An Examination of Herd Behavior in REITs:
Hong Kong
Jiani Xia
Bachelor of Economics and Business The University of Amsterdam
The Netherlands
1. Introduction
During the last two decades, the entire financial system has been challenged by several international crises, especially the international financial crisis in 2008. The prices of many financial products in different markets experienced drastic fluctuations within that period. According to Philippas et al. (2011), there were abnormal market sentiments and irrational price movements in some financial markets, and those could be attributed to not only fundamentals, but also many other market frictions, such as investors’ behavioral biases. In this paper, one important behavioral bias, (irrational)
herd behavior, is going to be investigated whether it existed in the Hong Kong real
estate market during the period from January 2006 to April 2013.
1.1 Introduction of Herd Behavior
Herd behavior is generally defined as a group of asset investors’ same-direction trading behavior for a period of time. It can be distinguished between two categories, namely rational herd behavior and irrational herd behavior. Rational herding stands for the principal-agent problem, in which managers follow the choice of the majority in the market and ignore their own information or preferences about the decisions with the purpose of protecting their reputations or their own profits (Scharfstein and Stein, 1990; Chang et al., 2000; Cipriani & Guarino, 2005). From the irrational perspective, on the other hand, many other researchers discussed the investors’ herd behavior pertinent to psychological reason (Rajan, 1994; Devenow & Welch, 1996; Nofsinger & Sias, 1999; Chang et al., 2000). It means investors just “disregard their prior beliefs and blindly follow other investors’ decisions” (Al-Shboul, 2013, p. 234).
In terms of the testing approaches, they are mostly based on two models previously created by Christie & Huang (1995) and Chang, Cheng & Khorana (2000). The first group of researchers stated a linear model to test whether herding existed under market stress. The model discovers the relation between the cross-section standard deviation (CSSD) and the average market returns. Lower value of CSSD
shows the smaller deviation of one individual equity return from the market index. With a significant fluctuation of market prices (returns), if there is herd behavior in the market, most individual investors would choose to follow market consensus and imitate other investors’ trading behavior. It implies that the corresponding CSSD would decrease or increase at a decreasing rate.
Chang et al. (2000) focused more on the tendency of nonlinear herding when there is an extreme market movement. The approach employed the cross-section absolute deviation (CSAD) model and tried to test the evidence of herding if the value of CSAD tend to decrease or increase at a decreasing rate with the change of market index. According to Al-Shboul (2013), the second approach is more appreciated because the usage of both least squares regression and quartile regression is more advanced to get fitting estimators. For this reason, this paper will discuss both methods later but only use the CSAD model to evaluate herd behaviour of the chosen sample.
1.2 Introduction of Hong Kong REITs market
A Real Estate Investment Trust (REIT) is defined as one kind of securitisation including the regrouping of real estate entities with the aim of delivering the income to individual investors or associations via the investments in a set of real estate such as retails, hotels, offices and private housing or real estate mortgage (Hang Seng Indexes Company Limited, 2013; Philippas et al., 2011). The first Hong Kong REIT was launched and listed in HK Stock Exchange (SEHK) at the end of 2005. By April 2013, there had been 9 (8 alive and 1 dead) REITs in the Hong Kong real estate market, with a total capitalization of approximately €15.294 billion (European Public Real Estate Association, 2012, P.2).
The growth of the REITs market is at a relatively low speed in Hong Kong, compared to the 75% rate of increase in the United State REITs market between 2002 and 2010. However, Hong Kong REITs market plays an eminent role among Asian
countries. The Link REIT (0823.HK), which was the first listed trust in Hong Kong, is currently the largest REIT in Asia with its market capitalization of 87.81 billion HKD. In addition, it is also one of the largest retail-focused REIT on a global scale. Furthermore, since Hong Kong is the administrative region of the People’s Republic of China, a good analysis of real estate market may help investors to explore the great potential in the mainland of China. Therefore the study of investors’ behavior in this relatively small but important market will still be valuable.
The rest of this paper will be structured as follows. Section 2 will give the relevant literature reviews, followed by the introduction of methodology and data in Section 3. The analysis of empirical results as well as the short conclusion will be indicated in Section 4 and 5, respectively.
