Gold in four forms

Four wrappers, one metal

Jewellery, coins and bars, a gold exchange-traded fund and a gold-linked government bond all track the same metal. They differ on five things: what you pay to get in, what holding costs, what you give up getting out, how quickly you can sell, and whether the holding pays you anything while you own it.

Only the last of those is obvious from the outside. The others are quiet, and two of them — the charge paid at purchase and the deduction taken at sale — are settled before the gold price has done anything at all. That is the whole subject of this article. Not which form is better, which is a question about what you want the holding to do, but what each one costs and when the cost lands.

One definition first, because the term does the work later. A scheme that holds the metal, keeps it with a custodian and issues units you can buy and sell on the exchange like a share is a gold exchange-traded fund, or ETF. A variant — a fund that holds units of that ETF, so you can buy it without an exchange account, at the day's closing value rather than at a live price — is a gold fund of funds, and it adds a second layer of charge on top of the first.

Nothing here says what gold will do, or what any of these forms will return. Why the metal moves at all, and why an Indian holder is carrying a rupee position they did not choose, is the subject of gold and real rates. This article assumes the metal does whatever it does and asks what each wrapper skims off on the way in, on the way out, and every year in between.

What a making charge actually buys

Start with the form most Indian households already own. A piece of jewellery is bought as an object: someone designed it, someone made it, metal was lost in the making, and a shop financed the stock while it sat in a display case. The making charge is what pays for all of that. It is quoted as a percentage of the metal value, added to the bill, and it is entirely real work being paid for.

It is also not gold. And that distinction is the one that decides the arithmetic, because the resale market pays for metal content and nothing else. Melt the piece and the design is gone; the buyer is bidding for grams.

Here is the round trip, on illustrative round numbers chosen so the multiplication is easy to redo. They are not observed charges, and no typical figure appears anywhere in this article.

Suppose the metal in a piece is worth ₹100 and the making charge is 12%. You pay ₹112, before any tax on the purchase, which raises the bill further. A buy-back that pays for metal content alone gives you ₹100 for it today. The gold price has to rise 12% before you are level — not before you profit, before you are back where you started.

Now add the second deduction, which is the one people do not expect. Many buy-backs also subtract something for melting and refining loss. On an illustrative 3% deduction, your ₹100 of metal returns ₹97, so the price now has to reach ₹115.50 for the round trip to break even — a rise of about 15.5%. Two charges, one on each end, neither of them visible in the gold price.

Purity is the third wedge, and it is usually a misunderstanding rather than a cost. Pure gold is 24 karat. Indian jewellery is commonly 22 karat, which means 22 parts in 24 by weight, or 91.6% gold, with the balance alloyed in for hardness. So a gram of 22 karat jewellery contains 0.916 grams of gold, and it is quoted and bought back at the 22 karat rate.

Nothing is lost in that conversion by itself. What is lost is the assumption underneath it: the rate you saw quoted for the day may well be the 24 karat rate, and it is not what your piece is priced against. A hallmark certifies purity; it is not a price guarantee, and it does not make a piece any more saleable at the rate on the board.

The specific mistake, then, is a small one with a large consequence: reading the day's gold rate as the price you would receive. It is the price of pure metal in a wholesale quote, before the making charge you already paid, before any refining deduction, before purity, and before tax. Four wedges sit between that number and your bank account, and three of them were agreed on the day you bought.

The five things that actually differ

Set the four forms against the five questions from the opening. The table states the shape of each cost, not its size — sizes move, and no figure here is quoted because no verified figure exists for it.

Two rows are worth pausing on. The cost to get out row is the only place where jewellery and coins genuinely part company: a coin or bar has no design to pay for, so the premium at purchase is smaller and the buy-back discount is narrower. The physical burden is otherwise identical — the same locker, the same insurance question, the same need to convince a buyer of purity.

The does it pay you row is the structural divide. Three of the four forms produce no income at all, which means their entire return is a price change, and the safe interest you declined by holding them is a cost you are paying whether or not you ever notice it. Only the bond breaks that pattern, and it breaks it by adding an issuer and a lock-in. No row in the table is free; every difference is about which cost you pay and when it lands.

