Why Time in the Market Beats Timing the Market

Staying invested has usually beaten attempts to time the stock market, and the reason is arithmetic rather than luck. Stocks beat Treasury bills in about two years out of three, so simply saying "stocks" every year is already right most of the time, and a timer who steps out gives up that return in every year it guesses wrong, so it needs a high hit rate to come out ahead. William Sharpe put the hurdle at roughly 7 correct calls in 10, and the simulation below lands close to that. This describes how hard timing is, not a promise that staying invested always wins; in some periods it did not.

Key takeaways

In Sharpe's 1975 study, a yearly timer needed 74% accuracy to beat a stock-and-cash mix of similar risk, and 83% to beat holding stocks.
On 1928–2025 annual data, a simulated timer needed about 69% accuracy to match buy-and-hold's 10.0% a year, and about 72% once each switch cost 1%.
For start dates from October 2016 to October 2025, waiting up to 12 months for a 10% dip left the average investor with 4.0% fewer shares than investing at once.
Vanguard found that investing a lump sum at once beat spreading it over three months 68% of the time, on global stock data for 1976–2022.
The CAPE valuation ratio explained only about a quarter of the variation in the next 10 years' returns, and very different outcomes followed similar readings.

The accuracy a timer needs

Market timing here means moving money between stocks and cash on a forecast of where the market goes next; buy-and-hold means staying in stocks throughout. William Sharpe asked how often that forecast must be correct in "Likely Gains from Market Timing" (Financial Analysts Journal, March 1975). As summarized in a 2021 peer-reviewed update by Buzzacchi and Ghezzi in the Journal of Risk and Financial Management, he used annual 1929–1972 returns on the S&P Composite and on cash equivalents (prime bankers' acceptances until 1942, then Treasury bills). His timer switched only at the start of each year, paid 2% of the portfolio per switch, and was equally accurate in good and bad years.

The timer beat a constant stock-and-cash mix with about the same risk only when it was right at least 74% of the time, and beat holding stocks outright only at 83% or more. Hence his advice not to try without calling the next year right at least 7 times in 10.

The 2021 update, using 1929–2018 data and a 1% commission, found lower hurdles against buy-and-hold: 71% in 1929–1972 and 61% in 1973–2018. The authors still concluded that the original finding holds. The exact threshold moves with the period and the benchmark, but it stays well above a coin flip.

A replication on 1928–2025 data

Aswath Damodaran's historical returns spreadsheet (NYU Stern) gives annual S&P 500 total returns and 3-month Treasury bill returns for 1928–2025. Over those 98 years, stocks compounded at 10.0% a year and T-bills at 3.4%. Stocks won in 66 of the 98 years, so always saying "stocks" was right 67% of the time.

The simulated timer holds stocks or T-bills for each calendar year, picking the better asset with probability p, independently of earlier years. Each value of p was run 10,000 times (random seed 2026), and the median path is compared with buy-and-hold.

Timer's return in year t:
  R(t) = stock return  if it holds stocks in year t
  R(t) = T-bill return if it holds T-bills in year t
  minus c (1% in the cost case) in any year its holding changes

p = probability of holding whichever asset did better that year
Annual return = (product of (1 + R(t)) over 98 years)^(1/98) - 1

Timer accuracy (p)	Median annual return, no cost	Share of paths beating buy-and-hold	Median annual return, 1% per switch	Share of paths beating buy-and-hold
50% (coin flip)	6.7%	0.1%	6.1%	0.1%
60%	8.4%	6.7%	7.9%	2.3%
70%	10.3%	58.9%	9.7%	38.0%
80%	12.1%	98.2%	11.5%	94.5%

Buy-and-hold earned 10.0% a year. The median timer matched it at 69% accuracy with no costs and at 72% with a 1% cost per switch. A coin-flipping timer earned 6.7% a year. Perfect foresight would have earned 15.8%, which is the prize that makes timing tempting.

