Transit Stud Rev (2013) 19:291–312 DOI 10.1007/s11300-012-0234-6 WORLD TRANSITION ECONOMY RESEARCH

Does Overconfidence Bias Explain Volatility During the Global Financial Crisis? Mouna Boujelbe`ne Abbes

Received: 11 February 2012 / Accepted: 30 July 2012 / Published online: 17 August 2012 Ó Springer-Verlag 2012

Abstract This paper explores the problem of the current global financial crisis, using a behavioral perspective. Particularly, the main objective of this paper is to test whether overconfidence bias can explain excessive volatility witnessed during global financial crisis in developed and emerging equity markets. Empirical results of EGARCH estimated models show an asymmetric effect of volatility for all equity market indexes. The relation between excessive trading volume of overconfident investors and excessive prices volatility is then estimated. The results indicate that conditional volatility is positively related to trading volume caused by overconfidence bias. This finding provides strong statistical support to the presence of overconfidence bias among investors in developed and emerging stocks markets. This cognitive bias contributes to the exceptional financial instability that erupted in 2008. However, during the subprime financial crisis period overconfidence bias cannot explain volatility because of the loss of confidence by investors in financial markets. Keywords Global financial crisis  Overconfidence  Behavioral finance  Volatility  EGARCH JEL Classifications

G01  G12  G15

Introduction The current global financial crisis following the advent of the subprime mortgage crisis in the United States is one of the most serious and dramatic international M. B. Abbes (&) Unit of Research in Applied Economics (UREA), Faculty of Economics and Management of Sfax, University of Sfax, Sfax, Tunisia e-mail: [email protected]

123

292

M. B. Abbes

financial crises of recent decades. The shock on the United States financial market was a starting point of severe financial turbulences. More notably, the current crisis which quickly spread into other market segments and countries is already seen today as one of the biggest financial crises in history (Ackermann 2008). Several studies try to answer to an essential question: What caused the current crisis? Many answers to this question circulating in the media. Obviously, there were many factors that contribute to this financial crisis such as the excessive use of credit, in particular in the housing market. The reasons directly related to financial institution liquidation such as the excessive leverage seems to be the most important. Overall, most of the advanced explanations ignore the behavioral elements of the crisis. Behavioral finance provides alternative explanation about global financial crisis. This explanation may be derived from psychology of judgment and choice such as the impact of overconfidence cognitive bias on the rationality of financial decision makers. Several studies show that investors do tend to have periods of irrational exuberance or excessive optimism that pervade financial markets. Overconfidence bias leads investors to be too certain of their views, a tendency that frequently results in their underestimating risk. Indeed, one of the main causes of this financial crisis is the excessive liquidity which leads to the enormous credit expansion. Having short memory investors become irrationally overconfident that liquid market would continue indefinitely (Shefrin 2009). Several studies suggest that overconfidence constitutes an important reason for excessive price volatility. Benos (1998) proposes a model in which the aggressive exploitation of overconfident traders’ profitable information, jointly with conservative trading strategy of rational traders, conducts prices to vary strongly in one or the other direction. In their model, Daniel et al. (1998) show that overconfident investors increases prices volatility at the time reception of private signals. The originality of this study is to investigate whether overconfidence bias can explain excessive volatility witnessed during global financial crisis in developed and emerging financial markets. We consider a large set of country including developed markets (US, Canada, France, United Kingdom, Swiss, Australia, Hong Kong, Japan) and emerging markets (Brazil, Mexico, Korea, Malaysia, Singapore, India and Kuwait) because, unlike previous financial crises, the US subprime crisis highly influences developed as well as emerging markets. The methodology here considers various empirical frameworks. First, this study examines the behavior of the index prices and volatilities for all sample period and during subprime crisis period. Then, this study employs an EGARCH model to study the leverage effect. Finally, this study estimates the conditional variance of EGARCH model by introducing two components of trading volume. The first component, due to past stock returns, is relating to investors’ overconfidence. The second component is unrelated to investors’ overconfidence. Following this introduction, Sect. 2 presents the literature review. Section 3 describes empirical data. Section 4 studies the behavior of index prices and volatilities. Section 5 examines the effect of overconfidence bias on market volatility. Section 6 provides concluding remarks.

123

Does Overconfidence Bias Explain Volatility?

