Appearance

Language

Long-term investing

Lump Sum vs Dollar-Cost Averaging in 2026: A Backtest of the S&P 500 and Taiwan's 0050, and Does the Day of the Month Matter?

With a sum already in hand, investing it all at once beat spreading it over 12 months in about three out of four start months for 0050 and four out of five for the S&P 500, in a backtest of the S&P 500 and Taiwan's 0050 with dividends reinvested. Which day of the month you buy on changed the long-run result by less than 1%, and the backtest also shows what happened to people who started at the worst times, late 2007 and early 2000.

Kelvin Wang · · 14 min read

Key points

  1. Same money, all at once vs over 12 months: for 0050 (start months 2004–2025) the lump sum ended ahead in 76.3% of start months, by 6.1% on average; for the S&P 500 (1993–2025, in NT$) in 79.7%, by 5.1% (79.4% and 4.9% in US dollars).
  2. When the lump sum loses, it loses big: starting in April 2008, 0050 bought at once was worth 32.7% less than the spread-out plan once that plan finished; starting in October 2007, it fell by more than half and took until January 2011 to get back to the money put in.
  3. If you invest from each month's pay, there is no lump sum to choose, so that comparison doesn't apply to you; whether you buy on the 6th, 16th or 26th, the annual return for 0050 and the S&P 500 differs by less than 0.1 percentage points.

These two questions usually get asked together, but they belong to different people. “All at once or spread out?” only matters if you already have a sum of money sitting in your account. “Which day of the month should I buy?” is for people who invest part of each paycheck. Work out which one you are first, then read the numbers for your case.

Share of start months where 0050 bought at once beat 12 monthly buys (262 starts, Feb 2004 to Nov 2025)
76.3%
Same comparison for the S&P 500 (394 starts, Feb 1993 to Nov 2025, in NT$; 79.4% in US$)
79.7%
How much more 0050 bought at once ended with, on average
+6.1%
0050 lump sum vs spread-out at the worst start month (April 2008)
−32.7%
Calculator Compound growth calculator Backtest your own amount and period on real monthly data for the S&P 500 and 0050. It shows NT dollars and applies Taiwan's tax rules by default (the US withholds 30% of SPY's dividends; 12% income tax bracket; a 0.1425% fee on each amount), because I built it in Taiwan; the S&P 500 backtest converts SPY at each month's actual USD/TWD rate. The floating Compound growth bar at the bottom of the screen is set to the S&P 500 and uses the same currency and tax.

Which situation are you in?

The same backtest means very different things depending on who is reading it. Start with your own situation:

Your situationHow to tellThe comparison that appliesWhat happened in the backtest
You invest from your payNo sum on hand; money to invest only appears each month after living costsMonthly investing is your only option, so lump sum vs spreading doesn't apply; see "Does the day of the month you buy matter?"Investing as soon as the money arrives already makes each payment its own "lump sum"; the day of the month moved results by under 1%
You have a sum of spare moneyA bonus, an inheritance, a home sale, a maturing fixed-term deposit: the money is already in your accountAll at once vs over 3–24 months0050 bought at once came out ahead in about three out of four start months, by 6% on average; at the worst starts it was 30%+ behind
You need the money within 3 yearsA down payment, tuition, money to start a businessBoth approaches carry the market's short-term swings0050 once fell 55.8% in 13 months, on daily closes from October 2007 to November 2008; even spread over 12 months, the worst case was worth 66% of the money when the last buy went in
You have no emergency fundLosing your job or falling ill would mean selling shares to pay billsBoth approaches can force you to sell at the bottomAnyone forced to sell in late 2008 never saw the recovery numbers in this backtest
A 30% drop would make you sellYou stopped or sold the last time markets fell hardSpreading out makes the paper loss smaller, at a lower average returnThe share of starts below the money put in when the spread finishes is similar (0050: 24.8% lump sum, 21.4% spread out); the worst cases are very different

The last column is from 0050's history with dividends reinvested [1] [2] [3]; for the S&P 500, the lump sum beat a 12-month spread in 79.7% of start months in NT dollars (79.4% in US dollars) [4] [5]. Past results don't predict future ones.

More than one row can apply. Someone on a salary who also just got a bonus faces the lump-sum question for the bonus, and is a plain monthly investor for the rest.

I’m a lump-sum investor myself, investing from Taiwan: when I have money to invest, it goes in at once rather than in installments.

