> ## Content Index
> Fetch the complete content index at: https://www.wealthyparrot.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Is it too late to invest at all-time highs?
- URL: https://www.wealthyparrot.com/investing-at-all-time-highs/
- Published: 2026-08-26T02:37:59.000Z
- Updated: 2026-08-26T02:37:59.000Z
- Author: Carmelo 
- Tags: Personal Finance

You have money to invest. You open the app, look at the chart, and see a line that has gone almost straight up. The S&P 500 closed at 7,652.86 on 24 August 2026, a rounding error away from its own record. Some part of your brain says the obvious thing: *surely not now. Surely I wait for a dip.*  
Then you ask the internet, and the internet answers in one voice. Records are normal. Buying at all-time highs actually produces *better* returns than buying on a random day. Stop worrying, hit the button.  
I went and checked that claim against seventy years of data. The first half is true. The second half is shakier than almost anyone repeating it seems to realise.  
The conclusion still lands in the same place - waiting is a bad plan - but for a reason nobody puts on the poster. And the difference between the real reason and the poster reason changes what you should actually do with the money.

## A record is not a rare event

Using the Federal Reserve's monthly series for US share prices, which runs from January 1957 to June 2026 - 834 months, just under seventy years - **27.9% of those months closed at a new all-time high**. More than one month in four.  
Widen the lens slightly and it gets stranger. **54.8% of all months closed within 5% of the running record.** More than two-thirds, 68.6%, closed within 10% of it.  
For most of modern market history, the honest description of conditions is "at or near a record." That is the resting state of a market that rises over time, which is what a market full of profitable companies is supposed to do.

![Horizontal bar chart of how often US shares sat at or near their running record, monthly 1957 to 2026. 27.9% of the 834 months closed at a brand new all-time high, 54.8% closed within 5% of the record, and 68.6% closed within 10% of it.](https://www.wealthyparrot.com/content/images/2026/08/2026-08-25-investing-at-all-time-highs-chart-2.png)

> **Fun fact:** the S&P 500 has spent **14.1% of the last ten years' trading days** closing at a brand new all-time high - 355 out of 2,512 sessions, per Federal Reserve daily data. Roughly one session in seven set a record nobody had ever seen before.  
> Which reframes the question you were actually asking. "I'll wait for a dip" sounds like patience. Applied to seventy years of history, it means declining to invest during more than half of all available months, while your money sits somewhere earning less.  
> Framed that way it is a strategy with a price tag rather than a display of patience. We will put a number on the price tag shortly.

## What actually happens after you buy at a record

I took every month in that seventy-year series, split them into two buckets - months that set a new record, and months that did not - and measured what the index did next. The numbers below are median annualised returns.

![Table of median annualised returns by holding period, US shares 1957 to 2026, comparing purchases made at an all-time high with purchases made at any other time. At one year, 7.62% after a record against 10.18% otherwise. At three years, 5.40% against 7.51%. At five years, 5.88% against 7.41%. At ten years, 6.17% against 6.55%.](https://www.wealthyparrot.com/content/images/2026/08/2026-08-25-investing-at-all-time-highs-chart-3.png)

Buying at a record underperformed at every single horizon. Not by a catastrophe - 2.55 percentage points at one year, narrowing to 0.38 at ten - but consistently, and in the opposite direction to what you have probably been told.  
The downside numbers are sharper. Measuring how often a holding period simply ended lower than it started: after a record, **14.1% of ten-year stretches finished underwater. From a non-record month, only 3.1% did.** The worst ten-year run after a record was -25.7%, and it started in March 1999, roughly a year before the dot-com peak.  
So the doomers get one point on the board. Entry price matters. Anyone telling you it makes no difference at all is selling reassurance rather than arithmetic.  
Now read the same table the other way, because both things are true at once.  
A median of 7.62% over the following year. 6.17% annualised over the following decade. Those are the returns from buying at *the worst possible-sounding moment*, and they are perfectly good returns. The market at a record has historically handed you something in the 6-8% range for your patience. It just handed slightly more to people who bought at other moments.  
The worst single year for someone buying at a record was **\-39.7%**, from an October 2007 purchase. The worst decade was -25.7%, from March 1999\. Both are real, both are brutal, and both sit at roughly one-in-seven rather than being the default.

