Market Efficiency changes due to the Global Financial Crisis?
A comparison for Banks in Emerging Markets and the United States
EVA R. KOOL*
ABSTRACT
This study examines the impact of the global financial crisis on the equity market efficiency of banks traded in both emerging markets and in the United States. It is expected that the degree of market efficiency improves due to the crisis. Stock return data from 2006 to 2012 are used. An event study on substantial positive or negative shocks in prices is conducted, followed by a regression using least squares estimates. The results are in line with the overreaction theory of behavioural finance, contradicting the efficient market hypothesis. The analysis shows that the degree of market efficiency in emerging markets does not significantly differ from the United States. Moreover, the degree of market efficiency coincides with the global financial crisis only for negative price shocks.
Search topics: efficient market hypothesis, global financial crisis, emerging markets, price shock, event study, banking sector
JEL Codes: G14, G01, G15
*
THE EFFICIENT MARKET hypothesis is challenged by criticism from the academic and the financial world, the main criticism deals with the inability to develop over time and invalidity of the assumptions. Therefore, the development of the market efficiency over time is an interesting topic for research. The efficient market hypothesis assumes that security prices are efficient forecasts of future prices, in the sense that they reflect minimum-variance price forecasts. Security prices are predicted from past prices and should price in all relevant information. This condition can, however, only be true for a given level of publicly available information (Ball, 2009). In May 2011, Richard Posner argued that there has been an unjustified faith in the efficient market hypothesis and the rational expectations theory. He renounced his faith in the efficient market hypothesis and argued that the Keynesian theory and the behavioural finance theory are soon to be the dominant theories (Cassidy, 2010). However, others argue that the financial markets have been unable to function efficiently due to distortions in the information markets. Crombez (2001) argues that pricing assets correctly is harder for most market participants due to increasingly complex financial products and private information. Furthermore, several empirical studies prove the existence of market anomalies that counteract the efficient market hypothesis. These anomalies recognize flaws in the functioning of the theory, such as underreaction or overreaction to information and price persistence (DeBondt & Thaler, 1985). The implicit allowance of asset bubbles to build up and burst is a specific concern. History has shown practical evidence on this phenomenon. Evidence on anomalies to the efficient market hypothesis and rational expectations theory raise practical as well as empirical questions on the functioning of the financial markets. Therefore, other studies test the development of market efficiency and show that market efficiency develops over time. These studies identify events that coincide with improvements in market efficiency. Amongst other determinants, they show departure from market efficiency in times of global crisis (Narayam and Islam (2012), Smith (2012) and Todea and Lazar (2012)).
The purpose of this research is to identify movement in the degree of market efficiency in the banking sector of emerging markets, comparing the period prior to the global financial crisis to the global financial crisis. The results are compared to the market efficiency of the United States, where I assume the United States to be most efficient. This research measures market efficiency by the short term price reaction following a predefined shock in prices. As a result, the results contribute to the existing literature on market efficiency with three distinct features:
Second, the focus of my research will be on the market efficiency of equity markets of listed banks, not on specific country equity indices. Most research is conducted on equity indices of countries, rather than specific industries. When focussing on a specific industry, the banking industry is interesting because the banking industry is the centre of attention in the current global financial crisis. The global financial crisis is partly caused by banks and therefore the market efficiency of that specific industry is especially relevant.
Third, this study focuses on emerging markets compared to the United States. A large body of literature compares developed markets to emerging markets in their degree of market efficiency. Most empirical evidence suggests that emerging markets are less informational efficient, meaning that the reaction of stock markets to the received information is slower, or not present at all. However, the evidence is mixed (Cajeuiro and Tabak (2004, 2005), Griffin et al. (2006), Kim and Shamsuddin (2006) and Risso (2009)).
Combining the three distinct features of this research, three questions about the market efficiency of emerging markets are asked and answered, following to be the three hypotheses. Are emerging markets less informational efficient than the United States? Does the global financial crisis positively impact the market efficiency in equity markets? Last, is the effect of the global financial crisis on market efficiency stronger in emerging markets?
The implications of this research are relevant for academics, regulatory authorities, managers and investors. In academics, empirical evidence on this topic has implications for financial theory and investment strategies and contributes to the current debate on the validity of the market efficiency theory. Moreover, the regulatory framework is affected by the understanding of different determinants influencing market efficiency. The relevance to managers in the banking industry relates to gaining information on the response of investors in their market. Knowing whether prices of stocks accurately reflect the underlying value of the stock is of importance not only to managers but also to investors. To investors, the arbitrage opportunities resulting from market inefficiencies in emerging markets might be a source of potential profits.
freely available to all market participants; there are no imperfections in the information market in this methodological setup. Other studies conduct a similar methodology, studying the short term price behaviour to events that are defined by sudden shocks in prices. This research adds to this body of literature because I include the global financial crisis (Alrabadi (2012), Cox & Peterson (1994), Mazouz et al. (2012) and Park (1995)). The sample I use consists of 12 emerging countries and one developed country over the period of 2006 to 2012.
The evidence on the event study and cross-section regression does not support the main hypotheses. Prior to the crisis, the degree of market efficiency between emerging markets and developed markets does not differ. Moreover, the results show that the global financial crisis coincides with an increase in market efficiency only for events that are negative price shocks. The evidence supports one of the anomalies to the efficient market hypothesis, namely the overreaction theory from behavioural finance. The investors in emerging markets as well as in the United States significantly overreact to information, resulting in a larger than appropriate price shock. Moreover, it seems that investors are more likely to overreact to negative news in the pre-crisis period.
The remainder of the paper is organised as follows: the first sections gives an overview of the existing literature and justifies the hypothesis. The second section provides a detailed description of the methodology followed by the data selection and description section. The fourth section provides the results of the analysis whereas the last section concludes the paper.
I. Theoretical Framework
The efficient market hypothesis serves as the underlying theory for this study. First, the efficient market hypothesis will be discussed, since previous literature tests and explains market efficiency in various ways. Second, various theories and evidence that challenge the efficient market hypothesis will be reviewed, as well as the impact of the financial crisis on markets. Finally, the hypotheses that are tested in this research are explained.
A. Efficient Market Hypothesis
The efficient market hypothesis is one of the fundamental theories on the functioning of financial markets. Currently, the effectiveness and validity of the theory is hotly debated in both the academic and corporate world. The characteristics, the implications and limitations are discussed first.
A.1. Theory and measurement
claims that the price of a security at any point in time is a reflection of the present value of the optimally forecasted future risky cash flows. Moreover, the efficient market hypothesis claims that investors learn to make correct inferences about the impact of new information (Brown, Harlow and Tinic, 1988).
