Statement of Business Task

Singapore has consistently ranked among the most expensive cities in the world. In fact, Singapore held the No.1 spot for 2017 to 2019, and No.2 in 2020. In 2021, it dropped out of the top 10, falling to No.13 but recovered to No.5 in 2022. The cost of housing plays a major part in keeping Singapore in such a high position.

It is important to understand that Singapore’s Housing market is broken up between the vastly expensive apartments and condominiums and more affordable HDBs favoured by most Singaporean residents. Condominiums or simply “Condos” typically have more facilities like pools, small indoor gyms etc. Apartments offer a mixture of facilities. Some have pools but not all and are usually within a gated area. HDBs are openly accessible public apartments with typically no additional facilities offered for the previous categories.

My analysis in this report will focus on the sale of apartments and condominiums that occurred between July 2017 and July 2022 in all 28 postal districts. I would have liked to incorporate the HDB market as well but the volume of data would make the analysis too long so I have decided to do a subsequent analysis when I have a chance.

The key question the analysis will attempt to answer is as follows: Are private housing prices rising outside the reach of average residents based in Singapore as it may seem? Or is there still hope? What is true and what is hype?

A small confession on my part is I have harboured dreams of moving to Singapore for a long time so my interests align with every other expat. But expat or not, the question will always be do you choose the more expensive condos or apartments with better facilities or choose the more affordable HDBs with less?

Data Source

The data I used for this analysis came from the website for the Singaporean Urban Redevelopment Authority itself. The URA’s goal is to make Singapore a great city to live, work and play so they are very keenly aware of all developments relating to the real estate market.

The data collected adheres to the ROCCC principle. It is reliable, original and complete because it is collected from each transaction and the data is collected directly by the URA. It is current, complete and cited because data is limited to the last 5 years, it is used by the URA themselves to conduct their own studies.

Included in the data set are the following with the data types:

Title Datatype Description
Project Name CHR Name of the development
Street Name CHR Street name where the development is located.
Type CHR Condominium or Apartment
Postal District INT Singapore has 28 Postal Districts. It is commonly referred to in its short form. D1 is District 1 and within each district are several townships.
Market Segment CHR There are 3 regions in the market segment: CCR, RCR and OCR.
Tenure CHR There are two major categories - Freehold and Lease. Most properties are considered leased from the government. The duration varies but typically it is a 99 year lease. 
Type of Sale CHR New Sale, Resale, Subsale
No of Units INT Number of units sold.
Price INT Price of property sold in Singapore dollars.
Nett Price CHR Price minus the value of any benefits in Singapore dollars.
Area sqft INT Property in square feet
Type of Area CHR The most common type is Strata meaning the development on the land can be of most types. For example, terraces, standalone homes etc.
Floor Level CHR The range of the floor level of the property sold in the high rise development.
Unit Price per sqft INT Unit price per square feet of property sold
Date of Sale CHR Date of the transaction

In order to have a little more context during our discussions, I have included the following table. It breaks down the township within each district and indicates which market segment they belong to. ( CCR - Core Central Region, RCR - Rest of Central Region and OCR - Outside Central Region)

District Name Market.Segment
1 Boat Quay, Chinatown, Havelock Road, Marina Square, Raffles Place, Suntec City CCR
2 Anson Road, Chinatown, Neil Road, Raffles Place, Shenton Way, Tanjong Pagar CCR
3 Alexandra Road, Tiong Bahru, Queenstown RCR
4 Keppel, Mount Faber, Sentosa, Telok Blangah RCR
5 Buona Vista, Dover, Pasir Panjang, West Coast RCR
6 City Hall, High Street, North Bridge Road CCR
7 Beach Road, Bencoolen Road, Bugis, Rochor RCR
8 Little India, Farrer Park, Serangoon Road RCR
9 Cairnhill, Killiney, Leonie Hill, Orchard, Oxley CCR
10 Balmoral, Bukit Timah, Grange Road, Holland, Orchard Boulevard, River Valley, Tanglin Road CCR
11 Chancery, Bukit Timah, Dunearn Road, Newton CCR
12 Balestier, Moulmein, Novena, Toa Payoh RCR
13 Potong Pasir, Machpherson RCR
14 Eunos, Geylang, Kembangan, Paya Lebar RCR
15 Katong, Marine Parade, Siglap, Tanjong Rhu RCR
16 Bayshore, Bedok, Chai Chee OCR
17 Changi, Loyang, Pasir Ris OCR
18 Pasir Ris, Simei, Tampines OCR
19 Hougang, Punggol, Sengkang OCR
20 Ang Mo Kio, Bishan, Braddell Road, Thomson RCR
21 Clementi, Upper Bukit Timah, Hume Avenue OCR
22 Boon Lay, Jurong, Tuas OCR
23 Bukit Batok, Choa Chu Kang,Hillview Avenue, Upper Bukit Timah OCR
24 Kranji, Lim Chu Kang,Sungei Gedong, Tengah OCR
25 Admiralty, Woodlands OCR
26 Tagore, Yio Chu Kang OCR
27 Admiralty, Sembawang, Yishun OCR
28 Seletar, Yio Chu Kang OCR

