Asset Classes by Personality

I

Meet the players

In Part I we built the toolkit: variance, the Kelly Criterion, geometric vs. arithmetic returns, systematic risk, correlation, and tail events. Now we apply those tools to eight asset classes that most investors actually consider. Each has a distinct personality — a different balance of return, variance, drawdown depth, and correlation behaviour.

II

Returns — the headline that hides everything

If you only look at returns, the ranking seems obvious. Bitcoin dominates, tech outpaces the broad market, real estate trails the equity ride, and the “safe” assets — bonds, T-bills, gold, commodities — look pedestrian by comparison. But Part I taught us that the headline return is the arithmetic average — the number that lies. What matters is the geometric compound rate, net of the variance tax.

The chart below shows the approximate growth of $10,000 invested in each asset class from 2015 to 2025. Log scale, so equal vertical distances = equal percentage moves. On a linear scale, Bitcoin would make everything else look like a flatline.

Over Jan-2015 to Dec-2025 — a window that starts near a Bitcoin trough and well into the post-GFC equity recovery, so be cautious about extrapolating — Bitcoin’s compound annual growth rate came in around 57%. The NASDAQ 100 delivered roughly 19%annualized (total return). The S&P 500 came in around 14% (total return). Real estate, including rental income, delivered approximately 8%. Gold compounded at about 11% over the same window — its 2024–25 surge on geopolitics and central-bank buying lifted the run. Broad commodities returned roughly 5% annualized despite a brutal 2020 and a 2022 spike. T-bills clocked approximately 2%. 10-year Treasuries approximately 0%nominal — the 2022 bond bear erased almost a decade of coupons. Shift the start date by a single year and most of these figures move several points; Bitcoin’s swings 15–25 points. Those are the numbers people share at dinner parties — usually without the window.

Now let’s look at what those numbers cost.

III

The variance tax — what returns cost

From Part I: variance silently destroys geometric growth. Two assets can have the same arithmetic return but wildly different geometric outcomes. The formula linking them is approximately:

How big is the variance tax for each asset?

This is the variance tax. The higher your volatility, the more of your arithmetic return gets consumed before it reaches your actual wealth. T-bills pay essentially no variance tax — their 0.5% σ means σ²/2 rounds to zero. Bonds and real estate sit in the low-tax band. Gold and broad commodities pay a meaningful but bounded tax. Equities pay more. Bitcoin pays the highest of any major asset.

On an annual basis Bitcoin’s arithmetic mean has run ~80–100% with σ ≈ 70–90% over 2015–2025, so σ²/2 alone consumes roughly 25–40 percentage points of the headline return before it ever reaches your geometric wealth.

And σ²/2 is the Gaussian approximation. Equities, real estate, gold, and bonds are roughly log-normal with modest excess kurtosis. Commodities have visible fat tails around supply shocks. Bitcoin is genuinely fat-tailed — kurtosis well above 3, with multiple 5σ-equivalent moves per year. For fat-tailed series, σ²/2 is a lower bound on the geometric drag; the true drag is worse than the formula suggests.

Real estate’s advantage is visible here: roughly 8–10% total return at a measured annualized volatility around 10% (Case-Shiller / NCREIF), so the variance tax is small and almost all the arithmetic return converts to geometric growth. Important caveat: that 10% figure is structurally understated because appraisal-based and infrequent marks smooth the series. Mark-to-market like equities and real-estate volatility is closer to 15–20%. The variance-tax win is real, just smaller than the headline.

Gold matches the S&P 500 on volatility (~15%) but with a different return driver — when you net out variance tax, gold and the broad market end up in similar Sharpe territory despite looking very different on the monthly chart. Broad commodities are noisier still (~18% σ) and the geometric return suffers accordingly.

Treasuries sit at ~7% σ — low for a risk asset, but not zero. The 2022 bond drawdown happened precisely because that 7% number is annualized over a long history that included 40 years of falling yields; in the rate-spike regime the realized σ jumped, and the duration drag landed on principal.

The S&P 500 sits in a comfortable middle. The NASDAQ 100 pays a higher tax than the S&P for its higher returns. Bitcoin pays the highest tax of all — but its gross returns have been so extreme that even after the variance drag, the net geometric return has been extraordinary. The catch: that only works if Bitcoin’s future arithmetic returns remain extreme. If they moderate toward equity-like levels while volatility stays elevated, the variance tax alone would make it a mediocre or negative geometric investment.

IV

Drawdowns — can you survive the valley?

Part I’s Kelly Criterion depends on ruin tolerance: how much can you lose before you’re knocked out of the game? Here is where the eight asset classes diverge most dramatically.

The NASDAQ 100 lost 83% from the 2000 dot-com peak to the 2002 bottom — and took over 15 yearsto recover. Bitcoin has experienced drawdowns of 80%+ at least three times. An investor who put 100% of their capital into Bitcoin at the 2021 peak saw it drop to $16,000 in 2022 — a 76% decline. The S&P 500’s worst was a 50% crash during the 2008 financial crisis, recovering in roughly 5.5 years. Broad commodities suffered a multi-year ~50% decline during the 2014–2020 cycle as oil, metals, and agriculture all detrended together.

Real estate’s maximum drawdown of roughly 33% during the 2008 housing crisis understates the real damage, because real estate is typically leveraged. A 33% price decline on a home bought with 20% down wipes out 165%of your equity — you’re underwater, owing more than the asset is worth.

