Forecasting national activity usinglots of international predictors:an application to New Zealand

We look at how large international datasets can improve forecasts of national activity. Weuse the case of New Zealand, an archetypal small open economy. We apply “data-rich” factorand shrinkage methods to tackle the problem of efficiently handling hundreds of predictordata series from many countries. The methods covered are principal components, targetedpredictors, weighted principal components, partial least squares, elastic net and ridgeregression. Using these methods, we assess the marginal predictive content of internationaldata for New Zealand GDP growth. We find that exploiting a large number of internationalpredictors can improve forecasts of our target variable, compared to more traditional modelsbased on small datasets. This is in spite of New Zealand survey data capturing a substantialproportion of the predictive information in the international data. The largest forecastingaccuracy gains from including international predictors are at longer forecast horizons. Theforecasting performance achievable with the data-rich methods differs widely, with shrinkagemethods and partial least squares performing best. We also assess the type of internationaldata that contains the most predictive information for New Zealand growth over our sample.[...] Sandra Eickmeier, Tim Ng