The comparison flatters buy-and-hold in one way. The timer sits in T-bills in roughly 40–45% of the years, so its returns were less volatile: about 14.5% a year, against 19.4% for stocks. Against a constant mix of about 75% stocks and 25% T-bills with the same volatility, the break-even accuracy falls to about 62% with no costs and 65% with costs. That is the closest match to Sharpe's 74% test.

The model is simplified. Real timers often act more than once a year and pay more in costs. Errors here are independent from year to year, so a timer whose mistakes cluster, such as staying out for a whole recovery, would do worse than the median path. Taxes appear only through the flat 1% cost.

The cost of waiting for a dip

The most common timing decision is holding new money in cash until prices fall. To test it, each trading day from October 13, 2016 to October 10, 2025 (2,261 start days) was treated as the day money became available, using daily closes of the S&P 500 index from FRED. The waiting investor buys on the first close at least 10% below the start-day level; if none comes within 12 months, they buy at the 12-month mark. The outcome is the number of shares bought, compared with investing on day one.

Waiting advantage = P(start) / P(buy) - 1

P(start) = index close on the day the money was ready
P(buy)   = index close on the day the waiting investor bought
A positive value means waiting bought more shares.

A 10% dip arrived within 12 months on 41% of start days, and waiting came out ahead on 43% of them. When the dip came, waiting bought 13.7% more shares on average; when it didn't, 16.5% fewer, and 42.8% fewer at worst. The average across all start days was 4.0% fewer shares. Nearly all the successful waits ended in one of four episodes: late 2018, February–March 2020, 2022 and April 2025.

The window covers one strong stretch, and FRED's series is a price index, ignoring both dividends missed while waiting and interest on waiting cash. As a rough longer check, the same rule applied to Robert Shiller's monthly S&P data for 1871–2026 found a 10% dip within 12 months 28% of the time and an average result of 3.7% fewer shares. Shiller's prices are monthly averages, which hide some intramonth dips. Starts in 2000–2009 roughly broke even on average (0.9% more shares), so the cost of waiting depends heavily on the decade.

Lump sum vs spreading it out

Spreading a lump sum over several months, called cost averaging or dollar-cost averaging, is a milder version of waiting. Vanguard's February 2023 paper "Cost averaging: Invest now or temporarily hold your cash?" compared investing all at once with investing in three equal monthly parts, measuring wealth after one year. On MSCI World Index returns for 1976–2022, investing at once came out ahead 68% of the time. For the US alone (Russell 3000, 1979–2022) the figure was 66.4%; other markets ranged from 61.6% (emerging markets) to 68.1% (UK).

The median gap after one year was 2.2% for an all-stock portfolio and 1.2% for a 40% stock, 60% bond portfolio. With interest on the waiting cash at the T-bill rate, lump sum still won 65% of the time. Vanguard links the result to U.S. stocks beating cash 76% of the time over 1976–2022.

Cost averaging still has a role. It beat staying entirely in cash 69% of the time, and Vanguard's utility model found that investors with high aversion to both risk and losses could prefer it. It narrows the worst outcomes and gives up some of the typical one.

Valuation signals: real but blunt

Some timers use valuation instead. The best-known measure is the cyclically adjusted price-to-earnings ratio, or CAPE: the S&P 500's inflation-adjusted price divided by its average inflation-adjusted earnings over the previous 10 years. Shiller's spreadsheet also lists, for each month, the inflation-adjusted total return over the following 10 years.

Across 1,630 starting months from January 1881 to October 2016, the correlation between CAPE and the next 10 years' real return was −0.48. That makes CAPE a genuine signal: higher readings were followed by lower returns. But CAPE explained only about 23% of the variation in those returns. That is close to the 0.24 that Asness, Ilmanen and Maloney report for 1900–2015 for the Shiller earnings yield (the inverse of CAPE) in "Market Timing: Sin a Little" (Journal of Investment Management, 2017). The ranges within each band are wide:

Starting CAPE	Starting months	Lowest next-10-year real return	Median	Highest
Below 10	229	1.8% a year	11.0% a year	20.0% a year
10 to 15	461	−4.2%	7.8%	17.6%
15 to 20	514	−4.6%	6.2%	16.1%
20 to 25	260	−4.0%	4.9%	14.2%
25 to 30	109	−1.2%	5.6%	11.8%
Above 30	57	−5.9%	−1.1%	4.5%

Because the 10-year windows overlap, 1,630 months contain only about 14 independent decades, a thin base for any trading rule. A CAPE of 27.7 in December 1996 was followed by 5.7% a year after inflation; 44.2 in December 1999, by −3.2% a year.

