Expectations
Read the consensus path, then read the revisions
The Survey of Professional Forecasters is the Federal Reserve Bank of Philadelphia's quarterly panel of professional macro forecasts. The shipped tables and unemployment vintage chart report the selected median statistic, never a blend and never a house forecast. The macro path, dispersion, and error panels are explicitly median-only. The recession panels instead summarize individual probability responses. The local SPF parquet runs from 1968Q4 to 2026Q2.
Data as of Philadelphia Fed SPF parquet through 2026Q2
2 because SPF horizon 1 is the prior quarter's preliminary value, not a forward quarter. RECESS uses a separate official mapping, with RECESS1 targeting the survey quarter itself. The tables and vintage chart honor the mean-versus-median switch. The three panels after the revision table stay median-only and say so plainly.Respondent recession probabilities
2026Q2 responses cover the current quarter through four quarters ahead
Each point is the mean probability assigned by individual professional forecasters to a quarter-over-quarter decline in real GDP. The dashed line is the median response and the band is the middle 50%. This is not the fraction of respondents predicting a recession, a FinObservatory model output, or a realized recession indicator.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters RECESS1 targets a decline from the prior quarter to the survey quarter. RECESS2 through RECESS5 target declines in the following one through four quarters. Counts include non-null responses only; no survey weights are applied. Methodology
Next-quarter history
The Anxious Index history keeps one event and one horizon fixed
Every point uses RECESS2, the Philadelphia Fed's Anxious Index: the probability respondents assign to real GDP declining in the quarter immediately after the survey quarter. Response counts vary across rounds, and null responses remain absent rather than becoming zeroes.
Latest, 2026Q2 for target 2026Q3: mean respondent-assigned probability 25.1%, median 20.0%, middle 50% 15.0% to 31.3%, n=24.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters The event is a quarter-over-quarter decline in real output, not an NBER recession call. The output concept is real GNP before 1992, fixed-weighted real GDP from 1992 through 1995, and chain-weighted real GDP thereafter. Methodology
Latest SPF path
2026Q2 quarterly path and calendar-year averages
Five headline SPF variables from the latest loaded survey: the current-quarter nowcast, the next four quarterly horizons, and the current and next two calendar-year averages where the survey carries them.
| Variable | Nowcast | Q+1 | Q+2 | Q+3 | Q+4 | CY | CY+1 | CY+2 |
|---|---|---|---|---|---|---|---|---|
Real GDP RGDP | bn real $ | 24,301 | 24,435 | 24,530 | 24,648 | 24,765 | 24,363 | 24,822 | 25,363 |
CPI inflation CPI | % | 6.0 | 3.0 | 2.5 | 2.5 | 2.4 | 3.5 | 2.5 | 2.4 |
Core PCE inflation COREPCE | % | 3.4 | 2.9 | 2.6 | 2.6 | 2.4 | 3.3 | 2.4 | 2.1 |
Unemployment rate UNEMP | % | 4.4 | 4.5 | 4.5 | 4.5 | 4.5 | 4.4 | 4.5 | 4.4 |
10-year Treasury TBOND | % | 4.3 | 4.3 | 4.3 | 4.3 | 4.2 | 4.3 | 4.2 | 4.2 |
Revisions
What changed versus 2026Q1
The revision table isolates three cuts of the same survey path: the current-quarter nowcast, the four-quarters-ahead read, and the next calendar-year average. A positive number means the 2026Q2 consensus moved up versus 2026Q1; a negative number means it moved down.
