Recorded home sales volume in San Francisco fell 17.8% on a year-over-year basis last month (410 recorded sales in November ’10 versus 499 sales in November ‘09), down 6.0% as compared to the month prior which was down 21.2% on a year-over-year basis.
For context, November sales figures for San Francisco from 2004 to 2008 were 682 (2004), 658 (2005), 568 (2006), 479 (2007), and 340 (2008). And on average, from 2004 to 2009 sales volume has declined 7.4% from October to November.
San Francisco’s median sales price in November was $680,000, up 4.6% compared to November ’09 ($650,000) and up 4.3% compared to the month prior.
For the greater Bay Area, recorded sales volume in November fell 11.2% on a year-over-year basis, down a nominal 0.2% from the month prior (6,111 recorded sales in November ’10 versus 6,878 in November ’09 and 6,122 in October ’10) as the recorded median sales price once again fell by 1.8% on a year-over-year basis, down 0.8% as compared to the month prior.
At the extremes, San Francisco recorded the greatest Bay Area drop in sales volume last month (a loss of 89 transactions) while Napa recorded no change in sales volume but a 21.1% decrease in median sales price on 104 transactions.
As always, keep in mind that DataQuick reports recorded sales which not only includes activity in new developments, but contracts that were signed (“sold”) many months or even years prior and are just now closing escrow (or being recorded).
∙ Bay Area November Home Sales, Median Price Down [DQNews]
∙ San Francisco Recorded Sales Activity Down 21.2% YOY In October [SocketSite]
∙ San Francisco Recorded Sales Activity Down 12.3% In August [SocketSite]
∙ He’s It’s Back: California’s $10,000 Homebuyer Tax Credit Returns [SocketSite]
∙ FHA: Not Out Of The Woods But Feeling A Little Less Lost [SocketSite]
It’s nice to see that large 4.6 percent price increase.
Although there are some alchemists on this board who will be able to explain mathematically that that 4.6 percent increase is really a dramatic drop.
I got an above average grade in Statistics, so I understand what median means.
“It’s nice to see that large 4.6 percent price increase.”
Excellent! You are ready for lesson 2 in statistics.
Do you see how the line wiggles around? How it goes up and down and up and down and up and down and up and down and up and down on short time scales?
That’s what we call … noise. It’s the imperfections and random fluctuations that sit on top of the signal and obscure our ability to see the signal.
Usually we wait for a change that is much larger than the noise before we conclude that anything has happened.
Given the noise that we see in the median plot the most that we can conclude is that median prices rose until 2005, were flat until 2008, then had a decline until 2009, and have been flat since.
According to redfin, home sales are down 16.7%, but condo sales are down 43.6% (not a typo) from the prior year.
So the median gets pushed up when prices fall because people are using the lower prices to trade out of condos into houses.
There is your higher median. Prices are down. Prices are falling and will fall further. Medians can rise in either scenario.
@tipster: I have to call it, but you are full of BS. Prices are not down YoY and you know it. I posted Condo and SFR homes in another thread straight from MLS. I’m sure you read it, but here it is again:
SFR-Sold Med. #
11/10 781,750 172
11/09 796,444 206
% Chg -1.84% -16.50%
Condo-Sold Med. #
11/10 687,500 134
11/09 676,000 193
% Chg 1.70% -30.57%
Also, sq. ft. prices for both types of properties are also flat YoY so the argument that people are buying bigger houses and that is why median is up won’t fly either.
SFR median is up Oct to Nov from $733k to $781k and is also up from the $710k in Sept, which puts Nov right on the 12 month moving average.
Condo median price was $687k in Nov, $690K in Oct and $649k in Sept, which puts Nov just a smidge over the 12 month moving average.
Finally, you should ask a RE agent to get you all the MLS data like I do in a nice spreadsheet form with graphs and all instead of relying on a secondary source like Redfin. I’m sure someone would do it for you if you asked nicely. I would post the nice graphs I get with 12 month moving averages, etc. if I could.
Skirunman,
not to be picky, but unless I’m reading your data wrong, or you misposted it, your data actually disproves your own point.
1)
Prices are not down YoY and you know it.
clearly your data shows that SFH prices are down YOY.
2)
also, your neither proves nor disproves the idea that it is mix that is primarily affecting the increase in OVERALL (combined SFH/TIC/Condo) median price over the last year.
in 11/2009 your data shows that there were 206 SFH sales with median of $796k, and 193 condo sales with median of $676k.
in other words, there were about equal numbers of SFH and condo sales. (51.6% SFH, 48.4% condo) And thus, the median will lie somewhere between the $676k and $796k figure… likely very close to right in the middle actually given the near 50/50 distribution of condo/SFH sales.
however, in 11/2010 you see that sales numbers are much more lopsided TOWARDS the higher priced SFHs.
176 SFH sales with median of $781k versus only 134 sales of $687k.
(56.8% SFH vs 43.2%)
Thus, the median will be somewhere between the $687k and $781k figure… but given that we have significanly more SFH sales the median will move towards the SFH median…
that said, the 2010 SFH median was lower than the 2009 median.
which is a stronger force? the decrease in SFH median price? or the change in ratio of SFH to condo sales?
I dunno… we’d need the raw data to calculate.
but clearly a change in ratio of SFH to condo sales can drastically affect overall median sales figures… in fact BOTH SFH median sales prices AND condo median sales prices can drop with a RISE in overall median prices!
but you already knew that.
(I’m happy to show a quick tutorial about how both SFH and condo median prices can fall with a rise in overall median for those who desire it).
my own personal thought:
SF RE improved dramatically from end 2009 to early 2010 based on governmental housing support programs as well as the “green shoots” technical recovery.
it is now at a “souffle” level, and incoming data will continue to worsen. (negative change in second derivative as predicted by me back in summer during the “we’re up 18% YOY” days has been ongoing for months… possible negative change in actual pricing early next year).
it all depends on what sugar mama govt gives us and if Grandma Fed can feed the mortgage monster and get those rates back down to a reasonable 3% (ha!) for a fixed 30 year mortgage or not.
SF real estate, like the rest of the country, is on government life support.
the machines are beeping like crazy right now… it’s a tossup what’s going to happen IMO.
sorry… wish there was edit function:
corrections:
clearly your data shows that SFH median prices are down YOY. (added median)
also, your data neither proves nor disproves the idea that it is mix… (added data)
added paragraph:
in addition to SFH vs Condo sales percentages affecting median price, we also have the affect of the change in TIC sales. As we all know, TIC’s sell at a discount to condos. The collapse of the TIC market will push median sales upwards.
For the people who claim r/e prices are up in SF and always complain about socketsite cherrypicking apples to apples, how about providing multiple apples to apples examples showing that housing is up? I am a casual observer and try to be objective about information but all I see is a lot more SFH’s and 2-unit buildings in the real SF languishing, being marked down, taken on and off the market, and very little trading. The properties that do trade are either down a reasonable amount or are very unique properties. Socketsite provides valuable info with the apples to apples even if it is cherrypicked– but as far as I can tell, it’s not.
applestoapples,
Here is a question for you, What is the difference between a brand new house and a hause that someone lived in for 3 years cooked in 5 days a week, stained the carpet, scratched the floors, dinged up the walls, put stickers on the doors, etc.
A: Nothing, it’s a perfect apple.
Volume is the big element here. The lower the volume the less significant the price action.
The current sluggish volume shows a stalemate that has long been debated here.
