This summary provides the key details from the document in 3 sentences: The document discusses a research project analyzing the relationship between stock performance and air pollution in Shanghai from 2012-2014. It examines stocks from heavy industry and environmentally-friendly industries, comparing stock returns to air quality measures like PM2.5, PM10, CO2 and SO2. The research found statistically significant relationships between worsening air quality and lower stock returns, especially for heavy industry stocks, though some limitations like omitted variables were noted.

1. FINC 625 Team 7 Project
Stock Performance and Air Pollution (haze,PM2.5) in
Shanghai
By
Dingteng Huang
Liang Shuang
Nan Li
Puyi Fang
Yiwei Yan

2. Abstract
There is a lot of research shows that air quality has significant influence on a nation’s economic
growth. Our research paper examines the relationship between stock return and air quality. Our main
target stocks are those of Heavy Industry and environmental friendly industry in Shanghai. Our main
air quality parameters are PM 2.5, PM10, CO2, And SO2. The time period we choose is from 2012
to 2014. After conducting numerous regressions analysis of different stock returns. We conclude
that air quality has significant influence on stock returns statistically and economically. However, we
also find some flaws with our regression models, one of which is the omitted variables. There are
also some observations that suggest the relationship between air quality and stock returns is not
simply linear. However, that is beyond the scope of our research paper.
Introduction & Background
In recent years, the big economic growth of Asia has attracted the rest of world’s attention.
There are increasingly a lot of investment opportunities because of the rapid growth of
economy. Especially, China plays an essential role in Asian economical development. From 1979
until 2010, China's average annual GDP growth was 9.91%, reaching an historical high of 15.2% in
1984 and a record low of 3.8% in 1990. China's nominal GDP by Expenditure approach surpassed
that of Italy in 2000, France in 2002, the United Kingdom in 2006 and that of Germany in 2007,
before overtaking Japan in 2009, making China the world's second largest economy after the United

3. States. In the annual year of 2014, China’s GDP growth rate is 7.4%, which is lower than before but
still a very attractive figure. Looking at the first quarter of 2015, the GDP growth rate of china is
about 7%.
Undoubtedly, China has plenty of opportunities to equity investors. Shanghai stock exchange
(SSE) is becoming a new rising star in the world’s largest stock indexes. Besides, Hongkong stock
exchange and Tokyo Stock exchange still play the major role in Asia area, and in November 14,
2014, China lunched “stock connection” between Shanghai and Hongkong stock exchange, relaxed
the restrictions that historically split the Chinese stock market between shares targeted at local
investors and those available to international investors. Another big issue that comes to people’s
attention is the air quality of China in recent years. Beijing and Shanghai suffer very terrible air
pollution in recent years. One of the air quality figures “PM 2.5” has become a very popular stuff in
people’s real life. Other air quality figures are PM10, SO2, NO2, and so on. The air quality certainly
has influence on people’s life, work, and health. In history, there has been a lot of research analysis
of among stock index, oil price, foreign exchange price, government policy and so on. However,
few people took analysis of the relationship between stock return and air quality. Our group is very
curious about whether the air quality has a significant relationship with stock returns. Based on this
idea, we conducted the regression analysis of the Shanghai stock index and Shanghai air quality
during the period of 2012- 2014.
Hypothesis
● Air quality, especially PM 2.5 has a significant effect on various stock returns.
● The more terrible the air quality is, the higher the stock return in heavy industry.
Data Collection Process
The first big part of our analysis is to find the relevant data and integrate it to the single excel
file. This is a very time consuming process. We gathered varies Shanghai stock price data during the
period of 2012-2014 from Chinese official economic website. We then collected the air quality data
of the same period. After having these data available, we begin conduct our research process. Before

