Reading Guides | AEM 4510

Baker, Bergstresser, Serafeim, and Wurgler (2018): U.S. Green Bonds

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

If Flammer asks whether green bonds affect what firms actually do, Baker, Bergstresser, Serafeim, and Wurgler ask how green bonds are priced and held in financial markets. That makes this the core asset-pricing paper in the green-instruments lecture and the natural complement to Flammer in our discussion of sustainable debt instruments.

The central issue is whether investors are willing to accept a marginally lower financial return in order to hold environmentally preferable assets. If they are, then green bonds can lower the financing cost of green projects, and private capital markets can support environmental investment through a channel that does not depend on regulation or subsidy.

The question

Do green bonds trade at a price premium relative to otherwise comparable ordinary bonds, and are they held by a distinctive investor clientele?

In yield language, a price premium is equivalent to a lower yield. If some investors derive utility directly from holding green assets, beyond the risk-and-return profile of the bond’s cash flows, they may accept slightly lower yields on green bonds than on otherwise similar conventional bonds. That wedge in yields, if it exists, is the empirical object of interest.

The market and the basic theory

The paper focuses on the U.S. green bond market, with particular attention to municipal green bonds. That focus is deliberate and useful. The municipal market is large, deep, and contains substantial within-issuer variation in bond characteristics, which makes it possible to compare green bonds to ordinary bonds issued by the same type of entity with similar maturity, tax status, and credit quality.

The authors organize the empirical analysis around a simple but powerful theoretical framework. Some investors care only about risk and expected return, as in a standard asset-pricing model. Other investors care about risk and expected return and about the environmental character of the assets they hold. The paper refers to this additional source of value as nonpecuniary utility from holding green assets.

That framework yields two sharp empirical predictions.

First, if green-preferring investors are marginal in the pricing of at least some green bonds, then those bonds should trade at higher prices and therefore lower yields than otherwise comparable ordinary bonds.

Second, green bonds should be held disproportionately by the investors who value the green label, producing measurably higher ownership concentration than is observed for ordinary bonds with similar risk characteristics.

What the authors do

The paper proceeds in three main steps. It first describes the structure of the U.S. green bond market: who issues green bonds, what kinds of projects those bonds finance, and how the market has grown over time.

It then studies pricing in both the primary market, where bonds are first placed with investors, and the secondary market, where they trade subsequently. The authors compare green municipal bonds to ordinary municipal bonds while conditioning on the factors that standard models of municipal yields account for: maturity, tax status, credit rating, issue size, issuer characteristics, and broad market conditions at the time of pricing.

Finally, the paper studies the ownership of these bonds. Using holdings data for insurance companies, mutual funds, and pension funds, the authors test whether green bonds are more concentrated in the portfolios of the kinds of investors who would plausibly value the environmental label, including specialized environmental funds and institutions with stated sustainability mandates.

How the pricing strategy works

The main empirical challenge is that green bonds are not randomly assigned. Issuers who choose to label a bond green may differ systematically from those who do not, and green bonds may differ in purpose, maturity, or risk from the ordinary bonds they are compared to.

The paper addresses this in two ways. First, it conditions on a rich set of bond- and issuer-level controls, including issuer fixed effects where the data allow, so that identification relies on within-issuer variation across bonds with different green status. Second, it makes use of an especially informative subset of cases in which the same issuer places a green bond and an ordinary bond at the same time, in a single bundled issue. Those bundled issues are particularly clean, because they hold the issuer and the market timing fixed by construction and isolate variation in the green label.

The research question is not whether the authors have a perfect experiment. The question is whether, after absorbing the standard determinants of municipal yields, a residual green premium remains, and whether that residual lines up with the predictions of the clientele theory.

How the ownership strategy works

The ownership analysis complements the pricing analysis in a theoretically disciplined way. If the green label truly matters to a subset of investors, green bonds should not only be cheaper for issuers. They should also be held disproportionately by the investors who value them.

The authors measure ownership concentration across institutional investors and test whether it is systematically higher for green bonds than for ordinary bonds with comparable characteristics. The theory predicts especially strong concentration for bonds that are small or nearly riskless, because those are the bonds for which a green-preferring investor can tilt their portfolio toward the green asset without taking on substantial undesired risk or absorbing a large share of the market themselves.

Main findings on pricing

The paper documents a green premium in U.S. municipal bonds. In the main large-sample regressions, after-tax yields at issuance are approximately 5 to 9 basis points lower for green municipal bonds than for otherwise comparable ordinary municipal bonds.

