and Roth accounts (balanced funds have always been my comfort zone).

Edit: I see that Vanguard has a Consumer Staples ETF, VDC. I guess that's my answer.

I'd say what's fairly consistent so far is overweighting Consumer Staples (Consumer NonDurables here) and Healthcare. It's generally leaving Utilities out. Lower long-term returns and relatively high volatility don't favour the criteria I'm using at the moment. And what's a bit surprising, although as rule of law guy says, is it tends to overweight Energy. So this portfolio optimises for a high Sharpe ratio and return (Sharpe x Arithmetic Mean Return). There's something somewhat Harry Browne about this. A combination of market risk, earnings stability and inflation protection, in similar sized chunks: (this test leaves the volatile period for Utilities out – the next one uses all the data.)

25% Energy
24% S&P 500
24% Consumer NonDur
22% Healthcare
5% Tech/Other
(34 month rebalance)



I'm using Evolution Strategies now. So unlike random portfolio weightings, it can now optimise and investigate effective designs. This portfolio has gold as an option. It's quite consistently using between 10 and 23%, and disregarding T bonds.

33% Healthcare
27% S&P 500
12% Gold
11% Consumer NonDur
8% Energy
4% Tech
(34 month rebalance)



So I'd say a Boglehead-friendly approach could be adding moderate Consumer Staples and Healthcare overweights, and a smaller Energy overweight. And, it seems, rebalancing less often.

Thanks for that! A lot to think about.

[Tellurius]

I'd suggest it's perhaps not entirely clearcut.

A perpetual bond can be regarded as a stock. In industries where governments subsidise and provide safety nets to businesses, there's a guarantee on future cashflows that at least has something bond-like about it. There's to what extent a government can be viewed as a business, and at what point a business might become more like a state. A business can fail, but if a whole market were to fail, might one not already be at the point where a government's failed? Wouldn't we default on debt before we let the Utilities sector go to zero, and plunge us all into the middle ages?

Why are you taking these extreme examples which are mostly theoretical? eg The last perpetual bonds of the UK government have been redeemed in the 2010s.

A government cannot be viewed as a business because it has the so-called “monopoly of violence” which leads to it having lots and lots of power, not least that for affecting finances. If the whole market were to fail, you have a different type of government.

[Logan Roy]

I'd suggest it's perhaps not entirely clearcut.

A perpetual bond can be regarded as a stock. In industries where governments subsidise and provide safety nets to businesses, there's a guarantee on future cashflows that at least has something bond-like about it. There's to what extent a government can be viewed as a business, and at what point a business might become more like a state. A business can fail, but if a whole market were to fail, might one not already be at the point where a government's failed? Wouldn't we default on debt before we let the Utilities sector go to zero, and plunge us all into the middle ages?

Why are you taking these extreme examples which are mostly theoretical? eg The last perpetual bonds of the UK government have been redeemed in the 2010s.

A government cannot be viewed as a business because it has the so-called “monopoly of violence” which leads to it having lots and lots of power, not least that for affecting finances. If the whole market were to fail, you have a different type of government.

From past experience. I've recommended funds that use perpetuals, ZDPs, junk, private debt, but warned that the risk is more comparable to stocks than bonds. So put it in the stock part of your portfolio. And I stopped recommending them, because people get lulled into thinking these give you market-like returns for cash-like risk, because volatility is how they think of risk (even CFAs).

Conversely, the market seems to put a lower risk premium on Utilities. Yet the sector's often more volatile than the market, presumably because of these long-duration characteristics, and demand changes so radically with economic risk. Across a portfolio, standard deviation does tell you something about how correlated or balanced things are – but risk clearly has distribution and concentration. I think the advantage of 100 year backtests is that concentration is more likely to show up either in standard deviation or total returns. So I'm getting quite satisfactory results optimising for Sharpe here, whereas with Portfolio Visualizer, due to generally having less data, I've not found it as useful.

[Logan Roy]

Thanks for that! A lot to think about.