| Variable | Nowcast revision | 4-quarter-ahead revision | Next calendar year revision |
|---|---|---|---|
Real GDP RGDP | +31 bn real $ 24,301 from 24,270 | −25 bn real $ 24,765 from 24,790 | −163 bn real $ 24,822 from 24,985 |
CPI inflation CPI | +3.3 % 6.0 from 2.7 | −0.0 % 2.4 from 2.4 | +0.0 % 2.5 from 2.5 |
Core PCE inflation COREPCE | +0.5 % 3.4 from 2.9 | +0.0 % 2.4 from 2.4 | +0.1 % 2.4 from 2.3 |
Unemployment rate UNEMP | −0.0 % 4.4 from 4.4 | +0.0 % 4.5 from 4.4 | +0.1 % 4.5 from 4.4 |
10-year Treasury TBOND | +0.2 % 4.3 from 4.1 | +0.2 % 4.2 from 4.0 | +0.2 % 4.2 from 4.0 |
Median-only latest paths
2026Q2 medians trace the nowcast through four quarters ahead
These four small multiples always show medians, even when the tables above are switched to means. CPI, unemployment, and the 10-year Treasury use published SPF point medians. Real GDP growth is computed for each respondent from paired adjacent level forecasts before taking the median.
Real GDP growth
2027Q2: 1.9716%Median-only path; target-quarter labels are derived from the survey horizon.
CPI inflation
2027Q2: 2.4000%Median-only path; target-quarter labels are derived from the survey horizon.
Unemployment rate
2027Q2: 4.4526%Median-only path; target-quarter labels are derived from the survey horizon.
10-year Treasury
2027Q2: 4.2000%Median-only path; target-quarter labels are derived from the survey horizon.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters Horizon 2 is the nowcast; horizons 3 through 6 are one through four quarters ahead. Methodology
Median-only forecaster dispersion
The middle half of one-year-ahead forecasts changes width across survey rounds
Each history is the horizon-6 interquartile range from anonymized respondent microdata. The latest quartile bounds and response count are stated separately for every variable. These distributions remain median-only context and do not change when the shipped tables display means.
Real GDP growth
2026Q2: median 1.9716%, middle half 1.7923% to 2.1554%, IQR 0.3630 percentage points, n=32.
CPI inflation
2026Q2: median 2.4000%, middle half 2.1370% to 2.9333%, IQR 0.7963 percentage points, n=31.
Unemployment rate
2026Q2: median 4.4526%, middle half 4.3167% to 4.5941%, IQR 0.2774 percentage points, n=31.
10-year Treasury
2026Q2: median 4.2000%, middle half 4.0012% to 4.3300%, IQR 0.3288 percentage points, n=29.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters Quartiles use continuous 25th and 75th percentiles. Real GDP is transformed respondent by respondent before summarization. Methodology
Median-only forecast errors
Signed misses show where one-year-ahead consensus forecasts ran high or low
Error is forecast minus realized, so values above zero were overforecasts and values below zero were underforecasts. Only CPI, unemployment, and the 10-year Treasury appear because each has a separately sourced realized mapping already on disk. The emphasized zero line marks an exact forecast.
CPI inflation
Target 2026Q2: forecast 2.6223%, realized 6.0717%, error −3.4494 percentage points.
Unemployment rate
Target 2026Q2: forecast 4.5000%, realized 4.2667%, error +0.2333 percentage points.
10-year Treasury
Target 2026Q2: forecast 4.1000%, realized 4.4219%, error −0.3219 percentage points.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters | Federal Reserve Economic Data, CPIAUCSL | Federal Reserve Economic Data, UNRATE | Federal Reserve Economic Data, DGS10 CPI uses annualized growth in quarterly-average CPIAUCSL; unemployment and Treasury use quarterly averages of UNRATE and DGS10. Methodology
Vintage chart
The unemployment misses are visible survey by survey, not hidden in one backtest line
Every thin path is one median SPF unemployment-rate forecast from 2019Q1 through 2025Q2, plotted from the survey quarter through four quarters ahead. The black line is the realized quarterly average unemployment rate from UNRATE, carried through 2026Q2. This is the most useful SPF exhibit because it shows both revision and error at once: which vintages saw the labor market softening, which did not, and how late the consensus turned.
Source: Federal Reserve Bank of Philadelphia, Survey of Professional Forecasters | Federal Reserve Economic Data, UNRATE Observed unemployment is the quarterly average of monthly UNRATE. Vintages shown are 2019Q1 through 2025Q2 so each four-quarter-ahead path has a realized outcome inside the local data span. Methodology