Weather Effect on stock trading: evidence from the Chinese market

(1)

Weather Effect on stock trading:

evidence from the Chinese market

by Weiting Tong

s1752979

University of Groningen Faculty of Economics and Business

(2)

Acknowledgement

(3)

Abstract

(4)

I. Introduction

The efficient market theory dictates that in an active market, where many well-informed and intelligent investors are present, stocks will be appropriately priced and reflect all available information. In the light of this statement, financial models such as the CAPM or Black & Scholes model are able to tie asset prices and the economic fundamentals together elegantly in one theory. Since they were supported by early empirical research in the 70s, the efficient market theory was widely considered to be proved beyond doubt. Yet faith in this theory was eroded by a succession of discoveries of long-term historical anomalies observed in the 1980s that appear to be contradicting to the theory. Starting with the calendar anomalies such as the January effect or day-of-the-week effect, and with the evidence of excess volatility of returns, many academics realized that the standard financial theories and models are unlikely to provide any satisfactory explanation to these phenomenons. As a result, they turned to research from a broader social science perspective including psychology and sociology. What they discovered is that people are subject to various psychological biases which inhibit their ability to make good investment decisions. In some cases, a particular biased decision could be commonplace in certain group of investors that in turn influences the entire market (e.g. noise trading).

(5)

empirical researches confirm that in general there is a positive relation between the weather variable and stock trading.

While these weather-related theories are becoming increasingly acknowledged and accepted, some scholars still question the validity of such explanations. In the recent debate kindled by Jacobsen and Marquering (2008), it is argued whether those researches are “sound enough to establish the link between weather induced mood changes and stock returns or results in nothing more than data-driven inference based on spurious correlations”. More specifically, while Jacobsen and Marquering (2008) confirms a significant seasonal anomaly in stock returns, they find that a simply winter/summer dummy serves better in describing the anomaly and they simply don’t think there is sufficient evidence in favor of the explanation that weather affects stock markets through mood change of investors. Consequently more research is required to discriminate between the possible explanations.

This paper contributes to the existing behavioral finance literature and the ongoing debate by focusing on the Chinese stock market. We investigate the influence of weather on two stock markets in mainland China in terms of stock return and trading volume. Moreover, unlike some of the previous studies which focus on a specific weather condition, this paper incorporates a group of weather variables in the analysis, namely: the sky coverage index, the adjusted daylight index and the human comfort index which measures thermal comfort of human body including air temperature, relative humidity and wind-speed.

The empirical results show that none of three weather variables has a strong impact on stock returns. However a robust and significant correlation has been found between trading volume of A-shares and human comfort index. Furthermore, evidence shows that trading volume of B-shares became significantly related to human comfort index after domestic investors were allowed to enter the B-share market. This effect remains after controlling for other weather variables and using a weekly sample, indicating that people trade more frequently under higher heat stress.

(6)

methodology used in this analysis, and it elaborates as well how the weather variables are constructed; Section IV provides empirical results with detailed interpretations while Section V presents the robustness checks of the results. The paper concludes in Section VI.

II. Literature Review

2.1 Weather, mood and decision-making

We have all experienced days of being cranky when it is too hot, or being listless even depressed when it is rainy and cold. Many people claim that that their mood shifts with weather. Apparently there is a great deal of truth in that belief - for a large body of psychological research has indeed confirmed a strong link between weather and mood. Back in the 70s, a survey of the impact of weather on mood, employing a large sample of 16,000 students from Switzerland, revealed surprisingly that nearly one-third of the girls and one fifth of the boys responded negatively to certain weather conditions (Faust et al, 1974). Another study by Howard and Hoffman (1984) tracked the mood of 24 college students over eleven consecutive days and found a significant effect on mood correlated with the weather. Specifically, they discover that humidity, temperature, and hours of sunshine have the greatest effect on mood. Such researches have been repeated in many countries, and all come back with similar results.

While being sensitive to weather, mood also has the power to affect our behavior. One common example is that people tend to make more optimistic judgments when they are in good mood. The reason is that good moods are connected with heuristic and less critical information processing whereas bad moods tend to stimulate people to engage in more analytical activity (Schwarz 1990). Hence, by inducing different forms of mood, weather conditions are able to go a step further to influence our behavior, including our financial decision-makings. The academic world has long been aware of this - as long as a century ago, Samuel A. Nelson (1902) reported: “During normal markets, brokers have observed that the psychological factor is so strong that speculators are not disposed to trade as freely and confidently in wet and stormy weather as they are during the dry days when the sun is shining, and mankind is cheerful and optimistic”.

(7)

effect. The first one claims that weather conditions induce changes in judgment of risk or uncertainty. Loewenstein (2000) suggests that people’s cognitive evaluations of risks often diverge from their emotional reactions to those risks. And when such divergence occurs, emotional reactions often drive behavior (Loewenstein 2001). In other words, people can make different risk assessment for the same thing under different emotions. The reason is that different emotions lead to different information processing, eventually to different decision-making under uncertainty. So when you feel upbeat on a sunny day, you tend to believe more about what the media says that the prospect of a certain industry is brightening, and thus likely to buy some related stocks. In contrast, on a gloomy day, you may be more skeptical about the report and tend to listen to others’ opinions before making any investment decisions.

The other explanation relates the phenomenon to misattribution that people sometimes attribute their feelings to the wrong source (Hirshleifer and Shumway 2003a). In a study by Isen et al (1978), they found that by putting people in a good mood at the beginning of an experiment they receive more favorable evaluations of a shopping experience than people give in a neutral mood. This explanation is somewhat similar to the first one in the way that they both claim weather exerts influence on decisions through mood even when the cause of the mood is unrelated to the decision being made. But misattribution has a broader scope that it involves more than risk assessment but many other aspects of decision-making, such as preference. It has been confirmed by the study of Mehra and Sah (2000) which found evidence that the small fluctuations in peoples’ mood can alter their subjective preference, leading to significant volatilities in stock prices.

Based on these two propositions, several attempts have been made by academics to find out the relations between stock trading and specific weather conditions, which can be roughly classified as: Cloud cover, temperature and daylight.

2.2. Cloud cover and stock market

(8)

researched, there is already a lot of evidence that sunshine influences markets. Saunders (1993) studies the relation between local weather in New York City and daily changes in indices of listed stocks (DJIA from 1927 to 1989; NYSE/AMEX from 1962 to 1989). What he discovers is that the level of cloud cover has a long history of significant correlation with major stock indices. This finding was confirmed by Hirshleifer and Shumway (2003) who basically replicate the study of Saunders over the index returns of 26 international stock exchanges from 1982 to 1997. Again, sunshine is found to be significantly correlated with daily stock returns.