Wealthy San Franciscans can afford to wait. Sellers with long term equity can afford to lower their expectation to sell. Bargain shoppers are getting decent deals. Buyers with windfall money still compete at very healthy levels on trophy property.
And a huge mass of people in the middle can still not afford the cost of buying here.
Now the key now is what’s going to happen to interest rates. If they hover around the lower 5s this stalemate can go on for a while. If they climb back up to 6% where they naturally belong, all bets are off: buyers will get disqualifies, and the famously derided “reset tsunami” might actually happen if the LIBOR follows the 10-Y note.
why is it so hard to understand … Medians are NOT prices.
The median figure only tells you what people are spending, it tells you nothing about what they are getting for those dollars.
The SF median was 680K, great! What does that 680K get you? 1 Bedroom condo? 3 Bedroom 2 bath SFH? A 3 unit TIC? Where they move in ready units/buildings or major fixer uppers?
If the median a year ago of 650k got you a one bedroom condo but the median of 680k this year gets you a 3 Bedroom condo have prics gone up or down?
The median is at best a noisy indicator of what buyers are spending and tangentially liqiudity, the ease of credit, in the mortgage market.
This is my favorite site, I read it every day. There are a lot of potential real estate buyers who read this site. I don’t want them to sit around hoping that an SFR in a nice neighborhood is going to plummet in price. The low appears to have been in March 2009 when there was blood in the streets. The stock market has nearly doubled since then. Bears who insist prices will drop are doing so because they very much want to own SF real estate, which is counterintuitively a bullish sign.
Prices are up 4.6% and these guys fill the board with anguished, astrologically-based explanations.
Sparky-b, that’s a ridiculous comment and you know it. First of all, most houses that are traded in SF are not brand new, they are “used”. Secondly, most “apples” are cleaned up for sale to somewhere near the condition they were in for the previous sale. The carpets are un-stained, the walls are un-dinged….
A house is not a car. Some people may overpay for a “new house” smell, but it doesn’t define the market in a place like San Francisco.
Well, the reason contractors are in a lot of pain across the nation is because many existing houses have become cheaper than what it would cost to build them. That and the fact that we have overbuilt in areas where people didn’t really want to live in the first place. Smart contractors will have built a nest egg during the fat years…
I don’t want them to sit around hoping that an SFR in a nice neighborhood is going to plummet in price.
The fact that bulls are worried about potential buyers sitting on the sidelines is a (not so)counter intuitively bearish sign.
There is some validity to sparky-b’s “new house smell value” point because there’s a subset of buyers who will only buy “new” or place a high value on new house smell. The question is what proportion of buyers see value in newness.
On the flipside there are buyers who see “new” as a red flag due to infant mortality of newly constructed systems. Usually the builder backs up their product with a warranty though it is not always easy to get a fair remediation when stuff breaks down.
But granted that the number of buyers who understand the bathtub curve are less than those who think “new” means “flawless”.
“Prices are up 4.6% and these guys fill the board with anguished, astrologically-based explanations.”
Median prices don’t tell the whole story when they’re down and don’t tell the whole story when they’re up either. Just look at Case-Shiller and the breakdown by Morgan Stanley. So keep repeating your party line, brother, but it’s meaningless.
sparky-b says: What is the difference between a brand new house and a hause that someone lived in for 3 years cooked in 5 days a week, stained the carpet, scratched the floors, dinged up the walls, put stickers on the doors, etc.
The new home seller hibernates over the holidays and relists next year. The homeowners who bought 3 years ago are desperate; they keep up their short sale listing as they are going into foreclosure anyway.
curmudgeon,
I don’t think that’s true. I think there is both the new unlived in versus someone has used this point, but I also think there is the point that people may clean up a place before they sell but they aren’t bringing it back to how they bought it.
They will mop the hardwoods but they won’t refinish them.
Probably clean but not repainted.
Get the carpets cleaned but not put in new carpet. That pee is still in the pad.
I could go on. In my mind an apples to apples sale is never actually apples to apples.
^^^
Put more colloquially, a house is a depreciating asset. The more people who realize this as sparky has, the better.
Yes, it is not the same thing. This surely explains why a 1M brand-new-in-2007 SOMA condo would sell for 700K today. Not apples. Not even close to be comparable. Nothing to see there. Prices are coming down because houses are getting older.
That is exactly right, lol. That is exactly what I’m saying. I will defer to you to speak for me from now on.
So when places were appreciating in value 15% each year, they were “really” going up by much more than that because the homes were falling apart so quickly! I’m amazed that even the real estate industry did not point this out in marketing materials.
Better up the maintenance cost projections to a far higher annual amount when doing those own vs. rent calculations.
Sparky’s point is valid, but at the same time, I’ve never known the owner of a new home who didn’t fix it up cosmetically, unless it was brand new.
So water faucets get swapped out with shiny new ones, a wall gets repainted here and there, door hardware gets replaced, electrical stuff you can see get replaced: light fixtures, old switches, that sort of thing. You don’t see that in the permits or the listing description, but it frequently occurs. It’s one of the reasons home buying is so critical to the economy: lots of other stuff gets bought in addition to the home.
So you have that on the upside. Unless a brand new place is being sold, the stuff that got improved probably about negates the stuff that got nicked and scuffed. On the other hand, people probably take care of a brand new place a little better.
In the end, I doubt there is much difference in price. Buyers know that floors can be refinished, stickers can be removed.
“In my mind an apples to apples sale is never actually apples to apples.”
While technically correct (and why programmers know that the expression “realValue1 == realValue2” will rarely ever be true), this statement is sort of splitting hairs.
Of course no property is exactly the same on different sale dates. Some degradation occurs. Some improvement might also occur. usually the slight degradation/improvement over the apple interval is just noise. It doesn’t mean that we cannot learn anything from SocketSite Apples.
The only case that really makes a significant difference is the “new home smell” case. When you’re comparing a house that was in a used state before to an unremodeled and still used state now, it is safe to assume that any degradation that may have taken place in the intervening years is insignificant and probably offset by minor owner improvements as well as pre-sales improvements.
It is good that readers challenge the choice of individual SocketSite apples. But it is really hard to claim that the methodology itself is significantly flawed.
“comps” in a market that sees huge investment in upgrades ? Now that is a methodology that is seriously flawed.
As my old Tao master use to posit:
Can a man step into the same MLS-listed home twice?
Indeed, can a man step into the same MLS-listed home once?
Then we would shrug and turn the Ravi Shankar tunes up as we continued with our spackling.
Right, I don’t think the degradation means that Socketsite apples are flawed, they are pretty good. I personally value the second sale as not as nice as the first and without the new car smell. We certainly hear if a sale is not apples because they improved anything of note (pulled any permits). People might do as much work to prepare for a sale as tipster says, but I don’t think most do. The methodology of apples is not flawed in general.
A.T. your Tao master spackled walls with you as you guys listened to Ravi Shankar…I think I know that dude.
apple falls to ground
who’s first – worm or polisher ?
newton just eats lunch
@ex-SF-er:
“clearly your data shows that SFH prices are down YOY.”
Data shows SFR median price is down YoY Nov-Nov 1.84%. However, I consider any +/-7% move on monthly data to be noise, so no, market pricing is flat YoY for SFR. This is also backed up by the fact that avg. sq. ft. pricing is also flat and we are tracking 12 month average median SFR price. I also consider +/-5% and 3% on quarterly and yearly data respectively to be in the noise for SF RE pricing.