2. Literature Review
Herding is one of the most important phenomena of bias, which shows the influence of psychology on financial markets. Most of the existing literature pays great attention on the irrational herd effects in different stock markets.
Christie and Huang firstly tried to investigate the US stock market in 1995. They created a linear model which focuses on the linear relation between the average market returns and the cross-section standard deviation (CSSD). The results of their test (significantly positive coefficients and confirmation of t-statistics) showed significantly higher dispersions between daily-based (and monthly-based) CSSD and market index under the market stress. The CSSDs tended to be larger. As mentioned above, the higher value of CSSD showed a larger deviation of one individual equity return from the market level, which further implied that the investors did not really follow the market consensus. Thus, Christie and Huang (1995) found no evidence of the existence of herd behavior for daily or monthly returns with respect to large price (return) fluctuations.
Based on their contribution, Chang et al., (2000) created a new nonlinear model called the cross-section absolute deviation (CSAD) to test the presence of herding with a global view (the U.S., Hong Kong, Japan, Taiwan and South Korea). That model mainly analyzed the relation between the CSADs and absolute value of average market returns as well as the square of the value of average market returns. They had the same findings for the U.S. as what Christie and Huang got. The coefficients for the market returns under both up and down down market were always positive and significant in three sub-models. Furthermore, the results for Hong Kong were similar as the ones for the U.S. However, for Japan, Taiwan and South Korea, the estimators displayed differently. For each country, most of the results were significantly negative in the sub-models, respectively. This obviously implied the declines of the corresponding CSADs. Therefore, Chang et al., (2000) concluded the strong evidence of asymmetric herding in two emerging markets, South Korea and Taiwan, and weak evidence in Japanese market, but no existence of herding in the U.S. and Hong Kong was supported by that paper, which was consistent with the results from previous research.
The current researches were mostly based on the previous two models. In 2006, Demirer and Kutan focused on the Chinese stock market, and found the market was efficient without any herding. Chiang & Zheng (2010) applied the CSAD method as well to 18 countries globally between 1988 and 2009. They found that herding was in the U.S. market and most of Latin American and Asian markets, while the asymmetry was stronger in Asia. In the same year, Philippas et al. (2010) extended attention to the REITs market for the first time. They examined the presence of herding in U.S. REITs market, and found a significant evidence for the herd behavior without asymmetry. In addition, investors’ sentiment and adverse shocks from funding conditions were proved as the factors which caused the herding.
3.1 Methodology
Christie and Huang method (1995)
The first model to test herd behavior was developed by Christie and Huang in 1995. They indicated that the herd behavior could be tested by the cross-sectional standard deviation of asset returns (CSSD) model, specified as:
(
)
1 1 2 , , − − =∑
= N R R CSSD N i t m t i t (1)where Ri,t refers to the REIT return on any trust i at time t. Rm,t is the return of the market index, which is represented by the equally-weighted average of N available trusts at time t, and N is the number of trusts.
The benchmark regression is shown as below:
CSSD
t=
α
+
γ
1D
tL+
γ
2D
tU+
ε
t (2)According to the definition given above, investors would have the tendencies to deviate from their individual information and follow the most common choice in the market when herding arises. As a result, the gap between individual asset return and market return would tend to be narrow due to this behavioral bias. Therefore, this method implies that REIT returns tend to converge around the market index level and the CSSD will be expected to decrease with the presence of herd behavior.
Chang, Cheng and Khorana method (2000)
However, this paper will apply an alternative method developed by Chang Cheng and Khorana in 2000. Chang et al. demonstrated that the new cross-sectional absolute deviation of returns (CSAD) was similar to the approach of CSSD; however, it did not
take full consideration of conditions because it was designed to examine not only the linear herding but also the non-linear one. The expected cross-sectional absolute deviation (ECSAD) is expressed as:
) ( * 1 1 , , 1 , , 1 , mt f t N i t m t i N i t i t E R R N AVD N ECSAD =
∑
=∑
− − = = β β (3)where AVD donates the absolute value of the deviation of asset i’s expected return from the portfolio return at time t. β and i,t β represent the systematic risk m,t measure of the trust and the market portfolio, respectively. In addition, E(Rm,t −Rf,t) stands for the expectation of the market premium.
