I went through the exercise of adding gold to Sharpe's World Bond Stock (WBS) portfolio. (This was a thought experiment... I don't currently hold any gold in my own portfolio.)
TL;DR: Publicly investable gold allocation is ~0.2% when added to Sharpe's WBS portfolio. (See the "4-Fund + Gold" tab
here.)
Hopefully this answers the "Do I need gold?" question for those who think that Sharpe's portfolio is a good idea. You probably don't need gold unless:
- you have already addressed larger gaps in Sharpe's WBS portfolio (e.g., high-yield corporate bonds), AND
- your portfolio is large enough that 0.2% of it is a non-negligible enough amount to bother about. (I'd only start to bother about it when 0.2% of my portfolio is a meaningful fraction -- say 5% or more -- of my annual expense.)
Okay, on to the details of how to find the right market cap for gold...
To start with, I know that we get some exposure to REITs just by holding VTI / VTSAX. I wondered if the same was true for gold. Turns out not to be true -- i.e., VTI's holdings do not contain GLD, IAU, AAAU, etc. I verified this by searching VTI's holdings
here. (A possible reason for VTI not covering these gold investments is that the gold ETFs are structured as "grantor trusts" which are not considered investment companies, and therefore(?) not tracked by the CRSP US Total Stock Market index.)
Having double-checked now that gold is indeed a gap in the WBS portfolio, we now move onto the problem of determining the size of the gap by finding the right market cap to use for gold...
(Aside: I should pause here to say that I wholeheartedly agree with the argument that gold is of limited industrial use and that stored gold produces nothing. But I disagree that this completely eliminates gold's validity as an investment. If the global market ascribes value to something, then it is worth considering, however mad it might appear at first glance... it's just a question of making sure that we don't
over-allocate to those bets. As we shall see, it is rather easy to over-allocate to gold if we aren't careful about what we measure. But if we measure the right thing, the result is reasonable -- 0.2% is a fairly conservative amount to allocate for a speculative bet.)
Simply googling for "gold market cap" gives an answer of ~$10T. Sharpe World Bond Stock market cap is currently ~$113T (see the "4-Fund" tab
here), so adding in ~$10T gold would result in a ~8% allocation to it.
But... $10T is in fact
not the market cap we're after if we use Sharpe's methodology. We're actually only interested in the
publicly investable portion of all the world's gold. It turns out that only ~2% of the $10T of the world's gold is held in publicly investable securities -- gold ETFs, mutual funds, and the like (e.g., GLD, IAU, AAAU) -- while the rest of the gold is held in jewelry, industrial use (fabrication), physical bullion / coin, and official holdings (e.g., central bank reserves) (
source). When determining the allocation for gold ETFs in one's portfolio, I believe Sharpe's methodology would count only the market cap of gold held in ETF-investable trusts which 1) is more precisely representative of what we are buying (e.g., GLD), and 2) has a stronger case for market efficiency. (Please correct me if I'm wrong on this crucial point!)
The World Gold Council releases monthly gold ETF market caps
here. I had to make a (free) account to access the data, and will track it going forward in my
WBS spreadsheet. (Sharpe's original "2-Fund" and "4-Fund" WBS portfolios are still preserved in the sheet; gold is added only to the "N-Fund" variant and the "4-Fund + Gold" variant.)
Btw, the nuance here about using only the publicly investable portion of gold is very reminiscent of searching for "bond market cap", and finding an answer of $100T or more (from
SIFMA data), while Sharpe's methodology (i.e., using the market cap of the indexes actually being invested in) works out to less than half of that (~$47T) for bonds. The reason is that the SIFMA market caps include things like private placement bonds that are inaccessible to the retail investor, just as the total gold market cap contains private jewelry, central bank reserves, etc.
This story is also similar to how WBS's real estate exposure is provided by (only) market-weighted public REITs, as opposed to over-weighting public REITs to try to match the market cap of private REITs and privately-owned real estate.
My main takeaway from Sharpe's original WBS portfolio as well as this extension to gold is:
buy what you measure, and measure what you buy.
P.S., I wonder how to extend this methodology to crypto. I found data sources for the overall crypto market caps
here, but have not yet figured out how to measure the "investable" portion -- is it the whole market cap, or are there investor-inaccessible portions blocked off by certain entities? Any suggestions on how to figure this out? (edit: Found a recent attempt at measuring "free float crypto"
here. My original market cap source claims to use
"circulating supply" to measure market cap, but its numbers for Bitcoin total supply was equal to the circulating supply, so I don't think any free float correction was actually done.) But note that, even with the whole market cap, Bitcoin would be no more than a 0.4% allocation of WBS + Bitcoin portfolio. So again, it's ignorable for most investors until their portfolio grows very large and they have addressed the larger gaps in the WBS portfolio, particularly in junk bonds.
P.P.S., This post was a thought exercise to see whether Sharpe's market cap methodology extends reasonably to other asset classes. It seems to do a reasonable job that aligned well with my own previously-held beliefs and preferences (e.g., I would balk at an 8% allocation to gold), but I suppose that your reaction to the results and methodology will depend on your own beliefs about things like gold and crypto. I am not invested in either gold or crypto myself. (I won't invest in crypto until I am confident that some of the practical details (e.g., tax reporting and beneficiaries/inheritence) work smoothly and conveniently.)
edit: fixed a link, grammar
AA = global stocks & bonds @ market weight (~60/40); EF = i-bonds; WR = -PMT(1%, 100-age, 1, 0, 1)