Real Estate price forecasts in the media

It's good that finally the media is including forecasts when discussing the future of Australian's real estate prices.

However we can improve on the forecasts being used. The CoreLogic Moody's forecasts that are quoted are regularly highly inaccurate; Vifortech Solutions has proved it can do better.

Comparing the percentage error of annual forecasts in the discussion below, we see:

  1. In only one case is the Vifortech Solutions forecast wrong by more than 25%; and
  2. In four out of five cases the CoreLogic Moody's forecast is wrong by more than 30%.

Contact me if you would like to discuss further.

Now... Let's dive into some details.

What is the claim?

A recent article (archive) by Shane Wright (no relation!) in the Sydney Morning Herald points to the Moody's Analytics property price forecasts with mentions of regions and expected movements:

In Sydney, Moody's is predicting an overall drop of 3.3 per cent in values. Areas such as Ryde (6.6 per cent) and the eastern suburbs (6.7 per cent) are forecast to endure substantially larger drops in value.

Whilst Moody's (and Moody's Analytics) are well known brands, there is no discussion of the (in-)accuracy of their previous work. Let's look and compare against the Vifortech Solutions forecast.

(For context, I should point out that I founded the CoreLogic International Analytics team, so I know something of their business.)

Forecasts: CoreLogic Moody's vs. Vifortech Solutions

Vifortech Solutions regularly outperforms CoreLogic Moody's

In the table below I have the performance of both the Vifortech Solutions and CoreLogic Moody's (CLM) forecasts for the 12 months ending Q1 2017 for the five largest Australian (capital) cities. That is, how accurate were the 12 month forecast?

(Yes, I know this is a while ago, but these are the only comparisons I'm able to publish at the moment.)

The table below lists the percentage error of the forecast 12 month change against the actual (observed) 12 month change. For those playing at home, the percentage error is the difference (i.e. forecast minus actual) divided by the absolute value of the actual.


Percentage Error

Vifortech Solutions CoreLogic Moody's
Adelaide   -24% -91%
Brisbane   40% -7%
Melbourne   -25% -32%
Perth   -12% 100%
Sydney   -18% -79%

What do we see?

In four of the five markets the Vifortech Solutions outperforms CLM. We didn't do so well in Brisbane.

In the closest market, Melbourne, the CLM estimate differs from the observed result by 32% compared to the Vifortech difference of 25%. That is, the CLM estimate is 28% (i.e. \(\frac{32}{25} – 1\)) more wrong.

Magnitude of errors

More telling is the magnitude of error compared to what actually happened:

  1. In only one case is the Vifortech Solutions forecast wrong by more than 25%; and
  2. In four out of five cases the CoreLogic Moody's forecast is wrong by more than 30%.

Perth, the most volatile of markets reinforces this point; the CLM forecast is incorrect by 100% – estimating zero change whilst the market fell by 4%.

The performance of the CoreLogic Moody's index

Below are the actual, published, CoreLogic Moody's results.


CoreLogic Moody's

Actual Forecast Difference
Adelaide   3.5 0.3 -3.2   (-91%)
Brisbane   4.5 4.2 -0.3   (-7%)
Melbourne   10.6 7.2 -3.4   (-32%)
Perth   -4.0 0.0 4.0   (100%)
Sydney   10.7 2.2 -8.5   (-79%)

Here is a plot of the Reserve Bank of Australia House Price Index values. The CoreLogic Moody's results are against their bespoke index – the results differ from the public data, but not materially enough to change the conclusions herein. The squares represent the equivalent CoreLogic Moody's forecast for this data, the triangles the Vifortech Solutions forecast.

/images/RBA-HPI-to-2017.thumbnail.png

Final thoughts

It's good to finally see some articles talking about forecasts based on analytics. It's a pity the easiest to consume analytic articles aren't using the best insights in the market.

The other question that we can't answer with the published results is the uncertainty of the results. How reliable are the forecasts? With four out of five forecasts wrong by more than 30%, I'm guessing not very.

If you'd like to know more, please reach out.

