Those of you who have been following me for a few years would know/realise that I have not been very active for the last year or so; in fact, I have not updated www.burbwatch.com.au for some 9-12 months (not sure how long exactly!).
Now, at long last, I have updated the charts site.
Yes, I’m Back…
OK, OK, where have I been?
I’ve been busy, sorry; life has take a change about a year ago, and, yes, life has been quite different – for the better, largely, but I have certainly been much more time poor.
I am sorry, dear reader, that I have not given you a BurbWatch fix for so long!
Blog Hits, Yeah!
Hey, check out those 26,000+ hits! Nice.
Thankyou all for supporting this humble site so well since mid-2010.
Reflections: MacroBusiness – QLD New Building Starts
Yes, I do like the content at MacroBusiness; I’ll admit to being a regular reader🙂
Here is a key chart (QLD is the green-line trend):
To those of you that are familiar with The ‘BurbWatch Project, you will likely also be familiar with my “Relative Sales” concept.
It compares the number of sales recorded with the amount of stock on the market.
The truth is, I don’t much like how raw sales figures are compared year-to-year, and sometimes significantly back in time, as these absolute numbers have different significance as time rolls on.
For example, selling 6000 properties this year, should not necessarily be seen as great compared to selling 5800 properties a few years ago. For, what about population growth? What about changes in the distributions and demographics of those who might buy properties? How many properties were on the market then, compared to now? And, how can we use sales information to better understand the state of the property market, if we are interested in doing so? Read more…
I’ll admit this is a little more of a “dry” post…
…but after looking at some info on the Australian Bureau of Statistics (ABS) website, I noticed the following in a section on “Average Weekly Earnings, Australia, Aug 2011“, relating to “Compositional Effects”:
Movements in average weekly earnings can be affected by changes in both the level of earnings per employee and in the composition of the labour force. Refer to paragraphs 25 and 26 of the Explanatory Notes.
It is good to see that the premier statistical body in our fine land also sees it as important to remind us that point-value metrics – such as average this, and median that – can change, not only because of actual shifts in per-unit values, but also because the shape of the distribution of the data concerned shifts to one side or the other.
This was in relation to Incomes, and Income distributions, but it also applies necessarily to Vendor and Sale Prices, just as readily, and this should be taken into account when considering the values and trends of point-value metrics – ie. the average, medians, modes, etc, that are stipulated and trended without prior illustration (ie. charting, often via a Frequency Histogram or similar) of the distributions from which those point-value metrics were obtained.
However, sadly, I have seen very, very few instances where data is released first showing the distribution – let alone time-trends of distributions! – before reporting the point-value metric (average, median, etc); the only example I can remember is the occasional ABS release where a distribution was shown for incomes for a particular period – fantastic!
But it is a sad state of affairs that we do not have it as a standard practice – and we make so much of so little when we consider these sorts of economic data when only point-value metrics are provided, without the distributions that actually give them their context and meaning.
I look forward to the day when this becomes standard practice.
I began collecting data from realestate.com.au for Gladstone (QLD), for an acquaintance in August 2011.
It has been 5 months since then, and I thought that it was time I (remembered!) collected some more data…so, please find the following, without much commentary at all, of some selected Gladstone data – from August 2011 to January 2011.
…observe the large shift to the right (more expensive) in the above chart…
…this is just a line version of the first chart, but you might be able to appreciate the differences between the August 2011 profile/distribution, and the January 2012 distribution; you may be able to see that the Vendor Prices appear to be centralising (ie. gathering or “huddling” more around a central value – in the case of January 2012, the BurbWatch calculated median is ~$507K, and the Trimmed Weighted Average is ~$559K.
The changes from August 2011 to January 2012 in terms of vendor price – which reflect Vendor expectations, or hopes, not necessarily the final sale prices (which might be higher or lower) – are conveniently reflected in the following BurbWatch Vendor Price Indices Chart (where I show all three types of my indices, for the sake of comparison):
…please note that the Un-trimmed price index has exactly the same value (so far) as the Trimmed price index, so they occupy the same line in the chart.
But what a boom in Vendor Prices (ie. expectations, demands, and hopes), from the point of view of the calculated median, with a ~5% rise in 5 months! Good, or Ouch?
Honestly, I would expect the median to be a better representation of such a data set that is “skewed”, so I will remain more attached to believing this higher (~5%) figure than the lower trimmed and un-trimmed figures, at least for now (the reasons such as are mentioned in my last post).
But it is still important to consider such point values in the context of the shape and location of the distribution – not only has the distribution very obviously surged to the right (more expensive), it has also centralised, with lower fractions at the cheaper end (quite noticeable), and lower fractions at the more expensive end (less noticeable).
It will be interesting to see what happens at Gladstone over the next few quarters, and I will keep you all (including my acquaintance, “Lefty”) informed of the results.
