Team # apmcm2303763

Team Number: apmcm2303763

Problem Chosen: C

2023 APMCM summary sheet

This paper addresses multiple issues pertaining to the development and impact of
new energy vehicles in China.
In the first problem, patent numbers, subsidies, per capita disposable income, and
charging station quantities from 2010 to 2022 are collected and used as influencing
factors. The sales volume of new energy vehicles is the dependent variable.A multiple
regression equation is established, with regression coefficients indicating the importance
of each factor. The order of importance is determined as subsidies, patents, per capita
income, and charging station growth rate.
In the second problem, seven indicators representing the electric vehicle indus-
try’s development, including market penetration rate, number of charging stations, and
production volume, are identified. Time series analysis using ARIMA models is con-
ducted on market penetration rate and sales volume, identified as pivotal development
indicators.
The third problem involves data collection on new energy and traditional vehicles,
employing the least squares method for analysis.A population competition model pre-
dicts future trends in the ownership and sales of new energy and traditional vehicles,
validated against original data.
For the fourth problem, the export volume of new energy vehicles is selected as
the policy-influenced indicator. Historical restrictions from major market countries on
China’s new energy development are investigated using a time series approach. A DID
model is employed to ascertain the strong negative correlation between foreign policy
implementations and China’s new energy vehicle exports.
Lastly, the fifth problem entails establishing a polynomial model based on urban
car statistics to derive data for a city of one million inhabitants. This data is used to
construct a model assessing the impact of vehicle electrification on the environment,
considering factors like gasoline emissions’ CO2 coefficients.

Keywords: New Energy Vehicles Regression Analysis Time Series Policy Impact
Environmental Impact Urban Car Statistics
Team # apmcm2303763

Contents

1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 . 1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 . 2 Problem Restatement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. Assumptions and Justifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Main Factors : What Affect the Development of NEEVs? . . . . . . . . . . . . . . . 3
4 . 1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4 . 1 . 1 Data Collection and Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . 3
4 . 2 New Energy Electric Vehicle Development Model. . . . . . . . . . . . . . . . . . . . 4
4 . 2 . 1 Influencing Factors and Index Setting . . . . . . . . . . . . . . . . . . . . . . . . 4
4 . 2 . 2 Multiple Linear Regression Model . . . . . . . . . . . . . . . . . . . . . . . . . . 5
4 . 2 . 3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5. What Will the NEEVs Be In the Next 10 Years? . . . . . . . . . . . . . . . . . . . . . . . 6
5 . 1 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
5 . 1 . 1 Data Collection and Data Pre-processing . . . . . . . . . . . . . . . . . . . . 6
5 . 2 Forecast model for the development of NEEVs in China . . . . . . . . . . . . . . . 7
5 . 2 . 1 Influencing Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5 . 2 . 2 ARIMA time series forecasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5 . 2 . 3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6. An Impact on the Global Traditional Energy Vehicle Industry . . . . . . . . . . . 13
6 . 1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6 . 1 . 1 Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6 . 1 . 2 Data Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6 . 2 Impact Analysis Model: New Energy Electric Vehicles on Traditional
Vehicle Market. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
6 . 2 . 1 Optimal Species Competition Model in Math Modeling . . . . . . . . . . 13
6 . 2 . 2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
7. Resistance to China’s NEEVs and the Impact . . . . . . . . . . . . . . . . . . . . . . . . . 16
7 . 1 Index Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
7 . 2 Difference-in-Differences Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
7 . 3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Team # apmcm2303763

8. Making Difference:Changes the Electrification Brings to the City of One

Million People. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
8 . 1 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
8 . 2 Composite Model of Urban Population and Carbon Emissions per Hun-
dred Kilometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
8 . 2 . 1 Urban Population and Fitting Model for Total Public Buses and
Cars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
8 . 2 . 2 The Fitted Model for Carbon Emissions and Vehicle Numbers. . . . . 20
9. An Open Letter:New Energy Electric Vehicles Moisten Your Life and Mine 21
10. Sensitivity Analysis and Error Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
11. Model Evaluation and Further Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
11 . 1 Model Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
11 . 1 . 1 Advantages of the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
11 . 1 . 2 Shortcomings of the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
11 . 2 Promotion of the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
12. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
13. Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Team # apmcm2303763 Page 1 of 31

I. Introduction

1 . 1 Background
In recent years, countries around the world have been calling for low-carbon and
environmental protection. Only by developing clean energy and new energy industries
can we promote the vigorous development of new energy in the world. In the face
of global greenhouse warming and worsening environmental pollution, fuel emissions
from traditional vehicles have become the main cause of greenhouse warming. In such
a general environment, new energy vehicles,especially electric vehicles(EV), emerged
as the times require.
With the development of the science and technology, China has continuously
increased its requirements for vehicular exhaust emission. New energy electric vehi-
cles(NEEVs), as a new, eco-friendly, and efficient industry, conforms to the concept of
green energy and can effectively reduce environmental pollution.[1] So the development
and impact of new energy electric vehicles are the two topics people more and more
concerned on.
But the development of new energy electric vehicles is affected by factors from
various aspects.They may from the government policies,production technology, infras-
tructures, value for money compared to substitute goods,and so on.Identifying the factors
that affect the development of new energy electric vehicles, making reasonable judg-
ments on the importance of each factor, and predicting future development trends and
social impacts are urgent issues that need to be solved at the moment.