Fixing Emacs' unknown and untrusted authority TLS error

XKCD: Real Programmers

I recently upgraded my MacOS version of Emacs to 25.3 and was looking to install a couple of new packages.

Installing packages

I have the following in my init.el (or equivalent):

(require 'package)
(add-to-list 'package-archives '("melpa" . "https://melpa.org/packages/") t)

Unfortunately when I ran M-x list-packages I got the following error:

Certificate information
Issued by:          Let's Encrypt Authority X3
Issued to:          CN=melpa.org
Hostname:           melpa.org
Public key:         RSA, signature: RSA-SHA256
Protocol:           TLS1.2, key: ECDHE-RSA, cipher: AES-128-GCM, mac: AEAD
Security level:     Medium
Valid:              From 2017-09-18 to 2017-12-17


The TLS connection to melpa.org:443 is insecure for the following
reasons:

certificate was signed with an insecure algorithm
the certificate was signed by an unknown and therefore untrusted authority
certificate could not be verified

Googling didn't help. There were all sorts of misunderstandings - including blaming the package prelude (which I don't have installed).

Checking melpa.org

Huh? Let's Encrypt isn't trusted? Is this a MITM attack? Nope.

You can check that melpa.org is legit by running the following on the command line

curl https://melpa.org

or

gnutls-cli --tofu melpa.org

or just visit https://melpa.org in your web browser.

The solution

It looks like the MacOS version of Emacs doesn't find the correct certificate authorities.

Step 1

Using homebrew (if you're on MacOS) install LibreSSL:

brew install libressl

Step 2

Set up Emacs to use the certificate authorities you just installed. Simple as adding the following to your init.el:

(require 'gnutls)
(add-to-list 'gnutls-trustfiles "/usr/local/etc/openssl/cert.pem")

This command tells Emacs to use the guntls package (which has, in the past had problems) and prepend the TLS CA certificates from LibreSSL to the list of trusted CAs.

Now when I install packages it all works.

What NOT TO DO

Please, please, please IGNORE any advice you see along the lines of

replace https with http

or

Start emacs with 'emacs --insecure'

Security is important. Don't just turn it off because you can't get something to work.

The flu as an analogy for risk management

I've recently suffered a bout of the flu, with 3 days of fever over 38℃ (that's over 100℉). It got me thinking.

When you get sick with the flu, people say things like:

Guess you didn't get the flu shot.

or

Looks like the flu shot isn't working.

...and it's intelligent, knowledgeable, people who say this.

Replace references to the flu with risk management and they're the same comments that get thrown around when many companies assess the value of risk management or when it fails.

Looks like risk management isn't working.

or

So... How much has Process X really saved us from losing?

It struck me that the same smart people who don't understand how the flu shot works are not that different from the people who don't understand how risk management works. Understand is definitely an emotive word here - what I'm talking about is the difference between knowledge and understanding. A smart person can know something, without understanding it – without being able to generalise their knowledge.

How does the flu shot/risk management work?

It's about managing and accepting risk. The cost – time, money, effort, etc. – of removing all risk (if it is even possible) is usually too high. It is, in some sense, the 80-20 rule. So we accept the risk of catching some (bloody annoying!) strain of the flu, with the confidence that we will avoid the majority of other strains that are likely to be around - all at a nominal cost.

It's the same with risk management – even the name gives a hint: management not removal or avoidance.

To avoid the flu or a loss, the flu shot, or risk management, needs to be right every single event. If it isn't, just one single time, then you catch the flu or make a loss.

How do you measure success?

It is ... difficult ... to quantify how much money you haven't lost because of good risk management practices. It is a lot easier to measure how much money you made by avoiding managing your risks. However, that's the (in-)famous strategy of

picking up pennies in front of a steamroller.

To measure success, firstly, don't trust your instincts. Virtually no-one can do statistics by instinct. Take the famous Monty Hall problem:

Suppose you're on a game show, and you're given the choice of three doors: Behind one door is a car; behind the others, goats. You pick a door, say No. 1, and the host, who knows what's behind the doors, opens another door, say No. 3, which has a goat. He then says to you, "Do you want to pick door No. 2?" Is it to your advantage to switch your choice?

