Nice…. I’m noticing the same thing. Zillow is ridiculously over-valuing the really hard hit areas like spring valley, encanto, eastlake. 30% over an actual sale price is the norm. Seems that their algorithm doesn’t work in a rapidly declining market, where 6-12 month old comps are really meaningless.

Look at this one.

http://www.redfin.com/stingray/do/printable-listing?listing-id=1227310

Listed for sale at $350,000. Zestimate: $613k. Lol. Maybe Zillow is partly to blame for their initial wishing price of $675k.

Price Reduced: 11/21/2007 from $675,000-$725,000 to $499,000-$500,000
Price Reduced: 12/15/2007 from $499,000-$550,000 to $450,000-$550,000
Price Reduced: 01/24/2008 from $450,000-$550,000 to $430,000-$550,000
Price Reduced: 02/07/2008 from $430,000-$550,000 to $399,000-$450,000
Price Reduced: 02/21/2008 from $399,000-$450,000 to $350,000-$399,000

Hi Kris,

This is cause for concern. Zestimates don’t typically have a bias. I’ll have our analysts look into the data and algorithms in San Diego but my bigger worry is that this news probably signals bad times ahead for San Diego home sellers.

Here’s why …

When all sales values are below Zestimate values it’s a sign that something other than normal sales transactions must be driving home prices down. That something is unfortunately probably foreclosures. (Do the sales you listed exclude foreclosures?) Sellers have a huge problem on their hands if foreclosures have started to make the market. If so, falling prices will beget cheaper foreclosures and so on and so forth – with prices in free-fall until foreclosure inventory is sufficiently absorbed so as to no longer make the market. Again, this is typically impossible and could only result from a massive foreclosure inventory – but seeing as that’s the case in most of CA – the impossible has unfortunately become the most likely explanation.

The Zestimate algorithm ignores foreclosure sales values. I assume you do the same thing when picking comps for your CMA’s. Foreclosures are not market-price transactions – at least not in a normal market.

I hope that gives you useful background for conversations you’re having with sellers. Unfortunately, it may also implies that the best advice to sellers right now may be not to sell unless you have absolutely no other option.

Funny aside: my anti-spam word as I post this is “laugh” – I’m sorry this comment is not more positive.

Jacob – You are right that the areas you mentioned are typically off more from their Zestimates, and these are also areas that have been hardest hit by the market correction. Which segues nicely into David’s comment.

David – What took you so long? You’re slipping.

Blaming the fact that all Zestimates are high on just foreclosure activity is wrong, I think. Wouldn’t that fact that the market is declining simply explain this, as Jacob suggests? If your data is from past performance, than it would reason that the Zestimates would continue to be high as prices continue to decline.

On the subject of foreclosures and other distress sales, we do not toss them in the trash pile when estimating probable sale price. As you said (and I agree), they are not relevant in a normal market. We are not in a normal market, therefore they do impact market dynamics and do constitute a “comp.” The appraisers consider them, the buyers consider them, and so the sellers must take them into account. In the little Scripps Ranch area I discuss, bank owned and short-sale homes on the market are measurable, but they do not constitute a majority (or even a quarter) of listings or sales. That my 9% is not terribly dissimilar to your own 7.2 percent for the metro area is case in point that we are not one of the communities most extremely affected.

So, I agree with everything you say, except I think the bigger issue is this: When home values are declining, and for whatever reason, Zestimates will tend to be high; when they are increasing, it is intuitive that Zestimates will tend to be low.

So, I think your anti-spam word might be appropriate. I wonder how much the Sellers will be feeling the love for their Zestimates when they perceive an undervaluation, at that point where we finally start to see a reversal of the current market trend.

Back to the point, I was simply reminding people of what you yourself have said on your site: Zestimates are sort of close a lot of the time, and they are not close at all a lot of the time, but at no time should they be considered 100% accurate or a substitution for the experience and eyes of the agent or appraiser. Read the label before use, do not take before bedtime, and consult a licensed agent if symptoms persist.

I think one reason the Zestimate is high is it ignores sales that it deems “too low”.

These two East Lake houses both sold recently quite a bit under their Zestimates, and both say “Transaction Not Included in Zestimate”.

http://www.zillow.com/HomeDetails.htm?zprop=51081745
http://www.zillow.com/HomeDetails.htm?zprop=17135453

Perhaps that “too low” threshold needs to be adjusted in bubbly areas.

