Is every government bond the same?

1One promise, many dishes

On 3rd Aug 2021 a reader put this to TradingQnA:

1] Do all GSEC series i.e. 2035, 2050, 2025 [Old/New, High Interest/Low Interest ] etc carry same default risk? 2] Is the default risk/credit risk of SDLs higher? Is there a scenario where buying SDLs makes better sense than GSECs?

Sage90210, TradingQnA, 3rd Aug 2021

A G-sec is a bond issued by the central government in New Delhi. An SDL, a state development loan, is a bond issued by one state. He wanted to know whether these are all equally safe, and when the state one is the better buy. The one reply sent him elsewhere to read. This chapter answers both parts.

Chapter 5 ended on a rule: match the bond's end date to the date you need the money. That still leaves the question of which government bond to pick, because there are several. A restaurant menu is a good way to picture them. One kitchen, several dishes, each suited to a different appetite. The state loans are the exception: they come from about 30 kitchens, one for each state and for union territories such as Delhi that borrow on their own.

The menu was not designed in one go. It grew in 3 stages, and this chapter follows them in order.

1. Learning to ask. For about 40 years the government did not have to ask anyone for money. It could make the RBI lend, and it could make banks buy its bonds at a rate it chose. Sections 2 and 3 cover how that ended and what replaced it: the Treasury bill and the central government bond.
2. Letting more people lend. Once bonds were sold at auction, the doors opened wider: to ordinary investors (section 4) and to the states borrowing on their own (section 5).
3. Dishes for particular needs. Then came bonds built for one worry each: rates rising, never wanting to sell, the environment, not wanting small payments, and inflation (sections 6 to 10).

A government bond is not one thing. Every dish carries the same promise to pay; each comes with different terms.

2The Treasury bill: lending for a few months

Start with the oldest problem any government has: its spending and its tax money do not arrive on the same days. Some weeks it is short of cash.

From the mid 1950s India solved that the easy way. The government kept its bank account at the RBI, and the account had a minimum balance. Whenever it fell below the minimum, the RBI was obliged to top it up and take an IOU in return. That IOU was called an ad hoc Treasury bill. From 1974 it paid just 4.6% a year, far less than anyone else would have charged. The RBI could not say no, so the IOUs kept piling up for 4 decades. In 1997 the Governor and the Finance Secretary signed the arrangement away:

The short loans that replaced them had to be sold to anyone willing to lend, at whatever price those lenders would accept. That is an auction: lenders bid, and the price comes out of their bids. The government had already started doing this. It auctioned a 182 day bill first, and in April 1992 a 364 day bill. Lenders who were free to say no asked for over 11% on the 364 day bill, more than double the 4.6% the RBI had been forced to take.

Figure 1

That is the Treasury bill, or T-bill, as it still works today. It comes in 3 lengths: 91, 182 and 364 days. It never pays interest along the way. Instead you buy it for less than its face value, the ₹100 it promises to repay, and get the full ₹100 at the end. The difference is your interest. A ₹100 gift voucher that you buy for ₹98.63 and spend at full value 3 months later works the same way.

On 7th Oct 2026, for example, a retail investor paid ₹98.6326 for every ₹100 of a 91 day bill. On ₹1 lakh of face value: ₹98,632.60 out, ₹1,00,000 back 91 days later, ₹1,367.40 earned.

3 traps with T-billsNo interest is credited. The interest arrives as part of the ₹100 at the end. The yield on the screen is a yearly rate. The yield is your return expressed as a rate per year. That ₹1,367.40 over 3 months was shown as 5.56% a year; chapter 4 works the arithmetic. 2 bills can end on the same day, so a bill is named by its length and its full end date.

The T-bill is the shortest dish: bought below ₹100, repaid at ₹100 after 3, 6 or 12 months. It replaced an IOU the RBI could not refuse with a loan lenders choose to make.

That solved the government's short term cash problem. Its long term borrowing still worked the old way, by decree.

3The central government bond: lending for years

For long loans, the government had a second captive lender: the banks. A rule called the statutory liquidity ratio made every bank keep a share of its deposits in government bonds, and the government chose the interest rate those bonds paid. By 1992 the share was 38.5%. On 29th Feb 1992 the Finance Minister told Parliament that the rule "locks up large quantities of bank funds in relatively low yielding Government securities". He cut it to 30% on new deposits. On 3rd Jun 1992 the RBI held its first auction of a long term government bond. From then on, the rate came from lenders' bids.

