Cash sitting inside a long-term portfolio has one cost you can measure and one benefit you can't. Start with the cost. Over the 126 years to the end of 2025, US Treasury bills returned 0.5% a year after inflation, against 6.6% for US equities. A dollar in bills grew to $1.80 in real terms. A dollar in equities grew to $3,296.
That gap is the entire argument in one line. Everything else is about whether the option value of cash earns it back — the ability to buy something cheap, or to avoid selling something cheap.
This is about strategic cash held deliberately inside an invested portfolio. It isn't about a reserve for job loss or a broken boiler, which is a different question with a different answer, covered in the guide to sizing an emergency fund against income volatility.
The arithmetic of a permanent cash sleeve
Take the Dimson-Marsh-Staunton long-run numbers at face value: 6.6% real for equities, 0.5% real for bills. A portfolio holding 10% cash permanently earns about 6.0% real instead of 6.6%. That's a drag of 0.61 percentage points a year.
Compounded, it's bigger than it looks. Over 30 years, the 10% sleeve ends up with roughly 16% less real wealth. At 20% cash the drag is 1.22 points a year and the 30-year shortfall reaches 29%. At 5% cash it's 8%. These are projections of a historical average, not forecasts, and the average blends wildly different regimes.
Nominal numbers flatter cash badly. US bills returned 3.4% a year nominally, against 2.9% inflation over the same 126 years. The headline looks like a return. Almost all of it was inflation.
Has waiting to deploy actually paid?
AQR tested this head-on in late 2025. Its Portfolio Solutions Group built 196 versions of a buy-the-dip strategy on the S&P 500 — four dip depths, seven dip lengths, seven holding periods — and ran them from 1 January 1965 to 30 September 2025. Whenever no dip signal fired, the strategy sat in cash, proxied by the ICE BofA 3-Month T-Bill Index. That is dry powder, mechanised.
The average version produced a Sharpe ratio 0.04 below passive equities, about a 16% degradation. Over 60% of the 196 versions underperformed on that measure. On the shorter window from October 1989, where S&P 500 total returns including dividends are available, the gap widens to 0.27 — a 47% degradation.
Total returns say the same. Passive equities beat the average buy-the-dip strategy by 1.1 percentage points a year, and beat the 25th-percentile version by 1.7 points. Even the 75th-percentile version, the flattering one, still trailed by 0.6 points.
AQR also offers a diagnosis. Buy-the-dip is a bet on reversal placed at a horizon where markets have historically trended. In their phrase, it's value investing at a momentum horizon. That is why the versions holding for a month or two fare worst.
The alpha test is the one that matters for dry powder. AQR regressed each strategy's excess-of-cash return against buy-and-hold. Average alpha was 0.5% a year. Only 16 of the 196 versions — 8% — cleared the conventional significance threshold. The authors point out that with 196 correlated tests, an honest threshold is around a t-statistic of 3.52, which none of them reached.
Two caveats belong right next to those figures. The 1965-2025 series uses S&P 500 price returns only, because daily total-return data starts in 1989, and AQR reports dip-buying looks worse over the shorter total-return window. The tests are also gross of fees and trading costs, which would hurt the higher-turnover versions most. Both caveats cut against dry powder, not for it.
Feeding cash in on a schedule doesn't rescue it
Vanguard ran the milder version of the question: hold cash briefly and drip it in, or invest it all now. Splitting a lump sum into three monthly instalments and measuring wealth after one year, the immediate lump sum won 68% of the time on MSCI World returns from 1976 to 2022.
The UK result is nearly identical. On FTSE All-Share data from 1986 to 2022, investing at once beat a three-month split 68.1% of the time and a six-month split 69.5% of the time. In the US the six-month split lost 73.7% of the time. Waiting longer made things worse, not better, in every market tested except Europe.
The mechanism is mechanical rather than clever. Over 1976 to 2022, US stocks beat cash 76% of the time and bonds beat cash 68% of the time. Time in cash is time not collecting a risk premium.
Paying interest on the uninvested balance narrows the gap without closing it. With cash earning the 3-month US Treasury bill rate, immediate investment still won 65% of the time for an all-equity portfolio. The higher the cash rate, the smaller the edge.
Vanguard is straight about where the cash-heavy path wins. Below the 25th percentile of outcomes — the bad markets — cost averaging finishes ahead. In a 60/40 portfolio the fifth-percentile outcome was $94,043 for cost averaging against $92,720 for the lump sum. That's the insurance being bought, and the median premium was about 1.8% of the pot over a single year.
The case for dry powder is better than the return series suggests
Now the strongest counter-argument, which has real evidence behind it. Cash isn't only a return stream. It's an option, and options don't show up in an annualised return table.
Mikhail Simutin tested exactly this on US actively managed equity funds from 1992 to 2009. The average fund held 4% of assets in cash, but the spread was enormous: the bottom decile held 0.18%, the top decile over 9%. Funds carrying high excess cash — cash above what their size, flows, expenses and portfolio risk predicted — beat their low excess cash peers by over 2% a year in raw returns, and by nearly 3% after adjusting for the Carhart four factors.
