Median values don't reflect the real world

'Tis interesting that you agree that medians do not necessarily reflect the real world. Will you refrain from using it as a stat in future?
I agree that the median has faults, especially when a particular demographic (like FHB) disappears from the market. However, until recently it was the best the ABS & others could come up with. What stat should we be using instead ? RPData have come up with the Hedonic Index which is an improvement over the median. It calculates the 'average' based on sales of similar types of houses. They release it towards the end of every month - see here.

As an example of the flaws of a median - 200 sales of 3 bedders and 100 sales of 4x2 in a suburb would give a median skewed towards the lower end of the range. If FHBs disappear, there will be only 100 3x1 sales, but 200 4x2 sales - consequently the median will jump towards the higher end of the range. So the composition of the data has an effect of the median, even if every single one of those houses sold for exactly the same at it did in the previous period.

A Hedonic Index takes the 'characteristics' of the dwelling into account. As an overly simplistic example, it takes the average of 3x1 sales & the average of the 4x2 sales for each period, and calculates an suburb average based on those figures. FHBs disappearing from the market will have little effect that type of Index.

Most house price index data providers use the OSR data which is accurate, but several months after the event. RPData collect their data from various sources earlier in the house buying process and that gives more timely information.
 
Hedonic Index

Sounds good Keith, I like the model that takes out improvements.

The fourth repeat-sales model of Goetzmann and Spiegel (1995) is motivated by the fact that in many cases the features of properties are not constant through time. Be it a fresh layer of paint, or the installation of air-conditioning, most houses undergo some level of revamping, often just prior to sale. The Goetzmann and Spiegel model controls for elements of price appreciation that are not related to the time between sales, and thus is ideal for the customer looking for a “pure” estimate of price growth in residential real estate.
All a lot of time and effort to compile and track this stuff. Matusik does a good job tracking resales for Brisbane and the Gold Coast, but there is still nothing comprehensive out there. When you get down to the individual property level it's interesting how different the performance of different areas of the same market can be.

I think there is a lot of untapped potential and likely paid demand for people to come in and do some quality and timely work on property data in this country.
 
I agree that the median has faults, especially when a particular demographic (like FHB) disappears from the market. However, until recently it was the best the ABS & others could come up with. What stat should we be using instead ? RPData have come up with the Hedonic Index which is an improvement over the median. It calculates the 'average' based on sales of similar types of houses. They release it towards the end of every month - see here.

It should be better but still introduces an element of judgement (eg what is 'similar' - should a 2x1 house on a very large block be considered equivant to other 2br houses or similar to 3br houses)?

I think the idea of price distribution of sales is important. Eg it would be good to have the upper quartile, lower quartile as well as the median. So you can see how each end of the market is doing. In some suburbs there would be too few sales each month for this to be reliable, but a 3 month or yearly figure should be reliable.
 
Problem is, though, the bigger the area, the harder it would be to get specific, useful information. But the smaller the area, the fewer buyers of the information there would be. These data providers won't do this information gathering and make a loss.

Which is why localised information source such as local agents, buyers agents, and your own legwork is so valuable with resi property.
 
Keith

Do you know if the hedonic index will be used in addition to the median index so news media will pick the one that suits them or if they'll stop tracking the median alltogether?

And if anyone else other than RPdata is using it?
 
It should be better but still introduces an element of judgement (eg what is 'similar' - should a 2x1 house on a very large block be considered equivant to other 2br houses or similar to 3br houses)?
They do take a lot of things into account (including block size), not just the number of bedrooms/bathrooms. I think there will always be an element of judgement when deciding the improvements of each individual property, however, this should a) be averaged out over a large sample, b) be reasonably consistent throughout time.

I think the idea of price distribution of sales is important. Eg it would be good to have the upper quartile, lower quartile as well as the median. So you can see how each end of the market is doing.
Agree. They do produce a Stratified Median Price Indices series, however, it costs $$$.
 
The main problem I have with the 'median' price is not the figure itself - I take it for what it is, a fairly anecdotal number that as Keith mentions, doesn't necessarily reflect the real world. The problem I have is when people use that as a basis to conclude that property is unaffordable.

Why use the median figure to conclude that property is unaffordable? Perhaps a better figure to use would be the cheapest figures of housing in the area you're analysing. Yes, the median of $400k may be classified as unaffordable, but does that matter if there are still thousand upon thousands of properties available at $250k? Yes people may desire the meidan property - but then that's exactly what it is, desire and aspiration, not necessity for the purpose of discussing affordability.
 
Do you know if the hedonic index will be used in addition to the median index so news media will pick the one that suits them or if they'll stop tracking the median alltogether?
The ABS and APM also produce house price series (including medians), so I guess the media will pick whichever is the most newsworthy :rolleyes:. Medians are easy enough for the masses to understand, so I can't see them disappearing any time soon. RPData also produce a Stratified Median - IMO to hard for the masses to understand.

