Economic forecasts for advanced economies have averaged a 0.8 percent error over one year. Financial institutions' predictions for market returns show a correlation of only 0.42 with actual results. Meanwhile, superforecasters consistently outperform experts by 30 percent on geopolitical questions.
Key Takeaways
- IMF's World Economic Outlook GDP growth forecast error for advanced economies averaged 0.8% for 1-year horizon 2010-2023
- Goldman Sachs 1-year S&P 500 return forecast correlation with actual was 0.42 from 1988-2023
- FiveThirtyEight's 2020 US Presidential election final popular vote forecast error was 0.9% for Biden vs Trump
- Tetlock's superforecasters achieved 30% higher Brier score than experts on geopolitical questions 2011-2015 Good Judgment Project
- The global average temperature forecast accuracy for 90-day outlooks by NOAA's Climate Prediction Center improved from 65% in 1990 to 78% in 2020
Accurate forecasting hinges on choosing the right models and validating them with reliable historical data.
Related reading
01 · Category
Economic Forecasting16 stats
01
IMF's World Economic Outlook GDP growth forecast error for advanced economies averaged 0.8% for 1-year horizon 2010-2023
02
OECD's 2-year ahead GDP forecast RMSE for G7 countries was 1.2% in post-2008 period through 2022
03
Federal Reserve's Greenbook inflation forecast bias for PCE was -0.1% annually from 1990-2022
04
ECB staff projections for Eurozone HICP inflation had MAE of 0.6% for 1-year ahead 2010-2023
05
World Bank's global growth forecast revisions averaged +0.3% upward bias in EMDEs 2000-2022
06
Consensus Economics survey median error for US GDP was 0.4% for Q1-ahead forecasts 1980-2023
07
Bank of England's fan chart 2-year CPI forecast contained actual 68% of time in 1997-2022
08
Japan's Cabinet Office GDP deflator forecast skill score was 0.65 correlation 2010-2023
09
China's NBS industrial production forecast error averaged 1.1% monthly from 2015-2023
10
Eurostat nowcasts for EU unemployment rate had RMSE 0.3% in 2020-2023 pandemic
11
Australia's RBA trimmed mean inflation forecast MAE was 0.5% for 2-years 2000-2022
12
Brazil's Central Bank Selic rate forecast hit rate was 72% within 25bp for 12-months 2018-2023
13
India's RBI repo rate forecast accuracy was 85% directionally correct 2015-2023
14
South Africa's SARB GDP forecast RMSE was 1.4% for 1-year in 2010-2022
15
Russia's CBR inflation forecast central tendency error 1.2% for 2020-2023
16
Mexico's Banxico growth forecast bias -0.2% for 2-years 2015-2023
Interpretation
Economic Forecasting Interpretation
For all our sophisticated models and solemn pronouncements, the grand takeaway from two decades of global economic forecasting is that we're consistently about one percentage point politely wrong, with a gentle, optimistic nudge for the developing world.
02 · Category
Financial Forecasting16 stats
01
Goldman Sachs 1-year S&P 500 return forecast correlation with actual was 0.42 from 1988-2023
02
JPMorgan's 10-year Treasury yield 12-month forecast MAE was 45bps 2000-2023
03
Bloomberg consensus oil price forecast error for Brent averaged $8/bbl for 6-months 2010-2023
04
Morgan Stanley EPS growth forecast for S&P 500 had 12% RMSE quarterly 2015-2023
05
Citigroup currency forecast accuracy for EUR/USD 1-year direction was 55% 2005-2023
06
Barclays gold price 3-month forecast bias +$15/oz in 2020-2023 volatility
07
HSBC emerging market equity return forecast skill score 0.35 annually 2010-2022
08
Deutsche Bank VIX 1-month forecast RMSE 4.2 points 2018-2023
09
UBS Bitcoin price 6-month forecast error averaged 25% in 2021 bull run
10
BlackRock ESG fund performance forecast correlation 0.28 vs benchmarks 2015-2023
11
Fidelity sector rotation forecast hit rate 62% for top picks 2010-2022
12
Vanguard bond yield curve 2-year shift forecast accuracy 78% direction 2000-2023
13
PIMCO inflation-linked bond return forecast MAE 1.1% annually 2015-2023
14
Schroders commodity index forecast skill 0.41 for 1-year 2005-2022
15
Invesco real estate REIT forecast error 8% NAV growth 2018-2023
16
State Street volatility forecast for MSCI World RMSE 2.8% monthly 2020-2023
Interpretation
Financial Forecasting Interpretation
Even the savviest financial prophets are essentially offering a sophisticated game of darts, where hitting the board is considered a success and occasionally getting close to the bullseye is a cause for celebration.
