
Suppose your business must pay a $100,000 invoice in six months. At EUR/USD 1.10, that payment would cost approximately €90,909. At 1.00, it would cost €100,000. The exchange-rate difference adds more than €9,000 to the bill.
Forecasting exchange rates means estimating a currency pair’s future value over a specified period. For businesses, that estimate can inform budgets and hedging decisions. For investors and traders, it can influence which currency exposures they take.
Interest rates offer one possible source of information. But turning that information into a useful forecast is surprisingly difficult. Many economic models have struggled to outperform a simple benchmark: predicting that the future exchange rate will equal today’s rate.
There are promising exceptions. Research by economists Zsolt Darvas and Zoltán Schepp found that a model incorporating long-maturity interest rates improved historical forecasts for the currencies and horizons they tested. Understanding how they achieved that—and what their results actually measure—is more useful than treating higher interest rates as an automatic currency-buying signal.
We’ll compare the main forecasting methods, work through the relevant calculations, and examine the evidence. Along the way, we’ll separate three things that are easily confused: a forward exchange rate, a prediction of the future spot rate, and a profitable trading decision.
Why Forecast Exchange Rates?
The business in our opening example needs to know whether an adverse currency move could wipe out its profit margin. A forecast becomes useful when it helps answer that kind of practical question.
The importance of forecasting exchange rates depends on who is using the estimate:
Importers and exporters use exchange-rate assumptions to budget costs, price contracts and assess future cash flows.
Investors consider whether currency movements could amplify or offset returns on foreign assets.
Policymakers examine how exchange rates could affect inflation, trade and economic activity.
For our hypothetical importer, a forecast of EUR/USD 1.10 provides one planning estimate. Testing alternative outcomes—such as 1.00 or 1.20—shows how sensitive the business is to being wrong. That scenario analysis can reveal whether the company has enough financial room to absorb an unfavorable move.
It may then choose to lock in a conversion rate with a forward contract. Doing so provides certainty about the contracted payment’s currency cost, even if waiting would eventually have produced a better rate. The purpose of that hedge is to control an exposure the business already has.
A useful forecast should therefore be judged against the decision it supports and the time horizon involved. Estimating next quarter’s import costs and predicting tomorrow’s currency movement are different tasks.
With that distinction established, we can compare the main forecasting methods—and the information each one uses.
The Main Methods of Forecasting Exchange Rates
Two analysts can examine the same currency pair and reach different forecasts because they use different information—or translate that information into predictions differently.
For practical purposes, we can group the main approaches to forecasting exchange rates into four broad categories.
Forecasting methods
Four approaches, compared
Compare the information each approach uses and the limitations to keep in mind.
| Approach | What it uses | Main limitation |
|---|---|---|
| Fundamental | Economic information such as interest rates, inflation, growth and trade flows, linked through an economic model. | A plausible economic relationship may produce weak forecasts, especially when its strength changes over time. |
| Technical and time-series | Past prices or returns, using trend rules, indicators or statistical patterns. | Patterns identified in historical data may fail when applied to subsequent observations. |
| Market-based | Today’s spot rate as a no-change forecast, or forward rates used as forecasting benchmarks. | A market price can serve a pricing or hedging function without accurately predicting the future spot rate. |
| Combined | Several forecasts blended through simple averages or estimated weights. | Combining forecasts cannot guarantee improvement, particularly when they share similar errors. |
Select an approach to compare its inputs and main limitation.
01Fundamental
- What it uses
- Economic information such as interest rates, inflation, growth and trade flows, linked through an economic model.
- Main limitation
- A plausible economic relationship may produce weak forecasts, especially when its strength changes over time.
02Technical and time-series
- What it uses
- Past prices or returns, using trend rules, indicators or statistical patterns.
- Main limitation
- Patterns identified in historical data may fail when applied to subsequent observations.
