Trend Inflation Estimates for Thailand from Disaggregated Data

This paper constructs a new trend inflation measure for Thailand based on the multivariate unobserved components model with stochastic volatility and outlier adjustments (MUCSVO) of Stock and Watson (2015). Similar to core inflation, the MUCSVO constructs a measure of the underlying trend based on disaggregated data, but with time-varying sectoral weights that vary with the volatility, persistence and co-movement of the sectoral inflation series. Based on the empirical results, the majority of sectoral weights show significant time-variation, in contrast to their relatively stable expenditure shares. Volatile food and energy sectors that are typically excluded from core inflation measures also turn out to be less volatile, more persistent and explain approximately 10 percent of filtered trend inflation rate movements. Compared to various other trend inflation measures, we show that the MUCSVO delivers trend estimates that are smoother, has narrower confidence bands, and are able to forecast 8 quarter-ahead average inflation more accurately both in-sample and out-of-sample, especially in the post 2000 period.