Essays in consumer heterogeneity and expectations

This thesis comprises three essays that examine the origins, cross-section, and consequences of household forecast errors. Drawing on harmonised micro-data from the Michigan Survey of Consumers, the New York Fed Survey of Consumer Expectations, and the ECB Consumer Expectations Survey, the first essay uncovers a single latent “pessimism” factor that declines monotonically with income. Next, the chapter investigates forecast pessimism at a more granular level, and demonstrates that the effect of income is quantitatively large. Exploiting individual level income shifts, it shows that the relationship between income and forecast pessimism survives. Lower-income households systematically over-predict inflation, unemployment, and other macroeconomic outcomes. The second essay investigates one explanation for forecast pessimism: ambiguity aversion. We ask: if ambiguity aversion drives forecast pessimism, what is the source of this ambiguity? The paper outlines two testable hypotheses and tests them. First, using survey-based Ellsberg experiments linked to expectation data, we find that more ambiguity averse agents have larger, positive, forecast errors, but that the elicited preferences covary with income themselves. Second, we find that the response of consumer expectations on receipt of a signal is asymmetric, and that this asymmetry declines with income. Taken together, these two results are consistent with low-income households acting as if they are more ambiguity averse, and the source of this ambiguity being the strength of signals households receive. The final essay embeds these distorted beliefs in a portfolio-choice framework with nominal risk. Non-homothetic robustness is employed, consistent with the conclusions of Chapter 2. When we calibrate the model, we match the empirical profile of inflation forecast errors, as detailed in Chapter 1. We next estimate the welfare cost of distorted inflation beliefs in the model and find that a modest aggregate forecast error has the potential to generate large welfare losses.