Current approaches used in prediction of a reservoir performance are less reliable and require much computational effort and time. In this study, system identification, which is common in economics and similar fields, is proposed for predicting performance of a reservoir under gas injection. Efficacy of the proposed method is verified using two synthetic case studies. Box-Jenkins, model structure which is common in system identification, is applied to capture dynamics of the reservoirs and predict their future performance. Box-Jenkins model with order BJ(2-2-2-2-1) is found to be efficient in terms of fitness during validation and having a reasonably less number of parameters. The model is validated using cross validation principle and the results found are auspicious for further investigation. Comparison of BJ(2-2-2-2-1) model with decline curve analysis, which is a well-recognised technique in oil and gas industry, shows that BJ(2-2-2-2-1) is superior in forecasting reservoir performance under gas injection. Copyright © 2016 Inderscience Enterprises Ltd.