Volatility Drag & Volatility Pumping

Volatility Drag Is the Price of Constant Risk, Not a Proof Against Leverage

The usual critique of leveraged portfolios starts with a true statement and ends with a false conclusion. The true statement is that volatility drag is real: because compounded returns are geometric, variance mechanically lowers long-run growth. The false conclusion is that leverage therefore should be avoided altogether. In fact, the existence of volatility drag says only that leverage must be sized, not abandoned.

In continuous-time portfolio theory, the optimal risky weight remains positive whenever excess returns justify it; what changes with higher volatility is the optimal scale of that exposure. Daily rebalancing fits naturally into this logic. It enforces a constant risk preference through time, preventing leverage from drifting upward after losses or collapsing after gains. The drag induced by reset is thus not pointless decay, but the necessary cost of keeping portfolio risk state-independent.

A more detailed write-up on volatility drag can be found in the first post and the more detailed write-up on the value of liquidity can be found in the fifteenth post.

Volatility drag is not an argument against leverage

Volatility drag is a statement about geometric compounding, not a theorem that leverage is irrational. In the standard continuous-time setup, let the risky asset follow

\(\frac{dS_t}{S_t} = \mu\,dt + \sigma\,dB_t, \qquad \frac{dM_t}{M_t} = r\,dt.\)

and let the investor hold a constant risky weight x, financed against cash when x>1. Then wealth evolves as

\(\frac{dW_t}{W_t} = \bigl[r + x(\mu-r)\bigr]dt + x\sigma\,dB_t.\)

By Itô’s lemma,

\(d\ln W_t = \left[r + x(\mu-r) - \frac{1}{2}x^2\sigma^2\right]dt + x\sigma\,dB_t.\)

So the geometric growth rate is

\(g(x) \equiv \frac{1}{dt}\,\mathbb{E}[d\ln W_t] = r + x(\mu-r) - \frac{1}{2}x^2\sigma^2.\)

The final term is the volatility drag. But its presence does not imply that the optimal risky allocation is zero. It implies only that the benefit of leverage is linear in x, while the compounding penalty is quadratic in x.

The correct conclusion is therefore not “never leverage,” but “there is a finite optimal leverage, and excess leverage is self-defeating.” In the same framework, the Kelly optimum which maximizes the geometric growth rate of capital is

\(x_{\mathrm{Kelly}} = \arg\max_x g(x) = \frac{\mu-r}{\sigma^2}.\)

while the more general CRRA optimum is

\(x^*(\gamma) = \frac{\mu-r}{\gamma\sigma^2}.\)

So using less than Kelly is not a rejection of leverage; it is just the expression of a higher risk-aversion parameter. That is exactly the same logic behind fractional Kelly.

This also gives the condition under which leverage still improves long-run growth. In any case, the mere existence of drag does not imply “do not lever.” It implies “do not over-lever.” The drag term is a curvature term, not a prohibition.

Why daily rebalancing can make perfect sense

Daily rebalancing is best understood as the operational form of a state-independent risk preference. In the CRRA setup used, the optimal risky fraction does not depend on wealth. That homogeneity property is one reason the model is attractive: the investor wants the same fractional risk exposure after gains and after losses. But a portfolio can only preserve a constant risky fraction by rebalancing. Without reset, realized returns mechanically change the risky share.

In one discrete period, if the starting risky share is x, the risky return is R, and the cash return is rf, then without rebalancing the next-period risky share becomes

\(x' = \frac{x(1+R)}{1+r_f+x(R-r_f)}.\)

So the investor who does not rebalance is not preserving a constant policy; they are allowing the realized state to rewrite their leverage. Rebalancing is therefore not an arbitrary source of path dependence. It is the mechanism required to keep risk preferences state-independent over time. This fits the note’s point that, under i.i.d. returns, rebalancing’s main contribution is to keep the portfolio at the optimal weight, rather than to mechanically “buy low and sell high.”

