YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF Free Download

Name: YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF
Author: rebeccawall1980

1 / 41

1 views•41 pages

YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF Free Download

YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF free Download. Think more deeply and widely.

YOU ARE MY SUNSHINE: EVIDENCE FROM

THAILAND STOCK MARKET

MR. PATTHARADANAI JATURAPORN

AN INDEPENDENT STUDY SUBMITTED IN PARTIAL

FULFILLMENT OF THE REQUIREMENTS FOR

THE DEGREE OF MASTER OF SCIENCE

PROGRAM IN FINANCE (INTERNATIONAL PROGRAM)

FACULTY OF COMMERCE AND ACCOUNTANCY

THAMMASAT UNIVERSITY

ACADEMIC YEAR 2019

Ref. code: 25626102042238OMU

YOU ARE MY SUNSHINE: EVIDENCE FROM

THAILAND STOCK MARKET

MR. PATTHARADANAI JATURAPORN

AN INDEPENDENT STUDY SUBMITTED IN PARTIAL

FULFILLMENT OF THE REQUIREMENTS FOR

THE DEGREE OF MASTER OF SCIENCE

PROGRAM IN FINANCE (INTERNATIONAL PROGRAM)

FACULTY OF COMMERCE AND ACCOUNTANCY

THAMMASAT UNIVERSITY

ACADEMIC YEAR 2019

Ref. code: 25626102042238OMU

(1)

Independent study title

YOU ARE MY SUNSHINE: EVIDENCE

FROM THAILAND STOCK MARKET

Author

Mr. Pattharadanai Jaturaporn

Degree

Master of Science (Finance)

Major field/Faculty/University

Master of Science Program in Finance

(International Program)

Faculty of Commerce and Accountancy

Thammasat University

Independent study advisor

Assistant Professor Wasin Siwasarit, Ph.D.

Academic year

2019

ABSTRACT

This paper is aimed to study the role of weather information which are wind

speed, temperature, and precipitation in forecasting the SET and SET50 return over the

period 2002 to 2008. The model AR(p)-GARCH(p,q) is used in this study in order to

compare the forecast ability between common and the weather models. Based on the

Wald test, the weather information does not depict any forecasting ability on the stock

return in Thailand.

Keywords: SET index, SET50 index, Weather effects, GARCH

Ref. code: 25626102042238OMU

(2)

ACKNOWLEDGEMENTS

The rough road of this degree would not have been possible without the

inspiration, support, and assistance of many people. I sincerely thankful to my family

and relatives for standing by me throughout all the terrible period, to my supervisor;

Assistant Professor Wasin Siwasarit, for your kindness and heartwarming help in every

single time we’ve met, to my friends for your support and warm love. I will not forget

all the thing I got and I hope everyone will have a glory on your own way.

Mr. Pattharadanai Jaturaporn

Ref. code: 25626102042238OMU

(3)

TABLE OF CONTENTS

Page

ABSTRACT

(1)

ACKNOWLEDGEMENTS

(2)

LIST OF TABLES

(5)

LIST OF FIGURES

(6)

CHAPTER 1 INTRODUCTION

CHAPTER 2 REVIEW OF LITERATURE

2.1 Weather and Mood

2.2 Weather and Stock Return

CHAPTER 3 RESEARCH METHODOLOGY

3.1 First Step: The Dummy Variables

3.1.1 Monday Effect Dummy

3.1.2 Weather Dummy

3.2 Second Step: Estimate GARCH Models

3.2.1 Common Model

3.2.2 Weather Model

3.3 Third Step: Evaluation the Models

3.3.1 Base Model Estimation

3.3.2 Data Plug-In

3.3.2.1 Collect the Data

3.3.2.2 Add-Drop the Data

Ref. code: 25626102042238OMU

(4)

Page

3.3.3 Model Estimation and RMSE Calculation

3.3.4 Rolling Windows

3.4 Fourth Step: Two-Sample Comparison

CHAPTER 4 EMPERICAL RESULTS

4.1 Common Model Results

4.2 Weather Model Results

4.3 Returns Forecasting Results

CHAPTER 5 CONCLUSION

REFERENCES

BIOGRAPHY

Ref. code: 25626102042238OMU

(5)

LIST OF TABLES

Tables

Page

4.1

Descriptive Statistics

4.2

Dummy Variable

4.3

The Estimated Results of Common Model

4.4

The Effects of Weather on SET Index Return

4.5

The Effects of Weather on SET Index Volatility

4.6

The Effects of Weather on SET50 Index Return

4.7

The Effects of Weather on SET50 Index Volatility

4.8

RMSE of SET Index Return from 1-period Ahead Forecasting

4.9

RMSE of SET50 Index Return from 1-period Ahead Forecasting

Ref. code: 25626102042238OMU

(6)

LIST OF FIGURES

Figures

Page

3.1

Rolling Window Pattern

4.1

SET Index and SET50 Index Returns

4.2

SET Return and SET50 Return Distributions

4.3

Weather Variable Graph

4.4

Weather Variable Distributions

Ref. code: 25626102042238OMU

CHAPTER 1

INTRODUCTION

Weather is the natural phenomenon that is hard to predict. In the summer

of Thailand, the temperature can be 25 degree Celsius or 35 degree Celsius. Changes

in temperature and the same level of precipitation can affect business activities such as

increasing regional agricultural productivity and economic activities. Kamstra (2003)

found that weather condition can affect investor’s mood. She examined the effect of

Seasonal Affective Disorder (SAD) on stock market which is caused by the weather

variation. Zuckerman (1994) investigated that depressive mood is associated with a

higher level of aversion risk. The seasonal variation can change the investment

preference according to the emotion variation state which means that stock market

returns are directly affected by behavioral changes. The relationship between stock

market returns and climate factors was described by Saunders (1993), Cao and Wei

(2005), and Floros (2011). The results of their research showed that the weather

variation has a significant negative correlation with stock market returns. However,

Hirshleifer and Shumway (2003) argued that if you consider the correlation only

between sunshine hours and market returns, there would be a positive correlation

among them. In case of Thailand market, the result of Khanthavit (2017) study

confirmed that the positive correlation of SET50, mai index returns, and Bangkok’s

weather is significant.

The results from the above studies confirm that weather conditions can

significantly affect the returns of the stock market. Therefore, the main objective of this

study is to investigate investment strategies and the weather information to find out

whether the weather conditions can relatively affect investors’ emotion and can also be

relatively correlated to the stock returns. The application of the findings should be able

to help improve the forecasting of stock return and its volatility. Consequently, the

weather conditions will be explored to see whether they also affect the investment

decision. Finally, the prediction performances of common model of returns are to be

compared with the prediction performances of the weather model during the similar

time between January, 2002 and December, 2018.

Ref. code: 25626102042238OMU

CHAPTER 2

REVIEW OF LITERATURE

2.1 Weather and Mood

It is believed by many people that the mood can be affected by the weather.

In the summer time, many people always dance and sing under the sunshine or jump

on on drenched streets. Previously, studies have been conducted on the relationship

between mood and weather. Denissen(2008) revealed that temperature, wind power,

and sunlight have a negative impact on our mood. He found that Vitamin D3 created

by the sunlight on human skin can change serotonin level in the brain. Keller’s study

result (2005) is also consistent with Dennissen’s (2008) that wind power can negatively

affect people's moods in spring and summer as people spend more time outdoors. Also,

Hoffman’s study (1984) describes the effects of climate on human mood on how

humidity, temperature, and daylight hours have the greatest influence on mood. He

projected that the higher level of humidity the lower the concentration score and the

increases of sleepiness being reported. Increased temperature can lower the anxiety and

scepticism mood score. This findings agree with Dexter (1904) on higher humidity

levels and lower barometric pressure reduced children's concentration scores in

classroom situations.

