The major contents of the study are consisted with the constructing a macro-econometric model emphasizing the agricultural sector, estimating the structural equations using the annual data, and performing historical and policy simulations to analyze the relationship between agricultural sector and general economy.
The structure of this paper is as follows. In section II we review the literature briefly. We construct the macro-econometric model in section Ⅲ. In this section, the results of estimation of structural equations are interpreted and discussed additionally. In section Ⅳ, we perform the historical and policy simulations. Finally section Ⅴ concludes the paper.
Since the analysis of the national economy based on macro-econometric models was initiated by Tinbergen, and Kelin-Goldber. Since then it has followed by various models such as Brookings model, FRB-MIT model, Warton model, DRI model, St. Louis model, and Chase econometric model. In case of Korea, more than 30 models have been developed since the first macro-econometric model was introduced by the Bank of Korea(BOK) in early 1970s. These models were policy-oriented because these models were developed under the supervision of central government or related organizations such as Korea Development Institute(KDI) and so forth, resulting in a strong tendency to enhance forecasting power instead of analyzing economic system or structures. It is no wonder that these models have the limitations caused by the fact that their attentions had to be paid to the public sectors in estimating the effectiveness of various kinds of policies.
The beginning of korean agricultural macro-econometric model was KASS(Korean Agricultural Sector Study) model in early 1970s, but it has too many equations to keep or amend its system. Recently KREI(Korea Rural Economic Institute) has developed KREI-ASMO(Agricultural Simulation Model) which is used for the forecasts on various agricultural items and farm household economy. These are conducted in the following methods: To begin with, one team is assigned to observe one agricultural item and conducts on-site researches and data analysis to figure out its supply and demand trend, and the research results are entered into the agricultural sector forecasting model (KREI-ASMO) and the item model of the Agricultural Outlook and Information Center(KREI-COMO) to predict the short and mid/long-term trends by item. The forecast results are transferred back to KREI-ASMO again and used in making an outlook for the agricultural industry and farm household economy. Later, feedbacks are given to each item team. However, KREI-ASMO don't include many macro-variables such as, consumer prices, GDP, exchange rate, etc.
The constructed model of this paper consists of six blocks-the final demand, labor/supply, prices, fiscals, finances, and foreign sector. Among these blocks, the final demand, labor/supply, prices blocks are designed as agricultural and non-agricultural sectors are divided and influenced mutually in system. The sample period for empirical study is 34 years between 1970 and 2003. The ordinary least squares(OLS) will be used for estimating each equation. And if needed, Cochrane -Orcutt method will be used to correct autocorrelation in the error terms. The six blocks are composed with 62 equations including 34 of behavioral equations and 28 of identities. The equations defined by 62 of endogenous variables and 32 of exogenous variables.
The simulation analysis intends to derive the values of endogenous variables by substituting estimated coefficients in an adequately established structural simultaneous equations, and to evaluate how well such derived values of endogenous variables represents empirically used data. The simulation analysis classified into the types of historical and policy ones, and ex-post and ex-ante forecasting. It is possible to evaluate the suitabilities and stabilities of the models by comparing the endogenous variables calculated by such methods. The typical evaluation standards include root mean square simulation error(RMS) and RMS percent simulation error(RMSPE). The RMSPE is designed to seek the average of the values achieved by squaring the deviations between the simulation values and the levels. According to the historical simulation, the RMSPE of the major variables is smaller, more or less, one decimal point compared with their levels which demonstrate they are considerably accuracy.
The policy simulation is basically designed to analyze how the changes in the exogenous variables affect to the endogenous variables in the future. That is to say, it comparatively analyzes the estimated values of endogenous variables calculated by behavioral simulation assuming that ad hoc policy variables or exogenous variables have changed by a certain ratio more than actual values. The policy simulation alternatives were selected on the assumptions of the following: the raise in the exchange rate of won to the dollar by 10%, increasing M3 by 10%, increasing in the agricultural investment expenditures of the government by 10%, downward readjustment of political interests in agriculture by 5%, increasing of agricultural import price by 10%.
This study has provided the basis in establishing macro-econometric models for analyzing the national economy, focusing on the agricultural sector. The number of previous works for this field has been performed, not including macroeconomic variables in internal system, by descriptive trend analysis, and such a way is confined by intuitive judgements. It resulted in insufficiencies for examining interrelations between macroeconomic variables and agricultural sector. The study has advantages in that it is able to analyze systematically on the whole systems, and explain how the changes in a specific variable of a specific sector affects other sectors.
However, it also has to be mentioned that the study is not a complete version of macro-econometric models for the economy emphasizing the agriculture. Therefore, in order to be supported the stability of the model, it is necessary to be confirmed by using the unit root test, and furthermore cointegration techniques to see long-run stable relationships of the models, and it may also be possible to realize short-run behavioral analysis by setting up error-correction modeling. More practically the macro-econometric models should be closely connected to specific economic policies including various agricultural polices.