However, since the ECSAD and E(Rm,t) in the model are unobservable, Philippas et al. (2011) simplified the previous model into the following formula:
∑
= − = N i t m t i t R R N CSAD 1 , , 1 (4)where similarly Ri,t donates the REIT return on any trust i at time t. Rm,t is the return of the market index, which is represented by the equally-weighted average of N trusts at time t, and N is the number of trusts in the sample we choose.
Chang et al. (2000) suggested the non-linear relationship between CSAD and market return as a more efficient approach to capture herd behavior than the linear relationship when the market becomes significantly abnormal. As mentioned before, when there is herd behavior in the market, REIT returns tend to fluctuate closely around the market index level and it should be decreasing in the return’s dispersion. The benchmark regression is shown as below:
t t m t m t
R
R
CSAD
=
α
+
γ
1 ,+
γ
2(
,)
2+
ε
(5)If the coefficientγ2is significantly negative, it implies that the presence of herd behavior is captured by the non-linear model. Therefore, the non-linear function is more powerful than the linear one in the CSSD model. Al-Shboul (2013, p.238) indicated that it is because of the increased herd behavior from investors when there are large price fluctuations in the market. As a result, “dispersions are expected to decrease or increase at a decreasing rate”.
Furthermore, the study of the asymmetry of non-linear herd behavior is another important aspect when examining herding in REITs market. In Chang et el.’s study (2000), they applied two separated regressions for the increasing and the decreasing market, respectively. However, Al-Shboul (2013) introduced an advanced alternative, which placed both two market conditions in one dummy-variable regression model without any restriction. Here, a dummy-variable regression is applied:
( )
mt t( )
mt t t t m t t m t tD
R
D
R
D
R
D
R
CSAD
=
α
+
γ
1(
1
−
)
,+
γ
2 ,+
γ
3(
1
−
)
, 2+
γ
4 , 2+
ε
(6)where Dt is the dummy variable taking the value of 1 at time t if the market return (Rm,t) is negative, or taking 0 otherwise. If the asymmetry exists when the market returns move downward, the cross-sectional deviations should be reduced more on days of market losses than on days of gains. Therefore, the coefficient γ is expected 3 to be significant and negative, and γ4 <γ3 <0 implies a stronger herding when the market is up. Many existing researches already provided evidence of the asymmetric herding in different market situations (Christie and Huang, 1995; Hong et al., 2007; Tan et al., 2008; Demier et al., 2010; Al-Shboul, 2013).
Finally, as the international financial crisis in 2008 was one of the most important events in the last decade, it is commonly believed that the crisis must, to some extent, have affected investors’ decisions in different markets, including REITs market. The general model to test the influence of crisis on herding in REITs market is expanded from the basic model (4):
t t m crisis t m t m t R R D R CSAD =
α
+γ
+γ
+γ
3 , 2 +ε
2 , 2 , 1 ( ) ( ) (7)where Dcrisis is the dummy variable taking the value of 1 if the trading days t are specified as the period of financial crisis, or taking 0 otherwise.
3.2 Data
The dataset of the daily prices and returns of all Hong Kong REITs for the period between November 2007 and April 2013 is sourced from the Hong Kong Stock Exchange (SEHK) and each company’s database. In the chosen sample, there are 9 REITs, while one of which has been closed since the end of 2010. So far, based on the obtained data, a proxy of the market index (Rm,t) on any date can be obtained by the calculation of the equally-weighted average of all available REITs returns as mentioned in the literature part. There are 1748 daily observations in the chosen sample.
In addition, all the data will be grouped into four different sectors, including Retails (R), Industrial/Offices (IO), Hotels (H) and Diversified (D). There are 2 Retail REITs, 2 Industrial/Office REITs, 1 Hotel REIT, as well as 4 Diversified REITs in the sample. Then the corresponding CSAD measures can be calculated for the purpose of testing the reactions from different sectors and within each sector.
4. Empirical results
4.1 Descriptive statistics
Table 1 reports the descriptive statistics of equally-weighted average, median, the maximum and minimum values, and the standard deviation of daily returns (Rm,t), as
well as the cross-sectional absolute deviation (CSADt) for all nine real estate
return ranges from -8.2647% to 10.1135%, while its standard deviation was low with the value of 1.1052%.
