The Block, by the Numbers
The mailer landed on a Tuesday, sandwiched between a grocery-store circular and a bill from my dentist. On the front, a glossy row of brick homes. The headline: Your Neighborhood Is Changing—Here’s What It Means for You. Inside, a tidy set of figures: median household income in this slice of Chicago had climbed 14 percent in three years. The share of adults with a college degree had jumped. The poverty rate had dropped by almost a third. The source, printed in type the size of a poppy seed, was the Census Bureau’s American Community Survey five-year estimates.
If you’ve ever watched a neighborhood change in real time—new coffee spots, rents inching up, the faces at the bus stop getting younger or older—the mailer felt true and a little too clean. It caught something, but it missed everything else. Census estimates work that way: a sharp lens that can bring a place into focus, as long as you remember it also flattens the picture.
What the Estimates Actually Measure
Let’s start with the bones. The American Community Survey—ACS for short—is the Bureau’s rolling replacement for the old decennial long-form questionnaire. Instead of a snapshot every ten years, you get two main flavors: one-year estimates for places with at least 65,000 people, and five-year estimates for everything down to individual census tracts. For a few square blocks, a ZIP code, a single tract, the five-year ACS is usually the only option.
The survey covers the basics: age, race, household type, income, education, housing costs, and more. It’s not a head count. It’s a rolling sample, pooled over 60 months to get enough responses for reliability. A neighborhood’s 2022 five-year estimate, for instance, pulls from surveys collected between 2018 and 2022. So the number is already a composite, smeared across time before you ever read it.
What You Can See: Long, Slow Shifts
Even with that smear, the ACS does catch genuine, tectonic movements. I’ve tracked a historically working-class neighborhood in Baltimore for years. Its five-year estimates show median rent rising steadily—from $850 to $1,250 over a decade—and a matching dip in the share of households spending more than 30 percent of income on housing. On the surface, that looks like a win: fewer people stretched thin. But cross-tab the data with tenure, and you see the owner-occupied share climbing while the number of renters drops. The neighborhood isn’t getting more affordable. It’s swapping renters for homeowners. The ACS lets you spot that trade—if you’re patient and skeptical enough to dig past the headline.
The survey is also good at illuminating broad shifts in education and work. In a gentrifying tract, you might watch the share of adults with a bachelor’s degree rise from 18 percent to 35 percent over three five-year periods, while the slice of people in service jobs falls by a similar amount. Those trends are real, and they often show up as the earliest statistical flicker of change—years before a boutique gym opens or a corner store becomes a wine bar.
What Gets Lost: The Margins, the Margins of Error
Now for the part that demands you slow down. Every ACS estimate arrives with a margin of error—a statistical range that tells you how much faith to put in the number. For big populations, like the total housing units in a county, the margin is often tiny. For a small subgroup inside a single tract—say, Black renters over 65—the margin can balloon until the estimate is basically noise.
I recently pulled five-year data for a tract in Queens where the estimate showed a 20 percent rise in the Asian population over a decade. The margin of error on that figure was plus or minus 15 percentage points. That means the actual change could have been anywhere from a 5 percent bump to a 35 percent surge—a spread wide enough to support completely different stories about who was moving in and why. Neighborhood change is often narrated through small, vulnerable subgroups, and those are exactly the groups the ACS captures with the fuzziest precision.
Also lost: the texture of turnover. Census estimates are net measures. They tell you the balance at two points in time, not who left, who stayed, or who arrived in between. If a tract gained 200 college-educated white households and lost 200 Black working-class households over five years, the ACS might show a stable population count and a higher median income. You’d never know a wholesale displacement happened unless you matched the data with address-level administrative records—something the ACS can’t give you.
Time and the Lag Problem
Five-year estimates are backward-looking by design. By the time the 2018-2022 data drops in late 2023, the midpoint of that window is already more than three years old. In a neighborhood changing fast, that lag can make the numbers feel like a photo of a party after everyone’s gone home.
Take a block in Austin that saw a wave of teardowns and new construction starting in 2021. The 2018-2022 estimates will include only a partial count of those new units—maybe 18 months’ worth, spread across the five-year sample. The median home value might look like it rose modestly, while prices on the ground actually doubled. A real estate agent would know that. A journalist glancing only at the topline figure would miss it completely. The ACS is a rearview mirror, and a smudged one. Useful for spotting a turn, not for steering through the curve as it happens.
