Gyroscopic Investing

A Pile of Numbers
I was looking around for different kinds of risks to take. That's the core idea behind any alternative investment, that you'll be taking different risks than normal stocks and bonds, which should improve your average returns. Theoretically, if some identifiable segment of the stock market worked differently than the others, then you could improve your risk adjusted returns by overweighting it. This is how factor investing works and why people overweight REITs. So let's look at all the sectors. I'm using VGSIX instead of XLRE here because XLRE data only goes back to 2015, but it should give us a decent idea.
A couple things stand out here. One is that most sectors have a fairly high R^2. So you would be better off buying whatever factors those sectors decompose to through a factor fund. The aggressive sectors (Consumer Discretionary, Tech, Financials, and Industrial) have higher R^2 than the defensive sectors (Consumer Staples, Health Care, Real Estate) and the commodities producers (Energy, Materials, Utilities). Real Estate is about 2/3rds explained by the factors and has a lot of negative alpha, so it probably isn't that great as a diversifier. Maybe if it had countercyclical behavior, it would be a good idea.

But the sector with the least R^2 is Utilities and not by a little bit. It is less than half explained by the stock and bond factor model, with 9 factors total. To put that in perspective, let's regress a whole bunch of assets.
Some of these assets are well known, but let's go over some of the more obscure ones:

DFSVX: The premier and longest running academic driven small cap value fund.
VWELX: Vanguard's Wellington Fund, a conservative large cap value stock and bond mix.
GNMA: Mortgage Backed Securities
USO: Petroleum Futures, so a major commodity fund.
MERFX: A merger arbitrage fund, so a market neutral hedge on whether companies will merge.
SOYB: Soybean Futures, rolled quarterly.

At an R^2 of 36.5 over the last ten years, XLU is standing right between USO and MERFX. Based on this R^2, it would be reasonable to guess that this is some weird alternative, maybe a commodity fund. But no, it is just a market float weighted index of Utility stocks. That is remarkable. Especially since it has a negative alpha of .33 annually, while the S&P 500 has a negative alpha of .28 in the same period.

Okay, but maybe this is just a weird quirk of the regression. Let's look at the correlations
Looks pretty independent to me. It has some correlations to consumer staples and total bonds, but not much to the core PP assets. Still, is there a fundamental reason why this should happen?

Why Would This Happen

Utilities are different. This is not a conjecture, this is a legal fact. Based on exciting supreme court cases like Federal Power Commission et al v. Hope Natural Gas Co. and Bluefield Water Works and Improvement Co. v. Public Service Commission of West Virginia , utilities get their profit determined by rates. IE, if a utility isn't making a profit, it can appeal this fact in court and raise prices, which the customers will inevitably pay. That's the benefit of monopoly. The converse is that if a utility makes too much of a profit, the Public Utility Commission can cut their returns. How does the profit margin get determined? By looking at other comparable investments (stocks and bonds) and saying "about that much".

So utilities aren't safe because of the market, utilities are legally distinct structures. In a way, they are a big insurance product. Governments don't have to try to run big public projects, instead they let a company collect some money from their taxpayers. In exchange, if something goes wrong with the utility, the investors are on the hook. This can't be a normal insurance product, since most of what could go wrong with a utility is uninsurable. Natural Disasters, regulatory issues, and long tail risks like if we discover that power lines cause fires (who knew?) are all outside what normal insurance can handle. Utilities are big enough projects that the reinsurance market may not be able to handle them even if it wanted to.

Investment Implications
So what does this mean for the Gyroscopic Investor? There are a lot of independent risks, why should this one matter?

• 1. Utilities have few if any fees, unlike most alternatives.
  2. Utilities are fairly independent of the big 3 risks (Gold, Stocks, and Bonds).
  3. Utilities are especially independent of SCV investments, which matters for GB types.
  4. If you're worried about bonds or currency devaluation, more real assets would make you feel better.

