Study on Fixed Effect versus Random Effects Modeling in a Panel Data Analysis; a Consideration of Economic and Political Indicators in Six African Countries

Individual heterogeneity can be controlled in panel data analysis to avoid bias in the final estimates. Panel data can be balanced or unbalanced, short or long panel. To identify the appropriate model, the fixed effects and random effects modelling approaches were applied to an economic data set, "Africa" in the R Amelia package. Both models were fitted to the data, taking into account the assumptions of the two models. The Breusch-Pagan Lagrange Multiplier test on the random model estimates revealed that the random model was acceptable for the data, however it had a poor coefficient of determination, of 0.48697. The fixed effect was then estimated using four different approaches (Pooled, LSDV, Within-Group and First differencing) and testing each against the random effect model using Hausman test. Our findings revealed that the random effect was inconsistent in all tests, indicating that the fixed effect was more appropriate for the data. with an of , the LSDV was found to have the best match among the fixed effects models.