philipmulder11@gmail.com
I am an assistant professor of risk and insurance at the University of Wisconsin Madison. My research agenda studies how households, insurers, lenders, and governments adapt to dynamic climate risk.
Links to my papers are below.
Published Papers
We describe the initial impact of risk-based flood insurance pricing in the United States. More than two-thirds of National Flood Insurance Program policyholders are paying higher premiums under "Risk Rating 2.0" (RR2.0), and increases will continue in the coming years. Premium increases are largest for policyholders classified as low risk on the Federal Emergency Management Agency's (FEMA's) flood maps, and the average premium gap between those in and out of the FEMA-mapped floodplain is narrowing. FEMA's flood maps, which fail to account for the new risk information, are still relied on by homeowners and regulators, blunting the adaptation benefits of RR2.0.
Working Papers
Most homeowners do not have enough insurance coverage to rebuild their house after a total loss. Using contract-level data from 24 homeowners insurance companies in Colorado, we show wide differences in average underinsurance across insurers that persist conditional on policyholder characteristics. Underinsurance matters for disaster recovery. Across households that lost homes to a major wildfire, each 10 p.p. increase in underinsurance reduces the likelihood of filing a rebuilding permit within a year by 4 p.p.. To understand why consumers purchase underinsured policies, we build a discrete choice insurance demand model. The results suggest that policyholders treat insurers that write less coverage as if they set lower premiums, forgoing options to get more coverage at the same premium from other insurers – a pattern we call coverage neglect. Our findings suggest that coverage limits are either not salient to consumers or difficult to estimate without the input of insurance agents. Under a counterfactual without coverage neglect, consumer surplus increases by $290 per year, or 10% of annual premiums, on average.
I study the effect of risk-based flood insurance pricing on welfare and climate damages in the U.S. I estimate my results with flood insurance policy data and a novel survey measuring property-level flood insurance demand, risk perceptions, and objective flood risk. To identify the effects of risk-based premiums, I use variation created by outdated flood maps that caused high-risk homes to be misclassified as low-risk. My findings show that higher premiums are both a price and information signal. Owners of high-risk homes misclassified as low-risk underestimate their current and future flood risk, invest less in risk-reducing adaptation, and buy less flood insurance despite having lower premiums. Embedding these estimates in a sufficient statistics model with dynamic risk and endogenous risk beliefs and adaptation, I find that accurately classifying a new high-risk home increases welfare by $25,000 and reduces its dynamic climate damages by 50%.
Global Banks and Natural Disasters (with Francisco Ilabaca and Robert Mann) R&R - Journal of International Economics
Draft (2024)
Natural disasters can generate large economic losses and disruptions for global financial institutions, raising the concern that these disasters may increase financial systemic risk. We use detailed data on the foreign claims and liabilities of large U.S. regulated banks to study how multinational lenders reallocate capital following large natural disasters. We find little evidence that international banks increase lending to countries after destructive disasters, which should increase the demand for funds. Instead, difference-in-differences estimates suggest that natural disasters lower cross-border lending to affected countries by 9% two years after large natural disasters. We hypothesize that damaging natural disasters exacerbate cross-border information frictions. To test this mechanism, we exploit within-country heterogeneity in monitoring costs between banks. Consistent with this mechanism, our results show that declines in aggregate lending are driven by banks with weaker economic connections to the affected countries. These findings suggest that information frictions both dampen the transmission of natural disasters and reduce the reallocation of capital through the international financial system.
Property Insurance and Disaster Risk: New Evidence from Mortgage Escrow Data (with Ben Keys)
Draft (2024)
We develop a new dataset to study homeowners insurance. Our data on over 47 million observations of households’ property insurance expenditures from 2014-2023 are inferred from mortgage escrow payments. First, we find a sharp 33% increase in average premiums from 2020 to 2023 (13% in real terms) that is highly uneven across geographies. This growth is associated with a stronger relationship between premiums and local disaster risk: A one standard-deviation increase in disaster risk is associated with $500 higher premiums in 2023, up from $300 in 2018. Second, using the rapid rise in reinsurance prices as a natural experiment, we show that the increase in the risk-to-premium gradient was largely caused by the pass-through of reinsurance costs. Third, we project that if the reinsurance shock persists, growing disaster risk will lead climate-exposed households to face $700 higher annual premiums by 2053. Our results highlight that prices in global reinsurance markets pass through to household budgets, and will ultimately drive the cost of rising climate risk.
