Commodity Currencies: How Tightly CAD Tracks Oil and AUD Tracks Iron Ore

A commodity currency is the money of a country whose export income depends heavily on commodities: the Canadian dollar (CAD) and oil, or the Australian dollar (AUD) and iron ore. The link has a real economic reason behind it, but the correlation you see on a chart swings a lot and sometimes turns negative. This article explains the mechanism, runs a synthetic experiment to show why the number is unstable, and ends with the use that actually pays off: not opening the same bet twice without noticing.

What is a commodity currency

The core idea is the terms of trade: the price of what a country exports compared with the price of what it imports. If what the country sells gets more expensive while what it buys costs the same, more money flows in from abroad. Exporters convert that income into local currency, the economy does well, and the market starts to expect higher interest rates. Every channel points to a stronger currency.

Currency	Commodity it is often tied to	Basic fact
CAD (Canada)	Crude oil	Canada is one of the world’s large crude oil producers and exporters, selling mainly to the US
AUD (Australia)	Iron ore, coal, gas	Iron ore is usually the largest export by value, and China is the main buyer
NZD (New Zealand)	Dairy	Dairy is the largest export
NOK (Norway)	Oil and gas	Oil and gas make up a large share of exports

The copper story around AUD is often told wrong. Australia does export copper, but it is not the main source of income. AUD and copper move alike because both reflect global industrial demand, especially from China. For NZD, traders watch the Global Dairy Trade auction, held roughly every two weeks.

Keep the quote direction in mind. A stronger CAD means USDCAD falls. If oil rises and CAD strengthens, oil and USDCAD move in opposite directions. AUDUSD and NZDUSD have the commodity currency in front, so a stronger currency means a rising chart.

What the research says

The paper that gave the topic its name is Chen and Rogoff (2003), “Commodity currencies”, which studied Australia, New Zealand and Canada. It found that the US-dollar price of their commodity exports had a strong and fairly stable effect on the real exchange rates of Australia and New Zealand, while the result for Canada was less robust. A later paper by Chen, Rogoff and Rossi (2010) found something many people don’t expect: these exchange rates helped forecast quarterly commodity price indices better than the other way round.

Both used quarterly data and real (inflation-adjusted) exchange rates. They do not say that if oil rises today, USDCAD falls tomorrow. They say that over long periods, export prices are one force that genuinely exists.

Why the correlation drifts: a two-factor experiment

CAD’s daily return does not come from oil alone. Write the simplest possible version:

CAD = β × oil + rate/dollar factor + noise

β is CAD’s sensitivity to oil. With β = 0.1, a 2% rise in oil pushes CAD up 0.2% on average. In this model the measured correlation is

ρ = (β + γ) × σ_oil ÷ √[ (β + γ)² σ_oil² + σ_rates² + σ_noise² ]

γ is how much the rate/dollar factor itself moves with oil, and σ is daily volatility. The formula says ρ can change while β stays put: the other factor only has to get louder, or start reacting to oil in the opposite direction.

We built 750 synthetic days in node with β = 0.1 fixed throughout, oil moving 2% a day and noise 0.25%, and changed only the second factor.

1. Days 1–250, calm rates. The rate factor moves 0.2%. The formula gives ρ = 0.2 ÷ √(0.04 + 0.04 + 0.0625) = 0.2 ÷ 0.3775 = 0.53
2. Days 251–500, noisy rates, as when the market can’t read the central banks. It moves 0.6%: ρ = 0.2 ÷ √(0.04 + 0.36 + 0.0625) = 0.2 ÷ 0.680 = 0.29
3. Days 501–750, oil lifts the dollar, as when expensive oil makes the market expect US rates to stay high. γ = −0.15, so β + γ = −0.05, with the factor moving 0.4%: ρ = −0.1 ÷ √(0.01 + 0.16 + 0.0625) = −0.1 ÷ 0.482 = −0.21

What the synthetic data produced:

• The whole phases came out at 0.53, 0.25 and −0.13, close to the formula. The gap is noise, since 250 days still carry a rough margin of ±2 ÷ √250 = ±0.13
• The 60-day rolling correlation (counting only windows fully inside one phase) ranged from 0.44 to 0.67 in phase one, −0.01 to 0.46 in phase two and −0.36 to 0.17 in phase three (some windows still positive). The whole line crossed zero 10 times
• In phase three the 60-day value first went negative on day 531, 31 days after the real change, because the window still held old data
• Over all 750 days the figure is 0.20, a single number that describes none of the phases

The economic link (β) never changed. The measured correlation depends on how loud the other forces are, so today’s value is a picture of a period, not a permanent property of the pair.

All of these numbers come from synthetic data, not a real market. The mechanism is similar to the example in Risk-on / risk-off, with one difference: there the link between the two series really changed, while here the link stays the same and only the noise from other forces changes.

Rate differentials are often the bigger force

In real markets the largest second factor is usually the interest-rate differential between the two countries and how it is expected to change. When the Bank of Canada and the Federal Reserve signal different paths, USDCAD can move more than oil explains in a whole month. The mechanism is in Why interest rates move currencies and gold, and the income side is in Carry trade.