03Market-based
- What it uses
- Today’s spot rate as a no-change forecast, or forward rates used as forecasting benchmarks.
- Main limitation
- A market price can serve a pricing or hedging function without accurately predicting the future spot rate.
04Combined
- What it uses
- Several forecasts blended through simple averages or estimated weights.
- Main limitation
- Combining forecasts cannot guarantee improvement, particularly when they share similar errors.
Fundamental and price-based approaches differ primarily in their information sets. Fundamental models draw on economic relationships; price-based methods search for predictive information in the exchange-rate series itself. Both require testing against outcomes that were unavailable when the forecast was made.
The boundaries can overlap. A model might combine interest-rate differences with recent currency returns. Machine learning can estimate relationships using either type of input, so the label “AI forecast” tells us little about the underlying economic logic.
Explaining a currency’s value and predicting its next movement are also different tasks. A valuation model may suggest that a currency is expensive without establishing when—or how quickly—it will adjust. CFA Institute makes this distinction when discussing long-run equilibrium value and the difficulty of short-term forecasting.
Combining approaches is worth investigating, but sophistication alone proves little. Federal Reserve research on model averaging found that combining economic information could improve some forecasts while still producing predictions close to the no-change benchmark.
That gives us a demanding starting point: before accepting a complicated forecast, check whether it improves on simply using today’s exchange rate.
Why Today’s Exchange Rate Is Hard to Beat
Suppose EUR/USD is 1.10 today. A forecasting model predicts 1.16 in one month, while a no-change forecast simply predicts 1.10.
If the actual rate turns out to be 1.11, the model misses by 0.05 dollars per euro. The no-change forecast misses by just 0.01. The model correctly anticipated the direction, yet produced the larger forecasting error.
That simple comparison explains why researchers take the random walk without drift seriously. Under this benchmark, the expected future exchange rate equals the current rate. Actual prices can still move sharply—the forecast assigns an expected change of zero. Darvas and Schepp use this benchmark to evaluate their model.
Why can it be so difficult to improve on?
Exchange rates respond to expectations about future conditions. Information about inflation, growth or interest rates may already be reflected in today’s price. Even models in which economic fundamentals determine exchange rates can generate behavior close to a random walk under certain assumptions.
Adding variables also means estimating more relationships, potentially introducing errors that outweigh the extra information.
The fair test is repeated forecasting using only information available at each forecast date. Researchers call this out-of-sample evaluation. Both the model and the benchmark should predict the same currency, dates and horizons, with their errors compared using a clearly specified measure.
Data timing matters too. A model tested using subsequently revised economic statistics may look different from one using the figures forecasters actually had. Federal Reserve research has documented substantial effects from data revisions and changes in the evaluation period.
This gives us a concrete standard for the interest-rate research: does incorporating those rates improve forecasts beyond what today’s exchange rate already tells us?
Can Interest Rates Predict Currency Movements?
Suppose comparable one-year investments pay 5% in U.S. dollars and 3% in euros. The dollar investment offers a two-percentage-point interest advantage. But a euro-based investor must also consider what happens when those dollars are converted back into euros.
To understand the forecasting implications, we need to separate two questions.
What happens when interest-rate expectations change?
A surprise shift toward higher expected U.S. interest rates relative to euro-area rates can strengthen the dollar immediately, other things equal. Research from the Federal Reserve Bank of Boston links exchange-rate movements to monetary-policy surprises and changes in expected policy paths, alongside changes in expected excess returns.
The surprise matters. A rate increase that investors already anticipated may provide little new information. A smaller increase than expected could even accompany a weaker currency.
What return should investors expect after that adjustment?
Uncovered interest parity, or UIP, compares investments whose currency exposure remains unhedged. Under its simplifying assumptions, including no currency risk premium, the higher-yielding currency is expected to depreciate enough to offset its interest advantage.
In our example, that implies approximately 2% expected dollar depreciation over the year. The extra interest compensates, in expectation, for the currency movement.