You probably know this problem and learned the answer, but as a test, try to come up with a solution and a justification (that you really believe). Research shows most people still do the wrong thing - even after being told the answer - knowledge vs. understanding again...

So, how do you measure the success of your risk management approach? Ask someone who understands risk management.

A flippant answer, that is isn't too far from the truth is — If you have someone in your organisation who is a contrarian, who disagrees with approaches and seems to always say "No, because ...", that's who you start with. If, in your organisation, you have a risk management expert who doesn't let their team speak, who doesn't have strong opinions or really stand up and fight for them, well, you're in the majority and at risk.

The best people in this area, don't think of risk management as an add-on, or as a separate entity. It is, and should be, part of the risk taking decision process. One of the best experts in risk I know is an ex-trader.

In the end

Coming full circle — remember to understand the risks you are taking, when something goes wrong (and you catch the flu) ask questions about the efficacy of your risk management, but remember, sometimes c'est la vie.

Real estate as an investment

I recently presented at the 2017 Q Group colloquium on

You can apply rigorous analytics to the real estate asset class (it doesn't have to be guesswork and slick salesmanship)

I want to help people see that the real estate asset class can be treated as an investment, rather than speculation. In my view, buying one house, or even a handful, isn't investing, it is speculating. However, saying that, I do know you can treat real estate as an investment.

I also pointed out most of the "everyone knows ..." conversations that keep happening about real estate are misleading and in many cases just wrong.


What does one need for an investment?

  1. Valuation of the underlying asset (& Income)
  2. Forecast / Prediction of future behaviour of the asset
  3. Default likelihood - Is the asset going to keep existing?
  4. Diversification / Correlation
  5. Derivatives, Hedging, Risk - How can I manage this asset?
  6. A market - There's no point if I can't acquire and dispose of the asset easily.

Unsurprisingly, Vifortech Solutions has tools for some of these and relationships with companies to help provide the valuations if you need them.

We view real estate as simply another asset class - like equities, FX, commodities, etc. With the tools that Vifortech Solutions has developed, the process of including real estate in current, multi-asset, investment strategies is straightforward.


Future posts will highlight some a couple of slides from my Q Group presentation in these areas. We've also got some fact checking on "everyone knows ..." type comments that are thrown around in this space.

Disasters and Real Estate

/images/flood-lights.jpg

Disasters happen to people

The recent Cyclone Debbie was a disaster. People were hurt and killed. The psychological impact is, and will be, dreadful.

I'm not going to talk about the human impact - only the economics. I could not do justice to the people involved. I'd suggest a visit to the Red Cross or similar organisation to help those impacted.

Disasters are "News"

The Cyclone Debbie disaster, naturally, made news worldwide (BBC, CNN, Guardian, NY Times) and of course real estate prices started getting discussed.

What always surprises me, when it shouldn't, is that there is so little content and so much hot air (AKA signal-to-noise ratio) in real estate discussions. When I was working at Hometrack I presented on real estate recovering from disasters at the ARCRC conference in 2011.

Historically what happens?

Rather than postulating wildly, I picked three recent disasters in Australia and reviewed what happening with prices and sales volumes.

In the charts that follow, I've normalised both the sales price index and the sales volume index so that they are 1 at the date of the disaster. That way we can easily see how much the market has moved.

1989 – Newcastle Earthquake

The 1989 Newcastle Earthquake was "one of Australia's most serious natural disasters".

Quoting Wikipedia:

The damage bill has been estimated at A$4 billion (including an insured loss of about $1 billion).

The effects were felt over an area of around 200,000 square kilometres (77,000 sq mi) in the state of New South Wales, with isolated reports of movement in areas up to 800 kilometres (500 mi) from Newcastle. Damage to buildings and facilities was reported over an area of 9,000 km2 (3,500 sq mi).

What happened to the market?