There is also an important distinction between unlevered fixed weights and levered target leverage. For 0<x<1, reset rebalancing is contra-cyclical in notional terms: after a loss the risky weight falls, so rebalancing buys risk; after a gain it sells risk. But for x>1, leverage rises after losses and falls after gains if the investor does not reset.

So maintaining constant leverage requires deleveraging after losses and releveraging after gains. In the leveraged case, daily reset is not primarily about harvesting mean reversion. It is about refusing to let leverage drift upward in bad states.

That is the right way to frame the “drag.” It is not evidence that daily resetting is stupid. It is evidence that demanding a fixed leverage target every day has a cost. That cost is the premium for a portfolio rule that says: my desired risk exposure is L, not whatever leverage the last price move happens to leave me with.

From Volatility Drag to Volatility Pumping

The same compounding mechanics that generate volatility drag can also produce what is known as volatility pumping. The classic illustration comes from Claude Shannon’s thought experiment known as Shannon’s Demon.

Consider a coin-flip game where:

• heads produces a +50% return

• tails produces a −33.3% return

The game has zero geometric return. A heads followed by a tails returns wealth exactly to its starting value. But something remarkable happens when the game is combined with cash and periodically rebalanced. Suppose an investor allocates:

• 50% to the coin-flip game

• 50% to cash

and rebalances after each flip. Despite both components having zero long-run compounded returns individually, the rebalanced portfolio produces positive growth over time. The reason lies in the fundamental inequality that the arithmetic mean exceeds the geometric mean.

The coin-flip game has positive arithmetic expectancy but enough variance to push its geometric return to zero. By mixing it with cash and rebalancing, we reduce variance faster than we reduce the arithmetic mean, allowing the portfolio’s compounded growth rate to become positive.

Volatility Targeting

The broader implication is straightforward. Volatility drag is not an argument for avoiding leverage. It is an argument for sizing leverage appropriately. The compounding penalty from variance means that leverage has a finite optimum; beyond that point, additional leverage becomes self-defeating.

Rebalancing plays a central role in implementing this logic. Maintaining a constant risky weight requires periodic adjustment as asset prices move. Daily rebalancing is therefore not a mechanical source of decay, but the operational mechanism that keeps portfolio risk aligned with an investor’s chosen policy. The drag associated with reset should be interpreted as the cost of preventing unintended leverage drift and preserving a stable risk target through time.

This framework naturally leads to volatility targeting. Since the growth-optimal allocation is inversely proportional to volatility squared, changes in observed volatility should mechanically alter the amount of exposure an investor takes to risky assets.

When volatility rises, optimal leverage falls; when volatility declines, optimal leverage rises. In practice, volatility targeting can help stabilize portfolio risk and has been shown in empirical research to produce higher realized Sharpe ratios across many asset classes.

However, this result should be interpreted cautiously. Volatility estimates are noisy, and the effectiveness of volatility targeting depends heavily on how volatility is measured and forecast. In addition, volatility targeting rules often induce trading patterns that are highly correlated with time-series momentum, since volatility tends to rise during market drawdowns and fall during sustained rallies.

As a result, some of the Sharpe improvement attributed to volatility targeting may simply reflect indirect exposure to the momentum factor rather than a pure volatility-management premium.

Disclaimer

The information provided on TheLogbook (the “Substack”) is strictly for informational and educational purposes only and should not be considered as investment or financial advice. The author is not a licensed financial advisor or tax professional and is not offering any professional services through this Substack. Investing in financial markets involves substantial risk, including possible loss of principal. Past performance is not indicative of future results. The author makes no representations or warranties about the completeness, accuracy, reliability, suitability, or availability of the information provided.

This Substack may contain links to external websites not affiliated with the author, and the accuracy of information on these sites is not guaranteed. Nothing contained in this Substack constitutes a solicitation, recommendation, endorsement, or offer to buy or sell any securities or other financial instruments. Always seek the advice of a qualified financial advisor before making any investment decisions.