A 5 to 9 basis point wedge is not large in absolute terms, but it is not negligible. On a bond with a duration of roughly ten years, the paper notes that a 5 basis point yield reduction corresponds to approximately a 0.5 percentage point higher bond price. In other words, investors are willing to pay a measurable premium for the green label, and that premium translates into real financing-cost savings for issuers over the life of the bond.

The bundled-issue evidence is also informative. When the same issuer places a green bond and an ordinary bond simultaneously, there is little green premium at the issue date, but a modest premium emerges later in the secondary market. That timing pattern helps the authors separate the roles of initial marketing and placement from subsequent trading in the secondary market, and it suggests that investor preferences for green assets become more sharply reflected in prices as the bonds clear initial distribution.

Main findings on ownership

The ownership results are consistent with the clientele theory. Green bonds are more closely held than ordinary bonds, and the concentration is especially strong for small or nearly riskless green bonds. That pattern fits the prediction that green-preferring investors can most easily tilt their portfolios toward green assets when those assets do not expose them to large residual risk or require them to absorb a disproportionate share of a deep market.

The ownership evidence reinforces the pricing result in an important way. A price premium on its own could in principle reflect many things, including liquidity differences, selection, or measurement issues. A price premium combined with concentrated ownership by a particular investor clientele is the joint pattern the theory predicts, and it is much harder to generate from alternative explanations that do not involve investor preferences for green assets.

The paper also examines the role of certification. A subset of labeled green bonds are certified by third parties and registered with the Climate Bonds Initiative, which imposes standards on project eligibility and use of proceeds. Certification appears to sharpen the market meaning of the label, because it gives investors a stronger basis to believe that the bond is genuinely financing green projects rather than simply being marketed as green.

Why the paper matters

The paper matters because it provides a clean example of how investor preferences can affect asset prices in a setting with a large, liquid market and credible controls. In the standard textbook view, prices reflect only risk and expected cash flows. This paper documents that, in this market, a nonfinancial attribute of the asset also appears to carry pricing weight, consistent with a meaningful clientele of green-preferring investors.

The result has a direct environmental implication. If investors accept modestly lower yields on green bonds, then green projects can be financed at modestly lower cost. The magnitude is not large enough to substitute for climate policy, but it is a real channel through which private capital markets can lower the cost of capital for environmental investment.

The paper is also important because it carefully separates two questions that are often conflated in discussions of green finance. One question is whether investors value green assets and are willing to pay for them. The other question is whether green finance improves real environmental outcomes at the firm or project level. Baker et al. speak primarily to the first question, which is why the paper pairs naturally with Flammer, who focuses on real-side outcomes.

What to focus on when you read

On a first read, keep three ideas clear.

First, understand the theoretical framework. A green premium requires the existence of a clientele of investors who derive nonpecuniary utility from the environmental attribute of the bond, and who are marginal in pricing at least some of the bonds in the sample.

Second, focus on the two empirical predictions and how they interact. Lower yields and more concentrated ownership are the joint signature of a clientele effect, and it is the combination of the two that identifies the mechanism rather than either one alone.

Third, pay attention to certification and market structure. The green label carries more pricing and ownership weight when investors have a credible basis for believing it, and certification through programs such as the Climate Bonds Initiative is the main mechanism through which that credibility is established in the U.S. market.

Terms to know

Greenium: the lower yield, equivalently the higher price, commanded by a green bond relative to a comparable ordinary bond.
Nonpecuniary utility: value that investors receive from holding an asset that is not captured in the asset’s financial return, such as utility derived from the environmental attributes of the issue.
Ownership concentration: the extent to which an asset is held by a relatively narrow set of investors, often measured by the share held by the largest institutional holders.
Primary market: the market in which a bond is initially placed with investors at issuance.
Secondary market: the market in which an outstanding bond trades between investors after issuance.
Certification: third-party verification that a green bond meets a stated standard for project eligibility and use of proceeds, often provided through programs such as the Climate Bonds Initiative.

Bernstein, Gustafson, and Lewis (2019): Disaster on the Horizon

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper is the sea-level-rise analogue to the classic hedonic housing papers we study earlier in the semester. It asks whether a long-run climate risk that is not expected to cause major local damage for decades is already capitalized into current house prices.

That question connects directly to several themes of the course. It links discounting, beliefs, and asset pricing inside a single market that students already understand reasonably well, and it provides one of the cleanest pieces of evidence that forward-looking capitalization of climate risk is not limited to specialized financial instruments. It shows up in the ordinary residential real estate market.

