Intertemporal Properties of Real Output: A Bayesian Analysis
用贝叶斯方法分析121个国家战后实际产出的单变量性质，允许分数阶积分，比较了ARIMA和确定性趋势模型，有助于理解冲击的长期持续性。
Abstract This article analyzes the univariate properties of real output and extends traditional analyses in three ways: (1) A Bayesian approach is taken. (2) Extensive cross-country comparisons are performed. Using data from Summers and Heston, cross-country comparisons are derived for postwar data for 121 different countries. (3) A model is developed that allows for the possibility that real output may be fractionally integrated. In addition, standard comparisons between autoregressive integrated moving average (ARIMA) and deterministic trend models are made. The class of fractionally integrated processes, although possessing certain properties useful for generalization of ARIMA models, has received relatively little attention in the macroeconomic literature. Such models are valuable because they allow for greater flexibility in estimating the long-run persistence of shocks. KEY WORDS: Fractional integrationPosterior oddsUnit root