What a measured effect on this site actually means

Every event page on this site that has enough public data behind it carries a “measured effect” section. It states, for example, that transits through a chokepoint fell by a stated percentage during the event window compared with an expected value from before it. It is worth being precise about what that number is, and is not, saying.

What gets compared

For each event, we take the daily count that a public source publishes, such as IMF PortWatch’s vessel transit counts through a chokepoint, and split it into two ranges: the event window, running from the event’s start date to its end date or to today if it is still ongoing, and a baseline, built from a stretch of days before the event started. The baseline is not simply an average of every day before the event. It excludes the seven days right before the event starts, on the theory that disruptions often cast a shadow ahead of their official start date. It excludes any day inside another recorded event’s own window, so one disruption does not contaminate the baseline for a different one nearby. It excludes public holidays from a curated calendar, since holiday traffic dips are a known, unrelated pattern. And it builds the expected value from the same day of the week as the day being compared, since weekday and weekend traffic can differ substantially at a busy chokepoint.

Why it can widen or shrink

If the baseline does not end up with enough clean days after all these exclusions, the method automatically reaches further back, a week at a time, until it has enough. There is a floor on how thin a baseline is allowed to get before we simply refuse to publish a number rather than publish one built on too little data, which the record states plainly as “insufficient baseline” when it applies.

Why “timing, not cause”

Every effect block on this site carries the same caveat, unchanged, wherever it appears: this shows a measured change in public data during the event window compared with a baseline before it. It shows timing, not cause. That sentence deserves its own explanation. A drop in transits during a canal closure is compatible with a story where the closure caused the drop. It is not proof of it on its own, because plenty of things move chokepoint traffic that have nothing to do with any single event: seasonal demand, weather unrelated to the event, other disruptions we have not yet recorded, or genuine noise in day to day ship movements. We only describe an event as having caused a measured change when an official source, cited by id, makes that claim. Otherwise the number stands next to the event in time, and the reader draws their own conclusion about causation, with the same data we used to compute it available to check.

Reproducing it yourself

The method is public and versioned. The current version’s full specification, including the exact constants used for baseline length, extension steps and the statistical floor applied to a thin or noisy series, is published on the effect method page, and a script in the pipeline can recompute any published figure directly from the same public series file the site itself reads. Nothing about how a number on this site was produced is meant to require trusting us.