“…Our choice of predictors follows Behrens et al (2018aBehrens et al ( , 2018b and Döpke, Fritsche, and Pierdzioch (2017). As financial predictors, we use the US federal funds rate, German money market rate (3 months), the term spread (the difference between the monthly averages of the yield on debt securities with a maturity of more than 3 years and the monthly average money market rate), and the continuously compounded year-on-year returns on the OECD share price index for Germany.…”
We contribute to recent research on the joint evaluation of the properties of macroeconomic forecasts in a multivariate setting. The specific property of forecasts that we are interested in is their joint efficiency. We study the joint efficiency of forecasts by means of multivariate random forests, which we use to model the links between forecast errors and predictor variables in a forecaster's information set. We then use permutation tests to study whether the Mahalanobis distance between the predicted forecast errors for the growth and inflation forecasts of four leading German economic research institutes and actual forecast errors is significantly smaller than under the null hypothesis of forecast efficiency. We reject joint efficiency in several cases, but also document heterogeneity across research institutes with regard to the joint efficiency of their forecasts.
“…Our choice of predictors follows Behrens et al (2018aBehrens et al ( , 2018b and Döpke, Fritsche, and Pierdzioch (2017). As financial predictors, we use the US federal funds rate, German money market rate (3 months), the term spread (the difference between the monthly averages of the yield on debt securities with a maturity of more than 3 years and the monthly average money market rate), and the continuously compounded year-on-year returns on the OECD share price index for Germany.…”
We contribute to recent research on the joint evaluation of the properties of macroeconomic forecasts in a multivariate setting. The specific property of forecasts that we are interested in is their joint efficiency. We study the joint efficiency of forecasts by means of multivariate random forests, which we use to model the links between forecast errors and predictor variables in a forecaster's information set. We then use permutation tests to study whether the Mahalanobis distance between the predicted forecast errors for the growth and inflation forecasts of four leading German economic research institutes and actual forecast errors is significantly smaller than under the null hypothesis of forecast efficiency. We reject joint efficiency in several cases, but also document heterogeneity across research institutes with regard to the joint efficiency of their forecasts.
“…Results from the combined recession probability forecasts show both in‐sample and out‐of‐sample ability for the two economies. Döpke, Fritsche, and Pierdzioch () using Germany data and boosted regression trees found that measures of the short‐term interest rate and the term spread are important leading indicators of recession and that whilst the importance of the former has declined over the years, the term spread and the stock market have gained in importance.…”
This paper decomposes the term spread into the expectation and the term premium components using a fractional integration approach and subsequently uses same with the economic policy uncertainty index to forecast the probability of recession in South Africa. We use different specifications of the probit model and quarterly data from 1990:1 to 2012:1. Our out-of-sample results show that the model that incorporates the expectation component and economic policy uncertainty provides the best forecast of recession. All three recession periods in our sample were accurately dictated by the prediction models and the best forecast occurred at the four quarters ahead horizon. A robustness check with a longer sample from 1946q1 to 2017q4 but excluding the factors and economic policy uncertainty due to data limitation provided justification for decomposing the term spread as the model with the expected spread turned out to be the best. We draw the implications of these findings.
“…Boosting techniques have been used by economists to model exchange rates (Berge, 2014) and to forecast output (Buchen & Wohlrabe, 2011;Robinzonov, Tutz, & Hothorn, 2012) and other macroeconomic variables (Wohlrabe & Buchen, 2014). Applications of BRT algorithms can be found in the research by Mittnik et al (2015), who use BRT to model stock market volatility, and Ng (2014) and Döpke, Fritsche, and Pierdzioch (2017), who use BRT techniques to forecast recessions.…”
We use a machine‐learning algorithm known as boosted regression trees (BRT) to implement an orthogonality test of the rationality of aggregate stock market forecasts. The BRT algorithm endogenously selects the predictor variables used to proxy the information set of forecasters so as to maximize the predictive power for the forecast error. The BRT algorithm also accounts for a potential non‐linear dependence of the forecast error on the predictor variables and for interdependencies between the predictor variables. Our main finding is that, given our set of predictor variables, the rational expectations hypothesis (REH) cannot be rejected for short‐term forecasts and that there is evidence against the REH for longer term forecasts. Results for three different groups of forecasters corroborate our main finding.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.