2.2 Weather and Stock Return

It is obviously known that serotonin created by the sunlight on our skin can

be directly affected to our mood by altering the brain serotonin levels. In Chinese

market, Feihui (2018) investigated that lower humidity and temperature have a positive

impact on stock returns. Accordingly, Markku Kaustia’s study result (2015) reveals the

temperature has a positive impact on trade volumes in the Finnish stock market. The

stock returns were also explained by the VAR model by Khanthavit (2017). He selected

seventh PC mood indices which included temperature, cloud cover, rainfall, humidity,

win speed, air pressure, and ground visibility to run with daily bond returns. The result

Ref. code: 25626102042238OMU

showed that humidity and wind speed can affect human mood. The findings are similar

to those of Hirshleifer and Shumway (2003) studying the effects of climate and the

return on 26 international stock exchanges. In addition to those, Saunders’ result (1998)

supports that the return would often be more positive on the sunny day than the cloudy

day. Hirshleifer’s and Shumway’ study (2003) added up on the model of Saunders

(1998) by including snow variable, which is not present in Thailand, but the result

didn’t have much change.

The return and volatility models described by climate variables are diverse.

Previously, the studies were using the simple regression, GARCH model and VAR

model. The model of return from Khanthavit paper in Thailand obtained a VAR format

with an external variable PC mood index (Vlady & Tufan, 2011), or VAR-X. The model

describing returns and volatility using a yield delay with weather indices. Lee, Jiang

and Indo (2002) mentioned that those models are incomplete because they are not based

on rigorous theories. Nevertheless, Khanthavit (2017) argued that his model was

sufficient to explain returns and volatility due to the external addition from the

decomposition theorem projecting that the exogenous variables do not change the well

diversified properties from the well diversified model. With reference to other

researchers like Juline (2016) who applied the OLS model and GARCH model to find

out a better one. She suggests that if you want to focus on the relationship between

weather effects and stock return, OLS can be applied. But in case of heteroskedasticity,

the GARCH model should be better in terms of forecasting volatility. In her model, she

chose the simple one, OLS model because it was better to describe the weather effect

on the return than the GARCH model. This disagrees with the model of Yoon and Kang

(2010). Their study concerns the effects of weather on Shanghai's returns and volatility

with the MA model. This is because the weather variable always had a high seasonality;

they recommended using de-seasonal technique to apply to each weather variable. Due

to rising global temperatures and the spread of global warming, the weather variable

should not be directly used. Yoon and Kang (2010) investigated the technique to resolve

this problem by using dummy variable of each weather variable. The process of this

technique included two steps: de-seasonal step and dummy variable step.

Firstly, de-seasonal step, Yoon and Kang (2010) subtracted the season

using the sum of the 15 forward and previous days. They then divided the result with

Ref. code: 25626102042238OMU

21; then the product of this step would be the average value of the weather variable

(Wt). The process can be described by Equation 1.

E(Wt) = 

(Wt-10 + Wt-9 + …+ Wt +…+Wt+9+ Wt+10)

(1)

Second, the dummy variable step suggests constructing a dummy variable

using mean + standard deviation (SD) as a benchmark for extreme climate variables.

For example, if the temperature is moving in the range of mean – one S.D. and mean +

one S.D., nothing might be happening. Here's how to describe a normal situation. In the

case of an extreme temperature, the temperature move is out of the range of one S.D.

(both in up or down boundaries), the dummy variable will be activated. If the weather

is moving up from the mean more than one S.D., the dummy variable for heat

temperature will be valued equal to one. If the weather is moving down from the mean

more than one S.D., The dummy variable for cold temperature will be valued equal to

one. The sample equation of dummy generating shows in below equations.

󰇟󰇛󰇜󰇛󰇜󰇠

(2)

󰇟󰇛󰇜󰇛󰇜󰇠

(3)

This technique was not the same in the models by Khanthavit (2018),

Hirshleifer and Shumway (2003) because they were using direct way to run their model

but the key is the seasonality. The weather variable has high seasonality (Yoon and

Kang, 2009) so, directly data will not be appropriated to apply with the weather model.

Ref. code: 25626102042238OMU

CHAPTER 3

RESEARCH METHODOLOGY

This study applies the GARCH model with weather data to forecast the

returns of the SET Index and SET50 Index in two separate model editions. The first

model is the general model, SET and SET50 index returns. These will be described by

lagged term of volatility and its return with the Monday effect dummy variable. As for

the second model, SET and SET50 will be described by the same components as in the

first model with the addition of a weather dummy variable. This chapter is presenting

4- step analyses as follows:

3.1 First Step: The Dummy Variables (Heat, Wet, Windspds)

There are two types of dummy variables in the model of Monday effect

dummy and weather dummy as the following.

3.1.1 Monday Effect Dummy

The value of Monday effect dummy is equal to 1 on Monday and 0

on the other day. The Monday dummy can reflect an unusual situation if there is a

Monday effect. This will absorb some of the unusual returns as mentioned in the

literature as the Monday effect.

3.1.2 Weather Dummy

Weather dummy was created following the guidelines of Yoon and

Kang (2009). Weather dummy variables are classified as windspeeds, temperature, and

precipitation. According to the work of Yoon and Kang (2009), the raw data of weather

variable should not be able to run the model because of seasonality effect. De-seasonal

effect technique of Yoon and Kang (2009) is firstly required to solve the problem.

E(Wt) = 

(Wt-7 + Wt-6 + …+ Wt +…+Wt+6+ Wt+7)

(4)

Where



is Weather variable value on day t

Ref. code: 25626102042238OMU

As in (1), the mean calculation will follow equation 1, the expectation

value is calculated from previous 7-day values and forward 10-day values. After

Finishing the calculation, the mean value is then used to estimate the standard

deviation.Then, the standard deviation can be calculated from eq. (2).

󰇛󰇜

󰇛󰇛󰇜󰇜





(5)

With the mean and standard deviation values, the dummy variables for each

weather variable are calculated; the precipitation is used to calculate the wet dummy

variable; temperature is used to calculate the heat dummy and wind speed is used to

calculate the Highwnds dummy variable.

󰇟󰇛󰇜󰇛󰇜󰇠

(6)

󰇟󰇛󰇜󰇛󰇜󰇠

(7)

󰇟󰇛󰇜󰇛󰇜󰇠

(8)

Where



is Precipitation (millimeters) on day t



is Temperature (°C) on day t



is Windspeed (knots per hour) on day t

The value of wet dummy variable depends on the value of the daily

precipitation. If the daily precipitation is greater than 󰇛󰇜󰇛󰇜, wet dummy value

will be 1, which is the precipitation on that day is more than the normal boundary (one

standard deviation), then it will equal to zero in the other cases. Heat dummy is defined

the same way as wet dummy. The boundary of the normal temperature is 󰇛󰇜

󰇛󰇜. If temperature on that day is greater than the thermal region, the heat dummy

Ref. code: 25626102042238OMU

value will be equal to 1 and 0 in the other cases. For the windspds dummy variable, the

creation will be the same approach as in wet and heat dummy variable.

3.2 Second Step: Estimate GARCH Models

GARCH model is the generalized autoregressive conditional

heteroscedasticity model. This model obtains the well-known characters of volatility

which is the volatility clustering, fat tails data distribution. This model was introduced

by Bollerslev in 1986.

3.2.1 Common Model

There are two indices that are SET and SET50 found in the stock

common return model. The returns and volatility of the present period are explained by

lag terms of itself and the Monday effect variable. The form of common model can be

presented as follows.

SET

SET50







 



(9)

󰇛󰇛󰇜󰇜

(10)

󰇛󰇜=  + 



 󰇛

󰇜+







(11)







 



(12)

󰇛󰇜

(13)

󰇛󰇜







 󰇛

󰇜









(14)

Ref. code: 25626102042238OMU

Where



is Constant term of return equation

is Constant term of volatility equation

is the conditional variance of rt

is the logarithm of returns

are the coefficient of lagged return

β1

is the coefficient of Monday effect dummy variable

β2

is the coefficient of lagged of conditional variance in mean equation

is the coefficient of lagged term of conditional variance in

variance equation

is the coefficient of lagged term of residual in variance equation

MON

is the Monday effect dummy variable; MON = 1 if the day is

Monday and 0 for otherwise

3.2.2 Weather Model

Concerning the weather model, the returns and volatility are

described by Monday effect and weather variables while the returns and volatility of

present period are explained by lag term of itself and weather variables.