4. getting there, we want to introduce the dependent and independent variables. We try to give you a
big picture of the subjects we are working for. A sample data sheet is provided in Exhibit1.
Dependent Variables:
● AdjClose is adjusted daily price of ShangHai Stock Exchange (SSE)
● SSEDiscete-n is the discrete stock return of (SSE)
● SSELogReturn is the log return of SSE
● SPCReturn is Shanghai petrochemical company stock return
● SHEReturn is Shanghai Electric Power stock return
● SPEReturn is Shanghai DaTun Energy Company stock return
● MotorReturn is Shanghai Motor company stock return
● BaoReturn is shanghai Bao Steel Company stock return
● PortReturn is Portfolio Return
● PorPorsiti-n is a binary variable that equal 1 if portfolio return is positive, otherwise is
negative.
● SEEPostive-n is a binary variable that equal 1 if SEE stock return is positive, otherwise is
negative.
● LKReturn and SHRreturn are two newly environmental company stock returns
● PM25PorRe is portfolio return, which includes two environmental company’s stock
Independent Variables:
● PM25Average is the Average daily figure of PM 2.5
● PM25Max is the highest PM 2.5 reached each day
● PM10 is the Average daily figure of PM 10
● SO2 is the Average daily figure of SO2
● NO2 is the Average daily figure of NO2
Seasonality
Q2, Q3, Q4 represents the second quarter, third quarter, and fourth quarter respectively. They
are dummy variables, whose base group is Q1, the first quarter. For example, if the data we collected
is in the first quarter of a year, then Q2=Q3=Q4=0. If data is belong to the second quarter, Q2=1, and
Q3=Q4=0. The seasonality has a significant impact on both stock price and air quality. For instance,
the air quality is much more terrible in the winter (the fourth quarter) and there is a so called
“January Effect” on stock price.
● Unheal-25100 represents if PM 2.5 is greater than 100, it is unhealthy. It is a dummy
variable. For example, if PM2.5 is greater than 100, then Unheal-25100 is equal one,
otherwise it is equal zero.
● ReturnSpread is the difference between the portfolio return and SSE stock return.
Correlation and Co-linearity

5. Because our independent variables that are related to air pollution is highly correlated to PM2.5
index, co-linearity problem is the first task we have to solve in order to get better regression results.
Correlation matrix of those variables is showing in Exhibit2. To solve the problem, we ran a
variance inflation factor test on all variables and targeting the variable that have a VIF higher than
hour. We than regress those variables against our main variable (PM2.5Average) and use the
residuals to replace the original value of targeted variables. The new variables are defined as follows:
● ResPMmaxOn is residuals of the regression process between PM max and PM 2.5Average
● ResPM10Onis residuals of the regression process between PM10 and PM2.5 Average
● ResSO2OnPM is residuals of the regression process between SO2 and PM2.5 Average
● ResNO2onPM is residuals of the regression process between NO2 and PM2.5 Average
Research Logic
To begin our research on the relationship between daily stock return and air quality in
Shanghai, we first need to make every possible regression from all the observations we have.
Different type of regression and variables modifications were used to enhance the coverage of our
research. Well-tuned filtering mechanisms were developed to compare and test regressions. Finally,
based on the “winning” regression results left in our pool, we were able to maintain meaningful and
significant conclusions. For better demonstration, our logic and process waterfall are showing as
follow:

6. Regression Analysis and Findings
Our analysis and findings are primarily built upon the regressions that passed all filters and
are in our final selection pool. Regressions in the end selection pool are divided into three major
categories include “OLS uses same-day Ys and Xs”, “OLS uses lagged Ys”, and “Probit Binary
Regression uses Lagged binary Ys”. In the following report, each category is denoted as “Tier I”,
“Tier II”, and “Tier III” accordingly. All regression result from the final pool is listed in Exhibit(R).
Tier I Same Day OLS
● SSE Composite Index Return (000001.SS)
SSEDiscreteReturn = 0.0007745 – 0.00136*PM25Average + 0.002541*Q4 + u
(0.0007228) (0.00111) (0.0009748)
For regression test result, please see Exhibit3. The regression have an F-stat(2,722) of 2.94
which is significant at 95% confidence level. Daily average PM2.5 index and dummy variable for
quarter 1 of each year are jointly significant. A negative beta of PM2.5 index, that is significant at
80%, aligned with our hypothesis that, within the same day, a 1 mg/m3increases in PM2.5 will
negatively impact the return of SSE Composite Index daily return by 0.136%. This effect is
depressed for the fourth quarter of each year due to the seasonality of PM2.5 index. This regression
will only pass Ramsey RESET test at 75% confidence level, which indicates omitted variable biases.
However, we expect this problem to be within most of our regression because our independent
variables only focus on air pollution and seasonality while its impact on Index return is only partial.
Therefore, we believe this type of problem does not conflict with our research topic very much. The
same logic applies to later analysis.
● Shanghai Motor Company (600688.SS)

