Forecasting GDP growth with financial market data in Finland: Revisiting stylized facts in a small open economy during the financial crisis

Abstract

This paper examines the ability of financial variables to predict future economic growth above and beyond past economic activity in a small open economy in the euro area. We aim to clarify potential differences in forecasting economic activity during different economic circumstances. Our results from Finland suggest that the proper choice of forecasting variables is related to general economic conditions. During steady economic growth, the preferred choice for a financial indicator is the short‐term interest rate combined with past values of output growth. However, during economic turbulence, the traditional term spread and stock returns are more important in forecasting GDP growth. The time‐varying predictive content of the financial variables may be utilized by applying regime‐switching nonlinear forecasting models. We propose a novel application using the negative term spread and observed recession as signals to switch between regimes. This procedure yields a significant improvement in forecasting performance at the one‐year forecast horizon.

Macroeconomic forecasting with mixed data sampling frequencies: Evidence from a small open economy

“…1 Recently, Monteforte and Moretti (2013) proposed a mixed-frequency model for daily forecasts of euro area inflation by combining a monthly index of core inflation with daily data from financial markets, and found that the mixed-frequency model displayed superior predictive performance with respect to forecasts solely based on economic derivatives. Kuosmanen and Vataja (2014) tested the ability of financial variables to predict future economic growth in Finland, and reported that, during steady economic growth, the preferred choice for a financial indicator was the short-term interest rate combined with past values of output growth whereas, during economic turbulence, the traditional term spread and stock returns were more important in forecasting GDP growth. Conrad, Loch, and Rittler (2014) used a mixed data sampling (MIDAS) approach to link the smooth component of daily return oilstock correlations to changes in monthly US macroeconomic variables.…”

Financial variables and economic activity in the Nordic countries

“…This may be because recessions with financial market origins have distinctively different characteristics from downturns driven by non-financial factors (Ng & Wright, 2013). Considerable evidence of instability and structural breaks in predictive relationships has been reported by, e.g., Binswanger (2000;, Stock and Watson (2003), Giacomini and Rossi (2006), Chinn and Kucko (2010) and Kuosmanen and Vataja (2014). In general, previous evidence suggests that linear models may be misspecified and that the time-varying predictive content of financial variables should be considered when forecasting economic activity.…”

“…We employ non-linear modeling to address the time-varying properties of financial variables. More specifically, we apply the inversion-recession signal first proposed by Kuosmanen and Vataja (2014) to distinguish among different growth regimes within the threshold autoregressive (TAR) modeling framework. The predictive ability of TAR models is compared to both the linear financial indicators model and to nonlinear Markov switching models.…”

“…However, this model specification does not take into account the possible time variation in the predictive ability of financial indicators, particularly during unsettled economic conditions. When forecasting GDP growth in the 2000s in the Finnish context, Kuosmanen and Vataja (2014) found it useful to distinguish among different stages of economic activity. It is reasonable to assume that this stage dependency also applies to the other Nordic countries.…”

“…It is reasonable to assume that this stage dependency also applies to the other Nordic countries. Kuosmanen and Vataja (2014) proposed applying the inversion-recession signal to forecast economic activity under changing circumstances. This approach combines the conventional definition of recession and the inversion of the yield curve as the signal to switch between regimes.…”