Live Data / OSHA + State License Databases

How the grade is built

Every Complifield grade is the output of four weighted signals drawn from federal OSHA enforcement records. No reviews. No self-reporting. No stars.

Grade = (Severity × 40%) + (Frequency × 30%) + (Response × 15%) + (Recency × 15%)

40% Severity Highest signal precision

30% Frequency Normalized per inspection

15% Response Time Revealed time preference

15% Recency Temporal discount

Grade Scale

A 90–100 Investment Grade Hire with confidence

B 70–89 Investment Grade Minor issues — ask questions

C 50–69 Speculative Proceed carefully

D 30–49 Distressed Significant concerns

F 0–29 Default Failed safety report

⚠ Score Overrides

Willful violation cap: Any willful citation locks the grade at C or below, regardless of other factor scores. Cannot be overridden.
Fatality cap: A fatality inspection within 36 months locks the grade at D or below. Applies even if all other signals are clean.

Economic & Game-Theoretic Foundations

The frameworks behind the grade

Each factor in the Complifield model is grounded in a named economic framework. These are not proprietary black-box weights — they map directly to peer-reviewed theory.

Framework 01 Expected Value with Temporal Discounting Intertemporal Choice Theory

Fama · Hansen · Shiller / 2013

The grade engine asks: how much should a past violation matter today? Temporal discounting answers this formally — past events decay exponentially with distance in time. A citation from 1983 carries fundamentally different information about current behavior than one from last month.

V(t) = V₀ · e^(−δt)

V(t) = present value of violation
V₀ = base severity score | δ = decay rate | t = age in years

Violation Age	Recency Score	Economic Meaning
0–1 year	0–60	Full present value
1–3 years	40–60	Partial discount
3–7 years	60–80	Moderate decay
7–15 years	75–90	Heavy discount
15+ years	~92	Near-zero weight

Framework 02 Bayesian Signal Aggregation Frequentist Approximation · Bayesian Inference

The grade is not a single measurement — it is four independent signals combined. Each signal has a different reliability (signal-to-noise ratio). The weighted linear model is a frequentist approximation that produces results consistent with a full Bayesian posterior under mild assumptions of signal independence.

Grade = Σ(wᵢ · sᵢ) where Σwᵢ = 1.0

wᵢ = weight of factor i (signal precision)
sᵢ = score of factor i (likelihood update)

Factor	Weight	Signal Precision
Severity	40%	Highest — OSHA-assigned, not self-reported
Frequency	30%	High — normalized rate controls for firm size
Response Time	15%	Moderate — influenced by legal contestation timing
Recency	15%	Lower for old data — high noise as predictor of current behavior

Framework 03 Lemons Markets & Information Asymmetry Information Economics

Akerlof · Spence · Stiglitz / 2001

George Akerlof's 1970 paper demonstrated that markets with asymmetric information can fail entirely. The contractor hiring market exhibits precisely this structure: the contractor knows their OSHA record; the homeowner cannot read it.

Akerlof Model	Contractor Market
Seller knows car quality	Contractor knows their OSHA record
Buyer cannot observe quality	Homeowner cannot read OSHA database
Lemons crowd out good cars	Unsafe contractors win bids on price
Market unravels	Homeowners rely on star ratings
Warranty = credible signal	Complifield grade = federal data, third-party

Framework 04 Reputation and Repeated Games Game Theory · Folk Theorem

Nash · Harsanyi · Selten / 1994

The response time factor — how quickly a contractor corrected each violation — draws directly from repeated game theory. Fast abatement reveals a high-δ agent who values their future reputation stream. Slow abatement reveals short-term extraction over long-term reputation.

V(cooperate) > V(defect) when δ > δ*

δ = patience / discount factor
δ* = threshold below which defection dominates
Cooperation = fast abatement, safety culture

Abatement Speed	Score	Revealed Preference
Same day / 1–3 days	95–100	High δ — reputation investment
1–2 weeks	80–95	Moderate δ — adequate
2 weeks–1 month	60–80	Lower δ — compliance-driven
1–3 months	30–60	Low δ — dragging feet
3+ months / open	0–30	Very low δ — defection

Framework 05 Risk-Adjusted Return & Structural Credit Models Merton Structural Credit Model · 1974

Fama · Hansen · Shiller / 2013

The A–F grade is structurally identical to a bond credit rating. Moody's, S&P, and Fitch compress complex default risk signals into letter grades using weighted factor models — the same architecture as the Complifield engine. "Default" in this model means a serious OSHA-cited incident that shuts down a job or triggers litigation.

DD = (V_A − D) / (σ_A · V_A)

DD = Distance to Default
V_A = asset value (safety culture)
D = compliance threshold | σ_A = asset volatility

Credit Rating	Grade	Score	Override
Investment Grade AAA–BBB	A / B	70–100	—
Speculative BB–B	C	50–69	—
Distressed CCC–CC	D	30–49	—
Default	F	0–29	—
Covenant breach	Willful cap	Max C	Hard lock
Technical default	Fatality cap	Max D	Hard lock

The Academic Foundation

"Complifield is a credible signaling mechanism that resolves the lemons problem in residential and commercial contractor markets by converting publicly available but practically inaccessible OSHA enforcement data into a standardized, temporally-discounted risk grade — collapsing information asymmetry and enabling efficient market pricing of contractor safety quality."

2001

Nobel Prize in Economics

Akerlof · Spence · Stiglitz

Asymmetric Information

1994

Nobel Prize in Economics

Nash · Harsanyi · Selten

Game Theory

2013

Nobel Prize in Economics

Fama · Hansen · Shiller

Asset Pricing

See how your contractor grades →

Full Framework Document

Download the economics PDF

The complete five-framework document with worked examples, formulas, and the Nobel Prize connection. Enter your email and we'll send it directly.