293

Literature Review The US and world economies are in the midst of a severe financial crisis. The magnitude and complexity of this crisis arouses the researchers’ curiosity to understand the causes as well as the repercussions of this crisis on developed and emerging equity markets. So¨hnke and Gordon (2009) suggest that all indices fall about 30–40 % in the period of mid-September to the end of October 2008. Anders (2010) finds that the spread of the subprime crisis affects European markets more than the Asian markets. Furthermore, the high level of comovements during times of international financial turmoil demonstrates the limited benefit of diversification in regional portfolios. Understanding the causes of the current global financial crisis requires a combination of approaches. One approach considers the problem inherent to the existing financial structure. The other approach called ‘behavioral finance’ considers the ‘irrational exuberance’ of market makers. Under the first approach, the fundamental cause of the subprime financial crisis in United States is the excessive use of credit, in particular in the housing market. Mortgage lenders (banks) made it easier to lend money to people with poor credit ratings under the assumption that housing prices would continue to appreciate in value. Mortgages no longer become affordable for many homeowners, and US property prices fall. Therefore, many banks have collapsed and stopped lending to each other, which substantially dry up liquidity (Berger and Bouwman 2008). Another factor which contributes to the crisis appearance is the excessive leverage of financial institutions. In fact, the use of new types of mortgages and the excessive investment of financial firms in derivative such as futures and credit default swaps increase their leverage and create a risk of bankruptcy. Since many of the assets were tied to mortgages, the fall in US property prices makes equity becomes worthless and the firms drop precipitously in value and default on its debt which creates a crunch credit. This crisis had grown into a serious slump of stock prices. As a consequence, in many developed and emerging markets equity prices have fallen drastically; major stock indexes have lost nearly a third of their values which generate a structural change in the capital markets volatility (Abdelhe´di et al. 2011). Under behavioural finance several aspects of irrationality on investor behavior may have played a role in setting the conditions for the appearance of the current global financial crisis. Shefrin (2009) suggests that ‘‘the root cause of the financial crisis that erupted in 2008 is psychological’’. Using five specific cases [(1) UBS, a bank; (2) standard & poor’s (S&P), a rating firm; (3) American International Group (AIG), an insurance company; (4) the investment committee for the town of Narvik, Norway, an institutional investor; and (5) the U.S. SEC, a regulatory agency), this study explains how psychological pitfalls affect judgments and decisions at various points along the supply chain for financial products, particularly home mortgages, in the crisis. Richard (2009) discusses how overconfidence bias can causes financial crisis. This cognitive bias leads investors to overestimate their performance and the

123

294

M. B. Abbes

precision of their beliefs. The author suggests that ‘‘the good news of the years up to 2006 may have led to the bad news that started in 2007’’ (Richard 2009). Indeed, performance reinforces the decisions of investors to buy structured securities and leads them to rely more on credit ratings.

Data and Descriptive Statistics Data The sample consists of market indexes of 15 countries which are United States, Canada, Brazil, Mexico, France, United Kingdom, Swiss, Australia, Hong Kong, Japan, Korea, Malaysia, Singapore, India, and Kuwait. This empirical study uses price and trading volume data of SP 500, S&P TSX Composite, Bovespa, IPC, CAC40, FTSE 100, Swiss Market Index, All Ordinaries, Hang Seng, Nikkei 225, Seoul Composite, KLSE Composite, Strait Times, BSE 30 and Kuwait Market Index. Using FTSE Group and Morgan Stanley Capital International criteria, this study classifies these countries into emerging and developed markets. Developed markets are US, Canada, France, United Kingdom, Swiss, Australia, Hong Kong and Japan. Emerging markets are Brazil, Mexico, Korea, Malaysia, Singapore, India and Kuwait. The choice of these countries allows us to examine the effect of global financial crisis in different regions. The data were drawn from DataStream for all market indexes except for Kuwait that we use data available in Kuwait Stock Exchange. This research uses monthly data for drawing index prices and volatilities figures and daily data for estimating empirical models. Most of the closing prices of indexes data cover the period between January 1999 and December 2009 except for Bovespa, S&P TSX Composite and Kuwait Market Index due to the unavailability of data. Brazil and Canada data cover the period between January 2000 and December 2009 and Kuwait data cover the period between Jun 2001 and December 2009. Descriptive Statistics Tables 1 and 2 present summary statistics on daily index returns and trading volume respectively for developed market and emerging markets. Results indicate that for developed market, the average returns are negative particularly for United States, France, United Kingdom, Swiss and Japan and it is positive but very low for the other financial markets. Also, the SP 500 index has the larger average transaction volume. For emerging market, Kuwait Market Index has the highest average returns and transaction volume. The normality test results show that the return and the transaction volume distributions are not normal. Indeed, the skewness coefficients of the return distribution are different from zero for all indexes. This reflects the asymmetry of returns. The high frequency of large negative returns compared with large positive returns can explain this result. Similarly, the coefficients of kurtosis are largely

