Certificate of Entitlement (COE) - When to avoid paying too much for COE?

When you want to buy a car in Singapore, you need to purchase a certificate of entitlement that allows you to hold on to that car for 10 years. This is a fixed quota determined by the Government to control the car population. So, in a way, the supply is inelastic but demand is ever growing due to population growth. Going by the demand-supply theory, we have a price-setter situation because the price is set by the Government. Nevertheless, if we dwell deeper, there are other factors affecting the price of COE as well. So, how to avoid paying too much for COE? This is the question that this article attempts to answer.

Data Preparation
In order to understand the factors affecting COE premium, I have taken a set of data from www.data.gov.sg and this contains COE results from 2010 to 2017. I have also gotten data for CPI, GDP, STI (Straits Times Index) and Resident Population from Department of Statistics and other reliable sources. For resident population, I have extracted only those who are 20 years old and above (kind of easy to figure out that anybody lower cannot really own a car). It took me a couple of days to pack all these together. Anyway, let's move on to the more exciting part. I have used Python with Matplotlib for the charts and sklearn machine learning package for this analysis. And, I only focus on Category A COE for this example.

Data Analysis
The above matrix chart looks cool right. Forget it! For the layman, it will be tough to understand the results. Well, if nobody understands what you have produced, it will be simply crap. So, I will attempt to put it into something that everybody understands.

a) Current COE premium is low (S$36k last week) but is it low enough?

Let's focus on the COE premium first. It seems that we have a 'normal' distribution for the COE premium (hello guys, this is statistics 101 - normal distribution, so don't ask me what is this normal thingy). The average Category A COE seems to be around the region of $55k. The current COE of $36k seems to be lower than the 25 percentile. Is this the best time to buy a car now? I am not sure. Buy it at your own risk. Don't blame me if you lose money.

b) COE premium is affected by the quota, so are the bids received

Next, we focus on the impact of the quota on COE premium. It is so obvious that the COE quota will affect the outcome of COE premium. Even my mum knows about it. But, let's try to figure out when you have to fork out sky-high price for a COE. See chart below. When the quota is less than 500, you will get to see sky-high premium. But, when quota goes beyond 1000, it will move closer to the 'norm' of S$55k. Of course, if you are the lucky few, you will be able to get one below $20k.

But, wait, I am seeing something interesting. There seems to be a correlation between quota and bids received for COE bidding. For most car buyers, we know that the car dealers will place the bids for you. And, how do they know the number of bids to place? Isn't it interesting? Somehow, all the car dealers will collectively submit less bids when the quota is low and more bids when quota is high. I will not attempt to interpret this because it is not right and fair to do it when I am not in this line. But, if any of the readers out there knows the reason, please let me know.

c) Is COE affected by CPI, Population, GDP and Stock prices?

The conventional wisdom is that COE premium will go up when market feels good. For example, when share prices go up, market will feel good and consumers will buy cars or change cars. It seems logical.  But, which factor (CPI, STI, etc) is more important than the others? And, what to look out for to avoid paying too much for COE?

When I compare STI with COE Premium, the relationship is not that obvious. It seems that COE will be mostly below S$50k when SIT falls below 3000. But, when STI goes above 3000, COE premium can go either way. But, one thing for sure, most of the expensive COE premiums ($70k and above) happen when STI is above 3000. So, buy car when STI is below 3000 if you do not want to overpay for the COE. Ha..ha.. I am not sure because STI is above 3200 now but COE is at $36k. Anyway, the facts are showing. Generally, STI, CPI and GDP are related to the health of the economy. So, they will have some impact on COE premium in one way or another.

Using RandomForest, I have done some importance ranking of the factors in the data set. It is kind of difficult to explain to you what this forest is about. Just think of this as a decision tree type of predictive machine learning logic that helps you to predict results. As a by-product, it also has this feature to rank the importance of the various factors that help in the prediction. As a reader, this is what you are looking for. COE quota ranks the most important. Next, comes CPI. The rest are GDP and STI.

Factors	Importance
Quota	42%
CPI	40%
GDP	4.90%
STI	4.70%
Bid Received	4.10%
Month	1.60%
Population	0.17%

Summary

So, in summary, many factors other than COE quota affect the COE premium. Generally, avoid COE when COE quota is low and the economy (STI, CPI, GDP) is doing good. Since you have taken the patience to read this long article, I would like to share some of my thoughts. COE quota is cyclical and it is much easier to forecast than the other factors. There is a low quota period and a high quota period. The last high quota period happened in 2007-2009. Going by the 10 year cycle, we are right now in the high quota period. If not for Uber and Grab, the effect of high COE quota would have already been very apparent. Nevertheless, we should see lower COE premium going forward before we are hit with the lower quota period again post 2018. The harder bit of the predictive model is to forecast the STI.  CPI is controlled and so we should not see too much fluctation. As for GDP, it should not deviate too much from our natural growth rate. So, the key deciding factors will be COE quota and STI. Watch these closely.

Lastly, buying a car is a big commitment. My analysis could be wrong. So, be careful and do it at your own risk. Do not blame me if you lose money.