CSO house price index built on complex foundations

For most of us, the monthly house-price index compiled by the Central Statistics Office (CSO) boils down to one thing: the headline rate of (hopefully) growth.

In other words: is my home going up in value or am I sinking into negative equity? Should I hold off buying or will I lose out by waiting?

The practice of measuring property prices is, however, a more complex business than many credit. Different datasets can generate significantly different results. Monthly figures can also be skewed on the basis of location or house type, resulting in erratic price shifts which run counter to the underlying trend.

Without “quality adjusting” and “smoothing”, the statistical equivalent of ironing, trends can be exaggerated.

For instance, average monthly price changes during the crash exaggerated the fall in prices within indices available at the time as purchases tended to involve smaller properties in less attractive areas.

Disquiet

Following the crash and an understandable disquiet about the number of estate agencies publishing house-price surveys, the task of producing an accurate data series was handed over to the CSO.

In 2011, it began generating monthly numbers on the back of mortgage transaction data from the six main lenders.

However, it quickly became apparent the CSO’s index was failing to capture a significant portion of the market, namely cash transactions.

In the immediate aftermath of the crash, cash buyers accounted for one in five transactions but, by 2013, this had risen to one in two, meaning the CSO’s index was recording activity in only half the market, prompting claims that its index was not representative.

To address the problem, the agency decided to switch from using mortgage transaction numbers to using stamp duty data. But, while this got over the problem of excluding cash buyers, it had other drawbacks.

"The figures were very limited in terms of describing the properties," says Gregg Patrick, chief statistician at the CSO's residential property prices division.

The stamp duty returns provided only limited locational data and no details about the size of the property being sold.

To collate accurate data, Patrick says, “You need to distinguish between qualitative price changes [those controlled for location and property type] and pure price inflation.”

To square the circle, the CSO decided to try and match the stamp duty data for 160,000 property transactions dating back to the 2010 with two other building databases – the Building Energy Rating directory, which gives more detailed floor space information; and the GeoDirectory, which provided the CSO with the relevant eircodes.

However, the agency had no direct way of linking the databases; effectively there was no unique crossover code.

“It’s very rare for addresses to be spelled the same in each database, you can have all sorts of variations,” Patrick explains, noting the Irish-English component combined with abridged rural addresses complicated matters.

The CSO used an algorithm to match the addresses of about 60,000 properties in the three databases.

But that still left the majority – 100,000 properties – unmatched. That would have to be done manually and the job was handed to two CSO staff – a painstaking process that took more than three years.

The end product was the CSO’s newly revamped Residential Property Price Index, the State’s most definitive property price measure to date, which was launched last month.

More severe crash

By revisiting all the transactions since 2010, the CSO was able to reveal that Ireland’s property crash had been more severe than previously estimated, with the peak-to-trough fall put at 54.4 per cent rather than 50.9 per cent.

On the upside, the new index also suggested recovery has been stronger than previously stated, with prices rising by an average of 43.2 per cent from the trough rather than 37.4 per cent.

With the information from the GeoDirectory, the CSO also, for the first time, published the average price paid for properties broken down by eircode, revealing huge price disparities not just between Dublin and the rest of the State but also within the capital itself.

Take, for example, the difference between Dublin 6, which covers Ranelagh and Rathgar, where the average price paid for properties last year was €733,006, making it the most expensive eircode in the State, and Dublin 10, which includes Ballyfermot and Cherry Orchard, where the average price was just €157,527.

Or the fact that house prices in Kinsale, Co Cork, are now rising faster than anywhere else in the State, up 40 per cent in the last 12 months, while prices in Cahir, Co Tipperary, slumped 15.8 per cent during the same period.

For all the revelations though, Patrick admits his biggest surprise was how much the new statistics mirrored the old figures. Though the margins were different, the overall trend was the same, capturing the overall decline plus the bottoming out in 2012, he says.

The next phase of the CSO's property price evolution is likely to involve the inclusions of land plots. This data will have to be obtained from Ordnance Survey Ireland, says Patrick, and again there is no obvious way of aligning the datasets other than doing it manually.

Under new Eurostat regulations, the CSO is also under an obligation to collate separate indices for new and second-hand dwellings.

Of the total transactions recorded in August, 3,053 (93.2 per cent) were for existing dwellings while just 224 (6.8 per cent) were for new builds. This illustrates just how small the market for new builds currently is.

By using stamp duty data, the CSO’s new index also reveals the total value of transactions on a monthly basis – €843 million in August.

From the CSO’s perspective, the Irish market is relatively small in terms of transactions, leaving it more exposed to statistical noise. That said, the CSO has clearly invested considerable time and energy in getting a better handle on the one of Ireland’s most sensitive and divisive set of numbers.