COE Analyzer Presentation

LYE Keng Fook
27 Sep 2015

COE analyzer is a data product to study & predict COE price trends. Certificate of Entitlement (COE Wikipedia) is Singapore's policy tool to manage motor vehicle population growth. One has to bid for a COE and register it together with a new car; used cars are sold together with the registered COE. Motor vehicles are categorized into 5 COE categories

COE Category Vehicle class
A CAR UP TO 1600CC & 97KW
B CAR ABOVE 1600CC OR 97KW
C GOODS VEHICLE & BUS
D MOTORCYCLE
E OPEN, transferable & can be used for any vehicle

Objective & Method

COE prices fluctuates according to market supply/demand, and is generally associated to a few causes:
1. no. of COE's available for bidding
2. no. of bids submitted
3. no. of deregistered vehicles (COEs of deregistered vehicles are recycled for future bidding)
4. economic performance (as cars are a discretionary spend)

Data Collection

COE price data were obtained & merged from below 2 sources:

Objective & Method (2)

Exploratory Analysis

Obtained data link were cleaned & transformed.
Cleaned data link retains COE price data from bi-weekly biddings from Apr 2012 till Sep 2015.
Note: Quota & bids received data for CAT B,C,D,E were simulated owing to lack of time.
Summary statistics of COE prices for the 5 vehicle categories presented below

     CAT.A           CAT.B           CAT.C           CAT.D     
 Min.   :    2   Min.   :  200   Min.   :    1   Min.   :   1  
 1st Qu.:15349   1st Qu.:16296   1st Qu.: 8495   1st Qu.: 687  
 Median :28331   Median :28022   Median :16409   Median :1102  
 Mean   :35088   Mean   :40022   Mean   :25479   Mean   :1564  
 3rd Qu.:57049   3rd Qu.:68302   3rd Qu.:45251   3rd Qu.:1856  
 Max.   :92100   Max.   :96210   Max.   :76310   Max.   :6801  
     CAT.E      
 Min.   : 3000  
 1st Qu.:17351  
 Median :28997  
 Mean   :40912  
 3rd Qu.:68356  
 Max.   :97889  

Objective & Method (3)

Statistical Modeling

Standard multivariate linear regression is used to model COE price movements. Example of model fit for CAT A COE


Call:
lm(formula = CAT.A ~ CAT.A.BIDS.RECV + CAT.A.QUOTA, data = coeDat)

Residuals:
   Min     1Q Median     3Q    Max 
-28577  -8409    565   8795  34109 

Coefficients:
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)     67562.173   1537.585  43.940  < 2e-16 ***
CAT.A.BIDS.RECV     5.082      1.751   2.902  0.00396 ** 
CAT.A.QUOTA       -30.223      2.329 -12.975  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 13090 on 320 degrees of freedom
Multiple R-squared:  0.6787,    Adjusted R-squared:  0.6767 
F-statistic:   338 on 2 and 320 DF,  p-value: < 2.2e-16

Conclusion

Reproducibility

All analysis's performed in this report are documented in GitHub (repository link)

To-do List

More work is required to complete this data product.
1. Complete data collection for CAT B,C,D,E.
2. Add variables to improve model fit (e.g. stock indices, economy forecast) 3. Improve plotting, current graph plots are horrendous!