1 . 2 Problem Restatement
Taking into account the background information and restricted constraints identified
in the problem statement, we need to settle the following questions:
★ Problem 1
Figure out the main influencing factors on the development of new energy electric
vehicles, establish corresponding mathematical models, and explain how the factors
affect the development of new energy electric vehicles in China.
★ Problem 2
Choose appropriate influencing factors that can reflect the development of China’s
new energy electric vehicles, select the indicators which affect the chosen factors
Team # apmcm2303763 Page 2 of 31

and collect more detailed data of them, establish rational mathematical models,
and forecast the development of new energy electric vehicles in China in the next
ten years.
★ Problem 3
Based on the findings of problem 1 and 2, and extend Chinese model to the world,
collect required data and improve the model, analyze the impact of the new energy
electric vehicle industry on the traditional energy vehicle industry.
★ Problem 4
In order to limit the rapid development of new energy electric vehicles in China,
some countries have introduced policies to boycott Chinese new energy vehicles,
establish a model that reflects the impact of these policies on the development of
new energy electric vehicles in China.
★ Problem 5
Analyze the influences of the urban new energy electric vehicles on the ecological
environment, plug in a city of one million people and get the result.
★ Problem 6
According to the model and conclusion of the previous question, write an open letter
to propagandize the benefits of new energy electric vehicles and the contribution
of their industry to countries around the world.

II. Assumptions and Justifications

Assumption 1
Assuming that the data obtained from the search has a certain degree of credibility
and rationality.
Assumption 2
Assuming in the optimal species competition differential equation model, when
new energy vehicles exist independently, they follow the logistic pattern of population
change.
Assumption 3
Assuming that the new energy electric vehicles in the optimal species competition
differential equation model are treated as pure electric vehicles.
Assumption 4
Assuming that all new energy electric vehicles in problem 5 are purely electric.
Assumption 5
Team # apmcm2303763 Page 3 of 31

Assume that urban vehicles in problem 5 do not include trucks, and all fuel vehicles
use gasoline as energy consumption.

III. Notations

Symbol Description

𝑢𝑛 The policy subsidy amount for the year n

′
𝑢𝑛 Revised policy subsidy indicators

𝑧𝑛 The number of relevant patents for the year n

𝑘𝑛 The median personal disposable income for the year n

𝑐𝑛 The number of electric vehicle charging stations as of year n

′
𝑐𝑛 The growth rate of charging stations in year n

𝑦 Objective function

𝑛 years

𝜂 market penetration
′
𝑦 The potential number of the entire market

IV. Main Factors : What Affect the Development of

NEEVs?

4 . 1 Data Description

4 . 1 . 1 Data Collection and Pre-processing

In order to analyze the main factors that affect the development of new energy
electric vehicles in China, relevant data sets need to be collected. Owing to various
statistical departments associated with distinct elements, the Table 1 below illustrates
the origins of the data sets utilized in this chapter alongside the explanations of variables.
Team # apmcm2303763 Page 4 of 31

Table 1 Data source

Production and sales Ministry of Industry and Information Technology of China

Government subsidy Ministry of Industry and Information Technology of China

allocation

Number of related Autopat website

patent applications

Personal annual National Bureau of Statistics

disposable income
(median)

Number of charging China Electric Vehicle Charging Infrastructure Promotion

piles in stock Alliance

Since most of the data obtained in the preliminary work is not in a form that can
be directly used, these data should receive some preliminary processing.
In order to eliminate the influence of different dimensions of each set of data and
facilitate the construction of a multivariate regression model, data for each variable
should be normalized. Due to the relatively stable nature of the data without extreme
outliers in its maximum and minimum values, this paper utilizes min-max normalization
to process the data. It can be represented by the following formula:
′ 𝑥 − 𝑚𝑖𝑛(𝑥)
𝑥 = (1)
𝑚𝑎𝑥(𝑥) − 𝑚𝑖𝑛(𝑥)
′
𝑥 is a specific index of the variable, 𝑥 ∈ [0, 1] , 𝑚𝑎𝑥(𝑥) and 𝑚𝑖𝑛(𝑥) respectively
represent the maximum and minimum values of the sample data.The variables mentioned
below are all normalized by default.

4 . 2 New Energy Electric Vehicle Development Model

4 . 2 . 1 Influencing Factors and Index Setting

Unless otherwise specified, the following data defaults to years as the time series.
Comprehensive analysis of research reports from major institutions and existing
research in this field, Sales of new energy electric vehicles (𝑦) are considered a mainly
indicator of development. It will be used as the dependent variable in the multiple linear
regression analysis in this chapter.
Team # apmcm2303763 Page 5 of 31

Combined with practical considerations and according to relevant research [2, 3] ,

the main factors affecting the development of NEEVs include subsidy policy provided
by the state (𝑢 𝑛 ) , number of related patents successfully applied for during this period
(𝑧 𝑛 ) , citizens’ personal disposable income (𝑘 𝑛 ) , number of charging piles nationwide
(𝑐 𝑛 ) .
It should be noted that the personal disposable income here is the median, not the
average. For the median represents the middle value when all incomes are arranged in
ascending order, offering a more robust measure that is not as heavily influenced by the
extreme values.

