Time series analysis of technical progress based on a Bayesian modeling

本文提出了一种基于恒定替代弹性（CES）生产函数的贝叶斯方法来分析要素增长型技术变化。为了估计资本和劳动力增加技术变革的趋势，基于平滑先验方法构建了一组贝叶斯线性模型。为CES生产函数构建的统计模型可以用时变系数的回归模型表示。然而，解释变量的时间序列数据之间的多重共线性给参数估计带来了困难。因此，我们用两个简化模型来表示原始模型。利用贝叶斯线性建模和极大似然法对各简化模型的参数进行估计。然后，基于贝叶斯模型平均，我们得到了一组时变系数的综合估计。

In this paper, we propose a Bayesian approach for analyzing factor-augmenting technical changes based on a constant elasticity of substitution (CES) production function. To estimate trends in capital- and labor-augmenting technical change, a set of Bayesian linear models is constructed based on a smoothness prior approach. A statistical model constructed for the CES production function can then be expressed in a regression model with time-varying coefficients. However, the multicollinearity between the time-series data for the explanatory variables makes parameter estimation difficult. Therefore, we express the original model using two simplified models. Estimates of the parameters for each simplified model are obtained using Bayesian linear modeling and the maximum likelihood method. We then obtain a set of synthetic estimates for the time-varying coefficients based on Bayesian model averaging.