Lump sum vs DCA

The "67% of the time" finding

Vanguard's 2012 paper Dollar-cost averaging just means taking risk later tested lump-sum vs 12-month DCA across decades of US, UK, and Australian market data. The result: investing the full amount immediately beat DCA roughly two-thirds of the time, with an average outperformance of about 2.4 percentage points over the first year. The 2023 update Cost averaging just means taking risk later reaches the same conclusion on broader data.

The Monte Carlo above reproduces the shape of that finding with calibrated parametric returns (a lognormal walk matched to each asset's documented annualized return and volatility, using the portfolio mixes from the PAC backtest and Monte Carlo tools). With default inputs you'll see the win-rate land somewhere between 60% and 75%, depending on the portfolio mix and horizon you pick.

Why lump-sum usually wins

Stocks have a positive expected return. Sitting on cash for 6, 12, or 36 months means a chunk of your money misses that expected return for the entire ramp period. The longer the DCA window, the bigger the average drag - which is why a 36-month ramp loses more often than a 12-month one in the simulation.

The math is simple: if expected monthly returns are positive, E[lumpSum_T] > E[DCA_T] for any T at the end of the horizon. What the Monte Carlo adds is the spread - the minority of paths where DCA wins are the ones where markets drop early, and DCA buys the dip on someone else's behalf.

When DCA actually wins

DCA wins on the bad-return paths. Look at the histogram: the left tail (red bars) shows paths where DCA ended ahead, and those are concentrated where the market fell hard in the first year or two. If you happen to deploy a windfall the week before a 30% crash, DCA materially limits the damage.

That's the regret-minimization argument for DCA: even if it has a worse expected outcome, it has a less painful worst-case outcome. If you'd sell in a panic after losing 30% of a lump-sum deployment, then DCA isn't a probability-optimal strategy - it's a behavior-optimal strategy. That's a perfectly valid reason to use it. Just know what you're trading.

Caveats

• This is calibrated parametric Monte Carlo, not historical bootstrap. Real markets have fatter tails and serial correlation that a lognormal walk understates - so the extreme-loss paths where DCA wins are slightly under-counted.
• Cash sitting on the sidelines during the DCA ramp earns 0% here. In reality a high-yield savings account would close some of the gap; we ignore it for clarity.
• No fees or taxes are modeled beyond the portfolio's documented TER. Trading costs and capital-gains drag would tilt the comparison slightly toward fewer transactions (i.e., toward lump-sum).
• Same per-month returns are used for both strategies inside each path (paired comparison). Win-rates are clean apples-to-apples; absolute numbers depend on the seed.

Source: Vanguard, Cost averaging just means taking risk later (2023). corporate.vanguard.com (retrieved 2026-05-07).