Stop underwriting on 2020 Census data

Nearly every population data vendor relies on the same source: the Census Bureau. They may repackage or license Census data under new brand names, but it’s still anchored to 6-year-old population counts.

Those numbers were established during COVID-19, under conditions that made an accurate count harder than usual. Underwriters who make decisions on 2020 data never had the chance to weigh the risk because nobody flagged it. It’s been sitting in their models the whole time.

The 2020 count was collected mid-pandemic. Not enough households responded on their own, which meant more of them needed a Census worker to show up in person and count them directly. In-person counting was delayed and scaled back at the same time.

The Census Bureau reviewed its own work afterward and found statistically significant coverage errors in 14 states. For example, New York was overcounted by 3.44%. Arkansas came in 5.04% below its true population, the largest state-level miscount.

In 2020, Census overcounted New York by 3.44% and undercounted Arkansas by 5.04%.

A 5% population error at the state level is not a rounding issue in an underwriting context. It propagates to every model, comp, demand forecast, and absorption assumption you use to price a deal. Since some markets were undercounted and others overcounted, no single adjustment can fix them all. A correction that works for one market makes another one worse.

CENSAI tracks migration using verified mover data refreshed every quarter, not a decennial snapshot. In Arkansas, where the undercount ran deepest, CENSAI data showed positive net migration every quarter since 2020. The people were there. They were showing up in address changes and household records the entire time.

The Census Bureau eventually updated its own population estimates to show net migration growth of roughly 2.5% by 2024. It took years for the official number to catch up. In the meantime, anyone underwriting off the original figure was working from markets that no longer existed.

Anyone underwriting with 2020 data is working from markets that no longer exist.

This is the practical argument for a shorter refresh cycle. A decennial count corrects itself on a ten-year schedule with revisions in between. Capital doesn’t move on that timeline. Neither do renters.

The American Community Survey is where 2020 Census data is easy to miss. ACS results land every year, but they’re weighted against the same 2020 population count. There has been no independent recount since.

A dataset can carry a current release date and still rest on a 6-year-old foundation. When most vendors in the industry stand on that same foundation, the problem stops being a vendor problem and becomes an industry-wide one. Comparing two data providers won’t surface it because both are modeling from the same inputs.

Before you trust a dataset that claims to be recent, ask the provider three things.

1. When was the data refreshed? This is a different question than when it was published. Only the refresh date tells you how current the underlying population count is.
2. Is the data modeled on top of Census data? A model layered over a flawed baseline not only inherits the flaw, but also carries it into predictions.
3. How is income calculated? From the same Census-derived model everyone else uses? A different model or estimate? Income drives rent-to-income ratios and affordability screens, so the answer shapes your entire underwriting output.

If every answer traces back to Census population counts, you haven’t found a fresher source of data. You’ve found the same baseline, repackaged.

CENSAI data is refreshed quarterly, with a 60-day lag for ingestion and quality assurance. That’s the full distance between what happened in a market and when you can see it.

25M+ annual moves. Quarterly data refreshes. Population intelligence in near real time.

Our verified, proprietary population and migration records cover more than 270 million U.S. adults and more than 25 million moves each year. We calculate income data for individual households, building affordability metrics from the ground up instead of relying on a Census national average.

Underwriting models are only as good as the data driving them. There is no reason to work from a 6-year-old, spotty baseline when you can access near real-time population intelligence.

CENSAI data is available through Cherre’s platform, giving underwriters direct access to a dataset built for faster, better-supported decisions.