Rates Trading Desk: How Steepeners, Carry and Swap Spreads Work

Type SOFR into two of the most recommended free guides for a rates trading interview, and both still describe it as a future change: one says the rate "will replace" LIBOR, the other tells candidates to know "LIBOR, or SOFR when it is phased out." LIBOR finished phasing out on June 30, 2023, three years ago.

What follows is the desk as it actually runs: a real Treasury curve pulled from the Treasury's own published data, the arithmetic behind a curve trade checked twice, and the swap spread math a strong candidate can walk through on a whiteboard.

What the rates desk actually trades

A rates desk trades the government bond market and everything priced off it: Treasury bills, notes, bonds and inflation-protected securities in cash form, and interest rate swaps and swaptions as the derivative overlay on top. Every one of those instruments is a bet on the same underlying curve, just expressed with different cash flows and different counterparties.

The four products break down like this:

• Government bonds: bills, notes, bonds and TIPS, the cash instruments the whole curve is built from.
• Interest rate swaps: an exchange of a fixed rate for a floating rate, now referencing SOFR rather than LIBOR.
• Swaptions: an option to enter a swap, used to hedge or express a view on volatility as well as on the level of rates.
• Inflation-linked bonds: TIPS sit as a section of this desk rather than a separate one, since the same curve and the same clients trade them.

Who is on the other side of every trade

Two client types dominate the flow. Hedge funds borrow in the repo market to hold a long Treasury position against a short Treasury futures position, a construction known as the cash-futures basis trade, and the desk prices and finances both legs. Real money managers, insurers and pension funds sit on the other side: they buy duration to match long-dated liabilities, and they are the natural counterparty when a hedge fund wants leverage rather than ownership.

• Hedge funds: leverage, often via the cash-futures basis trade, financed in repo.
• Real money (asset managers, insurers, pensions): duration to match long-dated liabilities.
• Corporates and other banks: hedging their own interest rate exposure through swaps.
• Official institutions: central banks and sovereign accounts holding Treasuries as reserve assets.

Corporates and other banks use the desk to hedge their own interest rate exposure through swaps rather than by reshaping a bond portfolio, and central banks and other official institutions hold and trade Treasuries as reserve assets. A junior on the desk learns to recognize these client types from the shape of an order alone, long before a name ever appears on the ticket.

The desk also trades against itself, in a sense: a swap taken on for one client often needs to be hedged with a Treasury future or a cash bond bought from a different client, so a single trade can touch three or four counterparties before the day is done.

Reading which client type is driving an order, real money reducing duration ahead of a Federal Reserve meeting versus a hedge fund putting on a basis trade, is most of what separates a strong market maker from an average one.

None of these clients trade in isolation from the others. A large real-money sale can widen the exact spread a hedge fund is trying to trade, and a desk that misreads which flow is temporary and which is a genuine shift in positioning ends up hedged against the wrong risk for the rest of the day.

How the desk makes its money

Market making is the core of it: quoting a bid and an offer, buying from one client and selling to another, and earning the spread between them. The Volcker Rule, in force since 2014 with full compliance required from 2015, otherwise bans a bank from proprietary trading.

It carves out three exemptions that keep a trading desk legal: market making, underwriting and risk-mitigating hedging. Holding inventory to serve client flow sits inside that market-making exemption; holding it purely to bet on rates for the bank's own account does not.

Most of what happens on this desk is flow: quoting two-way prices in on-the-run Treasuries and standard swaps, and turning the resulting inventory over the same day rather than holding it.

Warehousing a client's less liquid bond for longer than that, or holding a curve or carry position because the desk has a view rather than because a client is still on the other side of it, is risk, the kind the market-making and hedging exemptions exist to keep separate from banned proprietary trading.

Financing is the other half of the business. A desk that holds Treasuries funds them in the repo market, and the gap between what a bond yields and what it costs to finance is carry, the same arithmetic a junior has to reproduce in an interview and a trader reproduces every morning on the real book.