And if anyone else other than RPdata is using it?
I believe the RBA is using (or at least evaluating) it. RPData have a list of organisations that could make use of it. The ABS data has appeared to confirm RPData with a month or so lag... that's probably why the RBA are interested.
 
Keith, your opening post is more than reasonable.

If someone other than RP Data was doing the hedonic indexing I would be happier but my bias against them may be overdone. I'm human too. :D

"Like sales" cannot take into account new dwellings so that could not work. Nor would it work where a suburb is gradually undergoing "change of use".

I have mentioned my own home as an example. I bought in '68. It would have been above median at the time but I did not know the term. It was a near new three b/rm high set weather board house but that was the norm in Qld at the time. Most new homes today would be 50% bigger with double L/U garage and ensuite. Ergo the home I bought which was above average, today is a bit of a shack so I can't do some research, find the median now and divide it by the median then, come up with a multiple and apply that to my $10,000 buy price and say that is what my property would have given an investor in cap gains. (That's a bit long-winded. Sorry :))

It may actually be worth more'n that because this "shack" is on a 1012 sm development block but the structure adds nothing to my sell price if viewed that way.

So it is not the actual "median" figure, but how it is used to calculate "profit" I have a problem with.

BTW another thought crossed my mind: In '68 there were very few apartments in town and they were modest structures. Today there are expensive high-rises coming onto the market every year. How do you hedonicly adjust for that? A rising median is a fair indication that people are willing to pay more for a house, I'll admit, but I have never questioned that.
 
With the properties we target, there is no "median" data available.

I've never been able to rely on something like - "well the median for this type of property in this area is X, therefore with this particular one for sale right now it should be X-Y if it's bad, or X+Y if it's good."

How much do people actually rely on the median price to guide them ??
 
If someone other than RP Data was doing the hedonic indexing I would be happier but my bias against them may be overdone. I'm human too. :D
I'm always sceptical too, Chris Joye is perceived as a bull.... however, the 'offical' figures do tend to agree with RPData's figures, although a month later. The RBA is interested too.

"Like sales" cannot take into account new dwellings so that could not work.
I believe that's where a hedonic index works well. The new dwellings would be classified in a separate category from the existing dwellings. The 'old' 3x1s & 4x2s would have an average each, the new dwellings would have their own average. And then the suburb average would be weighted (somehow) to account for the different composition.

From wiki

Hedonic index is any price index, which uses information from hedonic regression. Hedonic regression describes how product price could be explained by the product's characteristics. Hedonic price indexes proved to be very useful when applied for information and communication products (e.g. personal computers) to calculate price indexes, because they can successfully mitigate such problems as new goods and rapid quality change.

Again from wiki
Hedonic models and real estate valuation

In real estate economics, it is used to adjust for the problems associated with researching a good that is as heterogeneous as buildings. Because buildings are so different, it is difficult to estimate the demand for buildings generically. Instead, it is assumed that a house can be decomposed into characteristics such as number of bedrooms, size of lot, or distance to the city center. A hedonic regression equation treats these attributes (or bundles of attributes) separately, and estimates prices (in the case of an additive model) or elasticity (in the case of a log model) for each of them. This information can be used to construct a price index that can be used to compare the price of housing in different cities, or to do time series analysis. As with CPI calculations, hedonic pricing can be used to correct for quality changes in constructing a housing price index. It can also be used to assess the value of a property, in the absence of specific market transaction data. It can also be used to analyze the demand for various housing characteristics, and housing demand in general. It has also been used to test assumptions in spatial economics.



Nor would it work where a suburb is gradually undergoing "change of use".
I'd agree that suburb indices wouldn't help especially in that situation. However, I'm sure RPData would be more than happy to sell data about the change in prices of houses of a specific quality (eg 3x1 on a 1000m block) within that suburb. That would be more useful than a suburb index.

I have mentioned my own home as an example. I bought in '68. It would have been above median at the time but I did not know the term. It was a near new three b/rm high set weather board house but that was the norm in Qld at the time. Most new homes today would be 50% bigger with double L/U garage and ensuite. Ergo the home I bought which was above average, today is a bit of a shack so I can't do some research, find the median now and divide it by the median then, come up with a multiple and apply that to my $10,000 buy price and say that is what my property would have given an investor in cap gains. (That's a bit long-winded. Sorry :))

It may actually be worth more'n that because this "shack" is on a 1012 sm development block but the structure adds nothing to my sell price if viewed that way.