Comparing Across Geographies: The Temptation Trap
It’s dangerously easy to put two census tracts side by side and declare one richer, whiter, or more educated. The pull is especially strong when a neighborhood’s edges line up with familiar markers—a highway, a park, a main street—that make the tracts feel like natural units. But tract boundaries are drawn for administrative convenience and roughly equal population size, not to reflect organic community edges. A single block group can straddle two different social worlds, and the ACS will average them into one number.
Even inside a single tract, the variation can be huge. I once mapped the ACS median income for a tract in Philadelphia and then walked the streets within it. On one block, renovated rowhouses sold for $700,000; two blocks over, families were doubled up in apartments with peeling paint. The tract-level estimate split the difference at a neat $65,000—a figure that described neither side. Aggregation is the ACS’s original sin, and no amount of statistical polish can fully erase it.
Practical Reading: What to Ask of Any Estimate
If you’re handed a census estimate about a neighborhood—from a mailer, a news article, a city planning doc—ask three questions before you buy its story.
First, what’s the margin of error? If the number is big relative to the estimate itself, treat the finding as suggestive, not settled. The Census Bureau warns against using estimates with coefficients of variation above 30 percent, but even a 15 percent CV can hide a lot. Always check the published margin before you repeat a statistic as fact.
Second, what’s the time frame? A five-year estimate is a period average, not a point-in-time measurement. Ask yourself whether the window makes sense for the change you’re trying to understand. If a neighborhood transformed in 24 months, a five-year estimate will understate the speed—and maybe the direction—of the shift.
Third, what’s the geography? Tract-level data can help you spot broad patterns, but it’s a blunt tool for block-by-block dynamics. If someone is making a claim about “the neighborhood,” find out exactly which tracts they’re using and whether those boundaries match what residents would recognize as their community.
Beyond the Tract: What Else to Watch
Census estimates work best alongside other signals. Building permits tell you where new housing is actually appearing, and in what form—single-family homes or apartments, rentals or condos. Property tax records reveal shifts in ownership and assessed value at the parcel level. School enrollment data, when you can get it, can show family composition changes years before the ACS picks them up. Even a simple change in the number of active voter registrations can hint at who’s arriving and who’s leaving.
None of these sources is perfect, and each carries its own biases. But together they form a mosaic that can check the census estimates against reality. In that Baltimore neighborhood I mentioned, the ACS showed a decline in the child population over a decade. School enrollment data, though, showed a sharp rise in kindergarten registrations in the final two years—a sign that families with young children were starting to move back, a trend the five-year average hadn’t yet caught. The census wasn’t wrong. It was just slow.
The Mailer, Revisited
That glossy mailer from my Chicago neighborhood wasn’t lying, exactly. The numbers it cited came straight from the ACS, and they reflected real shifts in income and education. But the mailer left out the margins of error, ignored the time lag, and treated the tract as one uniform place. It told a story of progress and uplift—a narrative that fit the sender’s purposes—without nodding to the people who’d been priced out during the same period, or the blocks that had changed very little.
Neighborhood change is messy, contested, and deeply personal. Census estimates can be an anchor in that mess, a way to ground conversations in evidence instead of anecdote. But they are not a mirror. They’re a blurry photograph of a moving subject. Reading them well means squinting at the edges, asking what’s just out of frame, and remembering that the real story often lives in the gaps between the numbers.
Frequently Asked Questions
How often are census estimates updated for neighborhoods?
The American Community Survey releases five-year estimates annually for all census tracts and block groups. Each release covers a rolling five-year period—for example, 2018-2022 data came out in late 2023—so a new set of estimates is available every year, but each one overlaps heavily with the previous version. One-year estimates are only available for larger areas with populations of 65,000 or more.
Can census estimates tell me exactly who is moving into my neighborhood?
Not directly. The ACS measures the characteristics of people living in an area at the time they respond to the survey, but it does not track individual moves. You can infer broad migration patterns by comparing demographic profiles over time, but you cannot distinguish newcomers from long-time residents, or see who left between two survey periods.
Why do some census tract estimates have very large margins of error?
Small population sizes and low survey response rates in a tract lead to higher sampling variability. When the number of survey respondents who share a particular characteristic—such as being a renter of a specific age and race—is tiny, the estimate for that subgroup becomes imprecise. The Census Bureau publishes margins of error for every estimate, and users should check them before drawing conclusions about small groups.