Thank you! I'd say obviously everything with a strong pinch of salt. I'm the first to say backtests are only a map of the past – and I'm only at an amoebal stage with what I want to do with this so far. But what I'm looking for is if there may be consistent principles emerging from the way an algorithm optimises these problems. And I think there seems to be something, so far, in overweighting the most stable or recession-proof earnings, offset against the most cyclical (and growth/inflation sensitive).

Which is somewhat what Buffett and Munger's portfolios look like (soft drinks and banks). Just that the recent moves into businesses like Apple and Amazon may represent a new kind of consumer defensive. I think iCloud and Apple Music subscriptions may prove more resilient than brand toothpaste.

[Forester]

It's silly to bet on industries vs buying one of the existing low vol / min vol ETFs. It's similar to creating a Value tilt by only holding the Energy and Industrial ETFs, as opposed to a value ETF which screens across all sectors.

I doubt one can even create a mix of Staples / Utilities / Healthcare ETFs in a backtest, which has a lower drawdown than SPLV or USMV since their 2011 inception.

[Logan Roy]

It's silly to bet on industries vs buying one of the existing low vol / min vol ETFs. It's similar to creating a Value tilt by only holding the Energy and Industrial ETFs, as opposed to a value ETF which screens across all sectors.

I doubt one can even create a mix of Staples / Utilities / Healthcare ETFs in a backtest, which has a lower drawdown than SPLV or USMV since their 2011 inception.

The reason I disagree with this is that fundamentals weighting has little practical application to portfolio design, and little reliable long-term data (as the way the market processes data changes over time as our relationship to information changes). Graham noted that security analysis made a lot more sense in the 1930s than it did by the 1970s. Markets adapted. People could obtain earnings reports, and analysts had gone from reading accounts to counting vehicles in retail carparks via satellite.

What sectors give you is predictable, causal relationships to things going on in economies. Markets don't necessarily get any better at forecasting recessions or inflation, so earnings underlying sensitive sectors still have to be reactive. And the best portfolio design tends to be grounded in macroeconomics (imo). Factors lead to portfolios designed on the abstract mathematical relationships. You need faith that markets don't simply adapt. And that volatility is an adequate measure of risk. In portfolio design, a volatile asset class with reliable earnings can be a best of both worlds diversifier (e.g. long-duration bonds).

[skinnybuddha]

I'm surprised that no one has raised the idea of using a dividend appreciation etf (like VIG or VIGI) for this. Those would overweight you in healthcare/consumer defensive/utilities, underweight tech/consumer cyclicals and pull out the companies in other sectors with more resilient earnings growth/quality, which I'd consider defensive. As opposed to a pure sector model, it would recognize that there are more and less defensive companies within given sectors. What does the community think of that in lieu of the 60/40?

Obligatory note that dividend appreciation is not "high dividend" - in fact VIGI has a way lower dividend than VXUS. It's about companies that have steadily increased their dividends for years, in fact cutting out the highest dividend payers in the sample because those are seen as likely unsustainable and leading to turnover in the index.

[Logan Roy]

I'm surprised that no one has raised the idea of using a dividend appreciation etf (like VIG or VIGI) for this. Those would overweight you in healthcare/consumer defensive/utilities, underweight tech/consumer cyclicals and pull out the companies in other sectors with more resilient earnings growth/quality, which I'd consider defensive. As opposed to a pure sector model, it would recognize that there are more and less defensive companies within given sectors. What does the community think of that in lieu of the 60/40?

Obligatory note that dividend appreciation is not "high dividend" - in fact VIGI has a way lower dividend than VXUS. It's about companies that have steadily increased their dividends for years, in fact cutting out the highest dividend payers in the sample because those are seen as likely unsustainable and leading to turnover in the index.

I'd be interested if we can find some long-term data on something equivalent. My sense so far is that the rebalancing effect is a major way the algorithms manage risk. And that bonds aren't particularly helpful (on past results). Here I'm now testing 20 year rolling returns, over 80-100 years, and optimising for risk and return over each 20 year period, then averaging. So this is maximising consistency, so that any 20 year period in this backtest has about the best balance of risk and return possible relative to every other, using a fixed asset allocation.

