Who Counts? Why Catastrophe Models See Buildings but Not People

Guest Author: Charlie Sidoti, Executive Director of InnSure.org

When a hurricane makes landfall, the catastrophe modeling industry springs into action. Within hours, firms like Moody’s RMS, Verisk, and CoreLogic produce loss estimates that move billions of dollars through insurance and reinsurance markets. These models are marvels of computational science—integrating atmospheric physics, structural engineering, and financial exposure data to estimate insured losses with remarkable precision.

But here’s what those models are actually measuring: damage to assets owned by people who bought insurance.

That’s not the same as measuring what the storm cost a community. It’s not even close.

If you’re a municipal planner or community leader trying to understand how climate risk threatens your jurisdiction, catastrophe models give you one piece of a puzzle and ask you to pretend it’s the whole picture. The piece they give you is important. But the pieces they leave out—the costs borne by renters, by uninsured and underinsured households, by the social fabric of communities that absorb repeated shocks without financial backstops—are where the real crisis is building.

What Catastrophe Models Were Built to Do

To understand the gap, you need to understand the purpose these models serve. Catastrophe models were designed to help insurers and reinsurers manage their portfolios. That’s their job. They estimate probable losses to insured assets so that carriers can set premiums, purchase reinsurance, and maintain solvency.

This is valuable work. Without it, insurance markets couldn’t function. But it means the entire analytical infrastructure—the data collection, the modeling assumptions, the output formats, the professional ecosystem—is oriented around a specific question: How much will this event cost the companies that insure physical assets?

The NAIC’s March 2025 Catastrophe Modeling Primer describes the standard components: hazard modules that simulate events, vulnerability modules that estimate damage to structures, and financial modules that translate damage into insured losses. At every stage, the unit of analysis is the asset—the building, the contents, the business income stream. The exposure database that feeds the model is a portfolio of insured properties.

If your property isn’t in the database—because you rent, because you couldn’t afford coverage, because your neighborhood is in the residual market—you don’t exist in the model. Your losses aren’t counted. Your costs don’t register.

The Missing Side of the Ledger

Here’s what this means in practice. When Hurricane Helene hit in 2024, the industry estimated roughly $50 billion in insured losses. But less than 25% of total economic losses had any form of prearranged financing. Private insurance covered 12-18%. NFIP covered 3-4%. FEMA disaster aid covered less than 1%.

Where did the rest land? On people and systems that catastrophe models don’t track.

It landed on renters—who make up about 35% of occupied housing nationally, but account for only 0.2% of all NFIP coverage. Approximately 41% of the nation’s rental stock—more than 18 million units—sits in counties that FEMA’s National Risk Index classifies as high-risk. Roughly 45% of renter households have no renters insurance at all.

It landed on low-income homeowners who are technically insured but functionally underinsured—households that can’t afford deductibles, that carry coverage limits well below replacement cost, that face policy exclusions for the specific perils threatening their homes. Research from First Street Foundation estimates that 39 million U.S. homes are insured at prices that don’t reflect their actual climate risk, representing a potential underinsurance gap of $28.7 billion annually.

And it landed on the broader community in ways that have no line item in any catastrophe model output: depleted municipal services, displaced small businesses, lost tax revenue, strained social services, reduced property values in neighborhoods where insurance retreated, and the slow erosion of community cohesion that follows repeated unrecovered shocks.

These aren’t speculative costs. They’re happening now, everywhere climate events hit communities where the protection gap runs deep. But they exist in a category that the modeling infrastructure wasn’t built to see.

The Language Problem

This is more than a data problem—it’s a language problem. The catastrophe modeling industry has given us a precise, standardized vocabulary for asset-centric losses. Average Annual Loss. Probable Maximum Loss. Occurrence Exceedance Probability. Return Period Loss. These terms are universally understood, consistently defined, and directly actionable for the professionals who use them.

For the social and distributional dimensions of disaster loss, we have no comparable vocabulary. The academic literature offers concepts—”well-being losses,” “socially contingent impacts,” “differential disaster burden”—but these terms aren’t standardized, aren’t operational, and aren’t available in formats that local decision-makers can use in real time.

A groundbreaking study published in Nature Sustainability modeled the distributional impact of a hypothetical earthquake in the San Francisco Bay Area. The finding was striking: poorer households suffered 19% of the asset losses but 41% of the well-being losses. The same event, the same geography—but radically different impacts depending on where you sit in the economic structure.

That’s an extraordinary insight. But it was published as a research paper. It doesn’t exist as a data feed. It doesn’t update when conditions change. It doesn’t connect to your planning GIS or your insurance market analysis. It isn’t available at the resolution of your jurisdiction.

A 2023 study in Communications Earth & Environment made the case explicitly: contemporary disaster risk assessments do not adequately account for equity, despite methods existing to do so. The tools are there. The integration isn’t.