Read the mechanism carefully, because it isn't the one dry-powder advocates usually claim. Simutin found no significant relation between raw cash levels and future returns. Only excess cash predicted anything. And the outperformance traced back to stock selection and to meeting redemptions without fire sales, not to timing market downturns. Cash was a marker of skill and flexibility, not a timing device.
Corporate balance sheets point the same way. Bates, Kahle and Stulz documented the average cash-to-assets ratio of US industrial firms rising 129%, from 10.48% in 1980 to 24.03% in 2004. Their explanation is precautionary. Cash flows got riskier, inventories and receivables shrank, R&D spending rose. Firms accepted the low return on cash to buy insulation from having to raise money at a terrible moment.
What a return series can't see
Critics of the pure-drag view make the liquidity argument best, and they're right as far as it goes. A return series records what cash earned. It doesn't record what cash prevented.
It can't see the sale you didn't have to make in March 2020. It can't see the rebalancing trade funded without selling into a falling market, the mechanics of which are set out in the analysis of rebalancing through a market crash. It can't see the buffer that keeps someone invested through a 30% fall, which matters given what the gap between fund returns and investor returns costs.
But notice what that concedes. Every one of those benefits is about not selling. None is about buying low. Waiting-to-deploy cash and don't-force-me-to-sell cash are different instruments wearing the same name, and the evidence supports the second far better than the first.
Scale is the other problem with the option. Vanguard UK published an illustration of an investor with catastrophic timing: £45,000 fed into the FTSE All-World in seven lump sums, each placed immediately before a crisis, from September 1997 to February 2026. That unluckiest investor finished with £197,963. The same £45,000 left in cash finished at £63,980.
Treat those numbers as an illustration, not a study. Vanguard proxies cash with Bank of England rates, notes that one month's rate was unavailable and carried forward, discloses no fees or tax on either side, and published the piece as client marketing rather than peer-reviewed research. The direction is still hard to dismiss. The worst plausible entry timing beat cash by roughly £134,000 over 28 years.
The 0.5% average hides two very different worlds
A 126-year average isn't a forecast for the next five years. Cash's real return is just the short rate minus inflation, and that number swings hard. There have been long stretches when bills lost real value every single year, and stretches when they beat inflation comfortably.
That matters now. UK CPI rose 2.6% in the 12 months to June 2026, down from 2.8% in May, on the ONS release of 22 July 2026. When deposit and money-market rates sit above that, cash earns a positive real return and the drag arithmetic softens. It doesn't vanish, because 6.6% real for equities remains the comparison, but the option gets cheaper to hold.
Plenty of capital has taken that view. US money market fund assets stood at $7.85 trillion in the week ended 29 July 2026, on Investment Company Institute data, split $3.08 trillion retail and $4.78 trillion institutional.
One warning about that figure. A large money-market balance is often described as fuel waiting to enter the market, but it's a stock, not a queue. Someone holds every pound of it at every moment. Its size says how attractive short rates look, not what equities will do next.
Where the argument gets slippery
Two framing errors do most of the damage.
The first is treating a decision never made as a decision made. Cash accumulates by default. A dividend lands, a bonus arrives, a holding gets sold, and the balance simply stays there. Vanguard calls this cash-hoarding by indecision, and it isn't a strategy — it's the absence of one. Written allocation rules are the standard remedy, which is the argument in the piece on pre-commitment through an investment policy statement.
The second is counting the wins and forgetting the wait. Anyone who deployed cash in March 2020 remembers it vividly. The years of drag beforehand are far less memorable. AQR's warning about datamining across 196 strategies applies to personal experience too. Given enough dips, some deployments look brilliant purely by chance.
What would change the conclusion
Four things, honestly.
First, a sustained regime of positive real cash rates alongside a compressed equity risk premium. The 6.6%-versus-0.5% gap is a 126-year average, and starting valuations matter. If the forward gap were two points rather than six, a 10% cash sleeve would cost roughly a third of what the history implies.
Second, evidence that dip-buying works at a horizon people can actually hold. AQR found that defining a dip over five years produced a higher average Sharpe ratio of 0.21 and average alpha of 0.8% — though only at a t-statistic of 1.0, and, as they note, few investors would call a five-year drawdown a dip.
Third, a household version of the Simutin result. His finding covers professional managers facing redemption pressure, and flexibility is what paid, not timing. Whether a private portfolio's cash buys the same flexibility simply hasn't been tested the same way.
Fourth, costs and tax. AQR's tests are gross of both. In a taxable account a strategy that repeatedly buys and sells looks worse still; inside a tax wrapper the picture is cleaner. Neither flips the sign of the result, but both change its size.
On the evidence available, portfolio cash is mostly drag with a small, genuine option attached, and that option is worth more for not selling than for buying. Where a cash position does a specific job, the size of the job sets the size of the position. Where it exists because a decision got deferred, the arithmetic above is the price of deferral. The part most portfolios skip is measuring the cash line as an allocation rather than a residual, which is what a tracker like LedgerTouch exists to do.