Short-term forecasting Swedish annual real GDP growth using SARIMA models : A study in forecasting current year Swedish annual real GDP growth using SARIMA models with the Box-Jenkins methodology as a general framework

Sammanfattning: Simulated current year annual real GDP growth forecasts for 2015-2021 are made using a chosen SARIMA model, with the Box-Jenkins methodology as a general modelling framework. The forecasts are compared to the actual outcomes and the Absolute Errors (AE) and the Mean Absolute Errors (MAE) are calculated for each year. Then the MAE is calculated for two different evaluation periods. The first period is 2015-2019, and the second 2017-2021. The Mean Absolute Errors for these periods are then compared to the Mean Absolute Errors achieved in these two periods by 11 different Swedish institutions and companies that regularly publish current year annual real GDP growth forecasts. For the period 2015-2019 the 11 forecasters achieved a Mean Absolute Error span of 0.31-0.58 and the chosen SARIMA model achieved a MAE of 0.58. For the period 2017-2021 the 11 forecasters achieved a MAE span of 0.37-0.87, and the chosen SARIMA model achieved a MAE of 1.13. The results indicate that for years without economic instability, a SARIMA model can provide adequate forecasting results. But when a major shock to the economy like the pandemic occurs, to forecast with a SARIMA model might not be very useful.