4 . 2 . 2 Multiple Linear Regression Model

Establish a multiple linear regression model between each factor and the sales:

′ ′
𝑦 = 𝛽1 + 𝛽2 𝑧 𝑛 + 𝛽3 𝑢 𝑛 + 𝛽4 𝑘 𝑛 + 𝛽5 𝑐 𝑛 (2)

′
Among them, The growth rate of charging stations in year n 𝑐 𝑛 is represented by
the following formula:
′ 𝑐 𝑛 − 𝑐 𝑛−1
𝑐𝑛 = (3)
𝑐𝑛
′ 3
The revised policy subsidy indicators 𝑢 𝑛 equals to 𝑢 𝑛2

4 . 2 . 3 Results

Use the regress function in Matlab to establish a multiple linear regression model
and obtain the estimated value of the regression coefficient 𝛽,the regression model can
be expressed as:

′ ′
𝑦 = −0.010 + 0.491𝑧 𝑛 + 1.192𝑢 𝑛 + 0.319𝑘 𝑛 + 0.012𝑐 𝑛 (4)

Table 2 Model Summary

R-squared F-test p-value Standard Estimate Error

0.9994 423.6054 0.0364 0.0004

The model summary shows that the R-square of the model is 0.9994,the large F
value and the p value of 0.0364 (less than 0.05) indicate that the model is significant
Team # apmcm2303763 Page 6 of 31

overall and the regression coefficient is credible.The overall fitting effect is good. The
multiple linear regression fitting plot and residual plot are shown below:

Residual Plot

0.15

0.1

Multiple Linear Regression Fitting Plot

0.05
1.2
fitting data

Residual
original data

0
1

0.8
-0.05
Normalized Index

0.6 -0.1

0.4 -0.15

0.2 -0.2

0
2017 2017.5 2018 2018.5 2019 2019.5 2020 2020.5 2021 2021.5 2022
1 2 3 4 5 6
Year
Case number

Figure 1 Multiple linear regression fitting plot and residual plot

V. What Will the NEEVs Be In the Next 10 Years?

5 . 1 The Data

5 . 1 . 1 Data Collection and Data Pre-processing

In this chapter, the sales data of NEEVs collected in the previous question are
continued to be used. The market penetration data was gathered from China’s Ministry
of Industry and Information Technology and supplemented by information sourced from
the China Association of Automobile Manufacturers.
Since it is observed that the initial data has exponential growth properties (refer to
the left pictures in Figures 2 and 3), we perform logarithmic processing on the original
data. The data after logarithmic processing is more stable and can make the time series
more stable (refer to the pictures on the right of Figures 2 and 3). These data are now
prepared and ready to be used for model fitting.
Team # apmcm2303763 Page 7 of 31

Time Series Plot of Sales Log Time Series Plot of Sales

7000 9

6000 8

7
5000
Sales (thousand vehicles)

6
4000

Log Sales
5
3000
4

2000
3

1000 2

0 1

2012 2014 2016 2018 2020 2022 2012 2014 2016 2018 2020 2022
Year Year

Figure 2 Sales time series plot and the plot with Log Transformation

Market Penetration Over Time Logarithmic Transformation of Market Penetration Over Time

0.25
−2

0.20
−3

Log(Market Penetration)
Market Penetration

0.15 −4

−5
0.10
−6

0.05
−7

0.00 −8
2012 2014 2016 2018 2020 2022 2012 2014 2016 2018 2020 2022
Years Years

Figure 3 Market penetration rate time series plot and the plot with Log Transformation

5 . 2 Forecast model for the development of NEEVs in China

ARIMA model is chosen due to its capability in handling time series data exhibiting
temporal dependencies and trends. This model is particularly suitable for our dataset
as it can account for the temporal nature and potential autocorrelation present in the
variables, providing a robust framework for time series analysis and forecasting.
The following diagram (Figure 4) illustrates the step-by-step process used to solve
the problem discussed in this article.
Team # apmcm2303763 Page 8 of 31

Figure 4 Step flow chart

5 . 2 . 1 Influencing Factors

Regarding the selection of influencing factors, this paper first conducts a correlation
test on the influencing factors selected in problem 1 and some factors that are objectively
believed to have strong relevance to the development of new energy electric vehicles.

Figure 5 Correlation matrix between feature variables

As can be seen from the above figure, the correlation between feature variables is
very strong. This also shows that some feature variables provide similar information
to other variables. Therefore, based on the selection rate of influencing factors in
mainstream research reports on the new energy electric vehicle industry, this article
selects sales volume and market penetration as indicators to reflect the development of
new energy electric vehicles in China.
Team # apmcm2303763 Page 9 of 31

Here comes the conception of market penetration(𝜂). It is related to sales data

′
for the year(𝑦),and the potential sales data of the automobile market that year(𝑦 ). The
formula is as follows:
𝑦
𝜂= ′ × 100% (5)
𝑦

5 . 2 . 2 ARIMA time series forecasting

The ARIMA (Autoregressive Integrated Moving Average) model is a widely used

time series analysis technique for forecasting future values based on historical data
patterns. The mathematical formula of an ARIMA(p, d, q) model is represented as:

(1 − 𝜙1 𝐵 − 𝜙2 𝐵2 − · · · − 𝜙2 𝐵 𝑝 )(1 − 𝐵) 𝑑 𝑋𝑡 = (1 + 𝜃 1 𝐵 + 𝜃 2 𝐵2 + · · · + 𝜃 𝑞 𝐵 𝑞 )𝑍𝑡 (6)

where 𝑋𝑡 is the time series data, 𝐵 is the back shift operator, 𝜙𝑖 and 𝜃 𝑖 are the
autoregressive and moving average parameters, 𝑝 is the autoregressive order, 𝑑 is the
differencing degree, and 𝑞 is the moving average order.
In order to build an ARIMA model, it is necessary to select appropriate parameters
𝑝, 𝑞, 𝑑.
★ Step 1: Differential processing and ADF test
Differencing aims to transform a non-stationary time series into a stationary one for
better model fitting. Perform Augmented Dickey-Fuller (ADF) test after each differenc-
ing. If p-value is ¡ 0.05, the test is successful. If not, continue differencing until p-value
¡ 0.05. After first differencing, sales volume remains non-stationary with a p-value ¿
0.05.

