عکس العمل شدید یا ملایم کوتاه مدت، پیش بینی و یا اجتناب از ابهام؟ مدارک و شواهد از هند

We examine the short-term price behaviour of three, size-conditioned Indian stock market indices, in response to informational shocks. A standard mean-adjusted returns model as well as the GJR-GARCH specification point towards underreaction to negative events in the medium and small capitalization indices. Also, the pre-event coefficients are generally negative and statistically significant, regardless of the sign of the shock, thus ruling out information leaks. We uncover a stable abnormal volatility pattern which increases monotonically a few days before the shock before suddenly decreasing in magnitude on the event day and beyond. We suggest uncertainty avoidance as a potential explanation of these features. The results are fairly robust across alternative event selection procedures, time, and size-conditioned shocks.

The basis of an ever-increasing behavioural finance literature is the idea that psychological biases on the part of investors systematically impact decision making and therefore asset prices in a predictable manner. In other words, at the very core of behavioural finance is the attempt to establish a direct link between documented psychological biases and resilient, (ideally) arbitrage-prone price patterns in financial markets. In this vein, underreaction and overreaction, two pervasive regularities (Barberis et al., 1998) are often at the forefront of the anomalies literature. Conceptual conservatism defined as the human tendency to cling to existing beliefs in the face of challenging new evidence (Nissani, 1994) partly underscores the idea of underreaction to events. Similarly, representativeness, defined as the tendency to view events as representative of a specific class, partly underscores the idea of overreaction. Less vehiculated is perhaps the idea of uncertainty avoidance, defined as the extent to which a culture programs its members to feel uncomfortable/comfortable in unstructured or surprising situations (Hofstede, 2001). The underlying dimension here is the tolerance for ambiguity, which can be found in individuals and which, in identical situations, leads some to feel more pressed for action than others (Hofstede, 2001). Cyert and March (1963) claim that people in uncertainty-avoiding cultures emphasize short-run reaction to short-run feedback rather than anticipation of long-run uncertain events, and solve pressing problems rather than develop long run strategies. This study proposes to analyze these ideas in the context of the Indian market over the period 2003–2010. Recently, this approach has attracted renewed attention but its research has been limited to either various developed markets (Spyrou et al., 2007 and Rieks and Lobe, 2008) or a large group of countries (Mazouz et al., 2009). There are very few (if any) studies focusing solely on key emerging markets. The starting point in these studies is the idea that a broad market index will adjust quickly and fully to shocks, regardless of their magnitude. As such, no persistent patterns, in the form of cumulative abnormal returns, are to be observed around market event days. The results generally point towards some form of market inefficiency. Spyrou et al. (2007) provide evidence that medium and small capitalization indices significantly underreact to both positive and negative shocks for several post-event days. They claim that this underreaction is still unaccounted for after the usual risk-factors, calendar effects, and bid-ask biases are considered. Their interpretation is that investors process extreme negative news optimistically and extreme positive news pessimistically, an idea raised in the past (Schnusenberg and Madura, 2001). Lasfer et al. (2003) also document short-term underreaction following both positive and negative news. They also differentiate between developed and emerging markets in terms of the amplitude of such a response. While studies such as Spyrou et al. (2007) compute the cumulative abnormal returns (CARs) based on a simple mean-adjusted returns model, Mazouz et al. (2009) employ both OLS regressions and an asymmetric GARCH in order to investigate the CARs following size-conditioned market shocks. The results differ across methodologies. As such, the OLS method provides strong support for return continuations following positive and negative market shocks with absolute values between 5% and 10%. There is also country-specific evidence of overreaction (return reversals) for very large positive and negative shocks. However, when a GJR-GARCH model is employed, market efficiency seems to prevail, particularly for medium and large shocks. Whatever research has been done on India on under/overreaction, it generally follows the popular approach of Jegadeesh and Titman (1993) and investigates the possibility of momentum and overreaction in class-conditioned stock returns (winners vs losers) over 3–24 months, without concern for short-run reaction to market shocks. As such, Rastogi et al. (2009) find that over a period of 3–12 months, there is evidence of underreaction and momentum profits in all size-conditioned portfolios. The only analysis of the Indian market's immediate response to market shocks has been so far only indicative, being part of a much broader study of global or regional behaviour. Within such context, Mazouz et al. (2009) uncover little evidence of market inefficiency except perhaps as a response to very large negative shocks (beyond 10% in absolute value), to which the market appears to overreact.1 However, this result is not conditioned on the average market capitalization of the stocks making up the index. Also, the study investigates a dataset which ends in December 2005, thus avoiding a period of unprecedented growth in the Indian market as well as the impact of the recent global financial crisis. Equally, the scope of the study is to document the presence of under/overreaction across 10 major Asian countries conditioned on the size of the shocks, without concern for the particulars of such phenomena in every country. This leaves unanswered a series of questions which the present study attempts to address. The choice of India is somewhat self-explanatory. India's economy is the eleventh largest in the world by nominal GDP and the fourth largest by PPP (purchasing power parity).2 For 2009/2010, India's growth stands at 7.4%, surpassing the previous year growth rate of 6.7%.3 Its growth is expected to reach 8–9% in the next two years, raising prospects that it may overpass that of China. The recent global financial crisis had a contained impact on the Indian economy. There were some outflows of capital in late 2008 but the trend was reversed in early 2009. If anything, the growing current concern in the region is the massive inflow of foreign capital in search of a quick return. All in all, India is the perfect example of a successful emerging market which, in less than two decades, has reversed its fortunes and set itself on the path of sustainable growth. Therefore, news assimilation in this buoyant stock market is of relevance. The current study investigates the Indian equity market by reference to three of its most representative, liquid and size-conditioned, indices: BSE Sensex-30, BSE Mid Cap, BSE Small Cap. We analyze their price behaviour in periods immediately surrounding significant price shocks. To enable comparison of the results with the like in Spyrou et al. (2007), we initially employ a mean-adjusted returns model to arrive at the cumulative abnormal returns (CARs). Often results in this type of studies are conditioned on the particular methodology employed. For this reason and in an attempt to cater for volatility clustering in the data, we also apply an asymmetric GARCH (GJR-GARCH) in a similar manner to Mazouz et al. (2009). We depart from previous studies by introducing pre-event CARs in our analysis, to enable us to better describe market behaviour ahead of (as well as after) shocks. Here, we are particularly interested in gauging the extent (if any) to which the market anticipates events or processes information leaks. Given that the customary sample selection cannot distinguish between expected and unexpected events, this provision is almost necessary. Exhausting data modelling, we turn to press headlines, in order to further investigate the validity of our interpretations. As we consistently uncover a stable pattern in pre-event returns, we gather data on net equity inflows of some of the largest players on the Indian equity market (foreign institutional investors and domestic mutual funds).4 We test for the relationship between the sign of the equity flows of these players and the sign of the market shock as delivered by jumps in Sensex (as a barometer of the Indian market). The overall results point in an interesting direction. We uncover a sustained pattern of negative, significant, pre-event CARs across shocks, indices and time. This is accompanied by a resilient abnormal volatility pattern which increases as the event day approaches, suffers a downward correction on the event day itself, and then declines in magnitude over the following 10 post-event days. The sign of the pre-event CARs (consistently negative) rules out the hypothesis of information leaks or market informed anticipation. Therefore, this result, corroborated with feedback from market analysts as reflected by the financial press on the day of the shock, suggests consistent short-run inefficiency in information processing. We suggest uncertainty avoidance as an explanation of such pattern. In periods of high uncertainty, a representative proportion of market players are compelled to act and this drives up pre-event volatility and can deliver negative returns. Indeed, we show that pre-event trading volume (for both index shares and index futures) is typically higher than normal (benchmark) trading volume and the null of equal volumes is statistically rejected in anything from 40% to 60% of all events. We also document a unidirectional Granger causality running from volume to volatility. This pre-event market dip is often followed by a sharp price correction, either up or down, as the uncertainty subsides. We also show that this hypothesis can be reconciled with the suggestion in Hofstede (2001), indicating India as a country with a relatively high tolerance for ambiguity. In the following we present the details of our methodology, followed by an interpretation of the empirical results and conclusion.