SET







 





(15)

󰇛󰇛󰇜󰇜

(16)

󰇛󰇜=  + 



 󰇛

󰇜+









(17)

Ref. code: 25626102042238OMU

SET50

Where



is Dummy variable for high temperature day; Heat = 1 if 

󰇟󰇛󰇜󰇛󰇜󰇠 and 0 for otherwise.

Wet

is Dummy variable for high precipitation day; Wet = 1 if 

󰇟󰇛󰇜󰇛󰇜󰇠 and 0 for otherwise.

WindSpds

is Dummy variable for high wind speed day; Windspds = 1 if

󰇟󰇛󰇜󰇛󰇜󰇠 and 0 for otherwise.

MON

is the Monday effect dummy variable; MON = 1 if the day is

Monday, and 0 for otherwise

3.3 Third Step: Evaluation the Models

3.3.1 Base Model Estimation

The first step of the model evaluation is basic model estimation which

starts by applying the GARCH model. The data for the base model start from Jan 2002

to Dec 2005.

3.3.2 Data Plug-in

3.3.2.1 Collect the Data

The data of weather variables are obtained in the morning (4.00

a.m. to 10.00 a.m.) prior to the stock market pre-opened period (9.55 a.m. to 10.00 a.m.)

and will be applied for one-day prediction.







 





(18)

󰇛󰇜

(19)

󰇛󰇜







 󰇛

󰇜











(20)

Ref. code: 25626102042238OMU

3.3.2.2 Add-Drop the Data

After data collection, the data obtained in the previous step will

be plugged into the model and the first date of January 2002 will be discarded from the

model. The new data will run from January 2, 2002 to January 1, 2006. This data will

then be used for further steps.

3.3.3 Model Estimation and RMSE Calculation

After the data has been changed to the GARCH model, an estimation

method will be applied to estimate the model.

Comparison between the common and weather models, the root mean

error table (R MSE) of the forecast is computed and RMSE is defined as follows

󰇛







 󰇜1/2

Where

is the total number of forecasting



is the true proxy of volatility





is the estimated value of volatility which is produced by model k at

time t

Start (Collecting period) End SET Pre-opened

4.00 a.m. 10.00 a.m. 9.55 to 10.00 a.m.

(Start) (End)

Dropped Added

1 Jan 2002 … 31 Dec 2005 1 Jan 2006

Ref. code: 25626102042238OMU

3.3.4 Rolling Windows

For 1 Jan 2006 forecast, the observation of estimation period starts

from January 2002 to 20 December 2005.This is the base model; but for the next day

(1 Jan 2006), the observation must be re-dated by adding 1 latest day (1 Jan 2006) and

dropping 1 first day (1 January 2002). After finishing re-dated process, the GARCH

model estimation will be applied again. This process will be administered for the year 2006,

2007, 2009, 2010, and 2018. The example can be depicted in the following illustration:

Figure 3.1 Rolling Window Pattern

3.4 Fourth Step: Two-Sample Comparison

After calculating the results of root mean square error, the two-sample

comparison method is required for the comparison of the two sample sets. The first

group is the root mean square errors from common models and the second is the root

mean square errors from weather models. The two-sample t-test hypothesis is illustrated

as follows.

H0 : 









Ha : 









For 1 Jan 2006

Start End

2 Jan 2002 … 1 Jan 2006

For 2 Jan 2018

Start End

3 Jan 2002 … 2 Jan 2006

For 3 Jan 2018

Start End

4 Jan 2002 … 3 Jan 2006

Ref. code: 25626102042238OMU

T = 



















Where







is Expected value of root mean square error in model i



is Variance of root mean square error in model i

is Statistic T-value

is Number of samples in group i

Summary Research Process

Collect

Weather data

SET50 index

SET index

Data cleaning

Calculate log return Replicate

Monday/weather

dummy variable

Estimate GARCH

2.1. Common portfolio

2.2.Weather portfolio

Rolling Windows Return

Forecasting

Two sample

comparation

Result

discussion

Start

End

Ref. code: 25626102042238OMU

CHAPTER 4

EMPIRICAL RESULTS

This chapter illustrates the results of the estimation based on the method as

shown in chapter 3 and this chapter also provides more details of each variable on its

significant level. The unit-root tests and descriptive statistic is depicted in in the first

part of table 4.1. The second part includes the value and distribution of each variable

which will be described from figure 4.1 to 4.4 as well as table 4.2. The last part, table

4.3 to 4.7, will represent the results of common and weather GARCH model while table

4.8 to 4.9 are the out-of-sample test of the model.

Table 4.1 Descriptive Statistics

Variables

Number of

observations

Mean

Median

Min

Max

S.D.

Skewness

Kurtosis

Unit root

ADF

SET50 Return

4,150

0.00038

0.00049

-0.17230

0.11431

0.01392

-0.64097

14.77175

-63.86100 ***

(0.00000)

SET Return

4,150

0.00038

0.00062

-0.16063

0.10577

0.01242

-0.81068

16.06744

-63.29000***

(0.00000)

Temperature

4,150

28.99000

29.10000

17.20000

35.00000

1.84000

-0.89000

5.78000

-17.83600***

(0.00000)

Windspeeds

4,150

16.01000

15.00000

0.00000

150.00000

5.86000

3.89000

70.14000

-43.78000***

(0.00000)

Precipitation

4,150

6.96000

1.50000

0.00000

123.30000

12.22200

3.56000

20.37000

-49.77778***

(0.00000)

Noted: This table shows the descriptive statistics value of SET index, SET50 index, and weather variable from 1 Jan

2002 to 31 Dec 2018.Unit root ADF column is the statistics test for stationary of each variable. The value in

parentheses represent p-value of the unit root test. *, ** and *** = significance at the 90%, 95%, and 99% confidence

levels, respectively. d.f. = degrees of freedom.

The data set used in the study are daily returns of SET and SET50 index

with weather information which include precipitation (millimeters), temperature (°C),

and wind speed (knots per hour). The SET index information is available on

SETSMART and the weather information has been collected from The Thai

Meteorological department at Don Muang Airport.

SET and SET50 log returns shown in figure 4.1 represent the volatility

clusters. SET and SET50 returns have equal mean of 0.00038. As for the median, SET

return median is greater than SET50. The Skewness of SET and SET50 returns are

Ref. code: 25626102042238OMU

negative or left-skewed. The kurtosis of SET and SET50 returns are greater than 3,

which represent the fat-tail in their distributions.

(a) SET Index Return

(b) SET50 Index Return

Figure 4.1 SET Index and SET50 Index Returns

Noted: This graph shows daily log returns of SET and SET50 index start from 1 January 2002 to 31

December 2018.

(a) SET Index

(b) SET50 Index

Figure 4.2 SET Return and SET50 Return Distributions

Noted: These two figures show the distribution of SET and SET50 returns.

Ref. code: 25626102042238OMU

(a) Wind Speed

(b) Precipitation

Figure 4.3 Weather Variables Graph

Noted: These figures show the daily value of wind speeds(a), precipitation(b), and temperature(c) from

1 Jan 2002 to 31 December 2018. The horizon line is the mean of each variable. The data includes 4,150

observations in total.

Ref. code: 25626102042238OMU

(a) Wind Speed

(b) Precipitation

Figure 4.4 Weather Variable Distributions

Noted: Figure 4.4 shows the distributions of wind speeds, precipitation, and temperature from 1 Jan 2002

to 31 December 2018, the data includes 4,150 observations in total.

The wind speed is valued between 0 to 35 knots per hour. It repeats the

cycle over the time. However, the precipitation graph shows the pattern of Thailand

seasonal patterns in which the amount of precipitation is usually high during the June

to August period. The value of precipitation is equal to zero on days without rain. In

conclusion, the weather variables show the unique characteristic of Thailand seasons.

