One of the side effects of the COVID-19 pandemic has been an increased awareness of statistics. Most Australians (and certainly every Victorian) today knows what a 14-day moving average is all about, for example. Statistics are likely to dominate economic discussion as well through much of 2021.
While there has been an understandable concentration on statistics as they directly reflect the effects of the pandemic, it is just as important to better understand the general environment with which the pandemic will interact. So there are advantages in taking a broader view of current stats, to better contextualise what is likely to happen in the future.
The beauty of charts
Statistical charts offer a way to quickly understand statistics. (We use the word "chart" at HNN, because the more common "graph" encompasses something a bit more complex in maths. Charts are a specific type of graph.) When used to aid understanding, rather than to shore up an established point of view, the best charts give a boost to our intuition. They are a bit like our favourite photos, really: there's "that" photo which shows how you are with friends, but another special photo that shows how you relate to your partner (or pet, family, car, house, etc.). They are all "true" (mostly), but each highlights a different aspect of who you are. The same applies to stats charts.
In terms of building statistics, one of the areas that really needs some help from charts is analysing building approvals. Approvals are very important for gauging the "mood" of the housing development industry, and seeing what that industry expects to happen in the housing market in another six to 12 months.
Chart 1 illustrates the monthly numbers provided by the Australian Bureau of Statistics (ABS) for building approvals up to the end of September 2020. We're using the consolidated numbers for houses and multi-unit dwellings, which include both private and public (government funded) construction.
The technical statistical term for a chart such as this is: "a bunch of squiggly lines going everywhere". In other words, it's really hard to tell what is going on in this chart.
Chart 2 shows a technique sometimes used to make these charts somewhat more readable, with the monthly figures consolidated into quarterly (or three-month blocks) of numbers.
This looks a little better, but does it really help? There is a real problem in taking lots of good data points and blurring them into fewer data points. You lose definition, and can suffer information loss.
Chart 3, which goes back to using the monthly numbers, shows a much better solution. Here all we've done is to take exactly the same numbers, and "stack" them.
So, for example, in January 2020 the number of houses approved was 8797, and the number of multi-dwellings was 7349. The bottom shading represents the 8797, and the top shading represents a further 7349, so the very top line for that month represents 16,146, the total dwelling approvals for January 2020.
The reason this chart works so well is that it clearly shows how the number of multi-unit dwellings relates to the number of houses, and represents clearly the total building approvals at the same time. What we see most clearly in this chart is that, while the housing approvals show some variance, the multi-unit approvals fluctuate much more, and are largely responsible for the peaks and valleys in the approvals data.
What a chart such as Chart 3 really gives us is a good place to start. It can help us to come up with more charts which provide confirmation and some extra details. One way of doing this is to shift to a much broader timescale. We can do that by using what statisticians calling a "trailing 12-month" period. For example, these ABS stats can end with the trailing 12 months from October 2019 to September 2020. (Sometimes "year-to-date" is used for this timescale, but it is ambiguous, as year-to-date September - for example - can also mean January to September.)
Chart 4 shows what those numbers look like. Trailing 12-month timescales all but eliminate seasonal variations, which helps to smooth the data.
It's very clear from this chart that our suppositions from Chart 3 are correct: house numbers are relatively stable, while multi-unit dwellings are more volatile.
To extend this a little further, we can chart the percentage changes between the trailing 12-month periods, which is what Chart 5 shows. There are some additional details this chart makes evident.
Perhaps the most interesting is that for the six years between 2013 and 2018 house proposals averaged 5.14% growth, with the lowest dip coming in 2017 at -1.74%. So, in terms of houses, this was not so much a period of decline, as some commentators have suggested, but mild, relatively stable growth.
The story for multi-unit dwellings is quite different. Over that same period average growth was 11.22%, while growth declined as much as 18%, and rose as high as 34%.
For the final two years, both dwelling types declined steeply in 2019, but then recovered, with houses managing to return to positive growth in 2020.
Volatility
So far, what we've been looking at are trends. The other element of interest in building approvals is what are sometimes called the "spikes": where the data goes sharply up or down.
