Dynamic analysis in productivity, oil shock, and recession
- Author(s): Katayama, Munechika
- et al.
The first chapter analyzes the predictive power of the interest rate for various industry-level measures of productivity growth. Although industry-level data are only available at an annual frequency, by using the state space model and the Kalman filter, it is possible to perform the Granger-causality test at a quarterly frequency. The results highlight the heterogeneous nature of predictive power and suggest that the nonexogeneity of the Solow residual reported by Evans (1992) is due to manufacturing industries. In addition, two case studies on industries in which we can obtain an appropriate measure of capital utilization rate show that the forecasting ability of the interest rate diminishes after taking account of variable capital utilization in TFP growth. The second chapter studies macroeconomic consequences of oil-price shocks. Output responses to oil-price shocks not only tend to be weaker, but also to peak earlier recently. This chapter builds a model that incorporates a realistic structure of US petroleum consumption and explores three possible explanations for the changes. The first is based on deregulation in the transportation sector, which has brought more competition and improved efficiency in the industry. The second is overall improvements in use of energy. The third is less persistence of the oil-price shock. Under realistic parameter values, it is demonstrated that all three factors play an important role quantitatively. These three factors together could account for a large portion of the changes over time. The third chapter examines forecasting models of recession probabilities. There are two margins to improve forecasting accuracy: including additional variables and using different functional form. Using out-of-sample and cross validation methods, I systematically compare the performance of various forecasting models that differ in terms of variables included and functional forms used. I find substantial gains from including additional variables, such as the Samp;P 500 and non-farm employment growth, together with the term spread. Additionally, we can further improve forecasting accuracy by utilizing a non- Normal distribution. I also explore this possibility by using the generalized Edgeworth expansion. Resulting predictions outperform ones from a probit model in all three measures of forecasting accuracy considered