Storage is a cost in every form; only some send you a bill

Metal has to sit somewhere. At home it is a security problem you are managing yourself; in a bank locker it is a rent, plus the insurance question the locker does not answer. Either way it is a recurring cost of owning the thing, and it does not shrink as the holding grows — a bigger holding needs a bigger locker.

A gold ETF has exactly the same cost. The scheme pays a custodian to hold and insure the metal, pays a fund house to run the vehicle, and recovers all of it through an annual charge deducted from the unit value a day at a time. You never write a cheque for it, which is precisely why it is easy to forget: the grams of metal standing behind each unit fall slowly and continuously, and nothing on your statement announces it. That charge is the storage cost, converted into a percentage and made automatic. The mechanics of how such a charge compounds are the subject of the expense ratio, and they apply here unchanged.

The gold fund of funds pays that charge and then adds its own, because it is a fund holding a fund. Whether the convenience of buying without an exchange account is worth a second layer is a question with a real answer for each buyer, and it should be answered knowing that a second layer is what is being paid for.

Then there is the cost that no form escapes, because it is not charged by anybody. Gold produces no income, so holding it means declining whatever a safe rupee asset would have paid. To give that a size without guessing at a market rate, take an administered one: the Public Provident Fund pays 7.1%, a figure the government sets and revises quarterly rather than one a bank quotes. It has its own lock-in and its own contribution ceiling, so it illustrates the scale of the forgone yield rather than naming an account anyone would switch into. Deposit rates are market outcomes, they move, and none appears here.

Decline a rate of that order for a decade and the forgone amount compounds exactly the way a return does. Over any holding long enough to be called an investment it is the largest cost in this comparison, and the only one nobody invoices you for.

One correction to that comparison before it is used, because the naive version is wrong. A deposit rate is a nominal rate, and gold is a claim on purchasing power rather than on rupees, so the honest comparison strips inflation out of the interest first. The full mechanism — why the real rate rather than the nominal one is gold's true carrying cost, and why the Indian real rate and the American one are different numbers — is in gold and real rates. It is the reason a bond that pays a coupon on a gold-linked holding is a structurally different proposition from metal: part of the forgone yield is handed back.

Turn a one-time charge into an annual one, and the forms become comparable

A making charge and an expense ratio look like different kinds of thing. One is a lump paid at the counter; the other is a fraction skimmed off every year. They cannot be compared as quoted, which is why they usually are not compared at all — and the reader is left with a vague sense that jewellery is expensive and a fund is cheap, without knowing by how much or under what conditions.

They become comparable with one step. Ask what annual rate of extra gold price rise would be needed to cover the one-time charge over the period you actually hold. On the illustrative 12% charge from earlier, that is the rate r for which (1 + r)n = 1.12. Every figure below is that arithmetic and nothing else.

The crossover sits at about eleven years. Before it, the one-time charge is the dearer arrangement; past it, the annual charge is, because an annual charge never stops and a one-time charge is spread over a longer and longer holding. The horizon decides the ranking, and the ranking genuinely reverses — which is why a flat claim that one form is cheaper than another, made without a holding period attached, is not a claim about anything.

Two corrections push the crossover out, and honesty requires both. Include the illustrative 3% refining deduction on exit and the one-time hurdle becomes about 15.5%, moving the crossover past fourteen years. And the annual charge is not the only cost on the fund side either: brokerage and the bid-offer spread are paid on both the buy and the sell, and while they are small they are not zero. Neither correction changes the shape of the argument, and both change the number.

Now the part that matters more than the table. An annualised cost is a summary, and summaries discard. This one discards four things at once, and each of them is nameable.

• It assumes you hold for the whole period and then sell. Sell in year three and you paid the year-three cost, not the twenty-five-year one, and no averaging undoes that.
• It discards the certainty gap. The making charge is spent on day one whatever happens next; the annual charge is only incurred for as long as you stay.
• It discards the sequence, since a charge paid up front is money that was never in the holding to compound at all.
• It discards the exit deduction, unless you deliberately add it back as above.

