Forecasting Inflation

Author:

Utrera, Francisco Javier

Area of Honors:

Finance

Degree:

Bachelor of Science

Document Type:

Thesis

Thesis Supervisors:

James Alan Miles, Thesis Supervisor
James Alan Miles, Thesis Honors Advisor
Dr. George David Haushalter, Faculty Reader

Keywords:

Inflation
CPI-U
Predicting
Prediction
Forecasts
Monetary
Fiscal
Policy

Abstract:

The main purpose of this paper is to analyze the accuracy of three main kinds of forecasts of CPI-U (Inflation) in the United States since 1982: Arbitrage-Free Model, Survey Models and Statistical Models. After analyzing the accuracy of the three models to predict CPI-U we found: 1. The statistical model ARMA (1,1) has significant predictive capabilities, which undermines the predictive power of professionals and experts in the matter. 2. Means are better predictive measures of CPI-U than Medians. Across the board we found that in all the forecasts that have Mean and Medians the Means were always more accurate than the Medians. 3. Arbitrage-Free Models produce the most inaccurate forecasts. This can potentially be attributed to the Liquidity and Inflation Premiums, in addition to the lack of predictive power of Short-Term T-Bills. More study within this topic is necessary to determine arbitrage opportunities within the TIIS Market. In conclusion, ARMA (1,1) was the best predictor of inflation over long periods of time, using three different frequencies (6 months, 3 months and 1 month). Even though the results were robust, more research on this subject is necessary to arrive at a definite conclusion on which is the best forecast of CPI-U. Specially, we see an increase in accuracy for the Arbitrage Free Model constructed with TIIS.