Combining all the different types of housing, SFR/Condo/TIC, in to a single number is not that useful in my mind and I never do any analysis based on this single number. Point I made was that if you look at the data broken out, in this case just SFR and Condo, you can see that both median price and sq. ft. price, and also the average price if you are interested, are flat YoY for both sets of SFR and condo data taken separately. Conclusion, market pricing is flat YoY and mix did not change the single housing number.
@badlydrawnbear: what is so hard to get that median, coupled with $/sf, is the best estimate of the market pricing. Apples-apples is really not useful IMO when looking at market pricing. It is useful when calculating an ROI for that particular owner for that particular piece of property, but only if you know everything about said situation. Unfortunately, posters on this site can only speculate on the ROI as we don’t have all the data. I have been waiting to do a post on apples-apples and maybe it is about time, but sparky-b has got the right idea. All consumable assets, like housing, depreciate over time and apples-apples does not take this critical fact in to consideration. In fact, wood structures need to be completely overhauled on average every 100-200 years or so and interior finishes every 20-50 or so. The market includes this mix of new, renovated and apples, and unless there is some outside reason why this mix would change dramatically, it is all captured in the market pricing estimates we use (median $, mean $/SF, mean $). Bottom line, it is expected some percentage of the housing stock will be modernized and some will need work. Trying to extrapolate 1 or even 50 apples to market pricing just does not work. As I have said before, I agree median is not perfect, but rarely are things in this world. However, median, coupled with $/SF, is the best measure we have for market pricing. It would be even better with the associated skew.
@Skirunman, the problem with your analysis is that a median being flat does not mean prices are flat. Median’s in real estate simply do not provide enough information to determine price movements in any reasonable level of accuracy. The same applies to $/sq foot medians.
They are useful, but you cannot arrive at a defensible conclusion about price relying solely on median information in real estate.
Skirunman – your systems depreciation timescales seem reasonable, but consider :
1. the apple interval is so relatively short that the wood structure depreciation is noise. Even the interior finish renewal cycle seems long in comparison
2. many owners make incremental improvements that “fly under the radar” of the apples selection criteria. For example many (most?) people do interior finish upgrades without permits. And pest inspection schedule 1 fixes don’t disqualify an apple.
So while I agree that structures do degrade over time the short apple interval as well as unaccounted for incremental upgrades nearly negate that effect.
@lyqwyd: I am not trying to defend any conclusion. I simply stating facts using accepted statistical methodologies to analyze data sets. By your argument, we can never make any conclusions about data sets using the 300+ years of mathematics behind the theories. I have an a degree in Mathematics and understand statistics and probability theory. As I have said before, median pricing and mean(not median) $/SF are the best tools we have to estimate market pricing. Note: I said “market pricing” and I try to be clear when I say this. Market pricing is “pricing” when we are talking about the market. If you are talking about pricing for a single property that is a very different beast. I am also not defending median or mean or any other statistical tool, we have theorems that do that.
My original post was simply to call BS on tipster’s allegations that the fact the median price for SF RE was flat was that the mix had changed. He has no way to support this point and I think I have shown it very unlikely, if not “proven”.
@Skirunman — What you’ve stumbled upon is that the concept of “the market” can mean different things to different people.
Using a hypothetical example, Imagine I buy a median priced car for $10k in 1980. Fast forward to 2000 when the country is in the midst of a SUV craze and the median car is now $25k.
This is great for car dealers since they are now getting commision off $25k vs $10k. So for them the market is better.
But what about the car I bought in 1980? Has it’s value increased 250% as the median price would indicate? It’s worth less then what I paid. Some of this price decline is wear and tear, but even if I never drove it and stored it in pristine condition for 20 years the market value of my car (and probably nearly all cars sold in 1980) is much lower in 2000 even though the median price has gone up 250%.
For people in the industry (RE agents, brokers, …) then the market really is determined by the median price and #sales. Thats the total pool of money from which any fees and commissions can be taken. There is some RE bashing on SS, but any business in their right mind would keep a close eye on their total available market. This is a perfect use of median price stats.
As the car example illustrates, the problem comes when people attempt to use median price changes (either in the short term or longer historical trends) to estimate the change in value of a house as an asset. i.e. what is my house worth now? What can I estimate it will be worth in 5 years? This is what a home buyer/owner (or REIC investor) considers the market. This is what repeat sales indexs or apples to apples comparisons are best at.
Median and repeat sales are two different indicators which each measure different things.
@Skirunman — “Market pricing is “pricing” when we are talking about the market. If you are talking about pricing for a single property that is a very different beast.”
That isn’t really it. It’s not a question of a few single outliers vs “the market”. It’s more a question of measuring “what’s selling now” vs “how have values changed for things sold in the past”
The car example illustrates that you can easily have a market where *every single* car sold depreciates down rapidly while the median price goes up.
Housing obviously has more correlation between median and repeat sales, but since both measures are easily available why intensionally use the wrong one for a given purpose?
You could measure flashlight brightnes in ounces, heavier flashlights are probably brighter on average, but why do that when you can measure in lumens?
Any appraisers out there? When valuing a home, do appraisers consider city-wide (or some smaller area) median sales prices or average $/sf at all? I would doubt it but maybe I’m wrong. When re refi’ed recently, the appraisal report never mentioned it, only a lot of “comps” in the area.
I think what most people on ss are talking about when they discuss “market price” is price changes on an individual property – medians, CSI, etc. are all just tools to try to gauge this. I use them, but as I’ve said before, I don’t think there is any better gauge than “apples” since those control for all the noise.
@tc_sf: Yes, I understand all of this, but I think your analogy has some flaws when compared to RE, especially in SF. First, we don’t take RE to the wrecking yard like we do with a car when it has reached the end of its useful life and RE has a much longer useful life, the land being almost infinite. Secondly, the housing stock in SF does not change very quickly and has definite limits to growth so when charting trends or comparing YoY you generally are looking at basically the same product and mix. I generally confirm this by analyzing the data at finer granularities (by # bedrooms, sq. ft. quarterly percentiles, etc.) In my experience, what is sold in SF does not vary that much over time, and therefore, the statistical tools do give a good sense of the market pricing.
I can see where this is going so I will stop here. I do agree that an investor ultimately only cares about ROI on the properties they purchase. As a homeowner I care a bit less as I consider the home I live in to be consumption, but hey, maybe that is just me.
@A.T.: your logic is flawed “… I don’t think there is any better gauge than “apples” since those control for all the noise” UNLESS you analyze every single apple sold or can accurately say your sample set is a statistically relevant sample set. As I have pointed out before on SS, what is posted on SS is HIGHLY unlikely to be a statistically relevant sample set, and IMO, is likely cherry picked by the Editor based on his own agenda, whatever it may be. As I said, this is just my opinion. Also, appraisers typically use very localized comps and not SF City wide market pricing data.
True, “apples” obviously provide the best gauge for that particular property and those that are near and/or similar to another property are the next best gauge for that other property.
There’s an open invite for people to post 2009-2010 apples that show the purportedly flat market, or the increase that many argue. Nobody seems to ever accept that invitation.
@A.T.: Sorry as I didn’t want my “your logic is flawed” statement to come out so harsh. I will restate it as “In my opinion, your logic may be flawed…”.
