Zelman on housing

[ERICOPOLY]

Zillow's pricing algorithm for my house...

 

Yesterday it was take asking price of $1,825,000 and add $24.  Zillow estimate:  $1,825,024

Today it is take asking price of $1,695,000 and add $2.  Zillow estimate: $1,695,002

 

Question:  did dropping the price $130,000 make Zillow feel more confident in our asking price so they narrowed the gap from $24 to just $2?

 

Why is asking price the input that Zillow uses for value?

[Gregmal]

2 hours ago, Dinar said:

Agreed, I ideally would buy in Oradell since it is a bit less snobbish than Ridgewood, but there is nothing for sale.   I mean Oradell, Woodcliff Lake, Haworth, Glen Rock, Ridgewood, Upper Saddle and Franklin lakes all work, but my preference is for Oradell/Glen Rock/Haworth due to much lower snobbism.  I came to this country with nothing, and I do not want to raise greenhouse kids.  I would walk two miles carrying 40 pounds of stuff in college to save a dollar on a bus fare, I would at least like my kids to not expect a BMW for sixteenth birthday.   

Glen Rock is a woke hell hole. You wouldn’t want to live there now. It’s not what it was 15 years ago. 

[crs223]

7 hours ago, ERICOPOLY said:

Why is asking price the input that Zillow uses for value?

 

If I were in charge of writing the zestimate algorithm, i would base the price on comps, sq footage, property tax assessment, lot size, and other “public” info.

 

None of this captures “private” information such as: does the house have mold/termites, was the kitchen remodeled, is it a fixer-upper, etc.  The asking price encompasses all of this.

 

The asking price was set by a professional agent and motivated seller who factor tangibles and intangibles.  The asking price is a goldmine of info and certainly better than some algorithm written by a computer geek 5,000 miles away.

 

My algorithm would slightly adjust the asking price based on recent local trends for homes to sell above or below ask.  Probably also look at the particular listing agent’s history of getting the asking price (“historically this agent asks for 7% too much”).

 

[ERICOPOLY]

1 hour ago, crs223 said:

 

If I were in charge of writing the zestimate algorithm, i would base the price on comps, sq footage, property tax assessment, lot size, and other “public” info.

 

None of this captures “private” information such as: does the house have mold/termites, was the kitchen remodeled, is it a fixer-upper, etc.  The asking price encompasses all of this.

 

The asking price was set by a professional agent and motivated seller who factor tangibles and intangibles.  The asking price is a goldmine of info and certainly better than some algorithm written by a computer geek 5,000 miles away.

 

My algorithm would slightly adjust the asking price based on recent local trends for homes to sell above or below ask.  Probably also look at the particular listing agent’s history of getting the asking price (“historically this agent asks for 7% too much”).

 

 

Makes sense.  But why the $2 discrepancy?

[crs223]

This is an issue of accuracy vs precision.

 

The zestimate algorithm (and any other numerical algorithm such as your car’s cruise control) produces overly “precise” results.

 

I bet the internal algorithm really wanted the zestimate to be $2.143 higher than your asking price.  of course such a number is not “accurate”.  But is sure is precise.  The algorithm is just adjusting your asking price by the 4 week moving average of asking-price discount in your zip code and adjusted by an additional penalty because your house had a price cut… historically people who reduce their asking price are going to get punished 0.3% at the negotiating table.

 

Of course zillow couldn’t adjust the zestimate by $2.143 because it would “expose” that the algorithm is overly precise and not necessarily accurate.  So they attempt to “hide” behind rounding the value to the nearest dollar.  Which is obviously still too precise when dealing with a $1.7M home.

 

I bet there is an outstanding issue in the inbox of a zillow developer: round to the nearest thousand instead of to the nearest dollar.  then some other algorithm PhD type is going to make the rounding value depend on price.  $1M homes are rounded to the nearest thousand.  $10M homes are rounded to the nearest $10k.

 

i work on algorithms like this at my job.  it’s a bunch of geeks guessing how to make an estimate… which produces “precise” results that are not always “accurate”… and then collaborating on the whiteboard about how to improve the algorithm to be more “accurate”.

[Pelagic]

10 hours ago, ERICOPOLY said:

Zillow's pricing algorithm for my house...

 

Yesterday it was take asking price of $1,825,000 and add $24.  Zillow estimate:  $1,825,024

Today it is take asking price of $1,695,000 and add $2.  Zillow estimate: $1,695,002

 

Question:  did dropping the price $130,000 make Zillow feel more confident in our asking price so they narrowed the gap from $24 to just $2?