The purpose of this study is to shed some light on the way investors on the Indian stock exchange react to market events. We are conditioning our investigation on several factors such as market capitalization, crisis vs non-crisis periods, and magnitude of shocks. We employ two different approaches, namely a CAR analysis based on a mean-adjusted returns model and a GJR-Garch model. We find evidence of post-event underreaction to negative events, detected mainly in the medium and smaller capitalization stocks, in all periods except post-crisis. During post-crisis, the market appears to over-react to bad news. On the other hand, the market seems to react more efficiently to positive events, overall. There is, however, weak evidence of overreaction to good news before the 2008 crisis which turns into underreaction to good news after the crisis. This conclusion is endorsed by the asymmetric GARCH approach. Equally, the pre-event CARs tend to be negative and statistically significant, independent of the particular methodology employed. Moreover, their corresponding abnormal volatility coefficients are significant, positive, and monotonically increasing up until the day before the event day and then monotonically decreasing from time 0 until up to 10 days following the event. We make use of the fact that, in the case of BSE Sensex, the analysis can be significantly extended beyond the original period, in order to subject our results to a time-test. Indeed, the results for Sensex over the period 1991–2010 validate the previous pattern. Moreover, conditioning events on the size of the shock does not alter the results. If anything, larger shocks give rise to larger pre-event volatility estimates as in Mazouz et al. (2009). We also document a uni-directional Granger causality between volume and volatility in the sense that volume does Granger-cause volatility but not the other way around. Also, we document a statistically significant and abnormally high trading volume in a large proportion of pre-event days, suggesting an increased traders’ appetite for selling (as returns tend to be negative and significant on these days). This result is confirmed by two alternative datasets: the total volume of Sensex shares being traded on pre-event days as well as the total number of Nifty futures contracts being traded on pre-event days. Moreover, a simple trend analysis looking at daily changes in price, volume, and open interest, suggests selling tendencies (for hedging, speculation, or long covering purposes) in pre-event days, all of which are indicative of market unrest. We suggest this pattern can be interpreted by reference to one particular psychological bias, namely uncertainty avoidance. In periods of heightened uncertainty, a representative group of market participants having low uncertainty tolerance are compelled to act (trade), driving up volatility and potentially driving down returns. As more information is released, uncertainty will subside and the market will undergo a correction (up or down) after which it remains largely efficient. Given that, on an ex ante basis, it is impossible to predict the individual events and their nature, this still conforms to a broad notion of market efficiency. We reconcile this hypothesis with the suggestion in Hofstede (2001) that the Indian market scores below average on a world uncertainty-avoidance scale.