Modeling of Social Economic Systems

Abstract

In this paper the difficulties, characteristic features, methodological characteristics, model verification in building socio-economic models are briefly reviewed. The purposes to build socio-economic models are summarized as: socio-economic systems analysis, policy appraisal, systems planning and control (management), forecasting. Main objectives of this paper are to introduce results obtainted at ISE in the subjects of small sample estimation and model structure changing systems which reflect some of current research efforts in the area of socio-economic systems modeling in China.For many developing countries like China the statistics available for modeling are usually short and incomplete and the data are actually generated by the time-varying or structure changing systems. Therefore methods of modeling under small (or limited) sample condition and methods to model structure changing systems should be studied and developed in order to build models fit the real systems better.From 1984 novel methods for small sample estimation are developed successively: first the LKL, a general method without using any prior information; later SSML, ESML and others based on ML function; and recently, the SSBE based on Bayesian method. Influenced by work at ISE other methods like minimax and MSAR are also studied in China. Model order estimation under small sample condition is also studied. Bayesian and non Bayesian methods to test changing points of structure changing models studied at ISsE are introduced. The former is studied for linear multiple regression models with single and multiple structure changing points, and also for multi-state threshold models. The latter is studied for linear regression models in both single equation and simultaneous equations forms