What is your total property return?

Measurement is a tricky thing. So many ways to get it wrong. So many of this incorrect or misleading measures are applied to property returns.

The 30 second view from a backward looking bird

The South African residential housing market has been entertaining to watch over several years now. At first, a sensible readjustment as confidence in our country and economy improved. Then, a further strong surge off the back of lower interest rates, themselves a result of imported deflation through a strengthening currency. The currency’s strength, hindsight shows us, was again due to improved confidence in our country, but with a large dose of commodity cycle boom. See Australia’s economy and currency for a similar pattern. There economy has a similar base to our own.

But then, with time, the property boom whizzed a little out of hand. Everybody (and their shoe-shine boy) was investing in multiple properties, any properties, as much as heavily geared 100% and more loans could buy.

Interestingly, well over a year ago First National Bank began tightening its credit granting criteria, and decreasing the concessions below prime offered. This was a clear decision to give up precious market share to maintain margins and risk levels. Well done FirstRand.

The more frightening eyes-forward reality check

And now the property market is looking rather sorry. Except for a few ridiculous estate agents punting the strong demand and limited supply (and nobody ever really believes them anyway) the market is on its way down. Standard Banks property measure (median of their sales) has been significantly down for several months now. ABSA’s (mean of their portfolio) is still positive, but only just.

Mean vs median vs trimmed mean

Standard Bank uses the median since it is more stable than the mean and less subject to outliers. This is true of the median, but I would be interested to see the results with a few other measures.  The trimmed mean is the average of values within a certain range (typically 10th percentile to 90th percentile, but different ranges are possible). This measure provides a better measure of overall central tendency, but limits the effects of outliers.

Weaving baskets

Both ABSA and Standard Bank use the ratio of the average (according to their respective measures) prices of houses within their portfolios acquired in month x to month (x-t) to measure the change in houses prices over period t from (x-t) to x. This captures three effects:

  1. Changes in average houses prices over the period
  2. Changes in the mix of houses sold nationally over the period
  3. Changes in the make-up of the sales to each bank over the period

Item 1 is what we want to understand. Items 2 and 3 are distortions to the numbers. Measurement errors, if you like.

Fortunately, since these portfolios are both significant portions of the total sales, the baskets are fairly representative of national sales. However, as economic conditions change, activity in different segments of the market would be expected to change. So it is clear that this is not only a source of increased error in measurement, but quite likely to be a bias.

Then, as touched on earlier in this post, banks will have different risk-appetites for mortgages at different points.

Incidentally, the competition commission would do well to notice these clear identifiers of competition between banks. These differences are the stuff of real cut-throat competition, with banks holding out as long as they dare before making a stand for profit margins, hoping that others will also breathe a sigh of relief and follow suit. This doesn’t reflect cartel-like behaviour, but rather tight competition where perfect competition pushes all participants in the great market game to the limit.

So as banks make decisions at different times, there share of the total property market will shift and change. This again introduces not only additional error but potential bias. Of course, it is conceivable that taking some sort of appropriately weighted average across all providers of mortgages could remove this problem. However, large as ABSA and Standard Bank are, they are by no means the entire market.

Glitch in the matrix

Standard Bank representatives have highlighted another specific problem area.

Unusual property market activity preceded the National Credit Act. The view put forward is that property prices experienced an upwards blip as market participants (blame spread between buyers, sellers, estate agents, mortgage originators and banks, but just not equally) rushed to process their purchases before the new rules came in to play and possibly restricted their transactions. The evidence would suggest that buyers were prepared to pay a little more to push through the deals. Or estate agents worked an extra few hours to earn their large commissions, since they had the most to lose with lower volumes of sales.

Like-for-like sales, a better alternative

Other property market indices in other countries have the luxury of more data. Their property indices are constructed by examining sales of the same house over time.

This may be best explained with an example. Two houses, A and B are sold for ZAR500,000 each in 2005. In 2006 (exactly a year later), house A is resold for ZAR550,000. A fair estimate of property price inflation over the year is 10%.  In the second year, house B is sold for ZAR616,000. The two-year return is measured as (616,000 / 500,000) = 23.2%.  The one year return over the second year is (1+23%)/(1+10%)-1 = 12%.  For those of you familiar with forward rates and spot rates, this should be ringing a bell about now.