2.3 Temperature

Besides sunshine, another weather variable that has been proved to directly affect human behavior is temperature. There are a number of researches that have focused on the link between heat and different types of human behavior (e.g. Rotton & Cohn, 2004; Cohn & Rotton, 2005; Anderson, 1987). Their findings suggest that extremely high or low temperatures tend to cause aggression. Furthermore, the effect of temperature on mood and task performance has also been an important research subject. Some researchers have gathered evidence that task-performing abilities are impaired when individuals are exposed to extremely high or low temperature (Allen and Fisher, 1978; Wyndham 1969). More recently, Seppänen et al. (2006) examines the effects of temperature on performance of office work. By including various objective indicators of performance, they calculated the percentage of performance change per degree increase in temperature, and statistically analyzed measured work performance with temperature. The results show that office-performance increases with temperature up to 21-22 °C and decreases with temperature above 23-24 °C and the highest productivity is at temperature of around 22 °C.

(9)

Moreover, similar to temperature, humidity (Howarth and Hoﬀman 1984; Allen and Fischer, 1978) and barometric pressure (Digon & Bock, 1966) have also been found to be associated with certain mood states. However, hardly any researcher has taken these two variables into account while studying the relation between weather and security trading.

2.4. Daylight and SAD

It is well established in the psychological literature that changes in daylight can trigger different emotions, which is typically characterized by depression during winter months when there is less daylight (e.g. Molin et al, 1996; Young et al, 1997; Rosenthal 1998 etc.). Such syndrome has been extensively documented and is commonly referred as the Seasonal Affective Disorder (SAD). Many scholars ascertain that SAD causes the biological clock of human body to go out of synchronization, therefore upsetting the body's routine, and possibly it can even affect certain hormonal levels in the body. Kamstra, Kramer, and Levi (2003) analyze the stock market returns around the globe and discover a seasonal pattern linking to the length of day. They attributed this seasonal time-variation of to SAD and suggest that it is associated with risk aversion induced by the depression due to diminishing daylight.

(10)

2.5 Thermal comfort index

We notice when reviewing the existing relevant literature that a majority of them appear to have ignored the interaction of climate variables, which particularly concerns the thermal complex (comprises all meteorological elements which have a thermo-physiological effect on humans). Meteorological literature suggests that air temperature alone is not the only determinant of the indoor thermal environment and certainly should not be expected to correlate well with people’s thermal sensations of comfort under all conditions (see Matzarakis and Amelung, 2008). There are many other important parameters also have the power to influence the thermal environment including relative humidity, air velocity, barometric pressure, clothing, and even activity. It is the combinations of these variables that eventually determine the thermal comfort. Take humidity as an example, when the air temperature is modest, it does not appear to have a major impact on thermal comfort. However, in the case of high temperature, high humidity environments tend to have a lot of vapor in the air, which prevents the evaporation of sweat from skin while evaporation of sweat is the main method of heat loss in human bodies. Likewise, small air movement in the office may be perceived as a draft. But if the air temperature is less than skin temperature it will significantly increase convective heat loss.

(11)

the human thermal sensation.

To sum up, there is a growing body of research exploring the effect of weather variables on stock market and the results appear to confirmative in most cases. At the same time, some scholars remain critical towards these findings. Nevertheless, such criticism doesn’t deny the existence of the weather effect, but only proposing interesting questions for further research, and this is exactly the motivation of this paper.

2.6 Weather and trading volume

One of the distinct features of the existing literature on weather effect is that most studies focused solely on the link between weather and stock return, whereas the link between weather and trading volume has hardly ever been mentioned. Such omission does not imply that trading volume is unimportant and not worth studying. In fact trading volume has proved to be a key indicator of the market trend. Zolotoy and Melenburg (2007) studied the relationship between trading volume, volatility, and stock returns across international stock markets. They discovered that lagged trading volumeis positively related to the stock market

volatility which has a negative and statistically significant impact on the serial correlation of

the stock market returns. Such findings revealed the importance of trading volume as an information variable. On the other hand, recent studies in behavioral finance suggest that trading volume also reflects investors’ trading behaviour. One well-known example is so-called noise trading, which refers to the trading based on rumours as if it were real information. For instance, Greene and Smart (2002) found that substantial increases in trading volume and significant but temporary abnormal returns occur when analysts recommend stocks in this column. Such swift response in volume clearly reflects investors’ risk-averse behaviour. Since psychological literature has established that weather is capable of inducing various emotions and influence people’s decision-making, we can hypothesize a link between trading volume and weather condition.

2.7 The Chinese stock market

(12)

(SZSE). Differing from the Hong Kong market which dates back to the middle of 18th century, SSE and SZSE are rather emerging markets (established since early 1990s). Despite of being young, these two markets have experienced tremendous growth in recent years and attracted a great deal of attention around the world. To date, the Shanghai and Shenzhen exchanges list more than 1,500 companies with a combined market capitalization of 2.7 trillion US dollars (2008), ranking the second largest stock market in Asia (after Tokyo Stock Exchange which has a total capitalization of 3.9 trillion US dollars, 2008).

(13)

III. Data Description

3.1 Stock Data

Although there are three major stock exchange markets in China, due to budget constraints, this study only concerns the Shanghai and Shenzhen market. We employ stock data regarding the two markets for a period of ten years - from 01-01-1998 to 31-12-2007, retrieved partly from Datastream and partly from QiangLong Stock database. The stock dataset consists of three parts: the first part contains daily/weekly prices of A-share/B-share indices in SSE and SZSE, both of which are major indices of each market and are reflective of the overall market performance. Based on this, the return can be calculated as follows:

i i i i

P

R





1 Equation (1) Where, Riis the arithmetic return on any given day i ; Piis the closing price of the index on day i; Pi-1 is the closing price on the previous day.

The second part of the dataset concerns trading volume, which measures the liquidity in a market. Previous studies of trading activity in stock market have proposed a number of measures of volume, such as aggregate share volume, aggregate dollar volume, aggregate turnover etc. In this study, we simply define it as the total number of shares traded on a given day (see Ying 1966 and Gallant et al, 1992). This variable is included since fluctuations of which are expected to shed some light on the impact of weather on trading behavior. For instance, if high temperature does induce aggression in risk-taking, one might expect extraordinary high trading volume. Moreover, for the purpose of serving the robustness test, we also collect daily data of a number of individual stocks which belong to sectors that are not easily influenced by weather conditions.

3.2 Weather data

(14)

budgeting constraints we have to reduce the research period to ten years, starting from 01-01-1998 to 31-12-2007.

Based on the raw weather data, we construct three indices: the Sky Coverage Index (SK), the Human Comfort Index (HC) and the Adjusted Daylight (AD) to cover different aspects of the weather effect. The Sky Coverage Index is calculated in a manner similar to Hirshleifer and Shumway (2002) - taking the average cloud cover of each day’s trading hours (between 9am and 3pm). The Adjusted Daylight index is used as a proxy for Seasonal Affective Disorder (SAD). It is calculated by subtracting the hours of any abnormal weather condition (such as rain, snow or fog during the daytime) from the length of the day. Prior literature has proposed two simple ways to measure SAD: the length of night and the length of day. Our method is based on the latter, only we make above mentioned adjustment so that it can be associated with not only the length of day but also the weather during the day. The reason for such adjustments is that despite being closely related to the length of day or night, SAD is also influenced by the daily weather conditions. For instance, with the same length of day, a sunny day should have much greater positive impact on depression than a gloomy and rainy day. Thus, by employing the AD index, we are able to take both aspects into consideration.

