Forecast Error Measurment: MAPE

[mscarlatos]

Following a regression I wish to test that model's utility as a forecasting tool. I then go to "Forecast" and insert the period I wish to compare the forecasted left hand side results with the actual left hand side data (2001Q1-2006Q4) and it give me a very large MAPE. Do I read the MAPE as a percentage (eg., is a MAPE of 179 read as 179%)? Secondly, am I correct in thinking that what eviews MAPE program does is to generate regression results through to 2000Q4 and then apply those estimated coefficients to forecast the 2001Q1 onwards which are then compared with the actuals? Finally, the forecast program defaults to the entire period, 1970Q1-2006Q6. Is this a forecast of 1970Q2 based on coefficients generated during the first 2 quarters? Thank you.