To measure the overall volatility of data (how it goes up and down) we typically use something known as "standard deviation" (STDEV). There is a bit of fancy maths to this, but the STDEV of a set of data basically provides an average for how much the data deviates from the data average. As a rule of thumb, something a bit under 70% of results will lie within one standard deviation from the average.
What we can do, then, is to work out the average and the STDEV, then plot only those months where the number of approvals exceeds the average plus one standard deviation.
To make it work better, we won't show any data points that are below one standard deviation, and we can express the ones we do show in terms of a percentage of one standard deviation, to keep everything in scale.
That is what is shown in Chart 6.
While this is an interesting chart, it's not all that effective. What we can see in the chart is that many of the data points are "clumped" together. What we could do - and in this case it will genuinely help - is to show the data not as monthly, but in quarters.
While we're at it, we are currently only showing data points beyond one standard deviation from the average - why not add data points for those that fall one standard deviation below the average as well? That's what is shown in Chart 7.
What becomes clear in this chart is that we can define four distinct periods. The first is a downside trend in spikes in the aftermath of the global financial crisis, from (in quarters) December 2008 to December 2009. The second follows the fading of the mining boom (a boom in terms of trade), from March 2011 to March 2013, also a downside spike.
The next two are upside spikes. There is a significant series of spikes from September 2014 through to September 2016, and from September 2017 to September 2018.
There is a very in-depth story to be told about how these spikes relate to the general market, and the actions of the Reserve Bank of Australia (RBA) in lowering interest rates, as well as the bank's influence on lending guidelines. HNN will be going deeper on this in the next issue of HI News.
COVID-19 pandemic
The big question, of course, is what kind of effects from the COVID-19 pandemic can be detected in the data, especially from March through to September 2020.
To tackle this question, the best tool to use is a month-on-month comparison of growth in building approval numbers. We can graph that over the three years to show how current numbers differ from past numbers during the same seasonal period. That's what we have done in Chart 8.
For people outside the hardware retail and home improvement industry, it might come as a surprise to see that, while the growth rate for detached houses did go negative in May, it recovered well in June, July and August, and increased by 28% in September.
The story for multi-unit dwellings is not as good. There was a brief upwards peak in July 2020, but this returned sharply to negative growth for both August and September. While the numbers are not as good as for houses, however, they are close to those for the same period in 2019.
One major factor in this is that the RBA dropped the cash interest rate by a total of 0.5% in March 2020. The federal government also launched its HomeBuilder program, which has made over $600 million in bonus payments available to new home builders and renovators.
Analysis
While this has been a fairly deep look at some aspects of the ABS Building Approvals stats, we really haven't gone nearly as deep as we could. The data the ABS provides is great, but the difficulty is to be able to extract the information that is relevant to the challenges your business faces.
The process of understanding statistics is seldom achieved by glancing at a single chart. For most of us, you really need at least two or three charts, examining data from different angles, and in the right time and statistical context, to get some sense about what data might really indicate.
At the moment, looking at the questions that COVID-19 pandemic has created, all that can be said statistically is that the support measures provided by the government to date seem to be working. The real difficulty, as HNN has suggested in the past, is going to be the state of the economy around May 2021.
A problem deeper than the pandemic, but which the pandemic recovery will need to play off of is that the housing market does not always function as it should. In a more efficient market, when demand decreases due to price rises, those prices should decrease to create more demand. The Australian housing market has become conditioned to the expectation that when it does encounter difficulties, and prices do begin to fall, the market will receive help, usually in the form of reduced interest rates.
Those interest rate reductions then feed into yet higher house prices, and the cycle starts over again.
With interest rates currently lower than they have ever been, and unlikely to decline any further, there will not be any help coming from that area, at least not for five or six years. With monetary stimulation all but gone, that just leaves fiscal stimulation, through programs such as HomeBuilder. Those are unlikely to go past $2 billion in total.
All this means that at some point, the housing market will need to reform itself, and start to act as a more efficient market. There is almost certainly going to be a downwards slide at the point where that occurs. The question is well that transition point can be managed.