 

Why is asking price the input that Zillow uses for value?

 

Fascinating. The same applies for a lot of houses for sale in my neighborhood with a $1 to $3 discrepancy between Zestimate and list price. It seems Zillow won't put a Zestimate higher than list by more than a couple dollars. Presumably their algorithm has it priced lower internally but they're not confident enough in the algorithm to say that. Then there are a handful of houses with Zestimates noticeably lower (20-50k) than list price so presumably the algorithm feels it is right or internally has a Zestimate even lower but is limited by a % deviation from actual list price.

 

Will have to do some more searching but I haven't found a house listed with a Zestimate noticeably higher than listed price. It seems to default to the $1-3 bump. Did just find one $309 higher on an $800k house.

 

On one hand a Zestimate higher than list would make the listing realtor look bad but on the other would likely have buyers flocking to it in droves thinking they're getting a deal. Every realtor I've ever talked to hates how fixated buyers can be over Zestimates. 

[KJP]

11 hours ago, Dinar said:

I probably should add it to the list, but could it be too high end?  Gladwyne seems to be a bit snobby.

 

That's probably fair, but you will find that in most of the areas you mentioned, e.g., Wayne.   In any event, LMSD, particularly the parts closer to the city like Merion and Penn Wynne, is quite overcrowded. 

 

One area you might add to your list is Lower Gwynedd.  It's in Wissahickon SD, which isn't as lauded as Tredyffrin, Radnor, or LMSD but I know parents that have moved there from LMSD and prefer it.

 

Newtown/Bucks County also has a much different feel than the Main Line.  Much more open space, orchards, etc.  And Doylestown and New Hope/Lambertville are nice options as far as those size towns go.  If you wanted a similar feel but west of the city, there are areas (i) between Newtown Square and West Chester, and (ii) northwest of Rt. 202 around Methacton and Worcester that have a similar feel.  They are probably a 1.25 hour (at least) commute into the city if that matters.

[crs223]

25 minutes ago, Pelagic said:

Every realtor I've ever talked to hates how fixated buyers can be over Zestimates

 

That is interesting and terrifying.  There is just no way a centralized algorithm can accurately price individual houses around the world.  but because it exists, it gets used (and apparently trusted).  Just the sort of thing an engineer hates, but management loves!

[Dinar]

21 minutes ago, KJP said:

 

That's probably fair, but you will find that in most of the areas you mentioned, e.g., Wayne.   In any event, LMSD, particularly the parts closer to the city like Merion and Penn Wynne, is quite overcrowded. 

 

One area you might add to your list is Lower Gwynedd.  It's in Wissahickon SD, which isn't as lauded as Tredyffrin, Radnor, or LMSD but I know parents that have moved there from LMSD and prefer it.

 

Newtown/Bucks County also has a much different feel than the Main Line.  Much more open space, orchards, etc.  And Doylestown and New Hope/Lambertville are nice options as far as those size towns go.  If you wanted a similar feel but west of the city, there are areas (i) between Newtown Square and West Chester, and (ii) northwest of Rt. 202 around Methacton and Worcester that have a similar feel.  They are probably a 1.25 hour (at least) commute into the city if that matters.

KJP, thank you very much.  Newtown/Buck County was a choice due to its Russian community, which values education and abhors political correctness.  

[Dinar]

9 hours ago, Gregmal said:

Glen Rock is a woke hell hole. You wouldn’t want to live there now. It’s not what it was 15 years ago. 

What a pity, thank you very much for the heads up.

[Pelagic]

5 minutes ago, crs223 said:

 

That is interesting and terrifying.  There is just no way a centralized algorithm can accurately price individual houses around the world.  but because it exists, it gets used (and apparently trusted).  Just the sort of thing an engineer hates, but management loves!

 

I think a lot of it comes down to Zillow infringing on their turf. Choosing good comps and providing an accurate valuation is one of the major value adds of a good realtor. And realtors are going to present their value estimate in the form of a range and usually provide a link to the comparable properties so you can kind of get a feel for how they arrived at their range. Zillow just puts up their Zestimate (they do present a range but you have to click through to see it) without much info on how it was derived. But Zestimates are so ubiquitous that every aspiring first time buyer browsing Zillow all day just accepts them as is.

