Intermarket Analysis and Changing Correlations

Three variables that reprice the instruments you chart

Lists in this subject are not priced in isolation, and the variables that move them are few. The first is the discount rate, which is what the long end of the yield curve sets: a higher risk-free rate lowers the present value of every future cash flow, so it compresses the multiples that stocks trade on. That is why rising yields are usually bad for long-duration equities — the growth names whose cash flows sit furthest in the future — and relatively less bad for banks and value, which is a rotation visible in relative strength before it is visible in any index. The second is the dollar, which reprices the earnings of companies that sell abroad: a stronger dollar reduces the translated value of foreign revenue, so large-cap multinationals feel it in their reported numbers, and commodity producers feel it directly because commodities are priced in dollars and become more expensive to foreign buyers as it strengthens. The third is the commodity complex, which moves with global demand and with the same inflation impulse that moves rates, so oil and copper are simultaneously economic data, inflation data and inputs to corporate margins. The relationships between these three and the instruments on your chart are real, measurable and — this is the whole lesson — not fixed. Their signs depend on which variable is doing the moving. If rates rise because growth is strong, equities can rise with yields; if rates rise because inflation has taken over, equities fall with them. A stronger dollar helps importers and hurts exporters, so the index effect depends on the composition of the index. Higher oil is demand-led strength for a producing economy and a margin tax for a consuming one. In each case the sign of the relationship is a function of the shock, not of the pair, and a trader who memorised the sign from one decade is holding a hypothesis rather than a rule. The practical discipline that falls out of that is a correlation check rather than a correlation assumption. Before leaning on any relationship — bonds hedging equities, the dollar inversing commodities, gold hedging inflation — ask what has been driving both series in the current regime, and how the relationship has actually behaved in the last few months. Correlations are estimable on any charting platform, which means the honest position is that a relationship is a measurement you can make rather than a fact you can recall, and that its stability is itself something to monitor. When the estimated correlation between two positions flips sign, the hedge you paid for is now a second exposure in the same direction, which is the exact risk that 2022 exposed in the most widely held portfolio in the world. What the shock decides — Rates up on strong growth: Equities can rise with yields — the discount rate rose, earnings expectations rose more · Rates up on an inflation shock: Equities and bonds fall together — both discounted by the same variable ← · Dollar stronger: A headwind for exporters and dollar-priced commodities, a tailwind for importers · Oil higher: Cost-push margin pressure and a sector rotation — bullish energy, bearish transport and consumers The list is not a table of constants. Each row is a statement about which shock is driving the move, which is what decides the sign of the relationship rather than the pair of instruments involved.

How to use it without memorising a false constant

There are three defensible uses. The first is as a pre-trade check on the position you are about to take: if you are long a long-duration growth name, you are implicitly short rates, and knowing that changes how you read a move in yields that would otherwise look like noise on a chart. The second is as a source of trade ideas, because divergences between related instruments — a commodity producer failing to follow the commodity, an exporter holding up while the dollar rallies — are the raw material of relative strength setups. The third is as a portfolio risk measurement: the correlation between the things you own is what determines whether you have five positions or one, and it is estimable rather than assumed. The misuse is the confident causal story. Financial commentary is full of explanations of the form “stocks fell because yields rose”, offered after the fact and unfalsifiable in the moment, and the honest version of that sentence is much weaker: yields and equities have been negatively correlated in this regime, and today they moved as that relationship would predict. There is nothing wrong with the observation; there is a great deal wrong with building a position on its persistence without checking. The same caution applies in reverse to the traders who dismiss intermarket work entirely: the relationships are not always stable, but they are usually present, and a chart read without knowing what the discount rate is doing is a chart read with one eye closed. Finally, the discipline of writing it down. The intermarket view belongs in the same place as the pattern and the invalidation: a sentence stating which variable you believe is driving the setup, what the relationship has been doing recently, and what would change your mind — because the failure mode here is silent. A position taken in the belief that bonds hedge equities does not announce itself as wrong when the regime flips; it simply stops working, and the loss arrives in the form of a hedge that cost money in the same week as everything else. • Use it as a check on what you already own: growth exposure is implicit short-rate exposure. • Use divergences between related instruments as a source of relative strength setups. • Measure the correlation rather than assuming it, and treat its stability as something to monitor. • Prefer the weak version of the story — “these have been negatively correlated” — to the confident causal claim. • Write the intermarket sentence into the plan, with what would change your mind. The specific trap is a hedge that has quietly become a second bet. A portfolio holding bonds against equities, or gold against inflation, or an energy position against a consumer position, is protected only while the regime that produced the negative correlation persists. When it flips, the position is twice the size in the same direction — and because the flip is silent, the next loss looks like bad luck rather than a structural change worth noticing.

The window writes the conclusion

A statement like “stocks and bonds are negatively correlated” or “the dollar and gold rise together” is a claim about a particular slice of history, and the slice is doing most of the work. Compute the same pair over daily returns for the last quarter, over monthly returns for the last decade, and over a rolling three-year window, and the three answers can differ in sign. The measurement is not a detail appended to the claim; it is the claim. Two mechanics explain why. First, the horizon. Daily correlation is dominated by shared one-day shocks: when a single piece of news re-prices everything, every pair moves together for a session and the daily number jumps. Monthly correlation, by contrast, reflects shared trends, so it is slower and smoother — and closer to what a holder experiences. Second, the span. A rolling window that is short enough to detect a regime change is short enough to be dominated by noise, and a window long enough to be stable will average several regimes into one meaningless blend. There is no window that is both. The noise is quantifiable, and doing so is the difference between reading a change and reading a coin toss. The sampling error of a correlation depends on how many independent observations went into it, and it is far larger than intuition suggests: with roughly a quarter of daily observations, the error band around an estimated correlation is wide enough that a reading of 0.3 is not distinguishable from zero, and a swing from 0.3 to −0.1 is inside the noise. Before concluding that a relationship has changed, compare the move with that band. A flip that is smaller than the sampling error is a flip in the measurement, not in the world. There is a further distortion worth knowing about, because it affects how the number is read. A correlation is not linear in the underlying relationship — a strong but non-linear link can produce a modest coefficient, while a single large outlier can produce a large one — and the tails behave differently from the middle of the distribution. That is why practitioners often work with ranks rather than raw returns, which removes the outlier’s leverage and answers a simpler question: when A was in its best decile of moves, where did B tend to sit. None of this is an argument against using the relationships; it is an argument for stating the measurement alongside the conclusion. “Equities and long bonds, sixty-day rolling, monthly returns” is a usable sentence. “Equities and long bonds hedge each other” is not, and it is the one that gets repeated. The transmission story is the model; the correlation is only consistent with it, and a number that could be zero is not confirmation of anything. So the procedure is a short one: pick a window by rule rather than by preference, state the confidence band that goes with it, and only call a relationship changed when it has moved outside that band and stayed outside for more than a few observations. That habit costs one line on a chart and removes the most common way a beginner is fooled by this subject — mistaking a rolling-window artefact for an economic event. Different windows answer different questions on purpose. A short window is a timing tool for positioning; a long window is a description of a regime. Using a long window to time a trade, or a short one to conclude that the regime has changed, is the specific error this read is meant to prevent.

Sources

• Intermarket analysis: bonds, the dollar, commodities and equitiesMurphy, “Intermarket Analysis”
• The discount-rate channel from yields to equity multiplesDamodaran, “The Little Book of Valuation”
• Correlation regimes and the breakdown of the stock–bond hedge in 2022MR17 in this curriculum and the 2022 inflation-shock literature