The U.S. Municipal Bond Market Does Not Price In Physical Climate Risk … Yet
Despite the rapidly accelerating concern about climate risk disclosure in the financial sector, a rigorous statistical analysis of ~800,000 U.S. municipal bonds accounting for more than $2.5T in outstanding debt issued from 2006 to 2021 shows no evidence that event-based physical climate risk is being systematically “priced in to” the interest rate cost of debt for issuers. Confining the analysis to bonds issued from 2017 to 2021, an era of significantly heightened climate awareness in the financial markets, shows the same insight. And finally, under the premise that investors might perceive climate change to be only a material risk in the distant future, we limit the analysis to the universe of outstanding long-dated (20+ year maturity) bonds and find the same result. This stands in slight contrast to two recent studies that suggest investors may demand very modest premiums for bonds exposed to dramatic long-term sea level rise.
The lack of clear climate risk pricing is troubling when juxtaposed with the reality that event-based physical climate risk is correlated with discounts in both property value appreciation and population growth over the past decade — the pillars of municipal market tax revenue and stability. This creates a “frog in a pot of boiling water” situation, wherein systemic risk is significantly underestimated, and the heat will at least turn up gradually, and maybe abruptly. This is especially true in the current market, where the treasury rate is historically low and there is more investor demand for bonds than there is supply, thus compressing yield spreads.
The municipal debt ecosystem has the apparatus to finance many aspects of climate adaptation in the U.S. All of the market’s stakeholders ultimately have an incentive to do so. Our goal is to provide all of the key participants — including issuers — not only insight into climate risk to assets but also tools to evaluate the financial and social cost-benefit tradeoffs of making specific, local climate resilience investments.
The municipal market is now reckoning with climate change
Climate risk as a whole has only surfaced to the consciousness of the U.S. municipal debt market in the last couple of decades; it leapt forward as a concern in particular after the disastrous hurricane season of 2017 and when Camp Fire decimated Paradise, CA, in 2018. Less acutely and formally, climate risk has been a growing concern and point of discussion since as early as post-Katrina and even more so post-Sandy. (All of this is not to say that climate risk did not matter before then: the first documented — but actually not well known — climate related municipal market default was driven by the infamous Great Galveston (TX) hurricane of 1900.)
The growing catastrophe insurance gap — created, counterintuitively, by the same risk transfer sector that purports to absorb climate shocks — increasingly leaves the burden of property damage on the shoulders of individuals and communities, in turn putting downward pressure on property values and municipal tax streams at risk, especially in smaller communities that lack strong economies in the first place.
Climate risk tools are relatively new to the municipal market
risQ’s physical risk models and a handful of other data products are relatively new to the municipal market. Clients have only been using our data since early 2020. A quickly growing number of municipal debt asset managers, bond insurers, and rating agencies are now employing risQ data on an operational and strategic basis. Even so, the data and focus on managing catastrophic risk, which has hit a point of relative maturity in the risk transfer sector, is still in its very early years for this market. On top of that, given the supply constraint dynamics of the market (generally, there is more money to be invested than there is debt issuance to buy), it may not be surprising that in the next section we show that physical climate risk has not been and is not priced into the cost for issuers to borrow capital.
Physical climate risk is not priced into municipal bond yields
We analyzed the yield of ~800,000 bonds (CUSIP9s) issued from 2006 to 2021 covering ~$2.5T of the current ~$3.9T in outstanding debt in the market; this is an era (post-Katrina and Stern Review) where it is reasonable to assume climate risk was broadly recognized as a potential issue. We first used a random forest statistical model to estimate the expected yield of all the CUSIP9s based on a variety of controls as well as external variables — but without using climate risk as an input. The basic controls include: issuance year (as proxy for temporal market dynamics and interest rate), original principal balance, bond duration, obligor type (e.g., county, school district, hospital, dirt deal, state, etc.), and security type (e.g., general obligation or revenue bond). The model works well, explaining the majority (~84%) of the variance in yield in out-of-sample tests (i.e., with bonds the model has not yet seen). Fully aware that the background socioeconomic profiles of issuers are given consideration in credit ratings, credit desk analysis, and consequently yield spread (more on this topic later), we also input risQ’s Social Impact Scores into the model. In particular, we input all 7 of its sub-scores: the Persistent Health Obstacles Score, the Nonwhite/Minority Score (clearly tied to our second hypothesis, see the next section for more), the Housing Unaffordability Score, the At-Risk Employment Score, the Low Educational Attainment Score, the Low Affluence Score, and the Poverty Concentration Score. All range from 0–100, where higher scores indicate more need for resources related to their themes; as they increase, a reasonable hypothesis would state that so does credit risk as conventionally and historically framed (more on this topic in a near-term future article).