Finally, the Human Comfort Index (HC) to account for the condition of thermal environment. It is derived from the Comfort Index originally developed by American Biometeorologist W·H·Terjung. Terjung’s model is characterized by taking both humidity and air temperature into account when considering the human thermal sensation. After having been introduced into China in the 1980s, it was modified by the China Meteorology Administration (CMA) in order to serve more suitably for the Chinese environment and population. These days, it is being extensively used by the media to provide bio-weather information. The formula is presented as follows:

32

25 .

3 )

1 )(

26

8 .

1 (

55 .

0

8 .

1 











T

RH

v

K

Equation (2)

Where, K represents the degree of comfort; T is the air temperature (°C); RH is relative humidity (%) and v is wind velocity (m/s). More details about the classifications of the comfort index are illustrated in table (1).

(15)

et al (2003a) regarding the adoption of outdoor data in the study of weather effect on stock market. In our opinion the use of outdoor data does not heavily affect the results. In fact, it is naive and erroneous to believe that people who stay indoor are free of any effect of weather. Rind (1996) studied people’s tipping behavior in a casino and discovered that those who have stayed all night are more generous in tipping in the next morning when they are told it is a sunny day outside. This finding implies that even if people have no knowledge of the weather condition outside, their behavior is still influenced by the sky conditions. On the other hand, although the thermal environment does appear to change a bit from an outdoor setting to an indoor setting, this difference comes largely from the change in wind-speed which is commonly assumed to be constant in office (0.1m/s, see Höppe1999), whilst there is very limited change in air temperature and humidity. Hence by using a constant wind-speed in calculation, the human comfort index can fairly account for the indoor thermal environment.

IV. Methodology

In the first step, we examine the relation between A-share indices and each weather variable individually. Following Saunders (1993), we first sort the matched data by the value of a particular weather variable in ascending order, and then divide the series into bins. For each bin, we calculate the mean return, the frequency or percentage of positive returns and trading volume (both in number of share and in RMB value). The bin-test is expected to give us the first impression about the relation between the weather variables and the stock market. Moreover, in order to compare the mean returns associated with the two bins covering the lowest and the highest temperatures, and we calculate the Z-score to determine whether this difference is significant. 1 2 1 2 1 1 , / / _ n n score Z k k k mean k        _{Equation (3)} Where k stands for the bin with the highest order; µ stands for the mean return; σ denotes the variance of return and n is the number of observations of the bin.

(16)

& Wei (2004), a lagged return is included in the regression. The reason of including this variable is that, according to the findings of Akgiray (1989), time series of daily stock returns exhibit significant levels of dependence. Therefore, by doing so, we are able to control for possible first-order autocorrelation in returns. Moreover, as we have discussed in previous section, a March effect has been confirmed in the Chinese stock market. To prevent the results from being influenced by this seasonal anomaly, we also include a March dummy in the regression. The regression equations are presented as follows:

t t t t t

R

Mar

SK

R





₀





₁ _₁





₂





₃





Equation (4) t t t t t

R

Mar

HC

R





₀





₁ _₁





₂





₃





R

Mar

AD

R





₀





₁ _₁





₂





₃





_{Equation (6)} Where Rtis the daily stock return based on the closing price of the A-share index on day t; Rt-1

serves as a lagged return variable to account for the non-synchronous trading effect; Mar is a dummy variable which equals 1 when day t is in the month March and zero otherwise; SK represents the fraction of the sky covered by clouds; HC represents the degree of human comfort; and AD denote the adjusted daylight index.

(17)

t t t t t

P

PMax

PMin

SK

V





₀





₁







₂





₃





₄





P

PMax

PMin

HC

V





₀





₁







₂





₃





₄





P

PMax

PMin

AD

V





₀





₁







₂





₃





₄





_{Equation (9)}

Where, Vt represents the total number of shares traded on a day t; Pt denotes the closing price

of the index on day t; ΔP represents change in closing price on a given day t in comparison with the previous day (ΔP=Pt -Pt-1); PMax is a dummy variable for past high price, its value

equals 1 when the current stock price is above 52-week highest price and 0 otherwise (the choice of 52 week is based on its prominence in business press, see Kumar, 2008); PMin is a dummy variable for past low price, its value equals 1 when the current stock price is below 52-week lowest price and 0 otherwise; SK represents the fraction of the sky covered by clouds; HC represents the degree of human comfort; and AD denote the adjusted daylight index.

Next, all the weather variables are tested together in one regression. The purpose is to determine whether the effect of a particular weather condition endures when other weather variables are also present:

t t t t t t t

R

Mar

CL

HC

AD

R





₀





₁ _₁





₂





₃





₄





₅





Equation (10) t t t t t t

P

PMax

PMin

SK

HC

AD

V





₀





₁







₂





₃





₄





₅





₆





Equation (11) Finally, we switch to the B-share market. A special characteristic of this market gives us a unique opportunity to examine whether weather really influences the stock market. As we have discussed in previous section, local Chinese investor were not allowed to trade B-shares, but after 19 February 2001 this restriction was withdrawn. Therefore, we divide the research period into two sub-periods by this date and re-estimate the regression with equation (10) and (11) for each period. Given the fact that local investors rushed into the B-share market after February 2001, we expect to see a notable change in the relationship between stock return/trading volume and the local weather in the second period.

IV. Results and interpretation

(18)

growing significantly during the research period. When comparing the weather, all conditions appear to be similar in both cities: average ratio of sky coverage are both around 4.7 Oktas (see table 1); average lengths of adjusted daylight are 4.8 hours (shanghai) and 4.3 hours (Shenzhen); average scores of the human comfort Index are both in the most comfortable zone, but it is somewhat higher in Shenzhen is than in Shanghai (72 and 62 respectively). This is easy to understandable given the geographical positions of the cities (Shanghai is located in the North Temperate Zone whereas Shenzhen is in the tropical zone). The distribution of the weather variables is illustrated in diagram (1) and (2) in which we can see that the daily comfort index and the adjusted daylight index are highly seasonal at cities, reaching lowest point around January and highest point around June. On the other hand, there is no obvious pattern in the distribution of sky coverage – it is rather scattered through the years.

4.1 Sky coverage and stock market

(19)

4.2 Human comfort and stock market

We calculated the daily Human comfort indices for both Shanghai and Shenzhen from 1998 to 2007. The resulted indices range from 31 to 86 (shanghai) and from 44 to 86 (Shenzhen). Accordingly we sort all the matched data by HC index in ascending order and divide the series into 6 (Shanghai) and 5 (Shenzhen) bins. Similar to what we discover with the bin-test for cloud cover, we observe no obvious relation between HC and stock returns: the returns of each bin appear to be random. Moreover, the percentages of positive return are also on the same level across the bins. Nevertheless a positive relation between HC and trading volume is observed. We notice that trading volume increases significantly with the degree of HC, in both number of shares and in RMB value. Comparing the results of two extreme bins, for Shanghai market, the lowest bin is associated with a daily trading volume of 8.5640 billion yen, whereas the highest bin has a daily trading volume of 24.5837 billion yen, which is almost tripled. Similar results are found for Shenzhen market, where trading volume increases sharply with HC index.

