Proximity factors that affect the resale price of public housing in Singapore
Singapore has one of the highest home ownerships in the world, ranking second globally with 91% of its citizens owning a property they call home. About 85% of the population live in Public Housing by the Housing Development Board (HDB), giving these properties its colloquial term “HDB”. Despite this, a report by CBRE Group, the largest commercial real estate services company in the world, ranks Singapore as the 2nd most expensive housing market in the world. Singapore not only has one of the most expensive houses, but is also known to be the most expensive country to own a car, where a Volkswagen Golf can cost you a cool $75,000 USD and up. As such, few Singaporeans drive, placing emphasis on the convenience of public transport, which translates to the importance of a strategically located home.
As such, this study aims to investigate how 3 proximity factors affect the resale price of a property. The factors are:
- Distance to the nearest train station
- Distance to the nearest shopping mall
- Distance to Singapore’s Central Business District (Raffles Place)
Methodology
Data Description
To better understand the resale prices of HDBs and how proximity factors affect its price, data was consolidated from several sources:
- The resale properties’ information and prices are available on the Singapore Government’s Data website as CSV files. https://data.gov.sg/dataset/resale-flat-prices
- OneMap Singapore, the authoritative national map of Singapore with the most detailed and timely updated information developed by the Singapore Land Authority was used to retrieve Geocodes/coordinates. OneMap provides API https://docs.onemap.sg
- Statista for Singapore Inflation Rates https://www.statista.com/statistics/379423/inflation-rate-in-singapore/
The raw data contains a wealth of information, providing data points on almost the exact location (the block, street name and storey range of the sold property without the exact unit number), the property size, age and its resale price.
A quick understanding of the data will show that the general trend in resale prices across the various flat types are relatively stable in their own bands and do not fluctuate much, with the exception of multi-generation flats. For simplicity, the study will focus on 4-Room Flats, the most common type of HDB Flats in Singapore.
Feature Selection
Crucial to the study, accuracy in distance measurements is fundamental in determining the importance of the proximity factors. To achieve this, distance calculations had to be made from address of the resold property to each of the three convenience factors — nearest train station, nearest mall and distance to CBD.
The list of train stations and malls were scraped from Wikipedia and the geocodes were retrieved from OneMap through its API based on each property’s address. Distances were calculated using geopy, a Python library, in kilometers.
The price of property was normalized to account for inflation, based on inflation rates from Statista and depreciation calculated linearly.
The above map uses Ang Mo Kio, one of the more established towns, to help visualise the problem. MRT stations are marked in red, malls in green and blocks in blue. Do note that a block can belong to Ang Mo Kio Town but be nearer to Bishan MRT Station.
Understanding trends
At this point, you might be wondering if there really is an effect on the price. Perhaps experienced buyers might have instinctively guessed it. Yes, the distance to the nearest MRT, mall and distance to Raffles Place, all have a negative correlation with price. Apart from the general downward trend, you might have observed and questioned the density on the bottom left quadrants of the graphs (A low price and a short distance to the MRT/mall), which goes against the hypothesis or some intuition. However, most of these data points belong to much older flats.
Note that the following graphs are presented — (Prices are calculated by Price-per-square-meter, distances are in kilometers)
Model
Given that we are performing an analysis on multi variables and proposing a trend these factors attribute to the price, I chose to conduct Multi-variable Regression.
In keeping things as simple, I started by testing the effectiveness of a Multi Linear Regression.
The R2 Score of the Linear Regression was 0.44, which is a promising start in proving the hypothesis that the proximity factors do influence price. However, I was confident a stronger case could be built. Hence, I proceeded to model with Polynomial Regression
I trained the model with poly-degrees up to 10, with 10 returning the best score of 0.6. However, we clearly can observe overfitting in some areas, predominantly in the initial two negative values at x=-4,000 and x=-1,900. “There is no such thing as a free lunch”. With no free lunch, there can’t be a free house.
Because the model trained itself to include the irrelevant outliers, I had to clean up badly predicted data points and retrained the model for an optimal poly-degree.
The resulting model after removing outliers and retraining yielded a score of 0.59 with poly-degree of 8.
Summary
In summary, resale prices of HDB flats in Singapore are influenced by a multitude of factors, and about 60% of the price is possibly affected by its proximity to a MRT station, a nearby mall or even Raffles Place. It supports the hypotheses and general buyers’ sentiments in Singapore but definitely highlights how the buyers’ population do not place that great of an emphasis on these factors as commonly heard.
Considerations
The three proximity factors are only valid for individuals who purchase a property based on the same premises as the hypotheses. I acknowledge that buyers’ habits can vary greatly and the study using these data points does not communicate the resale buying’s ground truth precisely.
- Distance to MRT — Whilst a shorter walking distance is preferable, the desirability can be affected by personal considerations such as train track noise and daily commute route, which might already be served well by alternative modes of transport like bus or car-pooling services.
- Distance to a Mall — A shopping mall brings about great convenience given the diversity of products and services offered. However, patronage is largely dependant on the relevance of the retailers to the buyers. Some considerations include the lack or limited choice of grocery chains and costlier food options.
- Distance to Raffles Place — It is a common trend for properties closer to the the center of a city to cost more, anywhere in the world. In Singapore, it is no different. However, the importance of staying close to Raffles Place largely depends on the lifestyle and daily commute of buyers.
Discussions
Other considerations to further improve the accuracy of the model can include precise modelling for the influence of inflation on housing prices and including the demographics of residents in each town. Furthermore, additional proximity factors could be added such as nearby educational facilities (in Singapore, children are enrolled into primary schools based on their residential address), Supermarkets and Hawker Centres, given how these are facilities more frequented as opposed to shopping malls.
Conclusion
As mentioned, I believe people purchase a property with their own laundry list of requirements. While the study pursued the investigation of common requirements, it has shown that while the proximity factors are crucial property selling points, it may not necessarily be the top-of-mind for everybody.
