분기 GDP 전망을 위한 초단기예측모형의 예측력 비교 연구

다국어 초록 (Multilingual Abstract)

This thesis compared the forecasting performance among models to promptly forecast GDP with slow announcement cycle after estimating very short-term model. The very short-term model is mainly used to forecast GDP of quarter based on co-movement with monthly economic variables that have short cycle in the formation of data.
The very short-term models constructed in this research are Intervention Model, Transfer Function Model, VAR Model, Factor Model and Simultaneous Equation Model, whose explanatory power is found to be good generally.
Considering the term of foreign exchange crisis as exogenous impact, Intervention Model established AR Model in which effect of intervention of foreign exchange crisis was estimated. The Transfer Function Model was constructed by taking composite indexes and business survey index, the representative monthly indicator of overall economy, as explanatory variables. Further, VAR Model, multi-variate model was constructed to evaluate forecast ability and many economic indicators were compressed into a few factor score through factor model so as to be used as explanatory variables of regression model. Finally, the Simultaneous Equation Model approached in the aspect of production of national account was estimated.
Evaluating the forecast ability of model, it was found that the forecast error calculated by MAPE of all models was within 1%, good value. In the case of Factor Model and Simultaneous Equation Model where monthly information can be reflected into model, the forecast error was decreased with the addition of monthly economic indicator. Especially, the forecast ability of model can be improved by Intervention Model when forecast is done by considering unexpected variables in the model, namely, policy variable with high exogeneity, international economic situation. Thanks to its merit to reflect many economic indicators, factor model showed relatively excellent forecast ability compared to other models. Further, it was confirmed that forecast value made by averaging all models further reduces forecast error than the method to utilize only a few models in forecasting GDP.

목차 (Table of Contents)

• Ⅰ. 서론 1
• Ⅱ. 초단기예측모형의 개요 4
• 1. 계량경제모형의 분류 4
• 2. 초단기예측모형의 필요성 5
• Ⅲ. 초단기예측모형의 방법론 8
• 1. ARIMA모형 9
• 2. 전이함수모형 13
• 3. 개입모형 16
• 4. 인자모형 20
• 5. VAR모형 21
• 6. 연립방정식모형 23
• Ⅳ. 모형설정 및 추정 결과 25
• 1. 개입모형의 추정 결과 25
• 2. reg-ARIMA모형의 추정 결과 31
• 3. 인자점수를 이용한 회귀모형의 추정 결과 34
• 4. VAR모형의 추정 결과 38
• 5. 연립방정식모형의 추정 결과 41
• Ⅴ. 모형의 예측력 비교 45
• 1. 모형의 예측력 평가 방법 45
• 2. 모형의 예측력 평가 결과 46
• Ⅵ. 결론 49
• 참고문헌 52
• ABSTRACT 55
• [부록1]모형별 예측모의실험 결과 그래프 58
• [부록2]연립방정식모형 추정결과 59