The result is the central government bond, the G-sec from the reader's question. Its name carries its terms. 7.06% GS 2041 means:

• 7.06% is the coupon, the interest it pays each year on ₹100 of face value, split into 2 payments, 6 months apart.
• GS means government security, issued by the central government.
• 2041 is the year it repays the ₹100.

New bonds are auctioned most Fridays, for anything from 3 to 50 years. The 50 year bond arrived on 3rd Nov 2023. Lenders offered ₹40,200 crore for the ₹10,000 crore on sale, and the coupon was set at 7.46% by the auction itself.

Loans with no end yearThe Union Budget once listed a 3% Loan of 1896-97. It had no end date; the government could repay it whenever it chose, on 3 months' notice. A 3% Conversion Loan of 1946 was repayable at the government's option on or after 16th Sep 1986. Both had been repaid by the 1990s. Every bond issued today carries its end year in its name.

Now the first half of the reader's question. Do a 2035, a 2050 and a 2076 bond carry the same risk of not being paid? Yes. They are the same promise from the same borrower. What differs is how far the price can swing before the end (chapter 5 showed that a longer bond swings more) and how easily you can sell one (chapter 8). A local train and the Rajdhani run on the same tracks, owned by the same railway; the ride is not the same.

TrapA government bond has no lock in. It can be sold on any trading day. What you cannot choose is the price, or whether a buyer turns up.

Every central bond carries the same promise. Since 3rd Jun 1992 the rate on new ones has been set by lenders at auction, not chosen by the borrower.

Once bonds were sold at auction, the obvious next question was who could bid. At first it was only banks and other large institutions.

4How a retail investor gets in

In January 2002 the government opened a window at its bond auctions for retail investors, meaning individuals rather than institutions. The RBI advertised it: "Gilts offer safety, liquidity, attractive returns for longer duration." (Gilts is an old name for government bonds.) In 2002-03 retail investors asked for ₹1,302 crore of the ₹4,050 crore set aside for them, about 32%. On 12th Nov 2021 the RBI went a step further and opened RBI Retail Direct: a free account held with the RBI itself, from which anyone can bid at an auction on a phone.

To see how a retail investor fits into an auction, picture a cloth seller with 100 metres to sell and a crowd at the counter. Some buyers haggle and shout their offers: ₹105 a metre, ₹104, ₹102. The seller fills the highest offers first and works down until all 100 metres are gone. The last offer he accepts, say ₹101, is the lowest price anyone pays. A few customers stand in a quiet second queue. They don't haggle; each asks for a couple of metres at whatever the hagglers paid on average. As long as that queue stays small, every one of them is served.

A government bond auction runs the same way. Banks, insurers and other institutions are the hagglers: they bid prices. The RBI accepts the best prices first and works down until the loan is filled. The lowest price it accepts is the cut-off. Retail investors are the quiet queue. They name no price and are filled at the average of the accepted prices. This is non-competitive bidding.

A retail bid starts at ₹10,000, rises in steps of ₹10,000, and stops at ₹2 crore per bond per auction. 5% of every central bond auction is kept for these bids, and 10% of every state loan auction. There are 3 ways to place one: RBI Retail Direct, your broker's bid screen, and NSE's goBID website.

The reserved share mostly goes unused.

2 traps at the auctionThe headline price is not your price. Retail pays the average, a shade above the cut-off: ₹98.60 against ₹98.58 on 25th Sep 2026. A bid is not a person. One app can send a single bid for many customers.

From ₹10,000, a retail investor can buy a new government bond at the average auction price. Few do.

The central government was not the only borrower using these auctions. The states use them too, and their loans are the second half of the reader's question.

5The state loan: why it pays more

Each state runs its own budget, and when its spending runs ahead of its taxes, it borrows. It does that the same way the Centre does, through RBI auctions, with the RBI acting as its banker. The loan is a state development loan, or SDL. It looks just like a G-sec: a coupon twice a year and ₹100 back at the end. The difference is who owes you.

The reader asked whether that makes it riskier. April 2020 shows how lenders treat it. India was in lockdown, tax money had dried up, and every state needed to borrow at the same time. On 9th Apr 2020 the Centre paid 6.53% for about 9.5 years. On 13th Apr Uttar Pradesh paid 7.65% and Maharashtra 7.60% for 10 years, more than 1% above the Centre. Andhra Pradesh found the bids for its 6 and 7 year loans too expensive and turned them all down; the RBI's notice records that it "has not accepted any amount".

