Explainer · numbers from this site’s backtest, updated with each run
Does the economy predict elections?
Mostly. Predicting each election from a model fitted without it, the economy line called 16 of the 20 popular-vote winners from 1948 to 2024. Its typical miss was ±2.8 points of the two-party vote.
What “predict” means here
Each model is fitted on 19 elections and asked about the 20th, which it has never seen. That is repeated for every year. It is a much harder test than fitting all 20 and reading off how well the line runs through them, and it is the only accuracy figure this site publishes.
| Forecast | Winners called | Typical miss |
|---|---|---|
| Economy line (Fair + Bread and Peace) | 16 of 20 | ±2.8 pts |
| Fair alone | 14 of 20 | ±3.4 pts |
| Bread and Peace alone | 15 of 20 | ±4.7 pts |
| No model: the party in power gets its average share | 11 of 20 | ±5.2 pts |
The last row is the yardstick. Knowing nothing about the economy, you would miss by about ±5.2 points. The economy line cuts that to ±2.8.
Where it went wrong
The economy line picked the wrong popular-vote winner in 1960, 2016, 2020, 2024. Each page shows what the data said and what happened.
Why it can’t be certain
- Twenty elections is a small sample. A model can look good partly by luck, which is why only out-of-sample results are shown.
- Candidates, scandals and wars matter and aren’t in the data. Bread and Peace counts war deaths; nothing counts scandals.
- These are forecasts of the popular vote. In 2000 and 2016 the popular-vote winner lost the Electoral College.
- Data published before an election often gets revised later. Wherever an archived first release exists, the backtest uses it, so the test is fair to what was knowable at the time.
What it says about 2028
Today the economy line gives Republicans, the party in power, 50.3% of the two-party vote and a 55% chance of winning it. See the forecast and which past year today’s economy most resembles.