Information reduces overconfidence in inflation forecasts

Researchers exploited a randomized information intervention within the Bank of Italy's Survey of Inflation and Growth Expectations (SIGE) to analyze how access to recent inflation data impacts managers' forecasts.

Firms treated with the latest inflation reading before submitting their forecasts displayed less underreaction at the consensus level and less overreaction at the individual level, compared to non-treated firms.

This intervention effectively moved their forecasts closer to the full-information rational expectations (FIRE) benchmark.

A model integrating noisy information with overconfidence in private information provided the best overall fit to the experimental data.

This framework outperformed alternative explanations such as diagnostic expectations, internal cognitive constraints, or over-extrapolation, suggesting that managers tend to overweight their private signals.

The study addresses a long-standing challenge to the full-information rational expectations (FIRE) assumption in macroeconomic models, which has been questioned by survey data on firms' and households' expectations.

Previous research has identified both underreaction in average forecasts and overreaction in individual forecasts to new information.

To reconcile these seemingly contradictory findings, the literature has explored combining information rigidities with other biases like noisy memory or over-extrapolation.

This paper contributes by leveraging a unique randomized controlled trial within an established survey, allowing for the assessment of how sustained access to information improves the efficiency of belief updating, thereby filling a critical gap in the understanding of expectation formation.