The question

Do homes exposed to future sea level rise sell at a discount today relative to otherwise comparable unexposed homes in the same local market?

The paper is not a simple question about current damage. Many of the exposed properties in its sample are not expected to flood in any given year for decades, and some not for close to a century. The relevant question is therefore whether buyers are capitalizing a long-horizon climate risk into current sale prices, rather than whether they are reacting to realized flood events.

Why this is complicated

Coastal housing is a useful but difficult setting because the coast provides both an amenity and a risk. A house near the ocean may have a valuable view, beach access, and higher demand precisely because of its coastal location. At the same time, being low-lying and close to the water makes the house more exposed to long-run sea level rise and to the associated tail risks.

The central empirical challenge is therefore to separate the positive capitalization of coastal amenities from the negative capitalization of future flood risk. A simple comparison of inland homes to beachfront homes would confound these two margins and would not identify the risk discount of interest.

Data and setting

The authors use detailed transaction-level housing data together with geographic information on property locations, elevations, and projected sea level rise exposure. For each property, they determine whether the structure would be inundated under a defined sea level rise scenario drawn from standard projections, and they record a rich set of hedonic characteristics.

The sample is restricted to properties close to the coast, and the core comparisons are made within narrowly defined local markets, typically using zip-code or finer geographic fixed effects. That sample restriction is important. The identifying variation is not “coastal versus inland.” It is closer to “two properties in the same local coastal market, with similar observable features and comparable distance to the waterfront, but different projected inundation exposure because one sits at a higher elevation than the other.”

What the authors do

The main regression compares sale prices of exposed and unexposed homes while conditioning on a rich set of controls:

distance to the coast,
zip-code or finer local fixed effects,
time-of-sale fixed effects at the month level, and
hedonic property characteristics such as living area, age, number of bedrooms, and number of bathrooms.

In economic terms, the authors compare properties that are similar in their local market, in their sale timing, and in the hedonic attributes that determine baseline value, and that differ primarily in projected exposure to future inundation.

The paper then pushes the design in three useful ways. It examines the rental market as a contrast to the sale market. It studies how the estimated discount evolves over time. And it asks whether the discount varies with buyer sophistication and with local climate concern, both of which proxy for the information inputs that a forward-looking capitalization story predicts should matter.

How the empirical strategy works

The strategy leans on the fact that two homes with very similar distance to the coast can differ meaningfully in elevation and therefore in projected long-run flood exposure. That comparison is powerful because it holds fixed much of what buyers like about coastal housing, including the view, the access, and the local amenity value, and isolates variation in exposure that arises from topography rather than from coastal proximity.

The heterogeneity analysis by ownership type is particularly informative. Non-owner-occupied properties are disproportionately purchased by investors and by buyers who are plausibly more attentive to long-run asset value than owner-occupiers making consumption-driven housing choices. If the exposure discount is larger among non-owner-occupied properties, that pattern is consistent with better-informed or more return-sensitive buyers pushing prices to reflect long-run climate risk more aggressively than less-sophisticated buyers would.

The rent test complements the sale-price analysis in a theoretically sharp way. If sea level rise is primarily a long-run ownership risk and not an imminent service-quality issue, it should be capitalized into sale prices, which reflect claims on the full future stream of housing services and resale values, but not into current rents, which reflect compensation for current housing services. Finding a price discount but no rent discount would therefore be strong evidence that the effect is about long-run asset pricing rather than about current livability.

Main findings

The headline result is that homes exposed to projected sea level rise sell for approximately 7 percent less than observably equivalent unexposed homes that are equally close to the coast. That exposure discount is the paper’s core number and is robust across a range of specifications that vary the geographic fixed effects, the control set, and the exposure definition.

A set of supporting results makes the interpretation especially convincing.

First, the discount grows over time across the sample period. That pattern is consistent with buyers learning more about climate risk, with the salience of sea level rise rising in public discussion, or both.

Second, the discount is concentrated among more sophisticated buyers. In particular, it is larger and more precisely estimated among non-owner-occupied properties, which are disproportionately held by investors. That heterogeneity fits a forward-looking information story rather than a mechanical amenity story.

Third, the discount is stronger in communities with greater measured climate concern, as proxied by survey-based beliefs about climate change. Beliefs therefore appear to be part of the pricing mechanism, not just a background attitude.