Balance sheet is not free: post-crisis leverage rules mean every Treasury a dealer holds requires capital behind it, which is why a desk cares as much about financing cost as it does about the bid-offer spread on a given trade.

The market by the numbers

US Treasury securities outstanding reached $30.8 trillion in the first quarter of 2026, up 7.9 percent from a year earlier, inside a total US fixed income market of $50.5 trillion. Globally, interest rate derivatives made up 79 percent of the $846 trillion in over-the-counter derivatives notional outstanding at the end of June 2025, by far the largest slice of that market.

Electronic trading has reshaped how the cash market itself trades: dealer-to-client Treasury volume executed electronically rose from 46 percent in May 2019 to 63 percent in May 2024, with average daily volume nearly doubling over the same five years.

$30.8tn

Treasuries outstanding

Q1 2026

26

primary dealers

current New York Fed list

63%

Treasury trading done electronically

May 2024, up from 46% in 2019

79%

of OTC derivatives notional is rates

BIS, June 2025

Electronification has not eliminated voice trading, particularly in size or in less liquid off-the-run issues, but it has compressed bid-offer spreads and pushed a growing share of the smallest, most standardized tickets onto a screen rather than a phone call. A junior on a modern rates desk spends more of the day managing electronic execution and less of it manually working a single order than a junior would have a decade earlier.

SOFR finished the job three years ago

USD LIBOR panels for the one, three, six and twelve month tenors stopped publishing for good on June 30, 2023, the step the Alternative Reference Rates Committee itself calls the end of the transition. The one week and two month tenors had already gone in December 2021. A synthetic version of LIBOR kept running afterward for legacy contracts that genuinely could not be amended, and even that stopped for good on September 30, 2024.

Every new swap on the desk today references SOFR, the Secured Overnight Financing Rate, a rate built from actual overnight Treasury repo transactions rather than a panel of bank estimates. That is the whole point of the switch: SOFR cannot be quoted by a trader with an incentive to move it, because it is measured from a market that already trades over a trillion dollars a day.

Reading the curve: from deep inversion to normal

On July 3, 2023, the two year Treasury note yielded 4.94 percent and the ten year yielded 3.86 percent, a spread of minus 108 basis points and the deepest point of an inversion that had begun in July 2022. Short paper paid more than long paper for two straight years, the kind of curve shape that has preceded every US recession since the 1970s, and this time it did not.

That inversion was also the longest since the 1970s, a stretch of about two years against the next longest run, in 1978, of about a year and eight months. The curve did not sit at its deepest point the whole time: it turned positive again on September 4, 2024, when the ten year yield closed a hair above the two year for the first time since the inversion began.

Yield by maturity. The maturity axis is on a square-root scale, so every point sits at its true maturity and the short end stays readable.

3 Jul 2023 (inverted)11 Sep 2026 (normal)

Curve	1M	3M	6M	1Y	2Y	5Y	10Y	30Y
3 Jul 2023 (inverted)	5.27%	5.44%	5.53%	5.43%	4.94%	4.19%	3.86%	3.87%
11 Sep 2026 (normal)	3.93%	4.07%	4.12%	4.35%	4.63%	4.78%	4.96%	5.35%

US Treasury par yields, two dated snapshots. Source: U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates.

A more normal, upward-sloping curve sits alongside that inverted one. On September 11, 2026, the two year yielded 4.63 percent and the ten year 4.96 percent, a spread of positive 33 basis points, close to what a healthy curve has looked like across most of the last fifty years.

Steepeners, flatteners and butterflies

A steepener bets the gap between two maturities widens; a flattener bets it narrows; a butterfly bets on the curve's curvature, the belly moving relative to both wings. All three are built the same way: long one maturity, short another, sized so a parallel move in the whole level of rates barely moves the position at all. What is left is a pure bet on the curve's shape.

Sizing that trade means matching DV01, the dollar change in a position's value for a one basis point move in yield. At typical yield levels, a two year note carries a DV01 of about $189 per $1 million face value, and a ten year note carries about $781, roughly four times as much, because the ten year has far more years of cash flow left to discount.

