What is carry trade? A data-driven look at FX yields

In operation, it is a funding, rollover, margin and execution structure moving through a market dominated by derivatives.

What is carry trade in practical terms? A trader borrows or synthetically funds in a low-yield currency, then holds a higher-yield currency or asset. The intended return comes from the interest-rate differential. But the rate gap is only the first line of the calculation. Spot movement, dealer financing, swap points, bid-offer spread, position size and volatility determine the realised result.

The mechanical error is treating a policy-rate spread as a yield. It is not. It is an input.

The mechanics of yield-based positioning

A currency carry trade is normally an unhedged cross-currency position. The trader is long the investment currency and short the funding currency. If the long leg produces higher financing income than the short leg costs, the position accrues positive carry over time.

The standard framework is direct:

1. Select a funding currency with low short-term rates or low implied funding costs.

2. Sell that currency against a currency with higher short-term rates.

3. Hold the exposure through daily rollovers, tom-next transactions, FX swaps or forward structures.

4. Collect the net financing differential if spot remains sufficiently stable.

5. Exit before spot losses, widening funding costs or execution friction consume the accumulated carry.

The key phrase is “if spot remains sufficiently stable.” A carry position can earn small daily financing credits for weeks and lose multiple months of accrued income during a single repricing event.

For an OTC retail account, the daily amount is generally visible as a swap or rollover adjustment. For an institutional book, the same economic exposure may be created through spot-plus-forward, FX swaps, cross-currency funding or derivatives against a higher-yielding asset. The wrapper changes. The exposure does not.

A retail platform may display a positive overnight swap on a pair. That does not establish a viable FX yield strategy. The displayed figure can change with interbank funding conditions, dealer mark-ups, account denomination, weekend treatment and the broker’s own rollover policy. A positive swap line is a quote. Not a forecast.

Carry is not the policy-rate gap. Carry is the residual after spot, funding, spread and leverage have taken their share.

The yield differential also has to be measured on the correct horizon. A central bank’s current policy rate may be unchanged while the forward curve has already priced several cuts or hikes. FX forwards and swap points react to expected rate paths, not only the current overnight setting. This is why a headline comparison of two policy rates is structurally incomplete.

Uncovered interest parity and the reality of FX returns

Uncovered interest parity, or UIP, provides the baseline objection to the carry trade. In its simplest form, a higher-yielding currency should depreciate by enough, over time, to offset its interest advantage. If that adjustment occurred consistently and cleanly, unhedged carry would not offer an excess return.

Observed FX markets do not behave that neatly at short horizons. Empirical evidence has repeatedly challenged UIP. This gap between theory and realised price behaviour is the operating space for carry strategies.

But the failure of UIP is not a free-arbitrage result. Covered interest parity concerns a hedged transaction using forwards or swaps. A conventional carry trade leaves the FX risk open. The trader is paid, in effect, for holding a position whose downside can be discontinuous.

A simplified one-period return can be stated as:

Realised carry return = net financing differential + spot return − transaction costs

Each component can reverse the intended outcome.

• Net financing differential: The rate advantage after the dealer’s rollover terms, funding basis and any financing mark-up.
• Spot return: Usually the dominant component during stressed sessions. A funding-currency rally can exceed annual carry in hours.
• Transaction costs: Spread, commissions, slippage, swap-roll friction and, for larger tickets, market impact.
• Position scaling: Leverage does not alter the percentage return on notional. It increases the percentage impact on posted margin.

Consider two pairs with the same apparent annual rate differential. One trades with dense two-way liquidity, narrow top-of-book spreads and stable implied volatility. The other trades through thinner liquidity and larger gaps around local fixing windows. The first may have a lower nominal yield but a better execution-adjusted profile. The second can offer more carry and less usable return.

This is where currency carry trade examples are often misread. Naming a low-rate currency and a high-rate currency explains the direction of the trade. It does not explain its quality. The required dataset is broader:

• Overnight index rates and the expected policy path.
• Forward points and FX-swap pricing across the intended holding tenor.
• Pair-specific realised and implied volatility.
• Spot sensitivity during risk-off sessions.
• Depth of market during Asian, London and New York overlap.
• Rollover charges as applied by the actual execution venue.
• Correlation between the currency pair and the rest of the portfolio.

A carry model that excludes forward pricing and realised volatility is not a model. It is a rate screen.

08:00 UTC: leverage, margin and the amplification of exposure

Carry accrues gradually. Margin is repriced continuously.

The CFTC illustrates the scale directly: a 2% margin requirement can control a $100,000 position with $2,000. That is 50:1 notional leverage. For non-major pairs under the U.S. limits referenced by the agency, a 5% margin requirement equates to 20:1 leverage. These are U.S.-specific regulatory references, not global standards. The mechanics, however, are universal.

At 50:1 leverage, a 1% adverse move in the currency pair is approximately a 50% move against the initial margin before financing, spread and any changes in margin requirement. A 2% adverse move can consume the original margin allocation. The carry earned over a few days is operationally irrelevant at that point.

The table is arithmetic, not a loss forecast. It excludes spread, slippage, financing and forced-liquidation effects. Those exclusions are precisely the issue. During normal conditions, carry traders tend to observe the small positive daily adjustment. During a disorderly unwind, the relevant figures are executable bid depth, stop trigger logic and price dispersion across venues.

Retail margin systems add another layer. A position may be economically rational under a long holding horizon and mechanically impossible under account-level margin rules. Mark-to-market losses can trigger liquidation long before the expected carry has time to accumulate. The strategy is therefore not just a view on rates. It is a balance-sheet design problem.

A robust position-sizing process starts with adverse spot distance, not yield.