7. MotorReturn = 0.002379 – 0.006*PM25Average – 0.00516*PM10 + 0.0031055*Q4 +
(0.0023685) (-0.00304) (-0.00467) (0.0017958)
0.0032212*UnhealthyPM25100 + μ
(0.002299)
For regression test result, please see Exhibit4. Although the F-test of this regression is only
significant at 80% confidence level, the t-stat of the coefficient on PM2.5 daily average reached -
1.97, which is statistical significant at 95%. This relationship is fairly that we can clearly tell from
Exhibit5. We find this interesting because people usually relate PM2.5 and any sort of air pollution
to automotive exhaust. And this negative relationship between auto maker’s stock return support that
common view. The other possible explanation of the relationship could be that when air quality,
especially PM2.5 is too high, or the index exceeds consumer tend to avoid outdoor activity. Thus,
fewer cars are sold during the day which indicates less return. In addition, stated on the previous
article, Good Day Sunshine: Stock Return and Weather (2001) David Hirshlefer and Tyler Shumway,
psychological evidence and casual intuition predict that sunny good weather is associated with
upbeat mood. Thus we believe this negative relationship between Shanghai Motor Company stock
return and daily PM2.5 index is economical significant.
● Shanghai Bao Steel Company
BaoReturn = 0.0002857 – 0.00153*PM2.5Averag + 0.001205*Q3 + 0.0037935*Q4 + μ
(0.0011943) (0.00153) (0.00137) (0.00133)
For regression test result, please see Exhibit6. The regression has an F-stat (3,721) of 2.91
which is significant at 95% confidence level. All regressors are jointly significant. Steel industry has
always been a large source of pollutions historically. And the view is widely shared by the general
public, especially after the pollution in China became a serious topic. The fact that PM2.5 index has
an negative impact on the return of Bao Steel, at some level showing the market’s reaction towards
air pollution in Shanghai.
Tier II Lag 1 Day OLS

8. ● SSE Composite Index Return (000001.SS)
SSEDiscreteReturn = -0.0016134 + 0.00507*S02 + 0. 00197*N02
(0.0011369) (0.00409) (0.00201)
+ 0.0013873*Q3 + 0.001821*Q4 – 0.0022324*UnhealthyPM2.5+ μ
(0.001049) (0.001087) (0.0011373)
For regression test result please see Exhibit7. The regression have an F-stat(5,710) of 2.10 which
is significant at 94% confidence level. Daily average SO2 index, average N02, dummy variable for
quarter 3 and 4 of each year, and dummy variable for unhealthy PM 2.5 index of that day are jointly
significant. A negative beta of unhealthy PM 2.5 index, that is significant at 95%, aligned with our
hypothesis that, if the former day’s PM 2.5 index exceed the healthy level, for the following day the
return of SSE Composite Index daily return will decrease by 0.22324%. This effect is amplified for
the third and fourth quarter of each year.
● Shanghai Petrochemical Company
SPCReturn = -0.0021294 + 0. 00975*PM10 + 0. 0083*NO2
(0.0021294) (0.00578) (0.00449)
– 0.0045505*UnhealthyPM2.5 + μ
(0.0026083)
For regression test result, please see Exhibit8. The regression have an F-stat(3,720) of 2.14
which is significant at 91% confidence level. Daily average PM 10 average, average NO2 and
dummy variable for unhealthy PM 2.5 index of that day are jointly significant. A negative beta of
unhealthy PM 2.5 index, that is significant at 92% confidence level, aligned with out hypothesis. If
the previous day’s PM2.5 index exceeds the healthy level, for the following day, the return of
Shanghai Petrochemical Company will decrease by 0.45505%. A positive beta of PM 10 index and
daily average NO2, that is significant at 90% confidence level. For the following day, a 1 mg/m3
increases in PM 10 will positively impact the return of SPC daily return by 0. 975%, a 1 mg/m3
increases in NO2 will positively impact the return of SPC daily return by 0. 83%. We assume that

