Skip to main content
eScholarship
Open Access Publications from the University of California

Forecasting Inflation in Real Time

  • Author(s): Jia, Mingyuan
  • Advisor(s): Chauvet, Marcelle
  • et al.
Abstract

This dissertation is intended to model the dynamics of inflation and forecast short-run

and long-run inflation using high frequency data. It first proposes a mixed-frequency

unobserved component model in which the common permanent and transitory inflation

components have time-varying stochastic volatilities. The key aspects of the model are its

flexibility to describe the changing inflation over time, and its ability to represent distinct

time series properties across price indices at mixed frequencies. More importantly, the

model is applied to builds short-run and long-run coincident indicators of US inflation at

the weekly frequency. The dynamics of the latent inflation factor shows that the persistence

of US inflation has reduced since 1990s due to different components over time. Next, it

proposes a nowcasting model for headline and core inflation of US CPI. The final selected

variables include daily energy price, commodity price, dollar index, weekly gas price,

money stock and monthly survey index. The model’s nowcasting accuracy improves as

information accumulates over the course of a month, and it easily outperforms a variety of

statistical benchmarks. Moreover, it uses a Nelson-Siegel Dynamic Factor model to fit the

monthly term structure of inflation expectation and describes its dynamics over time. The

extracted inflation factors correspond the level, slope and curvature of the term structure

of inflation expectation. It shows that a decomposition of the yield curve spread into

its expectation and risk premia components helps disentangle the channels that connect

fluctuations in Treasury rates and the future state of the economy. In particular, a change

in the yield curve slope due to expected real interest path and inflation expectation path, is

associated with future industrial production growth and probability of recession.

This dissertation adds to the literature by building a mixed-frequency model that can

track inflation in real time and produce better nowcasting results than the existing method,

by fitting the inflation expectation with a dynamic factor model that can describe the

dynamics of the whole term structure and by proving the usefulness of both inflation

expectation slope and real yield spread in predicting future economic activity.

Main Content
Current View