(20)

4.3 Daylight and stock market.

Kamstra et al. (2003) suggest that seasonal affective disorder (SAD) plays an important role in the seasonal variations of stock returns. In their study, SAD is measured by length of night in hours. Instead of using such conventional method, in this study we use the adjusted daylight index to measure SAD. We consider it to be superior for it also takes other daily weather conditions into account, such as cloud cover, rainfall or snow, which are also believed to contribute to the negative effects of SAD (Jack et al 2005).

The results of the bin-test are presented in Table (8). The distributions of the stock returns do not show any obvious pattern; the z-scores of the extreme bins (6, 1) are insignificant for SSE and SZSE market. On the other hand, trading volumes decreases slightly with the adjusted daylight. It might imply that people are less aggressive in terms of security trading when there is more sunshine and the daytime is longer. To further establish this proposition, a regression is to be performed. However, the results in table (9) indicate no strong relation between the adjusted daylight and stock return or trading volume. The coefficient of total daylight is negative and statistically insignificant for both SSE (coefficient: -0.03, t-value: 1.41) and SZSE (coefficient: -0.01, t-value: 1.38).

4.4 Joint effect of weather conditions

So far, we have discovered that among the three weather variables, only the human comfort index has a close association with stock market, specifically with the trading volume which has a significant positive relation with human sensation of thermal comfort. It is now necessary to find out whether such effect is still present after controlling for other weather variables. For this purpose, we put all three variables: cloud cover, human comfort and total daylight in one regression against stock return and trading volume. The results are presented in table (10). Consistent with our previous findings, none of the three weather variables appear to have statistically significant relation with stock returns. As for the trading volume in both markets, after controlling for cloud cover and SAD effects, the coefficient of HC remains positive and significant, once again confirmed the hypothesis that people are more willing to make a deal in stock trading in hot thermal environment.

(21)

price movement has a close relation with trading volume. In general, trading volume increases when stock price are moving upwards and vice versa. Also, we notice that trading volume largely increases when the current price has exceeded a historical high point or hits a historical low point, indicating that these past extreme stock prices are heavily weighted in people’s trading decisions. This result reflects the anchoring effect in trading behavior, which is referred to the tendency to attach or "anchor" one’s thoughts to a reference point - even though it may have no logical relevance to the trading decision at hand.

4.5 B-share and weather

As we have discussed in section II, previous literature argues that weather can influence stock market by inducing different emotions in local traders. Following the logic, if a stock is not traded by local traders but by people living elsewhere, for instance foreign brokers or brokerage firms who bid or sell via remotely internet or telephone, then it should be free of any local weather effect, or at least the impact should be minor. One way to demonstrate this requires attention to the so-called B share in Chinese stock market. B shares are securities listed in mainland China (SSE or SZSE) but traded in foreign currency. Before China opened the B-share market to Chinese domestic investors on February 19 last year, B-share was only allowed to be traded by foreign investors. Therefore, it is interesting to find out whether and how weather influences the B-share market. To do this, we divide the research period into two parts: before and after February 19, 2001. Then we run a regression for each of these two periods. We expect to see a significant relation between weather conditions and B-share trading after this date, whereas before this date, no such relation should exist.

(22)

strengthened in the second period (coefficient: 5.16; and t-value is 2.60). Moreover, the R-square value is also much higher in the second period (before: 0.02; after: 0.18) indicating higher explanatory power of the model when local investors are involved. The result is exactly as we expected - after the B-share market became open to domestic investors, it subsequently became more influenced by local weather.

(23)

(coefficient: 0.006; t-value: 0.0027). It can be interpreted as the result of an influx of domestic investor which evidently diluted the impact of weather conditions in Hong Kong.

4.6 Discussion of the results

So far, we have examined the link between weather and A-share market as well as B-share market. Evidence shows that the sky coverage, thermal comfort and daylight have little influence on the stock returns. The results are monotonic across markets. But it doesn’t deny the existence of weather effect on stock market, for the other important indicator of stock market – trading volume is found to be significantly correlated to weather, especially human thermal comfort. For both A-share markets, trading volume appears to increase with the degree of heat stress. More compelling evidence comes from the B-share market, which shows that after the restriction of domestic B-share trading was withdrawn, trading volume started being closely related to local weather, while such association did not exist during the period when local investors were not allowed to enter the market. All in all, it is confirmed that the thermal environment has the major impact on stock trading volume.

To explain this finding, we need further understanding of the Chinese stock market. First of all, it is common knowledge that both SSE and SZSE are characterized by a vast number of amateur investors who contribute heavily to the total trading volume of the market. These investors have very limited financial knowledge and make trading decisions based solely on the old market

wisdom: buy low, sell high. As a result,they often set an entry- point or exit-point for a stock

(24)

that the speculator would trade more frequently. The collective result of this effect leads eventually to a higher trading volume in the market.

V. Robustness Test

5.1 Weekly data

Jacobsen and Marquering (2008) criticize the study of Cao & Wei (2005) and Kamstra et al. (2003a) for relying solely on daily data because it could lead to spurious significant results. Moreover they also mention that “it is well know that daily data are considerably noisier than for instance monthly data”. Taking this into account, we decide to re-estimate the regressions using weekly data instead of daily data. As suggested by Jacobsen and Marquering, monthly data is also an option, but given the fast changing nature of the weather conditions, weekly data seems to be more suitable here.

The results of the regression for A-share are presented in table (14) to table (16), in which we can see that the relation between trading volume and human comfort remains positive and significant (for SSE: coefficient is 0.11, t-value is 2.36; for SZSE, coefficient is 0.1, t-value is 2.57). We also re-test the regression for B-share, the results are in line with our previous findings – the relation between human comfort and trading volume has been strengthened after the government opened the market for domestic investors (For SSE, in the period from 01-01-1998 to 16-02-2001, coefficient is 0.002 and t-value is 1.62; during 19-2-2001 to 31-12-2007, coefficient is 0.003 and t-value is 2.87). Finally, we notice a great enhancement in R-square value when switching to weekly sample, showing that daily data is indeed much noisier than weekly data.

5.2 Adjusted Index

(25)

The second part of the robustness test is designed to tackle this problem. The idea is to strip off the stocks that are easily impacted by weather and re-estimate the regression as we do in the previous section. It would be very accommodating if SSE and SZSE could provide sector indices, which is indeed the case; however the history of these indices does not back that long, some from 2007, mostly from 2008, which gives us no choice but have to seek alternatives. Fortunately we noticed that in the Qianglong system, all the stocks listed in SSE and SZSE markets are readily divided into 32 categories covering various industries and vast lines of business. Taking this advantage, we hand-picked stocks from each of those sectors and using these stocks reconstructed the major sector indices of SSE and SZSE. The stock in the selection must fulfill following criteria: first, it must be listed throughout the period 1998 to 2007; secondly its major business/product should have no obvious relation with weather conditions. We exclude a number of sectors that are clearly influenced by weather such agriculture or power etc. The resulted indices include media, real estate, financial, pharmaceutical, equipment, IT, chemical, medical device, construction, metal, electronic and retail, 12 sectors in total.

