Forecasting Norwegian GDP : an empirical analysis of categorized macroeconomic data

The topic of this master thesis is forecasting of Norwegian quarterly GDP growth. We aim to research whether a dataset of many variables can forecast Norwegian GDP growth accurate in the period 2014q2 to 2018q1, with forecast horizons of 4-quarters, 8-quarters and 12-quarters. Accuracy will in this thesis be defined as minimizing the root mean square error. Further, we are analyzing which group of categorized variables, based on economic content, that forecast GDP growth most accurately. The forecast is performed based on 148 variables, where we categorize the variables based on economic content, and then perform a Principal Component Analysis within each category. Finally, we investigate whether an index of leading indicators based on the Norwegian economy can forecast accurately. The index is created using the same method as The Conference Board Leading Economic Index for the United States, using corresponding variables for the Norwegian economy. We find that using Principal Component Analysis in forecasting is able to outperform the benchmark of an Autoregressive model. Further, the analysis shows that a category containing production measures forecasts most accurate for all horizon. The forecast model with all 148 variables included performs second most accurate forecasts. Further, the findings suggest that the created index of leading economic indicators for the Norwegian economy is not accurate in terms of forecasting Norwegian GDP growth in the period 2014q2 to 2018q1