Intermarket Analysis in Forex: Which Correlations Hold Up

Intermarket analysis is usually presented as a table. Gold up, dollar down. Oil up, Canadian dollar up. Yields up, currency up. The tables are easy to memorise and that is most of their appeal.

The difficulty is that a table states relationships as though they were permanent. They are not. Each one is a measurement taken over a period, and it holds only while the conditions that produced it hold. A shift into stagflation is a change of those conditions rather than an observation inside them. Relationships between markets are often explained by reference to the economic cycle, though the phase itself is identified only long after the fact.

The most widely published cheat sheet on this topic includes the claim that stocks and bonds are positively correlated. That relationship has reversed twice in thirty years, and the reversal is documented by the Bank for International Settlements.

What Intermarket Analysis Actually Claims

The claim is narrower than the tables suggest. Markets are not independent, because the same conditions affect several of them at once.

An interest rate decision moves bond yields, changes what equities are worth, and alters the attractiveness of holding one currency against another. The relationships appear because a shared cause is acting on all of them. One such shared cause is quantitative easing and tightening, which acts on bond yields and currencies together.

That framing already contains the limitation. A correlation between two markets is evidence that something has been acting on both, not evidence that one causes the other.

When the shared cause changes, the observed relationship can change with it, and nothing about the two markets themselves has to change for that to happen.

This page assumes the mechanics of measuring a correlation are already familiar. Our page on how correlation is measured covers the coefficient itself and how it behaves between currency pairs.

Why Every One of These Relationships Is Conditional

There is a precondition that almost no page on this topic states, and without it none of the numbers can be checked.

A correlation is computed over a chosen window. Change the window and the answer changes, sometimes in sign.

Two markets can be strongly positively correlated over the last 20 trading days and negatively correlated over the last 200, because different influences dominated the two periods. Both figures are correct.

A correlation quoted without its window is therefore unfalsifiable. It cannot be verified, and it cannot be contradicted, which is what makes the memorised table a poor tool.

The practical rule is to treat the period as part of the number. A relationship measured over a window that does not match how long positions are held is describing a different problem.

Bond Yields and Currencies: The Mechanism, and When It Inverts

The standard account is that higher yields attract capital, so a currency strengthens as its government bond yields rise. The mechanism is real, and it is incomplete.

What matters is the yield relative to other countries rather than its level, which is the same quantity that drives the interest rate differential behind carry positions.

The reason the yield moved matters just as much. A yield rising because growth is expected to be stronger is a different signal from one rising because inflation is expected to be higher, and from one rising because investors are demanding more compensation to hold that government’s debt.

The third case can move the currency in the opposite direction to the textbook relationship. Rising yields alongside a weakening currency is the recognised signature of it.

Expectations rather than announcements are what the market prices, which is why a change in tone can move a currency without any rate changing. Our page on central bank policy stance covers that separately.

Commodity Currencies and What the Production Data Supports

Commodity-currency claims are the easiest part of this topic to check, because they rest on production figures that are published by official agencies.

They are also where the widely copied tables are weakest. One frequently cited cheat sheet supports its gold-related currency rules by asserting that Australia is the third largest gold producer, that Canada is the fifth, and that New Zealand is a large producer, without citing a source for any of them.

The United States Geological Survey publishes the figures annually. Its 2026 Mineral Commodity Summaries give estimated 2025 gold mine production in metric tons as follows.

Australia at third is right. Canada at fifth is wrong; it is fourth, ahead of the United States. New Zealand does not appear among the named producing countries at all, so the premise behind the New Zealand dollar rule is not supported by the source that would establish it.

The correction matters more than the ranking. If the reasoning is that a currency follows a commodity because the country exports it, then the export position is the load-bearing fact and it has to be checked rather than assumed.

The same discipline applies to gold as a reserve asset. The Swiss National Bank publishes its holding as 1,040 tonnes, unchanged for several years, split roughly 70 per cent in Switzerland, 20 per cent at the Bank of England and 10 per cent at the Bank of Canada.

It publishes no figure for gold as a share of its reserves. The commonly repeated claim that around a quarter of Swiss reserves are gold therefore does not appear here, because it could not be confirmed at the source.

Risk-On and Risk-Off, and Why the Labels Move

Risk-on and risk-off describe a pattern in which some currencies are bought when investors are confident and others when they are not.

The pattern is real and observable. What is unstable is which currency belongs in which group.