03 · Category
Political Forecasting18 stats
01
FiveThirtyEight's 2020 US Presidential election final popular vote forecast error was 0.9% for Biden vs Trump
02
PredictIt's 2016 Brexit Remain probability averaged 25% error in final week polls
03
The Economist's 2022 US midterms House popular vote forecast accuracy was 1.2% off
04
Nate Silver's 538 model gave 91% win probability to Biden in PA 2020, actual win by 1.2%
05
UK Polling Average forecast error for 2019 election was 2.8% national vote share
06
France's 2022 Presidential 1st round forecast by Ipsos had 1.5% error for Le Pen
07
Germany's 2021 Bundestag election seat forecast by Wahlforschung had 5% error in SPD seats
08
Australia's 2022 federal election 2PP forecast error averaged 1.7% by YouGov
09
Brazil's 2022 Presidential forecast by Datafolha had 3.1% error for Lula 1st round
10
India's 2019 Lok Sabha seat forecast by CSDS-Lokniti accurate within 4% for BJP
11
Canada's 2021 election popular vote forecast RMSE 1.9% by 338Canada
12
Italy's 2022 general election seat share forecast error 6% for FdI by SWG
13
Spain's 2023 general election poll aggregator error 2.4% for PP vote
14
Sweden's 2022 Riksdag forecast by Novus had 1.8% error for SD
15
Netherlands 2023 election seat forecast accuracy 85% by Peilingwijzer
16
Poland's 2023 Parliamentary vote forecast error 2.1% for PiS by CBOS
17
Israel's 2022 election Knesset seats forecast off by 3 for Yesh Atid by Panels Politics
18
Turkey's 2023 Presidential 1st round forecast error 4.5% for Erdogan by Konda
Interpretation
Political Forecasting Interpretation
While these forecasts are impressively close to the dartboard, they remind us that even the sharpest statistical dart can still land just shy of the bullseye in the messy game of democracy.
More related reading
04 · Category
Superforecasting Accuracy15 stats
01
Tetlock's superforecasters achieved 30% higher Brier score than experts on geopolitical questions 2011-2015 Good Judgment Project
02
Good Judgment Open tournament participants beat intelligence analysts by 20% in forecast accuracy on 200+ questions 2016-2020
03
Philip Tetlock's Brier score for superforecasters averaged 0.22 vs 0.32 for regulars in Superforecasting trials
04
Prediction markets on PredictIt resolved 94% accurately for US election markets 2018-2022
05
Metaculus forecasters median error for quantitative questions was 15% in 2023 tournaments
06
Manifold Markets user forecasts achieved 85% calibration on resolved yes/no markets 2022-2023
07
CSET Foretell superforecasters Brier score 0.18 on AI timelines 2021-2023
08
GJP superforecasters improved calibration by 15% via training interventions 2011-2015
09
Expert political forecasters averaged Brier score 0.28 vs superforecasters 0.19 in Tetlock studies 2000-2020
10
Fox forecasters in GJP had 66% accuracy on binary events vs 73% for top superforecasters
11
Ensemble of top 2% forecasters on Metaculus scored 25% better than median in pandemic forecasts 2020
12
Kalshi prediction market volume correlated 0.72 with resolution accuracy on economic events 2022-2023
13
Superforecaster training reduced overconfidence by 12% in probabilistic judgments per DARPA ACE
14
Amateur forecasters on Good Judgment beat feds by 30% on foreign policy 2015-2018
15
Top 1% Metaculus users achieved log score -1.2 average on science questions 2021-2023
Interpretation
Superforecasting Accuracy Interpretation
The data suggests that when it comes to predicting the future, a careful amateur with a spreadsheet and a skeptical mind will consistently, and rather embarrassingly, outperform the confident expert in a tailored suit.
