Please rotate your device to landscape mode to view the charts.

Belief Distortions and Macroeconomic Fluctuations

Journal: American Economic Review

Date: 20220701

Author: Bianchi, Francesco; Ludvigson, Sydney C.; Ma, Sai

Abstract:
This paper combines a data-rich environment with a machine learning algorithm to provide new estimates of time-varying systematic expectational errors (belief distortions) embedded in survey responses. We find sizable distortions even for professional forecasters, with all respondent-types overweighting the implicit judgmental component of their forecasts relative to what can be learned from publicly available information. Forecasts of inflation and GDP growth oscillate between optimism and pessimism by large margins, with belief distortions evolving dynamically in response to cyclical shocks. The results suggest that artificial intelligence algorithms can be productively deployed to correct errors in human judgment and improve predictive accuracy.

Link: Google Scholar

Key Findings

Machine Learning Outperforms Survey Forecasts

The machine learning algorithm produces more accurate forecasts than survey respondents across all surveys (SPF, SOC, BC) for both inflation and GDP predictions, with MSE ratios consistently below 1.0

Overreliance on Private Information

Survey respondents consistently place too much weight on their own forecasts relative to publicly available information, with coefficient estimates substantially below 1.0

Dynamic Response to Cyclical Shocks

Survey expectations exhibit initial under-reaction followed by delayed over-reaction to economic shocks, with particularly strong patterns in inflation expectations

Machine vs Survey Forecast Accuracy

  • Machine learning forecasts outperform median survey forecasts across all three surveys
  • For SPF, machine MSE ratio is 0.85 for inflation and 0.88 for GDP growth
  • Largest improvements seen in SOC forecasts with MSE ratio of 0.62 for inflation

Forecast Performance 2013-2018

  • Machine forecasts show largest improvements in recent 5-year period (2013:Q2-2018:Q2)
  • SPF median GDP growth forecasts were 34% higher than actual growth
  • Machine achieved 30-42% improvement in forecast accuracy during this period

Response to Cyclical Shocks

  • Initial under-reaction followed by delayed over-reaction to economic shocks
  • Inflation expectations show greater delayed over-reaction than GDP growth expectations
  • Machine forecasts achieve MSE ratios of 0.6 for inflation and 0.87 for GDP growth during shock periods

Contribution and Implications

  • Demonstrates that artificial intelligence algorithms can effectively detect and correct systematic errors in human economic forecasts
  • Provides new methodology for measuring belief distortions using machine learning in real-time forecasting contexts
  • Suggests practical applications for improving institutional forecasting by complementing human judgment with AI algorithms

Data Sources

  • MSE Forecast Chart: Constructed using data from Table 1 comparing machine and survey mean-squared forecast errors
  • Recent Performance Chart: Based on performance metrics reported in text for 2013:Q2-2018:Q2 period
  • Shock Response Chart: Derived from dynamic response patterns shown in Figure 9 of the paper