Cherry-picking

The reporting threshold

Every row on the chart cleared a line somebody drew. What didn’t clear it isn’t zero — it’s absent.

a.k.a. the de minimis cut · the coverage floor · the scope threshold · the eligibility rule · mandatory reporting limits · the registry cut-off · ships above 5,000 GT · facilities above 25,000 tonnes · firms with 250 or more employees · the regulatory dataset

Most of the best data in the world exists because somebody was obliged to file it, and an obligation has to say who it applies to. So every register carries a line: above this tonnage, this headcount, this turnover, this emissions rate, you report — below it, you do not exist. The line is almost never a judgement about the data. It is a judgement about paperwork, made once, in an impact assessment, by weighing what a return costs the filer against what the regulator gains. Then the register is published, and it is superb: mandatory, audited, comprehensive, unbiased in every respect the word normally means. The trouble starts when it is read as a picture of the world rather than a picture of the filers. Because the threshold is a size rule, what it removes is not a random sample — it is everything small, which is usually most of the units and a minority of the quantity. A line set to capture 90% of the tonnes can easily hold half the objects, and both of those sentences are true of the same rows. Absence then does the rest, because a category with no row does not look like missing data: it looks like a zero, or like nothing worth a row. And thresholds move — a rule is amended, a limit is not indexed and inflation drags units over it — so the same series can step up or down in a year when nothing in the world changed but the eligibility clause.

How to spot it

Ask who was obliged to file this, and what the obligation said. Every mandatory dataset has a scope clause, and it is normally one sentence: above 5,000 gross tons, over 25,000 tonnes of CO₂e a year, 250 employees or more, facilities of this capacity and no smaller.
Read the denominator. “42% of the industry” computed inside a register means 42% of the filers, and the filers are the ones above the line.
Look for the category that is missing entirely rather than small. Small things arrive as short bars; excluded things arrive as no bar, and no bar is the one value the eye reads as agreement.
Check whether the threshold and the question are about the same quantity. A line drawn to catch emissions catches emissions; asked how many ships, how many farms, how many firms, the same rows answer a question they were not built for.
Watch the publisher’s own caveat. Regulators are usually candid about coverage — “covers around 90% of emissions, around 55% of ships” — and that sentence, in the front matter, is frequently the only place the chart’s limits appear.
A step in a time series with no event behind it. Ask when the scope last changed: units brought in, a limit raised, an exemption expiring. The picture moves because the rule moved.
Watch for bunching just above the line and a cliff just below it. Where reporting is costly, some units stay small on purpose, which makes the smallest reported size class its own artefact.
Ask whether anything below the line has been estimated elsewhere. Where a second source models the excluded tier — an AIS study, a survey, a satellite inventory — the gap between the two is the size of what the register cannot see.

The fix

Say what the rule is, in the caption, the way you print a unit: this chart shows ships above 5,000 gross tons, facilities above 25,000 tonnes, firms with ten or more employees. Then draw the line on the chart. Where the excluded tier has been estimated by anyone — and for most big registers somebody has, because the regulator needed to know before it set the line — give the estimate a bar of its own, in its own shade, labelled as an estimate: a modelled bar with a stated method beats white space, because white space is read as zero and a labelled estimate is read as an estimate. Where nothing has been estimated, write the sentence anyway, since “units under this size are not in this data” is short and it is the whole finding. Match the register to the question rather than the other way round: a threshold chosen to capture a quantity is excellent evidence about that quantity and poor evidence about how many things there are, so a census question needs a census, or at least a line saying which one you had. Keep the threshold in view when the series moves, too — annotate the year the scope changed and, where the numbers allow, draw the old scope and the new one as two series rather than one with a step in it. And never let a share computed inside a register be quoted as a share of the world: the denominator has the rule baked into it, and so does every percentage taken off it.

