Myth
Investing at all-time highs: is it a bad time to buy?
Busted: buying SPY on a record close has not led to worse returns. After the 675 record closes with a full year of data since 1994, SPY's median return over the next 12 months was +16.5%, against +14.8% from any day.
The myth, and why people believe it
The idea feels like common sense: if prices are at a record, you are paying more than anyone ever has, so a fall must be close. An all-time high is a closing price above every earlier close. Buying on one feels like buying at the top, and everyone remembers someone who bought in early 2000 or late 2007, just before a crash.
The test
How we tested this
We took every trading day of SPY, the S&P 500 in one fund, from 29 January 1993 to 20 November 2026. A day counts as a record when its closing price matched or beat every earlier close; the first year is left out, so the early days don't all count as records. For each record day we measured what money invested that day was worth 12 months, 3 years and 5 years later, with dividends reinvested, and compared it with the same measure from every trading day after the first year. We fixed the verdict lines before running it: the myth is busted if record-day buyers did no worse than any-day buyers by more than a point over 12 months, half a point a year over 5 years, and two points in how often they were up after a year.
The result
Records are not rare. SPY closed at a record on 8.5% of trading days, 705 in all, and they come in clusters: a rising market makes record after record. After a record, the next 12 months were usually good, and a little better than after an ordinary day:
- Record closes (675 with a full window, 705 in all, 8.5% of trading days): median 12-month return +16.5%, positive 85.6% of the time, worst tenth −5.0%
- Any day: median +14.8%, positive 81.4%, worst tenth −12.6%
- Fell 10% or more below the price paid within 12 months: 28.4% of record closes, 32.1% of all days
- Longest wait for a new record: 24 March 2000 to 1 June 2007 (7.2 years)
| Window | Median, records | Median, any day | Positive, records | Positive, any day | Worst tenth, records | Worst tenth, any day |
|---|---|---|---|---|---|---|
| 12 months | +16.5% | +14.8% | 85.6% | 81.4% | −5.0% | −12.6% |
| 3 years (a year) | +10.5% | +11.9% | 91.9% | 83.1% | +3.2% | −6.0% |
| 5 years (a year) | +12.0% | +11.1% | 84.0% | 84.3% | −2.0% | −1.2% |
Record closes on plain closing prices, returns on dividend-adjusted closes. Indices Lab study on SPY, dividends reinvested, 29 January 1993 to 20 November 2026. Data: EODHD. Hypothetical results: past performance does not predict future returns.
Buying at a record did slightly better on the median and was up after a year more often. The worst tenth of 12-month outcomes was also milder after a record, because the deepest losses in the history came from buying partway through a crash, not at the top before it. And within 12 months of buying, SPY fell at least a tenth below the price paid 28.4% of the time after a record, against 32.1% after any day.
Where the myth has a point
Three places:
- The two famous peaks. SPY's record of 24 March 2000 was its last until 1 June 2007, a gap of 7.2 years. Anyone who bought at those records waited years to be ahead. Over 5 years, the worst tenth of record-day starts did a little worse than the worst tenth of any-day starts (−2.0% against −1.2% a year), and record days were up after 5 years slightly less often (84.0% against 84.3%).
- Since late 2009. Split the start dates in two and the edge flips. In the first half, record days beat any day by a wide margin. In the second, the median 12-month return after a record was +14.6% against +16.0% from any day, and record days were up after a year less often (81.9% against 89.7%). In that stretch an ordinary day, often one after a fall, was the better time to buy.
- Three years out. The 3-year median after a record was lower than from any day (+10.5% against +11.9% a year), even though record-day starts were up after 3 years more often (91.9% against 83.1%).
