Signaling Games

Overview

One party knows something the other cannot verify. Cheap talk fails — anyone can claim high quality. A signal works when its cost differs by type: only a capable worker can bear four years of demanding study; only a quality manufacturer can afford a lifetime warranty. The signal conveys truth exactly when it would be unbearable for the wrong type to fake it. Spence (1973) formalized this; Akerlof (1970) showed what happens when no signal exists (lemons).

Composes with prisoners-dilemma, repeated-games-reputation, and pricing-strategy.

When to Use

Apply when:

• One party cannot verify what the other knows — hiring, fundraising, M&A, B2B, branding, diplomacy
• A market is failing to price quality correctly (Akerlof lemons problem) and you need a fix
• You're designing a credential, certification, warranty, or brand investment
• You're trying to break into a market with a signal competitors can't cheaply imitate
• You need to detect real quality among multiple claimants
• You're reading geopolitical or policy moves as costly signals — chip export controls, sanctions, AI-capex commitments, or a rival's public capability demo read as signals of hidden capability or resolve

When NOT to use: information is symmetric; signal cost is the same for both types (expense, not signal); one-shot low-stakes interaction where direct verification is cheap.