Roberts (1967) makes a distinction between three versions of the efficient market hypothesis, aiming to identify the level of information at which the hypothesis breaks down. The three testable versions of the efficient hypothesis are weak form efficiency, semi-strong form efficiency and strong form efficiency. In weak-form efficient markets the future prices cannot be forecasted from past prices because information arrives randomly. Therefore, current prices are the best estimators of future prices. Semi-strong form efficient markets reflect all publicly available information in the stock prices. In strong-form efficient markets all public and private information is reflected in the stock prices. When talking about the efficient market hypothesis we usually refer to the semi-strong form of market efficiency. Reaching strong-form market efficiency is an unreachable benchmark, since complete market efficiency requires a perfect market. Moreover, the concept of perfect markets is not economically sensible since there are always trading and information costs in a market (Grossman and Stiglitz, 1980).
Moreover, the efficient market hypothesis states that in developed markets prices are efficient reflections of the value of the underlying asset. The efficient market hypothesis implies that in such cases the prices that are set in the financial markets are fair to all. Therefore, the key question to investors is in which markets the financial markets function efficiently. Accordingly, it is important to measure market efficiency and research the determinants of market efficiency. The key determinants of market efficiency will be discussed below, followed by the tools of measurement.
There are various studies that research the possible determinants of market efficiency. Lagoarde-Segot and Lucey (2008) investigate the informational efficiency of seven emerging middle-eastern North African stock markets. The determinants they considered in their research are market development, corporate governance and the overall degree of institutional and economic liberalization. Their results show that the weak-form market efficiency is mostly influenced by differences in stock market development and corporate governance. In addition, Lim and Brooks (2007) conduct research on the potential determinants of the degree of market efficiency over the period 1995-2005. They conclude that stock markets are more efficient in countries that have liberalised stock markets, have a high degree of institutional collectivism, have a good corporate governance structure, allow short-selling and have a high number of security analysts. Concerning the economic variables, the countries with higher market efficiency tend to have higher GDP, trade openness and a lower level of inflation.
elements of institutional characteristics, regulation and macro-economic environment. It is valid to assume that due to changes in the various determinants of market efficiency the degree of market efficiency evolves over time.
Concerning the tools of measuring market efficiency, the most commonly tested aspect of the weak-form market efficiency is the unpredictability of stock returns, based on past prices. In these studies, the notion of efficiency is defined as the unpredictability of stock returns, measured by the random walk theorem. The central idea is that the available information is not restricted and will be immediately reflected by stock prices. This means that the stock prices randomly fluctuate since the information enters the market randomly. A second strand of literature testing weak-form efficiency of financial markets is pursued by event studies. The classic event study methodology is introduced by Fama, Fisher, Jensen and Roll (1969) and has become a standard for measuring the direct price reaction of a security to the announcement, or occurring, of an event.
To conclude, due to various reasons, such as institutional change, regulatory and technical changes, the degree of market efficiency can change over time. The evolving of market efficiency over time is a product of the determinants of the market efficiency degree.
A.2. Empirical evidence
There is an extensive body of literature on the market efficiency of specific countries, in both developed and emerging economies. In the case of developed countries Evans (2006) and Groenewold (1997) show evidence for efficient financial markets. Many of the studies compare developed markets and emerging markets, to assess the relative market efficiency. The evidence from previous studies on the market efficiency for emerging markets is mixed, varying for different countries and methodologies.
on various methodologies, they conclude that there is no evidence that a better regulatory environment, legal structure and governance structure lead to higher market efficiency. They conclude that, with respect to past prices and earnings announcements of public information, emerging markets are at least as efficient as developed markets.
Concluding, the conventional literature (Cajeuiro and Tabak (2004, 2005), Kim and Shamsuddin (2006) and Risso (2009)) find evidence for a lower degree of market efficiency in emerging markets, whereas Griffin et al. (2006) find contradicting results. The results on market efficiency are therefore mixed. It should be noted that these studies are conducted using the market indices in the countries. In the context of this research, which is specifically focused on banks, the literature presents no specific evidence.
B. Anomalies to the efficient market hypothesis
Many existing studies attempt to identify abnormal profit opportunities and thereby to find anomalies to the efficient market hypothesis. Research shows that in practice, investors have been able to select stocks that generate abnormal returns, posing challenges to the efficient market hypothesis. The two main strands of literature concerning these anomalies are investors’ reactions in behavioural finance and price persistence in the accounting literature.
B.1. Behavioural finance
The literature on overreaction and underreaction, predicts overreaction or underreaction to new information. Since the underlying basis of the information is unknown in this research these price shocks could be the results of overreaction or underreaction by irrational investors. If this is the case there will be direct price reversals following the price shock, indicating negative abnormal returns directly following a positive event and positive abnormal returns directly following a negative event.
B.2. Price persistence
According to the efficient market hypothesis prices directly reflect the information that becomes available. Since the past stock prices should not provide information about future prices, prices should not persist. Price persistence is the tendency for rising asset prices to keep rising whereas falling asset prices have the tendency to keep falling. Current studies explain the price persistence mostly by behavioural models, assuming that investors are irrational (Barberis et al. (1998) and Daniel et al. (1998)). Other studies explain price persistence on the basis of underreaction to specific assets. Mazouz et al. (2012) shows that the price reaction to stocks over 1992-2007 persists. Moreover, they show that stocks in their sample with low systemic liquidity risk respond efficiently whereas stocks with high systemic liquidity risk tend to underreact to shocks. The underreaction is shown by a long persistence of price increases or decreases.
Besides the assumptions on irrationality of agents, Crombez (2001) concludes that imperfections can arise due to distortions in the information market, for example the availability of information and consistency of information sources. Assuming investors rely upon historic price changes and information presented by market experts, the differences between the opinions of experts result in a slower incorporation into the asset prices. These results indicate that anomalies contradicting the efficient market hypothesis may relate to imperfect availability of information, not necessarily irrational expectations.
Concluding, previous studies show market anomalies to the efficient market hypothesis. My research focuses on price shocks, meaning that the information is available to all market participants, since prices are publicly available. Price persistence would be more likely when the price changes are based upon information from expert agents and when the information market is not perfect.
C. Global Financial Crisis
information increases and the market becomes more efficient. As previously discussed, empirical studies show mixed evidence on this hypothesis (Lim, Brooks and Kim, 2007). Third, a possible contributing factor to changing market efficiency is the occurrence of a market crash or financial crisis. The body of existing literature on this topic is relatively small and provides mixed results. Lim, Brooks and Kim (2007) empirically investigate the impact of the Asian financial crisis on the stock market efficiency of the related Asian countries. They find that the large collapse of the stock prices was followed by investor’s panic, which adversely affected the market's ability to price stocks efficiently. Hoque et al. (2007) examines the effects on stock market efficiency in the Asian financial crisis (1990-1997) and post-crisis periods. Their results are mixed, since six out of the eight countries remain to have inefficient stock markets. Nevertheless, Taiwan has shown to have increased market efficiency between the crisis and post-crisis period. Moreover, Kim and Shamsuddin (2006) find similar results in their research on the Asian financial crisis for both developed and emerging markets. During this crisis, only Taiwan and Singapore experience improvements in the degree of market efficiency.