Other forces that often drown out commodities:

• The broad dollar. US news makes USD stronger or weaker against everything at once, which you can see in the Dollar Index DXY
• Market sentiment. AUD, NZD and NOK tend to weaken when markets are scared, even if export prices haven’t changed. NOK is less liquid than the majors, so it swings hard at such times
• The price the country actually gets. Much of Canada’s crude is heavy oil, which usually sells below the WTI price on your screen, and that discount widens and narrows over time

To separate a CAD move from a USD move, use a Currency Strength Meter, which measures each currency across many pairs. If CAD is strong against everything, the story is probably Canada or oil. If everything is strong against USD, the story is the dollar.

Does oil really “lead” CAD?

A common belief is that you watch oil move first, then enter USDCAD because CAD will follow. Think about it as a market. If the oil return right now reliably predicted CAD’s move in half an hour, whoever saw it first would buy CAD immediately, CAD would move at once, and the gap would close. Whatever remains, if anything, is usually smaller than the spread and is collected by systems far faster than a retail trader.

Another trap is mismatched closing times. The daily close of an oil contract and the daily close of the FX market are often not at the same moment. We simulated 998 days, each split into 8 sub-periods, in which CAD and oil move together only within the same sub-period, so nothing leads anything. Then we tried two versions: both closed at the same time, and oil closed two sub-periods before CAD.

• With the same close time, the same-day ρ is 0.53 and every other day sits inside the noise band of ±2 ÷ √998 = ±0.06
• With oil closing earlier, the same-day value drops to 0.41 and a 0.12 appears between today’s CAD and tomorrow’s oil. Split the data in half and you get 0.11 and 0.13, stable in both halves

That 0.12 is pure measurement. The last two sub-periods of the day count toward today for CAD but toward tomorrow for oil. The numbers match a hand calculation: an overlap of 6 out of 8 gives 0.53 × 0.75 ≈ 0.40, and 2 out of 8 gives 0.53 × 0.25 ≈ 0.13. If you find a lead-lag in daily data, the first thing to check is the closing time.

How to test it yourself

1. Take oil and USDCAD prices from the same source at the same time, for example the H1 close at the same hour every day
2. Compute daily % changes and flip the sign of USDCAD so it reads as “CAD strength”
3. Calculate the same-day correlation, and the correlation of yesterday’s oil with today’s CAD
4. Compare with ±2 ÷ √n. With 500 days the margin is ±0.09, and anything inside it can’t be told apart from chance
5. If it passes, split the data into two periods and check it holds in both. Remember that ρ = 0.1 explains only ρ² = 1% of the variation, so ask whether anything is left after the spread. If you tried many assets and many lags, read Multiple testing before you believe it

How to use this when trading

The most concrete benefit is not guessing direction. It is counting your risk correctly.

Buying USDCAD profits when CAD weakens. Selling oil profits when oil falls. If oil falls and CAD weakens with it, both trades win or lose together: one bet placed twice. Example: say each trade’s daily profit and loss swings by 100 dollars (one standard deviation). The combined swing is √(100² + 100² + 2 × ρ × 100 × 100), where ρ equals the oil–CAD correlation, because both trades are on the opposite side of it.

ρ in that period	Combined daily swing
1 (exactly the same bet)	200.0 dollars
0.53 (like phase one)	174.9 dollars
0.25 (like phase two)	158.1 dollars
0 (unrelated)	141.4 dollars
−0.13 (like phase three)	131.9 dollars

Working through the 0.53 row: 10,000 + 10,000 + 2 × 0.53 × 10,000 = 30,600, and the square root is 174.9, much closer to 200 than to 141.4. The full formula is in Correlation and portfolio risk.

If two trades tell the same story, treat them as one, and plan the combined risk assuming ρ is close to 1 on the day the market panics.

Rules of thumb:

1. Group your bets. CAD with oil, NOK with oil, AUD and NZD with metals and sentiment: count each as one group, and cap risk per group, for example at 2%, not just per trade
2. Measure before combining. Calculate the correlation of daily returns over the last 20 and 60 days. If the two differ a lot, the relationship is changing, so don’t rely on a number you remember
3. Ask what the main driver is right now. Before trading CAD on oil news, check the economic calendar for Canadian or US rate events nearby. If there is one, oil may not be in charge
4. Watch pairs that mix several forces. AUDJPY carries China, sentiment and Japanese rates at once, so your result may come from a force you never meant to bet on
5. Position size still comes from the Stop loss. None of this changes how you calculate lots. Keep risking 1% per trade, then check whether the group is over its cap

Common mistakes

• Forgetting the quote direction. Oil rises and you buy USDCAD because “CAD looks good”, when that trade is actually a bet on CAD weakening
• Using one correlation from a book or website. That number is an average of one period. In the model above, the overall 0.20 matches none of the phases
• Treating two trades as diversification. Long USDCAD with short oil, or long AUDUSD with long NZDUSD, is usually a single bet
• Believing a lead-lag from daily data without checking closing times. A stable-looking value can be a measurement artefact
• Ignoring rate differentials. In many periods the central banks move the currency more than commodities do over a whole month