Fourth, and especially importantly, the paper finds no comparable effect in rental rates. The rent null is central to the interpretation. If current flood damages or current livability were driving the pattern, rents should move as well. Instead, the evidence looks like a long-horizon asset-pricing effect that shows up in ownership claims but not in the price of current housing services.

Finally, the paper finds a discount even for properties that are not projected to flood in any given year for close to a century. That is precisely the piece of evidence that connects the paper to the discounting lectures. Very delayed expected damages still move current prices.

Why the paper matters

The broad lesson is that residential real estate markets can be forward-looking over long horizons. Buyers do not need to wait for frequent realized flooding before adjusting prices. If future resale values will be lower because later buyers will face greater exposure, then current buyers rationally pay less today, and that capitalization is observed in the price data.

That result positions the paper as a bridge between hedonic pricing and climate finance. The asset is a house, but the pricing economics is close to that of a long-duration claim on an uncertain future cash flow stream. A future stream of expected damages changes today’s asset value, just as it does for the municipal bonds studied by Painter or the leasehold properties studied by Giglio, Maggiori, and Stroebel.

The paper is also important because it shows that beliefs matter. The exposure discount is not uniform across buyers or locations. It is stronger where the underlying information or concern is stronger. That is useful evidence that price formation in this setting depends on who is active in the market and on what they believe about long-run climate trajectories, not only on physical exposure measured in feet of elevation.

What to focus on when you read

On a first pass, focus on four things.

First, understand the core comparison. The relevant contrast is between exposed and unexposed coastal homes that are otherwise very similar in location, timing, and hedonic characteristics, not between coastal and inland homes.

Second, understand why the rent result matters. The absence of a rent effect strengthens the interpretation that the paper is about long-run ownership value rather than current housing services.

Third, understand the sophistication result. The fact that the discount is driven disproportionately by more informed or investment-oriented buyers connects the paper to information-based asset pricing and helps rule out simple amenity explanations.

Fourth, keep the timing in mind. The paper is not about homes that are flooding today. It is about how distant, low-probability future damages are capitalized into current prices, and that is what makes it so directly relevant for the discounting and climate-finance themes of this part of the course.

Terms to know

Capitalized into prices: reflected in current asset values through the forward-looking pricing decisions of buyers and sellers.
Sea level rise exposure: the vulnerability of a property to future inundation under a specified sea level rise scenario.
Hedonic comparison: comparing prices of observably similar properties that differ in a specific attribute of interest, in order to identify the implicit price of that attribute.
Non-owner-occupied property: a property that is not occupied by its owner, typically held as an investment or rental.
Rental rate: the current market price of housing services, as distinct from the sale price of the underlying asset.

Calel, Colmer, Dechezleprêtre, and Glachant (2021): Do Carbon Offsets Offset Carbon?

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper is the central empirical reading for the offsets lecture because it goes directly at the hardest issue in carbon offset markets: additionality. An offset credit is only economically meaningful if the project that generated it would not have occurred without the revenue provided by the credit. If the project would have happened anyway, then the credit is not buying any new emissions reductions.

The difficulty is that additionality is inherently counterfactual. The researcher, and the regulator, cannot simultaneously observe the same project both with and without offset revenue. Calel, Colmer, Dechezleprêtre, and Glachant design the paper around that identification problem and use the structure of the Indian wind sector to build a conservative, observable-based lower bound on the scale of non-additional credits.

The question

Did the Clean Development Mechanism (CDM) award offsets to wind projects in India that would have been built even without offset payments?

If the answer is yes, then the program is not delivering additional emissions reductions for the credits it issues. Worse, if regulated firms in capped jurisdictions use those credits to justify incremental emissions, then global emissions rise by the amount of those incremental emissions, even though no corresponding reductions have occurred elsewhere.

Why the setting matters

The CDM was established under the Kyoto Protocol and became one of the largest international offset programs ever implemented. It awarded tradable carbon credits, known as Certified Emissions Reductions (CERs), to projects in lower-income countries whose sponsors claimed they would reduce emissions relative to a counterfactual baseline scenario. CERs could then be used by regulated entities in capped jurisdictions, including installations covered by the EU Emissions Trading System, to meet part of their compliance obligations.

India’s wind sector is a strong test case for the integrity of the CDM. It is often viewed as the kind of sector in which the program should work relatively well. Wind projects generate a clean, measurable low-carbon output, and unlike some notorious industrial-gas cases, there is limited scope for developers to create additional emissions simply in order to be paid for reducing them. If additionality problems are large in Indian wind, that is bad news for the broader credibility of offset markets, because the setting is close to a best-case scenario.

