A fractionally integrated panel data model with a multi-level cross-sectional dependence is proposed. Such dependence is driven by a factor structure that captures comovements between blocks of variables through top-level factors, and within these blocks by non-pervasive factors. The model can include stationary and non-stationary variables, which makes it flexible enough to analyze relevant dynamics that are frequently found in macroeconomic and financial panels. The estimation methodology is based on fractionally differenced block-by-block cross-sectional averages. Monte Carlo simulations suggest that the procedure performs well in typical samples sizes. This methodology is applied to study the long-run relationship between energy consumption and economic growth. The main results suggest that estimates in some empirical studies may have some positive biases caused by neglecting the presence non-pervasive cross-sectional dependence and long-range dependence processes.

提出了一种具有多层次横截面相关性的分数集成面板数据模型。这种依赖是由一个因子结构驱动的，该结构通过顶级因子捕捉变量块之间的联动，以及这些块内的非普遍因素。该模型可以包括固定和非固定变量，这使其足够灵活，可以分析宏观经济和金融面板中常见的相关动态。估计方法基于逐块横截面平均值的分数差异。蒙特卡洛模拟表明该过程在典型样本量中表现良好。该方法用于研究能源消耗与经济增长之间的长期关系。