United States Government Budget
United States · Annual series · EC-US-GBP
Latest, 2025
-7.1 pts
+0.8 pts from 2024
Forecast
Forecast for 2028: -22.1 pts, with an 80% interval of -68.8 pts to -13.0 pts.
| Period | Forecast | 80% low | 80% high |
|---|---|---|---|
| 2026 | -10.4 pts | -10.6 pts | -10.3 pts |
| 2027 | -23.8 pts | -25.0 pts | 14.3 pts |
| 2028 | -22.1 pts | -68.8 pts | -13.0 pts |
AI confidence
Needs refreshThe data refresh is due. Confidence will be reassessed against the refreshed history.
Model: Hierarchical neural network (neural network)
- Accuracy vs naive
- n/aNot evaluated
- Direction
- n/aNot evaluated
- Turning points
- n/aNot evaluated
How this assessment is made
AI confidence describes historical evidence for this forecast, not a probability of being correct. Each backtest fits on data available before its test window. The baseline carries the last value forward. Supported evidence requires at least six complete, non-overlapping tests, an error ratio below 0.90 (rounded to two decimals), and wins in a majority of tests. Ratios from 0.90 to below 1.10 are on par; 1.10 or above is weak evidence. Inconsistent wins produce mixed evidence. For a selected model, only the recent tests excluded from model selection determine this label. These are historical simulations using the available data, not a live trading track record; revisions and foundation-model pretraining can limit how closely they reproduce a past decision. Past performance can change.
Model comparison
Previous comparisonThe stored comparison uses earlier data or a previous model roster. An eligible nightly batch can compare both candidates, subject to call budgets and retry limits.
Models tested
Hierarchical neural network scored lowest on the 6 older tests (0.86) and was chosen. The 6 recent tests played no part in the choice and give its grade (0.93). The search for a model that beats the naive forecast on both sets of tests goes on. Scores are MASE: a model's errors divided by the naive forecast's, so 1 is a tie.
4 model classes completed the tests: Neural network, Foundation model, Statistical, Machine learning. Previous model roster.
| Model | Class and compute | Older tests | Recent tests | Role |
|---|---|---|---|---|
| Hierarchical neural network (neural network) | Neural network · CPU | 0.86 | 0.93 | Serving |
| State-space model, 167M (foundation model) | Foundation model · CPU | 0.96 | 0.88 | Tested |
| Naive forecast (last value held) | Benchmark · CPU | 1.00 | 1.00 | Benchmark |
| Agentic Investor TSFM (foundation model) | Foundation model · CPU | 1.02 | 0.95 | Tested |
| Exponential smoothing (statistical) | Statistical · CPU | 1.22 | 1.00 | Tested |
| Gradient-boosted trees (machine learning) | Machine learning · CPU | 1.35 | 1.03 | Tested |
Reading
It is above 44% of the 9 earlier readings since 2016.
Recent history (24 observations)
| Period | Value |
|---|---|
| 2025 | -7.1 pts |
| 2024 | -7.8 pts |
| 2023 | -7.9 pts |
| 2022 | -3.8 pts |
| 2021 | -11.5 pts |
| 2020 | -14.1 pts |
| 2019 | -5.8 pts |
| 2018 | -5.3 pts |
| 2017 | -4.8 pts |
| 2016 | -4.4 pts |
| 2015 | -3.5 pts |
| 2014 | -4.0 pts |
| 2013 | -4.6 pts |
| 2012 | -8.1 pts |
| 2011 | -9.7 pts |
| 2010 | -11.0 pts |
| 2009 | -13.2 pts |
| 2008 | -6.6 pts |
| 2007 | -2.9 pts |
| 2006 | -2.0 pts |
| 2005 | -3.1 pts |
| 2004 | -4.2 pts |
| 2003 | -4.8 pts |
| 2002 | -3.8 pts |
Source
Data: public market data feeds.
More from Agentic Investor
- Open United States Government Budget in the app: the interactive chart and a fresh forecast
- World and commodity series: every stored series
- Vietnam indicators
Also stored for United States
- United States 15 Year Mortgage Rate
- United States 30 Year Mortgage Rate
- United States ADP Employment Change
- United States API Crude Imports
- United States API Crude Oil Stock Change
- United States API Crude Runs
- United States API Cushing Number
- United States API Distillate Stocks
- United States API Gasoline Stocks
- United States API Heating Oil
- United States API Product Imports
- United States Asylum Applications