@TMOD: I think the depreciation has a greater price effect on a single property than you think. The government allows 27.5 years straight-line for residential and 35 for commercial and there is a reason behind these numbers. I personally believe to really keep things apples to apples you would need to re-invest 3-5% of the property value each year on average. If not invested, this single apple in theory then drops in value as some logrithmic function of the amount not invested.
IMO you need paint every 5-10 years, flooring and kitchens every 10-20 years, bathrooms and roofs 20-30 years, plumbing and wiring every 40-50 years, etc. Therefore, I believe only sq. ft. additions or major re-configuring of existing sq. ft. should really rule out an apple for comparison purposes. Remodels that include these other type of changes are expected. I don’t know the average age of the SF housing stock, but I would guess it to be at least 50 years old for SFR and 30 for condos. Anyone know?
Combining all the different types of housing, SFR/Condo/TIC, in to a single number is not that useful in my mind and I never do any analysis based on this single number.
I agree with you here.
I also agree that price/sq ft is probably a more accurate data point than is median pricing with some caveats that I may expand upon later.
@A.T.: If you are asking for someone to post 2009 buy and corresponding 2010 sale apples that makes no sense to me. If not, I’m not sure what you are asking as there are plenty of apples posted even on SS that show pricing up since their previous sale. Most are 2005 and earlier buys as I think we all know the market did not go up much in 2006/2007 and is down since 2008. The issue is the time frame of hold by buyer. Single apples are NOT the best measure of market pricing and trends. They are the best measure to calculate the ROI for that buyer/seller over their hold time on a single property. Like I said, unless you have a statistical relevant data set to analyze for apples you will likely will come to false conclusions about the market trends. Yes, the statistical measures can also lead to false conclusions and hence you need to analyse the data in different ways and use different measures to help confirm the outcomes and verify your underlying assumptions about the data sets are not flawed.
I believe A.T. is asking you to post some apples that support the claim that prices are flat over the last year or so, as that is what you appear to be saying.
If you are not saying that prices are flat, then never mind.
Right, any number of down 2009-2010 “apples” have been featured here. As tipster frequently points out, if “the market” were flat (or up) then one would expect to find a fair number of up apples to counter-balance all the down ones. Yet nobody can ever seem to find them.
^^makes sense for a few reasons. if you bought a great house you don’t need to move. if you bought a marginal one hoping to be able to trade up later and are now selling into this market you will probably suffer.
otoh, if you bought a fixer and did a good job you can still sell for a profit even in this market.
Going back to aging apples.
I know for a fact that every place I have bought I have left it in better shape than when I moved in, either minor work or major. Permits don’t say everything neither. In general I’d say the overall housing pool in SF seems to improve simply because the city’s population is becoming wealthier and has higher standards. You should see what people in the 50s used to live with in their kitchen. A big stove slapped here, a fridge in a corner. No real unity. I still see a few like these, but less and less.
Sure apples are not perfect and some people will put a price on being the first to move in that brand new place. I am not sure it’s as much as contractors or flippers would like it to be though. But the mass is old houses. These are getting better year after year in small steps or in giant leaps.
“I personally believe to really keep things apples to apples you would need to re-invest 3-5% of the property value each year on average.”
No comment. I just wanted to meditate on this statement for a while. It contains profound truths.
“otoh, if you bought a fixer and did a good job you can still sell for a profit even in this market.”
Yes.
otoh, if you bought a fixer in 2003 and the only “improvement” you made was to drop your pants and take a big dump in the middle of the living room you could have sold it for a profit in 2005.
ha ha ha! True!
Here’s a 9 month apple at Millenium tower that sold for $737,500 and just got reduced by $19K to $680K, off about 8%. In 9 months! A 12% per year decline and only then if it sells at asking (having just been reduced by a massive 2.8%)!
Now add realtor commissions, closing costs, transfer taxes etc, an I’ll be very surprised if they don’t end up losing around $100K, all in 9 months! On a 1/1 condo!!
Think of what you could have done with $100K in the last 9 months!!! And that doesn’t even include the mortgage payment! WOW!!
http://www.redfin.com/CA/San-Francisco/301-Mission-St-94105/unit-17H/home/28486417
Wow. No less than 10 bangers. S-A-N-C-T-I-M-O-N-Y!
^ We’ve moved beyond the weak ad hominem to the even weaker “ad postinem”.
The downfall of the realtors on this site is even sadder than the losses the people who listened to them are taking up the rear. The realtors used to be so confident: “it’s on a busy street!” “Bifurcated market! – only the bayview will fall” “The upper crust has plenty of money and prices in that stratum will never fall.”
Now, the most they can do is to nitpick the number of exclamation points in a wholly accurate post of the decline in value of real estate, as they and their developer cohorts desperately try to cling to misrepresentative statistics like median and price per square foot, while the life’s savings of even their most recent clients (like those who bought 9 months ago) evaporates as fast as you can say the words “agent commission”.
So sad.
In general I would expect any property that turns in less than 1 year to sell for less than market pricing. This holds true in a flat pricing market like we have now, or even an up or down market. The simple reason being that IMO a 1, or even a 2 or 3, year hold likely indicates a distressed sell of some type. Could be they are relocating or can just no longer can afford said property because of personal issues or whatever. I believe these short term holds average out with the longer hold properties that generate higher prices on average.
Of course this is not always true for every short hold property, but with average ownership hold times of around 10 years (quoted as 8-12 years in multiple sources not including refinances) I would expect short term hold properties to have more pricing pressures than properties held longer. I would also expect them to be in general “inferior” properties to those held longer. I admit this is just my “theory” at this time as I don’t have all the data to back it up right now, but if I have time I will see if I can get it.
Regarding hold times, I could not find specific data for San Francisco, but my guess is they significantly more than 10 years in SF. If you just look at this simplistically and take avg of 5-6k # of transactions per year across the 115k properties you can see about 5% turn over per year. If anyone has this data please post.
If I were a self described brilliant real estate investor, who was profiting even in a soft (at best) market due to my self proclaimed smarts, I sure wouldn’t be wasting my time on a real estate blog trying to convince a bunch of anonymous posters of my brilliance and their ignorance.
No, I wouldn’t. Because I wouldn’t give a flying f@#k about what other nameless, faceless people thought about the real estate market. And I certainly wouldn’t be wasting my time composing twisted rationalizations based on the typical questionable data filled diarrhea that a certain type of MBA loves to excrete in order to “prove” just about anything, but especially one’s own financial or investing brilliance.
Instead, I would be schussing down a skirun, man. Or, kicking back on a beach in Kauai, dude. Or any number of other activities where I could enjoy the fruits of my smart investing.
Or if I were a workaholic, I would be focused on my own smart investments instead of trying to convince anonymous posters on a blog of how wrong their take on the market is. Because, why would I care how wrong they are???
Just sayin’…….
so nnona, what then is your motivation to post if not “trying to convince anonymous posters on a blog of how wrong their take on the market is. Because, why would I care how wrong they are????”
sounds like another shoutdown of any who dare oppose the ss scripture.
you do realize that people who like sfre also like to talk about it, no?
“In general I would expect any property that turns in less than 1 year to sell for less than market pricing. This holds true in a flat pricing market like we have now, or even an up or down market.”
LOL!!!
nnona,
Who are you talking to in that post? self described brilliant real estate investor?
“There’s an open invite for people to post 2009-2010 apples that show the purportedly flat market, or the increase that many argue. Nobody seems to ever accept that invitation.”