Records bunch into long runs and then stop for years, as the count by year shows:
- 705 record closes in 8,260 trading days (8.5%)
- Full years with no record: 2001, 2002, 2003, 2004, 2005, 2006, 2008, 2009, 2010, 2011, 2012, 2023
- Last record close: 6 October 2026
| Year | Record closes | Trading days | Share |
|---|---|---|---|
| 1994 | 4 | 234 | 1.7% |
| 1995 | 73 | 252 | 29.0% |
| 1996 | 35 | 254 | 13.8% |
| 1997 | 42 | 253 | 16.6% |
| 1998 | 44 | 252 | 17.5% |
| 1999 | 29 | 252 | 11.5% |
| 2000 | 5 | 252 | 2.0% |
| 2001 | 0 | 248 | 0.0% |
| 2002 | 0 | 252 | 0.0% |
| 2003 | 0 | 252 | 0.0% |
| 2004 | 0 | 252 | 0.0% |
| 2005 | 0 | 252 | 0.0% |
| 2006 | 0 | 251 | 0.0% |
| 2007 | 7 | 251 | 2.8% |
| 2008 | 0 | 253 | 0.0% |
| 2009 | 0 | 252 | 0.0% |
| 2010 | 0 | 252 | 0.0% |
| 2011 | 0 | 252 | 0.0% |
| 2012 | 0 | 250 | 0.0% |
| 2013 | 48 | 252 | 19.0% |
| 2014 | 52 | 252 | 20.6% |
| 2015 | 11 | 252 | 4.4% |
| 2016 | 18 | 252 | 7.1% |
| 2017 | 54 | 251 | 21.5% |
| 2018 | 19 | 251 | 7.6% |
| 2019 | 34 | 252 | 13.5% |
| 2020 | 33 | 253 | 13.0% |
| 2021 | 71 | 252 | 28.2% |
| 2022 | 1 | 251 | 0.4% |
| 2023 | 0 | 250 | 0.0% |
| 2024 | 58 | 252 | 23.0% |
| 2025 | 40 | 250 | 16.0% |
| 2026 | 27 | 224 | 12.1% |
Record closes per year on plain closing prices. Indices Lab study on SPY, dividends reinvested, 29 January 1993 to 20 November 2026. Data: EODHD. Hypothetical results: past performance does not predict future returns.
None of these crosses our line, and none is a signal you could use in advance: on the day, you can't know which record is the last one for seven years.
What it means for your portfolio
A record high on its own has not been a reason to wait. The cost of waiting shows up in two other myths we tested: buying the dip found that waiting for a fall left the buyer behind or level in most start months, and the golden cross found that a timing rule gave up growth to cut the worst fall. What a record should change is not when you buy but what you hold: if the worst fall in your history would make you sell, the mix is too risky for you at any price level. Max drawdown explained shows how to measure that, and how to backtest a portfolio shows how to test your own mix over every start date.
What this test can't show: one fund, one country, about three decades. Record days overlap heavily, because records cluster, so they are far fewer independent bets than their number suggests. Valuations, interest rates and the reason for the record are not in the test.
Questions
Should I invest when the market is at an all-time high?
In our test of SPY since 1994, buying on a record close did no worse than buying on any day: the median 12-month return was +16.5% against +14.8%. That is history, not a promise about the next record.
What usually happens after the stock market hits a record?
More records, most of the time. Records cluster in rising markets, and SPY was higher 12 months later after 85.6% of its record closes.
Should I wait for a pullback before investing?
Waiting has its own cost. Our buy the dip test found that waiting for a fall left the buyer behind or level in most start months since 1993.
Is it better to invest all at once or spread it out at a high?
This test only covers single purchases. Spreading money over months lowers the regret of buying just before a fall but, in a rising market, usually leaves money in cash for longer.
How often does the S&P 500 hit an all-time high?
SPY closed at a record on 8.5% of trading days in our test, with long gaps: none from 24 March 2000 to 1 June 2007.
About this page. Figures on this page come from Indices Lab’s own data, powered by EODHD, and every number is recalculated after each market close; the text is reviewed by a person before it is published. Indices Lab is a research and educational tool: everything here is hypothetical and informational, not investment advice and not a recommendation to buy or sell any security. Past performance does not predict future results. See our methodology.