Securitized Products: How the TBA Market and Negative Convexity Work

A thirty-year mortgage bond can behave like a two-year bond one month and a six-year bond the next, and the maturity date never changes. That whiplash is not a glitch. It is the defining fact of the securitized products desk, the least explained business on the fixed income floor, and nearly everything about the job comes back to it.

The desk trades agency mortgage-backed securities, non-agency residential mortgage-backed securities, commercial mortgage-backed securities, and asset-backed securities built from auto loans, credit cards, and student loans. Collateralized loan obligations are built the same way, from pooled corporate loans rather than pooled mortgages, but banks run them on a separate desk.

What the securitized products desk actually trades

Agency mortgage-backed securities are pools of home loans bought by Fannie Mae, Freddie Mac, or Ginnie Mae and turned into pass-through bonds, where a slice of every borrower's monthly payment flows straight to the bondholder. Ginnie Mae pools carry loans insured by the FHA or guaranteed by the VA. Fannie Mae and Freddie Mac buy conventional loans that meet their own underwriting standards instead.

The most heavily traded pass-throughs are the thirty-year and fifteen-year pools, quoted on a trading desk as FN30 and FN15 for Fannie Mae paper and G230 for Ginnie Mae. Since 2019, Fannie Mae's and Freddie Mac's thirty-year pass-throughs have settled under a single Uniform MBS structure, so a pool from either agency can fill the same trade.

• Agency MBS: pass-throughs and CMOs guaranteed by Fannie Mae, Freddie Mac, or Ginnie Mae
• Non-agency RMBS: residential loans outside agency standards, jumbo mortgages the most common example
• CMBS: loans against offices, apartments, and retail property, in both an agency and a private label form
• ABS: auto loans, credit card receivables, and student loans, sliced the same way a mortgage pool is

Non-agency residential mortgage-backed securities carry no government backing, and jumbo mortgages above the conforming loan limit are the most common collateral behind them. Commercial mortgage-backed securities pool loans against offices, apartment buildings, and shopping centers rather than houses. Freddie Mac guarantees a large agency slice of that market through its multifamily program; a separate, larger private label conduit market carries no such guarantee.

Asset-backed securities apply the same technique furthest from housing, turning an auto loan pool, a credit card receivable pool, or a student loan pool into bonds the same way a mortgage pool becomes one. The mechanics are identical across all four families: pool the cash flows, sell claims on them, and let a special purpose entity stand between the loans and the seller's own balance sheet.

Who trades with the desk, and what they want

Real money asset managers, banks, insurance companies, and pension funds are the natural buyers of the safest, most senior tranches, because a guaranteed or highly rated bond funds a long dated liability well. Hedge funds and specialist credit managers sit further down the structure, buying mezzanine and first loss risk for the yield.

Mortgage REITs occupy a category of their own. Firms such as AGNC Investment and Annaly Capital Management exist specifically to hold agency mortgage-backed securities funded with short term repo, so their entire business model is a bet on this one desk's product.

Mortgage originators are a different kind of client altogether. Between locking a homebuyer's rate and actually closing the loan, an originator is exposed to a move in rates with nothing yet to sell. Selling a TBA contract forward locks in today's price for a pool it has not yet created, which is the reason the TBA market exists at all: it lets the primary mortgage market and the secondary bond market clear at the same price.

Unlike the rates desk, which trades a single government curve, the securitized desk prices an embedded option inside every bond it touches, which is why its clients skew so heavily toward specialists who model prepayment behavior for a living.

How agency mortgage-backed securities actually work

The word "agency" gets used as if it describes one uniform government promise, and it does not. Ginnie Mae is a government owned corporation inside the Department of Housing and Urban Development, and its guarantee is the explicit full faith and credit of the United States. Fannie Mae and Freddie Mac are government sponsored enterprises, privately owned even while under conservatorship, and their guarantee has never been a statutory obligation of the federal government.