/images/1989-Newcastle-Earthquake.thumbnail.png

Sales Volumes

The market dropped December 1989 – January 1990, but I wouldn't conclude much from that. Sales volumes generally fall over the New Year period. Some of it would definitely be a result of the earthquake, but then the market starts improving.

However, look at how much the sales volumes fell from August 1988. There was something happening in the market prior to the earthquake.

Price indices

Units - The market didn't fall. It actually went UP after the earthquake, but longer term the market flattened.

Houses - The long term trend is up, up, up!

2009 – Black Saturday bushfires

The 2009 Black Saturday bushfires was the worst bushfire disaster that Australia has experienced.

More than 3,500 buildings were destroyed — 2,029 houses were among them. The burned area was 450,000 hectares (1,100,000 acres).

What happened to the market?

Here I just looked at a part of state — Kinglake, Murrindindi and Marysville. It was the largest of the many fires burning on Black Saturday, destroying more than 330,000 ha (820,000 acres). It was also the most destructive, with over 1,800 houses destroyed and 159 lives lost in the region. (Wikipedia)

/images/2009-Black-Saturday.thumbnail.png

Sales Volumes

This may, at first glance, be a surprising result - they're up. It definitely appears that the long(er)-term average sales volume is higher after the disaster than before.

However, sales volumes without a view as to the prices (i.e. the index) can be misleading.

Price indices

Units - It doesn't appear that there was much impact in the prices of units.

Houses - House prices, however, had a dramatic fall - so after the fire disaster there were more houses (increased sales volume) selling for less (decreased price index). If you were in the sensationalist media you would print a headline like:

Homeowners losing money as they flee disaster region!

...but again this would be wrong. Yes, there is a fall in housing prices and yes, sales have increased, but the trend in house prices was falling well before the fires. Looking at the index (and remember that values in indices are relative!) the price index fell from 1.1 to 1 from October 2008 to November 2008. That is a 9% fall in value in one month. The index looks to have fallen from 1 to 0.95 in the month after the disaster - that's (only!) a 5% fall. What's been happening in the market? From the peak of 1.4 in September 2008 housing prices had fallen almost 29% before the disaster!

Squinting at the chart (the very best way to do robust analytics!) I'd say, you'd be hard pressed to claim this bushfire disaster had a broad impact in the housing market.

2010–2011 – Queensland floods

Again, quoting Wikipedia:

A series of floods hit Queensland, Australia, beginning in December 2010. The floods forced the evacuation of thousands of people from towns and cities. At least 90 towns and over 200,000 people were affected. Damage initially was estimated at around A$1 billion before it was raised to $2.38 billion. The estimated reduction in Australia's GDP is about A$40 billion.

During the flooding it was reported that more than three-quarters of Queensland was affected by flooding.

How much of the state was impacted?

/images/2010-11-Queensland-floods-postcodes.thumbnail.png

At the time, I worked out, based simply on postcode, how many properties were in the impacted regions and their market value.

I found that 1.1 million properties were impacted - depending on who you ask (and that is another upcoming post!) there are around 10 million residential properties in Australia, so almost 10% of the country's residences were in postcodes impacted by this disaster.

I also found that the total market value of the impacted postcodes was $490B. The market value of all Australian residential properties was $5.9T in 2015, so 2011 would have been somewhere around $5B... We again see that around 10% of the country's residential property value was impacted.

What happened to the market?

Good question! Let's look at the Brisbane Local Government Area (LGA).

/images/2010-11-Queensland-floods.thumbnail.png

Sales Volumes

Sales volumes were already trending down. I'd suggest that sales volumes started falling mid-2009, but from Oct-2010 the fall in sales accelerated. Remember, the disaster was in December, so this fall preceded the flooding.

The volume of sales in January had more than halved, compared the two months prior. So the almost halving post disaster is hard to assign to either the disaster, or the prevailing economy.

