MisleadingCharts
Back to the gallery

The six dogs per name that put two Lunas in one waiting room

Showing the misleading chart

Be the first to star this exhibit

A vet’s waiting-room bulletin we drew divides New York’s 86,321 licensed, named dogs by their 14,651 names and gets 5.9 dogs per name. On that average, two patients in a 14-seat room should share a name only 0.6% of the time, so the practice drops surnames at check-in. But 60% of the names belong to a single dog, and Luna belongs to 1,059. With the names the city’s dogs actually have, a full room holds two namesakes about one time in eight.

01The claim

New York’s 86,321 licensed, named dogs share 14,651 names, so the average name belongs to just 5.9 dogs, and it has stayed between 5.3 and 7.1 in every register the city has published. With names spread evenly over all 14,651, two patients in a full 14-seat waiting room should share a name only 0.6% of the time, so a vet can safely call dogs by first name alone.

02The trick

Every number on the bulletin is right; the problem is the word “average”. Dogs per name is a mean, 86,321 dogs divided by 14,651 names, and a mean shares a total out evenly among members who don’t share it evenly. Dog names have a long right tail. 8,839 names, 60% of them, belong to exactly one dog (a Prufrock, a Dinglehopper, a Fafnir, and some typos). At the other end, Luna has 1,059 dogs, Bella 906, Coco 724, and the 125 names carried by more than 100 dogs each cover 28,092 dogs, a third of all named dogs. Biscuit, the bulletin’s “practically unique” example, has 114. The mean of 5.9 falls between those two groups and describes neither: 85% of names sit below it, yet 78% of dogs carry a name above it, and only 296 names have exactly six dogs. The bar chart of past years makes the figure look settled, but its wobble mostly tracks the size of each extract: the 2019–21 register, the largest at 137,658 dogs because it spans three years, gives the highest mean, 7.1, because extra dogs mostly add more Lunas rather than new names. The waiting-room odds rest on the same even picture. The clinic’s 0.6% gives each patient an equal chance of any of the 14,651 names. For whether two dogs in a room share a name, the city’s dogs behave as if there were only about 680 equally common names. (We drew this bulletin and its practice as a demonstration; the counts are the city’s.)

03The fix

The redraw shows every name instead of their mean. It groups all 14,651 names by how many dogs carry each one and draws two bars per group on one shared scale: how many names fall in the group, and how many dogs those names cover. The name bars pile up at one dog. The dog bars pile up between 11 and 500 dogs a name, where Shadow (155), Bruno (266) and Biscuit (114) live, with the Lunas and Charlies in the last row. The mean of 5.9 is drawn as a line between the groups, where few names and few dogs sit. Two medians tell the story better. The median name has 1 dog. Half of all dogs share their name with at least 37 others. For the waiting room, simulating 500,000 full rooms of 14 dogs drawn at random from the 86,321 puts two namesakes in the room about 12% of the time, one room in eight, or about twenty times the clinic’s 0.6%. The general rule: for anything with a long right tail, such as names, incomes, citations or followers, quote a median or a share, or show the whole spread, and name the outliers separately. And keep the surname at check-in.