Safe-haven status is not a permanent property. It reflects a currency’s role at a point in time, and that role depends on the country’s rates, its external position and what the alternatives are.

Currencies have moved between the two groups as those conditions changed. Treating a list as fixed is the same error as treating a correlation table as fixed, and for the same reason.

Positioning data offers one check on whether a pattern is currently in place, though with a lag, which our page on positioning data explains.

The measure most often cited alongside these labels is the VIX volatility index, which reads what S&P 500 options are priced for. It is an equity measure rather than a currency one, and it carries the same conditionality as the rest of this page.

The Stock and Bond Relationship That Flipped

This is the clearest documented example, and it is the one the memorised tables get wrong.

The BIS Quarterly Review of December 2023 carried a study of the correlation of equity and bond returns by Marco Jacopo Lombardi and Vladyslav Sushko. It records that the correlation between US equity and government bond returns switched sign in mid-2021, from negative to positive.

It also notes an earlier switch in the opposite direction in the late 1990s, from positive to negative. So the relationship has held both signs within the span of a single career.

The explanation given is mechanical rather than statistical. When inflation is high and volatile, the inflation outlook is what shapes the expected path of policy rates, so an inflation surprise pushes bonds and equities the same way at once.

The study reports that the coefficient on inflation surprises turned positive and statistically significant in mid-2021, while growth-related factors ceased to be significant. The driver changed, and the correlation followed.

The lesson generalises past this one pair. Any table asserting that two markets are correlated is reporting which driver was dominant when it was compiled, whether or not it says so.

How to Check Whether a Relationship Is Currently Holding

The useful question is not which markets are correlated in general. It is whether a specific relationship is holding now, and that is checkable rather than a matter of belief.

Three things make it checkable. State the window being measured, so the figure means something. Compare a recent window against a longer one, since a difference between them is the signal that conditions have shifted. Identify which driver the relationship depends on, because that is what will break it.

The third step is the one usually skipped, and it is the one that carries the information. A correlation that weakens tells you something changed; knowing what the relationship rested on tells you what.

Scheduled releases are where these shifts most often become visible, which is one use for an economic calendar beyond tracking individual events.

A relationship that has broken is not necessarily gone. It may be dormant while a different driver dominates, and the honest position is that its current state is unknown rather than that it has been disproved.

What This Can and Cannot Be Used For

Intermarket analysis describes context. It says what environment a currency is trading in and what else is moving alongside it.

It does not produce entries, exits or timing, and nothing above should be read as a signal. A relationship that held historically is a description of the past, not a forecast.

The failure mode is specific and worth naming. Because these relationships are usually stated as rules, they are easy to act on without checking, and a rule that has quietly stopped holding produces confident decisions built on nothing.

Anyone wanting a list of pairings to apply directly is better served by not using this at all. The value is entirely in the conditionality, and stripping that out leaves a table that is wrong an unknown proportion of the time.

Used properly it is a check on context rather than a source of trades, in the same way that leading and lagging indicators describe what a measurement can and cannot tell you.

Frequently Asked Questions

What is intermarket analysis in forex?

Intermarket analysis reads a currency in the context of other markets rather than on its own chart, on the reasoning that bond yields, equity prices and commodity prices reflect the same conditions that move exchange rates. It describes context rather than producing entries, and every relationship it uses is conditional on the environment that created it.

Do bond yields drive currency prices?

The usual account is that a rising yield attracts capital and supports the currency, and that mechanism is real. It is also conditional, because what matters is the yield relative to other countries and why it moved. A yield rising because inflation is expected to be higher is not the same signal as one rising because growth is expected to be stronger.

Is gold always inversely correlated to the US dollar?

No, and treating it as a constant is the common error. The relationship is frequently negative because gold is priced in dollars, but it has broken down repeatedly, particularly when both are bought at once during periods of stress. A relationship that holds most of the time is not the same as one that holds by definition.

Why do intermarket correlations break down?

Because they are outcomes rather than rules. A correlation reflects what has been driving both markets over the measured period, so when the dominant driver changes the correlation can weaken, disappear or reverse. The equity and bond correlation is the clearest documented case, and the BIS attributes its reversal to inflation replacing growth as the dominant influence on policy expectations.

What timeframe should I measure a correlation over?

There is no single correct window, which is exactly why the window has to be stated. The same two markets can show one correlation over 20 trading days and the opposite over 200, so a correlation quoted without a period cannot be checked or contradicted. Choose a window that matches how long positions are held and report it alongside the number.