Logarithmic Differences of Log_Market_Penetration Over Time

1.4

First Order Log Difference Plot of Sales 1.2

1.0
1.4
Logarithmic Differences

1.2 0.8
First Order Log Difference

1.0 0.6

0.8
0.4
0.6
0.2
0.4
0.0
0.2 2012 2014 2016 2018 2020 2022
Years
0.0
2012 2014 2016 2018 2020 2022
Year

(b) First difference of logarithm of market pen-

(a) First difference of logarithm of sales volume etration

Figure 6 First order difference of logarithmic time series

Team # apmcm2303763 Page 10 of 31

Then perform the second-order difference on the sales volume, and the p value is
less than 0.05.

Second Order Log Difference Plot of Sales

1.0

Second Order Log Difference

0.5

0.0

−0.5

−1.0

2014 2016 2018 2020 2022

Year

Figure 7 Second order difference of logarithmic time series of sales volume

The data is as follows:

Table 3 ADF statistics and p-value

sales volume market penetration

ADF statistics p-value ADF statistics p-value

first-order differencing -1.7733 0.3938 -4.4307 0.0002

second-order differencing -3.2121 0.0193

★ Step 2: Determination of 𝑝 and 𝑞

Due to the limited amount of data, accurately determining the orders 𝑝 and 𝑞 of
the ARIMA model from the autocorrelation function (ACF) and partial autocorrelation
function (PACF) can be quite challenging. Hence, the use of the Akaike Information
Criterion (AIC) and Bayesian Information Criterion (BIC) is adopted to ascertain the
values of 𝑝 and 𝑞.
AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) are
statistical measures used for model selection among different candidate models.
AIC balances a model’s goodness of fit with its complexity to find the best trade-
off between them. A lower AIC value indicates a better model. It’s calculated as
2𝑘 − 2 ln ( b
𝐿), where 𝑘 is the number of model parameters and b
𝐿 is the maximum
likelihood function value for the model.
Similarly, BIC also aids in model selection but penalizes complex models more.
It aims to choose simpler models by giving a heavier penalty for complexity than AIC.
Lower BIC values denote better models. The BIC formula is −2 ln ( b𝐿) + 𝑘 ln (𝑛), where
𝑛 is the sample size.
Team # apmcm2303763 Page 11 of 31

This paper explored various combinations of different 𝑝 and 𝑞 values, calculating

their corresponding AIC and BIC values. Ultimately, it pinpointed specific 𝑝 and 𝑞
combinations that resulted in the lowest AIC or BIC values.
For sales volume, two sets of 𝑝 and 𝑞 values meet the criteria: 𝑝 = 3, 𝑞 = 1 and
𝑝 = 0, 𝑞 = 1 are both valid. For market penetration, there are also two sets: 𝑝 = 3,
𝑞 = 0 and 𝑝 = 3, 𝑞 = 1, both meeting the conditions.
Now, the appropriate parameters 𝑝, 𝑑, and 𝑞 have been determined.

5 . 2 . 3 Results

The ARIMA(3,2,1) and ARIMA(0,2,1) models are employed to fit and fore-
cast sales data for the forthcoming decade. Simultaneously, the ARIMA(3,1,0) and
ARIMA(3,1,1) models are utilized for fitting and forecasting market penetration.

Table 4 Forecast of the development of NEEVs in China in the next 10 years

sales volume(000 vehicles) market penetration

year ARIMA(3,2,1) ARIMA(0,2,1) ARIMA(3,1,0) ARIMA(3,1,1)

2023 9034.13 10060.69 0.4148 0.3946

2024 10543.80 13249.39 0.4625 0.4219

2025 13200.22 16438.08 0.4862 0.4537

2026 16383.31 19626.78 0.5367 0.5223

2027 18609.43 22815.47 0.6170 0.5917

2028 20462.89 26004.17 0.6932 0.6293

2029 23134.30 29192.86 0.7415 0.6502

2030 26070.49 32381.56 0.7712 0.6821

2031 28315.21 35570.25 0.8058 0.7243

2032 30379.75 38758.95 0.8540 0.7585

Based on the projected data, the sales volume over the next decade is anticipated
to surge from 9 to 10 million to a range of 30.3 to 38.7 million units. This represents a
growth in sales data by a factor of 3.36 to 3.85 over the ten-year period.
Team # apmcm2303763 Page 12 of 31

ARIMA Model Forecast for Future Sales ARIMA Model Forecast for Future Sales
40000
Actual 30000 Actual
Forecast Forecast
35000
25000
Sales (thousand vehicles) 30000

Sales (thousand vehicles)

25000 20000

20000 15000
15000
10000
10000
5000
5000

0 0
2010 2015 2020 2025 2030 2010 2015 2020 2025 2030
Year Year

Figure 8 Sales forecast result of p=0,q=1(left) and p=3,q=1(right)

In the next ten years, new energy electric vehicles will usher in a blowout develop-
ment. Market penetration rises rapidly from low levels to extremely high growth rates.
In 2023, the market penetration is expected to be around 40 percent, which is less than
half. By 2032, the market penetration of NEEVs is projected to increase significantly
to between 75 percent and 85 percent.

