Herding Dynamics in Exchange Groups: Evidence from Euronext
Fotini Economoua, Konstantinos Gavriilidisb, Abhinav Goyalc and Vasileios Kallinterakisc,
aOpen University of Cyprus, PO Box 12794, 2252, Nicosia, Cyprus
bUniversity of Stirling Management School, Stirling FK9 4LA, UK
cUniversity of Liverpool Management School, Chatham Building, Chatham Street, Liverpool L69 7ZH, UK
Abstract
This study investigates in the context of Euronext, if joining an exchange group affects herding both within and across the group’s member-markets and if this phenomenon varies with domestic and non-domestic (group-specific and international) market states and the outbreak of financial crises. All four equity markets of the Euronext (Belgium; France; Netherlands; Portugal) exhibit significant herding following their merger into the group, with herding post-merger either becoming significant (Belgium) or exhibiting an increase in the robustness of its pre-merger significance under various market conditions. Theoverall post-merger significance of herding exhibits variability in its robustness when controlling for the euro-zone sovereign debt crisis. With a few exceptions, cross market herding is overall significant regardless of the period examined. Our findings are robust whencontrolling for abnormal positive and negative market returns.
JEL classification: G12, G15, G31
Keywords:Herding; exchange groups;Euronext;euro-zone sovereign debt crisis; market states.
- Introduction
In the last decade, financial integration has prompted considerable cross-border co-operation among stock exchanges worldwide. Perhaps the most visible form of such co-operation is the forging of regional and global exchange groups, whereby the member-markets of any group share a common institutional framework in their regulatory design and auniform trading protocol based on asingle trading platform. Although this development has given rise to new trading dynamics by facilitating cross-market trading within each group (investors from any member-market can trade more easily in the group’s other member-markets),the implications of thison investors’ behaviour havebeen largely overlooked. In this study we primarily focus on the behavioural pattern of herding which has often been documented in investors’trades globally andhas been found to impact onsecurities’ returns[1].An interesting issue arising in this context is whether membership in an exchange groupproduces an effect overinvestors’herdingboth within and across the group’s member-markets.A second issue iswhether this effect varies with thestate of a market’sdomestic and non-domestic(group-specific and international)environment, while a third issuerelates to whether this effect remains robust when considering the outbreak of the recent sovereign debt crisis in the euro-zone. The latter two issues involve factors i.e. market conditions and financial crises, that have been widely identified in research as relevant to herd behaviour.
From a theoretical perspective,herding refers to individuals choosing to mimic the behaviour of others following observation of each other’s actions and actions-payoffs (Hirshleifer and Teoh, 2003). Depending on the motives underlying imitation,herding can be classified into intentional and unintentional/spurious (Holmes et al., 2013; Gavriilidis et al., 2013). Intentional herding is reflected in correlated actions motivated through the anticipation of (mainly information- or career-related)payoffs in settings characterized by actual or perceived asymmetry. If an investor considers himself to be informationally disadvantaged relative to other investors due to the bad quality of his information source orhis inadequateinformation-processing skills, he has every reason to bridge the informational gap by copying the actions of those he considers better informed(Devenow and Welch, 1996).In the extreme, if the tendency tofollow others instead of focusing on their own private signals becomes widespread among investors, this will give rise to cascades (Banerjee, 1992; Bikhchandani et al., 1992), leading the public pool of information to grow poorer. Career reasons can also motivate herding intent, something particularly relevant to finance professionals, such as fund managers.With fund managers’ performance in the finance sector beingassessedrelative to their peers’(or the industry-average),bad-quality managers are tempted to mimic the trades of good-quality ones to improve their image and protect their career prospects (Scharfstein and Stein, 1990). Correlated actions can also emanate from commonalities in the background and trading conduct of market participants, thus leading to unintentionalherding. Finance professionals, for example, can herd due to relative homogeneity (DeBondt and Teh, 1997) in their background (in terms of their education, the information they receive and its processing;Wermers, 1999) and the regulatory framework they are subject to. Characteristic trading[2]also increases the correlation of investors’ trades, thus generating a false impression of them imitating each other withoutimitation actually being present (their common strategy inevitably leads them to trade similar stocks).