D. Short term price responses to shocks
Determining the evolvement of market efficiency over time is relevant to the current debate and financial theory. Many studies show a significant lower degree of market efficiency for equity indices in emerging markets (Cajeuiro and Tabak (2004, 2005), Kim and Shamsuddin (2006) and Risso (2009)).There is no specific research on stock markets of banks. Assuming there are no significant differences between the indices and the stocks of banks the following hypothesis is formulated.
HYPOTHESIS 1: Compared to the United States, the equity markets of banks is less informational
efficient in emerging markets
As various studies describe the determinants of evolvement of market efficiency, a market crash or financial crisis appears to be one of the factors impacting the degree of market efficiency. The global financial crisis started in the United States but spread to all other countries in the world. Most empirical evidence shows a positive relationship between the global financial crisis and the degree of market efficiency (Narayan and Islam (2012), Smith (2012) and Todea and Lazar (2012)). This leads to the following hypothesis.
HYPOTHESIS 2: The global financial crisis positively affects the market efficiency of equity
markets in the banking sector
The results presented in the literature by Narayan and Islam (2012), Smith (2012) and Todea and Lazar (2012) provide evidence on the coinciding effect of market efficiency when the global financial crisis hit. They show that in the comparison between developed and emerging markets the emerging markets tend to react stronger to the global financial crisis in terms of market efficiency. Since the developed markets already operate in highly liberalized markets with appropriate regulatory framework the impact of the financial crisis on the degree of efficiency in emerging markets is expected to be stronger.
HYPOTHESIS 3: The effect of the global financial crisis on the market efficiency in the banking
II. Methodology
The methodology employed in my research is an event study methodology based on the studies of Brown and Warner (1980, 1985) and MacKinlay (1997). Event studies serve the purpose of testing the null hypothesis of markets efficiently incorporating information. Moreover, under this hypothesis event studies examine the impact of an event on a firm’s wealth (Binder, 1998). A short term event study methodology captures the price reaction in the days directly following the event. In case of weak-form market efficiency the current prices are not affected by past prices, meaning that the prices of today cannot be based upon the prices of yesterday. Many previous event studies limit their scope to a specific range of events. By defining an event based on a predefined level of a price shock, the scope of the research is much broader. Moreover, the underlying cause of the event is unknown, creating an equal foot comparison of market efficiency. This research focuses on the short-term price response to any substantial event, testing the market efficiency of a specific stock market.
A. Event definition
The definition of a price shock is based upon the movement in prices of individual securities. Previous literature focuses on various levels of price shocks, varying from 5% shocks in prices such as Alrabadi (2012), to 10% price changes by Atkins and Dyl (1990) and Bremer and Sweeney (1991) and 20% by Mazouz et al. (2009). The level of the price shock should be high enough to capture those price movements that represent a substantial change in fundamentals or market beliefs. For this study, the definition of an event is based upon 20%, or more, positive or negative price shocks in individual stock returns. In other words, daily price changes of individual securities in excess of 20%. To avoid confounding effects when the event windows would overlap, any shocks within five days following a given shock are excluded from the event list. The data is divided into a panel of positive price shocks and a panel of negative price shocks. Separating these two types is necessary because the overreaction to both positive and negative shocks may cause the results to become zero on average, leading to a false perception of market efficiency.
B. Estimation and event windows
estimation window should be sufficient to appropriately estimate the alpha and beta, of the market and risk adjusted model. The estimation period will range from 100 days prior to the price shock to the day of the price shock [-100,-1]. The estimation window is used to estimate the market parameters for the different methods of calculating the abnormal returns, estimation of the normal movements in the stock returns and the correlation with the market stock index, MSCI World (Brown and Warner, 1985).
C. Model of normal returns
In order to specify the abnormal returns a model of normal returns must be specified. Three different models will be used to estimate normal returns: the mean adjusted model, market adjusted model and the market and risk adjusted model (Brown and Warner, 1980, 1985). The results of these models, the returns aggregated over time and firms, are used as a reference to identify the abnormal returns within the event window.
The daily returns are calculated by the logarithm of the closing price of the event day, minus the logarithm of the closing price of the previous day.
(1)
Where, is the actual rate of return on security i on time t, and are daily closing prices for security i on time t and t-1. The shocks, events, are defined on the basis of the returns calculation.
The first model to estimate the normal returns is the mean adjusted model. This model incorporates the average of the returns on security i to determine the normal returns. Given by:
(2)
(3)
Where is the abnormal return on security i at day t, is the realised returns on security i at day t and is the mean of the realised returns on security i. The estimation window is [-100,-1] days.
The second model is the market adjusted model. In this model the daily rates of return on each security are compared to the daily return of the market index. The market index that is used is the MSCI World index.
(4)
Where is the abnormal return on security i at day t, is the realised returns on security i at day t and is realised return of the MSCI World index, on day t.
the analysis will be based on this model. The returns based on the market model parameters are calculated by: (5) Given, (6) (7)
Where is the abnormal returns on security i at time t, is the realised return index on security
i at time t, is realised return of the market index, MSCI World, on day t, and are the market
model parameters.
When comparing the days following the event over the full sample of stocks the average abnormal return (AAR) is used. Calculated by,
(8)
Where is the abnormal return in security i on day t and n is the total amount of events in the panel of positive or negative shocks.
The sum of abnormal returns over the event window are calculated using the cumulative abnormal returns (CAR), which is the sum of the abnormal returns over the event window, where different lengths of event windows can be used. I will be using an event window of three days and five days following the event.
(9)
Where or are the cumulative abnormal returns for respectively day 1 to day 5 and day 1 to day 3 and is the abnormal returns on security i at time t.
D. Tests
significantly differs from the hypothesised value of zero. In testing the abnormal returns a significant, positive, abnormal return indicates that the returns following the event are higher than the benchmark returns. The benchmark returns are estimated using one of the three models of normal return. When the abnormal returns are negative and significant it means that the returns following the event are lower than the benchmark model. When the abnormal returns are insignificant they are not significantly different from zero, thus in line with the efficient market hypothesis. In the context of my research, the smaller the abnormal returns following a price shock, the faster the adjustment of the available information and thus the higher the degree of market efficiency.
