Anomaly detection in financial market indices in East African stock markets

Abstract

This study investigated the detection and impact of anomalies in stock market indices within East African financial markets, focusing on the Nairobi Securities Exchange (NSE), Uganda Securities Exchange (USE), and Dar es Salaam Stock Exchange (DSE). Utilizing a descriptive and explanatory research design, the study integrates neural networks with statistical methods to identify and analyse anomalies such as market manipulation, insider trading, and reactions to macroeconomic events. By analysing both historical and real-time data, the research aimed to understand how these anomalies affect market behaviour, investor confidence, and overall market stability. The research was anchored on The Efficient Market Hypothesis. The population for this study encompassed the Nairobi Securities Exchange (NSE), Uganda Securities Exchange (USE), and Dar es Salaam Stock Exchange (DSE). The sampling frame for this study consisted of daily stock market index values from the NSE, USE, and DSE, along with annual inflation rates and GDP data for Kenya, Uganda, and Tanzania. This study utilized secondary data, spanning the period from 2013 to 2022. The primary sources of data included official publications and financial reports from the respective stock exchanges, such as annual reports and historical index data. Additional data was obtained from reputable databases and organizations, including the East African Community financial publications, Bloomberg, Reuters, and DataStream. Descriptive statistics such as mean, maximum, minimum, and standard deviation were employed to summarize and describe the key characteristics of the dataset. To evaluate the relationships between variables and detect patterns, the study employed multi-linear regression analysis using SPSS. Anomaly detection in the stock market indices was conducted using two complementary methods: Z-scores and a Multilayer Perceptron (MLP) Neural Network. The analysed data were interpreted and presented using graphs, tables, and charts generated from Microsoft Excel and SPSS. The study found that a unit change in inflation yields a 6.635 negative change on Stock Index. The p-value of 0.001 means that the inflation has positive and statistically significant effect on the Stock Index in East Africa. The study also found that a unit change in GDP yields a 4.443E-11 positive change on Stock Index in East Africa. The p-value of 0.000 means that the GDP has positive and statistically significant effect on the Stock Index in East Africa. The study concluded that while inflation has a negative impact, GDP exerts a positive and more substantial influence. Further, it was concluded that GDP had the greatest influence on the stock index values in East African markets, while inflation had the least effect on the stock index values in East African markets. The research suggested that policymakers address macroeconomic factors, such as inflation and GDP fluctuations, which significantly influence stock market anomalies. Effective monetary and fiscal policies were deemed necessary to control inflationary pressures and foster economic growth, which in turn could stabilize stock markets. The research also suggested that regional collaboration between stock exchanges—NSE, USE, and DSE—be strengthened to address shared economic and financial challenges. Establishing shared databases and analytical tools to monitor regional market trends and detect anomalies was recommended as a way to create a more integrated and stable financial ecosystem.