Comparison of the predictive performance of ARIMA, VAR, and VECM models in cointegrated time series
Predictive models are tools used by scientists and business analysts to anticipate trends. Predicting a pattern of future behavior brings competitive advantages that can transform into better quality public policies and new products. Although the econometric protocols for estimating forecasting models are clear, in areas as (i)fp&a, (ii)data science and, (iii) risk, there is wide use of ARIMA models in cointegrated data series solving business problems. The object of study of this work is to compare the predictive capacity of classical models: (i)ARIMA, (ii) VAR, and (iii) VECM applied in simulated cointegrated data to evaluate the performance of these models in data out of development sample.