FX Mathematics, Arbitrage and Forecasting Analytics

Introduction and Context

Foreign-exchange decisions depend on numerical precision. A treasury manager must distinguish a spot quotation from a forward quotation, determine whether a forward price contains a premium or discount, test whether apparently inconsistent market prices create arbitrage, and assess whether a currency forecast is sufficiently reliable for a hedging or investment decision.

This chapter develops a unified framework for these tasks. It treats exchange rates as prices, emphasizes quotation conventions, derives covered-interest and cross-rate relationships, and introduces fundamental, technical, econometric and scenario-based forecasting. The central principle is that an FX calculation is meaningful only when the quotation direction, time period, compounding convention and transaction costs are explicit.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Interpret direct and indirect exchange-rate quotations.
2. Calculate spot, outright forward, swap-point and forward-premium or discount measures.
3. Use cross rates and identify triangular, covered-interest and locational arbitrage.
4. Explain the assumptions underlying interest-rate parity and purchasing-power relationships.
5. Apply fundamental, technical, econometric and scenario-based forecasting methods.
6. Quantify forecast uncertainty and distinguish a point forecast from a decision range.
7. Evaluate the governance, model-risk and practical limitations of FX analytics.

1. Exchange-Rate Conventions and Core Mathematics

1.1 Quotations and units

A quotation must specify the price currency and the base currency. For example, EUR/USD = 1.0850 means that one euro costs 1.0850 US dollars. The euro is the base currency and the US dollar is the quote currency. The unit is USD per EUR.

A direct quotation expresses domestic currency per unit of foreign currency. An indirect quotation expresses foreign currency per unit of domestic currency. If EUR/USD = 1.0850, then USD/EUR is:

Bid and ask prices must also be interpreted correctly. A dealer buys the base currency at the bid and sells it at the ask. Thus a dealer quotation EUR/USD 1.0848–1.0852 gives the dealer’s bid and ask, respectively. A customer buying euros pays 1.0852 dollars per euro.

1.2 Cross rates

A cross rate is inferred from two quoted rates that share a currency. If EUR/USD = 1.0800 and GBP/USD = 1.2600, then:

The result means that one euro equals approximately £0.8571. In practice, bid and ask rates must be combined conservatively: the executable cross-rate bid uses prices that make a dealer’s purchase least favourable, while the executable ask uses prices that make its sale least favourable.

1.3 Spot, forward and swap rates

The spot rate, S₀, is the exchange rate for near-immediate delivery. A forward rate, F₀,T, is agreed today for delivery at time T. A currency swap, in the dealing sense, combines a purchase and sale of the same currency for two different value dates. The difference between the forward and spot exchange rates is often quoted in points or pips.

If EUR/USD spot is 1.0850 and the three-month forward is 1.0905, the forward points are:

For a four-decimal quotation, this is 55 points. A positive difference is commonly called forward points at a premium; a negative difference is called a discount. The terms premium and discount here describe the forward price relative to spot, not necessarily an expected appreciation or depreciation.

The annualized forward premium on the base currency is:

where d is the number of calendar days to maturity. Using d = 90:

The 360-day convention is a market convention. A 365-day convention may instead be required by the contract or institution. The convention must not be changed halfway through a comparison.

2. Interest-Rate Parity and Forward Pricing

2.1 Covered interest parity

Covered interest parity (CIP) states that, absent transaction costs, capital controls and credit differences, the forward exchange rate eliminates an arbitrage opportunity between investing in two currencies while covering exchange-rate risk.

Let S₀ be the domestic-currency price of one unit of foreign currency, F₀,T the forward rate for T years, i_d the domestic interest rate and i_f the foreign interest rate. With simple annual interest:

With annual effective compounding over one year, the corresponding form is:

With continuously compounded rates r_d and r_f:

The formula is dimensionally consistent because rates are expressed per year and T is measured in years. If the domestic interest rate exceeds the foreign rate, the domestic-currency price of foreign currency is normally above spot; the foreign currency trades at a forward premium in domestic-currency terms.

Worked Example 1: Forward price from interest rates

Suppose USD/CHF spot is 0.9000 CHF per USD, the annual CHF interest rate is 3%, the annual USD interest rate is 5%, and maturity is six months. Using simple interest:

The forward rate is approximately CHF 0.8912 per USD. Because the quote is CHF per USD, the currency with the higher interest rate, the USD, is cheaper forward in CHF terms.