So this is sort of the 'expected' portfolio – I think it's saying Utilities are not much use, as expected returns are low. Earnings stability seems to be important – it's about 75% in Consumer stocks and Healthcare (recession trades). And this 25% in Energy reminds me of the Harry Browne portfolio. If I optimise slightly more for Sharpe, and less for mean or total returns, it does use Utilities a lot more. Also not using gold as much as I'd expect – but then gold had a long period of just behaving like the dollar here.

Consumer NonDur - 25%
Healthcare - 45%
Energy - 25%
Tech/Other - 5%
(Rebalance every 20 months)

[Logan Roy]

So I fixed a silly mistake. I was wondering why the market seemed to mean revert to the above portfolio so perfectly – and realised that it had to do with backtesting all these rolling portfolios, and the 'result' portfolio (it prints) being a collage of these. So now I test as before, but run the winning portfolio through from start to finish once the test is done. It outperforms much more – unsurprisingly, given that Staples and Healthcare have tended to be the highest returning sectors over the 20th century.

What's interesting is it's coming up with the same portfolio, testing from 1926, as it does from 1940. So that's one tick for out-of-sample testing. Yesterday I tried running this experiment using Efficient Frontiers – which is much simpler to calculate. Unfortunately (so far), Efficient Frontiers hasn't come up with any portfolios comparable to the brute force portfolios. Unsurprising, as you realise how little information EF is working with: average returns and a covariance matrix. The mud at a wall approach inherently factors in macro, trend, and potentially every undiscovered mechanism in markets – much the same way evolution comes up with solutions to problems it doesn't grasp. These results are heavily determined by how I'm measuring risk and success. So better results will come from better ways of measuring or conceptualising them. Efficient Frontiers is effectively doing the same thing, but only optimising for a single period, which is then averaged. It makes sense if we accept markets are a truly random walk. Measuring the distribution of market returns over all periods, the results aren't anywhere near flat enough to suggest randomness. Economic cycles, momentum, mean reversion, would all be mechanisms that spoil that view – whether they're predictable or not. Over the past 100 years, having run probably 200 million portfolios, defensive sectors have proven better at managing risk (relative to return) than gold or bonds.

45% Healthcare
25% Cons NonDur
25% Energy
5% Tech
(Rebalance every 24 months)

[TheDoctor91]

So I fixed a silly mistake. I was wondering why the market seemed to mean revert to the above portfolio so perfectly – and realised that it had to do with backtesting all these rolling portfolios, and the 'result' portfolio (it prints) being a collage of these. So now I test as before, but run the winning portfolio through from start to finish once the test is done. It outperforms much more – unsurprisingly, given that Staples and Healthcare have tended to be the highest returning sectors over the 20th century.

What's interesting is it's coming up with the same portfolio, testing from 1926, as it does from 1940. So that's one tick for out-of-sample testing. Yesterday I tried running this experiment using Efficient Frontiers – which is much simpler to calculate. Unfortunately (so far), Efficient Frontiers hasn't come up with any portfolios comparable to the brute force portfolios. Unsurprising, as you realise how little information EF is working with: average returns and a covariance matrix. The mud at a wall approach inherently factors in macro, trend, and potentially every undiscovered mechanism in markets – much the same way evolution comes up with solutions to problems it doesn't grasp. These results are heavily determined by how I'm measuring risk and success. So better results will come from better ways of measuring or conceptualising them. Efficient Frontiers is effectively doing the same thing, but only optimising for a single period, which is then averaged. It makes sense if we accept markets are a truly random walk. Measuring the distribution of market returns over all periods, the results aren't anywhere near flat enough to suggest randomness. Economic cycles, momentum, mean reversion, would all be mechanisms that spoil that view – whether they're predictable or not. Over the past 100 years, having run probably 200 million portfolios, defensive sectors have proven better at managing risk (relative to return) than gold or bonds.

45% Healthcare
25% Cons NonDur
25% Energy
5% Tech
(Rebalance every 24 months)

Sorry I didn't notice anywhere else in the thread but did you also try including LTTs in to your optimization?