An immediate appreciation and subsequent expected depreciation can therefore coexist. In a stylized model where the currency’s long-run expected value stays unchanged, favorable interest-rate news can lift its value today, leaving room for depreciation afterward.
The empirical difficulty is that currencies frequently fail to depreciate by the amount UIP predicts over short and medium horizons. That gap helps explain the appeal—and risk—of carry trades, which favor higher-yielding currencies.
Interest-rate differences provide a theoretical starting point. Their forecasting value still needs testing.
They also enter a more mechanical relationship: the calculation of a forward exchange rate. Working through that calculation shows why its meaning matters as much as its arithmetic.
Worked example · Covered interest parity
The forward exchange rate: price, not prediction
Two currencies, two interest rates, one hedged return. Calculate the one-year EUR/USD forward, then see what an unhedged dollar investment would be worth if the future spot rate changed.
- Spot EUR/USD
- 1.1000
- USD interest
- 5.0%
- EUR interest
- 3.0%
- 1-year forward
- 1.1214
Illustrative rates, annual compounding and matched one-year maturities. Forward displayed to four decimals.
What if the future spot rate differs?
Same $1,155 at maturity · two ways back into EUR
Try a weaker euro, a rate near parity pricing, or a stronger euro.
At 1.1214, the outcomes are virtually equal. The €0.04 difference comes from displaying the theoretical forward to four decimals.
The distinction: CIP prices a hedged exchange; it does not tell you where spot EUR/USD will trade next year.
See the €1,000 cash-flow proof
A · Stay in euros
Invest €1,000 at 3% for one year.
€1,000 × 1.03 = €1,030
B · Convert, invest and hedge
Exchange €1,000 for $1,100, invest at 5%, and sell the maturity dollars forward.
$1,100 × 1.05 = $1,155
At the full-precision parity forward, F = 1.1000 × 1.05 / 1.03 = 1.12135922… USD per EUR. Selling the $1,155 forward produces $1,155 ÷ F = €1,030—the same return as the euro investment. The displayed 1.1214 is a rounded quotation; dividing by that rounded number gives about €1,029.96 instead.
Because USD interest is higher than EUR interest, EUR/USD is above spot in the forward market. That is a euro forward premium in USD-per-EUR terms, offsetting the dollar interest advantage under these simplifying assumptions.
Why actual forwards may differ · sources and assumptions
The example assumes equal-risk investments, matched maturities, frictionless borrowing and lending, annual compounding, and no bid–ask spread. In live markets, cross-currency basis can reflect funding demand, balance-sheet costs and limits to arbitrage. Day-count conventions, discount factors and transaction costs also matter; the observed executable forward is not necessarily the textbook CIP price.
For further reading: BIS: Covered interest parity lost and CME Group: Forward FX pricing and basis.
Forward-rate lab · Interactive calculator
A forward price. Three possible outcomes.
Price a one-year EUR/USD forward, then test the euro cost of a future dollar payment. The forward fixes the exchange rate; it doesn't forecast future spot.
1 Pricing assumptions
Comparable one-year rates; simplified parity pricing.
Spot 1.1000 · USD 5% · EUR 3%
Theoretical one-year forward
1.1214 USD/EUR
Euro forward premium: 1.94% versus spot.
2 Future payment scenario
Try a preset, or enter any hypothetical future rate.
Euro cost of the USD payment
- Hedged · forward
- €89,177.49
- Unhedged · future spot
- €90,909.09
The hedge costs €1,731.60 less than waiting to convert at the selected spot rate.
Show the calculation and forward premium
Covered interest parity: F = S × (1 + iUSD) / (1 + iEUR) for the simplified one-year case.
1.1000 × (1.0500 ÷ 1.0300) = 1.121359… ≈ 1.1214 USD/EUR.
