Forecasting consumption expenditure using a dynamic panel model with cross-sectional dependence: the case of Japan

In this study, we predict the future trends of consumption expenditure in disaggregated age groups in both the within-sample and out-of-sample periods. In addition, we incorporate the estimation of a dynamic panel model with cross-sectional dependence into our forecasting methodology. As a whole, our dynamic panel model generates accurate forecasts for within sample. In particular, the accuracy is better in the 40–49 age group, while it is the most inaccurate for the over-70 age group. The out-of-sample period forecast results show that the dynamic panel model generates more accurate than the AR model in almost all age groups. Further, the impact of the COVID-19 shock in 2020 will be retained in many age groups for some time, leading to a decline in consumption. However, after a while, this impact will gradually disappear, and consumption will increase for most age groups. On the other hand, the out-of-sample period forecast results show that the older age group drags out the COVID-19 shock longer than the younger age group and will take longer to recover its consumption levels. In addition, aging of the heads of Japanese households will make it difficult for these households to maintain their current consumption levels unless some measures are taken to deal with the older age group.