A simple approach to nowcasting GDP growth in CESEE economies

Given the publication time lag inherent in national accounts data, we explore the informational content of higher-frequency indicators that become available during a quarter in nowcasting current-quarter GDP growth rates for 11 Central, Eastern and Southeastern European (CESEE) economies. Building on recent findings, we restrict our choice to three model classes: (1) principal component models, (2) bridge equations and (3) simple autoregressive (AR) models without higher-frequency variables. Moreover, we propose a variety of forecast combinations to arrive at the highest possible forecast accuracy. Our estimation sample starts in the first quarter of 2003, and our evaluation period ranges from the second quarter of 2012 to the fourth quarter of 2017. We find that higher-frequency indicators contain useful information for predicting current economic activity in most of the economies in our sample. Using forecast combinations of models with and without higher-frequency variables yields additional gains in predictive accuracy. The best performers ultimately selected vary strongly across countries: we find 10 different models for 11 countries. Eight country models produce a statistically significantly smaller forecast error than the benchmark. Calculating a CESEE-11 country aggregate based on the individual country forecasts yields a forecast performance that is highly superior to that of the benchmark.