Hedged euro cost: USD payment ÷ full-precision forward. Unhedged euro cost: USD payment ÷ hypothetical future spot. Calculations use the unrounded rate; amounts are displayed to the nearest cent.
Assumptions, risks and what this does not predict
All rates, payments and future scenarios are hypothetical. The forecast does not assign probabilities or claim that any spot outcome is likely. Hedging reduces exchange-rate uncertainty; it does not guarantee the lowest eventual cost.
Real FX forwards reflect bid–offer spreads, timing and day-count conventions, credit and funding conditions, and potentially a cross-currency basis. The simplified parity result is not an executable quote.
BIS — Covered interest parity and the cross-currency basis (2016)
How Long-Maturity Interest Rates Enter the Forecast
So how could the same ingredients produce a useful forecast?
In their 2024 study, Zsolt Darvas and Zoltán Schepp calculate synthetic forward exchange rates using the spot rate and three-, five- and ten-year interest rates. These calculated rates become inputs to an error-correction model.
The idea is that an exchange rate may gradually adjust when it moves away from its underlying economic value. Under the authors’ theoretical assumptions, a long-maturity synthetic forward rate can help measure that gap. They estimate its relationship with the following month’s currency movement, then repeatedly apply the model to generate longer forecasts.
A ten-year interest rate can therefore help generate a one-month forecast. The maturity describes the input; the forecast horizon describes how far ahead the model predicts.
The published analysis covers three currency pairs linking the US dollar, British pound and German mark, with the euro extending the mark series from 1999. The historical forecast evaluation runs from January 1990 to January 2023, comparing the model against the no-change benchmark using observations beyond each estimation window.
That puts the question where it belongs: did this approach actually reduce forecast errors—and by how much?
How Much Better Were the Forecasts?
The gains were modest at one month and much larger over several years. For the paper’s German mark/euro–dollar series, the combined model reduced mean squared forecast error by approximately 0.9% at one month, 11.6% at one year and 42.4% at five years, compared with the no-change benchmark.
The chart expresses the benchmark’s error as 100 at every horizon. A model value of 88.38 therefore means its mean squared forecast error was 11.62% lower. Look for how the gap grows as the forecast horizon lengthens.
Mean squared forecast error, or MSFE, squares each prediction error before averaging the results. That gives large misses extra weight. It measures the size of forecast errors; directional accuracy and investment returns require separate evaluation.
Now switch the chart to “Across historical periods.” For three-year forecasts evaluated during 2007–2010, the model’s error index reached 310.51—more than three times the benchmark. The strong overall performance included a substantial failure around the global financial crisis.
That matters when a forecast informs a real decision. A business facing a foreign-currency payment still needs to withstand an adverse move during a period when its model struggles.
The results give long-maturity interest rates a credible role in exchange-rate forecasting. The next question is whether that improved accuracy translates into useful trading returns after costs.
Research evidence · Exchange-rate forecasting
Forecast accuracy—and a crisis exception
Compare a combined FX forecasting model with a no-change forecast. Lower MSFE is better; 100 means equal forecast error.
MSFE than no-change at five years
The combined model’s MSFE index fell from 99.14 at one month to 57.58 at five years. The improvement is horizon-dependent.
- 1 month99.14
- 3 months97.75
- 6 months95.32
- 1 year88.38
- 2 years79.31
- 3 years69.69
- 4 years63.35
- 5 years57.58
Eight horizons. Index 100 means the model and no-change forecast have equal mean squared error.
MSFE than no-change in 2007–10
The model beat the benchmark in three subperiods—but the 2007–2010 error index jumped to 310.51. These periods all use a three-year horizon.
- 1993–9944.10
- 2000–0646.19
- 2007–10310.51
- 2011–2356.24
The historical-period chart uses a different horizontal scale (0–350) to show the large crisis-period result.
Method, data and interpretation
Source: Darvas and Schepp (2024), Table 6, combined-model MSFE results for DEM/USD. The authors used the German mark through 1998 and a euro-derived mark-equivalent series thereafter. The combined forecast gives equal weight to models using synthetic forward rates of 3-, 5- and 10-year maturity.