2.2 Forward points as a swap quotation

Dealers frequently quote swap points rather than a complete forward rate. If points are quoted as 18/22 and spot is 1.2500, the outright forward bid and ask are 1.2518 and 1.2522 when the points are added. If points are quoted as −18/−14, they are subtracted, producing 1.2482–1.2486. The direction must be confirmed from the market convention and the relative bid–ask structure.

2.3 Non-deliverable forwards

A non-deliverable forward (NDF) settles the difference between the contracted rate and a fixing rate in a convertible currency rather than delivering the restricted currency. The same pricing logic may be used, but fixing source, settlement currency, netting direction and regulatory rules become essential contract terms.

3. Arbitrage in Foreign Exchange

Arbitrage is a set of transactions intended to produce a riskless or near-riskless profit from inconsistent prices. In real markets, arbitrage is constrained by bid–ask spreads, funding costs, settlement risk, credit limits, taxes, capital controls and execution speed.

3.1 Locational arbitrage

Locational arbitrage arises when the same currency pair is quoted at inconsistent prices in two locations. If Bank A sells EUR/USD at 1.1000 and Bank B buys EUR/USD at 1.1010, a trader could buy euros from A and sell them to B, subject to simultaneous execution and costs. The gross profit is 0.0010 USD per euro.

The opportunity disappears when the executable prices converge. Retail or delayed screen prices should not be treated as arbitrage opportunities without checking settlement and transaction costs.

3.2 Triangular arbitrage

Triangular arbitrage uses three currencies when the cross rate implied by two markets differs from the directly quoted third rate. Suppose:

• EUR/USD = 1.1000
• USD/GBP = 0.8000
• EUR/GBP = 0.8700

The implied EUR/GBP rate through USD is:

The direct rate of 0.8700 is inconsistent with the implied rate before spreads. A trader would determine the profitable direction by starting with one currency and tracing both possible conversion cycles. The correct test is not simply whether mid-rates differ; it is whether executable bid and ask rates generate more than the starting amount after completing the cycle.

A convenient logarithmic representation is:

In an efficient market, the sum is approximately zero after allowing for spreads and costs. A non-zero result may indicate mispricing, but it may also reflect asynchronous quotes.

3.3 Covered-interest arbitrage

Covered-interest arbitrage occurs when the observed forward rate differs sufficiently from the CIP-implied forward rate. A trader borrows in one currency, converts at spot, invests in the other currency, and uses a forward contract to reverse the future conversion. The exchange-rate risk is covered.

The direction depends on the comparison:

• If the market forward is too high relative to CIP, borrowing the foreign currency, investing domestically and selling the domestic proceeds forward may be profitable, depending on quotation direction.
• If the market forward is too low, reverse the currencies.

The apparent profit must be calculated using borrowing rates, investment rates, bid–ask spreads and actual settlement dates. Using a central bank policy rate instead of the institution’s funding rate can create a false arbitrage conclusion.

3.4 Arbitrage and market frictions

The law of one price is an analytical benchmark, not a guarantee. Persistent deviations from CIP have been observed during periods of financial stress because of dollar funding scarcity, balance-sheet constraints, counterparty risk and regulatory capital charges. Therefore, a deviation is better described as an arbitrage signal until executable net profit and operational feasibility are established.

4. Currency Forecasting Frameworks

No forecasting model consistently predicts all exchange-rate movements. Forecasting is therefore a model-risk exercise involving assumptions, data quality, validation and uncertainty communication.

4.1 Fundamental analysis

Fundamental analysis links exchange rates to macroeconomic and financial variables, including:

• inflation differentials;
• interest-rate expectations and real yields;
• economic growth and productivity;
• current-account and capital-flow conditions;
• fiscal sustainability and public debt;
• commodity prices for commodity-exporting economies;
• political risk, reserves and external liquidity.

Purchasing power parity (PPP) provides a long-run benchmark. In relative form, if P_d and P_f are domestic and foreign price levels, the domestic-currency price of foreign currency is approximately:

The change form is often written as:

where s is the logarithm of the exchange rate and π_d and π_f are domestic and foreign inflation rates. PPP is a long-run tendency, not a reliable short-run trading rule, because goods are non-traded, transport costs exist and prices adjust slowly.