From an empirical perspective, the presence of market herding has been tested internationally in both developed and emerging stock exchangesand for periods of various characteristics with overall evidence being mixed[3].Christie and Huang (1995) report the absence of herding in the US during periods of extreme market returns, both at the aggregate market level and for different sectors. Chang et al.(2000) find significant herding for the emerging markets of their sample onlyafter controlling for market movement, size effects, daily price limits and market liberalization whileHwang and Salmon (2004) report significant herding for the US and South Korean markets irrespective of market conditions and fundamentals indicators. Caparelli et al. (2004)document mixed evidence for herding in the Italian market, while Gleason et al. (2004) document no evidence of intraday herding among US exchange-traded funds.Demirer and Kutan (2006) find no evidence of herding in the Shanghai and Shenzhenstock markets both at the aggregate market and the sector levels after controlling for extreme market returns, the Asian financial crisis and several regulatory changes;conversely,Tan et al. (2008) and Chiang et al. (2010) document the presence of significant herding for both A- and B-shares in Chinese markets[4].Regarding the euro-zone’ssouth European markets (Greece; Italy; Portugal; Spain), Economou et al. (2011) find evidence of significant herding across all of them and also signs for cross-market herding after controlling for periods of positive/negative market performance, high/low volatility/trading volume and the euro-zone sovereign debt crisis. Using an international sample of eighteen markets, Chiang and Zheng (2010) document strong evidence of herding in most of their sample markets (with the exception of Latin America and the US);their results suggest that herding is strong primarily during up-markets inAsian countriesand thatinvestors herd significantly towards the US market, aside from their domestic markets.Blasco et al (2012) find that intraday herding rises with volatility in the Spanish market, while,Gebka and Wohar (2013) find limited evidence of herding at the aggregate market and sector levels across thirty-two markets.
Despite the above wealth of research on herding, very little attention has been devoted to its presencein context of exchange groups, which are products of the global financial integration witnessed during the past two decades. The liberalization of international portfolio flows and the proliferation of cross-listings in the 1990s changed the landscape for stock exchanges, which thereafter started vying internationally for more listings and higher volumes of trade in order to ensure their liquidity and profitability. To enable themselves to compete under these conditions, several exchanges opted for more flexible governance structures – demutualization (Aggarwal and Dahiya, 2006)[5]. The profit-seeking incentives of demutualized exchanges prompted them to commit more resources to advanced technologies and financial innovation in order to handle the increasing volumes of trade and attract more cross-listings (Nielsson, 2009). With thecosts of these investments rising,many marketsentered cross-border alliances with other exchanges in the pursuit of synergies (Hasan and Schmiedel, 2004), thus gradually giving rise toregional and global exchange groups[6] characterized by uniform regulatory frameworks and trading systems.
Membership in an exchange group can affect a market’s herding, since it exposes the market to a wider clientele (traders from the group’s other markets are more likely to consider investing in it due to feasibility or familiarity reasons). If so, the market will be monitored by more investors, thus increasing the opportunity for correlated actions in that market (Andrikopoulos et al., 2014).This possibility can further be facilitated by the lower transaction costs (Schmiedel et al., 2006) and enhanced liquidity (Nielsson, 2009) anticipated in exchange groups, which will render herding more feasible. Membership in an exchange group is also capable of affecting cross market herding within the group itself; if member-markets grow more sensitive to signals regarding other member-markets or the group as a whole, it is likely that the herding of each will exhibit greater correlation with the herding of others, particularly if a group’s markets share similar fundamentals or strong economic links (Dornbusch et al., 2000).Institutional investors(the key players in modern markets) can contribute to this depending on the incentives driving their decision-making. If a manager is risk-averse and at the same time treats the group as a single market in his portfolio, then a negative announcement regarding one of the group’s marketscan lead him to reducehis exposure to the group as a wholesince he might anticipatethe announcement to cast a negative effect over other member-markets of the group.[7]Similarly, if a manager invests in only one of the group’s markets (market A) and his information about that market is imperfect, then a negative announcement regarding another member-market of that group (market B) can prompt him to reduce his exposure to market A, ifhe considers signals from the group’s other member-markets to be informative for his decisions regarding market A.[8]Career reasons are alsorelevant because an imperfectly informed manager may viewhis peers’ trades in a group’s markets asinformative and follow them rather than go-it-alone.
The present paper investigates the effect of exchange group membership onherding dynamicswithin and across the group’s markets. Besides, given the mixed evidence in the literature on the effect of different market states on herding, we also examine how herding varies with the domestic and non-domestic(group-specific and international) environment. Finally, we also analyse the impact of the outbreak of financial crises on herding within a market group. Our sample consists of the four equity markets (Belgium, France, the Netherlands and Portugal) of Euronext – one of the first cross-border exchange-groupsestablished. To proxy for different domestic and non-domestic market conditions, we use the daily (up and down) movement of each market’s average return andtrading volume, the return of CAC40 (the blue chip index of France, Euronext’s lead market),the return of the S&P500, the US VIX values andthe average return of the Euronextgroup (i.e. all four markets) as a whole.[9]Our current research aims to address the following questions:
a)Does membership in an exchange group have an effect on herding both within and across the group’s member-markets?
b)Does this effect vary with different states of the domestic and non-domestic (group-specific and international) environment?
c)Does this effect remain robust when controlling for financial crises?