On 10th Mar 2026 Delhi borrowed in the market for the first time: ₹1,000 crore for 10 years. Lenders offered ₹5,600 crore, and Delhi paid 7.45%, the same as Madhya Pradesh in that auction. A brand new borrower got the same price as an established one.

Put together, those episodes say that the extra a state pays is driven mostly by supply. It jumps when many states borrow at once, and a newcomer pays the same as an old hand. The RBI's own measure, which compares loans of the same length, agrees. States paid 0.31% more than the Centre in 2023-24 and 0.30% more in 2024-25. In 2025-26, the year state borrowing jumped, they paid 0.50% more.

Why one state pays more than another

States also pay different rates from each other, for 2 reasons. Both show up in a single auction. On 10th Mar 2026, 17 states and union territories borrowed on the same day.

The first reason is length. A longer loan pays more, for states just as for the Centre. Tamil Nadu borrowed for 3 different lengths that day:

The second reason is the borrower. Line up loans of the same length, sold the same day, and the rates still differ:

For the same 15 years, Jammu and Kashmir paid 0.18% more than Andhra Pradesh. For the same 10 years, Sikkim paid 0.12% more than Tamil Nadu. That extra is what lenders charge for the state itself: how much it borrows, how often its loans change hands, and how its finances look to them. The RBI tracks the average of this gap on fresh 10 year loans. It was 0.03% in 2023-24, 0.04% in 2024-25 and 0.08% in 2025-26.

The gap stays small because every state loan sits on the same footing. A state borrows on the security of its own revenues, under Article 293 of the Constitution, and if it still owes the Centre money it needs the Centre's permission to borrow more. The RBI runs every state's borrowing, and banks treat every state loan as safe as a central bond when they measure their risk. That reading is ours; the RBI publishes the numbers, not the reasons. Bonds issued by a state company with only a state guarantee are a different thing, and are not on this menu.

Part of what states owe was once their power companies' debt. Under the UDAY scheme of November 2015, states took most of that debt onto their own books, where some of it had been costing 14 to 15%, and refinanced it more cheaply. In 2015-16 alone the RBI issued ₹98,959.97 crore of special bonds for 8 states.

So the answer to the second half of the reader's question: a state loan is a different borrower that pays a little more, mostly for being plentiful and hard to sell. It makes sense when its end date is your date and you will hold it to the end. On 6th Oct 2026 Punjab borrowed until 2041 at 8.02%; days later the Centre sold its own 2041 bond at 7.49%. On ₹1 lakh, that extra 0.53% is about ₹529 a year.

TrapThe extra pays for real costs. A state loan is hard to sell before it ends (chapter 8). Brokers can let you use bonds you own as security for trading, a process called pledging, but in October 2026 one large broker's list of accepted bonds had no state loans on it at all (chapter 10). One forced sale can cost more than the extra 0.5% earns.

A state loan is a different borrower paying a little more, mostly for being plentiful and hard to sell. It suits money that will stay put until the loan ends.

T-bills, central bonds and state loans are the basics of the menu. Everything after this was built to deal with one particular worry. The first worry: what if interest rates rise after you buy?

6The floating rate bond: a coupon that moves

Chapter 5 showed what rising rates do to a bond with a fixed coupon: its price falls, because new bonds pay more. A floating rate bond, or FRB, avoids that by changing its coupon every 6 months to follow rates in the market.

The change follows a published formula, and the fine print matters. On 24th Feb 1999 the RBI worked out the next coupon on one of these bonds. Its formula was the average yield on recent 364 day T-bills, plus 1.25%. T-bills were yielding 10.41%, so the formula gave 10.41% + 1.25% = 11.66%. But the bond also promised it would never pay less than 13%, a minimum called a floor. So holders received 13%.

Today's FRBs use the average yield of the last 3 auctions of 182 day T-bills, plus a fixed extra set when the bond was first sold. FRB 2033 shows the sum. The 182 day auctions of 2nd, 9th and 16th Sep 2026 averaged 5.6189%, 5.6147% and 5.7167%; together, 5.6501%. Add the bond's fixed 1.22% and the coupon is 6.87%, the figure the RBI announced for the next 6 months.

