- Non arms-length transactions such as foreclosures, sales to family, etc.
- Properties whose use has changed, such as from single family to condo
- Transactions immediately preceeding or subsequent to substantial remodel or physical changes
- Suspected data errors when the magnitude of price change is abnormally large
- High turnover frequency, or too many sales too soon of a single home
After suspect data is removed, the remaining "sales pairs" are considered. A sales pair is two prices for the same home sold twice. The goal is to isolate price changes without too many external forces other than overall economy in the mix. That's why the characteristics listed above will cause a transaction to be removed from the data set. Then they go further and assign weights to bring the most accurate sale pairs to the top with more importance. Weighted factors include:
- Importance weight is reduced if price change is out of the norm for the area, but not so great as to be removed
- Long intervals between the two sales drop the importance weight, as the home could have been remodeled
- Importance weight increased the closer the initial price is to the area aggregate/average value (normal for area)
Those are the high points of the index construction and data selection. As you can see, a great deal of care was taken to arrive at a "sales pair" of prices for a home that had not been significantly modified or sold in other than an arms length transaction. The goal is to get the result closest to a purely economic price change without other influences.
