Year Over Year Cost Movement by Care Type and State

How to read this

These are modeled medians, not a price index. The comparison is like for like because both periods run through cost model cost-model-v1.4.0, which means the movement reflects the underlying survey and the wage index applied to it rather than a change in our own arithmetic.

Care type figures are the median of the state changes, not a national average weighted by population or by beds. That keeps one large state from setting the national line on its own, and it is why the state count is published against every care type row.

A survey whose respondent panel shifts between vintages can move a median without any provider changing a price. Nobody working from the same published survey can separate those two effects, and we do not claim to.

Only states priced in both periods are compared. A state that entered or left the model between periods is absent rather than carried forward, so a coverage change cannot show up here as a price change.

Methodology

We compare the modeled monthly median for the two most recent model periods we publish, at the state level, for every care type present in both. Only states priced in both periods are compared, so a state entering or leaving the model between periods cannot masquerade as a price change.

The comparison is only like for like when both periods run under the same cost model version, so the page reads the versions behind both periods on every request and compares them rather than assuming they match. When they match, the change reflects a change in the underlying survey vintage and the wage index applied to it, not a change in how we model. When they do not match, the piece says so before it reports any movement, and it carries the version behind each side in the table and the download.

Care type figures are the median of the state changes for that care type, so one large state cannot pull the national line on its own. The count of states behind each figure is published beside it.

Sitewide cost and data method: Methodology · Data sources · Corrections

What this cannot tell you

This is movement in a modeled median between two survey vintages. It is not a price index, it is not seasonally adjusted, and it is not what any single community charged in either year.

A survey with a different respondent panel between vintages can move a median without a single provider changing a price. We cannot separate that effect from real price movement, and neither can anyone using the same published survey.

A comparison across two model versions is not like for like, because part of the movement would then be a change in our own arithmetic rather than a change in the survey. The page checks the versions behind both periods each time it renders and names what it found. If they ever differ, it leads with the mismatch instead of the movement.

Memory care carries an applied premium band on top of the assisted living base rather than a separately surveyed rate, so its movement tracks the assisted living movement closely by construction.

A jurisdiction priced in only one of the two periods is excluded rather than carried forward at its old value, so the state count can be lower than the number of jurisdictions we price overall.