In the gallery

Dark navy briefing slide, masthead “Port decarbonisation desk · ships calling at EEA ports” on the left and “EU maritime MRV returns · reporting year 2023” on the right, headlined “Twelve thousand ships call here, and five types are four fifths of them.” Beneath the headline sits a bordered chip reading “Demonstration · drawn by us from published figures”. A standfirst says every figure is the European Commission’s own published summary of the 2023 returns under the EU Maritime MRV Regulation — verified by accredited verifiers, flag-neutral, one reporting year, one method; that 12,344 ships filed an emissions report for 2023 and together emitted 126.7 million tonnes of CO₂ on voyages to, from and between ports of the European Economic Area; that reporting is mandatory and every return is checked by an accredited verifier before publication, so nothing has been sampled, smoothed, indexed or rescaled; that the axis is linear, zero-based and evenly ticked and the shares are the Commission’s own at the precision it published them; and that the Commission publishes these figures in prose and in its own charts while the bars are ours. The main panel, headed “The fleet that calls at EEA ports, 2023” and subtitled “Share of the 12,344 ships filing an emissions report, by ship type · the five named types and the ten others · zero-based linear axis, evenly ticked, unbroken · shares as published”, is a horizontal bar chart on an axis running 0% to 50% in even steps of 10: bulk carriers 29%, oil tankers 15%, container ships 15%, chemical tankers 11%, general cargo ships 9%, and a sixth bar in a darker gold for the other ten reporting types at 21%. A caption reads “Shares as published · every ship in the 2023 register · one scale throughout”. A second panel, “And where that fleet’s carbon is”, draws the 126.7 Mt of reported CO₂ as one stacked bar: container ships 28%, oil tankers and bulk carriers 24%, the other twelve reporting types 48%, with a line beneath noting the top three ship types together at 52% of reported CO₂, bulk carriers alone at 11% — their lowest share on record — and 41 million tonnes of fuel burned by the reporting fleet. A third panel, “What each ship reports”, lists in two monospaced columns what a return carries — fuel consumed by fuel type, CO₂ emitted at sea and at berth, distance travelled and time at sea, cargo carried and transport work — alongside what is derived and published with it: average energy efficiency (EEOI, AER), technical efficiency (EEDI or EIV), ship type, deadweight and gross tonnage, and the verifier and company by name; a note says coverage is voyages to, from and between EEA ports plus emissions at berth, and that methane and nitrous oxide join the return from 2024 while the 2023 year reported here is CO₂ only. Six cards run down the right. “The fleet, in one figure” prints 12,344 ships in large type and says they filed for 2023, the second-highest count since the register opened in 2018 and 5.4% below 2022, reporting 126.7 Mt of CO₂, 7.9% less than in 2022 and about 13% below the pre-pandemic years. “Read-out for the desk” says five ship types are 79% of the arrivals and the top three are 52% of the carbon, that bulk carriers, tankers and container ships are the fleet, that no other named ship type on either chart reaches a tenth of it and small craft do not register at all, and recommends pointing the shore-power and berth-emission programme at the container terminals and the tanker and bulk berths because a small-craft programme has nothing on this evidence to aim at. “What the register is” explains that under the EU Maritime MRV Regulation ships above 5,000 gross tons monitor and report fuel use and CO₂ for voyages to, from and between EEA ports including emissions at berth, that the rule is flag-neutral, and that every report is checked by an accredited verifier before publication ship by ship in EMSA’s THETIS-MRV. “The five and the ten” tabulates the shares. “How the year moved” notes that twelve of the fifteen reporting ship types emitted less in 2023 than in 2022, bulk carriers falling furthest at 4.5 Mt or 23% with 12% fewer active, container ships down 6% while about 2% more of them reported, and passenger ships up 6% to their highest since 2018. “Basis of measurement” credits the European Commission’s 2024 Report on CO₂ Emissions from Maritime Transport, SWD(2025) 38 final of 12 February 2025 — the sixth annual report, covering the 2023 reporting year under Regulation (EU) 2015/757 — with shares and totals as published, notes that the exact count of reporting ships is the 2023 MRV data extract of 19 September 2024 as cited in the Commission’s supporting study while the annual report itself rounds it to “almost 12,300”, and gives a retrieval date of 11 September 2026. A method line runs along the foot: published figures as reported; one reporting year, one method; verified returns, flag-neutral; zero-based linear axis, evenly ticked; nothing sampled, smoothed or rescaled. A footnote credits the source and states that this slide is a demonstration drawn by us in the manner of a port authority’s decarbonisation briefing — the desk, the read-out and the recommendation are invented, and every figure behind them is real.

The 8,525 ships the chart has no row for

The reporting threshold