The Titanic crew who ranked last, and ahead of third class among both sexes

Dark navy exhibition panel, kicker “Gallery 3 · Lost at sea · The night of 14–15 April 1912”, headlined “The crew fared worst of anyone aboard”. A subhead says the Board of Trade inquiry counted 2,201 people aboard the Titanic and 711 survivors and that, ranked by where they belonged on the ship, the men and women who worked her fared worst of all, below even the third-class passengers in steerage. A horizontal bar chart labelled “Survived, % of each group aboard” runs on a zero-based axis from 0 to 100%: First class, 325 passengers, 62.5% (203 of 325 survived); Second class, 285 passengers, 41.4% (118 of 285); Third class, 706 passengers, 25.2% (178 of 706); and Crew, 885 aboard, 24.0% (212 of 885), the only bar in red. Below, a line reads “673 of the 1,490 people lost were crew, more than from any class of passenger. Third class lost 528, first class 122.” Three cards on the right: a red “24.0%” captioned “of the crew survived: 212 of 885, the lowest rate of any group on board”; a gold “885” captioned “crew aboard, the largest single group on the ship and 40% of everyone the inquiry counted”; and “Every soul counted: all 2,201 people in the inquiry’s tally, in four groups, on one zero-based scale. No sample, no estimate, no one left out.” The footer credits the British Board of Trade inquiry (1912) as tabulated by R. J. MacG. Dawson in the Journal of Statistics Education (1995), says the gallery panel is invented and every figure is real, and a badge reads “Demonstration · drawn by us from public data”. — Showing the misleading chart

Cream-background chart, kicker “The same 2,201 people · split by sex”, headlined “The crew fared better than third class among men and among women”. A subhead says that pooled, third class edges the crew 25.2% to 24.0%, but split by sex the crew come out ahead in both halves, by 5.0 points among the men and 41.0 among the women, and that the total sides with third class because 97% of the crew were men against 72% of third class, and men were far less likely to survive. A legend marks crew in blue, third class in orange, and first and second class in grey for context. Two dot plots on zero-based 0–100% axes. “Men and boys, 1,731 aboard, 64 of them boys, 21.2% survived”: First 34.4% (62/180), Second 14.0% (25/179), Third 17.3% (88/510), Crew 22.3% (192/862), noted “Crew ahead of third class by 5.0 points, and second only to first class.” “Women and girls, 470 aboard, 45 of them girls, 73.2% survived”: First 97.2% (141/145), Second 87.7% (93/106), Third 45.9% (90/196), Crew 87.0% (20/23), noted “Crew ahead of third class by 41.0 points.” Below, a panel headed “Where each total lands” explains that a group’s total is its men’s and women’s rates averaged by how many of each it carried, so the dot moves toward the women’s rate only as far as its share of women. For the crew a line runs from men 22.3 to women 87.0 with the total dot at 24.0, almost on the men’s end, labelled “97.4% men · 2.6% women”; for third class a line runs from men 17.3 to women 45.9 with the total dot at 25.2, labelled “72.2% men · 27.8% women”. Three cards on the right. “One group, two kinds of people”: 862 of the 885 crew were men and 23 women, 510 of the 706 in third class were men and 196 women and girls, and across the ship 21.2% of men and 73.2% of women survived. “Give both the same mix”: the crew’s rates applied to third class’s 510 men and 196 women, (510 × 192/862 + 196 × 20/23) ÷ 706 = 40.2%, against third class’s own 25.2%, and at the whole ship’s mix 36.1% against 23.4%. “What the split does not settle”: third class carried 79 children and the crew none, though adults alone still put the crew ahead (22.3% v 16.2% of men, 87.0% v 46.1% of women); some surviving crewmen were sent to man the lifeboats; the 23 crew women are too few for a precise rate; and sources disagree on exact counts. The footer credits the Board of Trade counts as tabulated by Dawson (1995) and distributed as the Titanic dataset in R, and says both drawings are our own demonstration. — Showing the misleading chart

Be the first to star this exhibit

A museum-style panel we drew from the Board of Trade’s Titanic counts ranks survival by class, and the crew come last: 24.0% against third class’s 25.2%. Split the same 2,201 people by sex and the crew are ahead among the men (22.3% v 17.3%) and among the women (87.0% v 45.9%). They lose the total because 97% of them were men, and the men’s rate is what their total mostly records.