MSFE is mean squared forecast error, shown here relative to a driftless random-walk/no-change forecast indexed to 100. Values below 100 indicate lower squared forecast error; values above 100 indicate higher error. The first view covers 1990–2023; the historical subperiods in the second view all assess 36-month-ahead forecasts. The paper uses log exchange-rate data.
These results describe a particular currency series, model and evaluation design. A 42.4% reduction in MSFE is not a 42.4% increase in directional accuracy or a trading return. The authors also caution that better forecast errors need not imply better trading returns.
Does Better Forecasting Mean Profitable Trading?
Profitability depends on the price you trade at, the eventual exchange rate and your costs. Predicting the direction is only part of the calculation.
Consider a hypothetical EUR/USD trade. Spot starts at 1.10, and you agree to buy €100,000 forward at 1.12. At settlement, spot reaches 1.11.
The euro appreciated, as you expected. However, your contract requires paying $112,000 for euros now worth $111,000: a $1,000 loss before additional costs. For this forward purchase, the exchange rate needed to exceed 1.12 to generate a profit.
Darvas and Schepp therefore evaluate trading strategies separately from forecast accuracy. Their 2024 study reports positive excess returns—returns above the risk-free rate—with stronger evidence of an advantage over carry trades at longer horizons. Those simulations exclude transaction costs and assume no leverage.
A 2025 follow-up extends the approach to daily GBP/USD data, evaluating forecasts from January 1990 to October 2024. It also reports positive model-based excess returns before costs and adds an estimate of how much transaction costs could absorb. For the illustrated daily strategy, the estimated break-even cost was 16.8 basis points, or 0.168%, per currency switch, assuming costs arose only when switching currency positions.
That assumption matters. Applying the research through a particular broker or instrument requires checking executable prices, spreads, commissions, funding or rollover charges, and collateral requirements. A cost allowance calculated for one trading setup needs reassessment for another.
The daily study adds useful evidence at shorter horizons. Establishing an intraday trading edge would require testing intraday signals and execution directly.
For a business, meanwhile, the useful outcome may be a more stable payment budget or a better-informed hedge. The practical test is whether the forecast improves the decision you actually face, after its costs and risks.
How to Judge an Exchange-Rate Forecast
Before using a forecast to price a contract, adjust a hedge or place a trade, ask five questions.
What exactly is being predicted? Identify the currency pair, quotation, forecast date and horizon. Check whether the target is an exchange rate on a particular day or an average over a month or quarter.
What does it beat? Look for comparison with an appropriate no-change forecast. Benchmark construction matters: RBA research shows that carrying forward the previous month’s average can make forecasting performance look stronger than comparison with the latest observed rate.
Could it have been produced at the time? Each historical forecast should use information available before its target date. Economic releases arrive with delays and may be revised. Federal Reserve research demonstrates that data revisions and sample selection can materially change measured forecasting performance.
Where does it struggle? Examine results by currency, horizon and historical period. Ask how large the improvement is, how uncertain it is, and whether a strong average conceals prolonged failures.
Does it improve your actual decision? For trading, evaluate returns after realistic costs and alongside losses and drawdowns. For a business, assess the effect on cash flow, margins and budget certainty.
Keep a dated record of forecasts and subsequent outcomes. Alongside the central forecast, calculate what an adverse exchange-rate move would cost. That turns a prediction into something you can evaluate—and gives you a plan when the outcome differs.
Can Interest Rates Help Forecast Exchange Rates?
Interest rates can provide useful forecasting information when incorporated into a tested model. Darvas and Schepp’s research offers evidence that long-maturity rates, combined with spot prices in an error-correction framework, improved historical forecasts—especially at longer horizons. Their results depend on how that information is modelled.