Property prices tend to go on the order of CCR, RCR and OCR. With CCR typically where the more high end properties are and OCR being on the other side of the spectrum. This provides a fairly high level view as townships can be in one or two of these market segments as the data will show later.

I could not find a clearly stated licence for the use of the data. However, in the terms of use, It was noted that the data provided in the website can be used as long as it is presented truthfully and without a claim of exclusive ownership of the data.

Process and Preparation: Cleaning and Manipulation of Data

The way the data was made available presented a challenge to extract. Although the information can be downloaded in csv format, there was a limit of searching for 5 districts at a time. The reason is when the data was selected, it was first presented on the URA’s website. The page that is presented for review is one of 4 pages of data. Depending on the district selected, there was a possibility of 20,000 rows of data per page. This made the page very hard to load and harder to download the desired csv file.

In order to resolve this issue, I had to download each district separately and combine all csv files later.

Once the data was combined into one file, I had to remove a few rows and columns. Each downloaded csv file had 5-10 lines of explanation as footnotes as well as few heading rows, and an index column to indicate row number which interfered with a clean analysis so they had to be removed. I also renamed the column names from their downloaded names to one written in snake case for ease of use during analysis.

For rows that indicated property sold in more than a single unit but as it was difficult to correctly estimate the cost of a single unit. I had to remove those rows as well as properties sold as just land as I wanted to keep my analysis to ready made homes.

The tenure data also presented data in a difficult way to perform analysis on as the lease duration and the date for the start of the lease are written in a long string. In order to simplify working with the data, I used excel and substring functions like LEFT() and RIGHT() to extract just the values I wanted and created two additional columns called lease and lease_start respectively. This allowed me to calculate the duration data I need to observe how long the property remained on the market.

This was important because a hot property market could suggest scarcity of supply. I also had to create a new column to duplicate the date of sale information as I needed a way to sort the date in a specific order or to fill by dates. The data when imported directly made it difficult to filter data to analyse and clean the data in excel.

Price per sqft of Recorded Sales in all 28 districts

By observing the scatter graph below, it is interesting to see that although prices of properties in the CCR regions and RCR do have a higher upper range above 6000 and 5500 Singaporean dollars respectively, both regions also recorded sales of properties that have a low range that match or cost lower than properties in the OCR regions.

When we plot the Price per sqft vs the Market Segments, the results were somewhat expected. The Core Central Region properties tended to have higher median prices, followed by the RCR and then finally the OCR. When taken together, CCR and RCR region prices tend to overlap slightly. This might be because certain townships within CCR and RCR regions are shared. This might lead to cost by association or simply proximity.

However, if you hover over the boxplot, it can be observed that both CCR and RCR has had transactions where prices were on the low end of the spectrum that occasionally overlaps with prices of the OCR region. Once again, it is the outliners that seem to power the CCR and RCR regions to higher highs.

The distribution of outliners is also an interesting thing to observe. Whereas RCR regions have transactions where their price per sqft had low priced outliners, CCR properties outliers are all well above the high median prices.

Another assumption I would have made is buyers in the higher priced CCR regions would purchasing properties that tended to be smaller due to the higher unit price per square feet. However, when looking at the area summary information, there are once again a lot of outlier data points.

When we take a closer look at the summary data, it is interesting to see that the median sizes of apartment and condominiums do not have such a high variance on average and to see that the larger properties were in the CCR and RCR regions whereas the smaller properties are in the OCR. This is the exact opposite to what I would have expected.

The perception is the property market has started to heat up since Singapore came out of COVID lockdowns and international borders started to open. This trend seems to be echoed all around the world. COVID caused issues in supply and also moved workforce further into the suburbs and away from the cities as the remote working trend took hold. These and many other contributing factors meant there was less supply of available homes for willing buyers.