9. this result is due to the fact that, SPC is the largest petrochemical company located in Shanghai,
which engaged in production of ethylene, fiber, resin and plastics. One of main pollution sources of
NO2 and PM 10 in Shanghai are come from SPC. Therefore, the increase of the pollutant indicates
that SPC’s production or sale increased for that day. As a result the company’s share goes up.
● Shanghai Electric Power Company
SHEReturn = -0.0045377 – 0.0037*PM2.5Max + 0.00299*PM10 + 0.01808*SO2
(0.0021054) (0.00447) (0.00433) (0.00709)
+0.00239*NO2 + 0.0034606*Q3 – 0.0060768*UnhealthyPM2.5 + μ
(0.00363) (0.0018325) (0.002122)
For regression test result, please see Exhibit9. The regression has an F-stat (6,717) of 2.31 which
is significant at 95% confidence level. Daily average PM 2.5 Max, PM 10, SO2, NO2 and dummy
variable for quarter 3 and unhealthy PM 2.5 index of that day are jointly significant. A positive beta
of SO2, that is significant at 99% confidence level (T-stat = 2.55), also a negative beta of Unhealthy
PM 2.5 index (T-stat = -2.88), that is significant at 99% confidence level, aligned with out
hypothesis. Scatter Plot shows the strong relationship in Exhibit10. If the former day’s PM 2.5
indexes exceed the healthy level, for the following day the return of Shanghai Electric Power
Company will decrease by 0.60768%. A positive beta of daily average NO2, that is significant at
99% confidence level. For the following day, a 1 mg/m3
increases in NO2 will positively impact the
return of SHE daily return by 1.808%. We assume that this result is due to SHE is the largest electric
power company located in Shanghai. The coal burning plants are the biggest pollution source of
PM2.5 and NO2. The biggest coal consuming factory at Shanghai is the SHE Company. Therefore,
the increase of the pollutant NO2 indicates that SHE’s production or sale increased for that day. As a
result the company’s share goes up.
● Shanghai Bao steel Company

10. BaoReturn = -0.0021287 – 0.000642*PM2.5 + 0.00519*NO2 + 0.0016734*Q3
(0.0014452) (0.00234) (0.00283) (0.0013792)
+0.0031895*Q4 – 0.0020612*UnhealthyPM2.5 + μ
(0.0016541) (0.0016541)
For regression test, result please see Exhibit11. The regression has an F-stat (5,718) of 2.38 which is
significant at 95% confidence level. Daily average PM 2.5, NO2 and dummy variable for quarter
3&4 and unhealthy PM 2.5 index of that day are jointly significant. The coefficient result is similar
to the SPC since they are both industrial based. And the similar results from same day OLS further
proves that the return of BaoSteel is strongly related to air quality in Shanghai.
● Industry Portfolio
PortReturn = -0.0005641 + 0.00657*SO2 + 0.0012949*Q4 –
(0.001037) (0.00524) (0.0014055)
0.0025843*UnhealthyPM2.5 + μ
(0.0014632)
For regression test result please see Exhibit12. The regression has an F-stat (3,720) of 1.62 which is
significant at 80% confidence level.
● PM2.5 Environmental Friendly Portfolio
PM2.5PortReturn = 0.0031607 + 0.00462*PM2.5 – 0.01482*SO2 – 0.0022936*Q2 + μ
(0.0021151) (0.00395) (0.001094) (0.0024741)
For regression test result please see Exhibit13. The regression has an F-stat (3,720) of 0.75
which is only significant at 50% confidence level. Although the F-stat of this model is low, the result
is interesting. This is the only model that the coefficient of SO2 is negative and the intercept is a
positive number. We think that the portfolio is consisted of environmental friendly stocks, if the SO2
increased by a 1 mg/m3
, the following day return of portfolio will decrease by 1.482%. The
environmental friendly industry is the emerging market in China nowadays, at meanwhile the

11. industrial firms are under strike. The high growth rate of Chinese GDP is contributed by the
industrial firms which heavily depend on the coal. However, this high growth brings China critical
pollution problems. Therefore how to balance the pollution of heavy industry and sustainable
development is the biggest economic issue in China now. Therefore, Chinese government economic
policy are becoming more and more favorable to the environmental friendly firms, and the heavy
industry firms are restricted by production limitation and forced to spend more on pollutant disposal.
As a result of this, the intercept proves that the returns of these environmental friendly firms tend to
be positive all the time while the returns of the heavy industry firms are not.
Tier 3 Binary Dependent Variable Probit Regression
In this part, we use the Logit regression to analyze relationship between the binary dependent
variables and the probability of several positive target returns.
Shanghai Electric Power (SHE)
Pr( SHEPositiveReturn =1| X1, X2, … , Xk ) = G(-0.2997554 -0.0055828*PM25MAX +
(0.1707965) (0.48503)
0.00149137*SO2 – 0.2963043*Q3 – 0.3226641*UnhealthyPM25100 + μ)
(0.72272)(0.1821412) (0.2013937)
For regression test results please see Exhibit14. The regression has a chi (4) of 7.64 with 90%
confidence level. Daily maximum PM2.5 index, average SO2, dummy variable for quarter 3 of each
year, and dummy variable for unhealthy PM 2.5 index of that day are joint significant. A positive
beta of average SO2 index with 96% confidence level, on the next trading day, a 1mg/m3 increase in
SO2 will positively impact the probability of positive return of Shanghai Electric Power (SHE) by
G(1.49%). This effect decreased on quarter 3 of each year by 2.96% with 90% confidence level.
Also, a negative beta of unhealthy PM 2.5 index with 89% confidence level, for the following
trading day the SHE will decrease by 0.3226641%.We conclude that the reason behind that is the
same as we discussed above in Tier 2.

