Why SA’s house-price indices are inconsistent

Recent news coverage of disparities in reported average South African residential property values points to the different indices used. But no single index can provide the whole picture, explains Mark Bechard.

Home-price indices are often used, or perhaps abused, to “prove” various things about the state of the South African residential property market and, by implication, the value of your property. But what do home-price indices actually tell you, and why do they come up with different numbers?

This month, low- commission estate agency HomeBid released an index that, it says, shows the “true state” of the property market. According to HomeBid’s index, home prices increased, on average, by a nominal 0.94 percent in 2015 (see table, right).

Two widely quoted indices, the Absa House Price Index and the First National Bank (FNB) House Price Index, put last year’s nominal home-price increase at six percent, while property research company Lightstone says the increase was 5.5 percent. For 2014, the figures were 7.1 percent (FNB), 9.3 percent (Absa) and 6.72 percent (Lightstone).

Neville Berkowitz, a property economist and adviser to HomeBid, says that, once the 0.94-percent nominal increase is adjusted for inflation, average home prices declined in real terms by 4.2 percent in 2015. Inflation, as measured by the Consumer Price Index, was 5.2 percent in 2015, according to the South African Reserve Bank.

Berkowitz says HomeBid’s index is a better measure of the property market, because the underlying data is every residential property transfer recorded at the country’s 10 deeds offices. The index is a simple average: the total value of residential transfers in nominal terms divided by the total number of transfers. The Absa and FNB indices are based on home loans approved by these banks, and Lightstone excludes properties bought new, which, he says, comprise 25 to 35 percent of sales.

Lightstone also uses deeds-office data, but it is “cleaned” to ensure it accurately reflects price growth in the developed residential market, Paul-Roux de Kock, the director of analytics at Lightstone, says. Lightstone uses a “repeat sales” methodology, which means a property must be resold to be included in the sample.

Berkowitz says that, according to research by Absa and Lightstone, 65.5 percent of homes in South Africa are bond-free. “The banks’ samples on which they base their home- price increases may not, therefore, be fully representative of the entire residential market,” he says.

According to the South African Property Transfer Guide, 289 613 homes were transferred in 2015 (290 257 in 2014). Neither Absa nor FNB would disclose the size of the sample on which their house-price indices were based, saying this information is confidential.

De Kock says delays in registration mean that sales that took place in 2015 are still trickling into deeds offices. He says Lightstone currently has 301 970 transfers of properties sold in 2015 in its database, and expects this number to increase.

FNB INDEX

John Loos, FNB’s household and property sector strategist, says it is incorrect to assume that an index that does not use deeds-office data is flawed, adding that there is a lot more to consider than sample size when constructing and assessing an index.

Deeds- office data is also not absolutely representative of market trends, because certain market segments trade more than others, he says. Low-income earners are less mobile, so they transact homes far less frequently. Therefore, there is an inherent bias in transaction data towards higher- income segments where homes are sold more frequently, Loos says. The Lightstone and HomeBid indices do not adjust for this bias.

He says one of the main challenges in compiling an index is the impact of short-term shifts in transaction activity across segments (such as price bands or property types). He says the FNB index’s “fixed-weight average” methodology significantly eliminates the impact of shifts in transaction volumes across the index’s segments and subsegments by weighting them (full title, sectional title, number of rooms, and building size).

FNB also applies the following filters to reduce the impact of outliers and anomalies:

◆ Full- title properties with a stand size of less than 200 square metres and more than 4 000 square metres are excluded;

◆ Properties with a sale price below R20 000 and above R10 million are excluded; and

◆ Properties where the sale price was more than 130 percent or less than 70 percent of FNB’s valuation are excluded.

Loos says a problem with deedsoffice data is the delay between the sale and when the transfer is registered in the deeds office. Using FNB mortgage approvals as the sample ensures that the FNB index for February will capture all the transactions that occurred in that month.

Deeds-office data has other shortcomings. One cannot be certain that all commercial property has been eliminated. It also does not distinguish between a vacant stand and a building. This can become a problem during boom years, when building activity takes off, Loos says.

FNB publishes other property indices that use different samples and methodologies. For example, its Area Value Band House Price Index, which analyses price trends in six metropolitan areas, is based on transactions recorded in the deeds offices. The index is weighted to account for shifts in transaction volumes across four value bands: upper income, middle income, lower middle income and low income. The data is filtered to exclude most commercial transactions.

In 2015, the sample for FNB’s Area Band Index was 113 070 transactions, and the average nominal home-price increase in the six metropolitan areas was 6.4 percent, Loos says.

ABSA INDEX

Jacques du Toit, the property analyst at Absa Home Loans, says Absa’s House Price Index is weighted according to the sample of the three categories of homes that comprise the index. These are small (homes of 80 to 140 square metres), medium (141 to 220 square metres) and large (221 to 400 square metres). Size refers to the buildings only, not the stand. In 2015, the sample excluded properties priced at more than R4.2 million.

At the end of 2015, nominal yearon-year price growth was 4.7 percent in the small-home category, 5.2 percent in the medium category and 5.8 percent in the large category.

Du Toit says the prices are adjusted and smoothed in an attempt to exclude the distorting effect of seasonal factors and outliers in the data. He says seasonal factors, such as public holidays, school holidays and the changing seasons, may affect transaction volumes. Many people are away during holiday periods or are less inclined to look at buying property during winter months, while this may even differ from region to region within the country. This means that figures for December or July cannot be directly compared with figures for the immediately preceding or following months.

LIGHTSTONE INDEX

De Kock says that, to compare apples with apples, Lightstone uses the historical purchase prices of properties to calculate changes in value. The gains or losses of each property in the sample are combined to calculate the average annual price growth.

De Kock provides the following simple example to illustrate the potential problems with an averageprice index. If five Swiss chocolates, which cost R30 each, and five South African chocolates, which cost R10 each, were sold during 2014, the average price of the 10 chocolates sold would be R20 (R200 divided by 10). If, in 2015, only one Swiss chocolate was sold for R33, while nine South African chocolates were sold for R11 each, the average price of all the chocolates sold would be R13.20 (R132 divided by 10). A direct comparison of the average price of all the chocolates sold in each year would conclude that the price of chocolate fell by 34 percent (R20 versus R13.20) between 2014 and 2015.

The repeat- sales methodology would compare the Swiss chocolate sold for R33 in 2015 with the same five chocolates sold for R30 in 2014, and the nine South African chocolates sold for R11 in 2014 with the five sold for R10 in 2014. It would therefore correctly reflect that the price of chocolates rose by 10 percent, De Kock says.

He says Lightstone filters the data it uses to exclude non-residential and non-standard transactions, such as sales between family members ( usually for below- market prices), areas with a lot of government-subsidised housing, and transactions where the price growth is extremely different to the norm. By not filtering these transactions and outliers, there is a significant risk that the sample will include large commercial transactions, deedsoffice errors, zero-value transfers and other transactions that can skew the index.

Lightstone publishes various indices that indicate how the residential property market is performing. Its value-band graph shows growth in four price categories: less to R250 000 (low value); R250 000 to R700 000 (mid value); R700 001 to R1.5 million (high value); and luxury: (more than R1.5 million). In 2015, the nominal average price growth in these value bands was 31.7 percent (low value), 8.2 percent (mid value), 4.8 percent (high value) and 3.6 percent (luxury).

De Kock cautions that looking at a national index in isolation can be one- dimensional, because many areas and specific homes far out-perform (or under-perform) national house-price inflation.

“Location remains one of the top determinants of performance, and it is important to look at the microeconomic factors at play when assessing the true market value and capital growth of a home,” he says.