Resumen
The study seeks to investigate whether non-linear patterns are present in the returns of two indices on the stock markets in Ghana and Nigeria between the period of 2011 and 2015.The results of applying four linearity tests on the returns concluded that the null of linearity is rejected on all four tests for the Ghanaian index but mixed for the Nigerian index. We modelled the indices under the non-linear self-exciting threshold autoregressive (SETAR) model. We compared the modelling performance of the non-linear SETAR model with that of the standard AR (1) and AR (2) by analyzing AIC values of the respective models. Our results show that the SETAR model ?ts the data well. Hence, modelling stock market returns from Ghana and Nigeria using linear models might lead to spurious conclusions.Keywords: Threshold models, Linearity tests, Self-Exciting Threshold Autoregressive (SETAR) modelJEL Classifications: C12, C13, C24