I spent $300 to find out where 286 strangers live

If you've ever read an article that says something like "according to a survey of Americans," you might have noticed a small line under the results:

n = 250 Americans

Have you ever wondered, "Who the hell are these random people taking surveys?"

I field several surveys a month and I ask this question all the time. I don't know anyone who "takes surveys."

This week I wanted to shed some light on this strange part of surveying, so I fielded a survey of 286 randomly selected Americans (18+) and asked them one question:

"What's your zip code?"

What I found won't completely demystify survey research. But it does show that "n = 250" is much less abstract than it looks — and it reveals some surprisingly interesting patterns in where Americans live.

Closer than you think

Let's say you randomly selected 250 American adults. What do you think the odds are that one of them lives within 10 miles of you right now?

There are about 240 million American adults spread across a piece of land roughly 2,800 miles wide and 1,600 miles tall (not counting Alaska and Hawaii).

You might assume the odds are small. But they're surprisingly high — about 50%.

And if you increase the sample size to 1,100 — closer to what national pollsters like Gallup use — the odds jump to around 90%.

Of course, it depends on where you live. But if you're reading this, there's a good chance you're within driving distance of a metropolitan area, in which case 90% is a safe bet.

So the people answering these surveys aren't "out there." They're someone you'd stand next to in line at the grocery store.

The closest & most isolated pairs

In the survey, I asked respondents to look up their ZIP+4. Most people know their ZIP code, but not the four-digit add-on the United States Postal Service uses for more precise delivery. So I included a USPS lookup tool where they could enter their address and retrieve it.

I then used an API from OpenCage Geocoder to convert each ZIP+4 into latitude and longitude coordinates. From there, calculating the closest pair was straightforward. (For privacy reasons, I'm not sharing their ZIP+4 codes.)

The closest pair lived in Manhattan — one on the Upper West Side and the other on the Upper East Side. They were about 1.3 miles apart, separated by Central Park. Roughly a 20-30 minute walk.

This isn't surprising. Both ZIP codes are among the most densely populated in the United States. 10028, on the Upper East Side, regularly ranks in the top five depending on the dataset.

In fact, nearly all of the closest pairs lived in large metropolitan areas — especially New York City.

At the other extreme, the most isolated respondent lived in Hawaii, 2,392 miles from the next closest respondent. After that came respondents in sparsely populated places like central Utah and western South Dakota. Interestingly, the fifth-most isolated respondent lived in El Paso, one of the most geographically isolated metro areas in the country.

(What would have been strange is if the closest pair had come from suburbs or rural areas. That actually happened twice — once near Worcester, MA and once in Bend, OR. I removed both. The proximity was suspiciously tight, suggesting the responses were fraudulent.)

Half the sample has a neighbor within about 20-25 miles

Imagine a circle around a respondent's home that expands outward until it touches another respondent. The radius of that circle is their "nearest neighbor" distance — a metric I calculated for every respondent in the sample using the latitude and longitude coordinates from each ZIP+4.

A few clear patterns emerge.

About 15% of respondents have another respondent within five miles. By 10 miles, that rises to 30%. At 50 miles, three quarters of the sample has at least one neighbor.

The jump from 10 miles (30%) to 50 miles (75%) reflects how concentrated Americans are around metropolitan areas. Even if you don't live in a city, there's a good chance you can reach one within an hour or so of driving.

After that, the curve flattens quickly and approaches 100% without reaching it because one respondent lives in Hawaii.

Next time you see 'n = 250,' picture your neighbors.

These aren't abstractions. They're real people, and they deserve good questions.

If you're fielding a survey and want a second set of eyes — on the questions, the flow, or whether you're actually measuring what you think you're measuring — that's what a Survey Roast is for. You send me a draft, I send back a recorded critique. No jargon, no slide decks. Just honest feedback.

You can see an example here.

Cheers, Sam