One positive legacy of the property boom and bust is that we now have more data on Ireland’s housing market than ever before. JOHN McCARTNEY reviews the information available.
The Residential Property Price Register, which went live in September 2012, has been something of a game-changer in providing transparent information on housing transactions at a local level. Likewise, the Private Residential Tenancies Board (PRTB) index, which was launched last May, fills an important gap in our knowledge of Ireland’s rapidly expanding rented sector. But the biggest explosion of new data has come from the multitude of residential property price indices (RPPIs) that have emerged over the last decade. While these indices are undoubtedly useful, their proliferation, and the fact that they occasionally appear to be sending mixed signals, have the potential to cause confusion. This article aims to cut through some of this confusion by outlining the key characteristics of the main indices, highlighting their similarities and differences, and indicating how these properties are likely to influence measured price growth. It should be noted that, as the compilation of RPPIs is a highly technical subject, this article only attempts to provide a broad overview of the key issues.
A ‘constant quality’ index
In general, the aim of a price index is to compare the price of an identical basket of items over time. By ensuring absolute consistency in the basket of items between periods, this type of ‘constant quality index’ isolates the pure effect of market price changes. If we deviate from this constant quality principle and allow the basket to contain a different mix of items in each period, we can no longer tell whether an observed price increase is due to an underlying change in market conditions or due to compositional change within the basket of items being tracked. To illustrate: imagine that small apartments accounted for the bulk of property sales in one quarter. In the subsequent quarter, simply by chance, large houses accounted for a much higher proportion of sales. The result would be an increase in the average price of properties between the two quarters. But this does not reflect any underlying change in market conditions. Instead, it simply reflects a change in the mix of properties traded. The object of a constant quality index is to avoid this type of misleading conclusion.
Unfortunately, it is difficult to achieve a constant quality index for residential housing. On one hand, houses tend to be traded infrequently. This makes it highly improbable that repeated sales of the same basket of properties can be tracked over time. An obvious way to work around this would be to try and compare transaction prices for ‘equivalent’ baskets of properties. However, no two properties are identical – for example, even apartments in the same block will have different views, which will affect their relative prices. As such, it is impossible to identify entirely equivalent baskets of properties.
This article aims to cut through some of the confusion by outlining the key characteristics of the main indices, highlighting their similarities and differences, and indicating how these properties are likely to influence measured price growth.
Repeated valuation methodology
Faced with these challenges, statistical compilers have come up with two broad approaches to achieving a constant quality RPPI. The first – and most straightforward – approach is to identify a particular basket of housing units and to track repeated observations of their hypothetical value over time. This is the methodology that several Irish estate agents use to compile their in-house residential price indices. By maintaining a constant sample this method isolates the effect of market price changes. However, there are a number of practical drawbacks. Firstly, the estimated ‘price’ for each property in the basket is based on a valuer’s appraisal, and this approach inevitably involves some element of subjectivity. Secondly, while the selected properties were probably very familiar to the base period valuer (often an agent who had recently sold the property), over time people move on. This can lead to the repeated valuations eventually being taken over by persons who have less detailed information about the properties’ specific attributes. This could undermine the constant quality characteristics of the index. Thirdly, it is not clear that this methodology adequately accounts for depreciation on, and/or improvements to, the specific properties in question.
An alternative approach is to deploy actual price data, but to control for differences in the mix of properties traded between periods using sophisticated econometric techniques. The most common way of doing this is the ‘hedonic mix adjustment method’. This identifies a number of factors that are known to affect house prices – location, number of bedrooms, property type (detached, semi-detached, terraced, apartment), etc. The combined effect of these factors on price changes is then calculated and excluded. In principle, this leaves the element of price change that is solely attributable to underlying changes in the market. Most of the mainstream RPPIs, both internationally and in Ireland, are based on variations of this method. This includes the benchmark CSO index, the old ESRI/PTSB index, and the indices produced by property portals Daft.ie and Myhome.ie. This mix-adjusted approach has the advantage of being based on real price data. However, it does have some disadvantages. Obviously, the complex methodology makes it something of a black box to many users. Moreover, this method requires a dataset that not only includes price information (sales price or asking price) but also hard data on the various factors that are expected to influence prices (location, bedrooms, etc.). In this context, it is frustrating that the Residential Property Price Register currently contains very limited descriptive information on properties that have been sold. While it is relatively straightforward to classify indices into our repeated valuation and mix-adjusted categories, significant variation exists within these groups. Much of this relates to detailed technical differences, and it can be difficult to assess the impact of these on the performance of the indices. But other differences are more obvious and their potential impact on the measurement of price growth is discussed below.