British Board of Trade inquiry into the loss of the Titanic (1912) — class, sex, age and survival for 2,201 people, as tabulated by Robert J. MacG. Dawson, “The ‘Unusual Episode’ Data Revisited”, Journal of Statistics Education 3 (1995), and distributed as the Titanic dataset in R’s datasets package, 1995Added October 14, 2026

01The claim

The crew fared worst of anyone aboard the Titanic. On a zero-based ranking of survival by where people belonged on the ship, 62.5% of first class survived, 41.4% of second, 25.2% of third and 24.0% of the crew, last of all, below even the passengers in steerage. (A gallery panel of the kind a maritime exhibition might hang, which we drew from the Board of Trade counts.)

02The trick

Nothing on the panel is miscounted: the axis starts at zero and all 2,201 people the inquiry counted are in it. What the ranking leaves out is the one thing that mattered most to who survived that night: whether the person was a man or a woman. Across the ship 73.2% of women survived (344 of 470) and 21.2% of men (367 of 1,731). And the four groups were made of very different mixes of the two. Of the 885 crew, 862 were men, 97.4%. Of the 706 people in third class, 510 were men, 72.2%, and the other 196 were women and girls. Compare like with like and the order flips. Among men, 22.3% of the crew survived (192 of 862) against 17.3% of third class (88 of 510), putting crewmen second of the four groups, behind only first class. Among women, 87.0% of the crew survived (20 of 23) against 45.9% of third class (90 of 196), a gap of 41.0 points. The crew are ahead in both halves and behind on the total. Nothing is miscounted: each group’s total is just its men’s rate and its women’s rate averaged by how many of each it carried. The crew’s average is weighted almost entirely to the men, so it lands at 24.0, close to their men’s 22.3. Third class carried enough women to pull its average up to 25.2, even though its women fared badly. The pooled ranking mostly records who was in each group. (Its “lowest rate of any group” is also only true at this grain: second-class men, at 14.0%, fared worse than any of the four totals.) That is Simpson’s paradox: a comparison that reverses when you split it by something that differs between the groups and also drives the outcome. (This is our own demonstration. We are not reproducing any published chart or exhibit.)

03The fix

Put sex back on the chart. Two dot plots, one for men and one for women, with all four groups on the same zero-based scale, make the reversal plain: the crew are ahead of third class by 5.0 points among the men and by 41.0 among the women. The panel worth borrowing for any reversal like this one shows where each total comes from: a line from a group’s men’s rate to its women’s rate, with the total sitting along it in proportion to the share of women. The crew’s total sits 2.6% of the way along; third class’s sits 27.8% of the way. If one number per group is needed, standardise: score both groups on the same mix of men and women. Apply the crew’s rates to third class’s own 510 men and 196 women and the crew come out at 40.2%, (510 × 192/862 + 196 × 20/23) ÷ 706, against 25.2%; at the whole ship’s mix, 78.6% men, it is 36.1% against 23.4%. Then say what the split does not settle. It removes one recorded difference, not every one. Third class carried 79 children and the crew none, but adults alone still put the crew ahead (22.3% v 16.2% of men, 87.0% v 46.1% of women). Some surviving crewmen were seamen sent to row and steer the lifeboats, so their higher rate partly reflects the jobs they were given that night: the split shows that the ranking reverses, not why. The panels also count boys with the men and girls with the women (64 boys and 45 girls in all), as the inquiry’s table allows. The 23 women of the crew are too few for a precise rate: two more deaths among them would move it almost nine points. And the primary sources disagree on exactly how many were aboard and how many were saved; these are the Board of Trade’s figures as Dawson tabulated them.