What’s the difference between a lump sum and dollar-cost averaging?

“Dollar-cost averaging” gets used for two different things: spreading a sum you already have over several months, and investing a fixed amount from each paycheck. They answer different questions, so they need to be kept apart.

  • Lump sum. All the money goes in at once, say 120,000 in whatever currency you use, into one ETF on one day.
  • Spreading it out. The same money is split and invested over several months, say 10,000 a month for 12 months. The rest waits in cash.
  • Regular monthly investing. A fixed amount each month, usually from your pay. The money doesn’t exist yet, so “all at once” is not an option.

The backtest compares the first two: two ways of investing the same money. The two assets are the S&P 500 and Taiwan’s 0050, which tracks the FTSE TWSE Taiwan 50 Index. A few terms:

  • Total return (dividends reinvested). The return after every dividend is used to buy more of the same ETF. Price alone leaves out the dividends, which come to about 2–4% a year for 0050 in most years, nearer 2% lately 1 3.
  • Rolling start months. Every month in the data is treated as a separate “the money arrived” moment, and we count how many of them favored the lump sum. A single start date would mostly tell you about that one year.
  • Win rate. In this backtest, the share of start months where the lump sum ended with more than spreading out. It is not the chance of making money.

With a sum in hand, all at once or over 12 months?

Method: each month-end is a start. The lump sum buys everything at that month-end; the spread-out plan buys 1/12 at that month-end and each of the next 11. Cash waiting to be invested earns nothing. The two are compared at the month-end of the last spread-out purchase. The data is the same monthly series behind this site’s compound growth calculator: 0050 from Taiwan Stock Exchange closing prices and dividend records, dividends reinvested 1 2 3; the S&P 500 from SPY’s adjusted close (dividends in, ETF fees out) 4, converted to NT dollars at FRED’s month-end exchange rate 5, because I built the backtest in Taiwan. In US dollars the results are close: 79.4% over 12 months. Brokerage fees and taxes are left out.

Share of start months where the lump sum beat spreading out %, by number of months spread over; 0050 starts from Feb 2004, S&P 500 from Feb 1993; the last start is the one whose spread finishes by Oct 2026
  • 0050
  • S&P 500 (NT$)

Dividends reinvested; waiting cash earns nothing; no fees or taxes. S&P 500 is SPY's adjusted close in NT$. In US dollars the four figures are 67.2%, 72.3%, 79.4% and 82.7%, and over 12 months the lump sum ended 4.9% ahead on average.

Source: twse.com.tw, yahoo.com, stlouisfed.org

The longer the spread, the more often the lump sum wins, because more of the money spends more time outside the market. Spread over 12 months, the money waits 5.5 months on average before it is invested: the first twelfth goes in at once, the last after 11 months.

How big the gaps are:

  • 0050 over 12 months. The lump sum ended 6.1% ahead on average (median 6.1%). When it won, it was ahead by 10.6% on average; when it lost, it was behind by 8.5%.
  • S&P 500 over 12 months. 5.1% ahead on average in NT dollars (4.9% in US dollars); in NT dollars, +8.1% when it won and −6.5% when it lost (+8.1% and −7.7% in US dollars).
  • Over 3 months. Much smaller: 1.1% ahead on average for 0050, 0.9% for the S&P 500.

On 100,000, 6% is about 6,000. That is an average over 262 start months, and the range for any one person is wide: at the best start (January 2009) the lump sum was 44.5% ahead; at the worst (April 2008), 32.7% behind.

Vanguard ran the same comparison in 2023 on the MSCI World Index from 1976 to 2022: against spreading over 3 months, the lump sum was ahead after one year 68% of the time 6. In this backtest the 3-month figures are 62.7% for 0050 and 68.5% for the S&P 500 in NT dollars (67.2% in US dollars), pointing the same way. Vanguard measures after one year and this backtest measures at the end of the spread, but the winner is the same, because once the spread-out plan is fully invested the gap between the two stops changing. The same paper splits the result by market:

Vanguard: how often a lump sum beat a 3-month spread, by market % of rolling one-year periods where the lump sum ended ahead; in each market's local currency (US$ for global and emerging markets)
  • UK (FTSE All-Share) 68.1%
  • Global (MSCI World, US$) 67.7%
  • Australia (S&P/ASX 300) 67.5%
  • Canada (S&P/TSX Composite) 67.2%
  • Europe (MSCI Europe) 66.5%
  • US (Russell 3000) 66.4%
  • Emerging markets (MSCI EM, US$) 61.6%

Periods: US 1979–2022, UK 1986–2022, Canada 1985–2022, Europe 1998–2022, Australia 1992–2022, emerging markets 1988–2022, global 1976–2022. The paper's main-text 68% also uses the MSCI World Index for 1976–2022; its US-dollar figure here is 67.7%. Appendix 1, Figure 6 [6].

