Is the CSO Property Price Index becoming misleading?

I have been a big fan of the CSO’s Residential Property Price Index since it’s launch in May 2011. Their figures have so far covered the period of Jan 2005 to Dec 2011.

However, they are now imo in dangerous territory of misleading the public on current residential prices (and drops.)

This is because not only are they based on transaction dates (at prices often agreed by buyer and seller weeks or months before, therefore most likely at slightly higher prices than currently achievable), they also do not include Cash Transactions.

The CSO use Revenue figures for Coverage of stamp duty returns versus mortgage drawdown returns. But they only have figures from 2005-2009. In 2009 just 6% of transactions (by volume) were in cash.

Latest reports are that Cash Transactions are now up at 15-25% of all sales. And cash sales are usually at lower prices than mortgage sales. If this is true, the CSO is understating, perhaps substantially, the decline in prices.

Namawinelake has also picked up on this and offers a commentary on the Dec 2011 figures here

He also mentions “The CSO expects to have monthly data from the Revenue Commissioners from mid-2012 and it expects that it may subsequently be able to show the market size with its monthly release of the residential index.”

This CSO updated data would be very welcome. At the very least, it now being Feb 2012, they should have the 2010 stamp duty return figures.

+1 - flawed from the get-go…

Indeed, and I think you have another month after closing during which to pay Stamp Duty, which I believe is the transaction date that will be provided to the CSO. In our own case, without even a mortgage involved to complicate things it was a full three months between paying booking deposit and paying stamp duty.

More detail on this CSO timeseries can be viewed here;

It is quite technical and it is on my list to review at some stage.

I have written lengthily and pretentiously on this elsewehere here as a basis for forecasting property prices.

The potential issues with this are:

• The underlying data is incomplete - they say they get stamp duty data so it should be complete, apart from data quality errors

• The underlying data is sparse - the number of transactions is low so the sample is not representative - this is certainly an issue

• The underlying methodology for calculating RPPI is faulty because of the complexity of property types - this may be an issue

The following describes how indices such as CPI are calculated. CPI is easier as the product mix is well-defined. Property is much less so.

Take a set of products and their weight within the overall CPI:

A B C D E F 1 Period 1 2 3 4 2Price 1 100 101 102 103 3 2 101 102 103 104 4 3 102 103 104 105 5 4 103 104 105 106 6 5 104 105 106 107 7Weight/Volume 1 10 10 10 10 8 2 5 5 5 5 9 3 7 7 7 7 10 4 8 8 8 8 11 5 3 3 3 3
The change in prices for this collection of products from Period 1 to Period 2 would be:

SUM i=1 to Product (Product_Price(Period 2,i) x Weight/Volume (Period 1,i)) ] /

SUM i=1 to Product (Product_Price(Period 1,i) x Weight/Volume (Period 1,i)) ]

which for the sample data listed above is:

(D2C7+D3C8+D4C9+D5C10+D6C11 ) / (C2C7+C3C8+C4C9+C5C10+C6C11) = 1.009836066

Similarly the change from Period 2 to Period 3 for the sample data listed above would be:

(E2D7+E3D8+E4D9+E5D10+E6D11) / (D2D7+D3D8+D4D9+D5D10+D6D11) = 1.00974026

However, with property there is no standard product type.

The pricing model uses simple variables such as:

• Location Dublin - postal district or local authority area
• Property Type - detached, bungalow, semi-detached or terrace
• Number of bedrooms - 1-2, 3, 4, 5+
• Floor area - Actual floor area in metres squared
• New or old property
• First time buyer

So, for example, all 4 bedroom terraced houses of 160 square metres in Dublin 6 are counted as the same.

If you have a situation where cheaper houses of this type sell more and more expensive ones sell much less frequently (for reasons such as because less people can affod them or sellers are less willing to sell and take their larger loss) then the index will generate an RPPI that is higher than actuality. So in general when cheaper houses of a given type sell more, the index gives a higher drop than reality.

Even within this classification, there is much room for variation.

It currently the best index because it is the only one that is based on actual sales prices rather than asking prices.

I do not view the RPPI as flawed. It does track other indices.

The CSO do not publish the underlying data, even anonymously. So it is not possible to perform your own analysis. The promised property price database will enable such analysis.