Lulu Cheng
Nov 18,2015
This article is to analyze the success of CARS program. California has the largest paid transactions, voucher amount and voucher amount per capita. Utah has the best fuel efficiency improvement. West Coast cusotmers significantly purchase more fuel efficient vehicles than other regions. Cusomers tend to buy more Japanese cars and Passanger Cars. To conclude whether the program is “widely successful”, we need more demographic and economic data. We cannot exclude the effect of gas price increase without an A/B testing experiment.
(Data is here)
Metrics to measure success of the CARS program:
One measurement we can use is total paid transactions for each state, which is to tell the market size of the program. According to the visualization, we can see that California performs the best (76,000 transactions) while Northern Mariana Islands performs the worst (7 transactions).
|
|
|
The main purpose of the CARS program is to maximize fuel efficiency. According to summary-statistics.pdf, New vehicles Mileage is 24.9 MPG on average and Trade-in Mileage is 15.8 MPG, which means the overall increase is 58% improvement. Using this metric, we can measure how successful the fuel efficiency improvement is for each state. From the visualization below, we can see that Utah has the best improvement (66.23%) while Virgin Islands has the least improvement (46.01%).
|
|
|
How profitable the program could be is another metric we should care about. The voucher amount we get from each state is different. California has the most voucher amount (321.55 million) while Northern Mariana Islands has the least (30.5k).
|
|
|
To get rid of the population size effect for voucher amount, we can also analyze the voucher amount we can get per capita. Population data is from 2010 Population Estimates. We can clearly see that California has the highest voucher amount per capita (17.15) while DC has the lowest (0.11). (Exluding states:GU, MP, VI)
|
|
|
Did West Coast (including CA, OR, WA, AK, HI) consumers purchase more fuel efficient cars than in other regions? First let's have a data explortary analysis with box plot.
We can see from the boxplot that both trade-in and new vehicles have higher mileage in West Coast than other regions. To further prove, we do a Hypothesis Testing.
Null hypothesis: MPG between West Coast and other regions has no difference.
Let's take a look at the trade-in car first:
##
## Welch Two Sample t-test
##
## data: old_0 and old_1
## t = -29.966, df = 132950, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -0.2488391 -0.2182857
## sample estimates:
## mean of x mean of y
## 15.66443 15.89800
P-value is smaller than 2.2e-16, which is far smaller than 0.05. So the testing result suggests statistically significant difference between two datasets. Thus we reject the null hypothesis. The trade-in cars from West Coast customers (MPG 15.898 on average) are more fuel efficient than other regions (MPG 15.664 on average).
Then let's see the new car they bought:
##
## Welch Two Sample t-test
##
## data: new_0 and new_1
## t = -67.642, df = 122450, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -1.609765 -1.519104
## sample estimates:
## mean of x mean of y
## 24.64316 26.20759
P-value is smaller than 2.2e-16, which is far smaller than 0.05. So the testing result suggests statistically significant difference between two datasets. Thus we reject the null hypothesis. The newly-bought cars from West Coast customers (MPG 26.208 on average) are more fuel efficient than other regions (MPG 24.643 on average).
For customers' behavioral pattern analysis, let's focus on their vehicle purchasing options:
We can easily see from the pie chart below that most people who trade in Passenge Cars still buy Passenger Cars. And most people who have Category 1 and Category 2 Trucks also prefer Passenger cars. And lots of people change from Category 3 Trucks to Category 2 Trucks. This program made lots of people change their car types. Average MPG for these types are: Passenger Car > Category 1 Truck > Category 2 Truck > Category 3 Truck. Thus customers are indeed tending to buy more fuel-efficient cars. One of the reasons might be the higer gas price.
Fuel efficiency distribution
|
|
|
Most of the customers prefer to buy Japanese cars. This might due to the more fuel-effenciency and lower price.
Fuel efficiency distribution
|
|
|
Car price distribution
Besides these analysis, we can also analyze the trade-in car made year, car country based on different types, etc. Due to time limit, I only analyze the above two patterns.
Data is not provided sufficient yet. We need to know more about demographic distribution in different states, such as the total amount of people who own cars, car make distributions, etc. Also it's better to do an A/B testing as a comparison. We need to tell whtether people buying fuel-efficient cars is really due to the CARS program, or due to the increase of gas price. It is very possible that people prefer fuel-efficient cars because the gas price is higher and higher over these years. Also there're some other economic factors we need to consider. Without knowing these, we can hardly conclude that the CARS program is “wildly successful”.