Similar to the original indices of SSE and SZSE, the adjusted indices are also calculated using a Paasche weighted composite price index formula:

1000

_







b it it

C

N

P

Index

Sector

_{Equation (12)}

(26)

VI. Summary and Conclusion

It has been well-documented in behavioral finance literature that weather can impact stock market, via the amount of sunshine (Saunders 1993), temperature (Cao and Wei 2004), the length of night (Kamstra et al 2003a) or other particular weather conditions. While previous studies are largely focusing on developed countries, this paper takes a deep look at the weather effect in the emerging Chinese stock market. We attempt to identify a relation between three weather variables (cloud cover, human comfort, and daylight) and stock trading (stock return and trading volume). To do this, we exploit daily as well as weekly closing prices and trading volumes of A-share and B-share indices in Shanghai and Shenzhen stock market. Both bin-tests and linear regression are employed in the empirical analysis which reveals following key findings.

First of all, contrary to expectations, we found no evidence of weather effect on stock returns. No significant correlation was found between daily index returns and any of the three weather variables, tested either singly or jointly. It remains this way when using weekly samples instead of daily sample.

However, we discovered a significant correlation between human comfort index and trading volumes of A-shares. Evidence shows that trading volume increase with the degree of human comfort index, indicating that investors are more willing to make a deal when they are under more heat stress. The results are monotonic across A-markets. Yet, the analysis of the B-share market provided even more convincing results. The Chinese B-share market was originally only accessible to foreign investor but opened to domestic investors after 19 February 2001. Taking this advantage, we broke down the research period into two parts and analyzed them separately. The results show that, for Shanghai stock market, human comfort had no impact on B-shares during the first period, but after the restriction was withdrawn it appeared to be significantly positively related to trading volume. As for the Shenzhen market, evidence shows that the human comfort index of Hong Kong city had a potent impact on trading volume in the first period during which the B-share investors were concentrated in Hong Kong. This impact was heavily reduced during the second period when domestic investors rushed into the market.

(27)

studies into account. For instance, Cao and Wei (2005) focus solely on temperature and ignore other important conditions that also have impact on human thermal sensation. In this study, we introduce a relative new weather variable – human comfort index into the research to incorporate humidity and wind-speed. Moreover, Jacobsen and Marquering (2008) question Cao and Wei (2005) and Kramer et al (2003a) for using daily data which is considered to be noisy and may lead to spurious results. Taking this into consideration, we also employ weekly sample in the analysis, which eventually delivers similar results as daily data.

(28)

Reference

[1] Akgiary, V., (1989). “Conditional Heteroscedasticity in Time Series of Stock Returns: Evidence and Forecasts”, Journal of Business, January 1989, 62, 55-80.

[2] Allen, A.M., & Fisher, G.J., (1978). “Ambient temperature affects on paired associate learning”, Ergonomics 21 (2), 95–101.

[3] Anderson, C.A., (1987). “Temperature and aggression: Effects on quarterly, yearly, and city rates of violent and nonviolent crime”, Journal of Personality and Social Psychology, 52(6), 1161-1173.

[4] Büttner. K., (1938). “Physikalische Bioklimatologie”. Akademische Verlagsgesellschaft

Leipzig.

[5] Cao, M., & Wei, J., (2004). “Stock market returns: A note on temperature anomaly”, Journal of Banking & Finance 29 (2005) 1559–1573.

[6] Cohn, E.G.., & Rotton, J., (2005). “The curve is still out there: A reply to Bushman, Wang, and Anderson’s (2005) ‘Is the curve relating temperature to aggression linear or curvilinear?’ ”, Journal of Personality and Social Psychology, 89(1), 67-70.

[7] Cunningham, M.R., (1979). “Weather, Mood and Helping Behavior: Quasi Experiments with the sunshine Samaritan”, Journal of Personality and Social Psychology 1979, Vol. 37, No. 11, 1947-1956

[8] Digon, E., & Bock, H. (1966). “Suicides and climatology”, Archives of Environmental Health, 12, 279– 286.

[9] Faust, V., Weidmann, M. & Wehner, W., (1974). “The influence of meteorological factors on children and youths: A 10% random selection of 16,000 pupils and apprentices of Basle City (Switzerland)”. Acta Paedopsychiatrica: International Journal of Child & Adolescent Psychiatry, 40(4), 150-156.

[10] Gallant, R., Rossi, P., and Tauchen, G., (1992). “Stock prices and volume”, Review of Financial studies, 5, 1992-42.

[11] Gao, L., & Kling, G., (2005). “Calendar effects in Chinese stock market”. Annals of Economics and Finance, Vol. 6, No. 1, pp. 75-88, 2005.

[12] Goetzmann, W.N., &. Zhu, N., (2002). “Rain or Shine: Where is the Weather Effect?” Yale ICF Working Paper No. 02-27; EFA 2003 Annual Conference Paper No. 304.

(29)

from the ‘Investment Dartboard’ Column”. Journal of Finance, Volume 54, Issue 5, P1885 - 1899

[14] Hirshleifer, D., & Shumway, T., (2003). “Good day sunshine: Stock returns and the weather”, Journal of Finance 58 (2003) 1009–1032.

[15] Howarth, E., & Hoffman, M.S., (1984). “A multidimensional approach to the relationship between mood and weather”, British Journal of Psychology, 75(1), 15-23.

[16] Isen, A.M., Shalker, T.E., Clark, M., & Karp, L. (1978). “Affect, accessibility of material in memory, and behavior: A cognitive loop?” Journal of Personality and Social Psychology, 36, 1 –12.

[17] Jacobsen, B., & Marquering, W.A., (2007). “Is it the Weather?” Journal of Banking and Finance, 32 (2008) 526–540.

[18] Kamstra, M.J., Kramer, L.A., and Levi, M.D., (2003a). ‘Winter Blues: A SAD Stock Market Cycle”, American Economic Review, March 2003.

[19] Kelly, P.J., & Meschke, F., (2007). “Sentiment and Stock Returns: The Sad Anomaly Revisited”, working paper, available at SSRN: http://ssrn.com/abstract=571144

[20] Keller, M, C.; Fredrickson, Barbara L.; Ybarra, Oscar; Côté, Stéphane; Johnson, Kareem; Mikels, Joe; Conway, Anne; Wager, Tor; (2005). “A Warm Heart and a Clear Head: The Contingent Effects of Weather on Mood and Cognition”, Psychological Science, 16(9), 724-731.

[21] Kumar, A., (2008). “Determinants of Stock Trading Volume: Evidence from Indian Stock Markets”, working paper, available at SSRN: http://ssrn.com/abstract=947429.

[22] Lo, A., and Wang, J., (2000). “Trading Volume: Definitions, Data Analysis and Implications of Portfolio Theory”, Review of Financial Studies 13, 257–300.

[23] Loewenstein, G.F., (2000). “Emotions in economic theory and economic behavior”, American Economic Review, Vol. 65, 426– 432.