Data and measurement

The data work is a substantial contribution of the paper. The authors construct a project-level dataset covering 1,350 wind farms built in India between 1992 and 2013. For each project they determine whether it was registered under the CDM, and they collect detailed information on project location, installed capacity, wind resource quality, and distance to the electricity grid.

Two pieces of this data are especially important for the identification strategy.

First, the authors use local wind-speed and meteorological data to estimate how productive each site should be in terms of expected electricity generation. A project sited in a windier location should earn more revenue per unit of installed capacity, all else equal, and is therefore more profitable on a purely electrical-market basis.

Second, they use detailed grid-infrastructure data to estimate how costly it is for each project to connect to the electricity network. A project that is farther from existing transmission capacity is more expensive to build and to operate, so it is less profitable than an otherwise identical project closer to the grid.

Combining expected revenue on the wind-resource side with expected costs on the grid-connection side lets the authors rank projects by observable profitability in a way that does not rely on self-reported financial models.

The core empirical problem

Additionality is, by definition, a statement about an unobserved counterfactual. A developer applying for CERs is asked to argue that the project would not have been viable without offset revenue, but the researcher cannot directly observe the unassisted world, and the developer has a clear incentive to claim that the project is additional.

Rather than attempting to identify every non-additional project, which would require a full structural model of investment decisions, the authors take a more conservative route. They identify a subset of projects that are obviously non-additional on the basis of observable project characteristics. They label this subset BLIMPs, for blatantly infra-marginal projects.

How the BLIMP strategy works

The logic is direct. Consider two wind projects built in the same Indian state and in the same year, so that they face similar market conditions, regulatory environment, and technology cost. One project received CER credits and the other did not. If the subsidized project is larger, located in a windier site, and closer to the grid than the unsubsidized project, then the subsidized project appears strictly more profitable than the unsubsidized project on every observable margin that would matter to a private investor.

If the weaker unsubsidized project was built without CER revenue, then it is very difficult to argue that the stronger subsidized project required CER revenue to be built. That stronger subsidized project is a BLIMP.

This logic yields a lower bound on the prevalence of non-additionality. The authors are not claiming to detect every non-additional project. They are claiming only that the projects labeled BLIMPs are so clearly more profitable than contemporaneous unsubsidized alternatives that calling them additional is not credible. That conservative design is a strength, because if even the observable-based lower bound is large, then the program has a serious integrity problem and the true figure is necessarily at least as bad.

Main findings

The quantitative results are severe. Among the 472 CDM-registered wind farms in the sample, the authors classify 265 as BLIMPs. Those BLIMPs account for approximately 52 percent of the offsets approved for Indian wind projects over the sample period.

The number is striking precisely because it comes from a deliberately conservative classification procedure. These are not all of the potentially suspicious projects; they are the subset that the authors can confidently label as clearly non-additional using observable project characteristics alone. The true share of non-additional offsets is at least this large.

The scale is also large in absolute environmental terms. The relevant BLIMP projects were expected to generate approximately 50 million CERs over their project lifetimes. If those credits are then used by regulated firms in capped jurisdictions to emit more than they otherwise would, the paper estimates that global emissions rise by approximately 28 million tonnes of carbon dioxide, which is roughly equivalent to operating a one-gigawatt coal-fired power plant for close to five years.

One of the most damaging observations in the paper is that a random lottery assigning subsidies to Indian wind projects would have directed fewer credits to BLIMPs than the CDM approval process actually did. That observation indicates that the problem is not merely one of imperfect screening; it reflects a systematic misallocation in which the CDM disproportionately credited the projects least in need of the subsidy.

Why the results are convincing

The authors run a range of sensitivity checks that push the analysis toward giving the CDM the benefit of the doubt. For example, they examine what happens if measured output from CDM projects is systematically overstated, or if grid-connection costs of CDM projects are systematically understated. Under a range of such adjustments, the main conclusion that a majority of approved offsets went to BLIMPs continues to hold.

This matters because a natural criticism of the approach is that CDM projects may face unobserved barriers that are not captured by publicly available data on wind resources, scale, and grid connection. The authors do not argue that such barriers never exist. They argue that, to overturn the main conclusion, these unobserved barriers would have to be implausibly large and systematically related to the observable profitability margins used to construct the BLIMP classification. That is a much stronger defense of the results than a simple correlation could provide.

