Link to the paper on SSRN:
https://papers.ssrn.com/sol3/papers.cfm ... id=2856435
The biggest contribution here is to extend the dataset backwards, looking at older data, and provide a perspective on some of the points made previously in the literature, particularly the famous Forton and Rouwenhorst (2006) that noted "equity-like returns" and the more skeptical Erb and Harvey (2006). Obviously there are potentially some data issues with figures that old and there is a question of how relevant it might be. But more data can help.
Following with prior precedent they construct equal-weighted hypothetical portfolios of commodity futures. (They also look at equal risk-weighted portfolios, but results and conclusions are much the same as for the equal-weighted.) For an individual investor, directly investing in commodity futures is a bit unwieldy based on contract size and the need for diversification, so most would use a fund, which would entail costs and an ER, a drag relative to the performance noted in studies like these.
Here's a look at the cumulative returns over time:
Here the returns in different conditions are examined:
Low inflation is defined as below the full-period mean; high is above. What's shown is excess average (arithmetic mean) return over cash. Because of the volatility, that 4.8% for the full sample is 3.3% in geometric mean terms. Recall that cash slightly beat inflation over the full period so that's a respectable real return.
Big difference between the different regimes: backwardation vs. contango, high vs. low inflation, expansion vs. recession. Interestingly the beta with bonds is not very negative under any of the conditions. But that is a large number of months for each; many justify commodities as a hedge for inflation shocks and those periods are rarer than any of these broader categories. Also note that there's a positive beta with stocks, just not a large one there either. So much for claimed negative correlation with stocks. On the flip side, those expecting commodities to crash every time stocks do, like they did in the financial crisis, may not be right either.
As noted elsewhere, excess spot return seems to be kind of negatively related with interest rate adjusted carry return. It holds for the alternate decomposition, not shown here, of spot return vs. roll yield.
I'm not sure what you might have expected, but the number of months of contango and backwardation are kind of similar, maybe not an obvious result based on what you see of some studies over periods that may be heavier in one (particularly the pro-commodities looks that might have focused more on the backwardation periods). And there were positive returns in excess of cash on average, even under the contango months, just lower than when in backwardation.
Some results when integrated in a portfolio:
A 10% allocation to commodities based on this history would have improved a 60/40 portfolio's Sharpe ratio by... 0.03. Before costs. Actually, notice how the bond Sharpe ratio is probably lower than you're used to seeing. Anyway, I think this table mostly speaks for itself.
All in all, I don't think this changes the mind of those of the world-has-changed, skating-where-the-puck-was, financialization-killed-it school of thought. After all, what does extending backward for more even-really-older data really say? On the other hand, I don't agree totally with that line of argument, and I think what we have here is a broader look at commodities futures under a wider range of conditions, which is not a bad thing.
Personally I don't see any particularly compelling evidence that one should be drastically offended or scared off by commodity (futures) allocations in Vanguard's Managed Payout fund or in some target retirement series funds from others. Also I don't see much strong, compelling evidence for portfolio inclusion for a retail investor.