Does Good News Wash out the Bad? Belief Updating and Climate Risk Information (with Lynn Conell-Price)
Draft (2024)
While property owners around the world have access to an increasing array of information about climate risks, it is unclear how such information affects their beliefs and participation in disaster insurance markets. This paper shows results from the first field experiment providing detailed, property-level flood risk information to U.S. homeowners. Our experimental design compares changes in risk beliefs and implied insurance demand between a treatment group who received individualized flood risk information from the First Street Foundation Flood Model, and control arms who received no individualized information. Our results show that while homeowners who overestimated their risk relative to the model revised their beliefs downward, homeowners who underestimated their risk showed little updating. Although "underestimators" are more confident about their pre-treatment risk beliefs, we do not find evidence that these results are driven by private information or adaptation. The belief updating we find implies economically significant changes in flood insurance demand, reducing predicted voluntary take-up by more than five percentage points among homeowners at lower risk and increasing adverse selection.
Money to Burn: Crowdfunding Wildfire Recovery (with Tony Cookson & Emily Gallagher) R&R - Journal of Finance
Draft (2023)
Crowdfunding is an increasingly popular way to raise emergency funding after disasters. However, for victims of a recent major Colorado wildfire, we find that crowdfunding raised more support for high income and prime credit beneficiaries rather than helping the most vulnerable. Specifically, beneficiaries with income above $100,000 receive 40% more social network support on GoFundMe than beneficiaries with income below $50,000. High income households are also 13 percentage points more likely to have a crowdfunding campaign at all. The regressive nature of disaster crowdfunding is not due to less severe losses by financially vulnerable consumers, nor is it fully explained by the degree of online campaign sharing. Our findings highlight substantial disparities in social network insurance, which, as we show, likely exacerbate income inequalities in the recovery process.
Neglected No More: Housing Markets, Mortgage Lending, and Sea Level Rise (with Ben Keys) R&R - Review of Financial Studies
Latest Draft (2022)
NBER Working Paper (2020)
Featured in NBER Digest (January 2021)
In this paper, we explore dynamic changes in the capitalization of sea level rise (SLR) risk in housing and mortgage markets. Our results suggest a disconnect in coastal Florida real estate: From 2013-2016, home sales volumes in the most-SLR-exposed communities declined 20% relative to less-SLR-exposed areas, even as their sale prices grew in lockstep. Only by 2019 did relative prices in these at-risk markets ultimately decline 5%. Over this period, home sellers accumulated an excess inventory of unsold properties as they maintained high list prices. Lender behavior cannot reconcile these patterns, as both all-cash and mortgage-financed purchases similarly contracted. We propose a demand-side explanation where previously neglected SLR risk became more salient in the home price expectations of prospective buyers than sellers. The lead-lag relationship between transaction volumes and prices in SLR-exposed markets is consistent with dynamics of prior real estate bubbles.
What's at Stake? Understanding the Role of Home Equity in Flood Insurance Demand (with Yanjun Liao) R&R - Management Science
Latest Draft (2022)
RFF Working Paper (2021)
Millions of homeowners are exposed to increasing financial risk from natural disasters. Yet, many households are uninsured against the costliest disaster: flooding. We show that low home equity depresses flood insurance take-up. To isolate the causal effect of home equity on flood insurance demand, we exploit price changes over the housing boom and bust. Insurance take-up follows house price dynamics closely, with a home price elasticity around 0.3. Mechanism tests are consistent with a debt over-hang channel, whereby mortgage default limits the flood risk exposure of uninsured households. As a result, households do not fully internalize their disaster risk.
The National Flood Insurance Program (NFIP) has been criticized for setting rates that inaccurately reflect flood risk, while on the other side of the market, studies have shown that homeowners typically hold inaccurate beliefs about their own risk. Despite cataloguing these sources of incomplete information, no research to date has studied asymmetric information in flood insurance. In this paper, I find evidence of adverse selection that emerges through dynamic changes in demand. I first show that experienced flood policy holders have higher claims conditional on rating factors relative to new purchasers. Initial buyers who drop their coverage have lower average costs than those that keep it. I show how this phenomenon can be understood as a form of dynamic adverse selection. To explore the possible sources and implications of dynamic adverse selection, I incorporate my empirical estimates into a model of insurance with learning to illustrate how much of flood insurance demand is driven by consumer-side risk type uncertainty. My findings suggest that 80% of initial take-up and two-thirds of consumer surplus stem from buyers’ uncertainty over their own risk relative to their knowledge after maintaining coverage for up to 15 years. In addition to their implications for flood insurance, my methods can be applied to understanding other markets that feature learning and dynamic demand processes.