Jakob -

Those were both trustee’s deeds upon sale. Foreclosures don’t suggest market values (at least not in a normal market.)

Hi Kris -

Mortgages are keeping me busy.

No, if the market were simply declining under normal market forces you would not see this pattern in Zestimates. Zestimates should fall evenly either side of sales values. Tomorrow’s market price is normally very well correlated with yesterday’s sales prices. The only reason Zestimates would be consistently high is if they’re blind to the sales that are setting the market – which brings us back to the discussion of foreclosures.

There isn’t much latency in sales transactions on Zillow (normally 2 to 3 weeks.) The Zestimate algorithm estimates the value “today” with an understanding of the fact that the date of the most recent sales information is in the (recent) past. The only trend that Zestimates couldn’t detect would be a radical and instant change in the market from appreciating to depreciating – which is near impossible – and even then, Zestimates would correct themselves in a matter of weeks.

I’ll let you know if we find a systemic issue in San Diego but I still suspect the culprit will be the fact that in this market, San Diego home values are now being set by foreclosures to some extent. The question I have for our rocket scientists is whether the algorithms should be trained on when to not ignore foreclosures.

I had this very zesty discussion with a client last week. Phoenix is a declining market except where their house- it went up 50k according to Zillow. I proceeded to show them home sales in the same neighborhood, closing last week- even got an opinion of value from my appraiser- you know the guy the bank calls for home valuation. But Zillow wins- no price reduction, I guess the spiders can stay in the vacant home for a few more months.

I do not know why it is rocket science on setting a fair price for homes. I simply take the last sale (pre-bubble or use one from pre-bubble years) and add 5% appreciation which is a pretty good deal. I have made a few offers using this methodology in the Ramona/Ryland home area which has not had a comp in 6 months. Unfortunately, I have not had any takers even though 2 owners had plently of equity to make the deal (one had a 30% and the other a 35% return). The offers came in 90K below their asking price. So, the whole neighborhood just waits for that all important comp that will, as one realtor put it, “set the price point”.

Kris, OMG, I saw your Ramona listing. I thought you guys don’t do Ramona. Nice house and I can try my offer: Last sale on Dec 03, 2002 for $450,000, so 5 years at 5% appreciation is 25%. So the offer would be 450K + 112K = $562,000. I guess we are over 100K apart. But, on my side, there are over 100 listings in this small area with many on the market for 6 months already. I am just waiting for a few people to decide that mayber 250 DOM is enough and I will get a $2500 a month mortgage so I can eat out once in a while.

Oh well….. It is 21 days, how long before the first price reduction?

kc

FYI, this is the one that may be the comp people are waiting for:
http://www.redfin.com/stingray/do/printable-listing?listing-id=1048470
Very comparable, with over an acre and upgrades galore…..

kc – This is where your science may break down. The Ramona home you reference was purchased as new construction – No landscaping, window coverings, etc. You can factor in some concession for the $100k in backyard improvements, and that might help reconcile the some of the difference. But, using your methodology, we aren’t comparing apples to apples.

You are correct, however, that San Diego Country Estates has a huge number of listings right now and pricing pressure continues downward. It is not a fun time to be selling a home in Ramona.

Back to your situation, if you have made several offers and are still without an escrow, perhaps you are being a bit too aggressive. I suppose you will argue that prices will eventually soften and catch up with you, but I suspect by then you will have adjusted your own palatable price downward and will again find yourself firing blanks. At some point, you may need to pull the trigger if you really want to buy a home.

As an aside, I hope everyone is appreciating the irony that I am allowing all of these links to Redfin. (It is well know that I don’t love their business model, but I will give them this – Nice search site.)

David G – Thank you for your last comment, which I just saved from the depths of my spam catcher. Funny how your previous comments were deemed worthy but the last one was banished. Perhaps my spam catcher doesn’t agree with your thesis?

Okay – I am starting to buy your argument for why the Zestimates are all tilting to one side. But whether they are all piled up on one side of the see-saw or evenly divided, the fact that they are not precise appraisals which can be taken to the bank still holds.

Thanks for the comments, and good luck “training” your algorithms! Let me know how that goes. I can’t even get my own children to clean their rooms.

Yes, thanks David for the insight into the algorithm.

