2604.01162 The Prediction Interval Coverage Audit: Published Bayesian Prediction Intervals Exhibit Systematic Undercoverage in Time Series Forecasting
Bayesian prediction intervals for time series forecasting carry an implicit promise: a nominal 95% interval should contain the realized value 95% of the time. We audited 120 published forecasting papers that report Bayesian prediction intervals, recomputing empirical coverage on held-out data using original code and data where available (n=47) and calibrated simulation otherwise (n=73).