Moreover, the first and second hypotheses are tested comparing the subgroups over time and over the two markets. In order to test whether the abnormal returns in the subgroups significantly differ from one another I will use an independent samples t-test and a Wilcoxon rank sum test/Mann-Whitney. The Wilcoxon rank sum tests and Mann-Whitney test are similar tests and are the non-parametric equivalent to the independent samples t-test. I will refer to the test as the Wilcoxon rank sum test/Mann-Whitney. When the t-test and Wilcoxon rank sum test/Mann-Whitney are significant this indicates that the subgroups significantly differ, compared over the mean and median. The sign of the differences must be interpreted to conclude whether or not the hypotheses are supported.
E. Cross-section regression
The following regression equation estimates the relationship between the cumulative abnormal returns and the above described factors. The regression is estimated on the panel of positive events and the panel of negative events, using each of the three models for normal return. Since the market and risk adjusted models is the most advanced model the results of the two other models will be included as robustness checks.
(10)
Where is the cumulative abnormal return over day 1 to day 5 for event i, is the dummy
variable for emerging markets, indicating 1 if the bank is located in an emerging market and 0 when located in the United States. is the dummy variable for the time effects, indicating 1 if the event occurs in the crisis period and 0 outside the crisis period and is the error term for event i.
III. Data and descriptive statistics
A. Sample description
The sample consists of listed banks in both emerging markets and the United States in the pre-crisis period and during the global financial crisis. The total sample covers seven years, from the 1st of January of 2006 to the 31th of December in 2012. For comparative purposes the sample will be composed out of two sub-samples: The first sub-sample is the pre-crisis period consisting of daily returns from 1st of January 2006 to 14th of September 2008. The second sub-sample, crisis period, consists of daily returns from 15th of September 2008 to 31th of December 2012. The structural break in the sample is marked by the fall of Lehman Brothers. At the 15th of September 2008 Lehman Brothers filed for bankruptcy in the U.S. This event appeared to be the starting of the exacerbated spill-overs to other markets and therefore I indicate the fall of Lehman Brothers as the starting point of the global financial crisis.
The data will be collected on a daily basis. The countries must have regular trading records. If there were to be irregular trading, the speed of adjustments to information will not be measurable.
B. Data selection
The data consist of the total return index of individual listed banks in the secondary emerging markets and in the United States in the time period of 2006 to 2012. The total return index assumes dividend is re-invested in the particular stock. The list of banks is retrieved from Bankscope. The selection criterion is that banks are listed, whilst both active and inactive banks are included in the sample. When only including active banks in the sample this might cause a bias towards surviving banks. Some banks are excluded from the sample due to overlapping ISIN and SEDOL numbers. Those banks that are listed on over-the-counter bulletin markets and banks that are listed on other over-the-counter markets are excluded. Moreover, the return indices are retrieved from Thomson
Datastream. Due to data availability problems with the data in the estimation period 50 banks have
been excluded. In total, 660 banks remain in the sample. From the sample of 660 banks, 405 banks are located in the United States. The other banks are distributed amongst the 12 emerging countries. The analysis is based on the group of emerging market countries rather than individual countries. The reason for this is that some of the emerging market countries have too little banks in the sample. If the individual countries would be compared this would cause non-generalizable results.
C. Data properties
Table I.
Distribution of events per period and market
The table presents the distribution of events categorised by time period and market. The structural break in the sample is the fall of Lehman Brothers on the 15th of September 2008. Positive shocks indicate price increases of 20% or more and negative
shocks refer to a fall in price of 20% or more. The mean and median of the shocks are given. The total sample consists of 1967 events.
PRE-CRISIS (01-01-2006 to 14-09-2008)
# events mean median
Emerging markets
Panel A - positive 64 35.18% 25.36% Panel B - negative 77 -34.53% -26.83%
Total Emerging markets 141
United States Panel A - positive 66 24.38% 23.09% Panel B - negative 51 -28.45% -25.97% Total U.S. 117 CRISIS (15-09-2008 to 31-12-2012)
# events mean median
Emerging markets
Panel A - positive 237 28.10% 24.84% Panel B - negative 198 -30.75% -25.95%
Total Emerging Markets 435
United States
Panel A - positive 650 26.65% 23.86% Panel B - negative 624 -26.34% -23.75%
Total U.S. 1274
As can be seen from Table I, there is a considerable difference in amount of shocks in the two time periods. In total there are 1709 price shocks that occur in the crisis period and only 258 in the pre-crisis period. Looking at the size of the shocks, the differences between the positive and negative shocks for each of the sub periods and markets is relatively small. For example, positive shocks in the pre-crisis period in emerging markets are on average 35.18% whereas the negative shocks in this subgroup are – 34.53%. Looking at the differences between the two markets for the pre-crisis period there are substantial differences. The average positive shock for emerging markets in the pre-crisis period is 35.18% whereas the average positive shock in the United States is only 24.38%. In the crisis period the between differences are smaller, the average positive shock in emerging markets is 28.10% whereas the average positive shock in the United States 26.65% is. Finally, when comparing the markets over time there is a noticeable result for emerging markets. The average positive shock in the pre-crisis period is 35% whereas the average positive shock in the crisis period around 28% is. The differences in United States over time are not substantial.
normality gives mostly significant results, indicating that the null hypothesis of the kurtosis and Skewness jointly being zero should be rejected. Also, when looking at the figures on Skewness and kurtosis it is seen that the data is mostly not normally distributed. Therefore I employ a non-parametric test besides the t-test.
V. Results
The analysis consists of three parts: first, the cumulative abnormal returns on the days following the events. Second, the analysis of the differences between the subgroups, namely pre-crisis or crisis and emerging markets or the United States and finally, a regression analysis on the cumulative abnormal returns. The cumulative abnormal returns are presented to give an indication on the market efficiency in the various subgroups. The statistical testing of the differences between the subgroups and the regression analysis are to test the hypothesis. Moreover, the cumulative abnormal returns are reported rather than the daily abnormal returns, this is done to give a concise overview. The more detailed daily analysis of each of the three components can be found in appendix B in Table B.II. until Table B.IV.
A. Cumulative abnormal returns
Table II.
Cumulative abnormal returns on event window [1,3] and [1,5]
The table presents the cumulative abnormal returns for each of the methods of normal returns in percentages. Two event windows are specified for both the panel of positive events and the panel of negative events. Significance is tested by a one
sample t-test and a non-parametric Wilcoxon signed rank test*.
