Using neural networks and cognitive mapping in scenario analysis: The case of Turkey's inflation dynamics

This paper has two objectives. First of all, it proposes a dynamic scenario analysis approach, which is a revised version of extended anomaly relaxation (EFAR) model [J. Operat. Res. Soc. 48 (1997) 793]; hereafter referred to as REFAR. REFAR aims to eliminate the basic drawbacks of EFAR and improve its efficiency. The basic steps of REFAR are presented and the improvement with respect to the original version realized in each step is emphasized. The second objective of the study is to estimate the future of inflation in Turkey through REFAR using data corresponding to the period January 1994-December 1998. The main reasons for selecting REFAR-based inflation estimation instead of adopting traditional forecasting techniques are explained. The basic scenarios finally reached through REFAR, the transition among each key scenario as well as among the scenarios grouped under each key scenario are explained in detail and the validity of the REFAR-based inflation estimation model is tested through several case studies.