That distinction matters for pricing. Ginnie Mae paper trades at a small yield premium to Fannie Mae and Freddie Mac paper precisely because the market treats the three guarantees as related but not identical. It also matters for anyone reading a careers page or a research note that treats "agency" as a single word: the desk trades three issuers whose promises differ, and the difference is legal, not cosmetic.

The TBA market: trading securities that do not exist yet

A TBA, or to-be-announced, trade settles on only six agreed terms: issuer, coupon, maturity, price, par amount, and settlement date. Neither side names the specific loan pools until later, because the convention treats any eligible pool of a given coupon and program as interchangeable with any other. That is what makes the market liquid enough to be, by trading volume, second only to US Treasuries among fixed income products.

Only single class, generic pass-throughs qualify. A CMO carves one pool's cash flows into several tranches with different risk profiles, so no two CMOs are interchangeable, and neither are non-agency bonds or interest only strips. All three trade specified pool instead, one deal at a time, with a real prospectus behind them.

A dollar roll uses the same forward mechanic for financing rather than investment. A holder sells a TBA for near month settlement and simultaneously agrees to buy back the same coupon and program for a later month at a lower price, the difference between the two months being called the drop.

The trade behaves like a repurchase agreement collateralized by the mortgage pool itself. When deliverable collateral is scarce relative to demand, the drop can move enough that traders call the roll "special."

How the desk makes money

Flow trading earns the bid-offer spread on TBA and pass-through business, the same way any market-making desk does. Holding inventory funded through repo earns carry, the gap between what the bond yields and what the financing costs.

The dollar roll adds a third source that is really a financing trade wearing a forward-market disguise, and it can be more or less profitable depending on how special the roll runs. On the origination side, structuring a CMO or a private label deal earns a fee for building and placing the tranches, work that sits closer to capital markets than to trading.

Why MBS has negative convexity

Prepayment risk is the risk that a borrower does not repay principal on schedule. It splits into two opposite failure modes: contraction risk, when a borrower repays faster than expected, and extension risk, when a borrower repays more slowly than expected. A CMO's whole purpose is redistributing that risk across tranches, so that a senior class carries far less of it than the pool as a whole.

Mortgage securities carry both kinds of risk at once because a homeowner effectively owns a call option on their own loan. Falling rates make refinancing attractive, so prepayments speed up and principal comes back early, right when an investor would rather keep the now-attractive coupon. Rising rates make refinancing unattractive, so prepayments slow and the investor is stuck holding a below-market coupon for longer than planned.

Because prepayments accelerate as rates fall and slow as rates rise, mortgage securities typically carry negative convexity.

One large mortgage REIT states the consequence plainly in its own SEC filings, and the pattern is worth turning into arithmetic.

The standard duration-convexity approximation makes the size of the effect concrete: the percentage price change is roughly minus the duration times the change in yield, plus one half times convexity times the change in yield squared. Take two hypothetical ten-million-dollar bonds, both priced at par with a duration of four. An ordinary, option-free bond carries positive convexity of 80; a mortgage bond with the same starting duration carries negative convexity of minus 300, because the prepayment option flips the sign.

If rates fall two percentage points, the option-free bond's approximation works out to roughly plus 9.6 percent; the mortgage bond's works out to roughly plus 2.0 percent, because the negative convexity term subtracts most of the gain the duration term predicted. If rates rise two percentage points instead, the option-free bond loses roughly 6.4 percent, while the mortgage bond loses roughly 14.0 percent, because now the same negative term adds to the loss instead of offsetting it.

The mortgage bond captures about a fifth of the option-free bond's upside and loses more than twice as much on the downside, from the same starting duration. That asymmetry, not a bigger number in either direction alone, is what a trader means by negative convexity. Compare it against an ordinary bullet bond's ordinary, positive convexity on the bond calculator to see the shape side by side.