A. Panel positive events (n=1017) B. Panel negative events (n=950)
PRE-CRISIS (01-01-2006 to 14-09-2008)
mean adj market adj market &
risk adj mean adj market adj
market & risk adj Emerging markets CAR [1,3] -3.38% -3.74% -3.21% 10.97% 11.18% 11.10% t-test (0.265) (0.221) (0.294) (0.000) (0.000) (0.000) Wilcoxon (0.073) (0.031) (0.084) (0.000) (0.000) (0.000) CAR [1,5] -2.99% -3.34% -2.85% 14.48% 14.57% 14.49% t-test (0.368) (0.317) (0.397) (0.000) (0.000) (0.000) Wilcoxon (0.097) (0.066) (0.168) (0.000) (0.000) (0.000) United States CAR [1,3] -1.67% -3.28% -1.43% 9.27% 8.11% 9.39% t-test (0.316) (0.039) (0.414) (0.001) (0.001) (0.002) Wilcoxon (0.361) (0.035) (0.335) (0.000) (0.000) (0.000) CAR [1,5] -0.23% -2.69% -0.12% 11.53% 9.45% 11.92% t-test (0.912) (0.160) (0.956) (0.000) (0.000) (0.000) Wilcoxon (0.618) (0.079) (0.455) (0.000) (0.000) (0.000) CRISIS (15-09-2008 to 31-12-2012) Emerging markets CAR [1,3] -4.66% -5.09% -4.93% 6.63% 6.52% 6.53% t-test (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Wilcoxon (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) CAR [1,5] -5.26% -6.14% -5.55% 8.01% 7.33% 8.00% t-test (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Wilcoxon (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) United States CAR [1,3] -3.32% -5.29% -2.79% 8.60% 6.99% 8.95% t-test (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Wilcoxon (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) CAR [1,5] -3.02% -5.90% -2.80% 10.41% 8.32% 9.73% t-test (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) Wilcoxon (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
*p-values in parenthesis, bold are significant at 5% level based on t-test and Wilcoxon signed rank test, grey shaded coefficient is only significant at 5% level for the Wilcoxon signed rank test.
Table II shows the results on each of the three models of normal return, I focus on the market and risk adjusted model.
degree of market efficiency in the emerging markets. However, it should be noted that the cumulative abnormal returns are insignificant for both markets and that there are thus no significant differences. Moreover, in the crisis period all the cumulative abnormal returns given in Table II, for both the emerging markets and the United States, are significant at a 5% level. Therefore, both markets are not efficient in the crisis period. In contradiction to the second hypothesis, the markets have become less informational efficient in the crisis period. In addition, the sign of the abnormal returns are negative, similar to the pre-crisis period. Negative abnormal returns following a positive event indicate that agents have overreacted to positive information in the market, resulting in a higher than appropriate initial shock. The overreaction needs to be compensated for, creating a negative reaction directly following the event. Therefore, this finding is in line with the overreaction theory. Moreover, when comparing the emerging markets to the United States in the crisis period, the cumulative abnormal returns are slightly larger in the emerging markets. This implicates a lower degree of market efficiency in emerging markets.
Concluding, when looking only at the panel of positive events the comparison between pre-crisis and crisis period shows that the abnormal returns for emerging markets and the United States seem to increase in the crisis, indicating decreased market efficiency. This evidence rejects the second hypothesis, there is no positive relationship between market efficiency and the global financial crisis. In the following subsection the differences will be tested in a multivariate setting, in order to be able to draw inferences about these findings.
time of the financial crisis, providing preliminary evidence that supports the second hypothesis. Finally, I compare the cumulative abnormal returns of the event window consisting of five days to the event window of three days, using the market and risk adjusted method of normal return. There is evidence of price persistence when the cumulative abnormal returns for the event window of five days exceed the cumulative abnormal return over three days. As a result, in the crisis period the price reaction persists in emerging markets as well as in the United States. The results are more pronounced in the panel of negative events. Based on this I use the event window of five days in the following tests.
The main conclusion is that the overreaction theory of behavioral finance is supported, given that the reaction to negative events gives positive abnormal returns and the reaction to positive events is negative. Moreover, Table II provides preliminary support on the first two hypotheses when looking at the panel of negative events. It is notable that agents are more likely to overreact to negative events, resulting in a less informational efficient response to negative events. In order to statistically test the differences between subsamples and provide grounded evidence on the hypothesis I conduct an independent samples t-test and the Wilcoxon rank sum test/Mann-Whitney on the sub-sample differences, the results are presented in the subsection below.
B. Sub-sample differences
Table III.
Subgroup difference analysis of cumulative abnormal returns
The table presents the difference between two subgroups and tests whether these differences significantly differ from zero using a t-test and the Wilcoxon rank sum test/Mann-Whitney, abbreviated by ‘M-Whitn’ in the table*. The cumulative abnormal returns for the event window of five days are used and calculated using the market and risk adjusted method of
normal return. EM refers to emerging markets and US to the United States.
A. Panel positive events (n= 1017)
EM US Difference t-statistic p-value M-Whitn p-value 1a. Pre-Cr: EM vs. US CAR[1,5] -0.028 -0.001 -0.027 -0.691 0.491 0.682 0.495 2a. Cr: EM vs. US CAR[1,5] -0.056 -0.028 -0.028 -1.711 0.087 1.314 0.189 Pre-crisis Crisis Difference t-statistic p-value M-Whitn p-value 3a. EM: pre-cr vs. cr CAR[1,5] -0.028 -0.056 0.027 0.862 0.390 0.897 0.370 4a. US: pre-cr vs. cr CAR[1,5] -0.001 -0.028 0.027 0.988 0.323 1.072 0.284
B. Panel negative events (n= 950)
EM US Difference t-statistic p-value M-Whitn p-value 1b. Pre-Cr: EM vs. US CAR[1,5] 0.145 0.119 0.026 0.570 0.570 0.467 0.640 2b. Cr: EM vs. US CAR[1,5] 0.080 0.097 -0.017 -0.926 0.355 1.489 0.137 Pre-crisis Crisis Difference t-statistic p-value M-Whitn p-value 3b. EM: pre-cr vs. cr CAR[1,5] 0.145 0.080 0.065 2.248 0.025 1.654 0.098 4b. US: pre-cr vs. cr CAR[1,5] 0.119 0.097 0.022 0.628 0.531 0.745 0.456
*p-values that are bold are significant at 5% level based on t-test and Wilcoxon rank sum test/Mann-Whitney, grey shaded coefficient is only significant at 5% level for the Wilcoxon rank sum test/Mann-Whitney.
The comparisons in the table are numbered; the numbers refer to the type of comparison and the letter to the panel to which they belong. To test the first hypothesis the market efficiency should be compared between markets. Comparison 1a and 2a and 1b and 2b compare the emerging markets and the United States in one of the time periods. The second hypothesis is tested comparing the market efficiency over time, therefore the final two comparisons, namely 3a, 4a, 3b and 4b; compare the cumulative abnormal returns between the pre-crisis and the crisis period.
