Public expenditure forecasting is an important task which attracts attention from many researchers. This forecasting problem is crucial for the success of future public financial management or budget management approaches in a country. The main purpose of this study is to develop some efficient forecasting models for public expenditure in order to reach high accuracy level. For this purpose, different linear and non-linear models based on autoregressive integrated moving average (ARIMA), exponential smoothing, and artificial neural networks are developed. This study applies various linear and non-linear models to the expenditures of 1980-2010 of two Turkish public institutions, namely, the State Planning Organization and the Court of Accounts to achieve accurate forecast levels. Different linear and non-linear forecasting models are applied to the related time series and obtained results are compared. As a result of the implementation, it is observed that the artificial neural networks provide very accurate public expenditure forecasts for these public institutions, suggesting that the artificial neural networks is a very effective tool for the public expenditure forecasting.