[crs223]

5 hours ago, Pelagic said:

But Zestimates are so ubiquitous that every aspiring first time buyer browsing Zillow all day just accepts them as is.

 

Monetize the unfounded trust: for $1,000 zillow will increase your zestimate by 10%.

 

…. actually that would destroy the trust… but there must be some way to monetize this blind faith.  perhaps it’s to get a job as an algorithm developer at zillow.

[Gregmal]

I recall a number of years ago there was some issues with realtors and brokerages potentially suing Zillow over the Zestimate stuff. I don’t recall the grounds but the gist was that Zillow was interfering with their business by misleading with the estimates. 
 

I personally don’t find them

much use. They still have rent estimates for some of my properties that are lower than 2013 prices. Like how do you see an MLS rental listing, entered, just like a for sale listing, and just ignore that data point? I recently pulled up one comparable unit and it said $1450 a month was market rate rent. Except in the entire area the cheapest available rental options all year were at least $2000.

 

Nevertheless we continue to see how Zillow is embedded in everything housing and it’s unavoidable. You’d think eventually just maybe they capitalize on some of it. 

[Blugolds]

1 hour ago, crs223 said:

 

Monetize the unfounded trust: for $1,000 zillow will increase your zestimate by 10%.

 

…. actually that would destroy the trust… but there must be some way to monetize this blind faith.  perhaps it’s to get a job as an algorithm developer at zillow.

 

 

 I just recently sold a lake property, I sold for over asking and the Zestimate. Initially I thought I was a little late to the party, I planned on listing last year but with supply chain issues, the new windows I ordered for the place took 6 months to deliver, then 2 of the 11 windows delivered with cracks, so it took 3 more months for them to send 2 replacement windows out. I wanted it listed last summer and ended up having to wait till this summer to list and it worked out all right. 

 

There was a slow down in the market that we felt. I honestly didnt thin kit would make it past the first week. Only property available initially when listed on this recreational lake, prime lake lot level to the water, its own peninsula and kitty corner from the local supper club etc. Arguably the best lot on the lake. 14 showings in 2.5 weeks...one offer. Initial offer was insultingly low and was immediately rejected...counter came back still low but was a start for negotiations. ↓↑ 5K increments until I wouldnt budge and they wouldnt budge. They walked away..Fed did the next rate increase and they came back with an offer that matched my final offer and I accepted. The offer they submitted before the fed rate hike was contingent on them securing financing for 30 yrs at 5.75%....the offer that matched my final submitted after the fed hike was contingent on financing at 6.25%. 

 

Meanwhile another place came on the market that they were also interested in. Asking about $50k more than I was with 700sq ft more living space, slightly dated, lot wasnt as nice but was a nice lot. I was concerned that they would start pursing that property. That property was listed at the Zestimate. 

 

Long story short, with the markets slowing down, as soon as my property closed they dropped their asking price the $50k to match mine...it sat for another 2 weeks and was pending...I got an email yesterday that it had closed/sold and the sales price was $50k LOWER than mine sold for and nearly $100k less than the Zestimate. 

 

I agree that first time home buyers or the unsavvy probably give more weight to the Zestimate, but at the end of the day, the market sets the price, regardless of what you ask, or what Zillow says. I think when properties were selling the first day and everything was going gangbusters maybe the people looked to zestimate for conformation, but many of those sold for OVER the zestimate...now that people are looking at things perhaps a little closer, I am seeing several properties selling for LESS than the zestimate. 

 

 I also have noticed that there are several changes in the zestimate numbers before you list, after sales in the area, after listing, during listing and after sale. I honestly dont think the algo is dialed in accurately. I have a duplex in another area that would confidently sell for over $100k more than the Zestimate just based on conservative cash flow numbers and other comps in the area. I dont think it accurately knows what the property is.  

 

Basically I think the Zestimate is trash on both ends. Could influence naïve buyers, but any decent realtor, or experienced buyer or seller probably doesnt pay attention to them. 

[lnofeisone]

14 hours ago, crs223 said:

 

If I were in charge of writing the zestimate algorithm, i would base the price on comps, sq footage, property tax assessment, lot size, and other “public” info.

 

None of this captures “private” information such as: does the house have mold/termites, was the kitchen remodeled, is it a fixer-upper, etc.  The asking price encompasses all of this.

Without disclosing the company, there are now efforts underway to incorporate image processing into pricing algorithms. Things like fixer-uppers, remodeled kitchen, etc. are now being detected with fairly consistent accuracy and recall. You can even get reasonably good numbers guessing what brand cabinetry the house has (Wellborn vs. Ikea, for example). Obviously, things like mold and termites are still detected by physical inspection only. 

