Download now Free registration required
Relying on University of Michigan data on consumers' inflation expectations, the authors establish some stylized facts on the process of inflation expectation formation across different demographic groups. Percentile time series models are employed to test for rationality and to study learning dynamics across the whole cross-sectional spectrum of responses. These display a significant degree of heterogeneity and asymmetry. Income, education, and gender seem to be rather important characteristics when forecasting inflation. In particular, high income, highly educated, and male agents produce lower mean squared errors. Moreover, socioeconomically "Disadvantaged" respondents assume as a reference point their specific consumption basket, while more advantaged respondents actually observe the general price level.
- Format: PDF
- Size: 1356.7 KB