The friendship paradox

Your friends have more friends than you do. In a real karate club, 29 of 34 members prove it.

An anthropologist mapped who socialised with whom in a university karate club. Before you look, guess how lopsided popularity was. Then see why this is not about that club at all.

Tap a person to inspect

Step 2

It is sampling, not friendship

Ask a random person how many friends they have, and you average over people. Ask a random person to name a friend, and you average over friendships. Run it yourself.

Each draw: one random person, then one random friend of theirs.

RANDOM PERSON: friends-

THEIR RANDOM FRIEND: friends-

random personnamed friendx axis: number of friends

Step 3

Who escapes the paradox?

All the data

Every member, ranked

Open the full table

Member	Friends	Friends' average	Friends ahead?

Click a column to sort. Click a row to inspect that member.

Sources

Karate club: Wayne W. Zachary, "An Information Flow Model for Conflict and Fission in Small Groups", Journal of Anthropological Research 33 (1977). Secondary source: the 78 friendships were read from the NetworkX function karate_club_graph (github.com/networkx/networkx, networkx/generators/social.py), which transcribes the paper's table. A friendship here means two members interacted outside class, as coded in the dataset.
Les Misérables: character co-appearance in the same chapter, from Donald Knuth's Stanford GraphBase, as shipped in miserables.json (github.com/d3/d3-plugins, graph/data). Secondary source. The file has one self-link and one character (OldMan) with no links; both were removed, leaving 76 characters and 253 links.
The paradox: first described by the sociologist Scott L. Feld in 1991 (Wikipedia, "Friendship paradox", secondary). All averages on this page are computed in your browser from the lists above.