Nobody knows how many people a water system serves

To say what share of Americans drink fluoridated water you first need to know
how many people each water system serves — and that number is an estimate
built on an assumption.

**The CWS service populations that states report to WFRS are estimates,
typically based on an assumed number of people residing in each household
connected to a water system.**

Household size varies; the assumption does not. **Because these estimation
methods are subject to either under- or over-estimation, CDC adjusts the
reported populations** rather than publishing them as given.

The scale of the data

More than 50,000 water systems report into the **Water Fluoridation
Reporting System (WFRS)**, the tool CDC maintains to help states manage their
fluoridation programmes. About 18,000 are reported as fluoridated, meaning
they either adjust fluoride to the recommended 0.7 mg/L, have **naturally
occurring fluoride at or above that level**, or **buy optimally fluoridated
water from a neighbouring system**.

Each year CDC works with state programmes to **identify and correct
discrepancies between WFRS and the EPA's Safe Drinking Water Information
System**, and biennially publishes the National Water Fluoridation
Statistics.

The correction

Two independent published estimates are combined to get a state total that
does not depend on the household assumption at all:

The USGS figure is the one doing the subtle work: it **accounts for residents
who receive water through private sources, such as private wells** — people
who are in the census but on no water system's books.

State total = Census state population × USGS percentage on CWS.

Then the same multiplier for everyone

That total is compared with the sum of what the state's systems reported, and
the ratio becomes a multiplier applied to **every individual system in the
state**:

c = (adjusted state total) ÷ (total reported CWS population)

**The fluoridated population is then the sum of the adjusted populations of
the fluoridated systems**, and the headline percentage is that sum divided by
the adjusted state total.

Note what this method does and does not claim. It does not find which systems
over-reported and which under-reported — it has no way to know. It corrects
the total, then distributes the correction proportionately, so a
system's share of its state is exactly what it reported. The individual
numbers are no more accurate afterwards; the state figure is.

Before 2016, a simpler method was used for the 2012 and 2014 statistics,
which is why comparisons across that boundary need care — the difference is
described in a published paper on the effect of the adjustment.

Source: Centers for Disease Control and Prevention.