123

USA SP 500

0.01

0.08

11.55

7,700.21

SD

Skewness

Kurtosis

Jarque–Bera

9,331.32

12.83

-0.56

0.01

-0.09

0.09

0.0002

9.2 9 10-5

Canada S&P TSX composite

1,992.70

Jarque–Bera

33,839.18

20.94

2.22

1.12E?09

3,430.87

8.83

1.78

31,585,195

12,574,800

3.14E?08

62,130,200

69,171,076

4,421.96

9.45

0.23

0.01

-0.078

0.113

0.0004

-4.8 9 10-5

Swiss Swiss market index

229.2566

3.49

0.72

75,493,330

448,516

5.31E?08

83,428,000

76,848,584

3,541.69

8.74

0.18

0.01

-0.09

0.11

0.0002

-2.84 9 10-5

France CAC40

50.53

3.04

-0.41

3.86E?08

1.38E?08

2.15E?09

1.55E?09

1.53E?09

4,637.98

9.62

0.05

0.013

-0.088

0.09

0.0002

-6.04 9 10-5

United Kingdom FTSE 100

27,823.72

22.05

2.55

3.69E?08

5,952,400

5.64E?09

7.00E?08

7.91E?08

7,828.65

11.47

-0.66

0.01

-0.08

0.05

0.0004

0.0001

Australia All ordinaries

2,136.07

6.70

1.67

1.08E?09

63,873,400

9.80E?09

3.90E?08

9.94E?08

7,461.73

11.46

0.19

0.017

-0.13

0.143

0.0003

0.0002

Hong Kong Hang seng

90.20

3.64

0.58

38,335.15

29,800.00

302,000

126,600

129,162.9

4,872.55

9.88

-0.13

0.01

-0.11

0.14

9.25 9 10-6

-9.40 9 10-5

Japan Nikkei 225

This table presents market descriptive statistics for daily returns (panel A) and daily trading volume (panel B): mean, median, minimum, maximum, standard deviation (SD), skewness, Kurtosis and Jarque–Bera. Sample periods is January 1999 to December 2009 for US, France, United Kingdom, Swiss, Australia, Hong Kong and Japan; January 2000 to January 2009 for Canada

5.39

Kurtosis

71,189,671

1.68E?09

1.65

SD

Skewness

9,473,300

1.15E?10

2.47E?08

Maximum

Minimum

1.57E?08

1.41E?08

2.24E?09

1.53E?09

Mean

Median

Panel B: market index trading volume

0.11

-0.09

Minimum

0.0003

Median

Maximum

-6.30 9 10-5

Mean

Panel A: market index returns

Countries Indexes

Table 1 Summary statistics of returns and trading volume of developed markets

Does Overconfidence Bias Explain Volatility? 295

123

123

Brazil Bovespa

7.10

1,619.64

Kurtosis

Jarque–Bera

2.32E?08

112,200

53,090,016

1.43

3.79

292.85

Maximum

Minimum

SD

Skewness

Kurtosis

Jarque–Bera

1,558.87

6.35

1.24

57,357,590

4,037,500

5.38E?08

1.02E?08

1.10E?08

1,645.45

6.92

0.25

0.01

-0.08

0.11

0.0010

0.0007

Mexico IPC

12,367

107.26

6.69

17,012,235

136,400

3.64E?08

489,800

7,857,988

1,221.88

6.41

-0.25

0.02

-0.12

0.12

0.0012

0.0004

Korea Seoul composite

11,229.61

11.35

2.51

1.01E?08

4,734,500

8.17E?08

71,353,600

1.07E?08

4,612.80

9.65

-0.34

0.01

-0.09

0.06

0.0003

0.0002

Malaysia KLSE composite

1,232.41

4.08

1.56

2.59E?08

7,237,491

1.82E?09

1.46E?08

2.53E?08

2,716.36

8.04

-0.31

0.014

-0.09

0.08

0.0003

0.0001

Singapore Strait times

146,653.6

49.28

4.58

14,492

800.0,000

217,600

21,800

24,852.62

1,339.39

6.55

-0.26

0.02

-0.11

0.0897

0.0012

0.0005

India BSE 30

11,058

351.77

17.93

8.55E?08

4,970,000

2.14E?10

1.84E?08

2.68E?08

1,414.94

7.04

-0.55

0.01

-0.05

0.05

0.0010

0.0007

Kuwait Kuwait market index

This table presents market descriptive statistics for daily returns (panel A) and daily trading volume (panel B): mean, median, minimum, maximum, standard deviation (std.dev.), skewness, Kurtosis and Jarque–Bera. Sample periods is January 1999 to December 2009 for Mexico, Korea, Malaysia, Singapore and India; January 2000 to January 2009 for Brazil and June 2001 to January 2009 for Kuwait

33,592,350

2,108,300

Mean

Median

Panel B: market index trading volume

0.02

-0.11

Minimum

0.06

0.14

Maximum

Skewness

0.0002

Median

SD

0.0005

Mean

Panel A: market index returns

Countries Indexes

Table 2 Summary statistics of returns and trading volume of emerging markets

296 M. B. Abbes

Does Overconfidence Bias Explain Volatility?

297

higher than three which confirm the high occurrence of extreme values. Consequently, the empirical distribution of all returns series is leptokurtic. This finding is consistent with Leo´n et al. (2004) validation. These authors suggest that stock return distributions exhibit excess kurtosis, which means that the market gives higher probability to extreme observations than in normal distribution. About the trading volume, all indexes have a skewness coefficients different from zero and kurtosis coefficients different from three. So, the Jarque–Bera test rejects the normality of the return and trading volume series.

