Chapter 36 — Further Reading

Sources are grouped by the book's three citation tiers (see the bibliography note). For a forward-looking chapter, much of the most useful reading is recent industry and regulatory material rather than settled canon — treated here with appropriate caution about specifics.

Tier 1 — Verified canonical (statutes, frameworks, named bodies and events)

  • The National Association of Insurance Commissioners (NAIC) — model guidance and bulletins on the use of artificial intelligence and big data by insurers, and the work of its committees on AI, big data, and catastrophe risk. The authoritative U.S. starting point for how regulators are approaching AI in underwriting.
  • Colorado SB21-169 — the Colorado statute requiring insurers to test their use of external consumer data and predictive models for unfairly discriminatory outcomes; the leading example of a state moving to regulate algorithmic fairness in insurance. (Owned conceptually by Chapter 35; named here as real.)
  • The National Flood Insurance Program (NFIP) — the federal flood program (Chapter 15), the largest living example of a public response to a private-market insurability gap, and a case study in both the promise and the peril of subsidized catastrophe coverage.
  • FAIR plans and state residual markets — the state insurers of last resort (and state-created catastrophe entities) that catch property risk the standard market won't write; their swelling rolls in stressed states are public record.
  • Hurricane Andrew (1992) and Hurricane Katrina (2005) — the catastrophe events (Chapters 6, 30) that reshaped catastrophe modeling, reinsurance, and the industry's understanding of correlated loss; the baseline against which the climate trend of this chapter is measured.
  • The Fair Credit Reporting Act (FCRA), the McCarran-Ferguson Act, GINA — the data-use, state-regulation, and genetic-information frameworks (Chapters 4, 8, 17, 35) that govern what the AI of 2035 may and may not use; named as real, with specifics deferred to their owning chapters.

Tier 2 — Attributed, specifics unverified (industry practice, reputable reporting)

  • Reinsurer and reinsurance-broker catastrophe and climate reporting — the annual catastrophe reviews, climate-risk papers, and renewal commentaries published by the major reinsurers and reinsurance intermediaries. Excellent for the direction of catastrophe trends and reinsurance pricing; treat any single figure as point-in-time, not canonical.
  • Consulting and actuarial-body material on AI in underwriting — reports from the major consulting firms and the actuarial professional bodies on predictive modeling, machine learning, large language models, and governance in insurance. Useful for the co-pilot/automation division of labor (§36.2); attribute honestly and avoid treating projections as facts.
  • Trade-press coverage of the California and Florida homeowners markets — reputable insurance trade journals and mainstream business reporting documenting carrier withdrawals, non-renewals, residual-market growth, and regulatory reform efforts. The qualitative narrative of Case Study 1 rests on this body of public reporting; no precise statistic is asserted.
  • Parametric and catastrophe-bond market commentary — industry and ILS (insurance-linked securities) market reporting on parametric structures and cat bonds for catastrophe and climate risk (Case Study 2). Good for the shape and growth of the product; specifics are market-dependent.
  • InsurTech and embedded-insurance industry analysis — reporting and research on digital MGAs, embedded distribution, and on-demand products (Chapter 34, §36.5), including the public stumbles that counsel humility about forecasts.

Tier 3 — Illustrative / constructed (this book's teaching material)

  • Harbor Steel & Fabrication, Inc. — the running Underwriting File; the 2035 forward look in this chapter and all of its figures are constructed teaching examples (Chapter 1 and the continuity ledger).
  • Figure 36.1 ("The risk that is modeled but not writable") and Figure 36.2 ("When the trigger and the loss disagree") — constructed teaching submissions; all numbers (the doubled AAL, the \$500,000 trigger, the 25-mile threshold, the scenario losses) are illustrative.
  • The constructed scenarios throughout §36.1–§36.5 (the sensored plant, the on-demand drone, the embedded shipping product, the coastal renewal) — all illustrative, built from real industry patterns but not drawn from any real account.

If you read only one thing: start with the NAIC's model guidance on the use of artificial intelligence by insurers. It is the single document that most concretely shows where the two forces of this chapter meet — the rise of AI in underwriting and the regulatory insistence that the human stays accountable for a fair decision — which is the whole argument of the chapter in one official source.