We then define a “Yield-to-Expectation Ratio” (YER) as the ratio of the actual original yield over the expected yield estimated from the statistical model. If climate risk has indeed been priced into the market, there would be a broad, positive correlation between our core measure of acute physical climate risk — the risQ Scores, which we have shown to correlate to adverse outcomes in the pillars underlying the municipal market, including property value, population change, and mortgage delinquency — and the YER. To test this in a robust statistical way, we perform a permutation test on the pairing of issuers’ YER and their risQ Scores. The permutation test works by assigning a new, randomly selected risQ Score to each issuer. This random assignment breaks any potential real correlation between the risQ Score and issuers’ YER. We repeat this random assignment 10,000 times, obtaining a Monte Carlo distribution of what the data would look like if there was no correlation between climate risk and residual yield. We then compare the original, real correlation with this “no correlation” distribution.
We find the same lack of correlation between the YER and the individual peril scores — the Wildfire risQ Score, Flood risQ Score, and the Hurricane risQ Score (Appendix Figures A1, A2, and A3, respectively).
Next, knowing that physical climate risk came to the forefront of the collective awareness of the market after 2017’s hurricane season, we repeat the exact same test but only for bonds issued between 2017 and 2021. The results in Figure 2 lend the same insight: we still observe no significant relationship between climate risk and YER.
Comparison to academic literature
In contrast, a recent peer reviewed paper by Marcus Painter did find some evidence that, only for bonds with longer-dated maturities, the underwriting cost of issuance and yield increase. Published in 2020, that paper is similar in its line of questioning to ours, but the data and methodology differ to an extent. That study focuses only on sea-level rise (versus the risQ Score, which considers wildfire, hurricane, and multiple types of flood, including coastal) and only examines county obligors. As such, it relies on what is ultimately a much smaller sample size. Whereas we use a random forest to allow for potentially complex, non-linear relationships between the variables we examine (e.g., the affluence and climate risk of communities are correlated) and yield, the 2020 study uses a simpler linear model. Finally, while there are many conceptual similarities between our analysis and Painter’s, the latter also relies on a slightly different set of independent variables and outcomes. For example, Painter uses credit ratings data directly, whereas our analysis incorporates Social Impact Scores as a proxy for what can drive credit ratings at least in part (crudely speaking, smaller communities with poorer populations have smaller tax bases and thus may often be treated as higher risk). Whereas Painter examines both yield and the total cost of issuance, we have only examined yield.
Another, even more recent paper by Goldsmith-Pinkham et al. (2021 — referred to as “GP et al.”), focused again strictly on sea-level rise, shows similar insights using a structural model of credit risk. In particular, the study estimates a (very modest) ~5 basis point price premium for every standard deviation in properties exposed to 6 feet of sea-level rise — overall, a more modest finding compared to Painter’s. Similar to Painter, GP et al. find a more significant effect when they narrow down to a sample of long-term bonds. The authors also make the case that Painter’s estimate may have been partially driven by the effects of the Great Recession since they only find significant results after 2012.
In total, this leaves one remaining key hypothesis for us to test that is centered around the long-term aspect: the idea that, after controlling for all else, bond investors have begun to demand a premium for the risk taken in holding debt that will exist long enough to feel more dramatic consequences of climate change. As a final test, we repeat our analysis for all bonds issued from 2006–2021, but only for those with a 20+ year lifetime at the year of issuance. Figure 3 shows the results; they do not corroborate the academic findings.
