The purpose of this study is to develop a new forecasting model that reduces measurement error by combining various existing accounting earnings forecasting models. The forecast errors from various existing time-series accounting earnings forecasting models are used as instruments in the combined model. However, this study discovered that using the signs of the respective forecast errors from existing forecasting models as instruments yields better results than using the raw (unprocessed) forecast errors as instruments. Additionally, this study verified that the newly developed forecasting model is particularly effective for small firms.