Source: vanguard.com

Does the gap shrink once the spread-out plan is fully invested?

No. After the last spread-out purchase, both people hold nothing but the same ETF, so from then on they rise and fall by exactly the same percentage, and “the lump sum is 6% ahead” stays 6% for good. Over 20 years the percentage stays put, but because the pot grows, the gap in money grows along with the pot.

In other words, the choice only matters during the months of the spread; for the 10 or 20 years after, both earn the same return.

In which years did the lump sum lose?

Lump sum vs 12 monthly buys, by start year %, average of the 12 start months in each year; positive means the lump sum ended ahead
  • 0050
  • S&P 500 (NT$)

0050's 2004 starts begin in February; 2025 starts run to November. Dividends reinvested, no fees or taxes.

Source: twse.com.tw, yahoo.com, stlouisfed.org

The years where the lump sum lost on average cluster just before big falls: 2007, 2008, 2011, 2015, 2021 and 2022 for 0050; 2007–2008 for the S&P 500, and 2000–2002 before the chart begins. In every other year the lump sum came out ahead, by double digits in years like 2020 and 2023 when a fall was followed by a sharp rise.

The catch is that you only know which year was which afterwards. In October 2007, the month 0050 peaked, the market looked like “it’s gone up a lot lately,” no different from many months that went on rising.

What happened to people who started at the worst times, late 2007 and early 2000?

Investing in 0050 in October 2007

From October 2007, 0050 bought at once vs over 12 months (per 100 invested) value per 100 invested, quarter-end, dividends reinvested, 0050 in NT$; the spread-out plan's uninvested cash counted at face value
  • Lump sum
  • Over 12 months

The spread-out plan buys a twelfth at each month-end from October 2007 to September 2008. No fees or taxes.

Source: twse.com.tw

  • Lump sum. Down to 46 per 100 invested at the end of January 2009, 54% below the money put in, and not back to 100 until January 2011: three years and three months.
  • Over 12 months. Bottomed at 56 in January 2009 and first got back above 100 in December 2009. The later purchases went in at 2008’s low prices, so once fully invested it stayed about 22% ahead of the lump sum.

The single worst start for the lump sum in the 0050 data was April 2008. The lump sum fell to 51 per 100 invested by the end of January 2009. The spread-out plan put half its money in between October 2008 and March 2009, near the lows, and was worth 90 when it finished at the end of March 2009, against 61 for the lump sum: 32.7% less.

Investing in the S&P 500 in January 2000

Early 2000 was close to the dot-com peak, yet measured in NT dollars, the lump sum stayed slightly ahead almost the whole way. All 12 spread-out purchases went in during 2000, while the S&P 500 was still drifting near its highs; the real fall came in 2001–2002. At the bottom, at the end of September 2002, the lump sum was worth 69% of the money put in and the spread-out plan 66%: 31% and 34% down. The lump sum was back above the money put in by August 2005, the spread-out plan by November 2005. Except at the end of February 2000, the lump sum was ahead every month.

In US dollars the picture is the same: the lump sum was ahead every month except February 2000, at the bottom it was worth 61% of the money and the spread-out plan 59%, and they were first back to the money put in by January and March 2006 respectively.

Spreading out only protects you from falls that happen during the spread. If the decline lasts longer than that, every spread-out purchase still lands near the top.

Would you sell after a 30% fall?

Every backtest number assumes both people sat tight. In practice, the person who invested everything in October 2007 and saw 46% of it left by January 2009, and the person who spread it out and saw 56%, both had a very hard stretch to sit through.

On average the lump sum ends with more, and spreading out makes the worst year hurt a little less; either way, the backtest numbers only happen to people who didn’t sell halfway.

Vanguard reaches the same conclusion: spreading out has lower expected returns, but for investors with very high aversion to both risk and losses, who might otherwise leave a windfall entirely in cash, it is still better than holding it all in cash: spreading over 3 months beat staying in cash 69% of the time 6. Its advice for those who do spread out is to keep the period short, “such as three months”.