Our results indicate significant herding for France, the Netherlands and Portugal both prior to and after their merger into Euronext, with herding in Belgium appearing significant post-merger only. These results are relatively robust when controlling for the effect of various states of the domestic and non-domestic environment on our results, with a few exceptions witnessed mainly pre-merger (in France, the Netherlands and Portugal) and less so post-merger (in France). Overall, these results show thatherding is significant in Euronext’s four markets following their merger into the group, with herding post-merger either becoming significant (the case of Belgium) or exhibiting an increase in the robustness of its pre-merger significance to various (domestic and non-domestic) market conditions (the other three markets).Cross market herding is significant for most market-pairs pre- and post-merger; most exceptions are noted for pairs involving Portugal and this may be due to its lower correlation with the rest three markets and its small size/high concentration giving rise to informational dynamics that lead its herding to evolve independently from that of the other markets. As thepost-merger period includes the euro-zone crisis,we split it into a pre- and a post-crisis’ outbreak period(November 2009 onwards) and repeat the post-merger tests for each market based on the splitto test for the crisis’ effect on our findings.Herding overall appears significant in Belgium, the Netherlands and Portugal both pre- and post-outbreak, while France shows significant herding pre-but not post-crisis; these results, however, exhibit variations in their robustness when controlling for different (domestic and non-domestic) market conditions. Cross market herding is significant pre- and post-outbreak, with most exceptions again being noted for pairs involving Portugal.
Our study contributes to the herding literature[10] by investigating the effect of exchange group membership on herding dynamics within and across an exchange group’s markets, and whether this effect varies with different states of the domestic and non-domestic market environment and the outbreak of financial crises. The results reported here are of interest to researchers, regulators and policymakers alike, given the ongoing consolidating activity among stock markets internationally and the relation of herding to systemic risk within, and contagion across, capital markets (Dornbusch et al., 2000). The evidence presented is also of interest to investors with a global investment outlookin view of the negative impact of herding over diversification (Chang et al., 2000). The next section presents an overview ofEuronext;section 3 presents the data and methodology employed alongside some descriptive statistics.Section 4 presents and discusses the results and section 5 concludes.
- The Euronext – group
The birth of Euronext can be traced back to March2000, when the merger of the Amsterdam, Brussels and Parisstock exchanges was announced with the merger of their equity, derivatives and clearing segments coming into effect from September 22ndof that year. The group expanded shortly thereafter when the LIFFE (London International Financial Futures and Options Exchange) and the Lisbon exchange joined it in 2002,[11] while 2007 saw it merging with the New York Stock Exchange to create the first intercontinental exchange in history. With its share in Euronext’s total equity turnover and capitalization hovering around 65 percent, France clearly emerges as the group’s dominant market, followed by the Netherlands, Belgium and Portugal[12].
Trading in the Euronext is based on a common platform modelled after the French Nouvelle Système de Cotation (NSC) – a hybrid system with a limit-order book emphasis. The harmonization of the four constituent equity markets (Amsterdam, Brussels, Lisbon and Paris) with that system came about rather easily, since all four of them shared rather similar trading systems (order-driven, electronic system for continuous trading with market-makers assigned to illiquid stocks)prior to their merger into the group (Nielsson, 2009).Stock trading on Euronextis continuous with a double-auction in placeand lasts from 09:00 to 17:25 Central European Time. The first auction (“start-of-day”) takes place before continuous trading commences and leads to opening prices being set while the second auction (“end-of-day”) takes place at day-close between 17:25 and 17:30 followed by a 10-minute window of trading to allow traders to transact at the “end-of-day” auction’s prices (Beltran et al., 2004). The system allows the placement of traditional (limit/market)and more sophisticated (fill-or-kill[13];must-be-filled[14];iceberg[15]) orders. Blocktrading is allowed for large-volume orders within price-limits set by the market (Beltran et al, 2004). Details on all orders/trades are available to traders pre- and post-trade, with trader-anonymitybeing ensured (Andrikopoulos et al., 2014).
- Data and methodology
The first attempt at estimating herding was that by Christie and Huang (1995) who used to that end the cross sectional standard deviation of returns, formally calculated as:
(1)
Here ri,t is securityi’s return on day t, rm,t is the market’s average return on day t calculated by averaging the return of all securities for day tand n is the total number of listed stocks on day t. Next we estimate herding using the following linear regression model:
CSSDt = α0 + α1DUP + α2DDOWN (2)
DtU(DtL) equals one if the market return lies in the extreme upper (lower) tail of the return distribution, otherwise it is zero. The crux of the model was that herding would be identified during extreme markets through a reduction in the CSSD (denoting a clustering of returns around the market’s average), reflected through significantly negative values for α1 and α2.
The above model contained a few drawbacks, the first one being that the CSSD is susceptible to the impact of outliers (Economou et al., 2011). Secondly, it assumes a linear relationship between CSSD and extreme market periods, ignoring the possibility that the occurrence of herding coincides with non-linear dynamics.Chang et al. (2000) proposed a modified approach over the Christie and Huang (1995)one which is based on the detection of herding through the cross sectional absolute deviation (CSAD) of returns, formally calculated as:
(3)