[ERICOPOLY]

Zillow has this on my house's listing:  "We estimate this home will sell faster than 87 % nearby."

 

It has been 2 full weeks now and we've had only one agent request a showing, and that was the day after we listed it. 

 

On Sunday we had an open house and only 5 couple showed up, all from the Sacramento region.  None from the Bay Area.  Unfortunately, homes like this one normally are supported by Bay Area buyers.  It was those buyers that pushed prices up in the first place.

[thepupil]

House near me. 3/3, 1800 sf, 40's build, updated but nothing spectacular, would have been $850-$900K pre-covid. Listed at $1.05mm, sold $1.15mm w/ 3 days of listing w/ multiple offers. the zestimate was about that price as of January 2022, but peaked at $1.3mm in May 2022. appreciation in my area was always a bit slower than the hottest of markets and i'd expect things to stagnate for a little given where rates are, but disaster has yet to hit the aspirational early to mid 30's beltway bandits. 

 

EDIT: all of this is driven by close to zero supply for something like this which is about 70% of the median SFH price in the area. Overall Inventory in the zipcode is down 20% YoY and probably much lower than 2 years ago. 

Edited August 26, 2022 by thepupil

[Spekulatius]

Do people really believe that something like a zestimate has less than a +/- 5% error? It’s probably less when you live in large community where houses constantly turn over and they are all more or less the same, but generaly, I think a 5% error margin is a good guess.

[Dinar]

In Manhattan, for roughly half a dozen apartments I looked at, it was off by 15% on average.

[valuehawk91]

I didn't know who Zelman was. I ended up coming across one her recent interviews/podcast, thought I would share.

https://macrohive.com/hive-podcasts/ep-122-ivy-zelman-on-the-coming-us-housing-crisis/ 

[Gregmal]

Zelman isn’t wrong but she’s also not right. The major crux of the thesis has to do with demographics and household formation. But they have for almost a decade. So it’s akin to folks saying the Fed has propped up the stock market for the past 10 years. Broken record. This time is different. The whole 9. If they propped up the market and it was so obvious why were you in cash or short?
Demographic and millennial trends didn’t change during COVID, what changed is the urgency and obviousness to many existing homeowners as well as potential ones with respect to what they want to deal with and where they wanna live. Is it really a coincidence Florida and Texas are the biggest winners? You need a lot of building to satisfy many of those markets. It doesn’t matter if the 30 year olds wait another 5-10 years to start families. My entire life, as a tri state area resident, its always been a given that folks leave and go to Florida at a certain age. If they didn’t have 2nd homes, they’d retire there. Now you have young and middle age families deciding to make the move. Literally everyone I know not smack dab in the middle of the 5 boroughs views the move as a matter of when and not if. 
 

Housing markets have always been regional. This has and will continue to be very apparent. Only so many people want to pay 3-4% of their home value in taxes every year while having to deal with woke politics. Two of the major reasons I’ve loved NJ is safety of my area and quality of school system. Start attacking those things and you literally have nothing left. 

[Spekulatius]

I think housing has seen a pull forward or demand during COVID-19 that is now waning. The fundamental LT issue is that we don’t build enough single family homes and likely won’t in the future, due to land use constraints.  I think in the future we need to go vertical and that means multi family homes.

 

The Southwest has issues with water scarcity and the Southeast with climate issues (rising sea levels, increasing temperatures and increased storm frequency and severity). Those are multi decade trends, but in my opinion, they are quite real.

[Gregmal]

https://www.zillow.com/homedetails/50-W-Sandlefoot-Ln-Inlet-Beach-FL-32461/251160975_zpid/

 

Another one, under contract inside a week after a super bearish price cut resulting in only a $100k profit from the price paid....a year ago. A nice, but totally unremarkable home as well. 

 

GFC 2.0. 

[ERICOPOLY]

On 8/28/2022 at 1:10 PM, Spekulatius said:

The Southwest has issues with water scarcity

 

It's a matter of priorities.

 

Irrigated agriculture is the largest user of water in Arizona, consuming about 74 percent of the available water supply.

https://new.azwater.gov/conservation/agriculture#:~:text=Irrigated agriculture is the largest,of the available water supply.

[sleepydragon]

Anyone looking at PHM? Seems cheap?