The first hypothesis predicts that the market efficiency is lower for emerging market banks than for banks in developed markets. The expectation is that developed markets incorporate the available information in stock prices more rapid, the abnormal returns should therefore be lower in the United States. In comparison 1a, the cumulative abnormal return is lower for the United States. Lower abnormal returns indicate higher market efficiency for the United States. In line with these results, comparison 1b shows that the cumulative abnormal returns are lower in the United States as well.
Comparison 2a shows that the emerging markets are less informational efficient than the United States. On the other hand, the results on comparison 2b show that the emerging markets are more informational efficient in times of crisis. However, neither of the comparisons is significant based on the independent samples t-test or Wilcoxon rank sum test/Mann-Whitney. As a result, I cannot reject the null hypothesis on the first hypothesis.
The second hypothesis predicts that the global financial crisis improves market efficiency. If the hypothesis holds, the cumulative abnormal returns in the crisis will be lower than the cumulative abnormal returns prior to the crisis. Comparison 3a shows a lower cumulative abnormal return in the pre-crisis period. Similarly, comparison 4a shows that the cumulative abnormal returns following a positive price shock are smaller in the pre-crisis period. These results give an indication that the markets for both emerging markets and the United States are more efficient in the pre-crisis period. However, the results for both countries in panel A are not significant. Therefore I cannot reject the null hypothesis of the second hypothesis.
On the contrary, panel B shows that the cumulative abnormal returns in both the emerging markets and the United States are smaller in de crisis period, see comparison 3b and 4b. Lower abnormal returns indicate higher market efficiency in the crisis period, supporting the second hypothesis. The results on the United States are not significant, unlike those on emerging markets, which are significant at a 5% level.
Concluding, the first hypothesis is not supported by the results and the null hypothesis cannot be rejected. Moreover, the evidence on the second hypothesis is mixed, for the positive events the hypothesis is not supported whereas the emerging markets in panel B does support the hypothesis. Therefore, this evidence does not fully support the second hypothesis. Based on this analysis no conclusions can be drawn on the third hypothesis. The third hypothesis predicts the positive relationship between market efficiency and the global financial crisis to be stronger for emerging markets. The third hypothesis will be tested using a regression model including interacting dummy variables. The results are presented in the subsection below.
C. Cross-section analysis of post-event cumulative abnormal returns
method are included in Table B.I, appendix B.
Table IV.
Regression estimation cumulative abnormal returns [1, 5]
The table presents the results of the regression analysis, based on equation 10, using least squares estimates*. The dependent variable is the cumulative abnormal return over the event window of day 1 to day 5. The dummy variable emerging is one for emerging markets and zero otherwise. The dummy variable crisis is one for the crisis period, after the 15th of September 2008, zero otherwise. The regression model is conducted both for the panel of positive price changes (panel A) and the panel
of negative price changes (panel B). The different model specifications contain a different set of variables.
Variables A. Panel positive events (n=1017) B. Panel negative events (n=950)
Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5
c -0.033 -0.026 -0.015 -0.001 -0.001 0.099 0.099 0.135 0.141 0.119 (0.000) (0.002) (0.436) (0.958) (0.964) (0.000) (0.000) (0.000) (0.000) (0.000) D Emerging -0.024 -0.028 -0.027 -0.001 -0.010 0.026 (0.099) (0.065) (0.467) (0.963) (0.562) (0.538) D Crisis -0.021 -0.027 -0.027 -0.041 -0.045 -0.022 (0.303) (0.186) (0.332) (0.060) (0.049) (0.519) D Emerging * D Crisis 0.000 -0.043 (0.995) (0.349) N obs. 1017 1017 1017 1017 1017 950 950 950 950 950 adj. Rsq. 0.000 0.002 0.000 0.002 0.001 0.000 -0.001 0.003 0.002 0.002
*p-value in parenthesis, bold when significant at 5% level
The dummy variables must be interpreted in reference to the benchmark group, measured by the constant term. The interpretation of the benchmark group differs when different dummy variables are included in the regression. The results will be discussed based on the panel of positive and the panel of negative events. Moreover, the adjusted R-square is very low, indicating little explanatory power of the model. However, in this methodological setup the dummies are used to determine the differences between the subgroups rather than adding explanatory variables to explain the source of the cumulative abnormal returns.
Moreover, in Table D.I. in appendix D the correlation matrix is given for the variables included in the regression. The correlation between each of the variables in both panel A and panel B are low. However, the correlation for the interaction dummy with the dummy for emerging markets is 0.850 for panel A and 0.804 for panel B. These high correlations are caused by the fact that the number of events in the crisis exceeds the number of events in emerging markets, causing a high correlation.
in line with the overreaction theory. Model two represents the analysis on the emerging markets, where the constant term represents the United States as the benchmark against the emerging markets. The dummy for emerging markets gives a negative coefficient, indicating that the total cumulative abnormal return in emerging markets is higher than in the United States. As a result, the emerging markets are less informational efficient than the United States. However, it should be noted that the coefficient for the emerging market dummy is not significant; therefore I cannot reject the null hypothesis. The third model focuses on the comparison over time. The model shows a larger cumulative abnormal return for the crisis period, indicating that the market efficiency is lower in the crisis period. This evidence, which is not significant, does not support the second hypothesis. Moreover, the coefficients in the fifth model are not significant. Therefore, the interaction term in the fifth model specification does not provide explanatory evidence. Concluding, the third hypothesis is not supported by the panel of positive events.
In panel B, the coefficients show a positive cumulative abnormal return following negative price shocks, supporting the overreaction theory. In the second model the constant term represents the cumulative abnormal returns over five days in the United States, in both the pre-crisis and crisis period. The coefficient on the emerging markets dummy indicates that the cumulative abnormal returns in emerging markets are equal to the United States. This means that there are no differences in market efficiency between emerging markets and the United States, not supporting the first hypothesis. However, these results are not significant. The third and fourth model compare the crisis period to the pre-crisis period. The coefficient on the crisis dummy in the fourth model is significant at a 5% level. The crisis dummy gives a negative coefficient, indicating that the cumulative abnormal returns are lower in the crisis period than the pre-crisis period. As a result, the market efficiency is higher in the crisis period than in the pre-crisis period. Based on this evidence I reject the null hypothesis for the second hypothesis based on the panel of negative events. A possible explanation for the asymmetry between panel A and B, concerning the effect of the global financial crisis, is that in a period of crisis investors have a higher urge to sell their securities. Therefore, in the crisis period investors might have a lower tendency to overreact to negative shocks.
