Comparing Pre- and Post-sale Estimates of the Price of a House
When I bought my house several years ago I was unsure how much money it would take to actually buy it. There was the list-price, of course, but I didn’t expect that to be the price the seller would accept. Zillow and Redfin estimated two different prices, both slightly higher than the list price, and my realtor suggested a forth based on comprable homes. In the end we made up a number combining all four prices, nudged it up a bit as a hedge, and had our offer accepted. It left me wondering if there was a better way to predict prices, and if Zillow and Redfin had been right with their predicted prices.
Recently a house in my neighborhood put up a “For Sale” sign which prompted me to look online for the listing. I couldn’t find one. None of the online real estate brokers had picked it up yet. I realized I had a chance to compare their current price estimates with the actual listing and sales price.
You can find the Jupyter notebook used to perform this analysis here (rendered on Github). The data can be found here.
I recorded the pre-listing estimates for four brokers: Redfin, Realtor.com, Zillow, and Xome. Xome provides both an estimate and a high and low range. All the others only present a single estimate. I also collected the same estimates after the listing was picked up and again after the sale was complete. The data is summarized in the table below:
I could not find Realtor.com’s estimate after the listing went up, so it is not included. Xome did not change their estimate after the listing was posted.
The house was listed at $948K and sold for $1070K.
Here a plot comparing the three estimates from each company to the list and sales price:
The three price estimates are:
- Pre-listing: Before the listing was picked up by the brokers, represented with a circle: ●
- Post-listing: After the listing was posted but before the sale, represented with a triangle: ▲
- Post-sale: After the sale was made public, represented with a square: ■
I have slightly offset the date points for each company—with the circle on left, the triangle in the middle, and the square on the right—to give a quick indication of how the price trended in time if you read from left to right. The Xome estimates include error bars for their high and low estimates. The listing and sale price are shown as lines. I have colored each company’s estimates to match their brand color. The companies are sorted from lowest to highest pre-listing estimate.
Zillow and Redfin strongly disagree about what the value of the house is initially, with a difference between their estimates of $200K. Both estimates poorly predict the final sales price, with Zillow underestimating by $132K and Redfin overestimating by $74K.
Both estimates revert towards the list price when it is posted, with Redfin’s slightly higher than the listing and Zillow’s slightly lower. This makes some sense, as the list price contains new information about the current market conditions and more importantly about the condition of the house and property relative to its neighbors. However, in this case the list price was obviously too low and likely intended to entice buyers.
Xome’s pre-list estimate is the closest to the sale value, missing by just about 3%, although they gave themselves a lot of room with their uncertainty. Here are the four pre-listing estimates ranked from lowest to highest absolute error:
|Company||Absolute Error||Percent Error|