This support the finding of Khanthavit (2018) which pointed out that the Thailand

weather variables have high seasonality.

As for temperatures, Bangkok temperatures are negative, which means that

most of the days, temperatures are less than the mean, which is accounted for 28.99 °

C. On average, Thailand's temperatures are always high in March to May and low in

the last quarter of the year.

The dummy weather variables for variations with precipitation (wet),

temperature (heat), and wind speed (high winds) are reported in Table 4.2. Wet dummy

Ref. code: 25626102042238OMU

equal to one in precipitation days is greater than the mean plus one standard deviation.

The total number of wet days in the model is 405 days, which is accounted counts

around 9.75% of the total precipitation observations (4,150). The average annual wet-

day is 24 days for the 34-day average high-temperature day-years, which is at about

14% of all observations. The number of hot days was approximately 34 per year

calculated from 2002 to 2016, but in 2017 and 2018, the number of hot days increased

to about 40 - 41 per year. Given the impact of global warming on Thailand's

temperatures in windspeeds, the average high windspeed days are 38 days per year,

which is counted for 15% of the total wind speeds noting the high wind speed days

increasing over time from 35 days per year and ends with 41 days a year.

Table 4.2 Dummy Variable

Year

Dummy = 1

Dummy=0

Percentage of total

MON

Wet

Heat

High

wnds

MON

Wet

Heat

High

wnds

MON

Wet

Heat

High

wnds

2002

194

211

199

203

18%

11%

16%

15%

2003

200

220

209

212

19%

11%

15%

14%

2004

197

225

208

213

20%

15%

13%

2005

199

223

218

206

19%

11%

16%

2006

198

223

212

208

19%

13%

14%

2007

201

218

213

208

18%

11%

13%

15%

2008

199

216

213

204

19%

13%

14%

17%

2009

196

219

207

204

19%

10%

15%

16%

2010

195

211

207

19%

13%

14%

2011

198

221

211

208

19%

14%

15%

2012

200

221

218

211

18%

10%

11%

14%

2013

201

217

211

205

18%

11%

14%

16%

2014

197

229

208

205

20%

15%

16%

2015

196

228

208

203

19%

14%

16%

2016

198

222

218

207

19%

11%

15%

2017

199

223

203

204

18%

17%

16%

2018

199

218

205

204

19%

11%

16%

17%

Total

783

405

582

638

3367

3745

3568

3512

321%

166%

238%

261%

Average

198

220

210

207

19%

9.75%

14%

15%

Noted: This table shows the data for the dummy variable, Mon is the Monday effect variable, the wet

dummy variable is precipitation, the temperature simulation variable Heat = temperature and the dummy

variable Highwnds = Windspeeds aggregated data. Up to 4,150 days from January 1, 2002, to December

31, 2018, the row of each year separated in 3 particles, dummy = 1 is the date that the dummy variable

we assumed is 1, dummy = 0 for that day our dummy variable is considered to be 0. And the percentage

of the total is the percentage of dummy = 1 divided by total observations.

Ref. code: 25626102042238OMU

The Monday Effect dummy variable has value equal to one for 783 days of

4,150 days, which is approximately 19%. These dates in each year do not include

holidays, and other non-working days. It includes only SET working days per year

which is around 244 days.

After the dummy variables replication is completed, the next step is to

disclose the results describing the GARCH approximation, both the common model

and the weather model, from 4.3 to 4.7. The durability test results are also shown in 4.8

to 4.11.

4.1 Common Model Results

The common model has two components which are mean equation and

variance equation as shown in the table 4.3. Conditional variance term, Monday effect

dummy variable and lagged of return are applied as an independent variable in the mean

equation. The variance equation is explained by lagged of residual term and lagged of

variance.

The results of SET index in the mean equation represent the significance

of the Monday effect dummy variable in 2005 and 2008 at 99% confident level and

2015 at 95% confident level respectively. Lagged of return coefficients are significant

in 2003 at 90% confident level and 2007 at 95% confident level.

In SET index variance equation, lagged of residual coefficients are significant

in 2004, 2010, 2011, 2012 and 2013 at 95% confident level. Moreover, lagged of

residual coefficients are significant in 2007 and 2008 at 99% confident level and in

2014, lagged of residual coefficient is significant at 90% confident level respectively.

Conditional variance terms or GARCH term coefficients are significant in 2002, 2003,

2004, 2007, 2008, 2009, 2011, 2012, 2013, 2014, 2015, 2016, and 2018 at 99%

confident level but in 2010, GARCH coefficients are significant at 95% confident level.

For SET50 index mean equation, Monday effect dummy variable

coefficients are significant in year 2005 and 2015 at 95% confident level and in year

2008 at 99% confident level. Lagged of returns are significant in 2003, 2007, and 2018

at 95% confident level.

Ref. code: 25626102042238OMU

For the equation of variance for the SET50 index, lagged of residual

coefficients are significant in 2004 and 2011 at 95% confident level. In 2007 and 2008,

lagged of residual coefficients are significant at 99% confident level. In 2012, 2013,

2014, and 2016, lagged of residual coefficients are significant at 90% confident level.

The conditional variance or GARCH term coefficients are significant at 99% in 2002,

2003, 2004, 2007, 2008, 2009, 2011, 2012, 2013, 2014, 2015, 2016, and 2018 but in

year 2005, GARCH coefficient is only significant at 90% confident level.

Ref. code: 25626102042238OMU

Table 4.3 The Estimated Results of Common Model

Year

SET

SET50

Mean Equation

Variance Equation

Mean Equation

Variance Equation





MONt

















MONt













2002

1.36911

-0.00355

0.03446

0.08547

0.72001***

0.77132

-0.00426

0.05482

0.10914

0.74802***

(1.01448)

(0.00222)

(0.07096)

(0.06851)

(0.19715)

(0.61488)

(0.00251)

(0.07320)

(0.07677)

(0.17719)

2003

-0.44989

-0.00283

0.13349*

0.03765

0.88236***

-0.91411

-0.00267

0.16982**

0.02842

0.86358***

(1.06389)

(0.00184)

(0.07132)

(0.04454)

(0.17685)

(1.55384)

(0.00209)

(0.06985)

(0.04485)

(0.24702)

2004

-0.01813

-0.00251

0.05261

0.11208**

0.84308***

-0.04350

-0.00265

0.04622

0.09802**

0.84977***

(0.26825)

(0.00204)

(0.06880)

(0.05192)

(0.06282)

(0.28427)

(0.00227)

(0.06776)

(0.04713)

(0.05638)

2005

11.77576

-0.00378***

0.10486

0.01123

0.48011

11.03324

-0.00369**

0.09264

0.01278

0.32403*

(20.58069)

(0.00131)

(0.06907)

(0.0206)

(0.42088)

(13.97861)

(0.00149)

(0.07253)

(0.01547)

(0.19529)

2006

0.31425

-0.00185

-0.0054

0.16881

0.21210

0.36488

-0.00126

0.00852

0.15204

0.20396

(0.53465)

(0.00148)

(0.06989)

(0.14539)

(0.31194)

(0.57249)

(0.00170)

(0.06886)

(0.13566)

(0.32839)

2007

-0.05271

0.00013

0.17035**

0.17221***

0.79089***

-0.01612

0.00048

0.15475**

0.16671***

0.80466***

(0.24355)

(0.00143)

(0.07572)

(0.05921)

(0.05843)

(0.23364)

(0.00166)

(0.07423)

(0.05595)

(0.05461)

2008

-0.02752

-0.00604***

0.07152

0.19661***

0.78979***

-0.01089

-0.00681***

0.05930

0.20177***

0.78538***

(0.19377)

(0.00203)

(0.06932)

(0.07495)

(0.06483)

(0.18764)

(0.00240)

(0.06947)

(0.07662)

(0.06636)

