We evaluate the forecasting performance of time series models for the production side of GDP, that is, for the sectoral real value added series summing up to aggregate output. We focus on two strategies that are typically implemented to model a large number of time series simultaneously: a Bayesian vector autoregressive model (BVAR) and a factor model structure; we then compare them to simple benchmarks. We look at point and density forecasts for aggregate GDP, as well as forecasts of the cross-sectional distribution of sectoral real value added growth in the euro area and Switzerland. We find that the factor model structure outperforms the benchmarks in most tests, and in many cases also the BVAR. An analysis of the covariance matrix of the sectoral forecast errors suggests that the superiority of the factor model can be traced back to its ability to capture sectoral comovement more accurately, and the fact that this gain is especially high in periods of large sectoral dispersion. JEL classification: C11, C32, C38, E32, E37

中文翻译：

---

预测GDP的生产端

我们评估了 GDP 生产方面的时间序列模型的预测性能，即部门实际增加值序列汇总到总产出。我们关注两种通常用于同时对大量时间序列进行建模的策略：贝叶斯向量自回归模型 (BVAR) 和因子模型结构；然后我们将它们与简单的基准进行比较。我们着眼于总 GDP 的点和密度预测，以及欧元区和瑞士部门实际增加值增长的横截面分布预测。我们发现因子模型结构在大多数测试中都优于基准，在许多情况下也优于 BVAR。对部门预测误差协方差矩阵的分析表明，因子模型的优越性可以追溯到它能够更准确地捕捉部门联动，并且在部门分散大的时期这种增益尤其高。JEL 分类：C11、C32、C38、E32、E37