Zillow Denver and Individual Neighborhood Estimates

[Kevin Grasse]

Hi All,

I was browsing a Zillow listing today and noticed that in the Home Value vs. Time graph, Zillow is no longer showing the 2 additional lines for Denver Average Price and the specific Neighborhood Average Price. Only the individual home value estimate line is being displayed. I'm hoping this is just a bug, or is being updated. 

Has anyone else noticed this and has it been like that for a while? 

I have an old data set of average neighborhood prices that Zillow used to provide for free download, but stopped providing it, probably around the time of the iBuys when they realized it was valuable data and misinterpreted it to lose millions, lol. I'm trying to add data for today's estimates, so that I can correlate to the old and see neighborhood trends. Unfortunately, I'm not seeing an easy way to do that now. Hoping this isn't their next step of taking away data...

Kevin

[Steve K.]

I guess if you're just looking at market trends it's not such a big deal, but I wouldn't use Zillow data for much these days. According to themselves, Zillow estimates have a less than 38% chance of being within 5% of the actual home value, and their average is 80%. In the Denver market where the average home price is north of $650k now, that means the best we can hope for is within ~$130k of the true market value, which is too far off to be valuable data. 

Anecdotally, they are currently off on all of my personal Denver properties by several hundred thousand. Ever since they famously lost half a billion during the second half of last year by trusting their own algorithms to buy properties, I've been cross-referencing their "Zestimates" every time I run comps just to compare, out of curiosity. They're often off by several hundred thousand and in one case recently the Zestimate for the subject property was $1.5M and we went under contract for $2.5M (off market deal so not even any competitive bidding in this case). It appraised at $2.55M, so Zillow was off by $1.05M.

I think it's time to ask whether statistical/AI-based price prediction is even worth looking at. We've always known that property pricing is based on many things besides number of beds and baths in a home, and that algorithms don't know what the difference in price is on one side of the street vs. the other, one side of the tracks vs. the other, city or county, one school district or another, updated inside or not, what the place smells like, etc. But it's gotten especially inaccurate lately with such dynamic market conditions. With the way prices have been rising so fast, even CMA's created last week are no longer accurate. Bidding wars on certain properties are completely unpredictable and can drive sale prices way up above what current comps support. It's impossible to create an accurate CMA right now without first calling all the listing agents of the comparable properties under contract to get a sense of what they're going to close at, and making that adjustment to the comp. Bots don't do that. I wouldn't count on Zillow for anything other than getting very rough, ballpark data, and with the understanding that it may not even be in the ballpark.

[Matt M.]

Have an agent pull historical data for you. 

Zillow is a lead generation company that sells leads to agents.

[Kevin Grasse]

@Steve K. Ya, I'm pretty sure my data was purely based on sales, no predictive analysis/Zestimate. Just an aggregate of the sales data for each neighborhood, which is still subject to the variabilities you describe, but I was just looking to find some neighborhood trends. I agree that there are too many variables to use any statistical/AI-based price prediction, though.

@Matt M. Is that just raw sales data? I was just looking to plug in a few numbers to update my spreadsheet and see neighborhood trends over the past couple years, not trying to build a whole model. 

[Duane Alexander]

Quote from @Steve K.:

I guess if you're just looking at market trends it's not such a big deal, but I wouldn't use Zillow data for much these days. According to themselves, Zillow estimates have a less than 38% chance of being within 5% of the actual home value, and their average is 80%. In the Denver market where the average home price is north of $650k now, that means the best we can hope for is within ~$130k of the true market value, which is too far off to be valuable data. 

Anecdotally, they are currently off on all of my personal Denver properties by several hundred thousand. Ever since they famously lost half a billion during the second half of last year by trusting their own algorithms to buy properties, I've been cross-referencing their "Zestimates" every time I run comps just to compare, out of curiosity. They're often off by several hundred thousand and in one case recently the Zestimate for the subject property was $1.5M and we went under contract for $2.5M (off market deal so not even any competitive bidding in this case). It appraised at $2.55M, so Zillow was off by $1.05M.

I think it's time to ask whether statistical/AI-based price prediction is even worth looking at. We've always known that property pricing is based on many things besides number of beds and baths in a home, and that algorithms don't know what the difference in price is on one side of the street vs. the other, one side of the tracks vs. the other, city or county, one school district or another, updated inside or not, what the place smells like, etc. But it's gotten especially inaccurate lately with such dynamic market conditions. With the way prices have been rising so fast, even CMA's created last week are no longer accurate. Bidding wars on certain properties are completely unpredictable and can drive sale prices way up above what current comps support. It's impossible to create an accurate CMA right now without first calling all the listing agents of the comparable properties under contract to get a sense of what they're going to close at, and making that adjustment to the comp. Bots don't do that. I wouldn't count on Zillow for anything other than getting very rough, ballpark data, and with the understanding that it may not even be in the ballpark.

 What sucks about this is that most home owners use the inflated zestimate to gauge the value of their properties which usually becomes their asking price.