However, judging by the scatter graph below, properties that were sold before the start of the first cases of COVID seem to have stayed on the market longer. There is a noticeable trend downwards. Perhaps, supply was not able to keep up with demand. Therefore, COVID might not have been the cause of the sudden rise of prices but it exacerbated it.

Property Prices over the last 5 years.

When examining the summary prices of properties sold in the last 5 years, the trend in all 3 regions seems to suggest that prices stayed relatively flat with only the RCR regions showing a little uptick near the end of our data set around June/July 2022.

However, by examining each region individually, the CCR region still maintains a healthy price range of between a little short of 2 million to around 7 million. These, however, are only the average highs and average lows. As observed in previous graphs we analysed, the CCR region has had a lot of property transactions closed outside the average range with prices rising beyond 15 million over the last 5 years.

If we take a closer look at the median price data between 1 to 5 million, it can be observed that there are instances like February 2020 where the median prices fell well to the 7 million highs and median prices fell to around a million. To put things into context, Singapore closed its borders in March 2020 so these price drops cannot really be attributed solely to COVID. Also, looking over the past 5 years, housing prices had closed a little over 500,000. This is well below the 2 million median prices which could suggest that in some places, there were properties available that were somewhat reasonable.

The data for the RCR seems to tell a slightly different flavour of the same story. The characteristics of the price data has a range but with a lot of outliers with prices once again extending well beyond the median prices. Some properties could possibly challenge those sold in the more expensive CCR region with prices going as high as 13 to 14 million. However these would be the exceptions and occur at a far less frequency seen in the CCR region.

The differences do not stop there. Whereas prices for CCR region properties tended to range between 2 and 7 million. RCR neighbourhoods were much more reasonable and cost between 1 to 2 million which is a much tighter price range. Outliers still exist and there are numerous transactions of property closing between 3 to 5 million but once again not to the same frequency as in the CCR.

When looking closer again, we can see that prices seem more to average from 0.5 to around 3 million but unlike CCR region properties, the median prices are ever so slightly creeping up with median prices showing around 1.4 and 1.5 million to just over 2 million recently in July 2022. The difference here is unlike CCR homes which might occasionally dip to 0.5 million, RCR homes seem to have prices from around there.

Perhaps the most surprising detail to come from looking at the summary median prices is OCR region homes sold for just under 1 million to around 1.4 million. This price range is reminiscent of the median price for RCR homes. It is very clear by looking at the data, the OCR region does lack the same abundance of outliers with prices going beyond 5 million only a handful of times.

However, when you go closer to the data from the OCR, there aren’t as many surprises there. Prices stayed relatively flat with median prices never really threatening the 1.5 million mark that was breached constantly by CCR and RCR region homes. Although, with some lows around 0.5 million and some highs around 2.5 million, it might be easy for a lot of buyers to think that perhaps all Singaporean properties cost at least a million dollars or more.

I think a summary here is: does the Singapore property market deserve the reputation as one of the most expensive in the world? Yes but as always the answer is a little more nuanced. In the truly high end market, in the city state where land is scarce. How much a property sells for is all about location and timing. Like any other market for any other commodity, being at the right place and the right time will yield you the best results. Looking at the data, it seems to suggest that Singapore isn’t as ruthlessly expensive as one might think. The sheer abundance of outliers in each of the market segments might give that illusion that everything in Singapore is astronomical and it is usually the truly outrageously unfathomable numbers that will make the headlines which add to the mystique.

We crave exclusivity and having the most expensive or most unique of anything is a human trait so is envy. However, as I said, this is a summary of the analysis not a conclusion as there is still so much more data to analyse. This story in data is only partially told. We still need to consider the HDB market where a majority of the population call home. In addition, aside from purchasing a property outright, it is more likely that a good portion of society are happy and/or only able to rent. Therefore a conclusion cannot be reached until that is completed.

For now, I would say no, thankfully even just looking at the high end scale of the property market in Singapore, it is not outside the reach of average residents and there is most definitely hope of home ownership is that is the intention. When you separate the hype, the data tells a slightly more positive view. Yes, there are super expensive homes but there are also some reasonable priced ones too. The ironic detail for me is that perhaps the best region to look for a home to buy in this end of the market might be in the RCR region and not the OCR as I first thought.

However, for the question of condo or HDB, that question still needs to be answered and I hope to do it in my next analysis when I look at HDB prices.