Option-adjusted spread, in plain terms

A plain yield spread treats a bond's future cash flows as fixed, which is a reasonable assumption for a Treasury and a poor one for a mortgage bond whose cash flows change with the prepayment path. Option-adjusted spread solves that by modeling many interest rate paths, valuing the prepayment option inside each one, and reporting the spread that is left over after the option's cost has been stripped out.

It answers the question a plain yield cannot: how much is an investor really being paid for taking on the risk that a borrower can prepay whenever it suits them.

How a securitization is actually built

Every securitization starts with a special purpose entity that buys the underlying loans from the original lender and issues new securities against them, structured so the pool stays bankruptcy-remote from the seller. Cash flows are then split into tranches with a strict order of priority: senior classes get paid first and absorb losses last, and a first loss class sits at the bottom, absorbing the earliest defaults in exchange for a far higher yield.

Freddie Mac's own K-Deal program shows the structure clearly, because it publishes a dated, illustrative example of exactly how a multifamily commercial mortgage deal is built.

Repaid from the top down. Each band is sized by its share of the structure.

Class A-1 and A-2 (AAA, guaranteed)82% · Guaranteed senior classes

Class A-M (unrated, guaranteed)14% · Guaranteed but unrated

Class D (unrated, unguaranteed)← you, if you buy the first loss piece5% · Sold privately, first loss

Freddie Mac Multifamily K-Deal Program Overview, investor presentation as of March 2026. Class sizes are illustrative; the percentages match the presentation's own subordination levels.

The guaranteed senior classes cover the top 81.5 percent of the structure and are eligible for repo funding because Freddie Mac stands behind them. The unrated but still guaranteed A-M class sits beneath that, and the unguaranteed Class D, sold privately rather than in a public offering, absorbs the first five percent of any loss.

Since the program began in 2009, Freddie Mac has securitized more than 29,000 loans totaling over 640 billion dollars through June 2026, almost entirely on this same basic template.

The US securitized market, by segment

Agency MBS

$10,689bn

Non-agency RMBS

$840bn

CMBS

$673bn

ABS

$1,585bn

SIFMA's own outstanding series for this market. Issuance from the same source is current.

The market has kept growing regardless. Full year 2025 US mortgage-backed securities issuance ran to 1.9 trillion dollars, up almost 19 percent on the year before, and asset-backed securities issuance reached 517.4 billion dollars, up a third.

Agency mortgage-backed securities alone stood at more than 9.4 trillion dollars outstanding by mid-2026 on loan-level data, split roughly 37 percent Fannie Mae, 33 percent Freddie Mac, and 30 percent Ginnie Mae. Whatever the exact current total, this is one of the largest fixed income markets in the world, and one of the hardest to measure precisely from published statistics alone.

How the desk changed since 2008

The 2008 crisis was, in large part, a private label mortgage crisis, and the regulatory response reshaped the desk more than almost any other on the floor. The Volcker rule bans short-term proprietary trading by a banking entity, but it explicitly exempts market making, underwriting, and risk-mitigating hedging, and it separately excludes loan securitizations from the definition of a covered fund.

A bank can still make markets in agency and non-agency mortgage bonds and sponsor a securitization vehicle. What it cannot do is speculate with its own balance sheet dressed up as either activity.

Dodd-Frank's credit risk retention rule forced every sponsor to keep skin in the game, retaining at least five percent of the credit risk on a deal it sells to investors. The rule phased in earlier for residential mortgages than for anything else: it took effect in December 2015 for RMBS and a full year later, in December 2016, for every other kind of asset-backed security.

$640bn+

Freddie Mac K-Deal volume since 2009

through June 2026

5%

Dodd-Frank risk retention

skin in the game

90%+

of agency MBS trades TBA

2021 trading volume

The Federal Reserve became the largest single holder of agency mortgage-backed securities during quantitative easing, buying them outright through its System Open Market Account. Those purchases fell to zero in November 2022 and have stayed negligible since, removing a buyer that had absorbed a meaningful share of new issuance for over a decade.