A single cost-per-year figure is a useful comparison and a poor description, and both of those are worth holding at once. That is not special to this table. Every summary statistic in investing throws something away, and the loss is usually specific enough to name: a standard deviation treats a 10% rise and a 10% fall as the same quantity of risk, and a ratio computed on overlapping windows counts nearly the same stretch of history many times over, which makes it understate its own error. The machinery is the subject of risk measures. The habit — ask what the one number threw away before using it — is the one this table needs too.

Jewellery is consumption with an investment story attached

The resale market pays for the metal in a piece of jewellery and not for the piece. So a necklace is an object you own that happens to contain gold, and the two halves of that purchase have to be counted separately — which is not a criticism of buying one.

A necklace at a wedding does a job. It marks an occasion, it is given and received, it carries family history, and it is worn. No fund unit does any of that, and no amount of cost arithmetic makes the object interchangeable with the exposure. The making charge is the price of that object — the design, the craft, the wearability — and it buys exactly what it says it buys.

What follows is only this. You made a single transaction and paid separately for an object and for a quantity of metal, and when you come to sell, the object's price is gone. That is not a hidden fee or a swindle. It is what happens to the price of any manufactured item at resale, and gold jewellery is unusual only in that a commodity sits inside it, which makes it feel like something that should hold its value in full.

The specific mistake is a bookkeeping one. It is entering the jewellery box on the family's net-worth statement at what was paid for it, or at the full retail value of the metal at today's rate. Neither figure is the realisable amount, and the two are wrong by different quantities.

Cost overstates the line by the making charge — on the illustrative 12%, ₹112 written down against ₹100 realisable, so about a ninth of the number is not there. Today's metal rate overstates it by the exit deduction instead, an illustrative 3%. Either way the correction arrives on the day the money is needed, which is generally the worst day for it.

There is a second, quieter consequence, and it is the one the arithmetic cannot reach by itself. The annualisation table above assumes a sale. Its whole comfort — that a 12% charge is only 0.5% a year over twenty-five years — belongs exclusively to someone who actually sells at the end of the twenty-five years. A piece worn until it is inherited, or exchanged for a new design that restarts the making charge on the replacement, was never sold. A cost spread over a sale that never happens has not been spread at all; it has simply been paid.

Whether that matters depends entirely on what the holding was for, and a purchase made for an occasion or out of custom never had to answer the question. It becomes pressing only when the same pieces are also counted as the household's gold holding, which is where the two functions get quietly conflated. Where a holding sits and what job it has is the subject of asset allocation; treating an object as a position is one of the common money mistakes precisely because both halves feel true.

What the paper forms give up

Symmetry demands the reverse case, because the two paper wrappers have costs of their own and one of them is a genuine constraint rather than a charge.

A gold ETF needs an exchange account — a demat and trading account — which not every household has or wants. And its price is a market price. The scheme's obligation to hold the metal, to value it by a stated method and to publish that value is regulation. What the units actually change hands for at 14:20 IST on a quiet Tuesday is not: it is supply and demand, and it can sit a little above or a little below the value of the metal behind each unit.

The mechanism that keeps those two close is worth knowing, because it explains when it stops working. Large intermediaries can exchange baskets of units against metal with the fund, which means a wide enough gap between the traded price and the underlying value is an arbitrage someone is paid to close. The gap widens when that arbitrage is slow — a thin trading day, an unusual hour, a scramble — and it is widest exactly when a seller is in a hurry. The bid-offer spread is the visible half of this and the drift from underlying value is the invisible half. Neither is a defect; both are the price of an exit that a jeweller's counter does not offer at all.

The sovereign bond trades the other way: it removes the annual charge and pays a coupon, and it charges for that with time. Its scheme terms belong to the article on sovereign gold bonds, which sets out what the coupon is actually paid on. What belongs here is only the cost comparison — and for that, the regulation and the practice must be kept strictly apart, because they are routinely presented as one thing.

• Regulated — the tenor, the dates on which early redemption is permitted, the formula and reference price used to value the bond at redemption, the coupon and the base it accrues on, and the fact that the obligation sits with the sovereign rather than with a jeweller or a fund house. Every one of those is a term of the specific issue, printed in its notification. This article states none of them, because they are scheme figures and no verified figure is available to it.
• Not regulated — what you get for the bond on the exchange before the redemption window opens. Secondary trading in these is often thin, and a thin market prices by negotiation, which is why the traded price can sit below what the redemption formula would give. That discount is a market outcome, not a rule, and nothing obliges it to be there or to go away.

