Homepage
PR Newswire
11-03-2015, 04:44 PM
(This post was last modified: 11-03-2015, 05:13 PM by HyperionTree.)
Property price indices are used by investors to gauge the general price level of properties in the overall market. There are many ways to calculate property price indices, each with pros and cons. Thus, understanding how property price indices are calculated is crucial to appreciate how useful are property price indices in estimating the price level the general market.
In this post, Hyperion and Tree will review the methodology and evaluate various common property market price indices in Singapore. Finally, Hyperion and Tree will suggest a new private condominium property price index based on a proprietary methodology invented by Hyperion for comments and suggestions.
Types of Property Market Index in Singapore
In Singapore there commonly used property market indices are:
1. URA Non-landed Private Property Price Index;
2. URA HDB Resale Price Index;
3. Singapore Real Estate Exchange (SRX) Non-landed Private Proeprty Index; and
4. SRX HDB Resale Price Index.
These indices are usually provided for three regions: Core Central Region, Rest of Central Region and Outside Central Region.
URA Non-landed Private Property Price Index
Based on news reports, the URA Non-landed Private Property Price Index is a weighted average of transacted property prices of different types of properties. To quote ST Property, "The weights used to compute the index are derived by taking the moving average of the values of properties transacted in each market segment over 12 quarters. Thus different months will have different weights. For instance, there could be many luxury homes sold in a quarter due to a fund exiting the market - a one-off exception as the high-end market has been languishing with anaemic sales. The averaging out of the weights across a period prevents this transaction from causing the city centre segment's weighting to jump sharply since this is an anomaly in the market's sales."
Source: http://www.stproperty.sg/articles-proper...s/a/113301
Recently, the government is reviewing the index and might change the methodology to a hedonic regression model.
URA HDB Resale Price Index
An article in Property Guru explained that the HDB RPI is "computed using the stratification method, with a representative basket of towns and flat models. Resale prices are grouped into segments based on flat types, models and regions. The average prices for each segment are then aggregated using 12-quarter moving average weights to derive the index.
Starting from Q4 2014, HDB will switch to the stratified hedonic regression method to compute the RPI. This method will control for variations in flat attributes, such as proximity to amenities, age and floor level, to derive the general price movements in each segment. These are then aggregated using five-quarter capital value fixed weights to derive the aggregate price change. To better reflect prevailing market structure, the weights will be updated once every three years."
As for what is hedonic regression method, we'll explore it in the discussion of SRX indices.
Source: http://www.propertyguru.com.sg/property-...-all-towns-
SRX Indices
SRX indices are constructed based on a hedonic regression. Hedonic regression is a method to quantify the effects of qualitative attributes, like distance to MRT, distance to schools etc, and quantitative attributes like area, on condo prices. A typical regression equation for hedonic regression is as follows: Ln (Price per sqf) = a(Ln(MRT distance)) + b(School) + C1(T1)+C2(T2)...
the variables C1, C2 which are coefficients to the time variable T1(month 1), T2(month 2) would be used to construct the property price index. For example, the average condo price change from month on month would be (e or 2.71828)^(C2-C1)
Source: http://static.streetsine.s3-website-ap-s..._paper.pdf
The main reason why hedonic regression is used to estimate condo price increases is to control for qualitative and quantitative effects. For example, if a new condo with larger area for each condo unit was recently built, and this is added into the property price index, it might show that property prices drop on a per psf basis. However, the reason it drop was not only due to more supply, but also because of larger area for each condo unit (condo units with larger area are cheaper on a Price psf basis usually). Thus there is a need to take out the effect of larger area to get the actual price decrease. This is why it is not possible to just compare changes in PSF prices month over month without adjustments to ensure you get an apples to apples comparison.
Other methods not used in Singapore
There are other methods that can be used. These are the repeated-sales method which is commonly used in the USA and the stratified median method. Most parts of the world uses the hedonic regression method to calculate the property price index.
New Property Price Index - Peer Group Price Index
Hyperion and Tree discussed the various property price indices mentioned above and decided to invent another way to calculate the property price index. Based on a proprietary methodology, Hyperion and Tree were able to group properties in Singapore into peer groups base on similar characteristics of the properties like area and distance to MRT etc. The properties in each peer group likely compete with each other for buyers and thus their prices would likely move together based on the supply and demand for the particular peer group. Based on these peer groups, Hyperion and Tree calculated the price per sqft changes over the last few years using the publicly available caveat data made available by URA. For example, the chart below consists of many peer groups for condos between 900 to 1100 square feet located at the 6 to 10 floor.