This pattern repeats across the academic literature. World Bank researchers have shown that conventional asset-loss metrics systematically understate the impact of disasters on lower-income populations. The UN’s Global Assessment Report 2025 documented that disaster loss databases are structurally biased toward insured losses because insurers have comprehensive data about their clients while uninsured loss data remains patchy. A study in Natural Hazards found that government disaster cost models that ignore social vulnerability produce inaccurate impact estimates—because they assume disasters hit all populations equally.

The research exists. The translation to municipal planning tools doesn’t.

Who Benefits from the Blind Spot

It’s worth asking: who does the current system serve?

When catastrophe models focus exclusively on asset losses, decisions naturally optimize for asset owners. Capital flows toward protecting insured properties. Resilience investments prioritize infrastructure that serves assessed valuations. Insurance innovation targets profitable customer segments. Rate adequacy becomes the organizing principle—making sure premiums cover expected losses to the assets in the portfolio.

None of this is malicious. It’s the logical outcome of a system designed to answer one question—What will this cost insurers?—being treated as though it answers a different question: What will this cost our community?

The gap between those two questions is where the equity crisis lives.

Consider what happens when a community faces rising insurance costs and undertakes resilience planning. The engineering firm models flood mitigation benefits to structures. The insurance analysis examines premium impacts for insured properties. The economic study evaluates property value effects for homeowners.

Who evaluated the impact on the renter household paying 50% of income on housing, carrying no insurance, and one disaster away from displacement? Who quantified the cost to the small business owner without business interruption coverage whose customer base just scattered? Who modeled the cascading fiscal impact on the school district, the social services budget, the healthcare system?

In most communities, nobody. Because the tools weren’t built for those questions, the data doesn’t exist at the right resolution, and the professional ecosystem has no economic incentive to provide it.

What Municipal Leaders Can Do Now

The absence of standardized social burden metrics isn’t a reason for inaction—it’s a reason for leadership.

Start by refusing to accept catastrophe model outputs as the full picture. When your risk advisors present loss estimates, ask: Whose losses? What percentage of our community’s actual exposure does this represent? What about the uninsured, the underinsured, the renters? What types of social burden impacts are modeled and how are the calculated? These questions won’t have clean answers. That’s the point. The discomfort of not having answers is more honest than the false precision of numbers that only count some of your constituents.

Push for distributional analysis in your resilience planning. When evaluating mitigation investments, require your consultants to show who benefits across income levels and tenure types. A seawall that protects $500 million in assessed waterfront property is valuable. But if the community on the other side of the road—the renters, the workforce housing, the small businesses—gets no benefit, that’s a distributional choice you should be making consciously, not by default.

Demand that your Total Cost of Risk analysis include the social costs that fall outside insurance. Displacement costs. Lost productivity. Emergency services. Mental health burden. Tax base erosion in neighborhoods where insurance retreats. These costs are real. They hit your budget. The fact that they’re harder to model than building damage doesn’t make them less important—it makes current models less useful.

And connect with the emerging community of practice around these issues. The UN’s Global Assessment Report 2025 called for fundamentally rethinking how we measure disaster impact. Academic researchers are developing well-being-based metrics. Some communities are beginning to pilot integrated assessments that go beyond asset loss.

This work is early. The tools are imperfect. But the direction is clear: the era of measuring climate risk purely through the lens of insured asset losses is ending. The question is whether municipal leaders will drive that transition or wait for it to happen to them.

The Real Question

Catastrophe models are extraordinary tools that answer the questions they were designed to answer. The problem isn’t the models—it’s our willingness to pretend those are the only questions that matter.

A community’s total climate risk includes every resident, every renter, every uninsured household, every business that would close permanently after the next event. It includes the social costs that compound when vulnerable populations absorb repeated shocks without recovery resources. It includes the long-term fiscal consequences of pretending that uncounted costs don’t eventually land on municipal balance sheets.

Until we build the analytical infrastructure to see those costs—with the same precision, the same credibility, and the same operational utility that we bring to insured asset losses—our planning decisions will continue to optimize for the people who already have protection.

And that’s a choice. One that community leaders should be making with open eyes, not one that’s being made for them by the architecture of models built to serve someone else’s questions.

Sources:

Housing, Climate Risk, and Insurance — NBER Reporter

NAIC Catastrophe Modeling Primer, March 2025

Quantification of disaster impacts through household well-being losses — Nature Sustainability

The importance of accounting for equity in disaster risk models — Communications Earth & Environment

The Insurance Protection Gap — Sastry, Sen, Tenekedjieva & Scharlemann (SSRN)

Renters Vulnerable to Climate Disasters Amid Insurance Gaps — Harvard Joint Center for Housing Studies

The growing void in the U.S. homeowners insurance market — npj Climate Action

Global Assessment Report 2025 — UNDRR

The cost of social vulnerability in disaster management — Natural Hazards