Skip to content

The Separating Equilibrium Test

Why It Matters

Most communication attempts fail this test. Cheap talk ("I'm trustworthy," "This is high quality," "I'm committed") is costless for all types to produce, so it carries no credible information. We waste enormous energy on uninformative signals and fail to recognize genuinely informative ones. The separating equilibrium test is a filter: does this action actually separate types, or is it just noise?

The Key Move

For any claimed signal of quality, credibility, or commitment, ask three questions:

  1. What does it cost the sender to produce?
  2. Would that cost differ if the sender were low-quality instead of high-quality?
  3. Is the cost difference large enough that low-quality types wouldn't want to mimic? If the answer to
  4. or
  5. is no, the signal conveys nothing.

Classic Application

Education as labor market signal. Spence's model: employers can't observe worker productivity directly, but education correlates with it. Education is costly (time, money, effort). Crucially, if high-productivity workers find education less burdensome than low-productivity workers (because they learn faster, for instance), then education can separate types. High-productivity workers signal through degrees; low-productivity workers find mimicry too expensive and don't acquire degrees. Employers use education as a credible signal despite it potentially teaching nothing job-relevant.

Surprising Application

Open-source contributions as hiring signals. Why do developers work for free on public projects? One answer: it's a separating signal. Producing quality open-source code is costly (time, skill). That cost is lower for genuinely skilled developers than for impostors (who would struggle to produce working code under public scrutiny). Employers treat open-source portfolios as more credible than resumes precisely because they're harder to fake. The "work for free" model exists partly because it solves a signaling problem.

Failure Modes

  • Ignoring audience beliefs: a separating equilibrium requires the receiver to interpret the signal correctly. If the audience doesn't know the signal's cost structure, it won't work.

  • Overlooking countersignals: sometimes high-quality types deliberately avoid costly signals to distinguish themselves from try-hards. Countersignaling requires even more common knowledge than signaling.

  • Assuming separation is efficient: just because a signal separates types doesn't mean the separation is socially valuable. Resources burned on signaling are often pure waste.

Go Deeper

  • Spence, "Job Market Signaling" (Quarterly Journal of Economics, 1973) - the foundational paper
  • Connelly et al., "Signaling Theory: A Review and Assessment" (Journal of Management, 2011) - comprehensive review across domains
  • Signaling - Information Economics: Reasoning Under Asymmetry and Uncertainty