ENGLISH ABSTRACT: This study investigates whether any exploitable pauems exist in a sample of tenAfrican stock markets that could lead to abnonnal gains. Southern Africa isrepresented by Botswana, Namibia. Mauritius and Zimbabwe, East Africa by Kenya,West Africa by Ghana and the BRVM, and North Africa by Egypt, Morocco andTunisia. Such evidence, if it exists, provides ground for refutation of the weak form ofthe efficient market hypothesis (EM H) as defined by Farna (1965. 1970).The thesis is predominantly empirical, but also provides an overview of African stockmarkets, the theoretical framework on which the study is based and the impact of theadvancement in information technology on market efficiency. The results show thatthe distribution of stock returns on African stock markets is not normal, and that thedeviation from normality is significantly pronounced with almost all the stocksrejecting nonnality using the Kolmogorov-Smimov test at the I % level ofsignificance.The stock price behaviour of the abovementioned stock markets is investigated bytesting the random walk hypothesis using the simple serial correlation and runs tests.The investigation is done using returns calculated on a trade-to-trade basis andadjusted for interval variability by weighting each trade-to-trade return by the numberof days between trades. While the first part of this analysis only includes the marketson which dividend information could be obtained, the second part includes all the tenmarkets with returns referring to capital gains. However, it is shown that dividendinformation does not have a serious impact on the results. While the majority ofstocks, especially those for Mauritius and Ghana, reject the random walk hypothesis,only Namibia, Kenya and Zimbabwe, can be said to be weak form efficient.While thin trading is known to cause econometric and statistical problems inempirical tests, thin trading has been taken as given in most studies. In this thesis, theseriousness of thin trading on African stock markets and its implications for efficiencytesting is empirically investigated. A comparison of the random walk test results whenreturns are calculated normally and when the trade-to-trade approach and its variant,the adjusted trade-la-trade approach, are used is carried out. It is found that thintrading is indeed a severe problem on African markets and that there are somedifferences in the random walk results due to the different methods used to calculatereturns.Investigating in-sample predictability using linear models appears to be the norm inmost tests of the EMH. This thesis argues that the return-generating process may notbe linear and if that is the case, the nonlinear models may outperform the linearmodels in out-of-sample forecasting. The random walk is considered a truedescription of stock price behaviour only if it is not outperformed by any of thealternative models in forecasting stock prices out-of-sample. This is empirically testedusing the indices data of the African stock markets in the sample. It is found thatalternative models, in most instances, outperform the random walk model in out-of-sampleforecasting.The random walk results are substantiated by the results on seasonal patterns andother anomalies to the efficient market hypothesis such as the finn size and price earnings(PIE) effects. Size and PIE ratios have been identified as significantpredictors of stock returns in other markets. In particular, it has been suggested thatsmall-size firm portfolios outperform large-size finn portfolios and that low PIE firmportfolios outperform high PIE firm portfolios. The size and PIE effects found in thisthesis are mostly exactly the opposite of those hypothesised in the literature.The existence of seasonal patterns contradicts the statement that stock prices behavein a random manner. This phenomenon is investigated on African stock markets usingindices returns. The study benchmarks the findings with those of South Africa'sJohannesburg Stock Exchange (JSE) Securities Exchange; other emerging markets,namely Brazil, Malaysia, Poland, Slovenia and Finland; and developed markets, suchas the United States of America (U.S.), Australia and New Zealand. Seasonal effectsare observed on some, but not all African stock markets and in most cases the patternsobserved are different from those observed on stock markets elsewhere.
