Keywords

Inflation forecasting; Time series analysis; Factor-Augmented Vector Autoregression (FAVAR); SARIMA; Principal Component Analysis (PCA)

Abstract

Understanding inflation—particularly across regions and categories—is crucial for effective policymaking, strategic business decisions, and safeguarding vulnerable populations, as it highlights the diverse drivers and impacts of price changes within the economy. This has become increasingly crucial in recent years between the volatile inflation conditions introduced by the COVID-19 pandemic, energy price shocks, and renewed trade tensions and tariffs. This thesis analyzes 77 U.S. monthly inflation time series from 2003 to 2023 using two forecasting approaches: an elementwise Seasonal Autoregressive Integrated Moving Average (SARIMA) model and a Factor-Augmented Vector Autoregressive (FAVAR) model. The data obtained from the Bureau of Labor Statistics includes monthly inflation indices, including national and regional consumer price indexes across key sectors such as food, housing, medical services, and energy.

Results show that both SARIMA and FAVAR outperform a naïve benchmark, with FAVAR demonstrating improved accuracy across regional datasets. SARIMA performed slightly better at the national level, suggesting that univariate models may capture broad inflation trends more effectively. The FAVAR model also offered a notable computational advantage, making it a scalable and efficient approach for high-dimensional time series forecasting.

Thesis Completion Year

2025

Thesis Completion Semester

Spring

Thesis Chair

Xu, Mengyu

College

College of Sciences

Department

Statistics and Data Science

Thesis Discipline

Statistics

Language

English

Access Status

Open Access

Length of Campus Access

None

Campus Location

Orlando (Main) Campus

Download

Plum Print visual indicator of research metrics
PlumX Metrics
  • Usage
    • Downloads: 17
    • Abstract Views: 6
see details

Share

COinS
 

Rights Statement

In Copyright