So let's look at some portfolios, starting with our backtesting options. XLU is just large cap utilities, adding the small caps should improve performance by capturing more of the market. VPU, the Vanguard Public Utilities etf only goes back to 2004, but Franklin Utilities goes back to 1948 and is completely unremarkable, a total closet index fund.
Now let's look at a simple swap, with FKUTX for VOO in the Golden Butterfly.
Lower total returns, but lower standard deviation, max drawdown, and a .2 drop in US market correlation. So if you're worried about US markets, this is a way to decouple while still investing domestically.

Next, let's consider a weird 60/40. Gold is basically a long duration, real rate, 0 coupon bond, so 20% Gold, 20% Long Bond, 20% Utilities, 20% US SCV, 20% International SCV is just a normal thing, right?
Doing pretty okay. Finally, let's take the big leap. 40 years has been a good long time for bond returns, will utilities do well enough that we can drop them?
Well, looks like we picked up a little more risk than return. Still, this was a very good period for bonds and there are very good reasons to be nervous about them. If you're thinking of diversifying with REITs or swapping out some long bonds, use utilities instead.

You might want to backtest beyond a period of only generally falling rates (which we have had for the last 35 to 40 years) since utility performance was drastically different (and drastically worse) during periods of rising rates and rising inflation (like the mid-1960s to 1980) than it was during the period of (generally) falling inflation and falling rates we have been in since the very early 1980s.

Use either FKUTX or the Fama-French "utility" category data and you will see that utilities got absolutely killed from 1965 to 1980.

Did they do worse than long bonds? Almost certainly not (they definitely did worse than regular stocks or an index like the S&P 500 or the TSM, though).

EDIT: I just went and checked and from 1965-80 LTTs did outperform FKUTX but underperformed the market-cap weighted utility index data from Ken French's data library; LTTs gave a CAGR of around 2.20%; FKUTX gave a CAGR of 1.41%; and the French "utility" category simulated index came in at 4.97% CAGR. FWIW the S&P 500 TR over this same period had a CAGR of 7.04% and T-Bills had a CAGR of 6.33%. All of the above numbers are nominal (i.e. not adjusted for inflation....which needless to say was rather high over this period).

I have monthly REIT total return data back to 1963 and thus I can tell you for a fact that utilities did far worse than REITs for the rising rate period starting in either 1965 or 1966 and going to 1980 or 1981.

Also, utilities fell (albeit generally not as much as a total stock index) in years like 2008, 2002, 2001, 1974, and 1966 when long bonds had returns that were anywhere from flat-ish to 25%+; you want at least one component of your portfolio to rise reliably when equities fall and LTTs--counting reinvested coupon income--have done that (even during rising rate periods the years 1957, 1960, 1962, 1966, 1970, 1973, 1974 were years where we had stock declines and LTTs provided positive nominal returns; in 1977 and 1981 LTTs were basically flat when stocks had slightly bad years both times); utilities often have not.

D1984 wrote: ↑Thu Nov 11, 2021 11:16 pm You might want to backtest beyond a period of only generally falling rates (which we have had for the last 35 to 40 years) since utility performance was drastically different (and drastically worse) during periods of rising rates and rising inflation (like the mid-1960s to 1980) than it was during the period of (generally) falling inflation and falling rates we have been in since the very early 1980s.

That backtesting makes sense, but I have worries about the long term data quality. Utilities (and especially electric utilities) were bubble stocks in the 20s and had a long, complicated legal fight through the 40s around the Public Utility Holding Company Act of 1935. State regulation and technological aging (could have) changed the nature of the asset, in addition to the highly leveraged public utility holding companies are in the return data. Those are not really comparable to what is present today.

If the tech platform giants end up utilities, that would be history repeating. The FAANG PEs are certainly unjustifiable if they end up rate takers and utility companies had a CAPE over 60 (briefly) in the 20s. What I would want is a data series separating the rate regulated industries from the companies which eventually ended up rate takers. GICS classification is by industrial activity, which is (I am guessing) is not as useful. The Fama-French data does the same. I also don't have a convenient setup to use that data, but I should fix that.