[Image: 2ci9qw8.jpg]
The y-axis is the price per sqft while the x-axis is the time. There are many peer groups, but Hyperion had put each adjacent peer groups together to form packets of 50 condos and calculated the 6 month moving average price psf. As such, each line represent 50 condos each.
As observe in the graph above, the peer group in grey had actually increased before falling, while other peer groups at the bottom did not experience any significant changes in prices.
These property price indices have the advantage of being able to reflect the real changes in prices of a group of properties with similar features and thus avoid the problem of calculating the weightings to average out the changes in price psf. Besides, investors can just track a particular peer group to see how the value of their property is changing.
Further using this method, Hyperion and Tree can identify peer groups for condominiums that have data. For example, for a condo at Blue Horizon in district 5 with an area of 1150 and floor of 16 to 20, the peers would be units at floor 16 to 20 with similar area at Hillsta, The Gardens at Bishan, Tree House, Central Grove, H2O Residences, Cassia View, Dunman View, Seasons View and Oleander Towers. Thus, the investor would need only to track the peers to estimate the value of its property.
Location
Interestingly, the peer groups reveal that for sub-urban condos, the condo peers are actually located in different districts. For condos in the central region, their peers are in the central. This means that if one were to value a sub-urban condo based on nearby transactions of other condos within a 1km radius, it will not be accurate because these condos are different. Comparing neighboring prices only works in central regions.
Why do condos that are near by be in different peer groups? It is likely that developers who build new condos besides existing condos will want to differentiate the condo to reduce price competition. When there is less price competition due to differentiation, developers can gain more profit. This is supported by a real estate theory called the Central Place Theory.
Conclusion
There is a need for more nuance property price indices so that investors can better gauge the price level in a market for a particular type of property. Further, buyers can actually find a list of properties to shop around, once they figured out which properties are peers. Valuing a sub-urban property based on nearby transactions are not accurate because the peers are actually further away.
Please let us know your views.
Thanks.
Cheers,
Hyperion and Tree
In this post, Hyperion and Tree will review the methodology and evaluate various common property market price indices in Singapore. Finally, Hyperion and Tree will suggest a new private condominium property price index based on a proprietary methodology invented by Hyperion for comments and suggestions.
Types of Property Market Index in Singapore
In Singapore there commonly used property market indices are:
1. URA Non-landed Private Property Price Index;
2. URA HDB Resale Price Index;
3. Singapore Real Estate Exchange (SRX) Non-landed Private Proeprty Index; and
4. SRX HDB Resale Price Index.
These indices are usually provided for three regions: Core Central Region, Rest of Central Region and Outside Central Region.
URA Non-landed Private Property Price Index
Based on news reports, the URA Non-landed Private Property Price Index is a weighted average of transacted property prices of different types of properties. To quote ST Property, "The weights used to compute the index are derived by taking the moving average of the values of properties transacted in each market segment over 12 quarters. Thus different months will have different weights. For instance, there could be many luxury homes sold in a quarter due to a fund exiting the market - a one-off exception as the high-end market has been languishing with anaemic sales. The averaging out of the weights across a period prevents this transaction from causing the city centre segment's weighting to jump sharply since this is an anomaly in the market's sales."
Source: http://www.stproperty.sg/articles-proper...s/a/113301
Recently, the government is reviewing the index and might change the methodology to a hedonic regression model.
URA HDB Resale Price Index
An article in Property Guru explained that the HDB RPI is "computed using the stratification method, with a representative basket of towns and flat models. Resale prices are grouped into segments based on flat types, models and regions. The average prices for each segment are then aggregated using 12-quarter moving average weights to derive the index.
Starting from Q4 2014, HDB will switch to the stratified hedonic regression method to compute the RPI. This method will control for variations in flat attributes, such as proximity to amenities, age and floor level, to derive the general price movements in each segment. These are then aggregated using five-quarter capital value fixed weights to derive the aggregate price change. To better reflect prevailing market structure, the weights will be updated once every three years."