Impact of Subprime Crisis on Index Prices and Volatilities This section focuses both on the behavior of index prices and volatilities during the subprime crisis period. A conditional GARCH variance constitutes a volatility measure. Index Prices and Volatilities Analysis of Developed and Emerging Markets: 1999–2009 Figure 1a, b illustrates the time path of the 15 monthly index prices from January of 1999 until December 2009 respectively for developed markets and emerging markets. Both developed markets and emerging markets index prices follow the same downward trend as the U.S. index. After mid-2007, price indexes of all markets lost about a third of their value. Such pronounced drops in stock market indexes prices are the typical result expected during this crisis (see Appendix Fig. 5). As consequence, the financial crisis originated in the US financial market spread quickly across several parts of financial markets in unanticipated ways, inflicting a sharp decrease of index prices. Figure 2a, b plot the index volatilities during 1999–2009 period respectively for developed markets and emerging markets. Figures indicate that the current financial crisis dramatically influenced the market volatility which has been high during mid 2007–2009, particularly during the 2008 period. This finding supports the argument of Black (1976) that stock volatility increases after stock prices fall. Volatility presents a peak during turmoil period higher for developed market than for emerging markets (see Appendix Fig. 6). Consequently, global financial crisis affects more developed markets, because of their highest correlation with US market (Claude et al. 1998; Chukwuogor 2008). Also, Figs. 1 and 2 present respectively the behavior of the volatilities and prices during Asian financial crisis. Stock markets of Korea, Malaysia, Hong Kong and Singapore present high level of volatility in 1999 which reflects persistence of 1997 East Asian crisis. Indeed, this finding supports the results of Kausik and Franc (2001) These authors examine whether the surge in volatility during the 1997–1998 Asian crisis has an effect on stock return volatility in the year 2000. Using a regimeswitching GARCH Model they find that in June 2000 most East Asian stock markets were still in the high-volatility regime initiated by the 1997–1998 crisis.

123

298

Fig. 1 Index prices 1999–2009, a developed stock markets, b emerging stock markets

123

M. B. Abbes

Does Overconfidence Bias Explain Volatility?

299

Fig. 2 Index volatilities 1999–2009, a developed stock markets, b emerging stock markets

Index Prices and Volatilities Analysis Through the Global Financial Crisis Period: 2007–2009 To further examine the current financial crisis, this subsection analyses the index prices and volatilities over the period January 2007 to December 2009 for developed and emerging markets. Figures 3 and 4 plot respectively the index prices and volatilities. The movement of these indexes shows that the trend can be divided into diverse phases. From

123

300

M. B. Abbes

January 2007 to mid-July 2007, figures indicate harmonized movements in the stock market indexes till July 2007. From mid-July 2007 to mid-January 2008 the American and European index prices have a significantly sharp decrease which is caused by the loss of confidence by investors in the value of securitized mortgages in the United States. While, during this period an increasing of price has been sawing by Asian markets. The period covering the end of January 2008 to the end of January 2009 is characterised by a decreasing of index prices and an increasing of volatilities. Indeed, the crisis spears, developed and emerging stock markets were affected and enter in a period of high volatility. The period of February 2009 to December 2009 is characterized by an improvement of the financial situation because of the corrective measures taken by governments and financial authorities in the most of countries. Indeed, Figures 3 and 4 show respectively an increasing of prices coupled with a decreasing of volatilities for all equity markets.

Fig. 3 Index prices 2007–2009, a developed stock markets, b emerging stock markets

123

Does Overconfidence Bias Explain Volatility?

301

Fig. 4 Index volatilities 2007–2009, a developed stock markets, b emerging stock markets

Impact of Overconfidence Bias on Market Volatility Volatility Asymmetry To study empirically the asymmetry observed on the indexes price volatility, this subsection estimates an EGARCH model taking into account an asymmetric effect in which a negative return shock increases volatility more than does a positive return shock (leverage effect). The EGARCH model (Nelson 1991) is as follows: rt ¼ lt þ gt

ð1Þ

gt =gt1 ; gt2 ; . . .; rt1 ; rt2 ; . . .;  GEDð0; ht Þ   jgt1 j þ jgt1 p ffiffiffiffiffiffiffiffi ln ht ¼ x þ f1 þ f2 ht1 ht1 where lt and ht represent respectively the expected return and conditional volatility. The volatility parameter, j, represents asymmetric effect in EGARCH model. If j \0, then conditional volatility tend to augment (to reduce) when the standardized residual is negative (positive). To let for the possibility of non-normality of the

123

302

M. B. Abbes

returns distribution, this study supposes that the conditional errors of EGARCH model pursue a Generalized Error Distribution, GED. This GARCH-type specification permits stocks returns into expected and unexpected returns.  decomposing   f1 j The statistic AD ¼ f1 þj measures the asymmetry degree. Table 3 presents the estimated coefficients of the EGARCH model. Panel A concerns developed markets and Panel B is relative to emerging markets. The significance of f1 for all indexes indicates that conditional volatility responds substantially to innovations shocks. A positive and significant f2 indicates that the volatility has a long memory. The asymmetric effect captured by the parameter estimate j is negative and significant suggesting the presence of a leverage effect. Thus, the volatility increases following  a previous drop in stock returns. Also, Table 3 shows that the asymmetry   degree ff11 j þj is more than one for all financial markets. This result reflects that developed as well as emerging markets are more sensitive to bad news than to good news despite the occurrence of the global financial crisis. Australia and Brazil present the highest asymmetry degree respectively for developed and emerging markets. Overconfidence and Volatility Several studies consider the proposition that investor overconfidence generates the high trading volume observed in financial markets (Odean 1998; Gervais and Odean 2001). Gervais and Odean (2001) and Odean (1998) theoretical models predict that high total market returns make some investors overconfident about the precision of their information. Although the returns are market wide, investors mistakenly attribute gains in wealth to their ability to pick stocks. Overconfident investors trade more frequently in subsequent periods because of inappropriately tight error bounds around return forecasts. Alternatively, market losses reduce investor overconfidence and trading, although perhaps not in a symmetric fashion. To study the effect of overconfidence bias on volatility, this article uses the relation between volatility and trading volume. This relation was the subject of many prior researches (Lamoureux and Lastrapes 1990; Schwert 1989; Benos 1998; Albulescu 2008). The contribution here is to distinguish excessive trading volume of overconfident investors from other factors that affect volatility. In this way, the Chuang and Lee (2006) methodology is considered. In the one stage of the test procedure, the trading volume is decomposable into one component relating to investors’ overconfidence (OVER) and another unrelated to the overconfidence (NONOVER) and it can be written as: " # p p X X Vt ¼ a þ bj rtj ¼ bj rtj þ ½a þ et  ¼ OVERt þ NONOVERt ð2Þ j¼1

