DICEMB Brief · 21 September 2026

How to Read a Market Reaction Without Confusing Correlation With Cause

A shared move is a correlation. A mechanism is a claim. Most post-event commentary upgrades the first into the second too fast.

A market reaction is a set of price changes after some timestamp. Correlation is those changes lining up: Bitcoin and equities both fell, or yields and the dollar both rose. Causation is a claim that one of those changes, or the headline that preceded them, produced the others. The claim can be plausible and still be poorly supported by a single afternoon. Reading well means labeling three different things: the correlation you can see, the mechanism you think is plausible, and the evidence you do not have yet.

This is a method article. It uses macro and crypto examples because those pages are where the mistake is expensive, including on Fed decision days. It does not offer a model that turns a headline into a trade.

Three labels, used on purpose

Label	You are allowed to say	You are not allowed to pretend
Correlation	These series moved together in this window	That they must move together next time
Plausible mechanism	Here is a channel that could connect them (yields, dollar, forced selling)	That the channel is the one that operated today
Evidence limit	We do not have positioning, flows or a counterfactual	That confidence can replace the missing measurement

Example: a rate decision and a Bitcoin candle

Correlation. In the hour after a statement, a Treasury yield rises and Bitcoin falls. That is a description of two series.

Plausible mechanism. A higher expected policy path can raise yields, support the dollar, and reduce appetite for high-volatility assets. Leveraged crypto books can amplify the first impulse. That mechanism is taught constantly because it has applied in some windows.

Evidence limit. You usually cannot see, in public and in real time, how much of the Bitcoin move was spot selling versus liquidations, or how much was already positioned for this statement. If equities did not fall, the “risk appetite” sentence is weaker than it sounds. If a large fund-flow headline landed in the same hour, you have a competing mechanism. The correlation survives; the exclusive causal claim does not.

Example: an inflation print and a narrative

Suppose CPI prints above consensus and Bitcoin rises that morning. The correlation is real for that window. One plausible mechanism is that the print was read as nominal growth, or that the move was a squeeze unrelated to the print. Another, contradictory, mechanism is the inflation-hedge story. You cannot pick between them from the candle color. The piece on CPI and real yields exists so the vocabulary is in place before the narrative starts. Vocabulary is not evidence, but missing vocabulary makes false certainty easier.

Example: two assets that “always” move together

Gold and Bitcoin, or Bitcoin and a stock index, sometimes move together and sometimes do not. A month of similar direction is a correlation over that month. A mechanism — shared liquidity, a shared narrative, a shared investor base — might explain it. The evidence limit is the number of independent episodes you actually checked, and whether you ignored the months they diverged. Choosing the months that fit is how a correlation becomes a slogan.

A short discipline after any headline

State the window in minutes or days. A cause that only works if you zoom until it appears is not a result.
State the correlation without adjectives. “Both down” is enough.
State at most two mechanisms, including one that would have predicted the opposite, so you can see you had a choice.
Name the measurement you do not have: order-book data, a full flow tape, a survey of positioning.
Demote the claim if a second headline landed inside the same window.

Writers who do this will publish fewer certain sentences and fewer corrections. That is the point. DICEMB’s job is to keep the reported move and the interpretation in different sentences so a reader can disagree with the interpretation without being told they misunderstood the fact.

What a careful paragraph looks like

A usable paragraph has a fact sentence and an interpretation sentence, and a reader can tell them apart without goodwill. Fact: “In the thirty minutes after the September 16, 2026 FOMC statement, which raised the target range to 3-3/4 to 4 percent, a named yield and Bitcoin either moved together or they did not — record which.” Interpretation: “One reading is that the path, not the quarter point, was the surprise.” The interpretation can be wrong tomorrow. The fact sentence should still be checkable against the Fed’s note and a chart. If you only have the interpretation, you do not have a market reaction written down; you have a mood.

Two failure modes show up constantly. The first is the single-cause story: one headline is asked to explain every asset that moved, including assets that moved earlier. The second is the invisible counterexample: the writer remembers the days the mechanism worked and files the other days under “positioning,” which is a word that can mean anything. Listing one counterexample you know about is enough to keep the mechanism in the “plausible” row of the table. You do not need a formal study to be more careful than a slogan.

A third failure mode is the renamed chart. Drawing an arrow from the headline to the candle does not add information; it repeats the timestamp. If the only evidence is “this happened after that,” say “after,” which is a sequence, and stop. Sequence is weaker than correlation across several episodes, and correlation is weaker than a mechanism you can watch in another market at the same time. Yields, the dollar and equity indexes are useful exactly because they are other measurements, not because they are automatically the cause. When they disagree with the story, keep the disagreement in the paragraph.

Key takeaways

Correlation is the co-movement. Cause is an extra claim.
A plausible channel (yields, dollar, liquidations) is not proof that the channel fired today.
Always note a competing mechanism and the data you do not have.
Fix the time window before you argue.
Use the Fed and CPI guides for definitions; use this page to stop the definitions from becoming slogans.