  • If you fear the worst case. Measured when the spread-out plan finishes, 0050’s worst case was 66% of the money put in over 12 months and 53% at once; for the S&P 500, 76% and 64% in NT dollars (70% and 57% in US dollars).
  • If you fear still being down once fully invested. The two are similar: at the month-end of the 12th purchase, the share of starts below the money put in was 24.8% (lump sum) vs 21.4% (spread out) for 0050, and 16.2% vs 19.0% for the S&P 500 in NT dollars (18.0% vs 19.8% in US dollars).
  • If you fear second-guessing yourself. A spread-out schedule is fixed in advance, so you buy on the planned dates during a fall, which is easier than deciding on the spot whether to go in.

Holding a sum back until the market drops is a different plan, and this backtest doesn’t test it, because “how far is a dip” has no objective definition and the result depends on whichever rule you pick. What the data does show is that every month the money spent outside the market put it, on average, on the losing side.

Does the day of the month you buy matter?

Hardly at all. The backtest uses daily closes: buy once a month on the 6th, 16th or 26th (moved to the next trading day when the market is shut, the same rule IBKR’s recurring investments use 7), dividends reinvested, valued at the end of October 2026. 0050 uses the exchange’s daily closes and dividend records from January 2010 1 3; the S&P 500 uses SPY’s daily adjusted close from February 1993, in US dollars 4.

Annual return by day of the month bought: 6th, 16th, 26th %, annual return accounting for when each payment went in (internal rate of return), dividends reinvested, no fees or taxes; day count 365.25
  • 0050 (Jan 2010 to Oct 2026, NT$)
  • S&P 500 (Feb 1993 to Oct 2026, US$)

A purchase date falling on a market holiday buys on the next trading day. S&P 500 is SPY's daily adjusted close.

Source: twse.com.tw, yahoo.com

This return is the internal rate of return: since each payment goes in at a different time, it is the annual return after accounting for how long each one was actually invested. The three days are within 0.1 percentage points of each other.

In money: 202 equal monthly purchases of 0050 from January 2010 grew to about 7.01 times the money put in by the end of October 2026 if bought on the 6th, 6.99 times on the 16th and 6.98 times on the 26th, at most 0.36% apart. Across every day from the 1st to the 28th, the best and worst days differ by 0.58%; for the S&P 500 since 1993, by 0.61%.

  • Early-month purchases end with slightly more because the money goes in a few weeks sooner. Every 6th-of-the-month purchase is invested about 20 days before the 26th’s, and spends about 20 more days in the market; compared by annual return, the 26th is actually a hair higher.
  • No day wins consistently. For 0050, the 26th had a slightly higher annual return in both 2010–2017 and 2018–2026, by under 0.15 points each time; for the S&P 500 in 1993–2009, the 6th and 26th tied and the 16th was slightly lower.
  • Stopping the plan matters far more than the day. From October 2007, 0050 lost half its value and took over three years to recover; pausing during that stretch changes the result far more than any choice of day.

So pick a date shortly after payday, when the money is sure to be in the account. If the cash isn’t there, that month’s purchase is skipped, which costs more than any choice of date. At IBKR, for example, a recurring investment that lacks funds is retried for the next two days, then moves to the next scheduled date; if the account is short again on that date, the instruction is canceled 7.

Does buying in 12 pieces cost more in fees?

Costs charged as a percentage of the trade, such as percentage-based commissions and transaction taxes, add up to the same total whether you buy once or 12 times. What differs is any fixed or minimum charge per order. Two examples for buying US ETFs from outside the US:

  • No commission. Firstrade charges US$0 for online ETF trades 8 and accepts fractional-share purchases of most ETFs from US$5 9, so splitting a sum into 12 orders costs nothing extra in commission.
  • A per-order minimum. IBKR Pro Fixed charges US$0.005 a share with a US$1 minimum per order, capped at 1% of the trade value 10. For a hypothetical ETF priced at US$50 throughout, US$6,000 buys 120 whole shares for US$1 in commission, while twelve US$500 orders buy 10 whole shares each and cost US$12 in total. IBKR’s recurring investments pay its standard commissions 7, but fractional-share orders have their own commission rules, so check those separately 10.