2009

-0.1716

-0.00136

-0.07025

0.06959

0.80847***

-0.23539

-0.00156

-0.07637

0.06023

0.82644***

(0.46762)

(0.00223)

(0.06961)

(0.05510)

(0.10235)

(0.48837)

(0.00255)

(0.06969)

(0.05084)

(0.08716)

2010

-0.12296

0.00020

0.07336

0.10575**

0.84404**

-0.19472

0.00034

0.07327

0.09559

0.85107

(0.25796)

(0.00135)

(0.06922)

(0.05030)

(0.06468)

(0.28337)

(0.00159)

(0.06974)

(0.04662)

(0.06514)

2011

0.25314

-0.00066

0.04307

0.15506**

0.77026***

0.33020

-0.00044

0.03733

0.15009**

0.77096***

(0.28759)

(0.00173)

(0.08005)

(0.06645)

(0.09893)

(0.30758)

(0.00198)

(0.08062)

(0.06385)

(0.09804)

2012

-0.25459

-0.00087

-0.03089

0.07167**

0.90979***

-0.31728

-0.00131

-0.04063

0.06596*

0.91921***

(0.34013)

(0.00127)

(0.06975)

(0.03492)

(0.04532)

(0.33531)

(0.00146)

(0.07026)

(0.03464)

(0.04359)

2013

-0.14960

-0.00202

0.02381

0.13109**

0.83633***

-0.10238

-0.00202

-0.00524

0.07444*

0.90042***

(0.24179)

(0.00183)

(0.07146)

(0.05742)

(0.06270)

(0.2522)

(0.00201)

(0.07145)

(0.04082)

(0.04817)

2014

-0.14207

-0.00025

0.00116

0.09401*

0.80494***

-0.11273

-0.00028

-0.00303

0.08682*

0.81096***

(0.26820)

(0.00094)

(0.07294)

(0.05031)

(0.06668)

(0.28052)

(0.00106)

(0.07713)

(0.04640)

(0.06059)

Ref. code: 25626102042238OMU

Table 4.3 Continued

Noted: This table shows the coefficient of each variable in the GARCH form, both in equation and variance, are the equations of the SET Index and SET50 Yield

Index, total data of 4,150 days from January 1, 2002, to December 31, 2018. The value in parentheses is the standard deviation *. , ** and *** = significant at 90%,

95% and 99% confidence levels, respectively, df = degrees of freedom.

The above results in table 4.3 are obtained from GARCH (1,1) model, which is added the Monday effect dummy variable, and

lagged mean equation variance for describing the returns. In the next part, the return will be explained by weather variables and Monday

effect dummy variable in mean equation and in variance equation. The variance is described by the weather variables shown in Table 4.4.

Year

SET

SET50

Mean Equation

Variance Equation

Mean Equation

Variance Equation





MONt

















MONt













2015

4.90964

-0.00262**

-0.02859

0.02146

0.83096***

2.57039

-0.00317**

-0.05502

0.03980

0.77927***

(5.56026)

(0.00122)

(0.05123)

(0.02728)

(0.08241)

(2.27126)

(0.00145)

(0.06469)

(0.03973)

(0.12508)

2016

0.30794

0.00013

0.00111

0.13440**

0.73034***

0.25685

-0.00012

-0.00529

0.08751*

0.83286***

(0.31100)

(0.00127)

(0.06932)

(0.06913)

(0.14115)

(0.32671)

(0.00150)

(0.06629)

(0.04553)

(0.07899)

2017

11.3468

0.00013

0.0178

0.00629

0.29925

5.46525

-0.02513

0.00500

0.15000

0.60000

(36.24144)

(0.00058)

(0.06196)

(0.02056)

(0.38405)

(1.19382)

(0.00188)

(0.06950)

(0.04900)

(0.14822)

2018

14.06729

0.00130

0.02152

0.01595

-0.55435***

9.89367

-0.08253

0.06017**

0.02070

-0.6105***

(34.34670)

(0.00104)

(0.03748)

(0.04142)

(0.16752)

(13.35845)

(0.11176)

(0.02981)

(0.03114)

(0.14700)

Ref. code: 25626102042238OMU

4.2 Weather Model Results

Table 4.4 The Effects of Weather on SET Index Return

Year

Lagged Variables

MONt

Weather variables

Wald

Chi-square









Heat

Wet

Windspd

2002

0.09004

0.12601

0.00014

-0.00154

0.00117

0.00028

1.25904

(0.30072)

(0.06982)

(0.00175)

(0.00191)

(0.00215)

(0.00256)

2003

1.00729

0.14149**

-0.00288

0.00280

0.00045

0.00036

1.19716

(1.34672)

(0.07082)

(0.00181)

(0.00279)

(0.00329)

(0.00369)

2004

-0.0612

0.06726

-0.00291

0.00361

0.00212

-0.00541*

6.42915*

(0.22176)

(0.06869)

(0.00194)

(0.00235)

(0.00294)

(0.00283)

2005

11.76817

0.09913

-0.00378***

-0.00249

-0.00061

0.00028

0.48062

(18.50395)

(0.06875)

(0.00136)

(0.00374)

(0.00356)

(0.00302)

2006

0.60877***

-0.04142

-0.00355**

-0.00013

0.00656***

-0.00242

21.31345***

(0.13434)

(0.05695)

(0.00140)

(0.00230)

(0.00144)

(0.00177)

2007

-0.07328

0.13611*

0.00028

-0.00536***

0.00233

0.00194

15.40771***

(0.15335)

(0.07304)

(0.00128)

(0.00169)

(0.00177)

(0.00138)

2008

0.16571

0.02663

-0.00320

-0.00360

-0.00027

0.00118

21.31345

(0.22716)

(0.05776)

(0.00433)

(0.00257)

(0.00287)

(0.00261)

2009

0.44763

-0.02908

0.25207

0.00445

0.005900*

0.00183

3.32760

(0.39370)

(0.06184)

(0.22372)

(0.00347)

(0.00353)

(0.00276)

2010

-0.64944*

0.01363

-0.00043

-0.00478**

-0.00384*

-0.00382*

8.72038**

(0.36411)

(0.06747)

(0.00143)

(0.00211)

(0.00220)

(0.00209)

2011

0.24090

0.01779

-0.00046

-0.00353

-0.00185

-0.00210

2.70794

(0.17692)

(0.07156)

(0.00182)

(0.00287)

(0.00259)

(0.00227)

2012

-0.43021

-0.07694

-0.00081

-0.00328

-0.00367*

-0.00202

9.40533**

(0.58237)

(0.07132)

(0.00122)

(0.00200)

(0.00196)

(0.00171)

2013

-0.18476

-0.00678

-0.00149

-0.00364

0.00132

0.00372*

6.74396*

(0.19411)

(0.07302)

(0.00166)

(0.00217)

(0.00246)

(0.00211)

2014

0.24043

-0.01774

0.00012

-0.00227

-0.00089

0.00097

3.49015

(0.26480)

(0.07790)

(0.00094)

(0.00119)

(0.00176)

(0.00126)

2015

26.20654*

-0.07275

-0.00302**

0.00227

-0.00107

0.00044

2.78529

(14.46924)

(0.05372)

(0.00127)

(0.00146)

(0.00253)

(0.00155)

2016

0.24043

-0.01774

-0.00094

-0.00227

-0.00089

0.00097

3.40085

(0.29796)

(0.06818)

(0.00123)

(0.00139)

(0.00183)

(0.00156)

2017

2.62984***

0.00800

-0.00007

-0.00033

0.00208

-0.00498

1.61006

(0.16736)

(0.32471)

(0.00215)

(0.00347)

(0.07005)

(0.00395)

2018

1.25700*

-0.04351

0.00169

0.00069

0.00836***

-0.00499

22.83552***

(0.76163)

(0.06287)

(0.00114)

(0.00160)

(0.00183)

(0.00141)

Noted: This table shows the coefficients of each variable in the mean of the SET index return equation.