[24] Loewenstein, G.F., Weber, E.U., & Hsee, C.K., (2001). “Risk as Feelings”, Psychological Bulletin, Vol. 127, 2001.

[25] Matzarakis, A., & Amelung, B., (2008). “Physiologically Equivalent Temperature as indicator for impacts of climate change on thermal comfort of humans”. Springer-Sciences and Business Media, 161-172.

(30)

environments”, Theoretical and Applied Climatology 38:43–49

[27] Mehra, R., and Sah, R., (2000). “Mood, projection bias and equity market volatility”, Journal of Economic Dynamics & Control, 26 (2002) 869–887

[28] Molin, J., Mellerup, E., Bolwig, T., Scheike, T., and Dam, H., (1996). “The influence of climate on development of winter depression”, Journal of Affective Disorders, Vol. 37 pp.151-55.

[29] Nelson, S.A., (1902), “The ABC of Stock Speculation”, Nelson’s Wall Street Library, Vol. V, New York.

[30] Parker, P.M., & Tavassoli, N.T., (2000), “Homeostasis and Consumer Behavior Across Cultures,” The International Journal of Research in Marketing, 17 (1), 33-53.

[31] Pilcher, J.J., Eric, N., Busch, C., (2002). “Eﬀects of hot and cold temperature exposure on performance: Ameta-analytic review”. Ergonomics 45 (10), 682–698.

[32] Rind, B., (1996). “Effects of beliefs about weather conditions on tipping”, Journal of Applied Social Psychology, 26(2), 137– 147.

[33] Robotti, C., & Krivelyova, A., (2003). “Playing the Field: Geomagnetic Storms and the Stock Market ”, Federal Reserve Bank of Atlanta Working Paper, No. 2003-5b.

[34] Rosenthal, N., (1998), Winter Blues: Seasonal Affective Disorder: What It Is, and How to overcome it, 2nd edition, Gilford Press, New York, NY.

[35] Rotton, J. & Cohn, E.G.., (2004). “Outdoor Temperature, Climate Control and Criminal Assault: The Spatial and Temporal Ecology of Violence”, Environment and Behavior, 36(2), 276-306.

[36] Sanders, J.L., & Brizzolara, M.S., (1982). “Relationships between weather and mood”, Journal of General Psychology, 107(1), 155-156.

[37] Saunders, E.M., (1993). “Stock Prices and Wall Street Weather”, American Economic Review, December 1993, 83(5) pp. 1337-1345.

[38] Seppanen, O., Fisk, W.J., & Lei, Q.H., (2006). “Effect of temperature on task performance in office environment”, Lawrence Berkeley National Laboratory. Paper LBNL-60946.

(31)

[40] Steadman, R.G.., (1979). “The assessment of sultriness Part I. A temperature-humidity index based on human physiology and clothing science”, Journal of Applied Meteorology 18:861–873.

[41] Tang, Y., Feng, Z.M., & Yang, Y.Z., (2008). “Establishment and Application of Human Settlements Environment Index Model (HEI) Based on GIS”, ACTA GEOGRAPHICA SINICA, Vol.63, No.12, December 2008.

[42] Terjung, W.H., (1968). “World Patterns of the Distribution of the Monthly Comfort Index”, International. Journal of Biometeorology, 1968, Vol. 12 number 2, pp. 119-151. [43] Thom, E.C., (1959). “The discomfort Index”. Weatherwise vol. 12, pp. 57–60.

[44] Wyndham, H.C., (1969). “Adaptation to heat and cold”, Environmental Research vol.2, pp. 442–469.

[45] Ying, C., (1966). “Stock market prices and volume of sales”, Econometrica, vol.34, pp. 676-686.

[46] Young, M.A., Meaden, P., Fogg, L., Cherin, E., Eastman, C., (1997). “Which environmental variables are related to the onset of seasonal affective disorder?” Journal of Abnormal Psychology, Vol. 106, pp.554-62.

[47] Zhang, Z.C., Wang, H., & Sun, W., (2007). “Financial Anomalies in Emerging Markets: The Case of China”, Euromed Marseille Working Paper No. 13.

(32)

VII. Appendix

Table (1) Description of the degree of cloud cover

Oktas Definition Category

0 Sky clear Fine

1 1/8 of sky covered or less, but not zero Fine

2 2/8 of sky covered Fine

3 3/8 of sky covered Partly Cloudy

6 6/8 of sky covered Cloudy

7 7/8 of sky covered or more, but not 8/8 Cloudy 8 8/8 of sky completely covered, no breaks Overcast

(Source: http://www.worldweather.org/oktas.htm)

Table (2) Description of Human Comfort Index

Grade Comfort Index Description

1 0 – 25 Extreme cold stress, highly uncomfortable

2 26 – 30 Strong cold stress

3 31 – 50 Slightly cold, quite comfortable

4 51 – 58 Cool, comfortable

5 59 – 70 Most comfortable

6 71 – 79 Slight hot, quite comfortable

7 80 – 85 Minor heat stress

8 86 – 89 Strong heat stress

9 ≥90 Extremely hot, highly

(33)

Figure (1) Pattern of local weather – Shanghai

Human Comfort Index

0 10 20 30 40 50 60 70 80 90 100 5- 1-19 98 5- 7-19 98 5- 1-19 99 5- 7-19 99 5- 1-20 00 5- 7-20 00 5- 1-20 01 5- 7-20 01 5- 1-20 02 5- 7-20 02 5- 1-20 03 5- 7-20 03 5- 1-20 04 5- 7-20 04 5- 1-20 05 5- 7-20 05 5- 1-20 06 5- 7-20 06 5- 1-20 07 5- 7-20 07 Date HCI

Cloud Cover Index

(34)

Figure (2) Pattern of local weather- Shenzhen Cloud Cover Index_Shenzhen

0 1 2 3 4 5 6 7 8 9

5-Ja

n-98

5-Ju

l-98

5-Ja

n-99

5-Ju

l-99

5-Ja

n-00

5-Ju

l-00

5-Ja

n-01

5-Ju

l-01

5-Ja

n-02

5-Ju

l-02

5-Ja

n-03

5-Ju

l-03

5-Ja

n-04

5-Ju

l-04

5-Ja

n-05

5-Ju

l-05

5-Ja

n-06

5-Ju

l-06

5-Ja

n-07

5-Ju

l-07

Date Oktus

Human Comfort Index_Shenzhen

0 10 20 30 40 50 60 70 80 90 100 5-Ja n-98 5-Ja n-99 5-Ja n-00 5-Ja n-01 5-Ja n-02 5-Ja n-03 5-Ja n-04 5-Ja n-05 5-Ja n-06 5-Ja n-07 Date HC degrees

Total daily sunshine_Shenzhen

(35)

Table (3) Descriptive statistics of data

Part A: Shanghai Stock Exchange, A-share index (January 1998 – December 2007)