I was going to run the same Ken French for the industry portfolios and see what I got as well and it would be a good intro to running the mean variance optimization in pyhon.

[Logan Roy]

Sorry I didn't notice anywhere else in the thread but did you also try including LTTs in to your optimization?

I was going to run the same Ken French for the industry portfolios and see what I got as well and it would be a good intro to running the mean variance optimization in pyhon.

I could only get 10 year Treasury data going back that far – from Shiller's data page. LTT would be ideal. If I could simulate long and short duration TIPS, I think that would really help confirm some things.

Just my 2c, but I did spend an evening playing with mean variance optimization. And I was very disappointed, properly testing the portfolios it came up with. I can appreciate the theory – if we really were dealing with random walks, then optimising over a single period (e.g. a month – which is what it's doing) should be the same as optimising over a year, a decade, etc. And you could say what I'm running into by not doing that is more back-fitting..

But if you take the relationship between stocks and bonds over a decent period (so you've got inflation, deflation, recession and growth, in several combinations) you could find stocks and bonds with a covariance of 0 (because of +ive and -ive periods cancelling out). But good portfolio design seems to involve a balance of things that offset each other in different environments. My takeaway is that there's nowhere near enough information going into mean variance optimisation to describe the dynamics of asset classes against the dynamics of an economy.

[quisp65]

I like the 3 defensive sectors portfolio and it beats my current portfolio that I'm retired on (below in red) in growth & volatility for the 18 years shown.
Both with bonds in blue & without bonds in orange.

I'm just not adventurous on my money to be that different at this point, but this defensive equity portfolio seems worthy to ride out in retirement.

It's Vanguard's 3 defense sector ETFs - 564 stocks total & 0.10% ER - so diversified enough & it's hard to imagine getting screwed in any of those 3 sectors.

That portfolio would give me peace of mind but maybe I missed one of the sectors risks in prior messages.

[Logan Roy]

I like the 3 defensive sectors portfolio and it beats my current portfolio that I'm retired on (below in red) in growth & volatility for the 18 years shown.
Both with bonds in blue & without bonds in orange.

I'm just not adventurous on my money to be that different at this point, but this defensive equity portfolio seems worthy to ride out in retirement.

It's Vanguard's 3 defense sector ETFs - 564 stocks total & 0.10% ER - so diversified enough & it's hard to imagine getting screwed in any of those 3 sectors.

That portfolio would give me peace of mind but maybe I missed one of the sectors risks in prior messages.

I agree. I'm running a version of this portfolio (which I started in August 2021), but I do mix in some broad market exposure, and some gold and TIPS, in equal measure.

What I ultimately found most successful and oddly reliable over all periods was this combination of 1/3rd each Consumer stocks, Healthcare and Energy. And that's the core of what I'm doing in my portfolio. Utilities make nice bond proxies, but tend to be priced accordingly, with low growth. The magic seems to be in having 2/3rds of the portfolio in these more stable sectors, and 1/3rd in something very cyclical – inflation-sensitive. That way you get this rebalancing effect, where you're regularly able to take some profits from Consumer and Healthcare, and buy Energy stocks on some really low valuations. And every once in a while, that part of the portfolio bounces back, and now Energy stocks are topping up everything else.

[Logan Roy]

Mini forward-test update. This thread was started at the beginning of September. I started the portfolio this was based on in August 2021. Three possible implementations.

The least impressive performing (portfolio 1) is closest to how I'm positioned with this idea. Partly due to discomfort with something like a >20% allocation to Energy (which is what's driven things so far) – knowing how much of a drag that could be in the future. But this may be a mistake on my part. But also partly to diversify currencies a bit more, with such a strong dollar in recent times, and potentially more attractive valuations in EAFE and EM. It's really divided into four stocks blocks: Staples and Healthcare (recession), Energy (inflation), and Value, US, EM, EAFE, as a basket of things that should benefit from improving global growth. The whole design is based on rebalancing.

