This study aims to examine the effect of government expenditure on economic growth in Ethiopia by using annual time series data covering from 1981 to 2015. This study had employed Johansen Co-integration and error correction technique in the context of Hendry’s general-to-specific modeling and Vector Autoregressive (VAR) specification with variance decomposition and impulse response analyses to provide robust short run and long run effects government expenditure on economic growth. The results show that transport and communication expenditure, defense and health expenditure variables have a long run relationship with the RGDP. This study indicates that the estimate of speed of adjustment coefficient found that about 8 percent of the variation in the RGDP is corrected within a year from its equilibrium level. The regression result of VECM reveals that government expenditure on agriculture and transport and communication were the variables significantly affecting the RGDP in the short run. Moreover, the variance decomposition and impulse response analysis corroborates a better picture of the short run dynamics. The analysis provided that the shocks to agriculture and transport and communication have persistent effects on the RGDP in the short run. Hence, economic growth to be sustained, government should optimize its spending more on transport and communication, while according on the findings, reducing the budget share of defense and health expenditure and transfer to highly productive sectors lead to higher productivity and hence higher economic growth. Keywords: Ethiopia, economic growth, Co-integration, Vector error correction