Week 40 Day 2: Volatility Drag: Why Losses Hurt More Than Gains Help

Start with $100. Gain 20% -> $120. Lose 20% -> $96. You did NOT break even. You lost $4. Now try: gain 50%, lose 50%. $100 -> $150 -> $75. You lost $25. Bigger swings, bigger drag. The same average return (0%) produces different outcomes depending on the magnitude of the swings. This is why volatility matters even if the average return looks fine.

The volatility drag formula: Geometric return (what you actually earn) is approximately equal to Arithmetic return minus (Variance / 2). With VTI: arithmetic average return approximately 11%, standard deviation approximately 16%, variance = 16%^2 = 0.0256. Volatility drag = 0.0256 / 2 = 0.0128 or 1.28%. Geometric (actual) return approximately 11% - 1.28% = 9.72%. With Bitcoin: arithmetic average return approximately 100%, standard deviation approximately 70%, variance = 70%^2 = 0.49. Volatility drag = 0.49 / 2 = 24.5%. Geometric return approximately 100% - 24.5% = 75.5%. The drag ACCELERATES with volatility. Doubling the standard deviation quadruples the drag (because drag depends on variance, which is sigma squared). This explains why high-volatility assets (crypto, leveraged ETFs, individual stocks) often underperform their headline average returns: the drag between the arithmetic average (what sounds impressive) and the geometric average (what you actually earn) is massive. Example: a 3x leveraged S&P 500 ETF has 3x the daily return but approximately 9x the variance drag. Over long periods, 3x leveraged ETFs often underperform the unleveraged S&P 500 despite having 3x the daily return -- volatility drag eats the leverage. The practical lesson: moderate, consistent returns beat wild, volatile returns over time. VTI's boring 10% beats Bitcoin's exciting 100% (adjusted for actual compounding, risk, and the probability of holding through the crashes).

The volatility drag is formally derived from the relationship between arithmetic and geometric means of a lognormal distribution: if arithmetic return = mu and variance = sigma^2, the geometric return g = mu - sigma^2/2. This is exact for continuously compounded returns and approximate for discrete returns. The practical implication: two assets with identical expected arithmetic returns but different volatilities will produce different long-run wealth. The lower-volatility asset will compound to a higher terminal value. This is the mathematical basis for the low-volatility anomaly (Baker, Bradley, and Wurgler, 2011): low-volatility stocks have historically delivered higher risk-adjusted returns AND higher absolute returns than high-volatility stocks -- contradicting the CAPM prediction that higher risk (volatility) should be rewarded with higher returns. The explanation: investors overvalue high-volatility assets (due to lottery preferences, overconfidence, and leverage constraints) and undervalue low-volatility assets, creating a persistent mispricing. For portfolio construction, the volatility drag insight supports: (1) diversification (which reduces portfolio variance below the weighted average of component variances, capturing the 'diversification benefit' of the drag reduction), (2) rebalancing (which mechanically sells high-volatility positions after they appreciate and buys after they decline, reducing the effective variance experienced), and (3) avoiding leverage (which amplifies variance by the square of the leverage ratio, dramatically increasing the drag).