I think the Zestimate should start including REOs when they become a more significant fraction of the local housing market. In socal we have as many foreclosures as sales. REOs will have a strong say in prices for some time. As such, Zestimes should account for them. Perhaps a simple weighting factor on REOs, where they are weighted proportional to the fraction of homes on the market that are REO.

reo_weight = reos_for_sale/total_homes_for_sale

If all the homes are reos, then an reo gets full weight. if no homes are reo, then a new reo gets zero weight.

So, when can you have this change in?

Jakob -

I really like your suggestion but unfortunately we don’t have the data we’d need to execute on it (i.e. all listings and REO’s.) I have passed this thread on to our statisticians – let’s see what they come up with. This is a highly unusual scenario and so it may not make sense to add this complexity to the algorithm if it’s very seldom applied. And because real sales are now competing with REO’s, they should start to inform the algorithm that prices have dropped further but in this case, Kris’ argument holds and the drop in Zestimates could lag the drop in home prices until REO’s stop making the market.

Kris -

Thanks for fishing me out. I totally agree that software can never replace a local expert when it comes to setting sales prices or deciding offers.

Hi Kris:

Thanks for the reply. Yes, Ramona is a buyers market. But there are so many “vanity” listings (“Excuse me buyer, but Flip this house told me that kitchen remodel is worth x amount of $$ and I will recoup every penny”, [poor listing agent]).

Why buyers would not take 100K profit today (based on a strong methodology to the offer, i.e historic norms) but rather wait 12 months in the hope of bubble pricing returning is beyond me.

I think my offers are in line with the historic variables (rent vs own, standard appreciation rates, etc). I am just trying to use some principals that I think all buyers should have:
1. Stay within about 30-40% of your NET income each month for mortgae payment (with tax incentive this is about $2500 for me)
2. Offer should be based on a above average appreciation rate of 5% per year (pre-2001). Sounds like a great deal, but only for sellers that did not use their house as a ATM.

BTW……….
Here is another topic for your blog. I have been looking for a home for the past few months, researching my target areas, developing a method for offering prices and working on a short list. I also have been working on the financing front as well. I do not want to waste an agents time or effort so I am doing this work since I know what I am interested in. (Yes, I do respect an agents time and would feel guilty going on ride alongs). When my short list and price gaps get to about 70K or so I will then interview a few agents.

So what I would want from what I want in a agent is really be an agressive negotiator and handle the rest. How do buyers agents approach a situation where the house hunter has done all of teh ground work (literally has found the house) and want to hand off the rest of the process to a professional agent?

kc – I guess the short answer is that it will depend on and is negotiable with the particular agent/broker. I will suggest to you what we suggest to all of our clients: Select the best representation for you, and then go about negotiating specifics. If you are not satisfied at that point, you can move on to your next choice. A home is such a huge investment, and (as you point out) the value of having a strong advocate on your side is tremendous. In other words, don’t let the cart drag the horse.

If Zillow can come up with an algorithm that will compare curb appeal, lot appeal, colors, smells, general ambiance and buyers emotions, then they might hit the valuation nail on the head…or not. Technology is a wonderful thing, but zeros and ones have their limits.

Mark, I have absolutely nothing to add to that. Well put.

[...] Gibbons from Zillow recently commented on a post here: The Zestimate algorithm ignores foreclosure sales values. I assume you do the same thing when [...]

In my opinion, the Zillow Zestimates are seriously flawed. Since the algorithm isn’t openly available for scrutiny, we can’t evaluate where the flaws are kicking in. That, in itself, is cause for concern.

We’ve been following this over at the UCR Sloan Center for Internet Retailing, and so far, Zillow doesn’t have a very good answer to the discrepancies.

See for yourself: http://sloan.ucr.edu/2008/03/09/why-are-zillow-zestimates-so-wrong/

Professor Donna L. Hoffman
UCR Sloan Center for Internet Retailing

Professor Hoffman,

Thank you for the comment and for the link to the fabulous blog I didn’t previously know existed! Our dilemma, of course, is that the consumers (our clients) think they carry a lot of weight.

[...] 5– Market Value = Zestimate. Keep Dreaming.By Kris Berg San Diego Real Estate Web Log. [...]

[...] Market Value = Zestimate. Keep Dreaming.By Kris Berg San Diego Real Estate Web Log. [...]