Financial variables and economic activity: forecasting issues

Abstract

The present research investigates the forecasting performance of various monthly indicators for the US industrial production growth rates in long horizon. Our focus in on financial variables that are often associated with future output growth, such as stock prices, interest rates and interest rate spreads. We also include housing and precious metal (gold) prices as well as commodity prices such as oil prices to see whether they produce marginal forecasting information. We use out of sample forecast evaluation and especially the iterated multistep approach, implementing forecast combination across observation windows of different lengths. For each observation window we use the sequential procedure of Bai and Perron (1998, 2003) to take into consideration the multiple structural breaks. The empirical results indicate that estimations without structural breaks lose in forecasting stability. Therefore we re-estimate taking into consideration five break points for each observation window. We also confirm the significant role of monetary policy in taking economic decisions.