Open any Brazilian video about Tesouro Direto (Brazil’s government bond platform) and time how long it takes for someone to say “passive income.” It rarely gets past ninety seconds. The pitch is always ready: the bond with semiannual coupons pays you every six months, the money lands in your account, you watch the return happening. You don’t have to wait until 2035 to earn anything. It’s concrete. It’s, according to almost everyone, an advantage.
And the question that comes out of it is always the same: which bond pays the highest coupon? The NTN-F (Brazil’s fixed-rate treasury bond with semiannual coupons) pays 10% a year. The NTN-B (its inflation-linked cousin) pays 6% a year on top of IPCA (Brazil’s official inflation index). Bigger numbers than their no-coupon siblings, which pay nothing until maturity. The conclusion seems to write itself.
The received wisdom
It’s worth saying what the pro-coupon case gets right, because it isn’t nonsense.
A bond with semiannual coupons returns capital along the way. That reduces duration — the price’s sensitivity to changes in interest rates. If rates spike and you need to sell before maturity, the coupon bond falls less than the equivalent zero-coupon. Less mark-to-market volatility, fewer nasty surprises on the statement.
It also generates cash without forcing a sale. For anyone living off their portfolio, that’s no small thing: it means not having to liquidate a position at the worst possible moment — real protection against sequence-of-returns risk.
And there’s flexibility. A coupon in your account is money with no obligatory destination. If a better opportunity shows up, you buy it. The zero-coupon locks you in.
Three legitimate arguments. The problem isn’t any of them. The problem is what gets left out.
The coupon isn’t a rate. It’s a price.
When you buy an NTN-F, the 10%-a-year coupon isn’t your return. It’s a fixed feature of the bond, set by the Treasury at issuance, identical for everyone who buys that bond — today, next year, or in 2031. Your return is the contracted rate, the yield you lock in and see on the Tesouro Direto screen at the moment of purchase, and it can be 11%, 13%, or 15% depending on the day.
How do two different things coexist? Through the price.
The bond has a face value of R$ 1,000 and pays R$ 48.81 per semester — always. If the market wants 13% a year and the bond only pays 10%, no one buys it at R$ 1,000. The price drops until those R$ 48.81 semiannual payments plus the R$ 1,000 at maturity deliver exactly 13% a year. Do the math, and the price is R$ 840.98. That’s the discount to par. If the market wanted 8%, the same bond would be worth more than R$ 1,000 — a premium.
The high coupon doesn’t hand you a bigger return. It makes you pay more for the same return.
The coupon isn’t how much you earn. It’s when you get paid. And “when” has a price.
(A bit of trivia that changes nothing in the argument: “coupon” is literal. Old bonds came as paper with detachable rectangles along the edges — every semester the holder clipped one with scissors and took it to the bank. “Clipping coupons” became slang for living off investments precisely because of this. The rectangles disappeared; the word stayed.)
You didn’t trade risk for safety. You traded one risk for another.
The coupon bond has a shorter duration. That’s true and it’s always said. What’s almost never said is the other side.
That 13% contracted rate that appeared on the screen? You only actually earn it if you reinvest every coupon at 13% a year until maturity. It’s baked into the bond’s math — an assumption, not a promise. If the yield curve softens and you can only reinvest at 9%, your realized return isn’t 13%. It’s less. The bond delivered what it promised; the market just wouldn’t let you put the money back in the same place. That’s reinvestment risk, and it shows up nowhere on the purchase screen.
Ten years, R$ 10,000, a hypothetical 13%-a-year contracted rate on both bonds. The only difference is the coupon. (“Selic” below is shorthand: what really drives the reinvestment rate on a ten-year bond is the long end of the curve, which moves with the Selic rate — Brazil’s benchmark interest rate — without being the same thing.)
| Reinvestment scenario | Zero-coupon (gross) | Coupon bond (gross) |
|---|---|---|
| Selic steady at 13% | R$ 33,946 — 13.00% p.a. | R$ 33,946 — 13.00% p.a. |
| Selic falls to 9% | R$ 33,946 — 13.00% p.a. | R$ 29,915 — 11.58% p.a. |
| Selic rises to 17% | R$ 33,946 — 13.00% p.a. | R$ 38,946 — 14.56% p.a. |
Look at the first row. Reinvesting at exactly the contracted rate, both bonds deliver the same cent — R$ 33,946. That’s not coincidence or rounding: it’s the definition of the contracted rate. The coupon was never additional return. It was always a schedule.
Now the other two. The zero-coupon column doesn’t budge — 13.00% in all three scenarios, because there’s nothing to reinvest. The coupon column swings almost three percentage points. On the way down, the coupon hijacks 1.42 points a year: you contracted 13% and took home 11.58%. On the way up, it gains 1.56.
So the coupon bond isn’t safer. It’s a directional position on interest rates. You’re long a rate rise — and it makes sense to build that position if that’s your read. What doesn’t make sense is building it while thinking you’re reducing risk, because whoever sold you the idea called the coupon passive income and not a bet.
The coupon you spent
Everything above assumes you reinvest every coupon, religiously, for twenty straight semesters. It’s worth asking how often that actually happens.
The money lands in the checking account alongside your salary, no label, no ceremony, and becomes dinner out. Nobody decides to stop investing — the decision never gets made. It simply doesn’t happen, six months at a time, and the compounding that was the whole point of the plan leaks out a drain the spreadsheet doesn’t show. I don’t have a number for this, and I distrust anyone who does — Tesouro Direto doesn’t publish coupon-reinvestment statistics. What can be said is that the product’s design pushes in one direction only. It’s the kind of mistake Morgan Housel describes: behaviour beats intelligence, and the coupon is a behavioural trap dressed up as a technical advantage.