In Taiwan, where I invest, the 0.1% securities transaction tax on ETFs is charged only when you sell, so it is the same either way 11.

Go deeper How to Buy US ETFs From Outside the US in 2026: Brokers, Fees, Dividend Tax and VOO, QQQ, SCHD Costs Opening an international account step by step, wire costs, and the EU and UK rules in detail.

What if the waiting cash earns interest?

The backtest assumes uninvested cash earns nothing, which tilts things slightly against spreading out. Assuming the waiting cash earns 1% or 2% a year (assumed rates, not any bank’s posted rate), the 12-month results become:

Interest on waiting cash0050: lump sum ahead0050: average gapS&P 500: lump sum aheadS&P 500: average gap
0%76.3%+6.1%79.7%+5.1%
1%74.4%+5.6%78.2%+4.7%
2%72.9%+5.1%74.6%+4.2%

Each waiting tranche earns interest compounded monthly at an effective 1% or 2% a year from the start month until it is invested. S&P 500 in NT$.

Interest narrows the gap only a little. Vanguard found the same: with 3-month US Treasury bill interest on the waiting cash, the lump sum still beat a 3-month spread 65% of the time for an all-stock portfolio 6. Stocks have returned more than cash over the long run, so the longer money waits outside, the more it gives up.

FAQ

I got a bonus. All at once, or add it to my monthly plan?

Both are cases the backtest covers: investing it at once is the lump sum, and adding it evenly to the next 12 monthly purchases is spreading over 12 months. In 0050’s history the lump sum came out ahead in about three out of four start months, but at the worst starts it was 30%+ behind. Which fits you depends on the conditions under “Which situation are you in?”, especially when you’ll need the money and whether you’d sell in a crash.

I already invest monthly and just got a sum of money. Should I pause the monthly plan?

In the backtest the two don’t interact: the monthly plan keeps buying on its usual date, and investing the sum at once or spread out is a separate comparison. Pausing the plan only means the monthly part buys nothing for those months.

Why are the 0050 and S&P 500 results different?

The start ranges differ (0050 from 2004, the S&P 500 from 1993), and the two markets rose and fell on different schedules. Converting the S&P 500 to NT dollars also adds currency moves: in US dollars the 12-month figure is 79.4% rather than 79.7%. Both point the same way: the longer the spread, the more often the lump sum wins.

Does the backtest include fees and taxes?

No. 0050 dividends are reinvested on the ex-dividend date at the ratio of the pre-ex-date close to the reference price, with no Taiwan tax deducted; SPY’s adjusted close doesn’t deduct the US dividend withholding that non-US investors pay either. Both methods buy the same ETF, so taxes affect them about equally. For how Taiwan taxes 0050, see Taiwan 0050 ETF in 2026; US dividend withholding and treaty rates are covered in the S&P 500 guide.

This is not investment advice. The backtests summarize historical data and don’t predict future results, and there are no affiliate or broker referral links.

Sources

  1. 1. Taiwan Stock Exchange: daily trading data (0050) · twse.com.tw · Jan 2010 to Oct 2026
  2. 2. Taiwan Stock Exchange: daily closing quotes (ETFs, for 2004–2009 monthly data) · twse.com.tw · 2004–2009
  3. 3. Taiwan Stock Exchange: ex-rights and ex-dividend results · twse.com.tw · 2004–2026
  4. 4. Yahoo Finance: SPY historical prices (adjusted close) · finance.yahoo.com · Jan 1993 to Oct 2026
  5. 5. FRED: Taiwan Dollars to U.S. Dollar Spot Exchange Rate (DEXTAUS) · fred.stlouisfed.org · 1993–2026
  6. 6. Vanguard: Cost averaging: Invest now or temporarily hold your cash? · corporate.vanguard.com · Feb 2023
  7. 7. Overview of the recurring investment feature · Interactive Brokers · checked 2026-11-18
  8. 8. Pricing · Firstrade · checked 2026-11-18
  9. 9. Fractional Shares Trading · Firstrade · checked 2026-11-18
  10. 10. Commissions: Stocks, ETFs and Warrants · Interactive Brokers · checked 2026-11-18
  11. 11. Taiwan Stock Exchange: ETF trading rules (in Chinese) · twse.com.tw · checked 2026-11-18

About the author

Kelvin Wang

Electrical engineer building tools with AI agents, sharing hands-on notes on engineering and investing.

More about me