1. Set the maximum loss in account currency. This must include a gap allowance, not merely a stop level shown on a chart.

2. Translate that loss into notional. Use pair-specific pip value and the plausible stress range for the chosen horizon.

3. Test margin utilisation after the adverse move. Initial margin is not a risk budget. It is collateral.

4. Add rollover and spread assumptions. Use the actual venue’s financing schedule where possible, not an interbank theoretical rate.

5. Reduce for correlation. Multiple carry positions can be one concentrated short-volatility book under different symbols.

The final point is frequently missed. Long several high-yield currencies against several funding currencies can look diversified by ticket count. Under a volatility shock, the positions may all require the same exit: buy back funding currencies, sell investment currencies, reduce leverage. Correlation approaches the same side of the book when liquidity is least available.

A carry book should be sized against its unwind path, not against its expected daily swap credit.

13:30 UTC: volatility and the mechanics of a sudden unwind

Carry positioning usually functions best in low-volatility conditions with stable funding assumptions. This is not a behavioural observation. It is a payoff observation. The income stream is limited. The open-ended FX loss is not.

The August 2024 sell-off showed the sequence in compressed form. The Japanese yen, a prominent funding currency during recent low-rate periods, appreciated while investment currencies including the Mexican peso depreciated. The direction was consistent with leveraged positions being reduced: funding shorts bought back, higher-yielding exposures sold.

The sequence matters more than the label.

1. Rate-path expectations shift. The expected return from the funding leg changes. A previously cheap currency can become less cheap before a policy move is fully delivered.

2. Implied volatility rises. Option pricing signals a wider expected range. Risk limits tighten. Margin requirements may also become more restrictive at some venues.

3. Spot liquidity thins. Top-of-book prices remain visible, but executable size declines. The gap between displayed and filled price widens.

4. Stops and risk limits activate. Leveraged accounts and systematic strategies create synchronous market orders.

5. Funding currency strengthens. Short-covering converts a gradual repricing into a directional move.

6. Carry income becomes immaterial. A few basis points of accrued yield cannot offset a multi-figure spot adjustment.

This is why carry trade risk metrics cannot be reduced to the interest-rate differential. The rate spread tells the trader what can be earned in a stable period. Volatility tells the trader what can be lost before that period finishes.

Useful monitoring variables are practical, not ornamental:

• Implied volatility term structure. A sharp rise in near-dated implied vol changes the cost of protection and signals greater gap risk.
• Realised volatility relative to carry. If a pair’s ordinary weekly range can erase months of expected carry, the risk/reward is thin.
• Funding-currency spot momentum. A persistent rally in the short leg is a direct impairment of the position.
• Swap-point repricing. Deterioration in forward points can reduce the expected financing return before the spot market confirms it.
• Depth of market by session. A carry trade held through an illiquid local window has a different liquidation profile from the same notional held during London-New York overlap.
• Cross-pair correlation. The relevant stress test is not one position in isolation. It is the simultaneous move across the whole funding complex.

Execution data is central here. Tick data can show whether a move was continuous or discontinuous. A continuous decline can often be reduced with controlled routing. A gap through multiple price levels turns the expected stop price into a reference point rather than an execution guarantee. The difference is slippage.

For institutional execution, this also raises routing questions. A FIX API feed may show multiple liquidity providers and nominally tight quotes. During an unwind, quote durability matters more than quote count. A venue that streams aggressive top-of-book pricing but rejects or re-quotes size is not providing usable depth. The carry model must be tested against fills, not screenshots.

16:00 UTC: quantifying the invisible carry flow

The global FX market is large enough to conceal intent. BIS data recorded average daily turnover of $7.5 trillion in April 2022, with FX swaps accounting for 51%. It would be incorrect to treat that figure, or the FX-swap share, as a measure of global carry exposure.

An FX swap can fund a carry position. It can also hedge trade settlement, manage corporate cash, roll a balance-sheet exposure, facilitate a securities transaction or serve other treasury functions. Standard market statistics record instruments and counterparties. They generally do not identify the economic purpose of each trade.

The same limitation applies to funding-currency borrowing. Large short positions in a low-yield currency may be related to carry. They may also reflect hedging, asset allocation, trade finance or derivative structures. The live global carry notional cannot be extracted cleanly from aggregate public data.

That does not make the market unobservable. It changes the method.

A working surveillance framework combines imperfect indicators rather than searching for a single definitive number:

The correct output is probabilistic. Carry conditions may be supportive, deteriorating or unstable. They are not “on” or “off.”

BIS has also noted that sizeable short positions in funding currencies can amplify exchange-rate responses to monetary-policy tightening. The mechanism is direct. A change in the cost of funding forces repricing. Leveraged books reduce exposure. The buyback of the funding currency adds to the initial move. This is structural convexity in the short leg.

The strategy therefore has an asymmetry. Gains are usually collected in increments. Losses can arrive through correlated exits, thin depth and accelerated short-covering. A performance report that shows only average daily carry hides the part of the distribution that matters.

The technical verdict: carry is a funding trade with an FX tail

Carry trade mechanics begin with interest-rate differentials. They do not end there.

A valid carry assessment requires four linked calculations: expected net financing, spot sensitivity, leverage-adjusted drawdown and executable exit capacity. Remove any one of them and the result becomes overstated. The rate gap alone is not return. The displayed rollover alone is not funding. The margin requirement alone is not risk control.

The cleaner operating rule is strict: treat carry as compensation for holding unhedged currency exposure through a potentially unstable funding regime. Enter only when the expected financing income remains meaningful after transaction costs and when the position can survive a volatility expansion without forced liquidation.

The market does not pay carry for patience. It prices carry against the probability that the funding leg will move first.