The last column is a test of the climate variable to test whether it is 0 or not. The value in parentheses is

the standard deviation *, ** and. *** = Significant at 90%, 95% and 99% confidence levels, respectively,

df = degrees of freedom.

Table 4.4 reports the impact of the weather on the return of the SET Index

for each year from 2002 to 2018. Returns are affected by delayed returns, delays

invariance, and weather variables, including heat, wet, and Windspds. The study results

showed that the coefficients of the climate variable were significant in 2006, 2007, 2010

Ref. code: 25626102042238OMU

and 2018. These results consisted of the Wald test that the climate variable was

simultaneously different from zero at a confidence level of 99. % The effect coefficient

was significant at 95% in 2006 and 2015 but was significant at 99% in 2005.

Table 4.5 The Effects of Weather on SET Index Volatility

Year

Weather variables

Wald

Chi-square

Heat

Wet

Windspd

2002

0.94268***

0.04653

0.00001

0.00001**

0.00009***

82.31551***

(0.03582)

(0.03378)

(0.00001)

(3.14E-06)

(0.00002)

2003

0.45722

0.00992

-0.00003

0.00002

0.00005

3.31934

(0.36018)

(0.05500)

(0.00003)

(0.00004)

2004

0.82636***

0.12251**

-0.00001

-0.00008**

0.00004

5.26427

(0.06303)

(0.05490)

(0.00003)

(0.00004)

2005

0.46575

0.01287

3.90E-06

2.39E-06

6.18E-07

0.21667

(0.34826)

(0.02085)

(0.00001)

(4.27E-06)

2006

0.25842

0.08241

0.00002

-0.00010***

0.00003

25.70262***

(0.23546)

(0.05401)

(0.00005)

(0.00002)

(0.00003)

2007

0.82879***

0.18423***

-0.00001

0.00001

-0.00002**

6.42510*

(0.03851)

(0.06571)

(0.00001)

2008

0.57893***

0.23170

-0.00018

-0.00020

-0.00015

1.78490

(0.16513)

(0.23002)

(0.00016)

(0.00023)

(0.00019)

2009

0.58076***

-0.06942*

-0.00018***

-0.00010

-0.00011***

15.43083***

(0.21222)

(0.03839)

(0.00006)

(0.00007)

(0.00004)

2010

0.60309***

0.25207

-0.00005

-0.00007

4.46884

(0.11362)

(0.22372)

(0.00003)

(0.00007)

(0.00006)

2011

0.89395***

0.07173**

0.00008**

-5.65E-06

3.12E-07

11.50751***

(0.04268)

(0.03515)

(0.00003)

(0.00004)

(0.00003)

2012

0.52460*

-0.00046

-0.00004

0.00002

2.11787

(0.28219)

(0.00182)

(0.00003)

(0.00005)

2013

0.85497***

0.13696**

0.00001

-0.00002

0.00003

2.33800

(0.05199)

(0.05363)

(0.00002)

2014

0.78801***

0.08492*

0.00001

9.27E-08

9.12E-06

3.84020

(0.05465)

(0.04447)

(0.00001)

2015

0.93790***

0.00198

-5.2E-07

1.01E-06

-3.5E-07

2.27007

(0.02953)

(0.00125)

(4.48E-07)

(7.19E-07)

(6.42E-07)

2016

0.78801***

0.10379*

0.00001

9.27E-08

9.12E-06

0.56909

(0.11943)

(0.05602)

(0.00002)

(0.00001)

(0.00002)

2017

0.58547***

0.14243***

-2.7E-06

-0.32759

-0.00001

1.42114

(0.12933)

(0.00319)

(0.00002)

(0.49057)

(0.00002)

2018

0.45795**

0.06410

-0.00002*

-0.00003***

-0.00001

37.62407***

(0.18983)

(0.05372)

(0.00001)

Noted: This table shows the coefficients of each variable in the SET index return variance equation. The last column

is a test of the weather variable to test whether it is 0 or not. The value in parentheses is the standard deviation *, **

and. *** = Significant at 90%, 95% and 99% confidence levels, respectively, df = degrees of freedom.









Ref. code: 25626102042238OMU

Table 4.5 shows the effect of inclement weather on volatility. The Wald

test showed that the results were significant for 2006, 2009, 2014, 2016, and 2018 at

99% confidence levels, but the results of each test were mixed at 95% confidence levels,

climate impacts on returns volatility were significant in 2011 and 2017 at 90%

confidence levels, and the weather affected volatility in 2010.

Table 4.6 The Effects of Weather on SET50 Index Return

Year

Lagged Variables

MONt

Weather variables

Wald Chi-

square









Heat

Wet

Windspd

2002

1.03701***

0.05251

-0.00395*

-0.00221

0.00028

0.00022

0.66737

(0.27327)

(0.0664)

(0.00214)

(0.00285)

(0.00295)

(0.00361)

2003

-5.70172

0.16148**

-0.00246

0.00069

0.00129

0.00300

1.49331

(6.31741)

(0.06659)

(0.00219)

(0.00274)

(0.00333)

(0.00256)

2004

-0.11295

0.05717

-0.00299

0.00396

0.00199

-0.005767*

6.17748

(0.23063)

(0.06642)

(0.00213)

(0.0025)

(0.00307)

(0.00322)

2005

9.67455

0.09478

-0.00369**

-0.00393

-0.00044

0.00023

0.69296

(11.00612)

(0.071638)

(0.00154)

(0.004836)

(0.004615)

(0.003765)

2006

0.63721**

-0.03347

-0.00312*

-0.00072

0.00734***

-0.00265

52.73021***

(0.2487)

(0.05809)

(0.00165)

(0.00275)

(0.00102)

(0.00181)

2007

0.08533

0.11029

0.00015

-0.00650***

0.00306

0.00059

13.64634***

(0.16574)

(0.07475)

(0.00154)

(0.002)

(0.00273)

(0.00188)

2008

0.11064

0.02333

-0.00608**

-0.00023

-0.00554

0.00132

2.70493

(0.30556)

(0.04166)

(0.00272)

(0.00323)

(0.00344)

(0.00335)

2009

0.19406

-0.02747

0.00006

0.00198

0.00455

-0.00142

1.43570

(0.62318)

(0.08014)

(0.00328)

(0.00464)

(0.00456)

(0.0038)

2010

-0.19566

0.04775

-0.00012

-0.00545***

-0.00160

-0.00084*

11.64771***

(0.21263)

(0.07743)

(0.00149)

(0.00174)

(0.00195)

(0.00047)

2011

0.27491

0.00816

-0.00011

-0.00400

-0.00207

-0.00241

2.79135

(0.19015)

(0.07208)

(0.00213)

(0.00324)

(0.00289)

(0.00255)

2012

-0.24527

-0.06210

-0.00119

-0.00220

-0.00274

-0.00177

6.77906*

(0.27968)

(0.07338)

(0.00142)

(0.00163)

(0.00182)

(0.0013)

2013

-0.07569

-0.03159

-0.00272

-0.00393

0.00138

0.00480**

8.22809**

(0.22127)

(0.07287)

(0.00555)

(0.00245)

(0.00294)

(0.00221)

2014

-0.56734

0.01305

-0.00061

-0.00008

0.00006

-0.002911*

3.30532

(0.46075)

(0.05307)

(0.00111)

(0.00153)

(0.00229)

(0.00169)

2015

13.48187**

-0.08682**

-0.003396**

0.00257

0.00038

-0.00011

2.39266

(5.87171)

(0.04395)

(0.00151)

(0.00166)

(0.00298)

(0.00182)

2016

0.18961

-0.02897

0.00022

-0.00289*

-0.00116

0.00153

3.49077

(0.227)

(0.06117)

(0.00159)

(0.00175)

(0.00231)

(0.00205)

2017

43.79174

-0.02145

-0.00027

-0.02599

0.07146

-0.00731

2.67745

(54.65792)

(0.03869)

(0.00305)

(0.03198)

(0.0804)

(0.00487)

2018

1.21655***

-0.03713

0.00170

0.00052

0.00851***

0.00168

18.47950***

(0.42601)

(0.06539)

(0.00128)

(0.00172)

(0.00205)

(0.00147)

Noted: This table shows the coefficients of each variable in the mean of the SET50 index return equation. The last

column is a test of the weather variable to test whether it is 0 or not. The value in parentheses is the standard deviation

*, ** and. *** = Significant at 90%, 95% and 99% confidence levels, respectively, df = degrees of freedom.