![Horizontal bar chart of the return penalty for buying at an all-time high, by holding period, US shares 1957 to 2026. The gap in median annualised return is -2.55 percentage points at one year, -2.12 at three years, -1.54 at five years and -0.38 at ten years.](https://www.wealthyparrot.com/content/images/2026/08/2026-08-25-investing-at-all-time-highs-chart-1.png)

## Why the claim everyone repeats is shakier than it looks

You have almost certainly seen the opposite finding. Buying at all-time highs beats buying on random days - it appears in bank research notes, in popular finance blogs, in the videos that racked up hundreds of thousands of views this year.  
The finding is real. It is also fragile.  
When I re-ran the same analysis on **1988 to 2026**, the window most of those studies use, the picture partly flipped. Records still lost at one, three and five years, but at ten years they *won* by 1.03 percentage points. Run it from 1995 and the one-year gap vanishes to nothing while the three-year gap widens to -3.52 points against records.  
Same method, same index, three windows, three different answers. A finding that changes sign depending on where you start the clock is a finding to hold loosely.  
Two limitations of mine, stated plainly, because a piece that just debunked someone else's numbers owes you its own.

**This is a price index. Dividends are excluded.** That matters, and it is the one caveat that does not rescue the popular claim - it deepens the problem for it. Dividend yields are highest after crashes, exactly the non-record months. Adding dividends back would lift non-record entry points *more* than record ones, which widens the gap rather than closing it.

**Monthly data smooths things.** Someone measuring daily closes, or using a total-return index, or defining "all-time high" slightly differently will get numbers that differ from mine at the second decimal.

There is a third problem, and it belongs to everyone working with market history. Long-run index series tend to flatter themselves. Elroy Dimson, Paul Marsh and Mike Staunton found that the standard long-run UK equity index had been assembled backwards from the companies that happened to be in it when it launched, quietly leaving out those that died along the way. Corrected, the real return from 1900 to 1954 was **3.8%** a year, well below what the published history implied.  
What survives all of that is the modest version: buying at a record produces solidly positive returns, somewhat lower than other entry points, with a meaningfully fatter tail of bad outcomes. Anyone claiming records are *actively bullish* is over-reading their own window.

## Is this just America?

Every number above comes from US data, and that is a genuine weakness, because the United States is the winner of the twentieth century. A pattern found there may be describing American good fortune rather than how markets work.  
Dimson, Marsh and Staunton built the dataset that tests this, covering sixteen countries from 1900 to 2000\. One unit of local currency put into equities in 1900, dividends reinvested and inflation stripped out, grew to **12.3** in Belgium, the worst of the sixteen, and to just under **1,700** in Sweden, the best. The United States sits among the winners, next to Sweden, Australia, South Africa and Canada.  
That is a 138-fold spread across sixteen developed markets, and the authors point out their sample omits stock market fatalities entirely - the exchanges that shut and never reopened. The true dispersion was wider.

> **Fun fact:** Japanese equities were **40% of world market capitalisation** at the end of 1988, briefly overtaking the United States. The Nikkei 225 peaked at **38,915.87** on 29 December 1989, bottomed at **7,054.98** in March 2009 - down **81.9%** \- and did not close above its 1989 peak again until **22 February 2024**, per Federal Reserve data. **34 years.**

None of that makes records dangerous. It means a rule extracted from one unusually fortunate market deserves to be held more loosely than the confident version you normally hear, mine included. It is also the argument for [owning more than one country](https://www.wealthyparrot.com/why-smart-investors-are-finally-going-global/).

## The real argument against waiting

If buying at a record is mildly worse, shouldn't you wait?  
No. And this is the part the poster version gets right by accident.  
The alternative to buying is not "buying later at a better price." The alternative is **holding cash until some future moment you have to correctly identify**, and that second job is the one that destroys returns.  
Nick Maggiulli ran the definitive version of this in *Just Keep Buying*, and the results are unkind to patience. Investing a lump sum immediately beat averaging it in slowly across **76% of all rolling twelve-month periods since 1997**.  
That is the easy finding. The devastating one is next.  
He simulated an investor with *perfect knowledge of the future* \- someone who knew the exact bottom of every crash in advance and held cash until it arrived. That omniscient investor **still lost to boring monthly investing about 70% of the time.** God-tier market timing, beaten by someone who just bought on the first of the month and went back to their life.

> **Fun fact:** relax that superpower even slightly - miss the exact bottom by **two months** \- and buy-the-dip underperforms plain dollar-cost averaging **97% of the time**, per Maggiulli's simulation of US market history from 1920 to 1980\. Two months of imprecision turns a winning strategy into a near-certain loser.