Most importantly, any potential debate around the subtler points of comparing this study with Painter’s and the work of GP et al. is the reality that even if there is a small pricing of dramatic, long-term sea-level rise, climate risk is still not priced appropriately given the existential challenge it presents.
Why isn’t climate risk priced into municipal bond yields (yet)?
Overall, the market is still in the early stages of reckoning with climate. There are at least several reasons that might partially explain why the municipal bond market has not priced in climate risk — yet. The answer is likely some weighted average of all of the below:
● “Muni bonds rarely default.” For multiple reasons, the municipal bond market is thought of as one that is low risk. Some of this is rooted in rules or conditions that protect lenders; e.g., many are state backed. Much of this is just rooted in data — compared to other asset classes, municipal bonds have indeed been historically less risky. Because of this, systemic risk in general (climate and otherwise) has not been nearly as central a concern to the world and culture of municipal bonds as it has been to insurance or mortgage-backed security markets.
● “Climate hasn’t historically caused defaults.” We are paraphrasing somewhat an argument that we hear less and less of as the climate crisis worsens. It’s also technically not true, as we pointed out earlier with respect to the 1990 Galveston hurricane. And climate risk is now the straw on the metaphorical camel’s back that is increasingly (and somewhat quietly) bankrupting small towns in the US. But in some ways as important as the reality of this argument is market perception of reality: since investors set the tone and ultimately drive yield expectations, if climate is expected to be a true credit risk, then yield spread should ultimately follow.
● There is more demand for municipal bonds than there is supply. Today, there is more capital for lenders to put into play than there is supply of municipal bond issuance. It is well known that this dynamic compresses yield spread in the market. While a large number of bond buyers are now operationalizing climate risk in their processes, the Miamis of the country don’t take a hit on credit spread yet.
● Issuers are not clearly incentivized or equipped to disclose climate risk on their own. Buy-side participants relying on issuers’ official statements will often be left in the dark for now. While more and more bond buyers are operating with objective climate risk data, a healthy number are still in this situation. Much is made of the need for issuers to disclose their climate risk — but realistically, for many without outside help, they lack the resources to do so, especially in a standardized, objective manner.
● Ratings agencies don’t bake climate risk explicitly into credit profiles of issuers … yet. But of course, in light of Moody’s recent acquisitions of 427 and RMS, for example, this is just a matter of time.
Why does it matter, and what should key stakeholders do about it?
Climate change has the potential to wreak havoc on the municipal bond market, gradually or abruptly, given its dependence on vastly underinsured property value. True reckoning with the reality of climate risk could occur through multiple potential trigger events, whether those events are actual climate events (analogous to Hurricane Andrew, which altered the very fundamentals and assumptions of P&C insurance in the early 1990s), policy adventions (per our partner at DeltaTerra Capital, the passage of FEMA’s Risk Rating 2.0 may upend residential flood insurance affordability and impact residential property prices sooner and more dramatically than is widely recognized), or sociological ones (migration in response to and/or even in anticipation of climate disasters).
To make matters potentially worse, the myriad protections that exist for bondholders in many cases could lead to a form of complacency that, in the long run, puts the entire system at risk. Consider an excerpt from a 2019 article that covers the case of Paradise in 2018:
“An estimated 80% to 90% of the town was ruined, including over 6,500 residences and 260 commercial buildings. Approximately 90% of the population was forced to leave.
In order to meet the debt service on its capital appreciation bonds, Paradise benefited from legislation that allowed the state to backfill payments due to lost property tax. Additionally, the state of California secured direct federal assistance to support affected communities such as Paradise. As a result, municipal bonds issued by Paradise were able to make all debt service payments as scheduled. This example of extreme damage highlights the many options municipalities may have to maintain financial solvency as they rebuild their communities.”