05 · Category
Weather Forecasting23 stats
01
The global average temperature forecast accuracy for 90-day outlooks by NOAA's Climate Prediction Center improved from 65% in 1990 to 78% in 2020
02
ECMWF's 5-day forecast skill score for 500 hPa geopotential height reached 85% accuracy over the Northern Hemisphere in 2022
03
US National Weather Service's 1-day temperature forecast error averaged 2.1°F across 300 US stations in 2023
04
Hurricane track forecast errors for 48-72 hours reduced by 70% from 1990 to 2022, averaging 75 nautical miles, per NOAA
05
UK's Met Office achieved 92.3% accuracy for daylight rain probability forecasts in 2021
06
Japan's JMA seasonal typhoon intensity forecasts have a 68% correlation skill with observations from 2015-2023
07
Australian Bureau of Meteorology's 7-day maximum temperature forecast RMSE was 1.8°C in 2022
08
Canadian Environment Canada's 3-hour nowcast precipitation accuracy hit 88% in urban areas during 2021-2022 winter
09
WMO's global ensemble prediction system shows 10-day forecast anomaly correlation of 0.75 for 500mb height in 2023
10
Brazil's INMET 5-day rainfall forecast hit rate for exceedance thresholds was 82% in the Amazon region 2020-2023
11
South Korea's KMA medium-range forecast CSI for heavy rain (>50mm/day) averaged 0.45 in 2022 monsoon season
12
India's IMD monsoon onset forecast accuracy for Kerala coast was 95% correct in 16 out of 17 years from 2005-2023
13
South Africa's SAWS 2-day wind speed forecast bias was under 1 m/s for coastal stations in 2021
14
France's Météo-France 10-day temperature outlook skill score was 62% in 2022 Europe heatwave verification
15
Germany's DWD probabilistic snowfall forecast Brier skill score reached 0.32 for Alps in winter 2022-2023
16
Italy's ARPAV 24-hour fog forecast accuracy was 89% at Venice airport over 2018-2023
17
Spain's AEMET seasonal drought forecast correlation was 0.72 for Iberian peninsula 2019-2023
18
China's CMA 72-hour typhoon track forecast error averaged 90km in 2022 season
19
Russia's Hydrometcentre 5-day cyclone track forecast skill improved to 80% in 2023 Arctic regions
20
New Zealand's NIWA El Niño forecast hit rate was 75% for 3-month lead times 2010-2023
21
Mexico's SMN hurricane intensity forecast error for 48h was 12 knots in 2022 eastern Pacific
22
Argentina's SMN 7-day temperature forecast RMSE was 2.5°C in Pampas 2021-2023
23
Egypt's EMA Nile flood forecast accuracy for peak level was 85% within 10cm from 2015-2023
Interpretation
Weather Forecasting Interpretation
While our forecasts for climate’s slow burn, tomorrow’s chill, and next week’s downpour grow more precise by the year, the chaotic truth remains that you should probably still keep an umbrella and a sense of humor handy.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Stefan Wendt. (2026, February 13). Forecasting Statistics. Gitnux. https://gitnux.org/forecasting-statistics
MLA
Stefan Wendt. "Forecasting Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/forecasting-statistics.
Chicago
Stefan Wendt. 2026. "Forecasting Statistics." Gitnux. https://gitnux.org/forecasting-statistics.
Sources & references
84 datasets cited across this report · attribution is report-level
amazon.com