A rent agreement that resets every 6 months to the going rate, plus a fixed top up, works the same way. It rises when rents in the city rise and falls when they fall. Unlike the 1999 bond, today's FRBs have no floor.

2 traps with floatersA steadier price is not a steady income. FRB 2033 has paid as little as 4.62% and as much as 8.51%. A new one is not sold every year. None was auctioned in 2026-27, so a retail investor could only buy one second hand, from another investor on the exchange.

A floating rate bond trades a fixed coupon for a formula. The income follows short term rates, and the price stays closer to ₹100.

The FRB still has a price that moves a little, because it trades. The next dish has no price at all.

7The RBI savings bond: the one you cannot sell

Many families hold an "RBI bond" bought at a bank branch. It is built for people who never want to sell: no market, no price on a screen, just interest until the end.

The 8% Savings (Taxable) Bonds went on sale on 21st Apr 2003. Anyone could buy them at a bank branch, from ₹1,000, with no upper limit. They could not be sold to anyone else. They ran for 6 years and paid interest either every 6 months or all at the end. They stayed on sale for almost 15 years, until this:

A 7.75% bond took its place on 10th Jan 2018. On 27th May 2020 that one was closed from the next day too. On 1st Jul 2020 the Floating Rate Savings Bond opened in its place. Each time, people who already held the old bonds kept their rate until the end; only new money was turned away.

The Floating Rate Savings Bond is the one on sale now. Its rate is tied to the NSC, the National Savings Certificate sold at post offices: it pays the NSC rate plus 0.35%, reset every 1st Jan and 1st Jul. It runs 7 years. It cannot be sold or used as security for a loan. It is held in an account with the RBI or a bank, not in a demat account, the electronic account where shares and most bonds are held. Investors aged 60 and above may leave early, after 4 to 6 years depending on age, giving up 50% of one interest payment. The rate was 7.15% in 2020 and 8.05% from July 2023. On ₹1 lakh at 8.05%, that is ₹4,025 every 6 months. In 2025-26 it raised ₹12,476 crore, of which ₹1,067 crore came through RBI Retail Direct.

The closest everyday match is a lease at a rent tied to the post office rate, which you cannot sell, sublet or break. With no market, its price never moves, which is how chapter 2 could say it has none of a bond's price swings.

Traps8.05% is a reset, not a promise. Below 60 there is no early exit. It cannot be pledged. Tax is deducted from the interest before it reaches you (chapter 9).

The RBI savings bond floats but cannot be sold. It suits money that will not be needed for 7 years, and the window to buy it can close at a day's notice.

The next dish is built for a different kind of worry: not about your money, but about what it pays for.

8The green bond

A sovereign green bond is an ordinary central government bond whose money is set aside for projects such as solar power. The first were sold on 25th Jan 2023: ₹4,000 crore each of 5 and 10 year bonds. Lenders offered ₹32,892 crore, about 4 times the amount, and the bonds sold at 7.10% and 7.29%. 2 weeks later more of the 5 year bond sold at ₹99.46, below its ₹100 face value. The green label did not stop the price from moving.

The government hoped some lenders would accept a slightly lower rate for the label, so green borrowing would cost it a little less. That hope was tested on 31st May 2024, when lenders offered ₹12,677 crore for ₹6,000 crore of a 10 year green bond, and again on 13th Jun 2025, when they offered ₹10,943.5 crore for ₹5,000 crore of a 30 year one. Both times the government sold nothing.

The RBI gave no reason. The likely one is that lenders would not accept the lower rate, and the government would not pay the normal one. Andhra Pradesh made the same call in April 2020: when the bids are too expensive, a borrower can walk away.

When green bonds do sell, the discount is small. The RBI estimated it at 0.0426% for the first 5 year bond and 0.0353% for the 10 year. On 1st Oct 2026 a 2056 green bond sold at 7.65% and a normal 2056 bond at 7.69%. That is about ₹36 a year less on ₹1 lakh. The comparison is ours, and the 2 bonds end a few months apart. It is the same dal served on a leaf plate, a few paise cheaper because some diners prefer the plate.

2 traps with green bondsGreen is not tax free. The interest is taxed like any other bond's. Green is not safer. Payments do not depend on the projects; it is the government's ordinary promise. Many recent green bonds also run 30 years, so their prices swing hard (chapter 5).

A green bond is the ordinary government promise with a label, paying a fraction less. Twice the government refused to sell one at the price lenders asked.