The defensive trio — Treasuries, T-bills, and gold — looks safer, but each has its own valley. Treasuries dropped roughly 25% from 2020 highs to 2023 lows as rates spiked from near-zero to 5%+ — the worst bond drawdown in 40 years. Gold drew down ~22% in the 2020–22 consolidation after the COVID rally; on a longer window (2011–2015) gold fell 45% from its post-GFC peak. T-bills hold the unique position of zero nominaldrawdown, but inflation-adjusted purchasing power dropped roughly 30% over 2021–23 — a real loss that doesn’t show up on any nominal chart.

Why max drawdown is a survival filter, not a risk metric

The maximum drawdown is not a risk metric. It is a survival filter. If you cannot hold through it — financially and psychologically — you have no business being in that asset class.

Applying Kelly’s logic: for an asset with 80% max drawdown, even a very high edge should lead to fractional Kelly sizing. No more than 5–10% of total portfolio in a single asset that can lose four-fifths of its value. This is not a philosophical preference — it is a mathematical constraint on geometric growth.

V

Correlation — the diversification question

From Part I: diversification works only when positions are uncorrelated. Two perfectly correlated assets provide zero diversification benefit — you’re just doubling your bet. So how do our eight asset classes correlate with each other?

Why S&P + NASDAQ is one bet, not two

The first finding from the matrix: NASDAQ 100 and S&P 500 are highly correlated (~0.90). Holding both provides almost no diversification benefit — they crash together and rally together. You’re paying for two tickets to the same ride.

Real estate, gold, and Treasuries — the three real diversifiers

Real estate has low correlation to equities (~0.25–0.35), making it a strong diversifier in normal markets. Its returns are driven by local housing supply, rental demand, and mortgage rates — largely independent of daily stock-market moves.

Gold sits closer to zero against equities (~0.10) and zero against Bitcoin (~0.20), so it adds genuine independent variance to a stock-heavy portfolio. Where it shines is regime change: gold has been the asset that does well precisely when stocks and bonds fail together, e.g., the 2022 inflation shock and the 2024–25 geopolitical surge.

Treasuries are the textbook diversifier — negatively correlated with equities (~−0.15 to −0.20) in growth scares: when stocks fall on slowing growth, bonds rally on rate cuts. But the textbook breaks in inflation shocks: 2022 saw both equities and Treasuries fall together as the Fed hiked aggressively. The negative correlation is regime-dependent, not structural.

T-bills are uncorrelated with everything — they are, by definition, the risk-free asset. Their diversification benefit is not in their variance contribution (which rounds to zero) but in their optionality: dry powder you can deploy when other assets crack.

Bitcoin’s diversification value is eroding

Bitcoin’s correlation story is more nuanced. Before 2024, it was genuinely uncorrelated with equities — correlation near zero. This made it an attractive portfolio diversifier despite its extreme volatility. But as institutional investors poured in through Bitcoin ETFs, the correlation jumped to 0.5–0.87. Bitcoin increasingly behaves like a high-beta tech stock rather than an independent asset. Its diversification value has eroded precisely as more people have adopted it for that purpose.

Commodities and gold — related but not identical

Gold and broad commodities show a moderate correlation (~0.40) because both are real assets, but the drivers differ: gold tracks monetary debasement and risk-off flows; commodities track industrial demand cycles and weather / supply shocks. Holding both gives you exposure to two distinct regimes — but understand the overlap, especially when commodity supercycles drive both up together (2003–2008, 2021–2022).

VI

Risk-adjusted returns — the Sharpe ratio reality check

The Sharpe ratio measures return per unit of risk: how much excess return (above the risk-free rate) did you earn for each unit of volatility you endured? A higher ratio means more efficient return generation. T-bills are excluded from this chart by definition — they are the risk-free rate.

Real estate occupies the most efficientposition on this chart: solid returns with the least volatility among risk assets. Its Sharpe ratio of approximately 0.7–0.8 is the highest of the eight, meaning each unit of risk buys more return than any other asset class here. Gold trails close behind at roughly 0.5–0.6 — surprisingly competitive once you net out its variance. The S&P 500 sits at roughly 0.65. Bitcoin, despite its explosive returns, lands at around 0.5–0.6 because its volatility is so extreme.

The NASDAQ 100’s Sharpe ratio is similar to the S&P 500’s — the extra return comes with proportionally extra risk. You’re not getting a free lunch from tech concentration; you’re getting a more levered version of the same ride. Broad commodities post the weakest Sharpe of the risk assets (~0.15–0.25): you take equity-like volatility for T-bill-like long-run return. The case for commodities isn’t standalone risk-adjusted return — it’s the diversification they offer in inflationary regimes.

Treasuries’ Sharpe over 2015–25 came in near zero — coupons earned over the decade were almost exactly offset by the 2022 principal crash. Run the same calculation over 1985–2020 and Treasuries’ Sharpe was >1.0. The lesson: bond Sharpe is regime-dependent. The 40-year disinflation tailwind is gone.

The Sharpe ratio answers the Kelly question indirectly: given the volatility you must endure, is the edge actually worth concentrating into?

For most investors without genuine stock-picking edge, the data argues for broad equity exposure (S&P 500), a real estate allocation for diversification and efficiency, a gold allocation for regime-change insurance, modest Treasury or T-bill exposure as dry powder, a small Bitcoin allocation for asymmetric upside (sized by Kelly logic — small, because the variance is enormous), and caution about NASDAQ 100 concentration unless you have a specific thesis about technology’s future that the market hasn’t priced in. Commodities are an optional inflation hedge with low standalone Sharpe — useful to hold, in modest size, if you don’t already have inflation protection through real estate or gold.