CAPE crossed 30 in July 2017 for the first time since 2002. From then through early October 2026 the index returned about 11.1% a year after inflation, dividends included, and CAPE was above 30 in 81 of those 112 months.

The AQR authors tested a rule holding 50% to 150% of a normal stock position depending on valuation, judged only on data available at the time. Before costs it did not beat buy-and-hold's risk-adjusted return over 1900–2015, and over 1958–2015 it earned less, partly because valuations drifted upward for decades and kept the rule underweight. It flagged the market as clearly overvalued in 1991, years before the late-1990s peak, a case the authors call "early equals wrong." The signal points the right way; the timing of its payoff is what makes it so hard to trade.

When moving money is not timing

Some portfolio changes follow a plan rather than a forecast, and the SEC's Beginners' Guide to Asset Allocation, Diversification, and Rebalancing describes each of them.

Rebalancing brings a portfolio back to its chosen mix, either on a calendar (such as every six or twelve months) or when an asset class drifts past a percentage set in advance. The trigger is drift, not a market call.
Cash needs follow the time horizon: the guide notes that holding only cash investments may suit short-term goals, so setting aside money for a near-term bill is planning.
Glide paths are built into lifecycle (target-date) funds, which shift automatically toward a more conservative mix as the target year approaches.

Frequently asked questions

Does investing part of every paycheck count as dollar-cost averaging?

Not in the sense Vanguard studied, which is spreading out money already in hand. Investing each paycheck as it arrives puts money to work as soon as it exists, so nothing is held back. Vanguard's paper makes the related point that investors without a windfall can still maximize time in the market by not delaying when funds are available.

Do the break-even figures include taxes?

Only roughly, through flat costs per switch (2% for Sharpe, 1% above). In a taxable account, selling at a gain realizes it, and under IRS Topic 409 a gain on an asset held one year or less is short-term and taxed as ordinary income, while one held more than a year is long-term and may qualify for lower rates. A timer that switches every year or two would often realize short-term gains, raising the accuracy it needs.

Would waiting for a smaller or bigger dip change the result?

Over the same October 2016 to October 2025 start days, a 5% dip arrived within 12 months 54% of the time and waiting came out ahead 55% of the time, yet the average result was still 4.2% fewer shares because the misses were large. A 20% dip arrived only 15% of the time, waiting came out ahead 25% of the time, and the average result was 6.6% fewer shares. In this window, a smaller threshold raised the hit rate but not the average outcome.

Does adding a trend signal fix valuation timing?

Asness, Ilmanen and Maloney argue that valuation timing has been fighting short-term momentum, the tendency of recent market trends to continue. In their tests, combining the two improved risk-adjusted results over valuation alone, but the results are hypothetical and before costs, and even the blend produced only modest outperformance. Hence the title: small tilts, not all-in or all-out moves.

Where is the current CAPE reading published?

Robert Shiller posts it monthly in the ie_data.xls file on his data site, with data back to 1871. In the version available in October 2026, CAPE was about 40.7 based on the October 2 close, against a peak of 44.2 in December 1999. The table above shows how loosely such readings have mapped to the following decade.

Before you move money

Write down the call you are making: which asset, which direction, by when.
Compare your honest hit rate with the bar: roughly 7 correct calls in 10, more after costs.
Count the costs per switch: trading spreads and any short-term capital gains tax in a taxable account.
Check whether the move is planned (rebalancing, cash for a near-term need, a target-date schedule) rather than a prediction.
If cash is sitting idle for a "better entry," set a date by which it gets invested regardless, or split it over three to six months if a sharp fall right after investing would lead you to abandon the plan.
To see how different return assumptions change a long-term balance, run the Investment Growth calculator with a range of rates, not one expected figure.