Can you Lose Money in Nifty 50 if you Invest for 10 Years?

"Equity is risky in the short run but safe in the long run" is the most repeated line in Indian personal finance. It's also rarely backed with an actual number. How safe? One-in-ten odds of loss? One-in-a-hundred? Zero?

This piece answers that with 34 years of Nifty 50 data, every trading day from July 1990 to October 2024, and a 2025 academic study that maps the same ground. The short version: the popular claim is roughly true, but only once you understand two things almost nobody mentions, the "trapping period," and the fact that when you bought has historically mattered more than how long you waited.

A note on what we're measuring: everything below uses the Nifty 50 price index (the headline index level), which excludes dividends. We look at CAGR (compound annual growth rate), the single yearly rate that takes your starting value to your ending value, because it lets us compare a 5-year and a 10-year outcome on the same scale.

First, the One-Year Reality: a Coin that's Lucky, Not Loaded

Over a single year, the Nifty has historically been positive about 74% of the time. Analysing daily Nifty data from 1990–2024, found the probability of a positive one-year return is 74%, with the most common outcome a gain of about 10.7%.

Flip that around: roughly one year in four ends in a loss. That's why a 12-month horizon isn't investing; it's a bet with decent but unreliable odds. The whole case for "long-term" rests on what happens when you extend the horizon. So let's extend it.

The "Trapping Period": the Idea that Explains Everything

For any holding period, you can ask: What was the single worst outcome in history? Not the average, the floor.

The study calls the stretch where that floor is still negative the "trapping period", the span over which an unlucky investor who bought at the worst possible moment was still underwater. Their finding, across the full 1990–2024 history: worst-case returns remain negative for up to ten years, defining a historical "trapping period." This horizon shortens to six years in the post-1999 period.

In plain terms, measured over all 34 years, you needed to hold roughly ten years to be certain (historically) of not losing money. But measured only from 1999 onward, that danger zone shrank to about six years. The market got faster at recovering.

Here are the study's worst-case, best-case, and most-likely annual returns by holding period:

Read the "worst CAGR" column top to bottom. Even the worst 10-year stretch in history only cost about 3% a year. And by year 11, the worst case in 34 years of data finally turns positive. The downside doesn't just shrink with time; it shrinks toward zero.

Now the Actual Probabilities: From the Raw Data

The study gives worst-case floors, not the odds of loss. So we computed those directly from the 8,325-day price series. The method: take every month as a possible starting point, look at where you'd be 5, 7, or 10 years later, and count what fraction of those windows ended below where they started.

Lump Sum: Probability of Ending with a Loss:

Nifty 50 price index. Full history: 328 seven-year windows, 292 ten-year windows. Monthly starting points. Dividends excluded.

This is the answer to the title. Investing a lump sum for 7 years, you lost money in about 1 of every 18 historical windows, and in zero of the 226 windows that started in 1999 or later. For 10 years, even counting the brutal 1990s, fewer than 1 in 100 windows lost money, and the worst of those was a loss of about 1% a year, effectively flat, not wiped out. Since 1999, no 10-year window has lost money at all.

Every single losing window, in both horizons, started in the 1990s, the era of the Harshad Mehta aftermath and a thin, young market. That's the trapping period showing up as a probability instead of a floor.

The Catch Nobody Mentions: SIPs Don't Lower These Odds

Most readers will assume a SIP, investing a fixed amount monthly, is automatically safer. On the size of a bad outcome, it is: spreading purchases smooths your entry. But on the raw probability of loss, the data says something blunter.

We ran the same windows for a monthly SIP, measuring each window's XIRR (the annualised return that accounts for money going in at different times):

At every horizon in the full history, the SIP was more likely to end in a loss, not less. The reason is mechanical: a SIP keeps buying right up to the end, so a chunk of your money has only been invested for a few months when the clock stops. If a downturn hits near the end of your 7 years, the lump sum has years of cushion built up; the SIP's recent instalments are sitting in the red with no time to recover.

This isn't an argument against SIPs; they solve a different problem (they stop you from betting everything at a single unlucky moment, and they match how salaries actually arrive). But "SIP makes long-term loss impossible" is not what history shows. Time in the market lowers your odds of loss. Averaging your entry lowers the severity of a bad sequence. They're different tools. (Note too that since 1999, SIP loss probability at 7 and 10 years is also 0.0%, the trapping-period effect dominates either way.)

The Real Lever: What you Paid Going in

If the holding period were the whole story, the advice would simply be "wait longer." But the study found a second variable that mattered more: the P/E ratio at the moment you invested. P/E, price-to-earnings, is how expensive the index is relative to company profits; a high P/E means you're paying more for each rupee of earnings.

Splitting history by entry valuation, the study found low valuations (P/E < 13) historically show zero probability of loss across all horizons, while high valuations (P/E > 27) correspond to unstable returns and extended breakeven periods. The full ladder is described:

The investors who got "trapped" weren't unlucky about time; they were unlucky, or careless, about price. They bought when the index was expensive. Buy when the market is cheap, and history shows essentially no loss at any horizon; buy at a euphoric peak, and you can wait years just to break even, even over a "long-term" holding period.

For context, the Nifty's long-run average P/E sits around 20–21, and it has spent meaningful time above 27, most recently in the liquidity-driven 2021 rally. Valuation isn't a market-timing crystal ball, but it's the closest thing the data offers to a dial you can actually read before investing.

Things to Keep in Mind

This is history, not a guarantee. A few honest limits:

• Price index, not total return. We excluded dividends, which makes these loss odds slightly worse than reality. With dividends reinvested, the picture is marginally better, so the conclusions are conservative, not inflated.
• Past patterns can break. "No 7-year loss since 1999" describes 226 historical windows. It is not a promise about the next one. A structural break, a prolonged earnings collapse, a decade-long stagnation like Japan's, would not be visible in this record.
• The 1990s were a different market. Much of the historical loss risk comes from a young, thin, lightly-regulated market. That's genuine information (it shows how an immature market behaves), but the modern Nifty is a different animal.
• One index, one country. This is the Nifty 50, large-cap India. A single stock, a small-cap fund, or a sector bet carries far higher and differently shaped loss odds. None of this transfers to individual stocks.
• Overlapping windows. Rolling windows share data, so these aren't 300 fully independent experiments. The direction is robust; treat the exact decimals as well-grounded estimates, not laboratory constants.

The Bottom Line

Is equity "safe in the long run"? For the Nifty 50, the honest, numbers-backed answer is: mostly yes, and increasingly so.

Hold a lump sum for 7 years, and your historical odds of loss were about 5% across all 34 years, and zero since 1999. Hold for 10 years, and the odds were under 1%, even counting the worst of the 1990s. The downside doesn't just become unlikely; when it did occur, it was shallow, a few per cent a year, not a wipeout.