That’s me — I’ve made this invitation several times, and no one ever seems to take me up on it. Once again, I encourage everyone who is saying that the editor cherrypicks apples to provide apples of their own.
“The market includes this mix of new, renovated and apples, and unless there is some outside reason why this mix would change dramatically, it is all captured in the market pricing estimates we use (median $, mean $/SF, mean $). Bottom line, it is expected some percentage of the housing stock will be modernized and some will need work. Trying to extrapolate 1 or even 50 apples to market pricing just does not work. As I have said before, I agree median is not perfect, but rarely are things in this world. However, median, coupled with $/SF, is the best measure we have for market pricing.”
As for Skirunman’s comments above, he is correct that some housing stock will be worked on, but the apple methodology account for that. We aren’t talking about 20-year apples where work might be necessary, but at best maybe 6-year apples (2004 and 2010 sales) where “work” might be a coat of paint.
In addition, we all know that mix does change dramatically at times, and I continually point to the Morgan Stanley data in the “Mix Pshaw!” thread as an indication of this.
If you really wanted to use median as a better indicator of price, you would need to use an adjustment factor, similar to Case-Shiller, where a short amount of time since the last sale weights a house higher than a long amount of time since the last sale, since a long amount of time makes it more likely that significant work was done on the house.
And again, Noe Valley Jim has taken the challenge on and delivered before but you always fail to mention that. Regardless, here’s one: 115 Spruce. Same exact price now as in 2005.
@nnona: I assume your comments are directed at me. If not, ignore everything below.
I have not and would never refer to myself as a “brilliant real estate investor” nor would I expect anyone else to do so.
As you seem interested in the reasons I visit SS here they are:
1) Hopefully to learn something I didn’t already know about SF RE happenings.
2) I have developed my own thinking regarding evaluating and valuing real estate and I like to test my ideas with others interested in similar subjects.
3) I enjoy debates and this forum definitely provides some outlet for that interest.
4) I have the time to do so, especially right now as I just finished up a large 18 month project. I also don’t have a “day job” and with RE slow downs this time of year it gives me even more time to type meaningless and long winded posts…
NoeValleyJim also doesn’t constantly complain that SocketSite’s apples are cherrypicked.
However, I thank fluj for his contribution of 115 Spruce, which appears to have been double-agent sold per MLS for $3.8M last Friday and also sold at the same price in December ’05 at the same price. Was the 2005 sale off-MLS?
I will note, however, that it was re-roofed between the two sales, which may or may not disqualify it as an apple. Re-roofing a $200K house in flyover territory is a very significant cost relative to the value of the house, whereas on a $3.8M SF house, it is a rounding error. As a result, I’m mixed as to whether re-roofing disqualifies an apple. It would most certainly disqualify a house in Antioch.
Not sure why a short hold would sell “under market”.
Why does the number of buyers or the amount they are willing to spend depend on the length of time a previous seller has held?
If anything shorter holds would have less potential for having an outdated design or “deferred maintenance” issues.
Having some accounting for maintenance/update costs is a great idea (Someone on a previous thread mentioned 3-5%/year). But you should be carefull not to double count. i.e. If you put 3-5% in your buy/rent spreadsheet you can’t also “discount” sale prices with long holds.
Also, the “market” is the collection of everything being sold so you can’t have everything selling “under market”. i.e. if you think short holds go under market due to “distress” and long holds go under market due do a need for updating and REO’s go under market… then what is selling above market to balance this all out.
I certainly call in to question the statistical validity of the apples presented on SS. I guess you can call it complaining if you want. How many apples are highlighted each month? (10-20 max?) With 400-500 real estate transactions alone each month this sample size is just not statistically relevant even without the Editor’s selection bias. Come on folks, this is a media site and he is trying to generate interest, and therefore, it is in the best interest of the Editor to select “interesting” apples. I have asked before for the Editor to present his methodology for selecting the apples presented to show that they are truly random without selection bias. Without this data one has to call in to question the validity of the apples presented. It is that simple even barring the small sample size.
@tc_sf: “Not sure why a short hold would sell “under market”.” My theory, and I admit it is simply a theory at this time without full data to back it up, is that short holds are:
1) more likely to be distressed sales than longer holds, especially 1-3 year holds. The median sold price for bank own properties over the last year was about $300k less than for non-bank owned.
2) likely to be inferior properties on average in their prospective price range. I believe better properties that command top dollar in their sub-market turn over less than less desirable properties.
I agree that the statistical measures of market pricing take this all in to account. However, the point I was trying to make was that I would not expect many (or any) 1 year hold apples (bought in 2009 and sold in 2010) to exist that show flat pricing even though I believe market pricing to be flat. People that buy and sell in 1 year IMO typically expect to take a loss, except maybe flippers during 2004-2007 bubble.
The NAR tweaks DOMs. It tweaks sale-vs-ask stats. It tweaks the final sale price when there’s cash back at closing. It’s an almost official body as it is a quasi-monopoly and its data is proprietary with no real oversight.
Having a couple of bloggers expressing their views of the market is just small compared to that bias. You’re giving more importance to SS than it probably has. The SF market has very deep imbalances and that’s the main reason for the current stalemate.
@Skirunman —
Re #1, I’m definitely willing to give accept that auction sales should sell at a discount. You are taking a chance and have less disclosure then a normal sale. But I’m less willing to write off bank owned houses sold through normal channels. Perhaps there is some correlation with bank owned houses and houses with poor upkeep, but one of the advantages of looking at individual house resales is that you can see the condition for yourself.
For #2, even if this is the case. The property was just as inferior 1 year ago.
@tc_sf: “For #2, even if this is the case. The property was just as inferior 1 year ago.”. Agreed, but it is my postulate that such such properties are more likely to take a LARGER % discount with a one year hold, i.e. their price is likely to vary more on the downside as they are less desirable. In any case, I don’t have the data or time to prove this right now so it is somewhat moot.
That roof is three years old now, parsing man.
skirunman,
If you wish “to question the validity of the apples presented”, all you have to do is list the omitted apples for a given time period. Either put up, or shut up. The editor appears (wisely) to not argue with those who will not produce data.
PS. MLS price sold errors that always err high are not noise, but bias.
@Delancey: I find the “put up, or shut up” comment to be quite hilarious and your PS comment only makes it even more clear that you lack the required understanding of statistics to analyze large and widely varying data, and hence, don’t really want to try to better understand what is happening with SF RE market pricing. Even if I were to produce 50 or 100 apples selected by me it would not provide a reliable answer to the question if market pricing is going up or going down. This seems to be the point many people just don’t get. In addition, Case-Schiller already provides a statistical model that attempts to account for apples. Unfortunately, it is currently not broken out for just the city of SF.
This will likely be my last post on statistics as my head is hurting from all the banging against the SS wall and it just goes to prove the old adage that most people just want to continue to believe what they want to believe. As I have said before, apples provide great value in understanding a specific property and possibly some value to another property that is a very close comp. They can also provide color to the statistical measures. However, any single or even 50 apples gives you no reliable insight in to SF RE market pricing trends without the guarantee that the number of apples is statistically relevant (it is not as presented on SS) and have been selected randomly (highly unlikely as presented on SS). I’m done.
