You've sold a property, received a bonus, or inherited money. The cash is sitting in an account and every day it sits there feels like a decision you haven't made. The usual advice is to feed it in gradually — a twelfth each month for a year — so you don't put everything in at the top. It sounds prudent. The evidence says it costs you money most of the time, and the reason is duller than the argument suggests.
What the data actually says
Vanguard tested the question across three markets over rolling one-year periods from 1976 to 2022. It compared investing everything at once against splitting the same sum across a phase-in schedule. Investing at once won between 61.6% and 73.7% of the time, depending on which market you look at.
The markets tested were the Russell 3000 for the US from 1979, the FTSE All-Share for the UK from 1986, and the S&P/ASX 300 for Australia from 1992. The multi-asset work used the MSCI World Index alongside the Bloomberg U.S. Aggregate Bond Index across the full 1976 to 2022 window.
The size of the gap matters more than the win rate. Measured over a one-year rolling period against three-month averaging splits, the median advantage to investing at once was 2.2 percentage points for a portfolio held entirely in equities, 1.8 points for a 60/40, and 1.2 points for a 40/60 equity-bond split.
Note the direction of that pattern. The more equity you hold, the more the delay costs. That's the whole mechanism in one line, and it isn't about timing at all.
Why it happens, and it isn't clever
Markets go up more often than they go down. That's it. If an asset has a positive expected return, then any period you spend out of it is a period you're not earning that return. Cost averaging is a schedule that guarantees you'll be partly out of the market for the whole phase-in window, by construction.
Over a twelve-month drip, your average exposure is roughly half the target. You're holding a portfolio that is half cash for a year, then switching. If you wouldn't choose a half-cash portfolio as your long-run allocation, it's worth asking why it's the right one for the twelve months you happen to be starting in.
This isn't a new observation. George Constantinides challenged the idea that averaging in reduces risk in the Journal of Financial and Quantitative Analysis in 1979, arguing the approach is suboptimal against alternative strategies. Knight and Mandell went further in Financial Services Review in 1992, testing it three ways — graphical analysis, historical stock market returns and Monte Carlo simulation — and finding that optimal rebalancing and buy-and-hold beat averaging in all three.
Neither paper needed a bull market to make the argument. The result falls out of the assumption that expected returns are positive, which is the same assumption you're making by investing at all.
The case for cost averaging, taken seriously
Here's the part the win-rate statistic hides. Investing at once loses roughly three times in ten. Those aren't random losses scattered evenly — they cluster in exactly the periods where markets fall over the phase-in window, which is to say the periods people are most frightened of.
If you put a large sum in and the market drops 25% over the following six months, the arithmetic says hold. The question is whether you will. An investor who abandons the plan and sells at the bottom converts a temporary paper loss into a permanent one, and the cost of that dwarfs 1.8 percentage points. We've looked at the size of that behavioural cost separately in the work on what bad timing does to investor returns, and the gap there runs at over a percentage point a year on the average dollar.
Read that way, cost averaging isn't a return strategy at all. It's insurance against your own reaction, and the median 1.8 points is the premium. Insurance that costs something and pays out in the bad state is a perfectly rational purchase. It just shouldn't be sold as the higher-returning option, because it isn't.
There's a second, narrower case. If the sum is genuinely large relative to your existing portfolio — a windfall that doubles your invested assets overnight — then the sequence of returns in the first year carries unusual weight. That's the same mechanism that makes the years around retirement dominate outcomes, which we covered in the piece on how return order changes results on identical returns. Phasing in reduces the variance of that first-year outcome. It doesn't improve the expected one.
Where the strategy genuinely earns its keep
Cost averaging's best historical showing comes from extended declines rather than sharp ones. A drip schedule through a market that grinds lower for years buys progressively more units, and the strategy's relative performance improves the longer the fall lasts. Japan through the 1990s is the case usually cited.
The catch is that you can't identify that regime in advance. If you could, the correct response wouldn't be a twelve-month drip — it would be not buying yet. A schedule chosen because you fear a decline is a weak version of a market call, and it commits you to buying through the decline anyway.
One distinction that gets lost: regular monthly investing out of a salary isn't cost averaging in this sense. That's just investing money as it arrives, and there's no lump sum sitting in cash for it to lose against. The comparison only bites when you already hold the full amount.