There are still decisions I've not quite ironed out – like whether the exposure to defensives means I could reduce allocation to gold and TIPS (however, knowing stagflation's still lurking, these still feel like good ideas). But, I think there's a reasonable portfolio design lurking in this – if one wanted to use sectors as diversifiers in this way.

https://www.portfoliovisualizer.com/bac ... mbol10=IDU

[smooth_rough]

Compare performance old original VWELX (60/40) vs VPU.

1 year
5 years
10 years
max number of years

Draw your own conclusions.

[GaryA505]

Compare performance old original VWELX (60/40) vs VPU.

1 year
5 years
10 years
max number of years

To get a better risk-adjusted comparison, I would suggest comparing Wellington with roughly 80/20 VPU/cash:
https://www.portfoliovisualizer.com/bac ... tion3_2=20

[smooth_rough]

Compare performance old original VWELX (60/40) vs VPU.

1 year
5 years
10 years
max number of years

To get a better risk-adjusted comparison, I would suggest comparing Wellington with roughly 80/20 VPU/cash:
https://www.portfoliovisualizer.com/bac ... tion3_2=20

Or maybe possible to equal (or do better than) wellington, with blend of VPU and money market fund? Your mileage may vary.

[Logan Roy]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

[smooth_rough]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

VPU is one of the great sector funds. You could blend it any way you want. I was just going for the simple illustration.

[the_wiki]

You guys are really good at predicting the past.

[GaryA505]

You guys are really good at predicting the past.

Yeah, it's easier to predict the past, unless you have a really good crystal ball (not those cheap ones they sell on Amazon and EBay).

[smooth_rough]

You guys are really good at predicting the past.

Wellington has history back to great depression. VPU inception 2004. S&P utilities index 1996. Back-testing has its limitations.

But if you like the characteristics of bonds then you need bonds for your AA because 'reasons'.

[Logan Roy]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

VPU is one of the great sector funds. You could blend it any way you want. I was just going for the simple illustration.

It's funny because until this thread, with the longer-term data, I'd always used Utilities in these models. But in simulations, its inclusion rarely led to optimum results.. The algorithms would include it when you wanted really steady results, but I would say, long-term, it always tended to be a bit of a drag on returns.

But as an alternative to total bond, I think it's compelling.

[Logan Roy]

You guys are really good at predicting the past.

Observing. I find the whole predicting part an unnecessary step.

I think we've covered the limitations of backtesting here.. The way I'd put it: the only thing worse than backtesting too much is not backtesting enough. Backtesting meant many investors went into 2022's stagflation knowing how bonds would respond. Much better position than not knowing. Unless you're one of these faith-based investors, which has always come out on top eventually.

[secondopinion]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

VPU is one of the great sector funds. You could blend it any way you want. I was just going for the simple illustration.

It's funny because until this thread, with the longer-term data, I'd always used Utilities in these models. But in simulations, its inclusion rarely led to optimum results.. The algorithms would include it when you wanted really steady results, but I would say, long-term, it always tended to be a bit of a drag on returns.

But as an alternative to total bond, I think it's compelling.

Remember, risk versus return. I would rather take the slight drag on returns than assume that my portfolio is sufficient without it, especially if I was talking about removing all holdings in bonds and TIPS. Utilities would act like an inflation protected annuity with credit risk (at least, the way I see it).

[secondopinion]

You guys are really good at predicting the past.

Observing. I find the whole predicting part an unnecessary step.

I think we've covered the limitations of backtesting here.. The way I'd put it: the only thing worse than backtesting too much is not backtesting enough. Backtesting meant many investors went into 2022's stagflation knowing how bonds would respond. Much better position than not knowing. Unless you're one of these faith-based investors, which has always come out on top eventually.

I know how they could have responded, so I stayed very short in duration after the rates collapsed in 2020-2021 (depending on the type).

[Logan Roy]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

VPU is one of the great sector funds. You could blend it any way you want. I was just going for the simple illustration.

It's funny because until this thread, with the longer-term data, I'd always used Utilities in these models. But in simulations, its inclusion rarely led to optimum results.. The algorithms would include it when you wanted really steady results, but I would say, long-term, it always tended to be a bit of a drag on returns.

But as an alternative to total bond, I think it's compelling.