As for what is hedonic regression method, we'll explore it in the discussion of SRX indices.
Source: http://www.propertyguru.com.sg/property-...-all-towns-
SRX Indices
SRX indices are constructed based on a hedonic regression. Hedonic regression is a method to quantify the effects of qualitative attributes, like distance to MRT, distance to schools etc, and quantitative attributes like area, on condo prices. A typical regression equation for hedonic regression is as follows: Ln (Price per sqf) = a(Ln(MRT distance)) + b(School) + C1(T1)+C2(T2)...
the variables C1, C2 which are coefficients to the time variable T1(month 1), T2(month 2) would be used to construct the property price index. For example, the average condo price change from month on month would be (e or 2.71828)^(C2-C1)
Source: http://static.streetsine.s3-website-ap-s..._paper.pdf
The main reason why hedonic regression is used to estimate condo price increases is to control for qualitative and quantitative effects. For example, if a new condo with larger area for each condo unit was recently built, and this is added into the property price index, it might show that property prices drop on a per psf basis. However, the reason it drop was not only due to more supply, but also because of larger area for each condo unit (condo units with larger area are cheaper on a Price psf basis usually). Thus there is a need to take out the effect of larger area to get the actual price decrease. This is why it is not possible to just compare changes in PSF prices month over month without adjustments to ensure you get an apples to apples comparison.
Other methods not used in Singapore
There are other methods that can be used. These are the repeated-sales method which is commonly used in the USA and the stratified median method. Most parts of the world uses the hedonic regression method to calculate the property price index.
New Property Price Index - Peer Group Price Index
Hyperion and Tree discussed the various property price indices mentioned above and decided to invent another way to calculate the property price index. Based on a proprietary methodology, Hyperion and Tree were able to group properties in Singapore into peer groups base on similar characteristics of the properties like area and distance to MRT etc. The properties in each peer group likely compete with each other for buyers and thus their prices would likely move together based on the supply and demand for the particular peer group. Based on these peer groups, Hyperion and Tree calculated the price per sqft changes over the last few years using the publicly available caveat data made available by URA. For example, the chart below consists of many peer groups for condos between 900 to 1100 square feet located at the 6 to 10 floor.
[Image: 2ci9qw8.jpg]
The y-axis is the price per sqft while the x-axis is the time. There are many peer groups, but Hyperion had put each adjacent peer groups together to form packets of 50 condos and calculated the 6 month moving average price psf. As such, each line represent 50 condos each.
As observe in the graph above, the peer group in grey had actually increased before falling, while other peer groups at the bottom did not experience any significant changes in prices.
These property price indices have the advantage of being able to reflect the real changes in prices of a group of properties with similar features and thus avoid the problem of calculating the weightings to average out the changes in price psf. Besides, investors can just track a particular peer group to see how the value of their property is changing.
Further using this method, Hyperion and Tree can identify peer groups for condominiums that have data. For example, for a condo at Blue Horizon in district 5 with an area of 1150 and floor of 16 to 20, the peers would be units at floor 16 to 20 with similar area at Hillsta, The Gardens at Bishan, Tree House, Central Grove, H2O Residences, Cassia View, Dunman View, Seasons View and Oleander Towers. Thus, the investor would need only to track the peers to estimate the value of its property.
Location
Interestingly, the peer groups reveal that for sub-urban condos, the condo peers are actually located in different districts. For condos in the central region, their peers are in the central. This means that if one were to value a sub-urban condo based on nearby transactions of other condos within a 1km radius, it will not be accurate because these condos are different. Comparing neighboring prices only works in central regions.
Why do condos that are near by be in different peer groups? It is likely that developers who build new condos besides existing condos will want to differentiate the condo to reduce price competition. When there is less price competition due to differentiation, developers can gain more profit. This is supported by a real estate theory called the Central Place Theory.
Conclusion
There is a need for more nuance property price indices so that investors can better gauge the price level in a market for a particular type of property. Further, buyers can actually find a list of properties to shop around, once they figured out which properties are peers. Valuing a sub-urban property based on nearby transactions are not accurate because the peers are actually further away.
Please let us know your views.
Thanks.
Cheers,
Hyperion and Tree