j¼1

where Vt is the index trading volume in t and rt is the index return in t.

123

United States

Mexico

Brazil

f2

j

f1

x

0.98****

(325.06)

(178.89)

(-9.48)

(-8.51)

0.96****

-0.08****

(9.45)

(7.98)

-0.08****

0.14****

(-8.38)

(-7.54)

0.11****

-0.27****

-0.37****

Panel B: emerging markets

2.50

3.22

AD

0.07

0.065

f1 þ j

-0.17

(551.87)

(434.88)

-0.209

0.99****

(-6.55)

(-9.49)

0.99****

-0.05****

(11.92)

(9.97)

-0.07****

0.12****

(-7.79)

(-7.93)

0.14****

-0.15****

-0.23****

Canada

f1  j

f2

j

f1

x

Panel A: developed markets

Countries

(295.72)

0.98****

(-6.22)

-0.04****

(10.30)

0.17****

(-7.85)

-0.27****

Korea

3.42

0.08

-0.27

(132.77)

0.97****

(-13.52)

-0.13****

(7.51)

0.12****

(-9.09)

-0.87****

Swiss

(107.2725)

0.9103****

(-7.6329)

-0.1115****

(17.2128)

0.3785****

(-12.8159)

-1.1013****

Malaysia

2.56

0.09

-0.22

(379.98)

0.98****

(-9.63)

-0.07****

(10.74)

0.15****

(-8.42)

-0.24****

France

Table 3 Volatility asymmetry: univariate EGARCH regression parameters: 1999–2009

2.82

0.08

-0.24

(364.8283)

0.9845****

(-8.2627)

-0.0517****

(13.1182)

0.1983****

(-10.3951)

-0.2881****

Singapore

(379.10)

0.99****

(-9.44)

-0.07****

(10.13)

0.16****

(-8.22)

-0.25****

United Kingdom

5.70

0.04

-0.24

(178.93)

0.95****

(-10.10)

-0.11***

(14.39)

0.26****

(-11.78)

-0.60****

India

(358.04)

0.98****

(-15.15)

-0.10****

(10.63)

0.14****

(-8.04)

-0.26****

Australia

1.81

0.10

-0.18

(446.85)

2.07

0.12

-0.25

(210.40)

0.97****

(-7.05)

-0.07****

(10.16)

0.19****

(-7.73)

-0.37***

Japan

(132.77)

0.93****

(-13.80)

-0.17****

(14.565)

0.30****

(-11.74)

-0.87****

Kuwait

0.99****

(-5.91)

-0.04****

(9.74)

0.14****

(-7.53)

-0.19****

Hong Kong

Does Overconfidence Bias Explain Volatility? 303

123

123

-0.23

-0.20

0.02

7.86

f1  j

f1 þ j

AD

1.8359

0.2669

-0.4901

Malaysia

t1

0.1465

-0.2501

Singapore

0.15

-0.368

India

0.08

-0.27

Kuwait

****, ***, **,* denote significant at the 1 %, 1 %, 5 % and 10 % levels, respectively

f1  j measures the sensitivity of volatility to negative shocks, impact of positive shock, f2 is the GARCH term that measures the impact of last period’s forecast  variance,    j þ f1 measures the sensitivity of volatility to positive shocks, the asymmetry degree (AD) = ff11 j t-statistic is reported into parenthesis . þj

1.69

0.13

-0.22

Korea

1.7068 2.43 3.42   ffiffiffiffiffiffi t1 þ f2 ht1 where j is the leverage effect parameters, f1 is the This table presents the results of the following EGARCH (1,1) model estimate: ln ht ¼ x þ f1 jgt1pjþjg h

3.58

0.06

Mexico

Brazil

Table 3 continued

304 M. B. Abbes

Does Overconfidence Bias Explain Volatility?