RETURN Volume (Billion RMB) SK (Oktas) HC AD (hours) Mean 0.000766 19.12528 4.723549 62.61945 4.836263 Median 0.000400 8.243359 5.000000 64.00000 5.200000 Maximum 0.396200 267.4522 9.000000 86.00000 12.80000 Minimum -0.084200 1.380998 0.000000 31.00000 0.000000 Std. Dev. 0.017031 34.38227 2.932507 12.99149 3.939334 Skewness 5.519622 3.781493 -0.329958 -0.157911 0.081357 Kurtosis 128.6839 17.98974 1.613693 1.835113 1.580109 Jarque-Bera 1554689. 27531.36 230.2332 142.2716 199.4907 Probability 0.000000 0.000000 0.000000 0.000000 0.000000 Observations 2344 2344 2344 2344 2344

Part B: Shenzhen Stock Exchange, A-share index (January 1998 –December 2007)

RETURN Volume (billion RMB) SK (Oktas) HC AD (hours) Mean 0.000731 13.16163 4.719517 72.03620 4.340824 Median 0.000209 6.195531 6.000000 74.00000 4.000000 Maximum 0.099988 140.8332 8.000000 86.00000 11.90000 Minimum -0.091161 1.121487 0.000000 44.00000 0.000000 Std. Dev. 0.016646 19.96262 2.782944 8.210740 3.599087 Skewness 0.331238 3.248943 -0.467339 -0.683383 0.184402 Kurtosis 7.033983 13.96107 1.758343 2.662778 1.580619 Jarque-Bera 1673.278 16257.06 241.8354 198.4248 215.3347 Probability 0.000000 0.000000 0.000000 0.000000 0.000000 Observations 2403 2403 2403 2403 2403 Notes:

(36)

Table (4)

Relation between Cloud cover and stock market returns/trading volumes – overall correlations Part A: Shanghai SK 0 1 2 3 4 5 6 7 8 return 0,0004 -0,0011 0,0013 0,0005 -0,0001 0,0015 -0,0012 0,0015 0,0017 +return 51,19% 46,94% 52,22% 50,82% 49,29% 55,68% 46,67% 57,68% 53,61% volume 1,8942 1,9168 2,0169 2,0154 2,0167 2,1691 1,9929 1,9897 2,1286 Vol.rmb 16,7271 17,9467 20,6437 20,1224 18,5051 21,1344 18,4829 18,1539 20,0081 mean score Z (8,1) = 1,1175 Part B: Shenzhen SK 0 1 2 3 4 5 6 7 8 return 0,0008 0,0008 0,0008 0,0008 0,0008 0,0008 0,0008 0,0008 0,0008 +return 51,70% 55,73% 54,24% 51,72% 49,48% 54,30% 46,70% 48,87% 50,41% volume 1,5201 1,7654 1,7591 2,0857 1,7866 2,1423 1,8617 1,4148 1,4195 Vol.rmb 11,9259 13,8489 13,3639 17,7086 14,7821 18,3367 15,7808 11,3650 10,0323 mean score Z (8,1) = -0,0626 Notes:

(1) This table reports bin-test results for the sample of A-shares and cloud cover index. The sample covers the period of 1998–2007 with 2344 observations for SSE and 2403 observations for SZSE. We report the mean return, the percentage of positive returns and trading volumes (in number of shares or RMB value) for each bin. .

(2) 1 2 1 2 1 1 , / / _ n n score Z k k k mean k       

(37)

Table (5) Market return, Trading Volume and Sky Condition

Panel A: Stock Returns

Shanghai Shenzhen

Coefficient T-statistics Coefficient T-statistics

Intercept -0.0002 -0.2287 0.0018 2.6177

Return(t-1) 0.0355 1.7195 * 0.0265 1.2999

March 0.0002 0.1940 0.0010 0.8124

SK 0.0002 1.5200 -0.0002 -1.9680*

R-squre 0.0023 0.0025

Panel B: Trading Volume

Intercept 1.6703 17.606 1.269275 23.622 ΔP 3.8569 3.2813*** 0.000289 1.2444 PMax 3.3937 18.328*** 2.734961 23.797*** PMin -0.8184 -2.4627** -0.660465 -4.1031*** SK 0.0193 1.1482 -0.011775 -1.2244 R-square 0.1528 0.2222 Note:

(38)

Table (6)

Relation between Human comfort and stock market returns/volumes – overall correlations Part A: Shanghai HC 3 4 5 6 7 8 return 0,0012 0,0012 0,0005 0,0001 0,001 0,0006 +return 48,89% 52,02% 51,88% 49,50% 52,31% 52,67% volume 1,2405 1,9890 1,6711 2,0548 2,2087 2,4329 Vol.rmb 8,5640 15,4974 15,3064 19,3309 23,0731 24,5837 mean score Z (8,3) = -0,3083 Part B: Shenzhen HC 3 4 5 6 7 8 return 0,0017 0,0018 0,0015 -0,0001 0,0010 +return 58,33% 52,36% 53,85% 47,59% 52,79% volume 0,8196 1,5989 1,6176 1,6489 1,7379 Vol.rmb 5,1618 10,6938 12,0305 13,6687 14,8720 mean score Z (8,4) = -0,1815 Notes:

(1) This table reports bin-test results for the sample of A-shares and human comfort index. The sample covers the period of 1998–2007 with 2344 observations for SSE and 2403 observations for SZSE. We report the mean return, the percentage of positive returns and trading volumes (in number of shares or RMB value) for each bin. .

(39)

Table (7) Market Return, Trading Volume and Human Comfort

Panel A: Stock Returns

Intercept 0.0017 0.9225 0.0046 1.4912

Return(t-1) 0.0359 1.7360* 0.0256 1.2552

March 0.0001 0.0884 0.0004 0.3206

HC -1.54E-05 -0.5488 -5.52E-05 -1.2952

R-square 0.0015 0.0016

Panel B: Trading Volume

Intercept 1.1630 4.8009 0.6749 2.8571 ΔP 3.9341 3.3522*** 0.0003 1.2547 PMax 3.3644 18.156*** 2.7437 23.897*** PMin -0.8068 -2.4303*** -0.6673 -4.1484*** HC 0.0096 2.5295*** 0.0075 2.29478** R-square 0.1546 0.2235 Note:

(1) This table reports regression analysis for A-shares and the Human Comfort index across markets using daily sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(40)

Table (8)

Relation between adjusted daylight and stock market returns/volumes – overall correlations Part A: Shanghai AD 0~2 2~4 4~6 6~8 8~10 10~12 return 0,0008 0,0030 0,0000 0,0014 0,0000 -0,0005 +return 52,75% 52,02% 49,42% 52,67% 50,73% 48,88% volume 2,1476 2,1761 2,0639 1,8547 1,9715 1,8800 Vol.rmb 20,4992 20,3948 20,3406 15,6476 19,3684 17,9926 mean score Z = -1,2154 Part B: Shenzhen AD 0~2 2~4 4~6 6~8 8~10 10~12 return 0,0006 0,0021 -0,0004 0,0003 0,0009 0,0017 +return 49,83% 55,06% 45,99% 48,42% 53,63% 56,19% volume 1,7089 1,6207 1,6431 1,7640 1,5502 1,3705 Vol.rmb 12,9048 12,9706 13,5697 14,9613 12,5675 12,6244 mean score Z = 0,7121 Notes:

(1) This table reports bin-test results for the sample of A-shares and adjusted dayligh index. The sample covers the period of 1998–2007 with 2344 observations for SSE and 2403 observations for SZSE. We report the mean return, the percentage of positive returns and trading volumes (in number of shares or RMB value) for each bin. .

