Bayesian VAR모형을 이용한 정보통신산업 수출 전망 연구

A Study on the ICT Industry Export Forecast Using Bayesian VAR Model

This paper forecasts Korean ICT exports by applying a Bayesian VAR model with quarterly data for the period between 2000Q1 and 2013Q3. The Bayesian VAR approach was known to provide accurate forecasts, compared to conventional macroeconomic models or alternative time series approaches. Bayesian VAR approach allows imposing prior restrictions on the model parameters reducing the dimensionality problem of VAR model, resulting in efficiency gains in the estimation of the parameters and in more accurate forecasts. In the analysis followed in this work, the empirical model is set by reference to previous studies, analysing the determinants of Korean ICT exports. The sample is spilt in two sub-samples: the first one, 2000Q1-2013Q4, is used to estimate the Bayesian VAR parameters and the second one, 2014Q1-2015Q3, is used to compute out-of-sample forecast error. In order to test accuracy of out-of-sample forecasts, we evaluate the Root Mean Squared Error (RSME) for the 7 periods from 2014Q1 to 2015Q3. The results show that the RSME of Bayesian VAR is only 1.64%. Comparing the forecast performance of Bayesian VAR with the unrestricted VAR, we find that Bayesian VAR shows a superior performance in forecasting as compared with VAR.