Price indices

This is again interesting. After showing ~6% growth in the year 2009, the market for both houses and units flattened off. Corresponding, unsurprisingly, to the drop off in sales. Post disaster, prices dropped slightly, but nothing out of trend. So, it looks like there wasn't a significant impact on property prices. The prices in the market seem to reflect the economy prior to the disaster. Of course, I'd like a longer time series to verify... Any love (and data) coming from Hometrack?

Hypothesis

In this particular case I'd offer a reason why the market impact was limited.

/images/2010-11-Queensland-floods-street-level.thumbnail.jpg

This image was created by one of the awesome members of my team at Hometrack. Each blue dot shows a property that reported damage because of the floods. Feel free to stare with awe at the rows of properties and streets appearing as if by magic.

Now that you're back reading this, what else do you see? Not all properties (even in a suburb) were impacted. More importantly, for someone buying a house in the unaffected areas - your new house survived the worst that nature could throw at it. Makes the surviving properties attractive doesn't it?

Why aren't real estate prices impacted?

This is the core question and conclusion from this work - and the answer is that I don't know. It's a question I'll leave to Economists to post-justify a result.

I have some thoughts around the market returning to normal:

  • People have short memories.
  • Governments throw money at disaster recovery and reconstruction so the physical impacts are mitigated quickly.
  • Insurance companies get a lot of pressure to resolve claims quickly — conversely there are lots of news articles to the contrary.
  • Perhaps some disasters are not particularly disastrous to property?

Do you have a hypothesis? Tell me!

Presenting at the 2017 Q Group Colloquium

I'm excited to be presenting this year at The Institute of Quantitative Research in Finance Inc. (also known as the Q Group) colloquium. It's being held on Thursday, July 20.

I'll be presenting on

You can apply rigorous analytics to the real estate asset class (it doesn't have to be guesswork and slick salesmanship)

The Q Group website is a good starting point to find out about the group. Alternatively, we're on LinkedIn.

I'm biased. I've been fortunate to be the President of this great organisation since 2011.

Honesty in forecasting

The Australian media seems to be in love with talking about real estate. In particular bubbles, and (even though they don't use the term) forecasts.

Rather than commenting about what is being written, I want to talk about what is good about these discussions.

Something positive - Honesty

BIS Oxford Economics (previously BIS Shrapnel) needs to be acknowledged as a leader in the forecasting field.

Unlike a huge number of organisations - worldwide - BIS publishes its performance. In the BIS Shrapnel’s median house price forecasts report you can see their performance over time.

Their honesty is refreshing. It should be aspired to by anyone who purports to have a view of the future. This honesty is why they are one of the main economic forecasting organisations in Australia.

Perth

Perth is an interesting city for forecasting. Where (according to BIS Oxford Economics) both nationally, and for Sydney, price increases have been relatively monotonic. Forecasting something to go up when it keeps going up is easy... However, Perth has growth spurts and then the market comes off. What FUN!

Let's look at a simple part of the forecast. I want a forecast for Perth three years out. I also want to see how that forecast performs over time.

BIS Oxford Economics - Perth

/images/HPI-Perth-3y-forecast-BIS.thumbnail.png

In this plot the short coloured lines show the performance of the 3-year forecasts. I've highlighted the forecast beginning 2005. It underestimates the actual market performance by half. Ouch

Vifortech Solutions - Perth

/images/HPI-Perth-3y-forecast-Vifortech.thumbnail.png

Here we see the performance of the Vifortech Solutions 3-year forecast, again, against actual. The thick horizontal bars show the data window included in the forecast, then highlighting the point of the forecast.

I'm comfortable saying that I trust the Vifortech Solutions 3 year forecast more than the alternatives and it looks more predictive.

However, without the upfront honesty shown here how would we know? More importantly, what aren't we being told about their performance by other seers?

On the other hand...

Now, you can draw conclusions about what is missing in discussions about real estate and forecasting from what I'm not saying.

...or you can wait for another blog post.

More changes at CoreLogic International

Kyle Evans, the Chief Data Officer at CoreLogic International is leaving. His destination is Quantium.

This comes after the recent resignation of CEO Graham Mirabito.

Will this result in a change of direction? Only time will tell.