The monetary approach emphasizes relative money supplies, income and interest rates. Modern forecasting usually combines such variables with expectations, risk premia and financial-market data rather than relying on a single theoretical equation.

4.2 Technical analysis

Technical analysis studies historical prices, returns and trading activity. Common tools include moving averages, momentum, support and resistance, volatility bands and trend indicators. A k-period moving average is:

where P is the observed price and k is the window length. A moving-average crossover may signal a trend, but it does not establish causality. Technical strategies are vulnerable to data snooping, transaction costs, regime changes and false discoveries. Out-of-sample testing and realistic execution assumptions are essential.

4.3 Econometric models

A simple autoregressive model for the exchange-rate change is:

where α is an intercept, β measures lag dependence, X_t is a vector of explanatory variables, γ is a coefficient vector and ε_t is an unpredictable error term. The model may be estimated by ordinary least squares under suitable assumptions, but residual autocorrelation, heteroskedasticity, structural breaks and endogeneity must be investigated.

A vector autoregression (VAR) models several jointly evolving variables. An error-correction model can represent short-run deviations from a long-run equilibrium such as PPP or an uncovered-interest relationship. Forecast evaluation should use rolling or expanding windows, not only in-sample fit.

Forecast accuracy can be measured by mean absolute error:

and root mean squared error:

where s_t is the realized value, ŝ_t is the forecast and n is the number of observations. MAE is easier to interpret in the original unit; RMSE penalizes large errors more heavily. Comparisons should use the same horizon, sample, quotation and loss function.

4.4 Volatility and uncertainty

A point forecast is incomplete without an uncertainty assessment. If a forecast error is approximately normal with standard deviation σ̂, an approximate 95% interval is:

This interval is only appropriate when the estimated error distribution, parameter uncertainty and model specification support it. FX returns often exhibit volatility clustering and fat tails, so historical or simulated quantiles may be more suitable than normal intervals.

A conditional volatility model such as GARCH(1,1) is commonly expressed as:

where σ_t² is conditional variance, ε_{t−1} is the previous innovation, and ω, α and β are parameters. Typically ω > 0, α ≥ 0 and β ≥ 0; covariance stationarity commonly requires α + β < 1. These are model conditions, not universal truths.

4.5 Scenario analysis

Scenario analysis is particularly useful when probabilities are unreliable or political and policy regimes may change. A treasury team may construct baseline, appreciation and depreciation scenarios, each with assumptions about inflation, interest rates, commodity prices and capital flows. Expected exposure cost can be calculated as:

where p_i is the assigned probability of scenario i, C_i is its cost and ∑p_i = 1. Scenario probabilities should not be presented as objective facts; they are decision assumptions and should be stress-tested.

5. Applied Case: Hedging a Foreign-Currency Payable

Background

This illustrative case concerns Meridian Components, a firm that must pay EUR 10 million in 90 days. Its reporting currency is USD. The firm is considering whether to lock the dollar cost or retain some flexibility.

Problem

The current EUR/USD spot rate is 1.0850. The 90-day forward rate is 1.0905. The treasury department’s baseline forecast is EUR/USD 1.0800 at maturity, but its scenario range is 1.0200 to 1.1600. The firm must compare the certainty of a forward hedge with the uncertainty of remaining unhedged.

Analysis

The forward hedge locks the dollar payment at:

If the spot rate at maturity is 1.0800, an unhedged payment would cost USD 10,800,000, which is USD 105,000 less than the forward cost. If the spot rate is 1.1600, the unhedged payment would cost USD 11,600,000, which is USD 695,000 more than the forward cost. If the spot rate is 1.0200, the unhedged payment would cost USD 10,200,000, or USD 705,000 less than the forward cost.

The forward therefore exchanges upside from a favourable currency movement for protection against an adverse movement. The relevant decision is not whether the forecast is correct, but whether the firm values budget certainty, has risk capacity and faces cash-flow or covenant consequences from adverse outcomes.

Approach and solution

A robust policy could hedge a high proportion of committed exposures with forwards, retain a smaller unhedged portion, and use options when participation in favourable movements is strategically important. The treasury team should document the exposure date, currency, notional, hedge ratio, counterparty limit, accounting treatment and forecast source.