Most of that data exists, for instance in https://cowles.yale.edu/sites/default/f ... 03-all.pdf here, but I have yet to find it in convenient form. I will probably end up digitizing what I can myself, if I get frustrated enough.

D1984 wrote: ↑Thu Nov 11, 2021 11:16 pm
EDIT: I just went and checked and from 1965-80 LTTs did outperform FKUTX but underperformed the market-cap weighted utility index data from Ken French's data library; LTTs gave a CAGR of around 2.20%; FKUTX gave a CAGR of 1.41%; and the French "utility" category simulated index came in at 4.97% CAGR. FWIW the S&P 500 TR over this same period had a CAGR of 7.04% and T-Bills had a CAGR of 6.33%. All of the above numbers are nominal (i.e. not adjusted for inflation....which needless to say was rather high over this period).

I have monthly REIT total return data back to 1963 and thus I can tell you for a fact that utilities did far worse than REITs for the rising rate period starting in either 1965 or 1966 and going to 1980 or 1981.

That is consistent with what I'd expect. FKUTX had higher fees and probably worse performance in the pre-Vanguard era, I was just using it for some quick and dirty backtesting. If the French "utility" category was still significantly non-correlated to the overall stock market and poorly explained by the factor model, then I would say they are diversifiers. They are still not as good as bonds, but they could qualify as a "liquid alternative" without the fees, active management, and general problems that plague the category.
Eyeballing Schiller's CAPE ratio for utilities in the late 60s puts them at a CAPE over 30 and higher than the Industrials CAPE. If utilities are stock like assets (which they are), then CAPE matters. This is relevant to today, since the utilities sector is at a significantly lower PE than the overall market.

I also doubt the data quality for REITs in this time period. They were very new, small, and weird in the time period. They also have a better explanation in the factor model and have high exposure to size and value which did exceptionally well at the time. But we can now buy those factors directly.

D1984 wrote: ↑Thu Nov 11, 2021 11:16 pm
Also, utilities fell (albeit generally not as much as a total stock index) in years like 2008, 2002, 2001, 1974, and 1966 when long bonds had returns that were anywhere from flat-ish to 25%+; you want at least one component of your portfolio to rise reliably when equities fall and LTTs--counting reinvested coupon income--have done that (even during rising rate periods the years 1957, 1960, 1962, 1966, 1970, 1973, 1974 were years where we had stock declines and LTTs provided positive nominal returns; in 1977 and 1981 LTTs were basically flat when stocks had slightly bad years both times); utilities often have not.

Negatively correlated is better than uncorrelated and weakly correlated. So they wouldn't replace long bonds. But utilities could be a useful supplement and diversifier, especially for factor investors like the GB.

It has been a bad eight months for the permanent portfolio. Well, a good month and then a bad half.

VPU (the Vanguard utilities sector etf) closed at $146.16 when I posted last. Today, it closed at $152.34. It paid 3 dividends since, totaling $3.163. With reinvestment, that is a return of ~6.4% in eight months. All PP assets are down in real terms, only cash is up in nominal. SCV is down for anyone with the Golden Butterfly.

After consideration, I settled on a portfolio in November. I can't fully implement it, my workplace retirement plan lacks SCV funds, but I am using market float weight funds there instead. It is:
1/5 US Utilities
1/5 Gold
1/5 Long Treasuries
4/25 US Small Cap Value
4/25 exUS Developed Small Cap Value
2/25 Emerging Value

Since there aren't good products for international utilities investing, I decided to tilt my SCV investing away from the US. Of my stocks, I am split 3/5 US, 1/3 Utilities. Two backtests, one with cash and one without.
The majority of the extra return is because of extra risk. But the Sharpe and Sortino Ratios are still higher. I'm not willing to move too far away from 1/N portfolios for fear of data mining, so I'm going to stick with my unoptimized weights. I may leverage the thing with treasury futures and the select sector future, if I can get enough money together (XAU is 70k a contract!). No more than 50% leverage, probably 35% as a target.