The consequence is a lock-in with a soft edge. Until the redemption window opens, your exit is the exchange at whatever it happens to pay; after it opens, the formula applies. Two different prices for the same holding, decided by the calendar. That is a real cost and it is paid in flexibility rather than in rupees, which makes it easy to leave out of a cost comparison entirely.

One more functional difference, and it runs the other way. Physical gold has a deep lending market: household metal can be pledged, and the RBI caps how much can be lent against it at 85% up to ₹2.5 lakh, 80% above that and up to ₹5 lakh, and 75% above ₹5 lakh. That is a genuine property of the physical form — money without a sale, and without paying a making charge away a second time — and it is covered in full in the gold loan article. Whether the paper forms can be pledged at all, and on what terms, depends on the lender and is not governed by those same ceilings.

Tax arrives on exit, and the wrapper decides which rule

Three of the four forms create no tax event at all while you hold them, for the simple reason that they produce nothing to tax. Gold pays no coupon, no dividend and no rent, so there is no annual accrual, and the whole gain sits unrealised until you sell. The only form that taxes you before you sell is the one that pays you — the bond's coupon is income when it arrives, and it is taxed in the year it arrives whether or not you wanted the cash.

That is not a defect of the bond and it is not a virtue of the metal. It is the same deferral mechanism that separates a fixed deposit from a fund: money that would have gone to tax each year stays invested and compounds until redemption instead. Over a year the difference is trivial. Over fifteen, deferral is doing real work — and it works in the same direction for jewellery, coins and an ETF alike.

Which rule applies on exit is where the wrapper matters, and the Indian classification machinery has a trap in it. The Income-tax Act sorts fund units by what the fund holds, not by what it is called. A fund is equity-oriented if more than 65% of its proceeds sit in domestic equity shares. Separately, a fund is a Specified Mutual Fund if more than 65% sits in debt and money-market instruments, and units of one bought on or after 1 April 2023 are taxed at slab rates, with the gain always treated as short-term.

Read those two tests together and the answer for gold falls out. A scheme holding metal satisfies neither — gold is not a domestic equity share and it is not a debt or money-market instrument. It sits in neither named bucket, and so takes ordinary capital-gains treatment rather than either special regime.

The common error is assuming that because a gold fund is obviously not equity, it must therefore be taxed like a debt fund. It is not, and the sentence in the statute that would make it so is about what the fund holds, which in this case is metal. The rates and holding periods that follow from ordinary treatment are deliberately not stated here — they change, and they belong in one place: capital gains tax.

Physical gold has a practical problem the paper forms do not, and it surfaces only at the exit. A gain is computed against a cost of acquisition, and a cost of acquisition has to be evidenced. Jewellery bought twenty years ago, or inherited, frequently comes with no invoice at all — which turns a tax computation into a documentation exercise at the moment of sale. An ETF holding has a transaction record by construction. That difference is worth nothing for years and then matters entirely on one day, which is the pattern planning exists to catch.

Where to look at this

The costs above have to be read off the documents in front of you: the making charge and buy-back terms from the jeweller, the annual charge from a scheme's factsheet, the coupon and redemption terms from a bond's issue notification. No tool substitutes for that.

What can be computed rather than argued about is what a scheme actually did. FNOTrader's Mutual Funds app runs on the full AMFI NAV history — around 34 million rows, updated at 22:30 IST each night — and simulates lumpsum and SIP on any scheme and period, reporting XIRR, invested against value, and maximum drawdown, alongside rolling-return distributions. A gold scheme's NAV series is already net of its annual charge, which means the drag described above is inside the numbers rather than sitting beside them.

What it cannot show you is the making charge on a piece in a locker, because that transaction never touched a scheme. That part stays arithmetic on a receipt.

Common questions

Continue reading

More in Advanced Investing · App: Mutual Funds · Definitions: glossary · Free tools: calculators · All: every article