How a house price index actually works
If you look up the house price index for your ZIP area, you get a number like 318.4. It is not a price. It is not an average sale price. It is not the value of any particular home. People reasonably assume it is one of those things, and then the number confuses them.
An index is a level. Somebody picks a starting period and calls it 100. Every figure after that says how far the thing being measured has moved since. An index of 318.4 means prices are roughly 3.18 times the base period. On its own that is nearly useless. The value is entirely in comparing two points.
The problem an index solves
Say you want to know whether homes in a city got more expensive last year. The obvious approach is to average what sold this year and compare it to the average last year. That approach is badly broken, and the reason is worth understanding.
Suppose a developer finished a tower of large, expensive condos and they all sold in the first quarter. The average sale price jumps. Nothing about any existing home changed. The average moved because the mix of what sold changed, not because prices did. This works in reverse too: a wave of starter homes coming to market can make a rising market look like a falling one.
Average sale price answers "what did people spend," which is a real question but a different one from "what happened to values."
Repeat sales
The fix is elegant. Instead of comparing different homes to each other, compare the same home to itself at two points in time.
A house sells in 2016 for $300,000 and again in 2024 for $465,000. You do not need to know anything about the house. You do not need its square footage, its condition, or its street. It is the same house both times, so the change is a clean signal about the market. Do that across millions of properties, weight it properly, and you get an index that is immune to the mix problem.
This is why an index can be honest about a market while knowing nothing about any individual address. It is measuring change, not level. It never has to answer "what is this house worth," which is a much harder question.
Where the FHFA index comes from
The index this site uses is the Federal Housing Finance Agency's All-Transactions House Price Index. It is built from repeat transactions on mortgages bought or guaranteed by Fannie Mae and Freddie Mac, which is an enormous sample going back to the 1970s. It includes both sales and appraisals, which is what the "all transactions" part means.
It is published quarterly, it is free, and it is in the public domain. Anyone can download the same file we use and check our arithmetic.
What it leaves out
Because the sample is loans bought or guaranteed by Fannie Mae and Freddie Mac, homes financed outside that system are underrepresented. That means cash purchases, jumbo loans above the conforming limit, and some government-backed lending. In an area dominated by very expensive homes or by cash buyers, the index is describing a slightly different slice of the market than the one you are in.
A repeat sales index also cannot see a home that has never sold twice, which quietly excludes new construction until it trades again.
Three-digit ZIP areas
We use the series published for three-digit ZIP areas: everything starting 282, for example, rather than 28210 specifically. FHFA labels this series developmental, and there is a good reason it is not finer grained. Cut the geography small enough and there are not enough repeat sales in a quarter to say anything, so the number starts measuring noise.
FHFA handles thin samples by substituting the wider metropolitan or state non-metropolitan index for that area. So an area figure can quietly be a regional figure. Every report on this site shows which area produced it.
What this means for your report
Your estimate is your purchase price moved by the change in your area's index. It inherits every strength and every limitation above. It is a reasonable answer to "roughly how much have homes like mine moved since I bought," and it is not an answer to "what is my house worth." Those really are different questions, and only one of them can be answered by arithmetic on public data.