IN a move to enhance the way private home prices here are tracked, the Urban Redevelopment Authority (URA) said it has revised its index methodology and expanded its data coverage to capture all private home transactions.
The revised approach for its property price index (PPI) maximises the use of data and is better able to distill a wide range of property attributes to measure a pure change in price.
This revision, in consultation with National University of Singapore real estate professor Lum Sau Kim, has come at a time when housing qualities in Singapore have widened with the ageing of existing houses and emergence of shoebox units since the index was last reviewed in 2000.
Starting from its first quarter computation of the PPI, URA has supplemented its existing data pool with stamp duty data from the Inland Revenue Authority of Singapore (IRAS). This is on top of using caveats lodged with the Singapore Land Authority for subsales and resales as well as the regular survey of developers for new sales.
URA has also switched to a more sophisticated method known as "stratified hedonic regression".
This new method works like a fine-tuning dial, accounting for fine differences between units such as size, age, and floor level to even micro-location factors such as proximity to MRT stations.
URA's previous "stratification" method involved sorting transacted property units based on housing type, tenure, completion status and region. The median prices for the categories are then aggregated using 12-quarter moving average weights of transacted values to compute the index.
While such a method has its own merits, it means that certain property attributes such as age and size are not accounted for. The use of moving weights also implies that changes in the index may be reflecting not just pure price changes but also a change in the weights. As an illustration, a sudden spike in the proportion of shoebox units transacted in one quarter, for instance, could possibly affect index movement in that period.
URA's revised method, however, is able to distill fine differences across property units to allow like-for-like comparisons. Transacted units are first sorted into five property types before a regression analysis is used to measure their price change during the period. Each of these property types is then assigned a fixed weight based on past five quarters of transacted values. The fixed weights are reviewed every three years.
This is similar to the revised methodology adopted by the Housing & Development Board for its HDB resale price index.
The index is also rebased from the fourth quarter of 1998 to the first quarter of 2009 to be consistent with the base period of the HDB Resale Price Index.
"We will also revise the methodology used to compute the private residential rental indices to the stratified hedonic regression methodology when we release the real estate statistics for first quarter 2015," URA said on Wednesday.
Back-testing of the revised approach shows that it does not change the direction of property price movements that were reported under the previous approach.
But under the revised approach, prices of non-landed properties would have fallen by a smaller 0.6 per cent quarter on quarter in the fourth quarter of 2014 instead of one per cent, while prices of landed homes would have dropped a larger 1.5 per cent instead of 1.3 per cent during the quarter.