(41)

Table (9) Regression of stock return/trading volume and adjusted daylight

Panel A: Stock Returns and Weather

Intercept 0.0011 1.8708 0.0005 0.8288

Return(t-1) 0.0357 1.7279* 0.0261 1.2813

March 0.0003 0.2165 0.0008 0.7114

AD -7.39E-05 -0.8283 3.55E-05 0.3706

R-squre 0.0016 0.0009

Panel B: Trading Volume and Weather

Intercept 1.9075 24.035 1.2618 29.399 ΔP 3.8660 3.2918*** 0.0003 1.2184 PMax 3.4291 18.562*** 2.7334 23.660*** PMin -0.8079 -2.4321*** -0.6490 -4.0079*** AD -0.0304 -1.4411 -0.0103 -1.3796 R-square 0.1567 0.2198 Note:

(1) This table reports regression analysis for A-shares and the adjusted daylight index across markets using daily sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(42)

Table (10) Regression of stock return/trading volume and all weather variables Daily data

Intercept 0.0007 0.3848 0.0055 1.7362 Return(t-1) 0.0355 1.7166* 0.0253 1.2408 March -1.45E-05 -0.0111 0.0005 0.4364 SK 0.0002 1.4197 -0.0003 -1.5033 HC -2.14E-05 -0.7183 -4.29E-05 -0.9811 AD 4.98E-05 0.4034 -8.34E-05 -0.6914 R-square 0.0025 0.0032

Intercept 1.3229 5.0539 0.7735 3.2342 ΔP 3.9393 3.3617*** 0.0003 1.2105 PMax 3.3678 18.213*** 2.7245 23.751*** PMin -0.7834 -2.3643** -0.6628 -4.1254*** SK -0.0315 -1.4022 -0.0357 -2.9734*** HC 0.0138 3.4353*** 0.0103 3.0628*** AD -0.0565 -3.2955*** -0.0302 -3.2021*** R-square 0.1590 0.2273 Note:

(1) This table reports regression analysis for A-shares and all three weather variables (sky coverage, human comfort and adjusted daylight) across markets using daily sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(43)

Table (11) Regression of stock return/trading volume and all weather variables (B-share, daily data, Shanghai Stock Exchange)

Before 19-02-2001 After19-02-2001

Intercept -0.0035 -0.6025 0.0025 0.8485 Return(t-1) 0.07450 2.0324** 0.1205 4.8219*** March 0.0002 0.0573 0.0025 1.2282 SK 0.0007 1.4238 0.0003 1.2916 HC -1.76E-05 -0.1978 -6.64E-05 -1.4457 AD 0.0004 1.2495 0.0001 0.6447 R-squre 0.0088 0.0196

Before19-02-2001 After19-02-2001

Intercept -79.913 -0.3941 124.08 0.9534 ΔP -4.6200 -0.1566 -1.7362 -0.2703 PMax 470.38 2.8884*** 1696.2 17.205*** PMin -245.17 -1.2860 -344.84 -2.7565*** SK 12.122 0.6973 23.079 2.1019** HC 4.6115 1.4682 5.1628 2.5997*** AD 5.2005 0.3999 6.3890 0.7545 R-square 0.0230 0.1765 Note:

(1) This table reports the regression analysis for B-shares on SSE and all three weather variables (sky coverage, human comfort and adjusted daylight) across markets using daily sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(44)

Table (12) Regression of stock return/trading volume and all weather variables (B-share, daily data, Shenzhen Stock Exchange)

Panel A: Stock Returns and Weather Before 19-02-2001 After19-02-2001

Intercept -0.0041 -0.4714 0.0073 1.4925 Return(t-1) 0.0897 2.4517** 0.0743 3.0149*** March 0.0021 0.6190 0.0051 2.7400** SK -0.0002 -0.5387 -0.0003 -1.1647 HC 8.01E-05 0.6785 -8.12E-05 -1.1904 AD -3.95E-05 -0.1168 0.0001 0.6161 R-squre 0.0097 0.0148

Panel B: Trading Volume and Weather Before19-02-2001 After19-02-2001

Intercept -25.886 -3.7683 53.849 2.8918 ΔP 0.1450 2.8777*** 0.0028 0.0787 PMax 44.112 8.6232*** 104.10 12.650*** PMin -11.830 -2.7454*** -31.200 -1.2462 SK 0.3480 0.9277 0.7293 0.8019 HC 0.5668 5.8962*** 0.1725 1.6601* AD 0.4999 1.8247* -1.6124 -2.2201* R-square 0.1995 0.1009 Note:

(1) This table reports the regression analysis for B-shares on SZSE and all three weather variables (sky coverage, human comfort and adjusted daylight) across markets using daily sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(45)

Table (13) Regression of trading volume of B-shares on SZSE and the Human Comfort index at Hong Kong, during the period from 1998 to 2007

Trading Volume

Before 19-02-2001 After19-02-2001

Intercept -22.453 -3.5075 62.627 3.7318 ΔP 0.1460 2.8978*** 0.0017 0.0476 PMax 44.098 8.6117*** 101.73 12.394*** PMin -11.358 -2.6367*** -34.323 -1.3700 HC 0.5484 6.5116*** 0.0006 0.0027 R-square 0.1953 0.0942 Note:

(1) This table reports the regression for trading volume of B-shares on SZSE and human comfort index at Hong Kong city using daily sample. We control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.

(46)

Table (14) Regression of stock return/trading volume and all weather variables A-share, Weekly data

Intercept 0.0084 0.7502 0.0300 1.8110 Return(t-1) 0.0444 0.9497 0.0128 0.2757 March 0.0020 0.3025 0.0009 0.1419 SK 0.0005 0.3258 -0.0017 -1.3723 HC -5.04E-05 -0.2983 -0.0002 -0.7924 AD -0.0009 -0.7462 -0.0011 -1.1363 R-square 0.0100 0.0093

Intercept 9.3579 2.9156 2.4809 0.9827 ΔP 0.0232 3.9561 0.0042 3.5089*** PMax 12.219 7.8966 10.255 11.891*** PMin -2.2628 -1.0254 -2.0076 -1.8815* SK -0.8137 -1.9171* -0.4937 -2.4774** HC 0.1096 2.3606** 0.0969 2.5753** AD -0.8940 -2.7306*** -0.4005 -2.5900** R-square 0.2295 0.3694 Note:

(1) This table reports regression analysis for A-shares and all three weather variables (sky coverage, human comfort and adjusted daylight) across markets using weekly sample. Panel A reports the result of regression on stock return, we control for lagged return variable (R (t-1)) and the March effect (Mar); panel B reports the results of regression on trading volume, we control for price change ΔP and include two dummy variables PMax and PMin to account for the impact of past price on current trading volume.