The formal version of this is older than the simulations. George Constantinides showed in the *Journal of Financial and Quantitative Analysis* in 1979 that dollar-cost averaging is suboptimal as an investment policy - not merely unlucky in backtests, but dominated on the mathematics.  
The mechanism is simple. Waiting costs you time in the market, and time in the market is where the returns live. To profit from waiting, you must be right twice - right that a fall is coming, and right about when to go back in. Almost nobody is right twice, including people who were right once and are now very confident.  
Meanwhile, look at what the waiting room pays. The US federal funds rate sat at **3.63%** in July 2026, while US inflation ran at **3.30%**. Your cash earns roughly **0.3% in real terms**. In the euro area, inflation was **2.9%** in July 2026 per Eurostat, and most instant-access savings accounts pay less than that outright. Cash still has a job - [the buffer you keep for emergencies](https://www.wealthyparrot.com/the-emergency-fund-paradox/) is not the money we are discussing here.  
So "waiting for a dip" carries a running cost: the gap between roughly 0.3% real and a historical 6-8% nominal, compounding for however long your discipline holds. If the dip arrives in three months you win a little. If it arrives in three years you have paid an enormous fee for a small discount.  
Ever moved money to a savings account "temporarily" while you waited for a better entry point? How long did temporarily turn out to be?

## What actually deserves your attention instead

The all-time-high question is loud and mostly unproductive. Three quieter things move your outcome far more.

### The price you pay, not the price on the chart

Howard Marks spends much of *The Most Important Thing* on a distinction that is easy to miss, and one that gets buried under [the Greek letters the industry uses for risk](https://www.wealthyparrot.com/the-greek-alphabet-conspiracy/). A high index level is not the same as an expensive market. The level reflects seventy years of accumulated earnings growth. Expensive means paying too much *per unit of those earnings*.  
His example is the **Nifty Fifty** \- the fifty companies American investors treated as unimpeachable in the early 1970s. Coca-Cola, IBM, Xerox. Genuinely excellent businesses. Investors bid them to price-to-earnings ratios of 80 to 90, and they subsequently fell to P/E ratios of 8 or 9\. Roughly a 90% loss, with nothing wrong with the companies at all.

> **Fun fact:** Marks also points out that between 1982 and 1999, the greatest drop the US market experienced was **5%**. An entire generation of investors reached the late nineties having never seen a real decline, which is precisely why they were positioned as though one could not happen.

How much valuation actually predicts is contested, and worth knowing before anyone waves a CAPE ratio at you. John Campbell and Robert Shiller showed in 1998 that valuation ratios forecast long-run returns. Ivo Welch and Amit Goyal later demonstrated that most such predictors fall apart out of sample, working beautifully in the period they were discovered in and poorly afterwards. Both findings have survived. Together they suggest valuation tells you something about the range of outcomes and very little about next year.  
A record level tells you almost nothing on its own. Valuation tells you something, and the two get confused constantly because one is printed on the front page and the other requires arithmetic.

### The mix you hold

You have probably met the claim that asset allocation explains "over 90%" of investment returns. It comes from Gary Brinson's pension fund studies, and it is one of the most misquoted numbers in finance. I had it wrong in an earlier version of this article too.  
Roger Ibbotson and Paul Kaplan untangled it in a 2000 paper whose title is the entire problem: *Does Asset Allocation Policy Explain 40, 90, or 100 Percent of Performance?* All three numbers are correct, for three different questions. About **90%** of the variability of a single fund's returns **over time** comes from its allocation. About **40%** of the difference **between** funds does. About 100% of the return level.  
Brinson measured the first. Nearly everyone quotes it as though it answered the second. Ibbotson and Kaplan put it plainly: the studies "are often misinterpreted and the results applied to questions that the studies never intended to answer."  
The 40% is your number, because you are choosing between portfolios rather than watching one wobble. Less dramatic than 90%, and still vastly more than your entry month is worth.  
If a record makes you uneasy, the productive response is to check whether your split between stocks, bonds and cash matches your situation, and whether you [rebalance it on any schedule at all](https://www.wealthyparrot.com/shannons-demon/). That is a real answer to a real anxiety. "Wait a bit" is not.

