OpsArtica
RESEARCH AT OPSARTICA
OpsArtica's research program audits the administrative datasets that shape occupational safety policy in the United States. The work is rule-based, reproducible, and publicly released as open code and open data. Active collaborations include the U.S. Census Bureau's Economic Statistical Methods Division and the BEACON industry-classification team.
I came to this work from the manufacturing floor, by way of the Navy and through the safety programs built to protect the people I worked with. Some of those programs were good. Many of them measured the wrong things, classified the wrong work, or missed injuries that never reached the books. The research below is what came out of taking those failures seriously enough to audit the federal datasets that drive the regulation.
LEAD STUDY
A rule-based validation pipeline applied to 1.46 million OSHA establishment-level injury records across four filing years, quantifying the scale, persistence, and economic consequences of industry-classification error in the federal injury-surveillance dataset.
53% of records trigger at least one detection gate; 382 establishments are chronic downshifters, repeatedly reporting under lower-risk NAICS codes across filing years. Named employers among the worst offenders include Dollar Tree (1,050 downward shifts), Tractor Supply (647), Costco (511), and BJ's (226).
44–65% of flagged records are material; they cross NAICS sectors or carry a DART injury-rate gap of more than 50%. Cross-sector misclassifications outnumber within-sector by 7:1. The flows systematically depress reported injury rates in high-risk sectors and inflate them in low-risk sectors.
27.7% of establishments with calculable premium gaps are underpaying relative to their true risk classification. The top distortion pattern moves establishments from Transportation into Professional Services, with rates of $0.21 per $100 payroll in place of $7.88, enabling adverse selection and cross-subsidization.
Unlike tax or financial audits, no federal mechanism currently validates employer-reported NAICS codes in the ITA system. BLS industry-level injury rates, the basis for OSHA enforcement targeting, are distorted by establishments filing under incorrect codes.
COMPANION PAPERS
The NAICS audit sits at the front of a wider research line. The four papers below extend the work into veterans' benefits adjudication, voluntary-compliance program effectiveness, and the pedagogy of plant-floor safety learning. State is named honestly: two are submission-ready, one is a published-data scaffold, and the deeper depth on each lives in the repositories and DOIs linked below.
Whether having an attorney, accredited agent, or VSO representative changes the odds of winning before the Board of Veterans' Appeals. Across 891,851 BVA decisions from 2010 to 2025, attorney representation is associated with 5.96-fold higher odds of a favorable outcome compared to going pro se, with inverse-probability weighting controlling for case complexity.
Which secondary-condition disability pairs the Board of Veterans' Appeals is granting, extracted from 1.18 million BVA decisions: 119,787 secondary service-connection pairs across 69,119 decisions. Of 12,121 granted secondary pairs examined, only 25.3% have supporting medical literature; the remaining 74.7% are granted without evidence-backed nexus, naming a downstream research opening on how representation status interacts with literature-light grants.
Whether OSHA's Voluntary Protection Program (VPP) produces better safety outcomes than inspection-based compliance. Matched difference-in-differences analysis across utilities and manufacturing, 2016 to 2022, finds VPP sites record 1.69 Total Recordable Injury Rate points lower per 100 full-time-equivalents than matched controls, roughly half the national baseline, and sustain that performance over four years. Sites driven by audit-only intervention show only short, statistically uncertain improvement.
A five-day arc of five-minute daily standups that revisit the same risk topic from different angles, applied at a southwestern-U.S. paint-manufacturing facility with 200 employees. The framework draws on spaced repetition, retrieval practice, question-driven facilitation, and near-hit anchoring. The headline result is durability: the program has sustained continuous implementation for 24 months without structural adjustment, alongside a practical AI-assisted workflow for generating facility-specific content at scale while preserving human review.
RESEARCH PROGRAM
Across the five papers, four themes recur. The administrative datasets that shape regulation are auditable, not just consumable. The systems that govern compensation, classification, and benefits make load-bearing decisions about workers' lives from data of uneven quality. Outcomes for those workers depend on whether the system's structure is examined from inside the structure. And practical interventions on the plant floor (pedagogy, voluntary programs, representation) measurably change those outcomes when designed with the science behind them.
Establishment-level audits of OSHA ITA, BLS QCEW, and Census County Business Patterns. Identifying classification, coverage, and reporting-quality gaps that affect downstream policy and enforcement targeting.
WCIRB rate-ladder analysis. Where misclassification creates premium-gap distortions, what the cross-subsidization patterns imply for correctly classified firms, and how adverse-selection pressures work through the underwriting market.
Administrative reporting quality as a proxy for organizational safety climate. Predictive surveillance, identifying high-risk establishments from reporting behavior before injuries occur, and connecting compliance pedagogy to retained behavior on the plant floor.
Decision-level analyses of Board of Veterans' Appeals data: how representation type changes outcome odds at scale, and which secondary service-connection pairs the Board grants with what evidence base. Aimed at the gap between system intent and system result for service-connected disability claims.
WHO IS DOING THE WORK
The work above is unusual in scope because the path to it was. The research is grounded in time spent inside the systems it audits, not above them. The arc below is the route.
The cross-discipline span is the work's load-bearing feature, not a biographical decoration. The papers reach across surveillance, classification, adjudication, and pedagogy because the path that produced them did.
FUNDING ALIGNMENT
The research program aligns with NIOSH strategic priorities in surveillance research and Total Worker Health, OSHA's stated interest in data-driven enforcement targeting, the ASSP Foundation's leading-indicators priorities, and workers' compensation research on adverse selection and rate adequacy. Datasets are public, methodology is reproducible, and the findings carry immediate translational potential for regulatory and insurance applications.
PRINCIPAL INVESTIGATOR
Principal investigator. Research Affiliate at Touro University. Patent-holder, Certified Safety Professional, and systems engineer with more than fifteen years of leadership across manufacturing, government, and defense.
Calvin's work blends organizational psychology, systems engineering, and rule-based data analysis. The NAICS misclassification audit grew from a research question about whether the federal injury-surveillance dataset can be trusted as a regulatory targeting tool. The answer carried it through a 17-stage pipeline, a 500-record validation study, Census Bureau engagement, and a $32.9B economic-impact estimate.