Outcome and learning points

The case demonstrates four lessons. First, a forward rate is a contractual price, not a prediction. Second, forecasts should be translated into cash-flow distributions or scenarios. Third, the hedge decision depends on risk appetite and operating consequences, not forecast accuracy alone. Fourth, bid–ask spreads, credit valuation adjustments, collateral and settlement risk can alter the economic result.

6. Professional Practice, Governance and Technology

A sound FX analytics process should:

1. standardize quotation direction and data timestamps;
2. use executable bid and ask prices rather than unverified mid-rates;
3. match day-count, compounding and settlement conventions;
4. separate market data, model assumptions and management judgment;
5. back-test forecasts out of sample and monitor model drift;
6. perform sensitivity, stress and reverse-stress tests;
7. preserve an audit trail of prices, rates, approvals and overrides.

Automation and AI can assist with data cleaning, anomaly detection, forecasting ensembles and scenario generation. However, machine-learning models may overfit, learn market microstructure noise, reproduce biased data and fail during regime changes. Human review remains essential for unusual events, illiquid currencies and policy-driven discontinuities. Model governance should include independent validation, explainability proportional to risk, access controls and documented use limitations.

Regulatory and ethical issues include market-abuse controls, suitability of derivatives, accurate valuation, fair dealing, confidentiality of client information and segregation of duties. A profitable-looking arbitrage strategy that depends on stale prices, confidential information or prohibited market manipulation is neither acceptable nor sustainable.

Challenges and Limitations

FX models face several persistent difficulties. Exchange rates are affected by expectations, so current prices may incorporate information before macroeconomic statistics are released. Relationships such as PPP and CIP can weaken during crises. Data are revised, non-synchronous and sometimes subject to survivorship or look-ahead bias. Forecasts may be statistically significant but economically unprofitable after costs. Finally, uncertainty is often asymmetric: a sudden devaluation may be more damaging than an equally sized appreciation is beneficial.

Accordingly, model outputs should be treated as conditional statements: given these data, assumptions and regime conditions, this is the estimated result. They should not be represented as certainty.

Best-Practice Checklist

Before approving an FX calculation or forecast, ask:

• What is the base currency and what are the units?
• Are bid, ask and mid-market prices being distinguished?
• Are dates, holidays, day-count conventions and compounding consistent?
• Does the calculation include funding, transaction and credit costs?
• Is an apparent arbitrage executable after spreads and constraints?
• Has the forecast been tested out of sample?
• Is uncertainty reported through intervals or scenarios?
• Are assumptions, overrides and model limitations documented?

Chapter Summary

FX mathematics begins with disciplined quotation conventions. Spot rates describe immediate exchange, forward rates specify future delivery, and swap points express the difference between forward and spot prices. Covered interest parity links forward rates to relative interest rates under assumptions of frictionless funding and hedging. Cross rates provide a basis for triangular-arbitrage tests, while locational and covered-interest arbitrage depend on executable prices and funding conditions.

Currency forecasting may draw on fundamentals, technical indicators, econometric models and scenario analysis. No method is universally dominant. Forecast quality must be assessed out of sample, with appropriate error measures and uncertainty intervals. In professional practice, forecasts inform risk decisions but do not replace them. Governance, transaction costs, model risk, regulatory compliance and managerial objectives determine whether an analytical result is usable.

Key Terms

Exchange-rate quotation: The stated price of one currency in units of another.

Base currency: The first currency in a currency-pair quotation.

Spot rate: The exchange rate for near-term delivery.

Forward rate: A rate agreed today for exchange at a future date.

Forward points: The numerical difference between a forward rate and spot rate.

Swap points: Market quotation of forward points for a currency swap or forward.

Cross rate: An exchange rate implied from two other currency quotations.

Bid: The price at which a dealer buys the base currency.

Ask: The price at which a dealer sells the base currency.

Covered interest parity: The no-arbitrage relationship between spot, forward and interest rates.

Locational arbitrage: Profit seeking from inconsistent prices in different locations.

Triangular arbitrage: Profit seeking from inconsistent prices among three currencies.

NDF: A non-deliverable forward settled by a net cash difference.

Purchasing power parity: A theory linking exchange rates to relative price levels.

MAE: Mean absolute forecast error.

