dc.description.abstract | A theoretical model for growth or inflation should be able to reproduce the empirical features of these variables better than competing alternatives. Therefore, it is common practice in the literature, whenever a new model is suggested, to compare its performance with that of a benchmark model. However, while the theoretical models become more and more sophisticated, the benchmark typically remains a simple linear time series model. Recent examples are provided, e.g., by articles in the real business cycle literature or by new-keynesian studies on inflation persistence. While a time series model can provide a reasonable benchmark to evaluate the value added of economic theory relative to the pure explanatory power of the past behavior of the variable, recent developments in time series analysis suggest that more sophisticated time series models could provide more serious benchmarks for economic models. In this paper we evaluate whether these complicated time series models can really outperform standard linear models for GDP growth and inflation, and should therefore substitute them as benchmarks for economic theory based models. Since a complicated model specification can over-fit in sample, i.e. the model can spuriously perform very well compared to simpler alternatives, we conduct the model comparison based on the out of sample forecasting performance. We consider a large variety of models and evaluation criteria, using real time data and a sophisticated bootstrap algorithm to evaluate the statistical significance of our results. Our main conclusion is that in general linear time series models can be hardly beaten if they are carefully specified, and therefore still provide a good benchmark for theoretical models of growth and inflation. However, we also identify some important cases where the adoption of a more complicated benchmark can alter the conclusions of economic analyses about the driving forces of GDP growth and inflation. Therefore, comparing theoretical models also with more sophisticated time series benchmarks can guarantee more robust conclusions. | |