Table 1 also reports the descriptive statistics of CSAD, whose values were located between 0.0765% and 5.5106%. However, the standard deviation of CSAD was at a lower level, at 0.5621%. According to Al-Shboul (2003), these results could reflect the possible presence of herd behavior because the dispersion of the overall return from market index is relatively low, and the standard deviation is low for that period as well, which means there are not many extreme movements of investments.
Table 1: Descriptive statistics of CSAD and market returns for overall market
Sample Period (Number of Observations) Mean (%) Median (%) Max (%) [date] Min (%) [date] St. Dev (%) 01/01/2006 - 30/04/2013 (1748) Rm,t 0.0263 0.0587 10.1135 [30/10/2008] -8.2647 [27/10/2008] 1.1052 CSA Dt 0.8538 0.7133 5.5106 [22/12/2008] 0.0765 [07/06/2007] 0.5621
This table summarizes the descriptive statistics of a proxy of REIT market returns with the
equally-weighted average measure from 2006 to April 2013. Also, the table reports the CSAD with respect to the calculated market index in that period. The formulation is specified as indicated in Equation (4).
Table 2, however, shows the similar descriptive statistics of CSAD and market returns for the sectoral markets instead of the overall market as above. There are totally four different sectors of the REITs, i.e., Retails, Industrial/Office, Hotel and Diversified. Nevertheless, the results of CSADH for “Hotel” sector cannot be obtained
because only one trust belongs to that sector, which means the investments under that sectoral market cannot deviate from one trust to others. The number of the available dates is between 693 and 1654. For each category, both sectoral market returns and CSADt are examined. As shown in Table 2, the average daily return is from the lowest
value of 0.0015% (“Industrial/Office” sector) to the highest value of 0.1135% (“Retail”), while “Retail” owns the lowest standard deviation of 0.8637%, compared to the highest standard deviation of 1.9011% from the “Hotel”.
In terms of the CSADt, the magnitudes of the average sectoral market CSAD can
be in the order as: CSADIO > RD > RR because 0.9183% > 0.7719% > 0.5217%.
Meanwhile, “Industrial/Office” sector shows the highest maximum value of 8.9520% with the highest value of standard deviation of 0.9604%, and “Retails” has the lowest maximum value of 3.2% with the lowest standard deviation of 0.4413%.
Table 2: Descriptive statistics of CSAD and market returns for sectoral markets
REIT Sectors Sample Period (# of Obs.) Sectoral returns & CSAD Mean (%) Median (%) Max (%) Min (%) St. Dev (%) Retails 21/04/2010 --30/04/2013 (693) RR 0.1135 0.1081 4.1847 -3.6585 0.8637 CSADR 0.5217 0.4222 3.2 0.0095 0.4413 Industrial/ Office 25/06/2007 --19/04/2010 (694) RIO -0.0015 0.0737 12.1161 -11.8493 1.8419 CSADIO 0.9183 0.6296 8.9520 0.0027 0.9604 Hotel 02/04/2007 --30/04/2013 RH 0.0148 0 15.2778 -17.9592 1.9011
(1442) Diversified 25/05/2006 --30/04/2013 (1654) RD 0.0246 0.0533 11.8763 -7.7647 1.2892 CSADD 0.7719 0.5843 6.1677 0.0683 0.6591
This table summarizes the descriptive statistics of four sectoral markets, i.e., Retails, Industrial/Office. Hotel and Diversified during the period between 2006 and 2013. The sectoral market returns are obtained by the equally-weighted average measure, and the CSAD is calculated using Equation (4).
4.2 Benchmark regression results
Table 3 reports the estimated coefficients of herd behavior with the benchmark regression model showed in Equation (4). In general, the herding only shows under the “Retails” sectoral market.
In panel A, there are 1748 observations for the overall market over the period. The estimated coefficients of independent variables are significantly positive (0.414***) for γ1 and negative (-0.931*) for γ2 with little significance. In addition,
the adjusted R2 is 0.3209. According to Chang et al. (2000), there is weak evidence to reflect the possible existence of nonlinear herd behavior under extreme market conditions since γ2 is negative but without strong significance. The dispersion
would decrease with the movements of the market return.