What this costs in practice
On £100,000, a median 1.8-point shortfall over the phase-in year is about £1,800 of forgone return in the middle case. That is real, but it isn't catastrophic, and it's roughly what a year of a 1% fund fee costs on a similar balance — a comparison we ran in detail in the work on what fees compound to over thirty years.
The asymmetry is worth stating plainly. The cost of phasing in is bounded and known in advance. The cost of investing at once and then panicking is unbounded and only revealed afterwards. People who know they're prone to the second problem are not being irrational when they pay to avoid it.
A worked example
Take £120,000 and a twelve-month drip of £10,000 a month. In month one you hold £10,000 invested and £110,000 in cash. By month six you're at £60,000 and £60,000. Averaged across the year, a little over half the money is invested.
So for twelve months you're running an allocation you didn't choose. If your target is 60/40, the drip has you at roughly 30/70 for the first half of the year, drifting up to target by December. Nobody would write that down as a plan. It's a by-product of the schedule.
Apply the median 1.8-point figure for a 60/40 and the shortfall is around £2,160 over the year. Whether that number frightens you is the whole question. It's a rounding error against a thirty-year horizon and it's a fortnight's income for a lot of people.
The comparison also assumes you actually finish. In practice a drip that starts into a falling market often stops — the schedule that was meant to remove the decision hands you twelve fresh chances to make it. That failure mode doesn't show up in any backtest, because backtests always complete the schedule.
The objections worth taking seriously
Critics of the lump-sum result make three arguments, and they're not equally strong.
The first is that the studies measure the wrong thing. They compare expected wealth, when what an investor experiences is the path. That objection is fair as far as it goes. Phasing in genuinely does cut the dispersion of first-year outcomes, and if your utility function punishes early losses more than it rewards early gains, the schedule can be the better fit even at a lower expected return.
The second is that the win rate is period-dependent — that four decades ending in 2022 flatter equities because the sample sits inside a long disinflation. This is the strongest version, and it's partly right. The honest answer is that the result also holds in the analytical papers, which don't depend on any sample at all, and that it held across three markets with different starting points and different crises.
The third argument is that phasing in lets you buy lower if the market falls. That one doesn't survive contact with the arithmetic. It's a claim about direction, and if you had a view on direction you'd act on it directly rather than through a schedule that also buys on the way up.
What would change the conclusion
Three things would move this, and one of them is measurable today.
First, if expected returns were negative, the ordering reverses immediately. Nobody holds that view about a diversified global portfolio over a multi-decade horizon, but it's the assumption doing all the work, and it's worth being honest that it's an assumption.
Second, the win rate isn’t constant across valuation levels. Work in the Journal of Financial Planning has tested whether the CAPE ratio flags periods when averaging does better, and reports that investing at once still wins at extreme valuations, just less often than its overall average. That’s a shift in odds, not a reversal, and it isn’t a licence to time.
Third, cash rates matter. Every one of these comparisons assumes the un-invested money earns a cash return. When cash yields 5%, the drag from sitting out is smaller than when it yields nothing, and the gap narrows. It doesn't close, because equity risk premia have historically exceeded cash by more than that, but the premium you're paying for the insurance falls.
The cash-rate point deserves an example. If the un-invested balance earns 4% and the portfolio's expected return is 7%, the opportunity cost of sitting half out for a year is roughly 1.5 points, not 7. When cash paid nothing, the same delay cost the full spread. The strategy's price tag moves with the deposit rate, which is a reason to check the arithmetic rather than inherit a rule of thumb from a different rate environment.
How to think about it
The evidence supports investing at once as the higher-expected-return choice, by a margin that's real but modest, and it has been stable across three markets and four decades. If you're confident you'll sit through a bad first year, that's the option the data points to.
If you're not confident, the honest framing isn't that cost averaging is safer. It's that you're buying a known, bounded cost to reduce the chance of an unknown, unbounded one. Price the premium, decide whether you want it, and don't pretend the trade is free.
The one thing the evidence rules out is the middle position — phasing in because it feels like the responsible default, without knowing what it costs. That's the version that gets you the lower return and the anxiety.