Intermarket analysis compares currency prices with specified bond, commodity and equity series. It can help formulate a hypothesis about a shared economic change, but simultaneous moves do not establish cause or a profitable trade. This lesson separates basket arithmetic, yield differences and currency conversion from empirical relationships that depend on the data and observation window.

What intermarket analysis actually is

Compare clearly identified markets rather than assume that they always move together. Interest-rate expectations, inflation, financing conditions and demand can affect several prices, while each instrument also has its own terms and participants. A hypothesis about those connections needs matched observations and a test; a chart alone cannot identify the motive behind a transaction.

For a currency trader, the practical question is whether another series adds information beyond the pair itself. A rise in US nominal yields does not by itself classify an EUR/USD uptrend as favourable or unfavourable: compare the maturity, real or nominal measure and changes elsewhere. This provides questions for the fundamentals of the currency market and technical analysis, rather than an automatic confirmation.

Which links matter most for forex?

DXY/USDX. ICE describes a geometrically weighted basket of six currencies, with a 57.6% euro weight. The index and EUR/USD therefore share an input and are not fully independent confirmations. With other components held constant, a higher EUR/USD reduces the index, but the basket is not every USD pair. Distinguish the cash index from a futures contract or a provider’s CFD.

Oil and CAD. Canada’s resource exports provide one reason to investigate commodity prices alongside the USD/CAD pair. The Bank of Canada also identifies relative interest rates, inflation and demand for financial assets. More expensive oil does not guarantee a stronger CAD. Identify the oil contract and any rollover adjustment; remember that CAD strengthening against USD means USD/CAD falls.

Gold and yields. First specify gold’s quotation currency and whether the yield is nominal or real. The US Treasury publishes separate nominal and TIPS real par curves. The Federal Reserve’s fitted same-maturity nominal-minus-TIPS spread measures inflation compensation, which can also contain risk premiums; subtracting a current CPI reading from an arbitrary yield is not the same series. Neither a higher yield nor the label “safe haven” fixes gold’s next move.

A currency-conversion check. These are hypothetical prices, without product costs. Gold rises from USD 2,000 to USD 2,020 per ounce, or 1%, while USD/PLN falls from 4.00 to 3.92 PLN per USD, or 2%. The converted price changes from PLN 8,000 to PLN 7,918.40 per ounce: 1.01 × 0.98 − 1 = −1.02%. Adding the two percentages would give the wrong answer. A PLN gold series constructed this way shares its FX input with USD/PLN; it is not an independent observation.

Apparent market divergence from delayed data

Before calling a relationship a correlation change, check whether observations share a timestamp. This model deliberately delays a copy of an identical series.

Scroll sideways to see the full diagram

Apparent market divergence from delayed dataThe current series at t0–t3 is 100, 101, 103, 102. A copy delayed one step is 100, 100, 101, 103. At the final step, the first falls 1 point while the delayed copy rises 2. This comes solely from timing, not different instrument behaviour.One series, two timestamp sets100101102103t0t1t2t3Current: 100, 101, 103, 102Delayed: 100, 100, 101, 103Final current change = −1Delayed change = +2Align observation times first.

Hypothetical model: arbitrary index points, equal time steps; the second copy repeats the first value at t0. This is neither a correlation measurement nor real market quotes. FRED documentation distinguishes observation dates from data-availability periods; this counterexample is an illustrative model.

Opposite final moves may reflect observations from different times.

Why are these correlations unstable rather than fixed?

A correlation coefficient describes a selected sample and transformation. State whether you compare price levels, percentage returns or yield changes in basis points, and give the frequency, window and missing-data treatment. A monthly sample of daily returns need not resemble five sessions of hourly returns. Pearson correlation measures a linear relationship; zero does not rule out a nonlinear relationship, and a high value does not identify cause or a leading signal.

“In a crisis everything correlates to one” is not a mathematical law. Specify which assets, quotation directions and sample are being compared before discussing a change in dependence. A past coefficient cannot guarantee future diversification, and a new coefficient alone is not a trading rule. The separate article on currency pair correlations in practice develops that portfolio question.

Intermarket relationships provide context and can change with the policy and risk regime.

How does a trader use this as context, not as a signal?

A hypothetical yield comparison. Over the same observation period, a US ten-year nominal yield moves from 4.00% to 4.15%, an increase of 15 basis points. A comparable German ten-year yield moves from 2.50% to 2.80%, up 30 basis points. The US-minus-German spread narrows from 1.50 to 1.35 percentage points, or from 150 to 135 basis points, despite the US yield rising. Suppose gold also moves from USD 2,000 to USD 2,020 and USD/CAD from 1.3500 to 1.3365: those changes are +1% and −1%. These two observations do not estimate a stable correlation or prove that one move caused another.

The hypothetical alignment provides a trading thesis, not a measured probability increase. Define an entry, invalidation level and position size independently, then test whether the proposed contextual filter adds value after costs. The same indicators can behave differently in another regime.

What to do tomorrow

  1. Define the series. Record the instrument, source, quotation currency and units. A cash index, futures contract and CFD may have different price histories.
  2. Align the information. Match observation times and check when each value was actually available. Today’s FX quote beside yesterday’s equity close is not a simultaneous comparison.
  3. Specify the calculation. Record the frequency, return or yield-change transformation, sample window and missing observations. Compare a later sample without changing the method to fit the result.
  4. Test a proposed filter. Define which observations would include or reject a trade, then compare both groups under the same entry, exit and cost rules. “The data do not provide useful context” is a valid result.

For further reading, the intermarket analysis section on forexmechanics.com discusses this layer of analysis. Keep the distinction between a documented construction, such as a basket or currency conversion, and an empirical relationship that requires a specified sample and test.