@Delancey: I find the “put up, or shut up” comment to be quite hilarious and your PS comment only makes it even more clear that you lack the required understanding of statistics to analyze large and widely varying data, and hence, don’t really want to try to better understand what is happening with SF RE market pricing. Even if I were to produce 50 or 100 apples selected by me it would not provide a reliable answer to the question if market pricing is going up or going down. This seems to be the point many people just don’t get. In addition, Case-Schiller already provides a statistical model that attempts to account for apples. Unfortunately, it is currently not broken out for just the city of SF.
This will likely be my last post on statistics as my head is hurting from all the banging against the SS wall and it just goes to prove the old adage that most people just want to continue to believe what they want to believe. As I have said before, apples provide great value in understanding a specific property and possibly some value to another property that is a very close comp. They can also provide color to the statistical measures. However, any single or even 50 apples gives you no reliable insight in to SF RE market pricing trends without the guarantee that the number of apples is statistically relevant (it is not as presented on SS) and have been selected randomly (highly unlikely as presented on SS). I’m done.
i appreciate the insight you bring to the site skirunman and i hope you do not leave b/c of all the shouting down that goes on here. tho it is like banging your head against the wall to change the ss viewpoint at least you are reaching the open minded around here.
^ i too hope that skirunman continues to post his muddled thoughts on statistics. It’s a useful reminder that “a little knowledge may be worse than none at all”. Lurking behind all the words about “statistically significant sample” and “random” is a profound lack of understanding of what selection bias really is in the real estate context. No need to elaborate – those who understand, already understand, and those who don’t, won’t.
– el bombero
@el bombero: I said I would not post on statistics again, but I feel compelled to respond. You presented two logical fallacies:
1) argumentum ad logicam (straw man)
I have not discussed sample selection bias in “the real estate context”. This is way too broad a subject. I have specifically called in to question the apples presented on SS as likely not being selected at random from all possible apples. Until the person presenting the data, in this case the SS editor, provides the selection methodology one can not assume the samples were in fact selected randomly, i.e. the burden of proof is on the data presenter. One of the two main types of sample selection bias is when a researcher selects the data in such a way as to prove a particular hypothesis. This can be done even without the conscious knowledge of the researcher. I believe this to be the case here. However, I would be happy to be shown otherwise.
2) argumentum ad hominem
Your personal attack lacks merit. I don’t know your depth of knowledge in the field of statistics and probability theory, but mine is pretty decent (degree in mathematics and first job out of school developing algorithms and software for anti-submarine warfare industry, which required extensive use of statistical theories and advanced mathematical topics such as Kalman filters). I am also happy to be shown where my “profound lack of understanding” actually exists in regards to any specific statement I have made.
Finally, you did mention “statistically significant sample (sic)”, but did not address this issue in the context of apples. To achieve reasonable confidence that your “randomly selected” apples accurately represented the data set you would need to present about 100 apples each month (assuming 400-500 real estate transactions). From my quick review it looks like about 10-20 unique apples are presented on SS each month. This is just too small of a sample size to provide meaningful insight.
Skirunman, I’m still glad to hear your thoughts.
…you would need to present about 100 apples each month (assuming 400-500 real estate transactions)
I’m curious why you think that you would need such a large sample of apples within that 400-500 transaction base. Putting aside the issue of selection bias, and also the issue as to whether all transactions in that base are created equal, I would expect a sample size of a few percent (10-20 apples out of 400-500 transactions) to be relevant.
I’m with El-D here — I am both interested in Skirunman’s thoughts (and have asked him about his thoughts on apples many times and am still interested in hearing it). Also, I would want to know why you would need 100 apples per month if there are 400-500 real estate transactions. Are you making an implicit assumption that if there were some smaller sample that it would be biased for some reason? I am open to the argument that a short-term apple may be statistically different from a non-apple, but it would take a bit to back that up.
The only one who needs 100 apples per month is a developer who doesn’t like the fact that they are showing a declining market.
And if there were 100, they’d be called outliers…
In all seriousness, 20 apples/month vs. 100 apples/month just gives you a larger margin of error. That’s why I’m curious why skirunman thinks you must have 100 apples/month to say anything.
MLS has selection bias too, as to data provided, as people have already mentioned. Garbage data in = garbage data out.
SFR, get a clue. He thinks you need that many because there aren’t that many.
You don’t get away from the selection bias merely by picking a “many” apples because each apple represents a “sold” property. If you’re trying to estimate the movement in statistics of a population you need to ensure that your sample set is randomly selected before you even get to whether the sample size is large enough to generate a reasonable confidence interval for the underlying movement. The decision to absorb the loss in a losing transaction is anything but random because human beings are loss-averse, thus you will expect that any sample size of sold real estate apples will exhibit a systematic bias to understate true losses. This is pretty elementary. It’s actually the potentially “distressed” sales like shorter than usual holds or foreclosures where this systematic bias in the sample is less likely to be present. Like i said, those who understand, do, and those who don’t, won’t.
– el bombero
Yes, in addition to being quite an accomplished architect 😉 Skirunman knows that he is BS’ing us with this sophistry. It goes back to the garbage in – garbage out point I raised earlier. Nobody who understands even basic statistics, and Skirunman does, could credibly argue that the incomplete, inconsistent, and biased sample of sold SF MLS properties yields anything too meaningful with respect to market prices, much less medians and $/sf. And then to argue that these flawed medians and $/sf are more significant than actualapples-to-apples sales is the step too far that convinces me Skirunman is just playing around with us.
Skirunman, if you spent any time actually trying to pull together apples, you would realize that there is no way that 1/4 of the sales in San Francisco qualify. For one thing, most of them were sold more than 5 years ago, which makes them pretty dubious for our purposes, partially for the reasons you state. There is a big difference between a house with a three year old kitchen and one with a 15 year old one.
It is also kind of hard to get the data, perhaps someone with MLS access can do a better job, but the ones I present are discovered by looking at the “for sale” listing on redfin. When I see one that has a sale listed in the last 5 years, I tag it for further notice. Later on, I will pull up a report on homes sold in the last six months and see if any of the ones I had tagged are listed. Then if one *is* listed, I go to the on-line DBI site and look to see if any permits were completed in the time frame.
This is all very labor intensive and then even after I go through all this, tipster usually claims that they are invalid for some made up reason. He often claims there is unpermitted work or that the seller and buyer engaged in some elaborate fraud to misreport the price on the MLS or some other bizarre fantasy of his.
But if you think you can pull together even 10 good apples a month, much less 100, knock yourself out. Please tell us your methodology for picking apples as well 🙂
BTW, I did in fact find some apples that had gone up back when Satchel was claiming that there weren’t any, but I haven’t found any with a short hold in quite a while. I mostly focus on the Noe, Glen Park and Mission Dolores neighborhoods, so there might be some out there elsewhere. But I haven’t seen one in a couple of years.
Hmm, just as I say this I go back and look at my ripening apples and see 110A Chattanooga, which recently sold for $2,025,000 after selling for $1,987,500 in Nov 2007. It has no permits in the online database and is relatively new in any case.
A few more apples (a meagre lot):
317 29th St #203 – $717k on 7/10, $769k on 8/06
3 Chenery – $895k on 8/10, $805k on 12/04
4175 Cesar Chavez – $1.1 M on 10/10 $1.1 M on 8/06
I have my doubts about 4175 Cesar Chavez, since looking at the pictures you see a new kitchen but there are no permits either before or after the sale for it online.
110A Chattanooga is actually kind of a surprising result, can anyone explain this one?
“110A Chattanooga is actually kind of a surprising result, can anyone explain this one?”