In accordance with the last post, I have put this one together for all the Australian states combined, for the Calculated Median Index (as opposed to the Weighted Average Index, from the last post):
Generally, for statistical distributions that are skewed (significantly “non-normal”) like these data sets are, median representations tend to be better than means/averages, as they are less subject to changes in the composition of the data itself; hence, medians tend to be less volatile and, thus, more “reliable” as a measure of the central tendency of the data set in question…
* Side Note: … even if, for data sets like these, I am a believer that: visual representations are an absolute must, at a minimum; and, for these sorts of skewed data sets, simply stating a lone median, or mean, etc, is a poor way of communicating the nature of the data – standard deviations (etc) should also be stated to assist in providing the “user” with a better idea of how well the median (etc) actually represents the data set … *
And here are some examples of how skewed these data sets are – for NSW, VIC, QLD (as examples); notice how different the shapes of the distributions are, even for these 3 major states…
It is interesting that NSW and QLD have some similarities (with NSW still being flatter and more broadly distributed), but VIC is tending more and more towards strong centralisation in the listed $300K-400K bracket – is this due to the surge of units coming onto the market in Melbourne, which we have heard so much about?
I’m not sure, but the $300K-$350K bracket has done nothing but increase its fraction of listing since I started recording the data in April 2011. Is it also the First Home Owner bracket for outer Melbourne? Are they finding things particularly stressful, and are now getting out? It’s hard to say.
Another point to consider is whether the composition of the Vendor Listings are having a leading effect on the sales trends – perhaps leading by 1-3 months? It’s a working hypothesis of mine, anyway, and I’ll look at ways of testing it over the coming months.
However, in light of this idea, I am wondering if the large surge (increase) in the NSW Calculated Median Vendor (Listing) Price Index (first chart) will cause an increase in subsequent sales index figures? That is, will the large swing to the upside in the weight of the distribution have a flow-on effect to seeing a rise (or mitigation?) of sales-indices in the 2-3 or so months following the August “turn around”? Subsequently, will the significant and rapid decline in the Vendor Price Index(s) (reflecting listings/supply price composition) following October then pre-empt similar sales index declines? We will see, I guess…
Is what I call “compositional bias” an issue? I am not talking about the bias of a collector, user or operator; I am talking about the actual makeup of the listings effecting subsequent sales results. Surely it is a significant consideration, since we are just not talking about changing buying habits of the property-buying population for a fixed supply profile – the supply/listings profile seems to change as dynamically as buying habits and preferences, as is suggested even by just the three listings profile time trends for NSW, VIC and QLD, above.
To expand a little more on the concept, this compositional bias notion suggests that changes in price-composition/structure of the listings available will influence the subsequent sales results. Honestly, it seems to make sense to me, and should work both ways: to either cause compositionally-biased decreases in sales prices, or increases; for, the more (greater supply) of a certain-priced property there is, the more likely it is that a property in that bracket will sell, and thus “drag” indexed-sales results in its own direction.
Additionally, it might temper/minimise larger trends either way – for example, compositional swings to the downside in a bull market could temper indexed-sales rises; and, similarly, compositional swings to the upside could temper indexed-sales falls in a bear market; and large compositional swings either way could see actual short-term counter-trend movements in sales-indices – for example, indexed rises in a longer-term trend bear-market, and what might possibly happen for the NSW market recently (or soon to be), as discussed above.
I am not saying that this is always a significant case; but I am saying that compositional bias should be considered when contemplating sales-indices values and trends, as it is part of the bigger picture…and a bigger picture is a better picture…and that it could be considered part of a predictive model.
Unfortunately, the major data providers do not make this sort of information readily available – if at all, or only for a significant price to a privileged few? – so, until they (and/or the govt data providers?) get their bums into gear and provide the populace with more information in this largely data-opaque industry, you the reader, are left with little ‘ol me, and my spreadsheets and charts to provide you with such information and insight!
The sooner it changes, the better, right?😉
Pick up your act data providers! There is a lot of hungry and interested people who want more diverse and transparent information for this industry. Surely there is a competitive motivator here somewhere?
Until next time,
It’s been a little while since I’ve posted, so I thought I would throw something out there into cyber-space, to let you all know I am still on-the-go.
So, here’s a short something to wet your whistle…
I don’t have data for actual sold-price, which is what all the media-reported indices are based on (and rightly so!), and is the stuff that you have to pay through the nose to get your hands on, one way or another. However, I do have ready access to Vendor/Listing price data, so that is what I use.
It can probably be used as some sort of leading or lagging indicator of the various sold price indices; it will always be inferior, and, I haven’t figured out yet whether it is leading or lagging…that will probably require getting my hands on some RPDate, ABS data, etc, and plotting them together, to see if there is any relationship, and how strong it might (or might not) be.
Nonetheless, below is a chart I’ve put together of the Trimmed Price Index (for short) for all of the states…or, for long, The Trimmed Weighted Average Vendor Sales Price Index.
“Trimmed” means I have manually removed the first and last price categories from the raw realestate.com.au data.
Why, you may ask? Because they are full of rubbish…everything that wasn’t entered properly by agents seems to find its way into either the lowest bracket ($0-50K) or the highest bracket ($10,000K+).
There is an untrimmed version, and there is also another type of price index I now use, called The Calculated Median Price Index – but I have not developed an all-states chart for either of those yet.
However, just so you get an idea of how the different/similar indices I have developed for Vendor Price tracking are, please find below the time-trends for NSW, ACT and VIC.
I hope that was somewhat interesting.