The fifth model is not significant. Therefore, the interaction term used to explain the third hypothesis does not add explanatory power and I cannot reject the null hypothesis in the third hypothesis, neither based on panel A nor based on panel B.
D. Robustness checks
D.1.Models of normal return
methods for each of the model specifications can be found in appendix B.
Similar to the market and risk adjusted method, presented in Table III, the results of the regression equation present significant results on the crisis dummy coefficients in panel B. This suggests that there is an asymmetry between panel A and B in stock price reaction.
When comparing the size and sign of the coefficient in the appendix to the regression, based on the market and risk adjusted method, the coefficients remain largely unchanged. Similar to the results on the market and risk adjusted method, the coefficients for panel A are negative whereas the coefficient on panel B are positive, indicating support for the overreaction theory. Relative to the market and risk adjusted method, the size of the coefficients varies slightly. The mean adjusted model estimates higher coefficients whereas the market adjusted method estimates slightly lower coefficients than the market and risk adjusted method.
In general, the results based on the mean adjusted and market adjusted methods of estimating normal returns do not substantially differ from the results specified by the analysis based on the market and risk adjusted method of normal return. Therefore, I conclude that the results are robust to different models of estimating normal returns.
D.2. Daily abnormal returns
The analysis is based on the cumulative abnormal returns of the event window. Based on the cumulative abnormal returns, conclusions can be drawn on the main hypothesis. However, the daily abnormal returns are more specific and are presented in appendix B. The analysis on daily abnormal returns is conducted as to verify the results on the cumulative abnormal returns and to be able to be more specific on the development of the abnormal returns. The appendix consists of the daily abnormal returns over the different subsamples, the analysis of differences between the subsamples and the regression estimation.
First, Table B.II. presents the daily abnormal returns for the market and risk adjusted method. The daily abnormal returns are not normally distributed and therefore the Wilcoxon signed rank test is conducted, in addition to the one sample t-test. As can be seen from the table, the abnormal returns in panel A are negative or close to zero. The reverse is true in panel B, where the abnormal returns are positive. These findings are in line with the overreaction theory of behavioural finance. In comparison to the results on the cumulative abnormal returns the daily abnormal returns do not present substantial differences.
abnormal returns as well. The table shows the development of the abnormal returns over the event window and provides no new evidence in comparison to the cumulative abnormal returns. In figure B.I. the stock price reactions following an event are graphically presented. The graphs show substantial abnormal returns following the events. Moreover, the stock price reaction persists in the days following the event.
Third, the regression estimation, to research the third hypothesis is conducted for the daily abnormal returns as well. The regression is based on the following regression equation. The dependent variable, abnormal return, consists of the daily abnormal returns of day 1 to day 5 following the event. The results are to be found in Table B.IV. in the appendix.
(11)
Where is the abnormal return over the first five days for event i, is the dummy indicating the event day for event i, varying from day 1 to day 5, the dummy variable indicating 1 for
emerging markets and 0 for developed markets, the dummy variable on the crisis period
indicates 1 for crisis and 0 for non-crisis period and is the error term. Since the regression equation holds dummy variables for each of the days that are included in the dependent variable, the set of variables is mutually exclusive; the constant term is removed from the equation.
There is a difference in interpretation between the cumulative and the daily returns regression models. The regression on daily returns does not hold a constant term and therefore the coefficients of the dummy variables of the subgroups must be added to the coefficients on the day dummies. The results in Table B.IV. show the development over the event window. The main conclusion is that the evidence supports the overreaction theory. The daily returns in panel A are negative and the daily returns in panel B are positive.
E. Further analysis on shock size
The analysis on the cumulative abnormal returns as well as the daily abnormal returns showed substantial overreaction to shocks. As a part of the overreaction theory, Brown and Harlow (1988) show a magnitude effect in their panel of negative events. This means that the higher the shock, the higher the overreaction and the higher the abnormal returns directly following the event. In order to specify the magnitude effect in my sample, I include the variable shock size in the regression estimation. The variable shock size is regressed in the fifth model specification of the regression equation. The regression is conducted for each of the three models of normal return.
from the table the estimation of the shock size variable are significant for both panel A and B, meaning that the variable shock size adds explanatory value to the regression model. The coefficients on the other variables in the model remain insignificant. The sign of the coefficients in both panels is negative, indicating a negative relationship between abnormal returns and the shock size. In other words, for panel A this means that when the shock is larger the abnormal returns become more negative. For both panels it shows, the larger the shock, the less informational efficient the market. The results are in line with the prediction of Brown and Harlow (1988) on the magnitude effect of the overreaction theory. They show a significant magnitude effect for only their panel of negative events whereas I show a significant magnitude effect for both panels. The size of the coefficients is around the same for each of the three models and both of the panels of events.
VI. Conclusion
This paper studies the evolvement of the degree of market efficiency of banks’ equity markets impacted by the global financial crisis. The empirical evidence of this paper is based upon stock markets of listed banks in emerging and developed countries in the years of 2006 to 2012. The analysis is conducted on two panels, consisting of respectively positive and negative price shocks.
Existing literature on the evolvement of equity market efficiency of indices provides mixed results. Literature on market indices by stochastic models is limited by their assumptions on the information market. On the other hand, the specific event studies are limited to a specific range of events. This study combines these methodologies and steers a middle course, employing an event study methodology where events are defined as 20% or higher price shock. The information content of the price shock is not specified, creating an equal foot comparison between countries in emerging and developed markets. Moreover, note that these findings are specifically related to the banking industry, the results however do not seem to differ substantially from the evidence presented on the market efficiency of indices in previous research.
The first hypothesis predicts equity markets of banks in the emerging markets to be less informational efficient than the United States in the pre-crisis period. I find that for panel B, the cumulative abnormal returns in emerging markets are higher than in the United States in the pre-crisis period. Higher cumulative abnormal returns indicate lower market efficiency. However, these differences are not significant. Moreover, the regression estimation does not present significant coefficients on the emerging markets dummy. Therefore, I conclude that there is no evidence to reject the null hypothesis for the first hypothesis, neither for the panel of positive events nor for the panel of negative events. These findings are in line with Griffin et al. (2006), since they argue that the market efficiency in emerging markets is at least as strong as in developed markets.
may impose structural changes, institutional reforms and enhance stock market development, impacting the market efficiency (Narayam and Islam (2012), Smith (2012) and Todea and Lazar (2012)). The empirical evidence in this research shows support for the second hypothesis solely based on the panel of negative events. The regression estimation shows that the crisis coincides with decreasing cumulative abnormal returns, relative to the pre-crisis period in both markets. This does not hold for the panel of positive events. A possible explanation for the asymmetry between the panels can relate to the higher urge of investors to sell the securities in a crisis period.
