The 8% Chance that the Iranian Regime will Fall

Four forecasting traditions, base rates, a structural risk model, prediction markets and a Bayesian inside view, each built into a working model and pointed at Iran.

Four independent forecasting traditions, each implemented as a working model and pointed at Iran, are averaged here. They converge near 8 percent for the probability that the Islamic Republic falls within the year, a number reported with a long accounting of why it should not be trusted.

What the confident predictions said

For three years, prominent voices have forecast the fall of the Islamic Republic in confident terms, most implying odds far above the single digits. Reza Pahlavi has been a prominent representative of the Iranian diaspora's movement to encourage regime change.

  • Reza Pahlavi (Spring 2026). Declared that “the Islamic Republic has reached its end and is in the process of collapsing” and that the unrest was “irreversible,” while presenting himself as ready to lead a transition. source
  • Benjamin Netanyahu (February 2026). Told the White House the Islamic Republic was “ready to fall” on the eve of the strikes; weeks later, with the regime still standing, he allowed only that its collapse was “possible” but “not guaranteed.” source
  • Donald Trump (February 2026). As the bombing began, told Iranians to “take over your government. It will be yours to take. This will probably be your only chance for generations.” source
  • Prediction markets (January 2026). During nationwide protests the traded odds that the regime would fall climbed to roughly 33 percent before sliding back into the low teens, the crowd’s own high-water mark. source
  • Foreign Affairs (2026). Ran the line “The Iranian Regime Could Fall,” part of a wave of commentary calling this the gravest threat to the Islamic Republic since 1979. source

The four methods

Outside view (base rate), 2%

Reference-class forecasting, the discipline Philip Tetlock's research champions: ignore the dramatic specifics and ask how often regimes like this one fall in an ordinary year. Authoritarian regimes break down at roughly 5 to 7 percent a year on average. Revolutionary regimes, which Steven Levitsky and Lucan Way show to be the most durable category of all, fall at perhaps a fifth to a third of that rate.

We take a 6 percent annual baseline for authoritarian breakdown and apply the revolutionary-durability discount. 6% x (1/3) = 2%.

Where it is weakest: It deliberately ignores everything specific about this moment, including a lost war and a dead Supreme Leader, on the theory that the specifics usually matter less than forecasters think. Sometimes they do.

Structural risk model, 13%

The Political Instability Task Force lineage built for the US government by Jack Goldstone and colleagues: a logistic model scores a country on a few structural risk factors and sorts it into a risk tier. Even the highest tier rarely carries an annual probability above 15 to 20 percent, because the underlying base rate of instability onset is only about 2 percent of country-years.

We start from that 2 percent country-year base rate, place Iran in the top risk quintile (which historically captures most instability onsets), and raise it for the compound stress of war, economic crisis, and protest. 2% x 4.3 (top-quintile concentration) x 1.5 (compound stress) = 13%.

Where it is weakest: The tier placement and the stress multiplier are judgment calls, and the model's own designers note that even severe flags pass 30 percent only in rare outliers.

Prediction markets, 10%

Skip the models and read the price. Prediction markets and forecasting platforms aggregate the money and judgment of many people into a single number, and on political questions they often match or beat the formal methods. This is a forecasting method in its own right, not a footnote.

We average the implied probability of regime collapse within the year across three live markets, after Ali Khamenei's death and the orderly succession of his son Mojtaba had already been priced in. (Polymarket 14% + Manifold 10% + Metaculus 5%) / 3 = 10%.

Where it is weakest: Thin markets, shared information, and headline-driven swings. The same Polymarket contract traded near 33 percent during the January 2026 protests before falling back.

Bayesian inside view, 7%

Start from the base rate as a prior, then update on the specific evidence under Bayes' rule, raising the odds for conditions that signal collapse and lowering them for conditions that signal endurance.

Prior of 2 percent. We raise the odds for the lost war, the currency collapse, and the sustained protests, and lower them for the one signal that has not appeared: a fracture in the security forces. The coercive apparatus held, and the succession to Mojtaba Khamenei went through in an orderly way. Prior odds 0.020, multiplied by 3.0 (war) x 1.5 (economy) x 2.0 (protest) x 0.4 (intact coercion) = 3.6. Posterior odds 0.073, a probability of 7%.

Where it is weakest: Every multiplier is a defensible guess, and a forecaster more impressed by the protests than by the security forces' cohesion could move this number a long way.

The number they produce

Averaging the four methods gives 8 percent. The methods are not truly independent, they lean on the same scarce historical cases, so their rough agreement is less reassuring than it looks.

Why even this should not be trusted

  • The sample is tiny. There are perhaps two or three dozen well-documented regime collapses in the modern record. Every model on this page leans on that handful, which is far too few to support a confident probability.
  • The inputs are soft. Military cohesion, elite loyalty, and the mood of a population are not measured the way GDP is. They are estimated, often from the outside, and small errors compound through the model.
  • The methods are not really independent. All four lean on the same scarce historical cases and the same observable conditions, so their rough agreement is less reassuring than it looks. It is four views of the same thin data, not four witnesses.
  • Rare events resist prediction. Collapse is non-linear and often turns on a single contingent moment, a defection, an assassination, a crowd that refuses to disperse, that no model can anticipate.
  • A forecast can change what it measures. A widely shared probability of collapse can itself alter the behavior of the people whose choices decide the outcome, a feedback loop no static model captures.

Live prediction-market readings

  • Polymarket: Iranian regime falls before 2027, 14%. ~$19M traded; strict resolution (core institutions dissolved)
  • Manifold: Iran's regime falls in 2026, 10%. ~25,000 trades
  • Metaculus: end of the Islamic Republic, 5%. stricter resolution bar; conservative forecaster base

Sources

  • Levitsky and Way, “Revolution and Dictatorship,” Princeton University Press, 2022 link
  • Geddes, Wright and Frantz, “Autocratic Breakdown and Regime Transitions,” Perspectives on Politics, 2014 link
  • Goldstone et al., “A Global Model for Forecasting Political Instability,” AJPS, 2010 link
  • Kendall-Taylor and Frantz, “How Autocracies Fall,” The Washington Quarterly, 2014 link
  • Polymarket: Will the Iranian regime fall before 2027? link
  • Manifold: Will Iran's regime fall in 2026? link

Open this deep dive in the interactive site →