Bundesbank Paper Finds US SPF Bin Width Changes Distort Measured Forecast Uncertainty

Key Takeaways

• A Deutsche Bundesbank discussion paper shows that changes in histogram bin widths in the US Survey of Professional Forecasters (SPF) created structural breaks in measured inflation and growth uncertainty, even where true uncertainty was largely unchanged.
• The paper is research, not binding regulation, but it matters for compliance and risk teams that rely on SPF-based uncertainty measures in stress testing, model validation, and scenario calibration.

Deep Dive

A Deutsche Bundesbank discussion paper shows that changes in histogram bin widths in the US Survey of Professional Forecasters (SPF) created structural breaks in measured inflation and growth uncertainty, even where true uncertainty was largely unchanged. The paper is research, not binding regulation, but it matters for compliance and risk teams that rely on SPF-based uncertainty measures in stress testing, model validation, and scenario calibration.

The paper's central finding is that bin width changes produced permanent shifts in measured uncertainty. Key dates are 1992q1, when bins for GDP deflator and real GDP growth were halved from 2 pp to 1 pp. In 2014q1, deflator bins were halved again to 0.5 pp. In 2020q2, real GDP bins were widened and made uneven because of COVID-19, and in 2024q2 most were narrowed again. The authors argue the 2014 drop in measured inflation uncertainty was an artifact of survey design. Core CPI and core PCE inflation, whose bin widths did not change, showed stable uncertainty around that date. They propose a break adjustment and horizon-specific bin widths so that measured uncertainty better matches underlying uncertainty. The authors also note that many prior studies of forecaster over- or underconfidence, forecast accuracy, and uncertainty may be affected. Their work rests on assumptions, including that changing the number of bins does not alter measured uncertainty, and the full structural break testing results were not included in the excerpt provided.

Recommended actions: (1) inventory internal models, stress tests, and dashboards that use SPF histogram-based uncertainty or volatility inputs; (2) test those series for structural breaks coinciding with the survey redesign dates above; (3) document the limitation, and consider break adjustments or restricting samples to periods with consistent bin definitions; (4) have model validation review conclusions about forecast confidence or tail risk drawn from unadjusted data; (5) monitor the paper's progress toward peer-reviewed publication and any SPF methodology changes by the Philadelphia Fed.

Risk implications are moderate and mostly confined to model and data risk. Unadjusted SPF series may understate uncertainty after 1992 and 2014 and distort it around 2020 and 2024, which could lead to miscalibrated scenario severity or overconfident risk estimates. There is no direct regulatory breach exposure, but weak data governance over external inputs could draw supervisory scrutiny under model risk management expectations.

Source: Deutsche Bundesbank