ARIMA Forecast of Market Penetration with Increasing Boost ARIMA Forecast of Market Penetration with Increasing Boost
Historical Data Historical Data
0.8 Predicted Values Predicted Values
0.7

0.6
0.6
0.5
Market Penetration

Market Penetration

0.4
0.4
0.3

0.2
0.2

0.1

0.0 0.0
2010 2015 2020 2025 2030 2010 2015 2020 2025 2030
Years Years

Figure 9 Market penetration rate forecast result of p=3,q=0(left) and p=3,q=1(right)

In the coming decade, New Energy Electric Vehicles (NEEVs) are poised to dom-
inate the automotive industry, capturing nearly 80% of the domestic car market. They
will progressively replace conventional fuel vehicles until the market reaches saturation.
Team # apmcm2303763 Page 13 of 31

VI. An Impact on the Global Traditional Energy

Vehicle Industry

6 . 1 Data Description

6 . 1 . 1 Data collection

The sales data follows the statistics from problem 1 and 2. The vehicle owner-
ship data is sourced from the International Organization of Motor Vehicle Manufactur-
ers(IOCA) and China News Network.

6 . 1 . 2 Data Pre-processing

In the quest for reliable modeling, data preprocessing stands as a crucial stage.This
process aims to fortify the dataset for modeling purposes.
For global new energy electric vehicles and traditional energy vehicles’ sales and
ownership data, we chose logistic regression for the former and the least squares method
for the latter based on their suitability and goodness of fit calculations. Logistic re-
gression is ideal for categorical outcomes, while the least squares method suits linear
relationships, aligning with the nature of the datasets. Moreover, preprocessing the data
offers significant benefits, such as improving data quality, refining data fit for models,
and enhancing overall analysis robustness.

6 . 2 Impact Analysis Model: New Energy Electric Vehicles on Tra-

ditional Vehicle Market

6 . 2 . 1 Optimal Species Competition Model in Math Modeling

The optimal choice of the species competition differential equation model in in-
dustries hinges on its capacity to replicate intricate interactions among competitors,
mimicking the rivalry between new energy electric vehicles and traditional counterparts.
This model effectively tracks evolutionary paths and competitive behaviors crucial for
assessing emerging technology’s impact on established markets. Using vehicle sales
and ownership as key metrics is motivated by their ability to portray market penetration
and consumer adoption rates comprehensively. Sales figures mirror immediate market
acceptance, while ownership signifies long-term commitment and market sustainability,
Team # apmcm2303763 Page 14 of 31

offering a holistic view of the industry’s response to the rise of new energy electric
vehicles.
⋄ Optimal Species Competition Differential Equation Model for Ownership
The optimal species competition differential equation model represents the dynamic
interaction between competing entities in an ecosystem. Mathematically, it is often
formulated as:
( d𝑁
1 (𝑡) 𝑁1
d𝑡 = 𝑟 1 𝑁1 (1 − 𝐾1 ) − 𝜎1 𝑁1 𝑁2
d𝑁2 (𝑡) 𝑁2
(7)
d𝑡 = 𝑟 2 𝑁2 (1 − 𝐾2 ) − 𝜎2 𝑁1 𝑁2

𝑁 d(𝑁 (𝑡) − 𝑁 (𝑡 − 1))

𝑏 = 𝑟 (1 − )= (8)
𝑘 d𝑡
We use the subscript 1 to represent the data from the traditional automobile in-
dustry, and the subscript 2 to the new energy automobile industry. In this section, 𝑁
represents the processed data of global holdings; 𝑟 means the maximum growth rate of
the population; 𝐾 represents maximum population size the environment can sustain; 𝜎
means the intensity of competition between entities. 𝑏 is the average annual growth rate
of a certain population when it exists alone.
Based on investigations into industry research reports and official news coverage, we
have found that the development of traditional energy has essentially reached saturation
in 2022[4] . The growth in ownership is relatively stable, 𝐾2 is infinitely close to
the maximum saturation of the market. So the indicators of traditional energy do not
conform to the logistic pattern of population change. That leads to the ratio parameter
𝑁1
𝐾1 =0.
According to the definition of each indicator, we associate the existing data with the
parameters in the formula: 𝐾2 equals to the ownership of traditional automobile industry
in 2022, which is 11,446 million. And the 𝜎 here represents the cost performance index.
𝜎1 is the average electricity cost per 100 kilometers of new energy electric vehicles and
the average fuel cost per 100 kilometers of traditional energy vehicles, and 𝜎2 is the
reciprocal of 𝜎1 .
Using the data above, the parameters in the model are determined. Bring these
parameters into the model to solve and get the results.
⋄ Optimal Species Competition Differential Equation Model for Sales Volume
The model for sales is roughly the same as the above model, with 𝐾2 changed to
7493.56.
Team # apmcm2303763 Page 15 of 31

6 . 2 . 2 Results

Using this model, we obtained competition outcome data. The data of next ten
years are showed below as data display, it demonstrates the development of new energy
electric vehicles(NEEVs) and traditional energy vehicles(TEVs).