2-year note

$189

10-year note

$781

The 10-year note moves about 4 times as much per basis point as the 2-year, which sets the hedge ratio below.

Buy about $4.13 million face value of the two year note and sell $1 million face value of the ten year note, and the two legs' DV01s match almost exactly: $4.13 million times $189 comes to $780.57 against the ten year leg's $781.

If both yields then rise 10 basis points together, the short ten year gains $7,810 and the long two year loses about $7,806, close enough to net to nothing. If instead only the ten year yield rises 10 basis points, the position gains the full $7,810, the curve bet working exactly as intended.

Carry and roll-down, worked through

Carry is what a position earns simply for existing: the bond's yield minus the cost of financing it in repo. On September 10, 2026, the ten year Treasury yielded 4.95 percent while SOFR, the standard financing-rate proxy, sat at 3.62 percent, for carry of 1.33 percent, or 133 basis points a year.

Roll-down is the second source of return on an upward-sloping curve. A bond bought today at the ten year point is, three years from now, a seven year bond, and the seven year point on the same curve yields less: 4.84 percent against 4.95 percent, an 11 basis point pickup.

Multiplied by the ten year note's modified duration of about 7.8, that comes to roughly 0.86 percent of price gain over three years from roll-down alone, or about 29 basis points a year.

Carry

133bp

Roll-down, annualized

29bp

Combined

162bp

Add the two together and a position held for carry and roll-down alone is underwriting about 162 basis points a year over its financing cost, on the assumption that the curve does not move, which it never quite does. The two numbers are a starting point for a trade, not a guarantee of one.

Swap spreads, and why they went negative

A swap spread is the fixed rate on an interest rate swap minus the yield on a Treasury of the same maturity. On September 11, 2026, the ten year swap rate was 4.858 percent against a ten year Treasury yield of 4.96 percent, a spread of about minus 10 basis points, negative rather than positive.

That sign is not new. The ten year swap spread ran positive until October 2015 and has spent most of the time since below zero; the thirty year spread has been negative continuously since 2008.

The mechanism is balance sheet, not credit: post-crisis leverage rules make banks hold capital against every Treasury on their books, while an offsetting swap position sits off balance sheet. Dealers who are net long Treasuries, which is the normal state of the market now, price that cost into the swap rate rather than absorb it.

The cash-futures basis, briefly

A related trade sits one step over on the desk's risk map, built directly on top of repo. A Treasury future settles against a single cheapest-to-deliver cash bond, picked out of a basket of eligible issues by a conversion factor.

The basis trade profits from the gap between that future's price and the cheapest-to-deliver bond's own repo-financed carry: buy the bond, finance it overnight in repo, sell the future short, and collect the difference as the two prices converge toward delivery.

Hedge funds run this at scale and with real leverage; one Federal Reserve estimate put related holdings at more than $300 billion accumulated since early 2022. It is a repo-financed trade first and a futures trade second, so the deeper mechanics of that financing leg belong to the repo desk.

Who the dealers are

The New York Fed's list of primary dealers, the firms that must bid at every Treasury auction and stand ready to trade with the Fed's own desk, ran to 26 names in 2026. Ten of them are the dealer banks named most often on this site, each running a rates business inside a broader markets division:

Every one of the ten is a primary dealer in its own right, alongside sixteen other firms, from regional and Canadian banks to specialist bond dealers such as Cantor Fitzgerald and Jefferies, that this site does not otherwise cover.

Being a primary dealer is a formal designation with formal obligations, not just a reputation; a firm that fails to show up and bid at auction risks losing the designation, which is part of why the list's membership shifts occasionally even though the ten household names on it rarely do.

A junior's day on the desk

The day is built around a small number of scheduled moments rather than a constant stream of activity. A junior arrives early to check overnight levels in London and Tokyo, load the swap pricer, and confirm the curve looks the way it should against yesterday's close. A morning meeting with sales and research sets the tone before the desk opens in earnest.