Electronification: what still trades by phone

TBA contracts, the most liquid instrument on the desk, are traded predominantly electronically and have been for years. In 2021, agency mortgage-backed securities carried roughly 281 billion dollars in daily trading volume, and TBA accounted for more than 90 percent of it.

Everything less standardized has lagged badly. A Coalition Greenwich study published in February 2025, based on interviews with US buy-side mortgage traders conducted the previous autumn, found that more than 75 percent of trading volume in CMOs, asset-backed securities, CMBS, and specified pools still runs over the phone or through a chat window.

The same electronification wave that reshaped corporate bond and Treasury trading over the past decade has barely touched anything on this desk that is not a generic pass-through.

Who the big dealers are

The desk exists at all ten dealer banks this site names, but the ten do not describe the business the same way.

No single league table of dealer market share in this market is reliable enough to repeat. The rankings that do circulate age quickly, since the last widely cited one still named Credit Suisse as a top competitor, and Credit Suisse has not existed as an independent dealer since UBS absorbed it in 2023.

What this business is called, and what it is not

American trading desks and job postings mostly say "securitized products," with a z. The industry's own US trade association makes the point explicitly: it recently renamed itself from the Structured Finance Industry Group to the Structured Finance Association, and its own site states that securitization is really a synonym for structured finance in how the market actually uses the words.

Europe leans on different spelling and a different emphasis. The European trade association's own market reports are titled "Securitisation," with an s, and pair it with "structured finance" rather than "securitized products." The US market for this business is also much larger than Europe's on any measure, a gap that traces to capital rules and investor base rather than to the mechanics, which are largely the same on both sides of the Atlantic.

A day on the securitized desk

The day is organized around the TBA settlement calendar rather than around any single market call. SIFMA's own good delivery rules set the fixed point in that calendar: pool notification 48 hours before settlement, which means pricing runs and the allocation of pools to trades have to clear ahead of that deadline on every settlement date, not whenever it suits the desk.

Structuring work runs on a different clock. Building a new CMO or private label deal means modeling how the pool's cash flows split across tranches before the deal ever prices, the same waterfall logic described above, worked out ahead of a sale rather than against a settlement date.

• Pricing runs across the coupons and programs the desk actively trades
• Pool notification and allocation ahead of the 48 hour settlement deadline SIFMA's rules set
• Cash flow modeling for a new CMO or private label deal, when one is in structuring

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

What the interview tests

The technical bar on this desk is unusually concrete, because the underlying product is genuinely more mathematical than most fixed income instruments. A strong candidate can explain prepayment risk and its two opposite forms, walk through why an embedded prepayment option produces negative convexity, and describe what option-adjusted spread is actually solving for.

Bond math questions on duration and convexity show up here more than almost anywhere else on the floor, so it is worth working through the bond math that underpins this desk specifically before an interview, alongside the broader fixed income interview preparation that applies everywhere.

• Explain the difference between contraction risk and extension risk
• Walk through why a mortgage bond has negative convexity, with the direction of the effect in both a rate rally and a selloff
• Describe what option-adjusted spread measures that a plain yield spread cannot
• Explain how a CMO's tranching redistributes prepayment risk rather than eliminating it
• Know why only generic agency pass-throughs, not CMOs or non-agency bonds, trade TBA

Where it leads, and how to prepare

Candidates describe several common exits from this desk: asset managers and credit-focused hedge funds that buy mortgage and asset-backed securities for a living, mortgage REITs whose entire portfolio is the product this desk trades, and occasionally the structuring or capital markets side of the same trade.

The skill that travels best is not any one bank's process, it is the ability to read a cash flow waterfall and price an embedded option, which few other fixed income products demand in the same way.

Preparation for this desk rewards depth over breadth. Working through the arithmetic behind negative convexity until it can be explained without notes matters more than memorizing every product acronym on the floor.

Understanding what a bank actually pays at each level, covered in the guide to fixed income trader pay, helps calibrate how hard to push in a compensation conversation. A structured run through likely questions on the interview readiness tool is a reasonable last step before a first round.