305

Formal overconfidence theories do not specify a time frame for the relationship between returns and trading volume, so this study lets the data determines the number of lags to include. Specifically, the study here set five lag (k = 5) based on both Schwartz Information Criteria (SIC) and Akaike Criteria (AIC). In the second stage, these two components of trading volume are including into the conditional variance equation of EGARCH model, as follows: rt ¼ lt þ gt ;

ð3Þ

g=ðVt ; gt1 ; gt2 ; . . .; rt1 ; rt2 Þ  GEDð0; ht Þ;   jgt1 j þ jgt1 pffiffiffiffiffiffiffiffi Inht ¼ x þ f1 þ f2 Inht1 þ f3 OVERt þ f4 NONOVERt : ht1 The parameter f3 represents the effect of overconfidence on volatility and the parameter f4 measures the effect of other factors on excessive volatility. If the excessive volatility observed during financial crisis is caused by the investors’ overconfidence, f3 must be positive and greater than f4 . To study the behavior of overconfident investors during global financial crisis, this section estimates the model for two periods. The first is relative to all sample period (January 1999–December 2009) and the second is relative to the crisis period (July 2007–December 2009). The choice of later period is based on the results of Chow breakpoints test which suggest that the subprime crisis was started in July 2007 (F-statistic = 514.525, probability = 0.000). This date is also considered by Abdelhe´di et al. (2011) and Avgouleas (2008). Table 4 reports the results from estimating Eq. (3) for 1999–2009 period. Panel A concerns developed markets and Panel B is relative to emerging markets. For most market indexes, the estimated f3 parameter is statically significant. But, overconfidence bias is not significant for the Japanese and Singaporean markets. This finding can be explained by the cross-cultural variations in the overall accuracy of probability judgments that bear upon more common practical decisions other than overconfidence (Yates et al. 1998). The significance of the parameter measuring the effect of overconfidence on volatility associated with the rejection of null hypothesis that f3 ¼ f4 suggests that overconfidence bias contributes to the return volatility on US, European, Hong Kong and most of emerging financial markets. Table 5 reports the results from estimating Eq. (3) for turmoil period (July 2007– December 2009). The parameter measuring the effect of overconfidence on index volatility becomes no significant for Canada, Australia, Hong Kong, Japan, Korea, Malaysia, Singapore and Kuwait. However, the parameter estimate is significantly negative for the United States, Mexico, France, United Kingdom and Swiss stock index markets. This result indicates that during global financial crisis the proxy of overconfidence bias affects negatively the market index return. This finding can be partially explained the reversal of index returns during subprime crisis period. Indeed, for most of the financial markets and especially for the United States, bigger companies having positive returns before the subprime crisis present negative

123

123

United States

5.04

(0.02)

Mexico

(0.00)

Brazil

(2.71)

(-6.04)

71.164

0.01 ***

(2.57)

(6.01)

-0.12****

0.07***

0.09****

(242.93)

(324.28)

(-6.28)

0.98****

(-9.80)

0.97****

-0.05****

(9.56)

(5.59)

-0.09****

0.11****

(-5.48)

0.08****

-0.23****

(-7.58)

f1

x

Canada

-0.28****

0.12****

(6.26)

(3.34)

(-6.56)

(-3.05)

0.09***

-0.44****

-0.48***

Panel B : emerging markets

v2 p value

f4

f3

f2

j

f1

x

Panel A: developed markets

Countries

(8.31)

0.15****

(-6.46)

-0.27****

Korea

(0.00)

14.94

(2.41)

0.01**

(4.22)

0.09****

(189.86)

0.98****

(-7.15)

-0.08****

(6.76)

0.15****

(-5.66)

-0.33****

Swiss

Table 4 Overconfidence and market volatility: 1999–2009

(20.92)

0.48****

(-12.77)

-1.79****

Malaysia

(0.00)

26.79

(4.81)

0.01****

(5.23)

1.11****

(276.69)

0.97****

(-8.77)

-0.08****

(8.04)

0.13****

(-8.60)

-0.32****

France

(11.09)

0.19****

(-8.64)

-0.36****

Singapore

(0.00)

9.59

(5.67)

0.03****

(3.19)

0.82***

(190.37)

0.97****

(-10.34)

-0.11****

(5.68)

0.08****

(-6.65)

-0.36****

Australia

(10.27)

0.29****

(-8.19)

-1.25****

India

(0.00)

16.23

(-3.27)

-0.02 ***

(3.67)

0.27****

(265.37)

0.98

(-6.76)

-0.07****

(5.54)

0.12****

(-4.71)

-0.19****

United Kingdom

(0.30)

0.31****

(-11.58)

-0.87****

Kuwait

(0.01)

6.33

(4.88)

0.01****

(2.69)

0.21***

(215.84)

0.98****

(-5.88)

-0.04****

(5.3715)

0.0801****

(-5.63)

-0.26****

Hong Kong

(0.99)

0.00

(0.03)

0.04

(0.80)

0.03

(0.16)

0.13

(-0.26)

-0.01

(-0.03)

-0.001

(-1.08)

-7.26

Japan

306 M. B. Abbes

(0.30)

(0.00)

p value

(0.00)

542.05

(7.49)

0.08****

(32.08)

0.46****

(60.36)

0.85****

(-2.97)

-0.05***

Malaysia

0.01

(2.46)

0.01**

(0.29)

0.052

(238.53)

0.98****

(-5.55)

t1

-0.04****

Singapore

40.77

(6.95)

0.07****

(6.59)

0.32 ****

(52.34)

0.88****

(-6.86)

-0.13****

India

5.31

(0.02)

0.0003

(2.23)

0.16**

(131.10)

0.93****

(-13.57)

-0.17****

Kuwait

****, ***, **,* denote significant at the 1 %, 1 %, 5 % and 10 % levels, respectively

statistic v2 with one degree of freedom is used to test the null hypothesis thatf3 ¼ f4 , and the p value is the probability finding the value of the v2 test statistic or higher under the null hypothesis. t-statistic is reported into parenthesis