I can. It is an awesome place! Zak/DeVito design and build really exceptional homes. You could draw broader conclusions from this very good result for the 2007 buyer if there were more places like this, but there aren’t.
Noe has held up better than just about anywhere else in the city. Still substantial declines, but relative to the shellacking everywhere else, it is doing better. As I have oft-stated, there are only degrees of bad in the current SF market, and Noe is probably the least bad.
Noe is probably the least bad.
You are not just saying that because you own property in Noe, right?
I’m not above that kind of thinking! But I actually own near Buena Vista Park, not Noe.
@NVJ: “if you spent any time actually trying to pull together apples, you would realize that there is no way that 1/4 of the sales in San Francisco qualify.” I am in agreement with you and that was the point I was trying to make. Apples as presented on SS, while interesting, are not all that useful when trying to understand overall market pricing trends. It does not mean we shouldn’t look at apples, but that we should look at them for what they are, data about a specific transaction(s).
@El-D and sfrenegade: I have simplified my response below, but I used Cochrane’s (1977) formulas for calculating appropriate sample size for continuous data. I assumed we are measuring %change of sale price from last sell and that data would be in the range of -35% to +15% and +/- 3 standard deviations of mean, an alpha of 5%, acceptable margin of error of 3%, and a population size of 450 (average number of closed deals in SF/month). This gives you a required sample size of 94 to be statistically significant. You can make other assumptions about the data set, margin of error, etc., but you will still likely end up with somewhere between 50 to 150 required samples.
@el bombero: I agree that sold properties are not selected randomly from all properties. However, it is moot as what we are discussing and trying to analyze is trends in the values for “sold” properties, i.e. market (transaction) pricing is the data set under analysis. Let’s also get back to the original point, which is that I believe that the apples as presented on SS don’t accurately represent market pricing and trends. Do you agree or disagree?
@A.T.: regarding your garbage in/garbage out argument, if I accepted it, which I don’t, it would also seem to invalidate the use of apples as well as data about RE transactions is “incomplete, inconsistent, and biased” as even data in public records can be manipulated. Of course the MLS data is far from perfect and especially $/sf data should be further analyzed, but if the errors are random, or even show a specific bias that does not vary much over time, then with the data set sizes we are using the error rates are likely not to be meaningful.
Finally, just to be clear, I am not arguing that prices are not down from the peak, I think they are down 15%-20% on average. I also believe pricing has been basically flat over the last 1.5 years. I’m going to stay away from public pricing predictions for 2011. Happy New Year!
“I also believe pricing has been basically flat over the last 1.5 years.”
Ahh, the wishful thinking of a developer…
All anyone ELSE needs to do is to look at the offers on 310 townsend the agent referred to on that thread. It’s been on the market for 9 months. Let’s see how “flat” that market was for her:
First she got a real offer for $500K. Then that fell out. Later, the best offer she could get was $450K. The bank wouldn’t agree to it, and now she says it’s fairly priced at $399K.
Flat market. LOL!! Only in a developer’s dreams. Does a real offer for $500K, then a later offer for $450K and finally a fair price of $399K look flat to you?
Me neither. Fortunately, skirunman didn’t support that assertion with any actual data, so I guess his “data” (there isn’t any) is “not inconsistent” with the real world data indicating a declining market.
Neither did you. You talked about one condo.
Yes, but that one condo I mentioned is in addition to the hundreds and hundreds of properties that aren’t selling and all of the other apples that declined.
@tipster: “Fortunately, skirunman didn’t support that assertion with any actual data…”. The data was presented in earlier posts in this thread as well as the chart posted by the SS Editor at the top of this thread that shows median pricing off 18.6% from peak.
Even better is the 12 month moving average as it smooths out noise from normal monthly swings in mix, which looks to be down about 17%-18% from peak (exact numbers are not posted, but definitely within 15%-20% range). Of course and as I have stated before, looking at a single number is not that useful as property classes and locations do vary greatly. I think best nabes (Russian/Telegraph Hill condos/Co-ops, prime Pac/Presidio Heights and parts of Noe SFH) are down more like 10%-15% off peaks and worse nabes including new SOMA tower condos are more like 30%-35% on average.
^You should use the 48 month moving average as it will include even more high priced sales!
Your continuous use of statistical methods to mislead is hilarious. 12 month moving average indeed!
Come on tipster, do I really have to explain why 12 month moving average is of interest? Real estate, just like many other businesses, does have annual seasonality. Looking at 12 month moving average, and monthly YoY data as I posted above, does help try to factor out some of the seasonality in the pricing trends. If you don’t like this data then look at month over month data. However, there is much more noise in this data. Here is a graph of last 12 months median pricing through November:
http://www.paragon-re.com/Docs/General/SixtyFortyImages/12-10_MedianPrice_All_Homes.jpg
I personally like 3 month trailing average median price as a nice compromise. This number is actually up through November, but still in my 5% noise factor.
[Editor’s Note: Unfortunately a change in median does not equal a change in values (up or down): Medians Are Up, But Don’t Confuse That With Increasing “Prices”.]
@Editor: JHFC not again. Your analysis in that thread is flawed as previously pointed out. Yes, medians are not perfect, but they are the best we have when trying to understand RE market pricing trends and certainly have low noise ratio when looking over 1 year’s worth of data. Add in other statistical measures such as mean $/SF and looking at pricing in quartiles and by locations (over 1 year to get enough data) and you get a good sense of the trends. I even agree that “randomly” selected spot checking of apples to provide additional color to the statistics is useful.
However, you continue to point to this flawed analysis and also continue to want to use the logical fallacy Ad Ignorantiam (burden of proof) in your argument that “Unfortunately a change in median does not equal a change in values”. You need to prove that change in mix was in fact responsible for a change in median pricing, not the other way around, which you have not done.
“medians are not perfect, but they are the best we have when trying to understand RE market pricing trends”
Nope, not when you’re only looking at homes that sold and ignoring everything else, and not when you have a pretty small sample each month and a large variation within those small samples. “Comps” (i.e. “apples”) are better, which is why appraisers use them to ascertain market values.
If you look at the medians for the entire year, rather than monthly medians, it is better but still pretty weak. Medians measure one thing, and it is not “market pricing.”
If I had access to the raw data, both of sales, and of the housing market in general, I could tell you whether or not medians as given by NAR/CAR and others is helpful. However, no one that I know of is in possession of this data. No one is providing in a single publicly accessible database an accurate enough record of all the housing units in SF, all the units for sale in SF, all the units sold in SF, etc. in order to do any sort of real analysis. This would require cobbling together MLS data, public records data, and even newspaper listings and certain other sources. Even MLS is not good enough, because a large number of units are off-MLS, at least according to our editor’s prior analyses, although perhaps MLS could be a large enough sample.
That’s why the best we have is Case-Shiller and an individual apples analysis, even with the relatively small number of apples. What’s next best is “comps,” although comps are always subject to spin, especially when everyone in SF is trying to assert that their nano-neighborhood is the best.
And in my opinion, the best Case-Shiller slicing and dicing I’ve seen was by Morgan Stanley in the “Mix Pshaw” thread, as I’ve mentioned way too many times. It separated out organic sales from distressed sales.
@A.T.: I think I already addressed the sold versus not sold issue in response to el bombero above. Also, you have the same issue with apples so what is the point? Discussing some hypothetical market of unsold homes is really not useful IMO.
