๐ A simple 1‑week‑prior forecast based on district vote share from the 2024 General Election
✅ Introduction
On June 3, 2025, South Korea held an unprecedented presidential election following the impeachment of the former president under martial law. Despite political turbulence and rapidly shifting polls, I found that a remarkably simple formula allowed for a surprisingly accurate prediction of the outcome. One week before the election, I shared my forecast with friends—and when it turned out to be nearly spot-on, they jokingly called it an “AI-level prediction.” This post, with help from ChatGPT, outlines the logic behind my estimate and compares it with the actual results. I also argue this model may remain useful for future elections.
๐ Why the 2024 General Election Was a “Prequel” to the Presidential Race
| Metric | Predicted | Actual (2025) |
|---|---|---|
| Registered Voters | 44,391,871 | 44,391,871 (source) |
| Turnout Rate | 75–78% | 79.38% |
| Votes for Lee Jae-myung | ~17,200,000 | 17,287,513 |
| Vote Share (Lee) | 49.7–51.6% | 49.42% |
➡️ Prediction of raw vote count was nearly exact; vote share decreased as turnout increased.
๐ฏ Prediction Formula
Lee Jae-myung's vote count was nearly fixed—around 17.2 million—based on:
- 2024 General Election Vote Share (50.56%)
- 2022 Presidential Election Turnout (77.1%)
- 2025 Registered Voters (44.39M)
The expected total votes = 44.39M × 77.1% = ~34.2M → 50.56% of that is ~17.2M votes.
The variable was not his supporters—but how many turned out for other candidates. Here’s how the share changes based on total turnout:
| Scenario | Total Votes Cast | Lee's Vote Share |
|---|---|---|
| 75% Turnout | 33,293,903 | 51.6% |
| 78% Turnout | 34,825,659 | 49.7% |
| Actual | 35,251,512 | 49.42% |
๐ Key Takeaways
- Lee’s raw vote count was predictable (within 1% error)
- His vote share decreased as total turnout increased
๐ Why This Matters for Future Elections
Korean elections show a consistent trend:
Local < General < Presidential in terms of voter turnout.
Thus, the previous General Election offers the most solid predictor base for forecasting the next Presidential election—unless disrupted by major scandals or crises.
๐️ Turnout History: 2002–2025
| Year | Election Type | Voter Population | Turnout |
|---|---|---|---|
| 2002 | Presidential | 34.99M | 70.8% |
| 2007 | Presidential | 37.30M | 63.0% |
| 2012 | Presidential | 40.51M | 75.8% |
| 2017 | Presidential | 42.48M | 77.2% |
| 2022 | Presidential | 44.20M | 77.1% |
| 2024 | General | 44.28M | 67.0% |
| 2025 | Presidential | 44.39M | 79.38% |
✍️ Final Thoughts
This simple prediction model proved highly effective.
By using general election data as a base and adjusting for turnout, one can reasonably estimate presidential election outcomes—barring unusual events.
I'll share more election prediction and AI analysis in future posts.
Feel free to leave a comment or email if you're curious!
๐ท️ Hashtags
#2025SouthKoreaElection #ElectionPrediction #LeeJaemyungVoteCount #GeneralElectionBaseModel #TurnoutAnalysis
๋๊ธ
๋๊ธ ์ฐ๊ธฐ