Spence's Signaling Theory: Why Credentials and Titles Persist Even When Everyone Knows They're Useless

60-Second Summary

Signaling Theory (Spence, 1973): when the buyer can't verify quality directly, sellers send costly signals that are cheaper for high-quality sellers to produce than low-quality ones.
Applied to hiring: a Stanford CS degree, a Google tenure, or a 'Staff Engineer' title are signals — not because they prove skill, but because they were expensive-enough to fake that only certain candidates bothered.
Signals persist even when everyone knows they're imperfect — because they still beat the alternative of no information at all.
Companies unintentionally create signal races: title inflation, degree filters, and 'must have worked at a top-tier company' become self-reinforcing selection mechanisms.
You have two levers: reduce information asymmetry (structured hiring, work samples) so signals matter less, or accept signals and pay the premium consciously.

Every experienced hiring manager has said, 'a degree doesn't prove anything' — and then, at 5pm on Friday with 400 resumes to screen, filtered to 'Bachelor's from a top-50 school'. That is not hypocrisy. That is signaling theory operating exactly as Michael Spence predicted in 1973.

What Spence actually proved

“Education may be a productive investment; but even if it is not, it may still be a rational investment for the individual as a signal of pre-existing ability.”
— A. Michael Spence, Job Market Signaling, Quarterly Journal of Economics (1973)

Spence's 1973 model asked: imagine an employer who cannot observe a job candidate's productivity in advance. Two candidate types exist — high-ability and low-ability. The employer would pay more for high-ability, but can't tell them apart. Now suppose education is costly, and — critically — costlier for low-ability people (they find it harder). Then high-ability people will get more education even if education teaches them nothing, purely to distinguish themselves from low-ability applicants. The employer, seeing the degree, rationally infers higher ability and pays more. Both sides act rationally; the outcome is a signaling equilibrium where a lot of money is spent on credentials that don't necessarily add productive value.

Spence shared the 2001 Nobel Prize in Economics with George Akerlof (the lemons problem) and Joseph Stiglitz (screening) for showing how information asymmetry reshapes markets. Job markets are the canonical example.

Four kinds of signal in HR

Where signals show up in hiring and promotion

1
Credential signals
University brand, degree level, certifications. Cheapest to check, easiest to over-index on. Google formally dropped degree requirements in 2020 — and still ~90% of hires have degrees, because applicants self-select.
2
Employer signals
'Ex-Meta', 'ex-McKinsey', 'ex-YC founder'. These are strong signals precisely because those companies invested millions in filtering. You are outsourcing your screen.
3
Title & tenure signals
'Staff Engineer at Series B startup for 4 years' signals more than the title itself. Duration matters — short tenures are read as negative signals even when the reason is neutral.
4
Behavioral signals
Open-source commits, conference talks, published writing, side projects. Costly to produce, easy for high-ability people to sustain — a Spence-style separating signal.

$500B+

annual global spend on higher education

OECD Education at a Glance — much of it Spence's 'signaling premium'

38%

of US bachelor's-degree jobs could be done without one

Harvard Business School / Accenture, 'Dismissed by Degrees' (2017)

27%

wage premium for graduates of 'elite' universities

controlling for admission ability — Dale & Krueger (2014), NBER

70%+

of that premium disappears

when you compare graduates who were admitted to elite schools but attended non-elite ones — signal, not learning

The signaling race — how it self-reinforces

Why signaling gets worse over time

How it starts

A few candidates use a signal (e.g. an MBA) to stand out
Employers observe correlation with ability
Employers reward the signal
Signal is a real distinguisher — moderate cost, moderate payoff

How it ends

Everyone needs the signal to compete
Signal loses distinguishing power
A stronger signal appears (top-3 MBA, then Ivy, then FAANG tenure)
Total cost of signaling in the economy rises with no productivity gain

This is why credentialism inflates over time even when everyone claims to want to move past it. A 'BA required' role becomes a 'Master's preferred' role, then 'PhD or equivalent'. The signaling race has no equilibrium unless someone intervenes.

How to break the signal dependence

Anti-signaling hiring design

1
Structured work-sample tests
Replace credential inference with direct evidence. A well-designed 60–90 minute work sample is 3–4x more predictive of on-the-job performance than credentials (Schmidt & Hunter meta-analysis, 85 years of data).
2
Blind resume review for early stages
Remove name, school, and employer for the initial screen. Amazon, Deloitte, and the UK Civil Service have all run controlled experiments showing significant candidate-pool broadening with no drop in hire quality.
3
Skill-based rubrics with anchored scoring
Interviewers rate observable behaviors against a rubric, not against their gestalt impression of the candidate's pedigree.
4
Publish your ladder criteria
If 'Staff' means 'demonstrated ownership of a system used by 3+ teams for 12+ months', that's evidence. If it means 'senior engineer at a big-name company for 4 years', you've just imported someone else's signal race.
5
Decide consciously when to pay the signal premium
There are real cases — brand-sensitive customer roles, board-facing execs — where the signal is the product. Pay for it deliberately, not accidentally.

Title inflation as internal signaling

Spence explains internal titles too. A 'Senior Director of Strategic Growth Partnerships' is often a signal to peers and future employers, not a description of scope. Titles inflate for the same reason degrees do: they are cheap for the company to issue and valuable to the receiver in the external labor market. That's why titles at growth-stage startups drift up 1–2 levels versus the same job at a mature enterprise.

Every signal is measuring something — but not always what you think, and rarely what you actually need.

Signal	What it's really measuring	When it's useful	When it fails
Elite degree	Ability to compete for admission ~18 years ago	Fresh grad roles with no other signal	Anyone with 5+ years of work history
FAANG tenure	Passed FAANG's filter + survived their culture	Roles requiring FAANG-style scale	Small-team, high-autonomy startup roles (mismatch)
High title	Someone else's calibration decision	Comparable-scope roles	Cross-industry moves — 'Director' varies 3 levels
Open-source contributions	Sustained voluntary work in the craft	IC engineering, DevRel, tech leadership	Roles where the actual work is closed-source (most of them)

The honest version

If your hiring bar is 'ex-Google + Stanford CS', you're outsourcing your screening to Google and Stanford's admissions office — and paying a signaling premium for it. That's a fine choice as long as it's a choice.

FAQ

Frequently asked questions

So are credentials worthless?

Not worthless — Spence shows they carry real information. They're just not the information most people think. A degree signals persistence, sortability, and ability at 18. It does not signal on-the-job productivity 10 years later, and using it that way is expensive.

How do I convince my hiring managers?

Run the Schmidt & Hunter meta-analysis at them: structured interviews (0.51 predictive validity) and work samples (0.54) crush unstructured interviews (0.38) and years of education (0.10). The data is 100+ years deep and it's not close.

Doesn't this just mean we should all skip college?

No — the individual return on a signal can be positive even when the social return is close to zero. Spence is about aggregate market design, not personal decisions. HR sits in the aggregate-design seat.

What's the fastest thing I can do Monday morning?

Add one structured work sample to your highest-volume role and blind-review the first pass. That single change usually broadens your funnel 30–40% with no drop in hire quality.

Takeaways

Credentials, brands, and titles persist because they're cheap information, not because they're accurate.
Every signal is measuring something — usually not what you assume it measures.
You can either replace signals with direct evidence (work samples, structured rubrics) or pay the signal premium consciously.
Title inflation and credential creep are not moral failures — they are Spence equilibria that require deliberate design to escape.