Also, the number of transactions is not small (400-500/month) and is much more significant than the 10-20 apples posted per month on SS. As I have posted above I think I have shown that apples as posted on SS have little statistical relevance to the overall number of market transactions. Also, what do you mean by “pretty weak”? Please give some statistical information instead of a this less than useful opinion. Medians do what they do, no more no less.
Finally, appraising is more of an art than a science, and they actually use median pricing and rely heavily on mean $/SF using relevant comps in the area. I just had a building re-appraised in Russian Hill last month with First Republic and went through the analysis with the appraiser in great detail. They will also consider construction replacement costs in some cases.
[Editor’s Note: A Quick Reminder Not To Bank On An “Appraisal” (Or Instant Equity)]
Ed: Shouldn’t your ‘note’ be tagged to ATs post since he’s the one who brought up appraisals?
@Skirunman — Hmm I’m not sure this is a correct usage of Cochran’s formula.
First off, I don’t think the formula is meant to provide a go/no-go line for whether or not a certain number of samples has any “statistical signifiance” or not. The more samples you have the better estimate you can make.
Secondly, I’m not sure about the calculation you showed. I have cochran’s formula as:
n_0 = ( t^2 * s ^2 ) / d^2
Using a t of 1.96 coresponding to an alpha of 5%
s = standard deviation of 3 as you indicate
d = margin of error of 3%
I get n_0 = 3.8 or 4 samples (since this is below 5% of the population).
You may think this seems low, but I think this is due to the standard deviation of 3 that you chose. With a standard deviation of 3 and a margin of error of 3, even with only one sample you would be within the margin of error ~70% of the time.
This brings up the interesting question of what is the actual standard deviation of home price changes over a fixed time period.(i.e. apples to apples for a given time period). When googling for the answer to this I found that you asked the very same question on SS a bit ago.
I couldn’t find a great answer to this.
All I found was chart 8 (Home Price Volatility — Standard Deviation of 12-Month Price Changes NSA) in:
http://www.standardandpoors.com/products-services/articles/en/us/?assetID=1245229752318
This has standard deviation peaking at 7. But I’m not sure this is measuring what we want. Specifically, this may be the standard deviation of the 12 month time series of house prices rather then the standard deviation of all 12-month old “apples” for a particular time.
I fairly certain that the numbers we want are generated in the process of computing the Case-Shiller index, but I could not find them published anywhere. Perhaps someone more plugged in then I could contact the analyst who wrote the linked paper.
I agree completely though that the apples picked by SS should in no way be assumed to be random. However as people seem to agree that out of 400-500 transactions at most 1/4 are apples, yielding 100-125 potential apples. It’s hard to dismiss 10-20 points out of 100-125 as outliers.
@tc_sf: this article has a good practical description on how to use Cochran’s theorem:
http://www.osra.org/itlpj/bartlettkotrlikhiggins.pdf
Skip down to the tables on page 6 if you want the cheat sheet version on the required number of samples. I used 6 for standard deviation (+/- 3) so sorry if that was not clear from my original post.
I don’t “dismiss” the 10-20 apples or even call them outliers. I find them interesting for what they are, which is information about a particular transaction(s). The point I am trying to make is that looking at 10-20 non-randomly selected apples each month is just not statistically significant when trying to analyze the entire market.
Also, I don’t think anyone has actually tried to show that about 1/4 of the monthly transactions are actually apples. I believe this was just someone’s guess. If I was to guess I would think it would be more like 3/4, but that is just my WAG. It would be interesting to calculate this number.
Finally, it is not entirely clear to me exactly what meets the criteria of an apple or not. For instance, does new paint and a new roof rule it out? These are just maintenance items so would not IMO. In fact, I would claim that only additions to square footage or significant interior changes to room layouts or significant structural upgrades should actually rule a property out as an apple. All other changes including new flooring, appliances, tile, or paint and redoing a bathroom or a kitchen, etc. are expected maintenance items and/or keep up with current style issues as a property ages and are expected by the market IMO. I am not trying to use apples to calculate a particular buyer/seller’s ROI, but instead trying to understand trends in market (transaction) pricing so this is how I would tend to look at apples. Obviously you would want to eliminate (or account for) all dollars spent on the property during ownership if you want to attempt to calculate ROI. In either case, there are some inherent issues with this approach that I touched on in an earlier post regarding accounting for depreciation and life cycles of materials.
Interesting paper, but it seems to confirm my recollection that you’d use Cochran’s rule more for designing a survey rather then tryng to determine the significance of a given number of samples from a population.
Regarding table 6 there is potentially some confusion with %error when the varriable being measured is also a percentage. Taking the 3% error that the table is built for, if i want a 3% error when measuring something with a mean length of 10in this gives me a range of 9.7-10.3. If I’m measuring mean appreciation with a mean of 10% this gives a range of 9.7%-10.3%. If I want a range of 7%-13% this is actually 30% error in this context. Notice on page 46 where they work an example and compute the “d” ( the acceptable error for the mean ) which is what is actually used in the formula by multiplying .03 by the 7 points of their survey questions.
Regarding the standard deviation, generally accepted terminology would call a std deviation of “+/- 3” be 3 rather then 6. A standard deviation of 6 would mean that 68% of results lie withing +/-6 of the mean. A standard deviation of 3 means that 68% of results lie within +/-3 of the mean
See the first graph in: http://en.wikipedia.org/wiki/Standard_deviation
If people are concerned if house prices are down 20% vs 18.6% then I think your numbers are ok. But generally there seems to be a much wider gap between the “market is flat” people and the “bloodbath” people.
To illustrate just how big a gap this is consider the following. If you assume that the “market is flat” people are arguing that the mean price change is 0% and you use a standard deviation of 3 then a data point (apple) at -20% is 6.7 std devations away from the mean!!
If you look again at the Wikipedia article ( http://en.wikipedia.org/wiki/Standard_deviation#Rules_for_normally_distributed_data ) you’ll see that you’d expect data points past -6 std deviations to occur with a frequency of 0.0000001973% /2 ( You divide by 2 since the table includes events less then -6 std dev and more then +6 std deviation). Basically 1 in a billion.
If I can find 10-20 samples which should be 1 in a billion outliers out of a population of 500 this is strong evidence that the model (mean=0%, std dev=3%) is wrong.
For those not finding Wikipedia an authoritative source, here’s wolfram alpha:
http://www.wolframalpha.com/input/?i=-20%2F3+sigma
Look at the “left tailed p value” showing the probability that a sample is less then the given value.
For all the BS on this thread, a Millennium Tower unit just sold as a just under one year apple for a hair under $100,000 off. 14% decline, plus commissions in just under a year.
Evidence of distress is nowhere to be found: it sold 5 months after it was listed: plenty of time to find a buyer.
http://www.redfin.com/CA/San-Francisco/301-Mission-St-94105/unit-17H/home/28486417
Geez, $15,000+ per month this whole thing ended up costing them. For a 1/1. Rent, fools!
“unit just sold as a just under one year apple”
“Evidence of distress is nowhere to be found:”
So you do not consider the fact that someone is trying to sell a home six months after buying it to be evidence fo distress? Personally I consider such a short hold in a declining market to be evidence of someone that wanted to bail out asap.
“it sold 5 months after it was listed:”
So? I bought a place that had been on the market for over 9 months, “plenty of time to find a buyer” as you say, but it was a very distressed sale.