Ref. code: 25626102042238OMU

Table 4.6 shows the effect of climate variables on the return of the SET50

index. The Wald test discloses the weather coefficients differed significantly from zero

in 2006, 2007, and 2018 at a 99% confidence level in 2009. The results, in contrast to

the SET yields, were insignificant at 95% confidence levels in 2009. The Monday

coefficient effect result was significant at 90% in 2002 and 2006, but was 95%

significant in 2005, 2008 and 2015

For the heat variable, the coefficient represents a negative sign. As

temperatures rise for one unit, the return tends to be significantly lower for 0.0065 units

in 2007 and 0.00545 units in 2010. The results consisted of the procedure (2017), with

most of the significant coefficients having a minus sign. The difference is the year that

is significant. In his paper, the coefficient of the temperature variable was significant

only in 1999, but the above results show a significant coefficient in 2007, 2010 and 2016.

For the wet variables, the variables were positive in 2006 and 2018,

consistent with Khanhwit’s results (2017), but contrary to the David and Tyler’s (2003)

which displayed significant negative precipitation coefficients.

For wind speed, the results were most significant at 90%. The remarkable

result was a significant coefficient of 90% with a negative sign, but at 95% the result

was significant with a positive sign. A significant coefficient of 95% agrees with the

results of Khanthwit (2017), showing a positive sign of 95% of the wind speeds variable

in his paper.

Table 4.7 proclaims the result of weather variable effect on SET50 index

volatility which indicates that weather variables can influence the volatility in 2002,

2006, 2009, 2011, and 2018, giving different results to the SET50 index. This means

that the volatility of the SET50 Index does not affect differently from that of the SET

Index. The impact of weather variable on volatility is the same for 2006 and 2018.

The results of the heat variable showed a significant negative relationship

among them, with temperature increases for one unit, the returns tend to be lower than

0.00017 units in 2009 and 0.00004 in 2014. (2017) where temperature coefficients are

negative.

Ref. code: 25626102042238OMU

Table 4.7 The Effects of Weather on SET50 Index Volatility

Year

Weather variables

Wald Chi-square

Heat

Wet

Windspd

2002

0.76806***

0.10447*

0.00005

-0.00011*

0.00021**

5.02220

(0.05525)

(0.05786)

(0.00003)

(0.00007)

(0.00010)

2003

1.01666***

0.00154

-2.16E-06

0.00001

1.38E-06

0.86505

(0.00114)

(0.00406)

(4.60E-06)

(0.00001)

2004

0.84165***

0.10272**

-0.00001

-0.00009**

0.00006

4.52862

(0.05627)

(0.04820)

(0.00003)

(0.00004)

2005

0.37462**

0.01579

0.00001

3.88E-06

8.51E-07

0.70340

(0.18366)

(0.01667)

(0.00001)

2006

0.24108

0.08331*

0.00002

-0.00015***

0.00004

25.72714***

(0.25581)

(0.05037)

(0.00007)

(0.00003)

(0.00005)

2007

0.85268***

0.14625***

0.00001

0.00004*

-0.00001

4.50097

(0.02959)

(0.03577)

(0.00002)

2008

0.55472**

0.12295

-0.00019

-0.00021

-0.00026

1.51072

(0.22996)

(0.12573)

(0.00023)

(0.00030)

2009

0.57897***

-0.01146

-0.00017***

-0.00013***

-0.00016***

562.65320***

(0.19793)

(0.07584)

(0.00004)

(0.00006)

2010

0.72402***

0.18142**

-0.00002

-0.00003

0.00007*

6.87641*

(0.09060)

(0.08575)

(0.00002)

(0.00004)

2011

0.89366***

0.07271**

0.00009**

-0.00001

2.07E-06

11.21457**

(0.04540)

(0.03595)

(0.00004)

(0.00003)

Noted: This table shows the coefficients of each variable in the SET50 index return variance equation. The last

column is a test of the climate variable to test if it is equal to 0. The value in parentheses is the standard deviation *,

** and. *** = Significant at 90%, 95% and 99% confidence levels, respectively, df = degrees of freedom.

For the wet variable results, significant negative sign coefficients were

shown in 2006, 2009, 2017, and 2018 at a 99% confidence level. The findings are

similar to those of Khanthwit (2017) in 2006 which marks the same coefficient sign.

The wind speed variable results showed a positive sign in 2009. These

findings are in contrast to the results of Khanthwit (2017) which did not significantly

show the wind speed coefficients in 2009, but his papers show a significant positive

coefficient in 2016.









Ref. code: 25626102042238OMU

Table 4.7 Continued

Year

Weather variables

Wald Chi-square

Heat

Wet

Windspd

2012

0.92846***

0.04690

-5.39E-07

0.00001

-0.00001

1.65823

(0.06088)

(0.03850)

(0.00001)

2013

0.89545***

0.08472*

8.46E-07

0.00001

0.21864

(0.05003)

(0.04470)

(0.00002)

(0.00003)

(0.00002)

2014

0.58220***

-0.05774

-0.00004***

-0.00004**

-0.00005**

21.68389***

(0.13207)

(0.04201)

(0.00001)

(0.00002)

2015

0.94783***

0.00273

-1.23E-06

1.95E-06

5.25E-08

2.51686

(0.03573)

(0.00182)

(1.03E-06)

(1.39E-06)

(1.58E-06)

2016

1.00849***

-0.01820

0.00002*

0.00001

0.00003**

28.57786***

(0.00023)

(0.00903)

(0.00001)

2017

0.03676

0.00087

0.00001

-0.00001***

1.55E-06

7.94218**

(0.05078)

(0.00186)

(0.00001)

(4.61E-06)

(1.55E-06)

2018

0.56885***

0.05523

-0.00002**

-0.00004***

-0.00002

21.79470***

(0.15477)

(0.04467)

(0.00001)

Noted: This table shows the coefficients of each variable in the SET50 index return variance equation. The last

column is a test of the climate variable to test if it is equal to 0. The value in parentheses is the standard deviation *,

** and. *** = Significant at 90%, 95% and 99% confidence levels, respectively, df = degrees of freedom.

The return and volatility results are simultaneously affected by the weather

variables in both the SET Index and the SET50 index, in 2006, 2007, 2010 and 2018.

Also, due to strongly significance of weather variables both in SET and SET50 index

volatility in 2009. In the next step, the year 2006, 2007, 2009, 2010 and 2018 are used

as the benchmark year to test the predictability of the weather variables on the return

forecasting.

4.3 Returns Forecasting Results

Table 4.8 RMSE of SET Index Returns from 1-period ahead Forecasting

Year

Month

2006

2007

2009

2010

2018

Common

Weather

Common

Weather

Common

Weather

Common

Weather

Common

Weather

RMSE

0.016344

0.015785

0.01422

0.014204

0.018204

0.01818

0.012628

0.012704

0.008671

0.008643

Two tail

Two

sample

T-test

0.762675

(0.446399)

0.361568

(0.717987)

0.458878

(0.646733)

-1.734516

(0.084105*)

0.883236

(0.377978)

Note: This table shows the full-year RMSE home, which is the result of the typical model daily SET index profit

destruction. (No weather) and weather model, the two performance results are shown in the last row, this table only

shows the year. From Tables 4.2 to 4.5, 2006, 2007, 2009, 2010 and 2018, the p-values for t-test *, ** and *** =

significant at a 90% confidence level. , 95% and 99% respectively.