(0.00)

4,696.33

(-2.53)

-0.01**

(1.72)

0.22*

(245.92)

0.98****

(-4.84)

-0.04****

Korea

(0.92) (0.00) (0.0213)   ffiffiffiffiffiffi t1 þ f2 Inht1 þ f3 OVERt þ f4 NONOVERt . The test This table reports the results of conditional variance equation estimate of the EGARCH model: Inht ¼ x þ f1 jgt1pjþjg h

v

1.05

(-0.95)

(-0.35)

10.67

-0.44

(4.99)

(3.25)

-0.00

0.03****

(151.32)

(50.02)

0.30 ***

0.96****

(-8.17)

(-4.27)

0.95****

-0.10****

-0.07****

Mexico

2

f4

f3

f2

j

Brazil

Table 4 continued

Does Overconfidence Bias Explain Volatility? 307

123

123

United States

j

f1

x

0.06****

(1.49)

(-0.78)

(0.20)

(1.54)

-0.03

0.01

(-15.29)

(-20.37)

0.08

-13.22****

-14.33****

Mexico

(0.73)

(0.00)

Brazil

0.11

(0.05)

(1.76)

21.94

0.12

(0.48)

(-4.67)

0.01 *

0.39

(8.85)

(39.20)

-1.63 ****

0.79****

(1.24)

0.94 ****

(-3.29)

(2.05)

0.05

(2.08)

-0.11 ****

0.14**

(-2.48)

0.08 **

-2.01***

(-2.46)

Canada

-0.63***

Panel B: emerging markets

v2 p value

f4

f3

f2

j

f1

x

Panel A: developed markets

Countries

(0.00)

9.89

(2.57)

(0.95)

0.02

(3.25)

0.10***

(-2.39)

-0.14**

Korea

0.03 **

(-3.07)

-1.23 ***

(38.98)

0.94****

(-0.74)

-0.02

(2.83)

0.18***

(-2.83)

-0.73***

Swiss

(0.00)

903.36

(-2.86)

-0.06***

(5.50)

0.15****

(-3.86)

-0.29****

Malaysia

(-10.33)

-0.002****

(-31.43)

-0.17 ****

(-10.66)

-0.002****

(15.76)

0.001****

(-32.85)

-0.001****

(-46.10)

-0.00****

France

(0.01)

6.33

(0.66)

0.02

(3.56)

0.18****

(-2.96)

-0.50***

Singapore

(-0.00)

-7.61E-10

(-2.52)

-0.003***

(57.35)

0.01 ****

(-0.96)

-0.0002

(2.31)

0.001**

(-1.61)

-0.002*

United Kingdom

Table 5 Overconfidence and market volatility during subprime crisis period (July 2007–December 2009)

(-2.57)

-0.09**

(3.83)

0.17****

(-4.14)

-1.40****

India

(0.20)

1.710

(1.57)

0.05

(1.38)

0.97

(10.05)

0.83****

(-0.37)

-0.01

(2.73)

0.24***

(-2.26)

-1.84**

Australia

(0.30)

1.08

(1.767)

0.062*

(1.30)

0.27

(7.94)

(-11.26)

-0.18****

(12.75)

0.32****

(-10.29)

-0.86****

Kuwait

0.8502****

(0.09)

0.003

(0.71)

0.04

(-1.44)

-1.30

Hong Kong

(1.33)

1.33

(1.57)

0.05*

(1.18)

1.81

(9.45)

0.82****

(-0.29)

-0.01

(2.38)

0.19**

(-2.15)

-1.68**

Japan

308 M. B. Abbes

(0.00)

(0.36)

0.83

(1.20)

0.01*

(-0.90)

-0.50

(124.66)

0.98****

Malaysia

0.001

(-1.61)

-0.01

(0.05)

-0.03

(51.21)

0.95****

t1

Singapore

46.63

(2.89)

0.04***

(6.98)

4.63****

(20.23)

0.84****

India

0.14

(0.13)

0.002

(0.40)

0.04

(118.96)

0.94****

Kuwait

****, ***, **,* denote significant at the 1 %, 1 %, 5 % and 10 % levels, respectively

statistic v2 with one degree of freedom is used to test the null hypothesis that f3 ¼ f4 , and the p value is the probability finding the value of the v2 test statistic or higher under the null hypothesis. t-statistic is reported into parenthesis

(0.00)

4,696.32

(-0.20)

-0.004

(1.01)

0.34

(157.09)

0.99****

Korea

(0.97) (0.00) (0.71)   jgt1 jþjg pffiffiffiffiffiffi t1 þ f2 Inht1 þ f3 OVERt þ f4 NONOVERt : The test This table reports the results of conditional variance equation estimate of the EGARCH model: Inht ¼ x þ f1 h

(0.00)

(-1.24)

(-2.08)

p value

-0.076

-0.36 ***

18.24

(-4.33)

(3.01)

14.24

-5.51****

(-6.36)

0.64 ****

-0.68****

(-8.94)

Mexico

-0.80****

v2

f4

f3

f2

Brazil

Table 5 continued

Does Overconfidence Bias Explain Volatility? 309

123

310

M. B. Abbes

returns during this crisis which cause the loss of confidence by investors in financial markets.

