How to read it

Each card opens one model. The top of the page describes the setup in plain language, the figures show the game and its equilibrium, and the panel on the right reports the equilibrium, explains what your last change did, and holds the sliders. Some models have several parts, shown as tabs under the title.

• Learn about this model: when you open a model you can read a few short pages first, or go straight to it. The pages give the question the author set out to answer, the explanations the author argues against, the words you need for the sliders and figures, and the author’s answer. The link under each model’s title opens them again.
• Equilibrium is what rational players do given the rules and payoffs you set. Green means the model predicts a peaceful or cooperative outcome, amber a risk of conflict, red a breakdown.
• What changed and why names the value you moved, the outcomes that shifted, and the mechanism in the article that explains the shift.
• Plots marked “click or drag” let you move the point directly.
• Try this prompts load a setup that shows one of the paper’s main results. Historical illustrations are the authors’ own examples, cited to the page or chapter; the models are not fitted to them.
• All payoffs, costs and probabilities are abstract. Numbers that stand in for an ordinal ranking or a shape the author left unspecified are marked notional.
• The address bar stores the model and every parameter, so a link reproduces the exact setup.

Related tool

Fearon’s 1997 comparison of tying hands and sinking costs is already built as module B of the Deterrence Lab, so it is not repeated here. The Deterrence Lab’s audience-cost module is a two-type, one-round simplification; the audience-cost model in this gallery is Fearon’s 1994 war of attrition itself.

What we simplified, and what we left out

• Fearon 1995. Risk-neutral states only; the article also allows risk aversion. For Claim 2 we draw B’s cost from a uniform distribution on [0, c̄], a stylized choice that has the nondecreasing hazard rate Fearon assumes. In the preventive-war part, A’s later demand is capped at 1 when the formula would exceed it.
• Powell 2006. In the strategic-territory part only the jump in p matters; the straight lines drawn on either side of it are notional. The guns-versus-butter mechanism (pp. 192-194) is described in the article but not built here. Powell (p. 188) notes the territory result depends on fixed war costs; if war destroys a fraction d of the flow, the jump must be at least d/(1 − d) to cause war.
• Brinkmanship. We could not open Powell’s 1990 book or his 2015 article, in which states choose the level of risk directly, so neither is implemented. The model here is Powell 1987, where each escalation adds a fixed increment of risk. Payoffs are rescaled so that prevailing = 1 and disaster = 0 for each state, which leaves every equilibrium unchanged. When both states can make the same number of bids, the tool resolves the tie by backward induction (I makes the last bid and prevails); Proposition 1 states only the strict cases.
• Fearon 1994. Linear audience costs and uniform priors on each state’s value for war; the article’s propositions hold for general distributions. The article reports each side’s odds of backing down; the stacked bar of who backs down first is our calculation from the equilibrium distributions.
• Jervis 1978. The article ranks outcomes but gives no numbers, so the payoffs are notional. The expected-value comparison formalizes his statement on p. 179; the “wait and see” switch formalizes p. 172; the repeated-game threshold is the textbook grim-trigger result, not Jervis’s. The four worlds are shown as he describes them, without numbers.
• Kydd 2000. Only the separating (reassurance) equilibrium is computed. Kydd’s four pooling equilibria are characterized in a formal supplement we did not obtain, so the tool says when the separating equilibrium fails rather than naming which pooling equilibrium takes over. We did not use his 2005 book.
• Slantchev 2003. The tool computes the separating equilibrium of Proposition 2 and checks the appendix’s conditions for it, including the one Slantchev checks numerically. When they fail, it does not compute the semi-separating or pooling equilibria the paper mentions but does not characterize.

Sources

Every source below except Fearon 1997, which the Deterrence Lab covers, and Schelling 1966, cited through Powell 1987, was opened and read for this tool; results are cited to the printed page. No outside data are used.