Panel B displays the estimated coefficients for four sectoral markets. The results for “Hotel” sector are omitted because of the single trust in that sector, therefore, no tendency to deviate among different choices in the sectoral market. For the other three sectors, all have the positive and significant coefficient of γ1, which rejects the
possibilities of linear herding in the market. However, “Diversified” has the negative coefficient of γ2 (-1.699***) with the only strong significance among three sectors,
which is an obvious signal of herding. In contrast, in other two sectoral markets, the estimated coefficients are both negative but not significant. The adjust R2 are 0.1177, 0.3397 and 0.5221, respectively. The explanation behind the results is that, CSADD
can achieve its highest value when RD (=
2 1
2γ
γ
− ) reaches its lowest value based on the equation (5)
CSAD
R(
=
α
+
γ
1R
R,t+
γ
2(
R
R,t)
2+
ε
t)
. Therefore, taking a maximum value of 11.8763% from descriptive statistics with γ1 =0.564, γ2 needsto be -1.699 or smaller.
Therefore, we can deduce that the nonlinear herd behavior could be found under the “Diversified” sectoral market with strong evidence. As for the other three markets, there is no evidence to support the presence of herding.
Table 3: The coefficients of herd behavior using the CSAD method
Panel A: Overall market
Period Obs. cons
γ
1γ
2 Adj. R2 01/01/2006 --30/04/2013 1748 0.00576*** (33.83) 0.414*** (16.53) -0.931* (-1.97) 0.3209
Panel B: Sectoral market
Sectors Obs. cons
γ
1γ
2 Adj. R2 Retails 693 0.00325*** (10.71) 0.342*** (5.38) -3.405 (-1.39) 0.1177 Industrial/ Office 694 0.00427*** (8.84) 0.435*** (9.38) -0.561 (-0.98) 0.3397
Hotel 1442 omitted omitted omitted omitted Diversified 1654 0.00338*** (19.68) 0.564*** (26.18) -1.699*** (-4.78) 0.5221 Note: * p < 0.05, ** p < 0.01, *** p < 0.001.
This table reports the estimated coefficients with the respect to the benchmark regression model: CSADt =α+γ1Rm,t +γ2(Rm,t)2+εt. CSAD is specified as the
dependent variable here, γ1 and γ2 are the estimated coefficients of two independent variables,│Rm,t│ and (Rm,t)2, respectively.
4.3 Dummy-variable regression results
In order to find the asymmetry of herd behavior under up and down markets, a dummy-variable regression model is applied in this part. The model is displayed as Equation (6): CSADt =α+γ1(1−Dt）Rm,t +γ2Dt Rm,t +γ3（1−Dt)
( )
Rm,t 2+γ4Dt( )
Rm,t 2+εt. As discussed before, the first step is to check whether herding exists. In Table 4, the coefficients of linear (and nonlinear) and adjusted R2 are presented. For the overall market, the coefficient γ1 and γ2 are both significantly positive, which implies nolinear herding under either increasing or decreasing market.
γ
3 is statistically negative as expected, but it is not significant with the value of -0.316. Therefore, even if the value of the coefficient γ (-1.740*) is smaller than4 γ3, there is only little evidence to prove the presence of herding under the down market. In other words, the asymmetry of herding cannot be investigated. Similarly, herd behavior is also found in the “Industrial/Office” sector only when sectoral market return moves down. Furthermore, in the “Diversified” sectors, herd behavior is found under both up and down markets, but herding is pronounced during the time of declining market since -2.631*** is smaller than -1.393***. No other asymmetric herd behavior could besupported under either up or down market situations in the rest of sectoral markets because there is no evidence for herd behavior at all.
However, so far, there is one obvious shortage in the analysis for the parts of sectoral markets. Since the Hong Kong REITs market is a fledgling market, not many listing REITs are in the real estate market yet. This causes the problem that in a certain sectoral market, only one or two trusts may be in that sector during the chosen testing period. Therefore, the dispersions within some sectors are already smoothed, and the results of CSADt in Table 2 are not really robust. The same problem also
matters the analyses of the regressions (5) and (6) because of the unrobust CSADt.
Nevertheless, since the outputs of the regressions for the overall market display a high degree of similarity of the results as in each of the different sectors, it serves as the evidence for the robustness of the results within different sectors. Therefore, the obtained findings above could also be considered credible.
