Dollar-Cost Averaging vs. Lump Sum: What the Math Actually Says

If you have a lump of cash to invest, you have probably run into the statistic that investing it all at once beats spreading it out most of the time. That statistic is true. It is also answering a different question from the one most people are actually asking.

By Kevin WinmillUpdated 5 min read

#Why lump sum wins on average

The argument is simple and correct. Markets rise more often than they fall over most multi-year periods. If you spread a lump sum over twelve months, part of it sits in cash while it waits its turn, and cash has historically returned less than stocks. Averaged across many historical periods, that waiting is a drag.

Studies of this question, most notably Vanguard’s, typically find lump sum ahead roughly two thirds of the time across rolling historical windows. The exact figure varies with the market, the period, and the assumed cash return, but the direction is consistent and the mechanism is not in dispute.

“Wins about two thirds of the time” also means it loses about one third of the time. That is not a rare event. It is a coin weighted two to one, and you get one flip.

#What that looks like in a decade where it lost

Here is the same $70,000 going into SPY across the 2000s, once as a single investment on day one and once as $10,000 up front plus $500 a month. Identical dollars, identical fund, identical decade. Only the timing differs.

The same $70,000 into SPY from 2000 to 2010, invested all at once versus spread across the decade.
PlanInvestedFinal valueAnn. return
All $70,000 invested on day oneSPY$70,000$63,882-0.91%
$10,000 up front plus $500 a monthSPY$70,000$73,048+0.75%
The same $70,000 into SPY from 2000 to 2010, invested all at once versus spread across the decade. Simulated from January 1, 2000 to January 1, 2010 against real daily closing prices, pre-tax and excluding commissions. Full methodology.
Portfolio value over time: All $70,000 invested on day one versus $10,000 up front plus $500 a monthThe lump sum starts at full size and spends the decade recovering from two crashes. The contribution plan climbs from nothing, buying through both of them.$0$20K$40K$60K$80K$100K2002200420062008
All $70,000 invested on day one$10,000 up front plus $500 a monthMoney you put in
The lump sum starts at full size and spends the decade recovering from two crashes. The contribution plan climbs from nothing, buying through both of them.

The lump sum finished below what was put in, at an annualized return of about negative one percent. The contribution plan finished ahead of it by roughly $9,000 and stayed slightly positive. Investing everything in January 2000 meant buying immediately before two of the worst bear markets in modern history, and a decade was not long enough to recover from it.

The contribution plan was not smarter. It simply kept buying through 2002 and again through 2008, accumulating shares at prices the lump-sum investor never saw. The QQQ versus SPY comparison shows an even more extreme version of the same effect.

#And in a decade where it won handily

Run the identical test over the last ten years, a period that rose nearly throughout, and the picture inverts.

The same $70,000 into VOO over the last decade, invested all at once versus spread across it.
PlanInvestedFinal valueAnn. return
All $70,000 invested on day oneVOO$70,000$295,929+15.51%
$10,000 up front plus $500 a monthVOO$70,000$178,251+15.74%
The same $70,000 into VOO over the last decade, invested all at once versus spread across it. Simulated from September 9, 2016 to September 9, 2026 against real daily closing prices, pre-tax and excluding commissions. Figures current as of . Full methodology.
Portfolio value over time: All $70,000 invested on day one versus $10,000 up front plus $500 a monthIn a decade that rose almost throughout, every month the contribution plan spent waiting was a month of gains it did not capture.$0$100K$200K$300K20182020202220242026
All $70,000 invested on day one$10,000 up front plus $500 a monthMoney you put in
In a decade that rose almost throughout, every month the contribution plan spent waiting was a month of gains it did not capture.

The lump sum ended roughly $112,000 ahead. This is the two-thirds case, and it is what most of the last decade looked like: every month spent waiting was a month of gains not captured.

Look carefully at the last column. The lump sum has the far larger ending value but the lowerannualized return. That is not an error. The lump sum had all $70,000 working for the full ten years, so it earned more dollars. The contribution plan’s money was invested for less time on average, so the rate it earned per dollar-year was slightly higher. Total value and annualized rate answer different questions, which is the subject of why IRR and CAGR give different answers.

#Why the average is beside the point

“Wins more often across history” is a statement about a distribution of outcomes. You do not get the distribution. You get one investment, on one date, and you find out afterward which third you landed in.

More importantly, the statistic ignores the decision people are actually struggling with. The question is rarely “which has the higher expected value.” It is usually “how do I make a decision I can live with and actually stick to.” Those have different answers, and only the first one is a math problem.

Regret is asymmetric in practice. Investing everything the month before a 30 percent decline feels considerably worse than spreading it out and missing some upside during a rally, and it is far more likely to push someone into selling at the bottom. A strategy with a better expected value that you abandon halfway through is not the better strategy for you.

#What actually determines the right answer

  • How would you react to investing everything today and watching it fall 25 percent over the next year? If the honest answer is that you would sell, the expected-value argument is irrelevant, because you will not be there to collect it.
  • Is this money genuinely long term? The lump-sum edge comes from more time in the market. Over a short horizon that edge shrinks and the timing risk does not.
  • How large is this relative to everything else you own? Doubling a portfolio in one transaction is a different decision from adding five percent to it.
  • Are you comparing against a real alternative? Spreading over three months is a very different plan from spreading over three years, and the drag grows with the delay.

#The middle path most people actually take

A common compromise is to invest a large portion immediately, capturing most of the expected-value benefit, and spread the remainder over a short window of three to six months rather than a year or more. This keeps the majority of the money working while limiting the damage from a single unlucky entry date.

There is no formula that produces the correct split, because the input is your own tolerance rather than a market parameter. But the shape of the tradeoff is easy to see: the longer the spreading window, the more expected return you give up, and the less any single date matters.

#Test it against your own numbers

Pick a ticker and the amount you are actually considering. Run it once with contribution frequency set to lump sum, then again as the same total spread monthly, across several different start years. Include at least one window that begins right before a downturn, such as 2000 or 2007.

Watching the range of outcomes across many starting points is a far more honest test than either the average statistic or a single favorable backtest.

Both tables here are pre-tax and exclude commissions, and neither accounts for the tax consequences of holding cash while you wait. Nothing here is financial advice. Why backtests can mislead you covers the limits of this kind of analysis, including the fact that two decades is a very small sample.