MIT let a lottery place married student veterans in neighbouring apartments in 1946. Seventy-six years later, researchers hunted for the same geography pattern inside Major League Baseball's draft. Both searches landed on it. Physical nearness predicts who gets chosen, trusted, or labelled a good fit, and that prediction holds even when skill measures say otherwise.
Baseball is the harder proof, because a university can shuffle apartments and watch friendships form. No league will reshuffle where scouting directors live, or run the draft twice under two different maps.
So the 2022 paper built a substitute. It used each prospect's recorded stats and scouting grades to stand in for the coin flip a housing study gets for free. That move, showing cause when a real lottery is impossible, shows up often enough in published work that this site keeps a whole section for it: Natural Experiments.
The rest of this page reads the two studies side by side. First the lottery that pinned down the mechanism. Then the draft that found the same lever moving careers. Last, four checks for any result someone still wants to call pure merit.
The story in four parts
In 1946 MIT placed married student veterans into apartments essentially at random, then asked each couple to name their closest friends in the complex.
Couples nineteen feet apart named each other a close friend four times as often as couples eighty-nine feet apart on the same hallway, with no other measured difference between them.
Holding a prospect's measured skill fixed with machine learning, a 2022 study still found scouts leaning toward players who lived nearby.
Prospects who gained from living near their scouting director were 38% less likely to appear in even one major-league game.
Leon Festinger, Stanley Schachter and Kurt Back were hunting for what actually produces friendship, and MIT handed them an unusual lab. Two new complexes, Westgate and Westgate West, had just gone up for married student veterans arriving on the GI Bill after World War II. Incoming families received units essentially at random, before anyone met a future neighbour or walked a floor plan. Months on, each couple named their three closest friends inside the complex.
Why did living nineteen feet from someone predict friendship better than anything about who they actually were?
Unit assignment worked like a coin flip, so friendship patterns that followed could not come from people already choosing neighbours they liked. Whatever drove the pattern had to sit in the building's layout.
41% of next-door neighbours, about 19 feet apart, named each other a close friend. Two doors down the share fell to 22%. Opposite ends of the same hallway, about 89 feet apart, sat at 10%. Roughly 65% of every close friendship named stayed inside the same building, even though residents could meet people across the rest of the complex just as easily.
One more detail tightens the mechanism. A door beside a shared stairwell or mailbox, where every resident had to pass, drew more friend nominations than raw distance alone would predict. Straight-line feet missed what mattered. Festinger and co-authors called the crossing of daily paths functional distance.
Two neighbours need only repeated, unplanned contact. Familiarity from those small collisions starts to feel like earned compatibility, even though a floor plan neither person chose set up the meetings.
Westgate's evidence is clean because MIT did something companies and leagues almost never do on purpose: it randomised who ended up physically close to whom. That is why the friendship finding holds weight, and why Major League Baseball can never give a researcher the same gift. No one will move scouting directors around a map and re-run the draft to watch picks change.
A 2022 paper asked whether the same lever, familiarity from repeated exposure standing in for a real merit judgment, was already shaping the MLB draft. Each team's scouting director works through thousands of prospects with shared stats, reports, and film, a process designed to reward talent across geography.
Showing that geography still mattered needed a stand-in for random assignment. Machine learning built a flexible prediction of each prospect's draft-worthy skill from everything else on file. The authors then asked whether living nearer a team's scouting director still predicted selection once that skill estimate was held fixed. The full method, and what it can and cannot prove, sits in Natural Experiments.
A player is 7.1% more likely to be drafted for every 1,000km closer he lives to a team's scouting director.
A second, smaller lift sat beside it: about 4.9% more likely per 1,000km closer to the team's home city. Both effects survived the machine-learning skill control, and both peaked in the draft's later rounds, where a scouting director has the widest personal discretion and the least outside review.
Players drafted with a proximity advantage were 38% less likely to appear in even one MLB game than similarly drafted players without it, and played in roughly 25 fewer games on average. Geography did more than reorder who got taken when. It pulled down the quality of who got taken.