Conclusion The global financial crisis of 2008–2009 indicates that in spite of witnessing new financial innovations; there are some problems which are inherent to the existing financial structure and to irrational exuberance or excessive optimism that pervade financial markets. This paper explores the problem of the current global financial crisis, using a behavioral perspective. Particularly, this paper examines empirically the ability of overconfidence bias to explaining excessive volatility during subprime crisis across emerging and developed stock markets. To further examine this current crisis, the movement of price and volatility of 15 stock market indexes is analyzed. Results indicate an otherwise struggling stock markets around the worldwide. Indeed, there has been a steep decline in index prices and a peak in index volatilities. Then, this study employs an EGARCH model to assess the asymmetric effect of volatility. The results show that the effect of bad news is greater than good news on volatility for all of indexes. Therefore, the study here examines the relation between excessive trading volume of overconfident investors and excessive return volatility. The trading volume decomposable into a first variable relating to overconfidence and a second variable unrelated to investors’ overconfidence. The relation between return volatility and these two variables has been estimated for all sample period and during global financial crisis. For all sample period, conditional volatility is positively and significantly related to trading volume caused by overconfidence bias except for the Japanese and Singaporean markets. This finding provides strong statistical support to the presence of overconfidence bias among investors in most of developed and emerging stocks markets. Generally, this psychological bias constitutes a confirmed explanation of the prices excessive volatility. Although, during global financial crisis period, overconfidence variable cannot explain volatility for developed markets and the most of emerging markets because of the loss of confidence by investors in the financial markets. Overall, behavioral finance proposes an excellent way to understand the fundamental causes of the current global financial crisis and to identify possible policy responses to remove the behaviors that produce instability in the financial markets.

Appendix See Figs. 5 and 6

123

Does Overconfidence Bias Explain Volatility?

311

Fig. 5 Index prices of developed and emerging markets, 1999–2009

Fig. 6 Index volatilities of the developed and emerging markets, 1999–2009

123

312

M. B. Abbes

References Abdelhe´di ZM, Boujelbe`ne-Abbes M, Boujelbe`ne Y (2011) Subprime crisis and volatility spillover. Int J Monet Econ Financ 4:1–20 Ackermann J (2008) The subprime crisis and its consequences. J Financial Stab 4:329–337 Albulescu CT (2008) The Impact of the New Financial Products on the Volatility of the Economic Growth. J Appl Econ Sci 2:12–19 Anders CJ (2010) Financial markets in East Asia and Europe during the global financial crisis. In: Working Paper Stockholm School of Economics Avgouleas E (2008) Financial regulation, behavior Finance, and the financial credit crisis in search of a new regulatory model. In: Working paper, http://papers.ssrn.com. Accessed 20 Nov 2008. Bachellier L 1900 Benos AV (1998) Aggressiveness and survival of overconfident traders. J Financial Mark 1:353–383 Berger AN, Bouwman CHS (2008) Financial crises and bank liquidity creation. http://ssrn.com Black F (1976) Studies of stock price volatility changes. In: Proceedings of the American Statistical Association, Business and Economic Statistics Section, pp 177–181 Chuang W, Lee BS (2006) An empirical evaluation of the overconfidence hypothesis. J Banking Financ 30:2489–2511 Chukwuogor C (2008) Stock Markets Returns and Volatilities: A Global Comparison. Int Res J Financ Econ 15:1–24 Claude B. Erb, Campbell RH, Viskanta TE (1998) Risk in emerging markets, contagion and other factors must be taken into account. In: The financial survey July–August, pp 42–46 Daniel K, Hirshleifer D, Subrahmanyam A (1998) Investor psychology and security market under and overreactions. J Financ 53:1839–1886 Gervais S, Odean T (2001) Learning to be overconfident. Rev Financial Stud 14:1–27 Kausik C, Franc K (2001) Have East Asian stock markets calmed down? Evidence from a regimeswitching model working paper, mimeo, University of Amsterdam, Amsterdam Lamoureux CG, Lastrapes WD (1990) Heteroskedasticity in stock return data: volume versus GARCH effects. J Financ 45:221–229 ´ , Rubio G, Serna G (2004) Autoregressive conditional volatility, skewness and kurtosis WP-AD Leo´n A 2004-13, Instituto Valenciano de Investigacines Economicas Nelson DB (1991) Conditional heteroskedasticity in asset returns: a new approach. Econometrica 59:347–370 Odean T (1998) Volume, volatility, price, and profit when all traders are above average. J Financ 53:1887–1934 Richard JR (2009) Too much right can make a wrong: setting the stage for the financial crisis FBR of Chicago Working Paper, No. 2009-18 Schwert GW (1989) Why does stock market volatility change over time? J Financ 44:1115–1153 Shefrin H (2009) How psychological pitfalls generated the global financial crisis voices of wisdom: understanding the global financial crisis So¨hnke MB, Gordon MB (2009) No place to hide: the global crisis in equity markets in 2008/09. J Int Money Financ (in press) Yates JF, Lee JW, Shinotsuka H, Patalano AL, Sieck WR (1998) Cross-cultural variations in probability judgment accuracy: beyond general knowledge overconfidence? Organ Behav Hum Decis Process 74:89–117

123