The dissertation is an investigation of the theory and measurement of inflationary expectations and inflationary uncertainty. The dissertation extends previous approaches to the measurement of these subjective magnitudes by analyzing the behavior of economic agents in a stochastic environment in which they are unable to distinguish between permanent and transitory changes. The dissertation uses a generalized method of moments procedure to estimate the parameters of linear rational expectations models and a Multi-State Kalman Filter to separate shocks to economic variables into permanent and transitory components.;Chapter II extends the analysis of Chapter I by developing a simple econometric strategy in order to assess if permanent-transitory confusion by the Livingston survey respondents can account for the detection of bias in this survey series. The analysis also compares the inflation forecasts from the Multi-State Kalman Filter to those from the Livingston respondents. The results indicate that while permanent-transitory confusion cannot account for the detection of bias in the Livingston series, the relative forecasting performance of the two predictors depends on the sample period selected.;Chapter III focuses on the measurement of inflationary uncertainty by autoregressive conditional heteroscedastic (ARCH) models. The analysis demonstrates that in finite samples permanent-transitory confusion by an econometrician can result in the detection of ARCH effects even though there are no ARCH effects in the population. The Multi-State Kalman Filter is used to generate forecasts of the quarterly growth rate of the United States CPI and to examine this time series for episodes of large permanent and large transitory shocks. The results indicate that ARCH effects previously detected in this time series are not heteroscedastic properties of the population.;Chapter I uses a generalized method of moments estimator to reexamine the rationality of the Livingston price expectations data and the SRC expected price change data. The analysis takes careful account of not only serial correlation and conditional heteroscedasticity in the disturbance terms but also measurement error in the reported concensus forecasts. The results strongly reject the absence of bias for the Livingston series but find an absence of bias for the SRC series.
展开▼
