07/06/2011: Residential property prices

An impressively decent dataset from CSO on residential property prices has been released for the second monthly installment, so here are the charts and some high level analysis.

• Overall Residential Property Price Index (RPPI) for April was 78.2 or 0.8 points below March levels. Hence, mom the index has fallen 1.013% and is now 1 point below its 3mo MA. Year on year the index has fallen 12.233% and relative to peak of 130.5 reached in September 2007 it is now down 40.077%.
• Overall RPPI has recorded its 8th month of consecutive declines having risen statistically and economically insignificant 0.11%mom in August 2010. Year on year, April marked 38th consecutive month of declines.
  
• April index for houses fell 0.9 points to 81.3, down 1.095% mom, or 1 point below 3mo MA. Year on year index has fallen 12.013%. The peak for this sub-index was reached in September 2007 at 132.0.
• April index for apartments fell to 60.4, down 0.6 points - a mom decline of 0.984% and a yoy decline of 15.288%. April reading was 1.233 points below 3mo MA. This sub-index peaked at 123.9 in February 2007.
• Dublin properties sub-index has fallen 0.5 points in April to 70.5, a decline of 0.704%mom or 12.963% yoy. The sub-index now stands 0.77 points below 3mo MA and 47.584% below the peak of 134.5 in February 2007

Charts to illustrate:
To summarize - the deflation of house prices continues, although the monthly rate of decline has now fallen below both 6mo and 12mo average. This, however, might be due to seasonality, since April marks a relatively moderate month in terms of price movements in every year since 2008. house prices have now fallen 38.41% since their peak, while apartments prices have declined 51.25% from their peak.

It is worth noting - not as a criticism of the CSO, since it cannot do anything about the data - that the index is computed based on mortgages drawdowns, hence excluding any share of transactions that might take place on the 'gray market' (tax evading payments, swaps etc), as well as cash-only purchases and mortgages issued by lenders other than the 8 largest lending institutions from which the data is available.

Another issue, again - little that CSO can do for this - relates to hedonic adjustments undertaken in index computation. Hedonic characteristics used by CSO exclude a number of relevant parameters, such as number of bathrooms and the site size, as well as existence of garage and/or off-street parking. This, alongside with the tendency - due to planning permissions restrictions - to under-report actual floor area and number of bedrooms - means that the hedonic model might be relatively weak.

Finally, CSO employes a Laspeyers-type indexation method, which is "calculated by updating the previous month’s weights by the estimated monthly changes in their average prices". However, like all types of indices, Laspeyers indices suffer from some specific drawbacks. In particular, these indices are weaker in periods of adjustment in the markets. Here's a quick non-technical discussion:

Laspeyers index is designed to answer the question: "How much is the sales price today for the house that is of the same quality as in the base year (2005)?" Quality is compared using the hedonic model mentioned above, based on specific size of the house (floor area), its amenities (number of bedrooms, house type) and location (note - we do not know the granularity of such 'location' adjustment, which can be critical. For example, I live in Dublin 4, but not the "fashionable" part of it. This means that if location code used is D4 for my house, it will receive signficantly higher locational weight relative to true value of my location than a house in a "fashionable" D4 locale.

One key objection to Laspeyers index is that it is computed while assuming that the base year (2005) house remains unchanged over time. Hence, quality is assumed to be constant for referencing, implying the index over-states inflation and under-states deflation.

In addition, index does not capture the effects of substitution in housing. In other words, Laspeyers index does not reflect conversions of house features to substitute away from more expensive options, etc, or purchases shifting in favour of smaller properties.

Index also assumes that geographical distribution of house sales does not change over time - a feature that introduces significant biases into the index when locational markets are not uniform (when there are significant differences within the markets).

Finally, the index overstates price appreciation at the peak of the bubble, since at that point, less desirable properties were disproportionately represented in the market as buyers chased any home available for sale. This is known on the basis of the US data where at the top of the markets 'gentrification' of lower quality locations in many states has led to Laspeyers indices understating price inflation.

For thes reasons, Laspeyers indices are known as 'constant quality' indices.

Chain-linked indexation, employed by CSO, helps addressing some of these issues, but it does not eliminate them. Of course, that too has its drawbacks, namely the more substantial data requirement, plus the lack of index additivity (you can see this indirectly in the first chart above by the gravitational pull of the houses index on overall index.