Macro · 8 min read
A strategic, annually-rebalanced gold allocation of 10–20% has not been a low-return insurance policy on the Nifty over the past 30 years — it has been a genuine free lunch. Backtesting 0%, 10% and 20% gold weights against the Nifty 50 from 1996 to 2026 shows the 20% mix produced both a higher compound return and a meaningfully shallower, faster-recovering drawdown profile than an all-equity portfolio. The case for gold in an Indian portfolio doesn't rest on timing its rallies — it rests on this 30-year mechanical record.
Gold just finished one of its steepest two-year rallies on record: from ~$2,618/oz in December 2024 to an all-time high of $5,247.9/oz in February 2026 — a near-doubling — before correcting 16.8% to ~$4,367/oz by September 2026. This is exactly the moment investors ask "did I miss it" or "is gold overbought now." The data below argues that's the wrong question: the case for holding gold was never about calling this particular rally — it's a structural diversification effect that has held for three decades, rally or correction.
| Allocation | CAGR | Annualised volatility | Max drawdown | ₹10L (1996) grows to |
|---|---|---|---|---|
| 0% Gold (pure Nifty) | 11.53% | 21.99% | −55.1% | ₹2.81 crore |
| 10% Gold / 90% Nifty | 11.95% | 19.72% | −49.1% | ₹3.16 crore |
| 20% Gold / 80% Nifty | 12.27% | 17.67% | −43.4% | ₹3.44 crore |
The variable, in plain English. "X% gold allocation, rebalanced annually" means holding X% of the portfolio in gold and (100−X)% in the Nifty, and once a year selling whichever asset has outgrown its target weight to buy back into the other. This mechanical discipline — not a view on where gold is headed — is what captures the extra return in the table above: it forces the portfolio to trim Nifty after equity has run up and add gold when it's cheap relative to equity, and vice versa.
Historical precedent. Gold's cushioning shows up most clearly around the two big Indian equity crashes in this sample. During the Global Financial Crisis (Oct 2007–Mar 2009), the pure-Nifty portfolio fell −48.8%, the 10%-gold mix fell −41.1%, and the 20%-gold mix fell −33.1%. During the COVID crash (Jan–Apr 2020), the equivalent figures were −17.6%, −14.6% and −11.7%. More strikingly, recovery time from the GFC trough dropped from 59 months (pure Nifty, recovered only by Oct 2013) to 22 months (10% gold) to just 13 months (20% gold) — a five-year wait for pure equity investors versus roughly one year with a fifth of the portfolio in gold.
Transmission mechanism. Gold's price is driven mainly by two forces largely unrelated to Indian equity fundamentals: US real interest rates (proxied here by the US 5-year yield, correlation −0.27 with gold — gold tends to rise when yields fall, since it carries no coupon and looks relatively more attractive) and the strength of the dollar (correlation −0.24 with USD/INR's own moves). Neither of these is a function of Nifty earnings, RBI policy, or domestic flows — which is exactly why gold's correlation with the Nifty sits at a near-zero 0.13, the structural source of the diversification benefit.
Cross-asset read. The hedge isn't perfect, and it's worth being honest about the exceptions. Looking at the Nifty's 10 worst individual months since 1996, gold (in Rupee terms) was flat or positive in 7 of them — including +10.4% in Jan 2008 and +5.8% in Mar 2020. But in October 2008 (Nifty −26.4%) and March 2026 (Nifty −11.3%), gold fell too — both were acute global liquidity/margin-call-driven selloffs where investors sold everything, including gold, for cash. Gold is a partial, not absolute, hedge.
| Metric | 0% Gold | 10% Gold | 20% Gold |
|---|---|---|---|
| GFC return (Oct 2007–Mar 2009) | −48.8% | −41.1% | −33.1% |
| COVID return (Jan–Apr 2020) | −17.6% | −14.6% | −11.7% |
| Months to recover GFC trough | 59 | 22 | 13 |
Most Indian investors build positions through monthly SIPs, not a single 1996 lump sum, so the same 0%/10%/20% gold weights were re-tested as a ₹10,000 monthly SIP into Nifty and Gold, unrebalanced, from Jan 1996 to Sep 2026 (369 months, ₹36.9 lakh invested in every scenario). The direction of the result holds: both the return and the drawdown benefit persist for SIP investors, though the recovery-speed benefit shrinks, because ongoing monthly contributions already buy units cheaply during a crash and do a lot of the recovery work regardless of allocation.
| Allocation | XIRR | ₹36.9L invested grows to | Max corpus drawdown | Months to recover GFC peak |
|---|---|---|---|---|
| 0% Gold (pure Nifty SIP) | 11.76% | ₹3.17 crore | −54.0% | 22 |
| 10% Gold / 90% Nifty SIP | 12.02% | ₹3.35 crore | −49.9% | 22 |
| 20% Gold / 80% Nifty SIP | 12.27% | ₹3.52 crore | −45.4% | 19 |
The lump-sum backtest's most dramatic finding — GFC recovery cut from 59 months to 13 — doesn't carry over one-for-one to SIP investors (22 months down to 19), simply because a SIP investor's fresh contributions are already the biggest lever pulling the corpus back to its old peak. But the return (XIRR) and drawdown-depth improvements are essentially the same size as in the lump-sum case, so the core allocation case is unchanged either way you invest.
For an Indian equity-heavy portfolio, this backtest supports a structural 15–20% gold allocation, rebalanced annually rather than tactically timed — the return and drawdown benefits above came from the discipline of rebalancing at a fixed weight, not from calling gold's rallies or corrections. The current post-rally pullback (−16.8% off the Feb 2026 high) is, if anything, closer to a scheduled rebalancing opportunity than a reason to avoid the asset class. Sizing beyond 20% wasn't tested in this backtest and shouldn't be assumed to extrapolate the same free-lunch effect — Nifty is still the portfolio's primary growth engine.
Three things would weaken this case. First, if gold's correlation with the Nifty rises meaningfully above its historical 0.13 — as it briefly did in October 2008 and March 2026 during acute liquidity shocks — the diversification benefit shrinks exactly when it's needed most. Second, if US real interest rates move structurally and durably higher, gold's opportunity cost rises and its standalone return could lag for an extended stretch, as it did through parts of 2013–2018 in this dataset. Third, if the annual rebalancing discipline isn't actually followed — investors chasing gold after a rally or abandoning it after a correction — none of the historical numbers above apply, since they depend entirely on the mechanical rebalance.
Gold's job in this portfolio was never to be the best-performing asset in any given year — it was to shrink the multi-year recovery times that follow equity drawdowns, and on this 30-year record, a 15–20% allocation has done exactly that, without giving up return to get it.