Cross-Market Herding: Do ‘Herds’ Herd with Each Other?

Abstract

Although herding constitutes one of the most widely researched behavioral trading patterns internationally, the possibility of cross-market herding has remained largely underexplored in the literature. Our study provides a detailed empirical investigation of this issue in the context of ten Asia-Pacific markets for the February 1995–March 2022 window. We find that all ten markets’ “herds” project significant relationships with each other, with causality being identified within a minority of those relationships. These results are robust when controlling for financial crises (Asian; global financial; global pandemic) and US market returns.

Keywords:

• Herding
• cross-market herding
• Asia-Pacific
• causality
• financial crises

JEL CLASSIFICATION:

• G4
• G11
• G15

Disclosure statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Notes

1 For a detailed discussion on the empirical identification of cross-market herding in the extant literature and how our study contributes to the debate on the issue, please see the appendix.

2 To the best of our knowledge, the only study that has explicitly attempted to assess the relationships (yet not causality) between different markets’ “herds” is Chiang et al. (2013); for more on their approach (and its innate shortcomings), see the discussion in the appendix.

3 For more on why other measures and approaches proposed in the herding literature present issues in cross-market herding estimations (and, thus motivate the employment of the Hwang and Salmon 2004 measure here), see the discussion in appendix 1.

4 The overwhelming number of research papers relevant to our analysis of herding behavior published in the last years is presented in summary in a table in appendix 2. The papers have been classified according to the scope of our research in: (1) those focused on the Asian continent, (2) the ones that study the role of different crises, (3) papers that consider different markets and their possible relationships and, finally, (4) those papers that are relevant but do not fit into any of the three previous categories (and are classified as "Others").

5 The two markets tend to bear differences in their institutional structures, hence choosing both of them, instead of only focusing on one to proxy for the Chinese market.

6 The choice of the starting date was motivated by the limited availability of data for some of our sample markets (the Shanghai market had very few listed stocks pre-1995; some risk-free rate data was also unavailable pre-1995).

7 East and South East Asian markets tend to generate market-wide herding much more often compared to their European and North American counterparts. This has often been ascribed to Asian markets’ larger average retail participation increasing the potential for noise trading patterns. See, for example, Chang et al. (2000), Chiang and Zheng (2010), Chen (2013), Chiang et al. (2013), Lam and Qiao (2015) and Chong et al. (2017). Investigating cross-market herding in markets with lower probability of herding would be counterintuitive, as we would potentially end up with some markets exhibiting no herding at all.

8 At least half of our sample’s markets (Shanghai; Japan; Hong Kong; South Korea; Taiwan) feature among the top 20 markets in the world in terms of capitalization (according to the World Federation of Exchanges); our sample, therefore, allows us the opportunity to assess whether cross market herding varies in terms of its origins/effects contingent on a market’s size.

9 More restrictions in terms of foreign investors would be expected to culminate in an enhanced retail investors’ base, something evident in many Asian (emerging, mainly) markets – see Chang et al. (2000), Chiang and Zheng (2010), Chen (2013), Chiang et al. (2013), Lam and Qiao (2015) and Chong et al. (2017). If so, and given retail traders’ noise trading tendencies, this may imply greater potential for herding (particularly so in emerging markets – see the discussion in the previous section). In addition, many of our sample markets tend to accommodate concentrated corporate structures in the form e.g., of interfirm networks (keiretsu) in Japan (Kim and Nofsinger 2005) or conglomerates (chaebol) in South Korea (Fitzgerald and Kang, 2022) and this may prompt greater response to foreign markets’ signals, if news related to those corporate formations’ sectors arrives from overseas markets (and potentially foments cross-market herding in their stock markets).

10 The indices used are the following: Shanghai Stock Exchange Composite (Shanghai); Hang Seng (Hong Kong); IDX Composite (Indonesia); Nikkei 225 (Japan); FTSE Bursa Malaysia KLCI (Malaysia); PSEi (Philippines); FTSE ST All Share (Singapore); KOSPI (South Korea); TAIEX (Taiwan); SET (Thailand).

11 In the vast majority of cases (China-Shanghai/Hong Kong; Indonesia; Japan; Malaysia; Singapore; Taiwan; Thailand), the risk-free rate used is the 3-month deposit rate of each market; exceptions include the Philippines (91-day Treasury Bill rate) and South Korea (91-day certificate of deposit rate).

12 See also the discussion in appendix 1.

13 Hwang and Salmon (2004) rationalized the choice of the monthly frequency as a trade-off between reducing biases in beta-estimation and obtaining a number of $h_{\mathit{\text{mt}}}$-observations large enough to allow for the detection of herding.

14 Results from the Dickey-Fuller stationarity tests are not reported here in the interest of brevity and are available on request from the authors.

15 Perhaps due to crises revealing groundbreaking fundamentals that render the pre-crisis consensus (and, hence, its herding) obsolete; see Andrikopoulos et al. (2017).

16 US market returns have been found to motivate herding internationally in several studies (Chiang and Zheng 2010; Economou et al. 2015a; Guney et al. 2017).

17 Extracting herding from the cross-section of the betas of a market’s stocks necessitates accounting for the potential impact of non-synchronous trading over beta-estimates. If a stock is infrequently traded and its price occasionally remains stale, the covariance estimate between that stock and the market will be downwardly biased, leading the estimated beta itself to be downwardly biased as well. As this issue has been observed among Asia-Pacific markets (e.g., Levine and Schmukler, 2006), we adjusted all betas estimated from Equation (9) for non-synchronous trading based on the methodologies of Scholes and Williams (1977) and Dimson (1979) and herding has been re-estimated. Results (not reported here in the interest of brevity and available on request from the authors) are clearly indicative of herding across all sample markets, thus denoting that the significance of the estimated single-market herding used for the cross-market herding tests holds when correcting for thin trading.

18 In that respect, if intra-regional portfolio investment is low (or varying among markets), this would suggest a less significant impact of each market’s investors on other markets’ equity trading in the region - and help account for the limited evidence of causality unearthed in our study. The lack of data-availability on intra-region equity trades in the Asia-Pacific, however, renders it impossible to verify this. One might argue that herding from more “open” markets in the region (such as Hong Kong and Singapore) would be more receptive to/influential for herding in other markets; this, however, is not confirmed via our results in Table 6. In addition, when considering the aggregate influence of the other nine markets’ herding on each market’s herding (see the last row of panels A and B in Table 6), we find that it is only significant for the herding of 6-7 countries (Indonesia, Malaysia, South Korea, Taiwan and Thailand for both panels, and Shanghai (Hong Kong) only in panel A (B)).

19 For more on herding in cross-border exchanges’ member-markets and its determinants pre and post membership, see Andrikopoulos et al. (2017) and Economou et al. (2015a).

Additional information

Funding

This work was supported by the Spanish Ministry of Science, Innovation and Universities, the Spanish State Research Agency (AEI) and the European Regional -Development Fund (ERDF) under Grant RTI2018-093483-B-I00; and the Government of Aragon under Grant S11_20R: Cembe.