There are many econometric methods for forecasting by different economic variables in the future. recently, the procedures of dynamic forecasting either for univariate or multivariate models were available for estimation on the software packages, i.e., EVIEWS, SAS, and SHAZAM. The research problem of the study, concerned with the different types of such dynamic models, with respect to, estimation, choosing the best fit model for forecasting by the economic variables, i.e., labor and wages on the agricultural and national level. So the objective study, is to concentration and determination the best forecasting model among univariate and multivariate dynamic time series models. The time series data on the agricultural and national level were collected from the ministry of planning during the period (1975-2002). The methodology framework discussed the theoretical and mathematical approach for the dynamic univariate models, i.e., autoregressive integrated moving average (ARIMA), and multivariate models, i.e., vector autoregressive (VAR), vector error correction model (VECM), and state space model (SSM). The dynamic models contain four stages that have, identification, i.e., stationarity and cointegration tests, model selection criteria for determination the lag length, causality test, and choosing the techniques of estimation,also estimation stage, diagnostic stage for model accuracy, and forecasting stage. The study estimated the dynamic models by maximum likelihood estimation (MLE) for (ARIMA) models, and by seemingly unrelated regression for (VAR) and (VECM) models, during the period (1975-2002), and forecasting by labor and wage through the period (2003-2012). The estimation and forecasting results, indicated that the agricultural labor will increase at decreasing rate, also the relative share of agricultural labor and the total agricultural wages will decrease during the period subject to forecasting. The national labor will increase at increasing rate. Finally the study recommended by cultivation crops, ...