Keep three questions separate: what price can you lock in through a forward, what exchange rate does your model predict, and what decision does that prediction support?
Return to the $100,000 invoice from the introduction. At EUR/USD 1.10, it costs approximately €90,909. At parity, it costs €100,000. A forecast helps you assess that exposure; a hedge can establish the exchange rate for the payment.
The useful outcome is a decision that fits your cash flow, costs and capacity to absorb an adverse move—even when the forecast misses.
FINANSIFIED · Knowledge check
Can you evaluate a currency forecast?
Eight questions on forward pricing, forecast testing and interpreting research. Check each answer, then see what you have mastered.
Review all eight answers
Finansified · Evidence base
References & further reading
The papers, central-bank research and primary guidance behind our forecasting models, forward-rate examples and currency hedging discussion.
01Main research2 sources
- 01
Darvas, Z., & Schepp, Z. (2024).
Exchange rates and fundamentals: Forecasting with long maturity forward rates ↗Journal of International Money and Finance, 143, 103067.
The main study behind the model and forecast-error chart discussed in this article.
Free full text ↗ - 02
Darvas, Z., & Schepp, Z. (2025).
Forecasting the daily exchange rate of the UK pound sterling against the US dollar ↗Finance Research Letters, 71, 106451.
The daily GBP/USD extension, including trading simulations and break-even transaction-cost estimates.
Free full text ↗
02Forecasting methods and evaluation4 sources
- 03
Nason, J. M., & Rogers, J. H. (2008).
Exchange Rates and Fundamentals: A Generalization ↗International Finance Discussion Paper No. 948 · Federal Reserve Board.
Explains why economic fundamentals can influence exchange rates while forecasting remains difficult.
- 04
Wright, J. H. (2003).
Bayesian Model Averaging and Exchange Rate Forecasts ↗International Finance Discussion Paper No. 779 · Federal Reserve Board.
Research on combining forecasts from different models.
- 05
Faust, J., Rogers, J. H., & Wright, J. H. (2001).
Exchange Rate Forecasting: The Errors We’ve Really Made ↗International Finance Discussion Paper No. 714 · Federal Reserve Board.
Examines how data revisions and sample selection affect measured forecasting performance.
- 06
McCarthy, M., & Snudden, S. (2025).
Forecasts of Period-average Exchange Rates: Insights from Real-time Daily Data ↗Research Discussion Paper 2025-09 · Reserve Bank of Australia.
Explains why forecast targets, data frequency and benchmark construction matter.
03Interest rates and forward pricing4 sources
- 07
Borio, C., McCauley, R., McGuire, P., & Sushko, V. (2016).
Covered interest parity lost: understanding the cross-currency basis ↗BIS Quarterly Review · September · pp. 45–64.
Covers the pricing relationship and why actual market quotes can depart from simplified parity calculations.
- 08
Stavrakeva, V., & Tang, J. (2015).
Exchange Rates and Monetary Policy ↗Working Paper No. 15-16 · Federal Reserve Bank of Boston.
Examines monetary-policy surprises, expected interest-rate paths and currency movements.
- 09
CFA Institute (2026).
Currency Exchange Rates: Understanding Equilibrium Value ↗CFA Program Level II Economics · Refresher reading.
Background on parity conditions, forward rates, currency valuation and carry trades.
- 10
Federal Reserve Bank of St. Louis (February 19, 2026).
The link between interest rates and exchange rates ↗FRED Blog.
An accessible explanation of immediate exchange-rate adjustments and expected subsequent movements.
04Business exposure and hedging2 sources
- 11
International Trade Administration.
Foreign Exchange Risk ↗U.S. Department of Commerce.
Practical background on currency exposure in international trade.
- 12
HM Revenue & Customs.
Understanding corporate finance: derivatives: hedging foreign exchange risk ↗Corporate Finance Manual · CFM13400.
A worked explanation of how forward contracts establish payment certainty.