### The length of time you can leave it alone

Bernstein's shortfall numbers take ten seconds to absorb. The probability that stocks underperform short-term government bills is **36% over one year, 13% over ten years, and 1% over forty years.**  
Your holding period does more work than your entry point, which is why [working out your actual timeline](https://www.wealthyparrot.com/fire-calculator/) beats agonising over a start date. Buying at an inconvenient moment with a thirty-year horizon is a rounding error. Buying at a perfect moment with an eighteen-month horizon is a coin flip regardless of how clever the timing was.  
Fairness demands the other side, because this one is genuinely unsettled. Paul Samuelson spent years attacking exactly this reasoning. A long horizon raises the *probability* of a gain, he showed, while raising the *magnitude* of the loss you can suffer. Repeating a risky bet does not cancel the risk, it enlarges the stakes. Both statements are true, and which one you weight depends on whether the odds of a bad outcome worry you more than its size.  
This is also why the ten-year column in my table matters more than the one-year column. The gap between buying at a record and buying elsewhere shrinks to 0.38 percentage points once you hold for a decade. Time does not erase the difference, but it grinds it down to something that barely registers next to the decisions you actually control.

## So what do you do with the money?

Invest it. The alternatives are measurably worse, and the gap between a good entry month and a bad one is smaller than the anxiety suggests.  
A few honest refinements.  
If the lump sum is large enough that a 30% drop would genuinely change your life, your allocation is too aggressive for your situation, and moving the entry date does nothing about that. Victor Haghani and James White set out the variable structure in *The Missing Billionaires*: the right position size rises with expected excess return, falls with variance, and scales to your own tolerance for risk. The date is not in the formula. Overshoot the size and you end up, in their phrase, "exponentially worse off than doing nothing at all." Maggiulli's finding points the same way: a lump sum into a conservative 60/40 portfolio produced better risk-adjusted results than slowly averaging into an all-stock one. Fix the mix, then buy.  
If you genuinely cannot press the button, splitting the purchase across a few months is a legitimate anxiety tax. It costs you expected return - 76% of the time, per the data above - and it buys you the ability to sleep and stay invested. A slightly worse plan you stick to beats a better one you abandon in month three.  
And if you find yourself certain the market is overvalued right now, notice that this is a forecast. Marks looked at the record of professionals making exactly these calls: the average forecaster missed the thirty-year bond rate by **96 basis points** six months out, which on a $1,000 bond is a $120 swing. These are people paid full-time to be right about this.

> **Fun fact:** just **4% of listed US companies** account for the entire net gain of the US stock market since 1926, according to Hendrik Bessembinder's 2018 study in the *Journal of Financial Economics*. The other 96% collectively matched Treasury bills. Which is the strongest argument going for owning the whole haystack rather than trying to time your entry into it.

## Practical takeaways

- **A record is the market's normal state.** 27.9% of months since 1957 set one, and 54.8% closed within 5% of one. Waiting for a dip means opting out of most of market history.
- **Buying at a record is mildly worse, not dangerous.** Median forward returns of 7.62% at one year and 6.17% annualised at ten - a bit below other entry points, with a fatter tail of bad decades (14.1% of ten-year runs ended lower, against 3.1% from non-record months).
- **Treat "highs predict higher returns" as unproven.** It flips sign depending on the window. Anyone stating it confidently has not tested it on more than one period.
- **Waiting is a strategy with a price tag.** Cash returned roughly 0.3% real in the US in July 2026\. Perfect-foresight dip-buying still loses to monthly investing 70% of the time; miss the bottom by two months and it loses 97% of the time.
- **Spend your worry on allocation, not entry.** Roughly **40%** of the difference between portfolios traces to allocation (Ibbotson & Kaplan, 2000) - not the "90%" everyone quotes, but far more than your entry month is worth.
- **One market's history is not a law of nature.** Across sixteen countries from 1900 to 2000, a unit invested in equities grew to 12.3 in Belgium and \~1,700 in Sweden. Japan took 34 years to regain its 1989 peak.
- **Lengthen the horizon and the question mostly dissolves - though not everyone agrees.** Stocks trail T-bills in 36% of one-year periods and 1% of forty-year periods; Samuelson's counter is that a longer horizon shrinks the odds of a loss while enlarging its possible size.
- **A sum large enough to frighten you is telling you about your allocation.** Adjust the mix rather than the date.
- Rules and account types differ by country - check your local regulations before acting on any of this.

## One last thing

The market spent more than half of the last seventy years within touching distance of a record. If you had treated that as a reason to hold off, you would have spent most of your investing life on the sidelines, waiting for permission that the data was never going to give you.

The dip you are waiting for will arrive. It always does. What the record cannot tell you - what nothing can tell you - is whether it arrives from a level well above where you are standing right now.