Ref. code: 25626102042238OMU

In 2006, 2009 and 2010, RMSEs of SET common model are greater than

the weather model. However, in two-sample T-test, the result doesn’t significantly

show the difference. In 2007 and 2010, RMSEs of common model are smaller than the

weather model but in 2010, the two-sample test representation is significantly different

at 90% confident level.

Table 4.9 RMSE of SET50 Index Returns from 1-period Ahead Forecasting

Year

Month

2006

2007

2009

2010

2018

Common

Weather

Common

Weather

Common

Weather

Common

Weather

Common

Weather

RMSE

0.01789

0.018149

0.014219

0.014270

0.018214

0.018236

0.012624

0.012727

0.008661

0.008658

Two tail

Two

sample

T-test

-2.324763

(0.020912**)

-0.476604

(0.634070)

-0.213613

(0.831028)

-2.152846

(0.032323**)

0.091637

(0.927061)

Note: This table shows the full-year average RMSE which is the result of forecasting the daily SET50 index return

across the general model. (No weather) and weather model The comparison results of the two examples are shown

in the last row. This table only shows the significant years from Tables 4.2 to 4.5, 2006, 2007, 2009, 2010, and 2018.

The values in parentheses represent the p-values of t-test *, **, and *** = significant at the level. Confidence 90%,

95% and 99% respectively.

In 2007 and 2009, the conventional model's RMSE was smaller than the

weather model, but in 2018, the conventional model's RMSE was greater than that of the

climate model. On the other hand, the results did not differ significantly. In 2006 and 2010,

the conventional model's RMSE was smaller than the weather model. The results of the

two tests showed a significant difference at a 95% confidence level.

Ref. code: 25626102042238OMU

CHAPTER 5

CONCLUSION

This paper provides empirical results on how weather variable affects the

Thailand SET Index and the SET50 index prediction using simulated variables of

temperature, precipitation, and wind speed for one-period ahead prediction.

The results from the GARCH (1,1) model reveal the impact weather

variables on the returns of the SET index in 2006, 2007, 2010, and 2018. The weather

variables affected the volatility of the SET index in 2006, 2009, 2014, 2016, and 2018.

The out-of-sample tests showed that weather variables were unable to improve the one

period ahead predictability. It is worth noting that in 2010, the weather variable raised

the RMSE as it was added to the model.

For the SET50 index, the results showed that weather variables are

significant in 2006, 2007, and 2018. The weather variables were able to influence the

volatility. However, the RMSE value has declined over some years, but it has also

increased in some years as weather variables are added to the model. As the results, two

tests confirmed that the weather variables were unable to support prediction because

the RMSE result was significantly smaller in the common model.

Based on this study, we can conclude that the weather information cannot

play a role in forecasting investor mood which is contradict to the works of Saunder

(1993), Cao and Wei(2005), and Hirshleifer and Shumway(2003). One explanation is

that these works are related to the stock markets located in areas with extreme climates.

Due to the extreme weather, shorter periods of sunlight have become an important

factor influencing people's moods (Keller, 2005) by changing serotonin in the blood

(Denissen, 2008).The main results of this paper is aligned with the paper of

Khanthavit(2017) which find out that the weather is not the main driver of investor

mood. The output that they obtain from the paper confirms that the weather only affects

certain years and does not repeat itself every year. It supports the fact that weather is

not the primary factor in regulating investor mood (David and Tyler, 2003). Not only

the sunshine hours that can control the mood of investors but also fundamental factors

such as of dividend yields, short interest rates also play a key role in predicting returns

(Paye, 2005), which is the sunshine during the darkest hour for investment prediction.

Ref. code: 25626102042238OMU

REFERENCES

Bakri Abdul Karim, M. H. (2018). Weather Mood and Stock Market Return in

Argentina. Accounting and Finance Research .7(4), 159-166.

Bodie, K. a. (2014). Investment (2 ed.). Mc Graw Hill Education.

Bollerslev, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity.

Journal of Econometrics. 31(1986), 307-327.

Cao, W. a. (2005). Stock Market Returns: A Note on Temperature Anomaly. Journal

of Banking&Finance. 29(6), 1559-1573.

Daskalakis, G. a. (2011). Does the Weather Affect Stock Market Volatility? Finance

Research Letters. 7(4), 214-223.

Floros. (2011). On the Relationship between Weather and Stock Market Returns.

Studies in Economics and Finance. 28(March), 5-13.

He, F. (2018). Weather Effect on Stock Returns: Evidence from The Chinese Stock

Market. Physica A: Statistical Mechanics and its Applications. 389(1), 91-99.

Hoffman, E. H. (1984). A Multidimensional Approach to the Relationship Between

Mood and Weather. Brittish Journal of Psychology. 75, 15-23.

Jaap J. A. Denissen, L. B. (2008). The Effect of Weather on Daily Mood: A Multilevel

Approach. American Psychological Association. 8(5), 662-667.

Kamstra, K. a. (2003). Winter Blues: A SAD Stock Market Cycle. American Economic

Review,working paper.13(a), 1-37.

Khanthavit, A. (2017). Instrumental-Variable Estimation of Bangkok-Weather Effect

in the Stock Exchange of Thailand. Asian Academy of Management Journal of

accounting and Finance.13(1), 83-111.

Khanthavit, A. (2017). Weather, Investor Sentiment, and Stock Returns in the Stock

Exchange of Thailand. ABAC Journal . 39(1), 1-14.

Khanthavit, A. (2018). Investors' Weather-related Moods and Stock Market. Human

Behavior, Development and Society .19, 7-9.

Khanthavit, A. (2018). Weather Driven Stock Return Correlations. Working Paper,

Thammasat University, Bangkok, Thailand .

Ref. code: 25626102042238OMU

Markku Kuastia, E. R. (2011). Does Mood Affect Trading Behavior? Journal of

Financial Market . 29(C), 1-26.

Mezard, S. C. (2006). Risk Minimization Through Portfolio Replication. The European

Physical Journal B - Condensed Matter and Complex Systems. 57(2), 175-180.

Nijhout, J. (2016). Mood and Stock Returns The effect of weather conditions and

biorhythm on the dutch stock market. Working Paper, Erasmus University,

Rotterdam.

Orskaug, E. (2009). Multivariate DCC-GARCH Model. Norwegian Computing Center

Working Paper.

Paye, B. S. (2005). Instability of Return Prediction Models. Journal of Empirical. 13(3),

274-315.

Sang Hoon Kang, Z. J.-M. (2010). Weather Effects on the Returns and Volatility of the

Shanghai. Physica A. 389(1), 91-99.

Sapienza, C. P. (2004). The Real Effects of Investor Sentiment. Corporate Finance:

Valuation, NBER Working Papers.

Saunders. (1993). Stock Prices and Wall Street Weather. The American Economic

Review. 83(5), 1337-1345.

Shumway, D. H. (2003). Good Day Sunshine Stock Returns and the Weather. The

Journal of Finance.58(3), 1009-1032.

Sogner, M. F. (2014, Jan 15). Weather& SAD Related Mood Effects on the Financial

Market, The Quarterly Review of Economics and Finance, Elsevier. 57(C), 11-31.

Tsay, R. S. (2005). Analysis of Financial Time Series (2 ed.). John Wiley & Sons.

Zuckerman, M. (2015). Sensation Seeking: Behavioral Expression and Biosocial Bases.

International Encyclopedia of the Social & Behavioral Sciences, 607-614.

Ref. code: 25626102042238OMU

BIOGRAPHY

Name

Mr Pattharadanai Jaturaporn

Date of birth

March 28, 1994

Educational attainment

2013: Thammasat University, Bachelor of

Economics

Work Experiences

Investment Consultant

Maybank Kim Eng (Thailand)

Ref. code: 25626102042238OMU

1 views·41 pages

YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF Free Download

YOU ARE MY SUNSHINE: EVIDENCE FROM THAILAND STOCK MARKET PDF free Download. Think more deeply and widely.

Uploaded by rebeccawall1980 on 4/9/2026

/41

100%