Predicting Gdp Growth in the Euro Area

A vast literature in finance and macroeconomics is dedicated to the forecasting ability of financial variables for real economic activity. Since GDP growth is one of the most important macroeconomic indicators and, consequently, the main subject of interest for both society and policymakers, forecasting GDP is probably one of the most discussed topics in the literature. However, empirical evidence is mixed and results are not robust with respect to model specification, sample choice and forecast horizon, as well as, to the variables that should be used.

GDP measures economic output, representing business activity and supporting the country’s level of productivity. On the one hand, economists rely on GDP data to determine whether we are in expansion or contraction, while on the other hand, monetary policymakers use GDP when measuring the state of the economy and inflation. This economic indicator gains therefore an enormous relevance for several agents’ interest in the economy’s wealth and future direction (expansion or recession). Finding a way to model such a variable is far from consensual and it has been intensively studied in the past. Hence, it is of interest to find a good model to predict GDP.

Empirical studies often choose financial variables that are considered as leading indicators of economic activity, such as stock returns, interest rates, interest rates spreads, monetary aggregates, and others. Banerjee et al. (2003) using an extensive list of leading indicators for output growth found that measures of short and long-term interest rates, as well as interest rate spreads are the best performing single indicators for GDP growth. Furthermore, Moneta (2005) found that the yield spread is a powerful variable for predicting recessions in the euro area, a result that was also confirmed by e.g Duarte et al. (2005), who used aggregated data for the euro area and observed the ability of the spread to predict output growth and recessions. However, there are studies that consider that the separate use of the long-term and the short-term interest rate is more powerful than the yield curve. Moreover, there is empirical evidence that the forecasting ability of the term spread has decreased over the past decade. For instance, Haubrich and Dombrosky (1996) and Dotsey (1998) confirmed, using US data and linear models, that from 1985 there is a sharp decrease in the predictability power of the term spread. In addition to these studies, we can further identify other works suggesting that the term structure and monetary aggregates are associated with future economic activity, e.g. Harvey (1988, 1997); Estrella and Hardouvelis (1991); Plosser and Rouwenhorst (1994) and Hamilton and Kim (2002).

The capability of the spread to predict recessions or economic activity can be explained using an example. Image a country that is currently enjoying a strong economic