Indices based on mortgage drawdowns
The old PTSB/ESRI index and the current CSO index are based on information provided by lending institutions about properties against which they have issued mortgages. There are two practical consequences of this. Firstly, up to three months can elapse between a purchaser’s offer being accepted (the point at which the price is set) and the mortgage being drawn down. This introduces a measurement lag into indices that are based on mortgage drawdowns. Therefore, in a rising market these indices may be expected to understate the extent of real-time price increases. Secondly, in recent years there has been a strong upsurge in houses being bought entirely with cash, and an estimated 58% of total sales in the first half of 2013 were all-cash transactions. As such, the CSO index is currently missing a large chunk of marketplace activity. If cash sales are concentrated in sections of the market where prices are rising at a different pace than elsewhere, this could bias the CSO’s measurement of house prices. In the first half of 2013, 75% of Savills’ residential sales in the €1m+ price bracket were cash-only deals, compared with only 39% of sales below €1m. Anecdotally, some agents believe that prices at this upper end of the market are also rising more quickly. If so, then the CSO index could potentially be understating price growth. The CSO has recognised the importance of including cash- as well as mortgage-based transactions, and is currently working on a number of initiatives to quantify the problem and address it if necessary.
Asking vs. selling price
The CSO index, the old PTSB/ESRI index and the new Daft sales price index are all based on actual selling prices. In contrast, the traditional Daft.ie index and the Myhome.ie index are based on asking price information, which has been harvested from online advertisements. Asking price data provide a window into the psychology of vendors and their expectations. However, as acknowledged in recent Myhome.ie reports, asking prices may tend to undershoot actual sales prices during a boom, and exceed them in a downturn.
The Residential Property Price Register
Because it does not keep a constant mix of properties between periods, the Residential Property Price Register cannot be considered a true price index. Despite this, however, analysts may be tempted to derive average prices from the register and track these over time. Statisticians caution against this for two reasons. Firstly, when there is a big shift in the mix of properties being traded, such an exercise can result in biased estimates of price growth. For example, many agents would say that there has been a flight to quality over the last five years as people who were previously priced out of bigger properties in better locations took the opportunity to pounce when prices collapsed. Therefore, not only did house prices fall during the recession, but the quality of the product that buyers were getting for their money also improved. A time series of average prices does not take this offsetting effect into account, and therefore underestimates the true level of house price decline during the recession. Secondly, because there can also be purely random fluctuations in the mix of traded properties between time periods, a simple average price index is likely to prove inherently volatile. From a practical perspective, this could make the results difficult to interpret in the short term.
Ireland’s property crash led to a recognition that better data were needed for market analysis. This has resulted in a proliferation of new RPPIs. While all of these indices add value to our understanding of the market, they differ in several important ways. This article simply highlights the key points of difference between the indices and suggests how these might affect measurement outcomes.
The first key message to emerge is that, in any given period, the idiosyncrasies of the individual indices can give rise to different signals about price growth. For this reason, analysts should be cautious about making hard-and-fast inferences about the Irish housing market based on data from any single period.
Critically, however, these anomalies tend to iron themselves out over a longer period, and it is clear from Figure 1 that all of the main mix-adjusted indices tell a broadly similar story over time. Therefore, a second key finding is that the indices do appear to give a reliable and consistent view of market trends over the longer run.
A third key conclusion is that caution should be exercised in deriving price indices from the Residential Property Price Register. As shown in Figure 1, such indices tend to be inherently volatile. Moreover, when the mix of properties being traded in the market is shifting, they may also be significantly biased. This notwithstanding, the Residential Property Price Register has enormous potential to be developed further as a data resource. With a new postcode system coming into place before the end of 2015, the potential exists for the register to gather more granular information on housing transactions by location. In addition, the register could be expanded to capture more detailed information on the characteristics of the properties being sold. These developments would allow for a more sophisticated mix-adjusted index that would not only be based on a larger number of transactions, but would also incorporate both mortgage- and cash-financed sales.
The author would like to thank Niall O’Hanlon of the CSO, Dr Ronan Lyons of TCD and Daft.ie, Graham Neary of Myhome.ie and Ray Hanley of Savills for their assistance. Any remaining errors are the author’s.