The zero-coupon doesn’t give you that chance. There’s nothing to spend until 2035.
The tax nobody puts in the spreadsheet
Go back to the first row of that table — the steady-Selic one, where both bonds deliver an identical R$ 33,946. Same credit risk, same issuer, same maturity.
After tax, they aren’t equal. And the gap isn’t small.
Income tax on Tesouro Direto follows the regressive table, counted from the purchase date. On a coupon bond, each payment is taxed on arrival, at the rate in force at that moment. The NTN-F pays on 1 January and 1 July; assuming you buy on a coupon date:
| Coupon received at | Days since purchase | Tax rate |
|---|---|---|
| 6 months | 181 | 20% |
| 12 and 18 months | 365 and 546 | 17.5% |
| 24 months onward | 730 and beyond | 15% |
The first coupon escapes the priciest bracket by a single day: a calendar semester is 181 days, and the 22.5% rate applies up to 180. Buy mid-semester and you don’t escape — the schedule depends on your purchase date.
Two problems live here, and the second is much worse than the first.
The first is obvious: the early coupons draw a short-term rate. The ten-year bond you bought to hold to the end pays 20% on the first coupon, as if it were a few-month deposit. The bond is long; the coupon, as far as the taxman is concerned, is not.
The second is the one that actually hurts. Every coupon arrives net, and only the net amount goes back to compounding. You didn’t lose 20% of R$ 580 once — you lost the nine and a half years of compounding those R$ 116 would have generated. And whatever’s left, as it earns, gets taxed again. The zero-coupon does none of this: the full amount compounds untouched for a decade and meets the taxman a single time, at 15%, at maturity. It isn’t an exemption. It’s deferral — and deferral, over ten years, is real money.
Same contracted rate, same gross return, Selic flat:
| Zero-coupon | Coupon bond | |
|---|---|---|
| Gross over 10 years | R$ 33,946 | R$ 33,946 |
| Net over 10 years | R$ 30,354 | R$ 28,862 |
| Net return | 11.74% p.a. | 11.18% p.a. |
R$ 1,491. About 4.9% of the final balance, evaporated in a scenario where both bonds earned exactly the same thing before tax.
And this cost doesn’t depend on the brackets. Even if every coupon were taxed at the 15% floor — the most generous case there is for the coupon bond — the bill would still be R$ 1,332, half a percentage point a year. The regressive table makes the problem worse. It isn’t what creates it.
And this is where the whole argument turns. Reinvestment risk is a bet: if rates rise, you win, and the table shows it. The tax isn’t a bet. You pay it on the way up, on the way down, and when nothing moves. You pay it even in the scenario where your read on rates was right — the bill simply comes out of the profit. Under current rules, there’s no scenario where the coupon is tax-efficient during accumulation; there’s only the one where the gain on the rate was big enough to pay the toll and still leave something over. The one plausible twist would be a future hike in income tax on fixed income, which would reward whoever paid early — a bet as directional as the other one.
Nobody puts that in the passive-income video.
Where this argument doesn’t apply
None of this makes the coupon bond a bad product. It makes it a product with a specific customer, who probably isn’t you yet.
If you live off your portfolio, the logic flips completely. The coupon stops being money that escapes compounding and becomes money you use — and the tax cost turns into the price of not having to sell an asset in a bad market. That’s genuine protection against sequence-of-returns risk. If you hold a firm view that rates will rise, the coupon is the right instrument to express it; just call the position by its name. And if you might be forced to sell before maturity, the shorter duration has real value, because a ten-year zero-coupon takes a beating when rates jump.
What defines the right side of the line isn’t your risk profile. It’s a two-part question: do you need this money in the next ten years, and will you hold to maturity? Whoever’s accumulating answers “no” and “yes” — and for that person the coupon is pure cost, charged in three currencies: tax paid early, reinvestment risk, and the semiannual temptation to spend.
The border between accumulating and living off income, by the way, is a lot blurrier than the two nouns suggest. Almost nobody wakes up one day and switches sides.
What this actually means
The question “which bond pays the highest coupon?” doesn’t have a wrong answer. It’s the wrong question — it measures the schedule and calls the result a return. The coupon decides when the money arrives, who taxes it along the way, and which risk you carry until 2035. None of those three things shows up in the comparison most people make.
If you’re still building wealth and don’t plan to sell before maturity, the default choice is the zero-coupon — Tesouro Prefixado or Tesouro IPCA+ Principal (Brazil’s fixed-rate and inflation-linked Treasury bonds, in their no-coupon versions). Not because the rate is better, but because it’s the same and the cost is lower. This speaks directly to the logic of Brazilian real interest rates: in a country that pays high real yields, the value lies in letting compounding work without interruption — and every coupon is an interruption, with the taxman standing at the door.
It’s worth modelling instead of believing. Run both paths through the Retirement Calculator with your real horizon and see what a 0.56-point-a-year difference does over twenty years — then ask yourself in which year, exactly, you stop accumulating and start needing the cash flow. That date, not the size of the coupon, is what decides which bond you should be buying.
This article is for general educational purposes and does not constitute investment advice. The figures are illustrative and assume a hypothetical contracted rate of 13% a year, purchase on a coupon date, holding to maturity, and full reinvestment of coupons — assumptions that rarely hold exactly. The model also assumes reinvested coupons meet the tax only in the tenth year — a generous assumption for the coupon bond, since each reinvestment restarts the regressive table: the real cost tends to be higher than calculated here. The rates shown are those in force in 2026. Confirm the current rules with Tesouro Direto and consult a qualified financial adviser before deciding.