Table 5 Forecast for the development of NEEVs and TEVs in the next ten years

Ownership(000s) year Sales Volume(000s)

NEEVs TEVs NEEVs TEVs

170.064 144356.98 2023 140.142 6600.78

253.653 144237.74 2024 208.285 6504.63

378.289 144059.90 2025 308.972 6363.42

564.081 143794.91 2026 456.989 6157.70

840.929 143400.33 2027 672.851 5861.73

1253.226 142813.37 2028 983.650 5444.14

1866.708 141941.48 2029 1421.927 4872.79

2778.356 140649.12 2030 2019.211 4128.23

4130.402 138739.71 2031 2789.012 3229.47

6129.517 135932.42 2032 3697.395 2263.65

The competition results of the population model for the respective holdings and
sales are as follows:

10 4 Changes in the number of two groups increasing number of two groups

15 7000

6000
annual inventory number

5000
10
annual increment

X 2032
X 2039
4000 X 2031 Y 3697.4
Y 75079.2
Y 3229.47
X 2038
Y 56165 3000

5
2000

1000
new
new
traditional
traditional
0 0
2010 2015 2020 2025 2030 2035 2040 2045 2010 2015 2020 2025 2030 2035 2040 2045
year year

Figure 10 Competition results of ownership(left) and sales(right)

Team # apmcm2303763 Page 16 of 31

Based on the development trends of the two, the number of new energy electric
vehicles and traditional energy vehicles will reach the same level between 2038 and 2039.
And in 2041, new energy vehicles will account for more than 86% of the automobile
market.
Based on the development trends of the two, the sales of new energy electric vehicles
and traditional energy vehicles will reach the same level between 2031 and 2032. And
in 2034, new energy vehicles will account for more than 88% of the automobile market.

VII. Resistance to China’s NEEVs and the Impact

Because China’s share of the new energy electric vehicle market exceeds 50%
(data statistics are as of 2022), other countries have introduced policies to protect local
companies and inhibit the development of Chinese companies.

Figure 11 Global new energy electric vehicle market share in 2022

Apart from China, Europe and the United States wield the greatest influence on
the market. Hence, we focus solely on evaluating the effects of European and American
policies on our country’s new energy vehicles.

7 . 1 Index Setting
Based on a survey of past relevant research papers, China’s outbound new energy
vehicle exports serve as a direct gauge of foreign policy impact on its electric vehicle
industry. These exports reflect market changes due to tariffs, subsidies, and regulations,
showcasing how international policies influence demand and industry growth.
Hence, this section selects China’s NEEV export volume as the yardstick to assess
the influence of foreign policies on the growth of China’s NEEV industry. We collect
Team # apmcm2303763 Page 17 of 31

the main policies from October 2021 to October 2023.

★ Policy 1 (May 2021):
The U.S. Senate Finance Committee passed the ”American Clean Energy Act”
proposal, which allocates $31.6 billion in electric vehicle consumption tax credits.
At the same time, it relaxes the 200,000-unit limit for automakers to enjoy tax
credits, excluding Chinese manufacturers.
★ Policy 2&3 (February-March 2022):
The U.S. House of Representatives passed the American Competition Act, which
plans to allocate $52 billion over the next five years to subsidize research and
manufacturing in key industries such as semiconductors and automobiles.
★ Policy 4 (June 2022):
The passage of the tougher ”American Innovation and Competition Act” has more
powerfully protected the development of domestic companies in the United States.
★ Policy 5 (September 2022):
The U.S. Inflation Reduction Act has officially become legislation. This bill con-
tains strong unilateralism and protectionist overtones. Only companies that are on
the list can enjoy preferential treatment.
★ Policy 6 (September 2023):
European Commission President von der Leyen announced in her State of the Union
address that she would launch a countervailing investigation into China’s electric
vehicles.

7 . 2 Difference-in-Differences Model
A difference-in-difference (DID) model is a statistical method used to estimate the
causal effects of policy interventions by comparing changes in outcomes over time
between experimental and control groups. The model is based on the traditional
difference-in-difference (DID) approach, incorporating multiple time periods and/or
multiple groups.
The time series settings of policy nodes in our DID model are as follows: when
there is a policy, the sequence value takes 1, and when there is no policy, it takes 0. Use
the DID model to test the causal relationship between export volume and whether the
policy is enacted, and give the correlation (negative correlation)
Judging from the existing monthly export volume data (October 2021 to October
2023), China’s export volume of new energy electric vehicles is gradually rising. There-
Team # apmcm2303763 Page 18 of 31

fore, we set the control group data that is not affected by the policy to be the arithmetic
sequence distribution from the first group to the last group of existing data, and conduct
DID model testing.

7 . 3 Results
The monthly export volume-policy diagram used in the DID model is as follows:

Exports Over Time

Exports
120 Policy Promulgation 1
Policy Promulgation 2
Policy Promulgation 3
Policy Promulgation 4
100 Policy Promulgation 5
Policy Promulgation 6
Exports (thousands)

80

60

40

20
Ap 21
Ma -21
Jun 1
Jul 1
Au 21
Se 21
Oc -21
No -21
De -21
Jan 21
Fe 22
Ma 22
Ap 2
Ma -22
Jun-22
Jul 2
Au 22
Se 22
Oc 22
No -22
De -22
Jan 22
Fe 23
Ma 23
Ap 3
Ma -23
Jun-23
Jul 3
Au 23
Se 23
Oc 23
3
y-2
-2

r-2

-2

r-2

-2

t-2
r-

-
g-

c-
-
b-

-
g-
p-

c-
-
b-

-
g-
p-
r

r
v

v
y

y
p
t

t
Ma

Date

Figure 12 Export volume-policy diagram

And the results of the goodness of fit and policy influence of the DID model we
used are as follows:

Table 6 Model Summary

DID model Policies

R-squared Adj.R-squared Prob (F-statistic) Coef P>|𝑡|

0.942 0.938 2.72E-19 -15.3499 0.049

The provided DID model exhibits a high goodness of fit, with an R-squared of 0.942
and an adjusted R-squared of 0.938, indicating that approximately 94.2% of the variation
in the outcome variable is explained by the included variables. The Prob (F-statistic)
of 2.72E-19 is significantly low, suggesting that the model as a whole is statistically
significant, indicating that the included variables jointly have a significant effect on
the outcome. Additionally, the coefficient of policies is -15.3499, and its associated t-
statistic yields a p-value of 0.049, which is marginally below the conventional threshold
of 0.05. This indicates a potential significant impact of policies on the outcome variable.
Team # apmcm2303763 Page 19 of 31

VIII. Making Difference:Changes the Electrification

Brings to the City of One Million People

8 . 1 Data Description
According to authoritative industry reports in the automotive sector and publicly
available national information, we have obtained the average electrification rates for
buses and cars (𝑀𝑏 ,𝑀𝑐 ). Additionally, we’ve acquired the energy consumption equiva-
lents for electric buses and cars (𝑎 1 , 𝑎 2 ), where energy consumption equivalents denote
the conversion of electricity consumption per hundred kilometers to fuel consumption
per hundred kilometers. Furthermore, we have the fuel consumption per hundred kilo-
meters for non-electric buses and cars (𝑏 1 , 𝑏 2 ), as well as the carbon dioxide emission
coefficient for gasoline (𝜔).

8 . 2 Composite Model of Urban Population and Carbon Emissions

per Hundred Kilometers
To evaluate the impact of automobile electrification on the ecological environment
in a city of one million inhabitants, it requires fitting into two nested models. Given
the composition of vehicles in the city, for model simplification, we assume the city’s
automobiles do not include trucks. We use carbon emissions, the most influential
indicator, to assess the impact on the ecological environment. The first step involves
analyzing the relationship between urban population and the quantities of electric and
non-electric vehicles. The second step involves gathering data to reasonably fit vehicle
quantities and carbon emissions, ultimately determining the carbon emissions after
electrification for a city of one million inhabitants.

8 . 2 . 1 Urban Population and Fitting Model for Total Public Buses and Cars

Due to the strong fit between the total number of public buses and the total popu-
lation (as shown in the figure 13(a)), the estimated total number of buses(𝐿 𝑏 ) in a city
of one million inhabitants is 1161.
Team # apmcm2303763 Page 20 of 31

Scatter plot of car ownership per capita

0.5

0.45

0.4

Car ownership per capita

0.35
10 4 The relationship between the number of buses and the number of urban population
0.3
2

0.25
bus number

1.5

0.2
0.5 bus number vs. Population number
fitting
Prediction boundary- 90%

500 1000 1500 2000 2500 3000 0.15

Population number
Residual plot

0.1
5000

0
0.05
b

-5000

-10000 0
-15000
0 500 1000 1500 2000 2500 3000 3500
500 1000 1500
p
2000 2500 3000
Population

(a) The Graph Fitting Bus Quantity to Urban (b) Scatter Plot: Car Ownership per Capita vs.
Population Urban Population

Figure 13 The Urban Population’s Relationship with Total Public Buses and Cars

However, due to the scattered distribution between the total number of cars and
the total population (as seen in the scatter figure 13(b)), we perform fitting on the total
count. Then, subtracting the bus count from the total count gives us the total number of
cars. After fitting, the estimated total vehicles in a city of one million inhabitants are
1.0465 million, from which the number of cars(𝐿 𝑐 ) can be calculated.

8 . 2 . 2 The Fitted Model for Carbon Emissions and Vehicle Numbers

We choose the total carbon dioxide emission as the representative indicator for
carbon emissions(𝑇) .Establish a model for carbon dioxide concerning the number of
urban buses and cars using the following formula:

𝑇 = 𝜔 × [(𝑎 1 𝑀𝑏 + 𝑏 1 (1 − 𝑀𝑏 )) 𝐿 𝑏 + (𝑎 2 𝑀𝑐 + 𝑏 2 (1 − 𝑀𝑐 )) 𝐿 𝑐 ] (9)

Finally, we get an estimate of carbon dioxide emissions per 100 kilometers for a city
with a population of one million.The calculation result of the model is 21545123.4782328
kilograms.
The figure of fitted model for carbon emissions and vehicle numbers is as follow:
Team # apmcm2303763 Page 21 of 31