Table 4: The coefficients of asymmetric herd behavior under positive and negative market
Panel A: Overall market
Period Obs. cons
γ
1γ
2γ
3γ
4 Adj. R2 01/01/2006 --30/04/2013 1748 0.00568*** (33.52) 0.450*** (15.46) 0.400*** (12.76) -0.316 (-0.55) -1.740* (-2.65) 0.3313
Panel B: Sectoral market
Sectors Obs. cons
γ
1γ
2γ
3γ
4 Adj. R2 Retails 693 0.00324*** (10.68) 0.321*** (4.59) 0.372*** (4.73) -1.864 (-0.64) -5.562 (-1.66) 0.1164
Industrial/ Office 694 0.00419*** (8.84) 0.442*** (8.32) 0.449*** (8.14) 0.739 (1.05) -1.962** (-2.65) 0.3643
Hotel 1442 omitted omitted omitted omitted omitted omitted
Diversified 1654 0.00334*** (19.06) 0.547*** (21.88) 0.605*** (19.76) -1.393*** (-3.58) -2.631*** (-4.03) 0.5226 Note: * p < 0.05, ** p < 0.01, *** p < 0.001
This table reports the estimated coefficients with the respect to the benchmark regression model:
( )
mt t( )
mt t t t m t t m t t D）R D R （ D R D RCSAD =α+γ1(1− , +γ2 , +γ3 1− ) , 2+γ4 , 2+ε , where Dt is the dummy variable taking value of 1 if market return is negative, or taking 0 otherwise. CSADt is specified as the dependent variable, γ1,γ2,γ3,γ4 are the estimated coefficients of four independent variables respectively.
4.3 Impact of financial crisis 2008
Eventually, this section will analyze the influence of financial crisis on herd behaviour. It is confirmed by many researchers that herding effects were more severe during the crisis period (Economou et al., 2008; Chiang & Zheng, 2010; Al-Shboul 2013). There are three definitions of the crisis period. Brunnermeier (2009) gave a general definition of crisis period, between August 2007 and December 2008, compared to a narrow definition since late-September 2008. In Philippas et al.’s paper, they gave a third one (July 2006 - April 2009), according to the turning points of the US home price index. Here, the third definition is used from July 2006 to April 2009 because Hong Kong real estate market experienced similar process of declining bottoming out within that period.
Table 5 reports the coefficients under the overall market, the “Industrial/Office” market and the “Diversified” market. The results for other two sectoral markets are
omitted because one sectoral market (“Retails”) started after the crisis period and the other one (“Hotel”) has only one equity. For the overall market, the significantly negative
γ
2 provides the strong evidence of the presence of nonlinear herding.However, the positive and significant coefficient
γ
3 shows that the cross-sectiondispersion would increase during the crisis period. Therefore, herd behaviour is not obviously found under the overall market during the period of financial crisis. For the same reason, the existing herding in the “Retails” and “Diversified” sectoral market could not be attributed to the crisis either.
Table 5: The coefficients of asymmetric herd behavior under positive and negative market
Panel A: Overall market
Period Obs. cons
γ
1γ
2γ
3 Adj. R2 01/01/2006 --30/04/2013 1748 0.00558*** (32.60) 0.483*** (17.70) -6.696*** (-6.32) 5.109*** (6.06) 0.3345
Panel B: Sectoral market
Sectors Obs. cons
γ
1γ
2γ
3 Adj. R2
Retails 693 omitted omitted omitted omitted omitted Industrial/ Office 694 0.00424*** (8.74) 0.446*** (9.13) -1.854 (-0.94) 1.195 (0.68) 0.3392
Hotel 1442 omitted omitted omitted omitted omitted
Diversified 1654 0.00319*** (18.61) 0.628*** (27.21) -6.723*** (-8.50) 4.518*** (7.08) 0.5359
Note: * p < 0.05, ** p < 0.01, *** p < 0.001.