One perspective on this echoes the tone of the article, which is a positive one reflecting the ability of municipalities to stay afloat despite exposure to existential risks. Under this lens, in the short term for the issuer, people residing within the issuer’s boundaries, and for the issuer’s creditors, this is good news. But another perspective shows plainly how it creates an incentive to rebuild in high-risk areas, or at least a lack of clear incentives for the municipal market to adjust to reality.
● Bond buyers are currently not being compensated for a systemic and increasing credit risk, all while operating in a low interest rate environment. Smaller communities and tax bases are particularly vulnerable, meaning higher yield funds likely hold outsized risk. But most importantly, all bond buyers are reliant on the stability of the debt market; climate change is a systemic risk that, left unaddressed, will eventually actualize and threaten the stability of the entire market. Buyers arguably have more leverage than any other entity in the US: in theory, they are in a unique position to demand that issuers not only disclose their climate risk but also act and invest to mitigate it.
● Bond insurers hold the bag on 10–30 year bonds and have no means of transferring that risk through a means that is comparable to the property and casualty industry’s reinsurance and catastrophe bond tools. So while events that trigger bond insurance payouts are rarer than those that trigger residential or insurance policy payouts, climate risk will at the least create an uptick in losses over time, and at the worst, a Black Swan event if many issuers were to default “in sync” (e.g., think contemporaneous, massive wildfires and/or hurricanes that devastate small, often insured and already credit impaired issuers — all of which would be amplified in a recessionary environment) and/or if the often tenuous backstop of federal disaster aid were to dry up. Since less than 10% of municipal bonds are insured, bond insurers could actually consider climate risk as a significant opportunity for top-line growth as investors become increasingly wary. Meanwhile, insurers should consider the spatially-correlated nature of climate risks — e.g., if drought and wildfires continue to plague the American west, issuers could experience impairment, population loss risk, and property value declines en masse, whether or not they are the direct victims of events.
● Regulators hold a key to catalyzing a shift in the municipal bond ecosystem at scale. If climate risk disclosure becomes mandatory, this would accelerate the flywheel of climate adaptation. Reticent issuers would be forced to confront climate risks, while bond buyers and insurers would eventually be operating on a more level and transparent playing field. In light of the previous anecdote about Paradise, regulators may also want to explore what legislation is in place that may actually be inhibiting
● Bond issuers will need to prepare for potential “sticker shock” in many cases — yields don’t reflect climate risk yet, but this is almost certainly a matter of when, not if. If they are able to, issuers should not be reluctant to disclose climate risk; proactivity around the problem is actually widely perceived as a credit positive from a governance perspective. As mentioned earlier, we argue that not all issuers are equipped to disclose, let alone adapt to, climate risk without additional resources. risQ ultimately exists not to antagonize issuers but to help them, and we encourage issuers to engage with us in whatever capacity is suitable. We are in the midst of quantifying both the economic and social benefits of an array of proposed flood defense investments for a Boston-area issuer, in partnership with an infrastructure engineering firm. This project aims to be the first of its kind, seeking to weigh benefits relative to costs, as one of the key bottlenecks to local climate adaptation has often been “who’s going to pay for it and why?” The municipal market then provides a natural channel for executing the financing of climate resilience.
Why the Municipal Market Is So Important to Addressing Climate
The inevitability of just how broadly and dramatically climate risk will reshape the US is borne out by the data, the trajectory of climate change, and the science that links it all together.
The fact that the municipal market has not factored climate risk into municipal bond pricing and evaluations yet is not surprising, as the data to do so has not been available until very recently. There’s simply no way this can hold up in perpetuity.
Arguably more than any other sector, the municipal market has a rare opportunity to exert massive leverage on how well and fast the U.S. adapts to the climate crisis. The apparatus and incentives for financing climate resilience reside within the system. As individual municipal market participants, we all have unique, high-leverage roles in how the U.S. responds to climate change; this is simultaneously a great privilege and responsibility.
Given the increasing pace and time pressure of climate change as well as the political friction around the topic at the Federal level, the sooner climate risk moves from being an unpriced externality to a central, material concern to all municipal market participants, the better.