Remember, risk versus return. I would rather take the slight drag on returns than assume that my portfolio is sufficient without it, especially if I was talking about removing all holdings in bonds and TIPS. Utilities would act like an inflation protected annuity with credit risk (at least, the way I see it).

How we define an optimal risk vs return is a trickier problem than you'd think. Sharpe and Sortino accommodate the relevant information, but if you let an algorithm optimise for either, it'll just put you in bonds. PV adds a 'minimum annual return'. But then you're only really optimising along 1 axis. As Douglas Adams said: you only really know what you mean when you have to try explaining it to a computer.

The three problems with Utilities, from an optimal portfolio point of view, were (if I recall) a very bad patch around the 1930s, too much stability (all optimal portfolios made heavy use of the rebalancing effect, such that all returned about as high as their top performing component, over 80+ years), and too little scope for long-term profit growth – being rather bond-proxy-like. What it came down to was that Staples and Healthcare give you a more optimal balance of earnings stability and profit growth, while Energy offers very strong inflation protection and opportunities to buy at very low valuations, quite regularly.

At the moment, none of my portfolios based on this work use Utilities. Gold was typically what the algorithm would use to improve Sharpe further. But Utilities come in just behind gold – in this model. Discretionary also packs quite a punch, in terms of optimal diversification benefits and profit growth.

[secondopinion]

80:20 Utilities and gold, or Utilities and Healthcare/Staples, achieves better Sharpe ratios. I'd say you always want to avoid cash. I'm genuinely surprised with the result from Utilities here though.

https://www.portfoliovisualizer.com/bac ... tion6_2=20

VPU is one of the great sector funds. You could blend it any way you want. I was just going for the simple illustration.

It's funny because until this thread, with the longer-term data, I'd always used Utilities in these models. But in simulations, its inclusion rarely led to optimum results.. The algorithms would include it when you wanted really steady results, but I would say, long-term, it always tended to be a bit of a drag on returns.

But as an alternative to total bond, I think it's compelling.

Remember, risk versus return. I would rather take the slight drag on returns than assume that my portfolio is sufficient without it, especially if I was talking about removing all holdings in bonds and TIPS. Utilities would act like an inflation protected annuity with credit risk (at least, the way I see it).

How we define an optimal risk vs return is a trickier problem than you'd think. Sharpe and Sortino accommodate the relevant information, but if you let an algorithm optimise for either, it'll just put you in bonds. PV adds a 'minimum annual return'. But then you're only really optimising along 1 axis. As Douglas Adams said: you only really know what you mean when you have to try explaining it to a computer.

The three problems with Utilities, from an optimal portfolio point of view, were (if I recall) a very bad patch around the 1930s, too much stability (all optimal portfolios made heavy use of the rebalancing effect, such that all returned about as high as their top performing component, over 80+ years), and too little scope for long-term profit growth – being rather bond-proxy-like. What it came down to was that Staples and Healthcare give you a more optimal balance of earnings stability and profit growth, while Energy offers very strong inflation protection and opportunities to buy at very low valuations, quite regularly.

At the moment, none of my portfolios based on this work use Utilities. Gold was typically what the algorithm would use to improve Sharpe further. But Utilities come in just behind gold – in this model. Discretionary also packs quite a punch, in terms of optimal diversification benefits and profit growth.

I do not discount that; utilities are not stock-like normally in their growth. However, if I knowingly want stock-like growth, I would not be buying a bunch of utilities. Different objectives call for different investments. What might be good for a 100 year portfolio does not equate to a 50 year portfolio or a 20 year portfolio or a 5 year or a 1 year.

[Hector]

Procter & Gamble, Coke and Pepsi makes up 1/3rd of Consumer Staples!

[quattro73]

Just take Wellington at 80% and MCD at 20%

[Logan Roy]

Procter & Gamble, Coke and Pepsi makes up 1/3rd of Consumer Staples!

I think that's fine. When you look at a business like P&G, and how many brands across different consumer sectors it owns, it looks like a whole private equity portfolio. It's one reason I was always a little dubious about there being a size effect.

[Muffin Master]

60/40 (timing) vs 100% invested (no timing)