Figure 14 The Fitted Model for Carbon Emissions and Vehicle Numbers

IX. An Open Letter:New Energy Electric Vehicles

Moisten Your Life and Mine
Dear Citizens,
As our world navigates through the challenges of environmental sustainability,
there’s a beacon of hope shining brightly on our roads: new energy electric vehicles.
The electrification of transportation, from cars to buses, holds immense promise in
transforming our cities into cleaner, healthier, and more sustainable spaces. It’s time to
recognize how significant these vehicles contribute to protecting our ecological envi-
ronment.
New energy electric vehicles (NEEVs) are not just a mode of transportation, they
represent a pivotal shift toward a greener future. The benefits they offer are multifaceted
and impactful. By choosing NEEVs, we collectively contribute to reducing greenhouse
gas emissions, curbing air pollution, and mitigating the effects of climate change. We
provide the future generations with a quieter, cleaner and safer living environment.
Countries around the world have been embracing the electric vehicle industry in a
more positive and open manner. They are adopting proactive measures and policies to
drive innovation and foster a revolution in transportation. From the Norway to China,
the United States to Netherlands , Governments and industries around the world are
increasing investment in NEEV infrastructure and technology. It is precisely because of
the accumulated efforts of all parties that the overall improvement in our quality of life
has been achieved today.
Electric buses,especially worth to be mention, are silently transforming our cities.
They offer They provide a quieter, less bumpy and pollution-free mode of public trans-
Team # apmcm2303763 Page 22 of 31

portation. Just imagine waiting at a bus stop surrounded by clean air, with buses silently
passing by, leaving behind no emissions. It’s a vision that’s becoming a reality in many
cities, thanks to the widespread adoption of electric buses.
The benefits of electric vehicles extend beyond environmental advantages. They
also bring economic opportunities, fostering job creation and innovation in clean en-
ergy technologies. Moreover, as the NEEV industry grows, the cost of these vehicles
continues to decrease, making them increasingly accessible to everyone.
However, this transition requires collective action. Each of us plays a crucial role in
accelerating the adoption of electric vehicles. Whether it’s choosing an electric car for
your daily commute or supporting initiatives that promote NEEV infrastructure, every
action contributes to a more sustainable future.
Let’s take pride in our collective efforts toward a cleaner and healthier environment.
Let’s embrace the potential of new energy electric vehicles and support the electric
vehicle industry, not just for ourselves but for the well-being of our planet and future
generations.
Together, we can drive change, one electric vehicle at a time.

X. Sensitivity Analysis and Error Analysis

When conducting multiple linear regression model in Problem 1, in order to explore
the sensitivity of the model,a 85%-115% confidence interval for the most influencing
factor should be given.
Observing the value of the regression coefficient 𝛽 ,we can get the biggest one is
belonged to the policy subsidy amount of NEEVs. It confirmed that policy subsidy
amount is the most influecing factor. Therefore, we fix the values of other variables and
perturb the subsidy coefficient, Obtain a 85%-115% confidence interval for the policy
subsidy amount.
Team # apmcm2303763 Page 23 of 31

Sensitivity Analysis Diagram

1.4

1.2

Normalized Index 1

0.8

0.6

0.4

original number
5%
0.2 10%
15%
-5%
-10%
-15%
0
2017 2017.5 2018 2018.5 2019 2019.5 2020 2020.5 2021 2021.5 2022
Year

Figure 15 1

The model’s conclusion highlights a consistently narrow range of solution fluc-

tuation, indicating a low standard deviation. The accompanying figure illustrates the
model’s low sensitivity and strong robustness.

XI. Model Evaluation and Further Discussion

11 . 1 Model Evaluation

11 . 1 . 1 Advantages of the Model

★ In the first question, our strength lies in identifying the factors influencing the
development of the new energy industry through relevant research and assessing the
extent of their impact, presenting a more intuitive display of their influence.
★ Evolving and refining the population competition model to better align with
the developmental patterns of industry competition. This model has the capability to
forecast the future trends of both new energy vehicle and traditional vehicle industries
under their mutual influence, enhancing its scientific validity.
★ We utilized a double difference model to analyze the impact of policies on export
volume, achieving a good fit. It holds more persuasive strength compared to traditional
regression models.
★ Building upon the foundation of the environmental impact of automobiles in
various Chinese cities, the model constructed is more scientific and representative. It
Team # apmcm2303763 Page 24 of 31

distinguishes between the electrification rates of cars and buses, avoiding a simple
aggregate of the two. This approach fully considers the varying degrees of vehicle
electrification between cities, rendering the model’s outcomes more comprehensive.

11 . 1 . 2 Shortcomings of the Model

★ The collected sample size of data can be expanded further, providing room for
additional optimization and improvement of the model.
★ In the fifth problem, the discussion focuses solely on two vehicle types: cars and
buses, presenting a relatively simplified analysis of vehicle categories.

11 . 2 Promotion of the Model

★ We can collect more indicators and data, and build a more complete and more
complex mathematical model for more accurate evaluation of various indicators reflect-
ing the development level of NEEVs.
★ We can collect more indicators and data, make a promotion of the Difference-in-
Differences Model. Analyze the duration of various policy lags, obtain new policy node
time series with policy lags, and establish a more complete DID model, which may lead
to better fitting results.

XII. References
[1] Zhai Jiatian. Analysis of development status and prospect of new energy vehicles.
Automobile Applied Technology, 48(20):193–198, 2023.

[2] Zhang Qingxin, Guo Mengmeng, and Liu Renyi. The impact of rd investment on
the development capacity of new energyvehicle enterprises. Journal of North China
University of Science and Technology(Social Science Edition), 23(06):27–32+60,
2023.

[3] He Lin. Competitiveness and industrial development trends of new energy vehicles
in the post-subsidy era and policy recommendations. China Academic Journal
Electronic Publishing House, 2022(04):16–23, 2022.

[4] Hu Ansheng and Feng Xiayong. Development and changing trends of the world
automobile industry. ChinaMachinery Electric Industry, (9):33–35, 2004.