Example 1 for measuring property price changes
House Sale t=0 Sale t=1 Sale t=2 Holding period return
A 500,000 550,000 10%
B 200,000 616,000 23.2%
Average 500,000 550,000 616,000

Now, for this example as provided, we could also take the ratio of sales in each year (year one: 550,000/500,000; year two: 616,000/550,000) and get the same growth. The two methods diverge when the house prices themselves are different from the outset. Let’s consider the scenario of houses C and D.

Example 2 for measuring property price changes
House Sale t=0 Sale t=1 Sale t=2 Holding period return
C 500,000 550,000 10%
D 200,000 246,400 23.2%
Average 350,000 550,000 246,400

We the figures from the above table, the results look very different on the two methods. The like-for-like method gives us identical returns to Example 1 10% and 12% for year one and two respectively. However, on the naive average of sale prices in each period, we get nonsensical results of 57.1% in year one and -55% in year two.

This example gives particularly poor results, because the change in mix of properties sold in each year are particularly different.

Almost the conclusion

So we have already explored some of the difficulties in measuring the actual change in property prices. The remaining step is to understand how property is measured as an investable asset in the financial press.

A typical approach is to consider the total capital growth or decline in property prices over the period. Sometimes commentators deduct CPIX off this nominal growth figure to determine the real capital growth or decline. If this result is negative, investors are deemed to have lost money in real terms. However, this is naive:

  • It ignores rental cashflows that would have been received over the period. Rental yields fluctuate, btu can range anywhere from 1.5% for high-priced residential houses to 10% for industrial rentals in unattractive areas. One must add this to the capital return to determine whether an investor has made a profit or loss over the period under consideration.
  • On the other hand, direct property is an expensive asset class to invest in. There can be an array of expenses incurred in the management of the investment. These should be deducted off the gross return calculated so far.
  • A more complex area is that of financing. Those who say property is a fantastic investment usually highlight the fact that “nobody is making any more land” (which is not strictly true, given Cape Town foreshore is reclaimed land, as well as parts of the Netherlands, Beirut and development around Dubai to name just a few of which I am aware) and that one gets to use the bank’s money to make money. If one considers the effects of gearing, one must also consider whether the financing is fixed or variable rate, and possibly into the realms of risk-adjusted returns
  • For taxable investors, one might want to consider tax and the extent to which expenses incurred are tax deductable, and the split between income tax on rentals and capital gains tax on, well, capital gains.

So the real conclusion is that measurement of returns to investors in property is a complex area. Although we may all feel we have a good understanding of property since many of us own property, that doesn’t provide instant certification to discuss the property market in the financial press and on financial websites.

Published by David Kirk

The opinions expressed on this site are those of the author and other commenters and are not necessarily those of his employer or any other organisation. David Kirk runs Milliman’s actuarial consulting practice in Africa. He is an actuary and is the creator of New Business Margin on Revenue. He specialises in risk and capital management, regulatory change and insurance strategy . He also has extensive experience in embedded value reporting, insurance-related IFRS and share option valuation.

Join the conversation


  1. David

    The one thibg about the so-called “property index” that has always amazed me is that it does not take into account the capital investement into the market.

    I buy a house for R600k, spend R500k fixing it up and then sell it for R1.2m. The bank’s index will view my house as having 100% growth in value in the period. I have barely met my carry costs.

    The extent to which the building activity (which mind you has slowed down a lot in the last few months) fueled the increase in house prices does not seem to be factored into the stats from 2002 to 2007.

    C ya


  2. I wish I had included that point in the main post. As you mention, this effect is likely to both bias the rate of growth in the index upwards and introduce a cyclical distortion over time as the extent of improvements changes.

    Some methods of adjusting the basket to reflect differences in erf size, number of rooms, house area etc. might take small elements on that into account. I’ve seen this done with GLM with mixed results. However, qualitative improvements (and knock-down-rebuild) will still be entirely missed.

Leave a comment

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.