RMSE: Root mean squared forecast error.

GARCH: A model of time-varying conditional volatility.

Scenario analysis: Evaluation of outcomes under alternative assumptions.

Model risk: Risk that a model is incorrectly specified, implemented or interpreted.

Review Questions

Short Answer Questions

1. What is the difference between a direct and an indirect quotation?
2. Why must bid and ask rates be used when testing practical arbitrage?
3. Define covered interest parity.
4. What do positive forward points indicate relative to spot?
5. Why is a forecast interval more informative than a point forecast alone?

Long Answer Questions

1. Derive the simple-interest covered-interest-parity forward formula and explain its assumptions.
2. Compare locational, triangular and covered-interest arbitrage, including the main implementation risks.
3. Explain how PPP, interest-rate differentials and capital flows may influence fundamental currency analysis.
4. Evaluate the benefits and limitations of technical and econometric forecasting methods.
5. Discuss how a corporate treasury department should govern FX models and forecast outputs.

Analytical / Application Questions

1. EUR/USD spot is 1.1000, the 90-day forward is 1.1120, the USD interest rate is 5% and the EUR interest rate is 3%. Using simple interest and a 360-day year, calculate the CIP-implied forward rate and determine whether the observed forward is above or below it.
2. A company expects to pay EUR 4 million in six months. Compare the dollar cost of a forward rate of 1.0950 with unhedged costs at maturity rates of 1.0400, 1.0950 and 1.1500. Interpret the risk-management trade-off.
3. Design a validation plan for a currency forecasting model that combines macroeconomic variables, moving-average signals and scenario analysis. Specify data splitting, error metrics, stress tests and governance controls.

Multiple Choice Questions with Answers

Question 1

If EUR/USD = 1.0800, what is USD/EUR approximately?

A. 0.9259

B. 1.0800

C. 2.0800

D. 0.0800

Correct Answer: A. 0.9259

Question 2

Spot EUR/USD is 1.0850 and the 90-day forward is 1.0905. What are the forward points?

A. −55 points

B. 5.5 points

C. 55 points

D. 550 points

Correct Answer: C. 55 points

Question 3

Under simple-interest CIP, if the domestic interest rate is higher than the foreign interest rate, the domestic-currency price of foreign currency will generally be:

A. Below zero

B. Above the spot rate

C. Exactly equal to spot

D. Unrelated to interest rates

Correct Answer: B. Above the spot rate

Question 4

Which activity is an example of triangular arbitrage?

A. Borrowing at a fixed rate only

B. Trading between two dealers in one currency pair

C. Using three currency conversions to exploit an inconsistent cross rate

D. Buying a bond after a rating upgrade

Correct Answer: C. Using three currency conversions to exploit an inconsistent cross rate

Question 5

If EUR/USD = 1.1000 and USD/GBP = 0.8000, the implied EUR/GBP cross rate is:

A. 0.3000

B. 0.8800

C. 1.3750

D. 1.9000

Correct Answer: B. 0.8800

Question 6

Which measure penalizes large forecast errors more heavily?

A. RMSE

B. MAE

C. Median price

D. Forward points

Correct Answer: A. RMSE

Question 7

A positive forward premium means that the forward rate is:

A. Lower than the spot rate

B. Equal to the interest rate

C. Higher than the spot rate in the stated quotation

D. Always an accurate forecast of appreciation

Correct Answer: C. Higher than the spot rate in the stated quotation

Question 8

Which condition is commonly associated with covariance stationarity in a GARCH(1,1) model?

A. α + β < 1

B. α + β > 2

C. ω < 0

D. All parameters must equal zero

Correct Answer: A. α + β < 1

Question 9

A forward contract for a EUR payable primarily provides the company with:

A. A guaranteed exchange rate for the contracted future transaction

B. Unlimited profit from favourable exchange movements

C. Elimination of all credit risk

D. A forecast of the central bank’s policy rate

Correct Answer: A. A guaranteed exchange rate for the contracted future transaction

Question 10

Which is a major limitation of purchasing power parity as a short-run forecasting tool?

A. It uses no exchange-rate concept

B. Non-traded goods, transport costs and slow price adjustment

C. It cannot compare price levels

D. It applies only to interest-rate swaps

Correct Answer: B. Non-traded goods, transport costs and slow price adjustment