12. ● SSEComposite Index Return (000001.SS)
Pr(SSEDiscreteReturn =1|X1, X2, … , Xk) = G(-0.1380309 +0.005589*PM25MAX
(0.2257337) (0.47463)
+ 0.0052701*NO2 - 0.2691055*Q2 + 0.2678245*Q3 – 0.1472397*UnhealthyPM25100 + μ)
(0.37009)(0.19131) (0.20083) (0.18815)
For regression test, result please see Exhibit15. The regression has a chi2 ( 5) of 9.49 with91%
confidence level. Daily maximum PM2.5 index, average NO2, dummy variables for quarter 2 and 3
of each year and dummy variable for unhealthy PM2.5 index of that day are jointly significant. A
positive beta of daily average NO2 index with 85% confidence level, a 1 mg/m3 increase in NO2 will
positively impact the probability of positive daily return of SSE Composite Index by G (0.52701%)
on the next trading day. We conclude that the higher discharging of NO2 usually will increasing the
probability of positive return of SSE index, the reason behind that is heavy industries is one of the
most economic lifeline in China, besides NO2 is the worst polluter in China, which means high level
of NO2 discharging usually stems from the rapid development of heavy industries.
● Environmental Friendly Portfolio
Pr(PM25PortPositiveReturn =1| X1, X2, … , Xk) = G(0.0655521 +0.0097548*PM10+
(0.0884624) (0.4317)
0.2549341*Q3 +μ)
(0.1777918)
For regression test result, please see Exhibit16. The regression has a chi2 (2) of 5.70 with 91%
confidence level. Daily average PM 10 index and dummy variable for quarter 3 of each year are
joint significant. A positive beta of daily average PM 10 index with 98% confidence level, a 1
mg/m3 increase in daily average PM 10 will positively impact the possibility of positive daily
return of our environmental friendly portfolio G(0.97548%) on the next trading day. This
situation amplified on quarter 3 of each year. The reason behind that is higher PM 10 index

13. usually arouse the focus on developing environment friendly industries from the society, which
would push forward the development of environment friendly industries. For the scatter plot,
please see Exhibit17.
Pr(PortPositiveReturn =1| X1, X2, … , Xk) = G(0.0782299 + 0.0058157*PM25MAX
(0.111321) (0.45879)
+ 0.21751*Q3 – 0.1988713*UnhealthyPM25100 + μ)
(0.180038) (0.1684623)
For regression test result, please see Exhibit18. The regression has a chi2( 3) of 4.70 with 81%
confidence level. Daily maximum PM 2.5 index, dummy variables for quarter 3 of each year and
unhealthy PM 2.5 index of that day are joint. A positive beta of daily maximum PM 2.5 index
with 80% confidence level, a 1 mg/m3 increase in daily maximum PM 2.5 will positively impact
the possibility of the positive daily return of our heavy industries portfolio G(0.58157%) on the
next trading day. The reason behind that is the same as we discussed above in Tier 2.
We concluded that all pollution indexes (such as daily average SO2, average NO2, average PM
10, maximum PM 2.5) have a positive relation with all the probability of positive target returns
(such as Shanghai Electric Power, SSE Composite Index, Environmental Friendly and Heavy
Industries portfolio) respectively.
Conclusion
We also find that “3rd and 4th quarter effect” of each year has appeared in most of our
meaningful regressions. Therefore we can conclude that, maybe due to the specific climate in
Shanghai, air quality in fall and winter is much worse than the rest of a year.

14. Bibliography
1. The U.S. Department of State, Reporting of Daily Air Quality-Air Quality Index (AQI )(2012
to 2014), http://www.epa.gov/ttn/oarpg/t1/memoranda/rg701.pdf.,
2. David Hirshleifer and Tyler Shumway†, Good Day Sunshine: Stock Returns and the
Weather∗ ,2001
3. Mitra Akhtari, Reassessment of the Weather Effect: Stock Prices and Wall Street
Weather,2011
4.