Finally, the third hypothesis that predicts that the positive relationship between the global financial crisis and market efficiency is stronger for emerging markets. I do not find any handles to support this hypothesis. The interaction terms in the regression estimation are not significant in both panels. The reason for this is that the market efficiency of emerging markets is, pre-crisis, similar to the market efficiency in the United States, suggesting that there is not more potential to increase market efficiency in the emerging markets.
Besides the rejection of the hypotheses, some interesting remarks are to be made concerning the findings of my research. The first is that the evidence supports the overreaction theory, challenging the efficient market hypothesis. The results clearly show negative abnormal returns following positive shocks and positive abnormal returns following negative price shocks. These findings may suggest that the rational expectations theory is not valid, and support the recent criticism on the efficient market hypothesis. Moreover, it is notable that there is an asymmetry in the stock price reaction between panel A and B in the pre-crisis period. In panel B the abnormal returns are significantly different from zero, showing a lower market efficiency compared to panel A. In panel A, in the pre-crisis period the cumulative abnormal returns were insignificant, suggesting efficient markets. In line with Brown and Harlow (1988) investors have a higher tendency to overreact to negative information, in the pre-crisis period. The underlying reason relates to risk adversity. Second, I find evidence that stocks with a higher initial shock are followed by larger reversals. This magnitude effect is measured by the shock size in the regression estimation. For both the panel of positive and the panel of negative shocks the shock size is significant. The larger the shock, the larger the reaction following the shock and therefore the larger the overreaction.
control for confounding events in the same period of time. Finally, the methodology does not allow for time-varying market efficiency in the time periods.
Appendix A. Descriptive Statistics Table A.I Descriptive Statistics ARs and CARs
A. Panel positive events (n=1017) B. Panel negative events (n=950)
Day 1 Day 2 Day 3 Day 4 Day 5 CAR (1,5) Day 1 Day 2 Day 3 Day 4 Day 5 CAR (1,5)
Appendix B. Robustness
Table B.I. Robustness models of normal return
The table presents the results of the regression analysis, based on equation 10, using least squares estimates*. The dependent variable is the cumulative abnormal return over the event window of day 1 to day 5. The dummy variable emerging is one
for emerging markets and zero otherwise. The dummy variable crisis is one for the crisis period, after 15th of September 2008, zero otherwise. The regression model is conducted both for the panel of positive price changes (panel A) and the panel
of negative price changes (panel B). The different model specifications contain a different set of variables.
Mean adjusted method A. Panel positive events (n=1017) B. Panel negative events (n=950) Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5
c -0.034 -0.028 -0.016 -0.004 -0.002 0.103 0.105 0.133 0.142 0.115 (0.000) (0.000) (0.358) (0.810) (0.926) (0.000) (0.000) (0.000) (0.000) (0.000) D Emerging -0.020 -0.023 -0.028 -0.007 -0.015 0.030 (0.137) (0.091) (0.424) (0.665) (0.353) (0.448) D Crisis 0.001 -0.026 -0.028 -0.035 -0.040 -0.011 (-0.034) (0.172) (0.271) (0.090) (0.060) (0.720) D Em * D Cr 0.005 -0.053 (0.889) (0.211) N obs 1017 1017 1017 1017 1017 950 950 950 950 950 adj. Rsq 0.000 0.001 0.000 0.002 0.001 0.000 -0.001 0.002 0.002 0.002
Market adjusted method
c -0.056 -0.056 -0.030 -0.029 -0.027 0.087 0.084 0.125 0.125 0.094 (0.000) (0.000) (0.069) (0.108) (0.247) (0.000) (0.000) (0.000) (0.000) (0.001) D Emerging 0.001 -0.003 -0.006 0.009 0.000 0.051 (0.967) (0.813) (0.844) (0.522) (0.978) (0.169) D Crisis -0.030 -0.030 -0.032 -0.045 -0.044 -0.011 (0.095) (0.092) (0.188) (0.024) (0.030) (0.709) D Em * D Cr 0.004 -0.061 (0.911) (0.135) N obs 1017 1017 1017 1017 1017 950 950 950 950 950 adj. Rsq 0.000 -0.001 0.002 0.001 0.000 0.000 -0.001 0.004 0.003 0.005
*p-value in parenthesis, bold when significant at 5% level
Table B.II. Daily average abnormal returns
The table presents the daily average abnormal returns for the market and risk adjusted method of normal returns. The abnormal returns are given for the five days following the event, for both the panel of positive events and the panel of negative events. Significance is tested by a regular t-test and a non-parametric Wilcoxon signed rank test*.
A. Panel positive events (n=1017) B. Panel negative events (n=950)
Day 1 Day 2 Day 3 Day 4 Day 5 Day 1 Day 2 Day 3 Day 4 Day 5
PRE-CRISIS (01-01-2006 to 14-09-2008) Emerging markets mean -1.89% 0.46% -1.78% 0.25% 0.12% mean 7.13% 2.49% 1.49% 2.97% 0.41% t-test (0.451) (0.630) (0.133) (0.933) (0.918) t-test (0.003) (0.050) (0.121) (0.162) (0.664) Wilcoxon (0.403) (0.921) (0.915) (0.938) (0.585) Wilcoxon (0.002) (0.177) (0.640) (0.847) (0.872) United States mean 0.09% -1.22% -0.30% 2.10% -0.78% mean 4.79% 3.56% 1.04% 2.38% 0.14% t-test (0.942) (0.152) (0.736) (0.026) (0.417) t-test (0.005) (0.011) (0.457) (0.044) (0.908) Wilcoxon (0.858) (0.245) (0.903) (0.010) (0.583) Wilcoxon (0.001) (0.011) (0.509) (0.037) (0.564) CRISIS (15-09-2008 to 31-12-2012) Emerging markets mean -2.76% -2.02% -0.16% -0.45% -0.16% mean 3.95% 1.76% 0.83% 0.70% 0.77% t-test (0.002) (0.009) (0.801) (0.457) (0.793) t-test (0.000) (0.016) (0.358) (0.344) (0.193) Wilcoxon (0.002) (0.008) (0.221) (0.302) (0.877) Wilcoxon (0.001) (0.089) (0.536) (0.161) (0.266) United States mean -1.34% -0.38% -1.07% 0.57% -0.58% mean 5.84% 2.04% 1.07% 0.30% 0.49% t-test (0.012) (0.403) (0.007) (0.138) (0.159) t-test (0.000) (0.000) (0.025) (0.519) (0.271) Wilcoxon (0.001) (0.679) (0.035) (0.159) (0.529) Wilcoxon (0.000) (0.000) (0.550) (0.047) (0.023)