Whatever the day's scheduled release is, a jobs report, a Treasury auction, a Federal Reserve decision, is when the real work starts: client orders arrive in a burst, hedges get executed in Treasury futures, and a junior is booking trades and updating the blotter in real time rather than reading about markets. The afternoon slows into position reconciliation, an end-of-day report, and a hedge check before the desk hands off to the next region.

A junior on the Treasury cash desk spends more time on auction mechanics and less on volatility than a junior on the swaptions desk, and a junior trading TIPS inside this desk watches inflation prints the way a corporate bond trader watches earnings. The shared thread across all of them is that the calendar, not the clock, sets the pace of the day.

How the rates day compares with a credit or securitized trader's, hour by hour, is set out in a day in the life of a fixed income trader.

What the interview actually tests

Bond math is tested on its own. Beyond it, three things come up on almost every rates trading interview: constructing a DV01-neutral curve trade, working through carry and roll-down on a real position, and explaining why a swap spread can run negative. All three are worked through above, with the arithmetic checked rather than just described.

An interviewer is less interested in a memorized definition than in whether a candidate can adjust the arithmetic on the spot: what happens to the hedge ratio if yields move first, what happens to carry if the Federal Reserve cuts before the position is unwound. Being able to rebuild the worked examples above from scratch, not just recite their answers, is what separates a strong technical round from a weak one.

The other question that comes up constantly is a live one: what happened in markets recently, and what did it do to yields. There is no fixed script for it, because the honest answer changes with every Federal Reserve meeting and every major data release; a candidate is expected to track the ten year yield and the last policy decision closely enough to have a real answer, not a rehearsed one.

Working the DV01 and carry arithmetic above by hand once is worth more than reading it twice. A bond calculator that shows price, yield, duration, DV01 and convexity together is the fastest way to check the numbers against a different note or a different yield level, and a broader interview readiness check covers the rest of the desk-agnostic material.

Where rates traders go next

Many stay in trading, moving up within the same desk or laterally to a bigger dealer. The most common outside move is to the buy side: a macro-focused hedge fund seat, often inside a multi-manager pod, values exactly the products and the market-moving events a rates desk trades around every day, and the analyst-to-portfolio-manager track at these funds draws heavily from sell-side rates traders.

Asset managers hire from the same pool for portfolio and execution roles, and a smaller number move into research or economic strategy roles at banks or official institutions.

• Another dealer's rates desk, at a more senior level.
• A macro-focused hedge fund, often inside a multi-manager pod.
• An asset manager, in a portfolio or execution role.
• Research or economic strategy, inside a bank or an official institution.

What is specific to rates, rather than to trading generally, is how directly the skill set transfers. A rates trader already prices duration, curve shape and volatility every day, which is the exact toolkit a global macro fund needs.

That is why the move from sell-side rates trading to buy-side macro is one of the shortest paths across the fixed income floor. How that path compares with the exits open to other fixed income desks, and what each one tends to pay, is its own subject.

How to prepare

One book comes up more often than any other across trading-desk forums and reading lists for this interview: Bruce Tuckman and Angel Serrat's Fixed Income Securities, recommended specifically for its early chapters on yield curves, duration and the building blocks worked through above. It rewards a slow read more than a fast one; the early chapters are dense enough that working through the arithmetic by hand, rather than skimming for the conclusion, is what actually sticks before an interview.

Reading a real Treasury par curve cold, and saying without prompting whether it is inverted and by how much, is worth more than any flashcard. The same goes for the arithmetic above: rebuild the DV01-neutral hedge ratio on a different pair of maturities, and work through carry and roll-down on a bond picked at random using that day's actual yields, until the steps come without reaching for a calculator.

The one thing worth refreshing the morning of the interview itself is what the last Federal Reserve meeting did to yields, since that particular answer expires the moment the next meeting happens.

The bottom line

The rates desk is the fixed income floor's biggest seat, because the products it trades sit underneath almost everything else on it: a swap spread, a curve trade and a repo rate eventually touch every other desk.

SOFR has been the market's reference rate for more than three years, the curve that inverted for two of those years is normal again, and the arithmetic behind a steepener or a swap spread is short enough to write on one page and worth knowing cold before the interview.