Essays on an emerging stock market: the case of Nairobi stock exchange (Statistical distribution of returns, market seasonality and reactions to dividend announcements
The general objective of this thesis is to test the well known market efficiency hypothesis using daily data from the Nairobi Stock Exchange. This high frequency data permits a thorough testing of the efficiency hypothesis because the very short¬period nature of the data, helps control for effects of other determinants of the stock market performance, which have been a persistent problem in previous studies. The analysis of data reveals that the distribution of daily compounded returns on ordinary shares is not normal, and unlike what some previous studies have shown, the distribution of stock returns exhibits long tails. The shape of this distribution implies that the actual data fluctuates with a bigger margin than what would other wise be expected from a standard normal distribution. It also renders linear models unsuitable tools for analyzing behavior of stock returns. There is strong evidence of volatility, clustering, and asymmetry of price dispersion, which further justifies the use of non¬linear models in the analysis of stock markets. With regard to asymmetry, it is found that big changes in returns follow big ones, and that small changes follow small ones, and negative changes in returns are more persistent than positive changes. On asset pricing models, the results show that the linear model fails to capture the relationship between daily returns on ordinary shares and market returns. As consistent with previous studies, there is evidence of ARCH effect, with TGARCH model outperforming the "'bLS, GARCH (1, 1) and the EGARCH models. On calendar anomalies, the study shows that though methodologies play an important role in outcomes of tests of the null of the market efficiency hypothesis, the various methods deliver similar trends, such that the calendar effect is only evident when large periods are considered. The implication of this is that though there is no evidence of day-of-the-week effect, there is a weak pointer towards existence of month-of-the-year effect, and strong evidence of quarter-of-the-year effect. The evidence that the quarter of the month effect exists, suggests that although investments in ordinary shares made on the basis of the day-of-the-week will yield capital gains by chance, profits from long term share investments are almost guaranteed. As to sensitivity of stock returns to an event, a non-parametric test of this sensitivity outperforms the regression test. The test results show that there is need to use short estimation periods, since longer ones are subject to data smoothing, in addition to increasing the chances of the event of interest overlapping with other events. There is evidence that at the Nairobi Stock Exchange, ordinary share returns are sensitive to dividend announcements, with the announcements triggering market volatility, followed by normalization in about a week. This pattern of performance implies that it takes only a short period for publicly available stock information to get to all investors, so that only the investors who react within the first one week can make abnormal profits on the basis of such information. Finally, it is found that most investors at the Nairobi Stock Exchange are speculators who have no allegiance to particular firms.