So it is not the actual "median" figure, but how it is used to calculate "profit" I have a problem with.
Land size & zoning & proximity to amenities/CBD are taken into account. I think for an individual property the median (then or now) isn't especially relevant..... putting it to auction would establish it's value. If there are a number of houses with similar characteristics (Older highset W/b on 1000sqm with dev potential) in the suburb, then I'd expect there would be index data you could buy from RPData. Or they would be able to tell you what each attribute was worth eg 1000sqm=$150K, close to CBD=$50K, favorable zoning=$50K, older 3br w/b=$10, etc ?

BTW another thought crossed my mind: In '68 there were very few apartments in town and they were modest structures. Today there are expensive high-rises coming onto the market every year. How do you hedonicly adjust for that?
I'd guess that new apartments are significantly different from the old ones, so a new category (sub-index?) could be created. The major attributes may be - views(?)=$80K, newness=$100K, 3br=$100K, proximity to beach=$40K.
 
With the properties we target, there is no "median" data available.
But you would break down the property into it's consituents & value them... much like a hedonic index. Eg 10,000sqm = $1M, Lotsa frontage=$200K, Level=$250K, blue chip tenant=$400K, long lease=$400K, Lardlord friendly lease=$100K, Close to Freeway=$750K. To come up with a value.
 
I'd agree with KeithJ that a hedonic index is a good thing.

The three main UK indices (Halifax, Nationwide and Land Registry) all use a similar methodology. The Halifax and Nationwide figures tend to correlate quite well.

_47605130_house_prices_mar466x310.gif


If you've got a big enough sample size then the median will probably tend towards being accurate as the sales made will approximate the composition of the market. So I think that it's more useful than just an anecdote.
 
Wow, finally it's sinking in.
But maybe you and others around here think that "it's the best that ABS and others can come up", cause I been posting it here for years that it ain't.
Admittedly in a bull market there is no need to think or analyse, prices just go up regardless of whatever criteria people think they may use.
There's even a book about it called "Fooled by ramdomness".
Of course nobody likes that idea and likes to attribute a boom market to their own doing, when the fact is that they were just "there".

An auction does not really establish worth or value, but a price that someone is willing to pay at that time (for a number of reasons).
If someone is willing to pay the same price of a house to buy a tulip, does that make a tulip worth the same as a house? Me thinks not.
Just as a company with no profits being worth more than IBM or Time Warner.
Many people, analysts, commentators and experts did at one time.
So price and worth are two different things.

And imo RPData's "Like sales" is like crappola with choc sprinkles on top.

With the properties we target, there is no "median" data available.
I've never been able to rely on something like - "well the median for this type of property in this area is X, therefore with this particular one for sale right now it should be X-Y if it's bad, or X+Y if it's good."
How much do people actually rely on the median price to guide them ??
In that world you need to think & analyse a little harder than reading the latest API for the next hot pick suburb.
Each investment has it's own "constituents" that interact with each other in different ways than the previous.
And a reason why the difference between value & price variations are much greater imo.
 
An auction does not really establish worth or value, but a price that someone is willing to pay at that time (for a number of reasons).
If someone is willing to pay the same price of a house to buy a tulip, does that make a tulip worth the same as a house? Me thinks not.
Just as a company with no profits being worth more than IBM or Time Warner.
Many people, analysts, commentators and experts did at one time.
So price and worth are two different things.

If there is one thing from my years of investing that i could emphasise to 'newbies' it would be this.

Work out your own interpretation of 'worth', dont let the market do it for you.
If you can do this and be reasonably correct (you dont need to be exact, just approximately correct), you will do well over time.

In regards to all this talk about the medium, its a useful general proxy, one amongst many, thats all.
 
If there is one thing from my years of investing that i could emphasise to 'newbies' it would be this.

Work out your own interpretation of 'worth', dont let the market do it for you.
If you can do this and be reasonably correct (you dont need to be exact, just approximately correct), you will do well over time.

We all start as noobs, but the reason we are noobs is that because our sense of "value" does'nt have >20 yrs of experiencing values changing.
How can you have an interpretation of values?
They are also bombarded with media and gurus and the whole RE industry telling them not to worry RE goes up 7% every year forever.
Of course the avg person on this forum is a lot savvier than that, but the general market don't seem that way.

The craziest example is LTCM, where they used less than 10 years of back data to work out their system. And these we're economics nobel prize winners FFS! Billions on the line, and what happened before 10 years is irrelevant. :confused:
Then of course the .con...
That's how the "experts" still think and the mass of investors seem follow them.
So telling a noob the opposite of what they see & hear on the the tele, radio, print etc is'nt an easy task.
C'est la vie.
 
if you think it's "value" then buy.

first impressions are not always what they seem, but then, if you've been looking for a while, you'll see the tips and traps and notice when something looks "cheap".
 
Back
Top