This table reports the estimated coefficients with the respect to the regression model: t t m crisis t m t m t
R
R
D
R
CSAD
=
α
+
γ
1 ,+
γ
2(
,)
2+
γ
3(
,)
2+
ε
where Dcrisis is the dummy variable taking value of 1 when the trading date is in the defined crisis period, or taking 0 otherwise. CSAD is specified as the dependent variable, γ1,γ2,γ3 are the estimated coefficients of independent variables, respectively.5. Conclusion
This paper examines the existence and the asymmetry of herd behavior in the Hong Kong REITs market from January 2006 to April 2013. In addition, the possible influence of 2008 international financial crisis is also tested.
Based on the literature review, the cross-section absolute deviation (CSAD) rather than cross-section standard deviation (CSSD) approach is applied to test whether herding occurred in that market. This is because CSAD method can capture both linear and nonlinear herding under both normal and extreme market conditions if herding is there. According to the obtained empirical results in section 4, only partial evidence is found to support the existence of herd behavior under the overall market, while there is strong evidence for the herding under the “Diversified” sectoral market. However, no evidence is found for other sectoral markets. The problem about asymmetry herding is also analyzed by a dummy-variable model. The results imply the evidence of asymmetric herd behavior only under the “Diversified” sectoral market that herding would become intense if market return is negative. Finally, the result of the influence of financial crisis is surprisingly far from the expectation. No herding is found in either the overall or four sectoral markets during the period of crisis.
and April 2013. The presence of herding in the overall market and the “Diversified” sectoral market, as well as the asymmetric situations when the market goes down may attribute to the relatively large amount of trusts compared to other sectors. However, Hong Kong REITs market is younger and smaller compared to either the U.S. market or many developed European markets. Some may argue that investors’ psychological behavior could not be completely displayed due to the small size of market and the short growing period of existing trusts. For example, there is only one trust in “Hotel” market and two trusts in “Retails” market, causing the omitted or unrobust empirical results. Therefore, it is better to allow this market some time to mature, and then apply another examination in the near future.
References
Al-Shboul, M. (2013). An Examination of Herd Behavior in the Jordanian Equity Market. International Journal of Economics and Finance, 5(1), 234-249.
Brunnermeier, M. (2009). Deciphering the Liquidity and Credit Crunch 2007-08.
Journal of Economic Perspectives, 23, 77-100.
Chang, E., Cheng, J., & Khorana, A. (2000). Examination of herd behavior in equity markets: an international perspective. Journal of Banking Finance, 24(10), 1651-1679.
Chiang, T. C., & Zheng, D. (2010). An empirical analysis of herd behavior in global stock markets. Journal of Banking & Finance, 34(8), 1911-1921.
Christie, W. G, & Huang, R. D. (1995). Following the pied pier: do individual returns herd around the market? Financial Analysis Journal, 51(4), 31-37.
Cipriani, M., & Guarino, A. (2005). Herd Behavior in a Laboratory Financial Market.
American Economic Review, 95(5), 1427-1443.
Demirer, R., & Kutan, A. M. (2006). Does herding behavior exist in Chinese stock markets? Journal of International Financial Markets, Institutions and Money,
16(2), 123-142.
Devenow, A., & Welch, I. (1996). Rational herding in financial economics. European
Economic Review, 40(3-5), 603-615.
Economou, F., Kostakis, A., & Philippas, N. (2011). Cross-country effects in herding behaviour: Evidence from four South European Markets. Journal of
International Financial Markets, Institutions & Money, 21, 443–460.
HONG KONG. 1-9.
Hang Seng Indexes Company Limited. (2013). Hang Seng REIT Index. Available: <http://www.hsi.com.hk/HSI-Net/HSI-Net>.
Hong, Y., Tu, J., & Zhou, G. (2007). Asymmetries in stock returns: statistical tests and economic evaluation. Review of Financial Studies, 20(5), 1547-1581.
Nofsinger, J. R., & Sias, R. W. (1999). Herding and feedback trading by institutional and individual investors. Journal of Finance, 54(6), 2263-2295.
Philippas, N., Economou, F., Vassilios, B., & Alexandros, K. (2011). An examination of herding behavior in REITs. International Review of Financial Analysis.
Rajan, R.G., (1994). Why credit policies fluctuate: a theory and some evidence.
Quarterly Journal of Economics, 436, 399–442.
Tan, L., Chiang, T. C., Mason, J. R., & Nelling, E. (2008). Herding behavior in Chinese stock markets: An examination of A and B shares. Pacific-Basin Finance