Every bond so far pays something along the way, or, like the T-bill, ends within a year. The next dish is for someone who wants nothing until the end.

9STRIPS: a bond that pays nothing till the end

The day before his question, the same reader asked which government bonds were "Zero Coupon": bonds that pay no interest along the way, just one amount at the end. The appeal is that there are no small payments to reinvest.

India's government issued a zero coupon bond only once, in 1996. From 1st Apr 2010 the RBI allowed ordinary bonds to be cut into pieces instead. Each coupon becomes its own small bond with one payment date, and so does the final ₹100. The pieces are called STRIPS, and each has a face value of ₹100. The RBI wrote that, with nothing to reinvest, "they can be attractive to retail/non-institutional investors".

Picture a booklet of coupons torn into single pages, each cashed on its own date. On 8th Oct 2026 there were 1,618 STRIPS outstanding, worth ₹8,99,667 crore, all cut from existing central bonds. State loans have been allowed to be stripped since 12th Jun 2025.

There are 2 different minimums, and they are easy to mix up. Cutting a bond up is a wholesale job: it has to be done in blocks of at least ₹1 crore of face value, by large holders. Buying a piece that already exists is much smaller. Since 16th Jun 2021 NSE has let STRIPS trade like shares in its capital market segment, in lots of 100 units, which is ₹10,000 of face value. The catch is that almost nobody trades them. On 4 trading days checked in late September and early October 2026, NSE's daily list of trades did not show a single STRIPS.

TrapA STRIPS on the exchange can be bought from ₹10,000 of face value, but only if someone is selling, and selling it later needs a buyer too. Chapter 8 covers what an empty order book does to your price.

STRIPS turn one bond into many zero coupon pieces. Creating them takes ₹1 crore; buying one on NSE takes ₹10,000 of face value, if a seller turns up.

The last worry is the oldest one for anyone living on fixed payments: prices rising faster than the payments.

10The inflation indexed bond

A normal bond promises a fixed number of rupees. If prices rise, those rupees buy less. An inflation indexed bond raises its payments when prices rise, the way dearness allowance raises a salary.

In 2013 prices were rising fast, and families were buying gold to protect their savings. On 15th May 2013 the RBI announced inflation indexed bonds. Their purpose, in its words, was to "protect savings of poor and middle classes from inflation and incentivise household sector to save in financial instruments rather than buy gold". The first was sold on 4th Jun 2013. A retail version followed in December 2013, sold at banks from ₹5,000.

The retail version raised ₹92 crore before it closed on 31st Mar 2014. The other matured on 5th Jun 2023, and none has been outstanding since. An earlier attempt, the Capital Indexed Bond, protected only the ₹100 repaid at the end, not the interest; it was issued in December 1997 and matured in 2002. Commentators blamed the weak response on the price index used, the yearly tax on the top up and the long lock in. The government has given no reason.

India has sold bonds that keep pace with prices, and few people bought them. None is on sale, though the government keeps the right to issue more.

11The menu card

Gold is the last item. In November 2015 the government launched Sovereign Gold Bonds, which pay interest and track the price of gold. No new issue has been sold since February 2024, so they can only be bought second hand. They have their own chapter: Sovereign Gold Bonds after the scheme closed.

The whole menu, one row per dish. Each row is a dated snapshot; the rules can change.

And the menu has changed often. Some dishes have stayed for decades; some came off overnight.

3 months after the first question, another reader left this on Varsity's chapter on government securities:

How to decide whoch Bond is suitable for my investment goals, between capital indexed, inflation indexed, savings bond etc. Where to find detailed read on Varsity for these different types of Gsec bonds

Paras Gupta, Varsity, 8th Nov 2021

To Sage90210: every central bond carries the same promise, whatever its year. A state loan is a different borrower that pays a little more, mostly for being plentiful and hard to sell. It makes sense when its end date is yours.

To Paras: choose by 2 things, when the money has to come back and what it has to do in the meantime. A few months: a T-bill. Fixed income for years: a central bond. Income that follows rates: an FRB, or the savings bond if you will never need to sell. And 2 of the 3 dishes he named, the capital indexed and inflation indexed bonds, are off the menu.

Every government bond is the same promise. Pick the dish by when you need the money and what it must do until then.

Where to find today's figures

Every figure in this chapter is dated. This is where to find the current one.