Why the paper matters

The paper matters because it makes clear why offsets are difficult to regulate well. Additionality sounds like an accounting rule, but it is a counterfactual judgment about project profitability. That is a demanding task for researchers with rich data and years to study the problem. It is substantially harder for regulators who must approve or reject project applications in real time and at scale, typically on the basis of information supplied by the project sponsors themselves.

The broader lesson is that offset markets can fail in a specific and damaging way. They can appear on paper to deliver emissions reductions while in practice raising global emissions, if credits issued to non-additional projects are used to authorize incremental emissions elsewhere. That is a much stronger critique than saying the program is somewhat noisy or modestly inefficient.

The paper is also useful because it demonstrates the value of conservative empirical design in a policy-relevant setting. By building a lower bound rather than overclaiming a point estimate, the authors produce a headline statistic that is difficult to dismiss on methodological grounds, and that places an informative floor under any subsequent evaluation of the CDM’s environmental integrity.

What to focus on when you read

On a first read, keep four things in mind.

First, understand the economic meaning of additionality. It asks whether the project exists because of the offset subsidy, and it is inherently a counterfactual statement about project profitability.

Second, understand the BLIMP construction. The authors compare subsidized projects to weaker unsubsidized projects built in the same state and year, using observable measures of wind resource, scale, and grid connection to rank profitability.

Third, remember that the resulting estimate is a lower bound. The paper is identifying only the clear cases, so the true share of non-additional offsets is at least as large as the 52 percent figure reported.

Fourth, keep the policy implication clear. Non-additional offsets do not merely waste money on infra-marginal projects. When those offsets are used for compliance in capped jurisdictions, they can raise global emissions relative to what would have occurred without the offset program.

Terms to know

Carbon offset: a tradable credit that authorizes emissions in one place on the basis of a claimed reduction elsewhere.
Additionality: the requirement that the credited reduction would not have occurred without the offset incentive.
CER: a Certified Emissions Reduction, the tradable credit issued under the Clean Development Mechanism.
Infra-marginal project: a project that would have been built even without the offset subsidy, so that any credits it receives are not associated with additional reductions.
BLIMP: a blatantly infra-marginal project, as defined by the authors using observable measures of profitability relative to contemporaneous unsubsidized projects.
Lower bound: a conservative estimate that is constructed to understate rather than overstate the underlying problem.

Flammer (2020): Green Bonds

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper is the central “does it work?” reading in the green-bonds lecture. A green bond is designed to channel financing toward environmentally beneficial investment, but that design claim is only economically interesting if issuing a green bond changes something real about what the firm does. If the label is purely cosmetic, then green bonds are a marketing exercise with no environmental consequences, regardless of how investors feel about them.

Flammer asks exactly that question. Are green bonds a label that firms use for reputation and marketing, or do they affect financial performance, environmental performance, or both? The answer matters for whether green-bond markets should be viewed as a meaningful instrument of environmental policy or as a classification that mainly reorganizes existing investment without changing it.

The question

Do green-bond issuers become better environmental performers after issuance, and does the market interpret these bonds as value-enhancing for the issuing firm?

This is the right question for environmental economics. We do not only care whether investors value the label; we care whether the instrument changes real outcomes, and in particular whether it reduces emissions and improves environmental performance at the issuing firm.

The broader market background

The paper opens with a set of facts about the scale of the global green bond market. Global green-bond issuance rose from approximately $0.8 billion in 2007 to $141.3 billion in 2018. Flammer also documents the rapid growth of the U.S. green municipal bond market, from approximately $0.6 billion in 2010 to $4.3 billion in 2018.

That growth is part of the motivation for the paper. Green bonds were becoming a quantitatively important component of sustainable debt financing at a very rapid pace, but the empirical literature on whether they had any real effects on firm behavior was thin. The paper is designed to fill that gap.

What the paper does

Flammer focuses on green bonds issued by publicly listed firms. That sample restriction is important because public firms are subject to detailed financial disclosure requirements and to established environmental reporting datasets, which makes it possible to track outcomes before and after issuance in a consistent way.

The paper studies three sets of outcomes.

First, it examines the stock market reaction around the announcement of a green bond issue. This is the short-run market test and speaks to whether equity investors view the issue as value-enhancing for the issuing firm.

Second, it examines longer-run financial performance using standard accounting-based measures, including return on assets and return on equity.

Third, it examines environmental performance using firms’ reported carbon emissions and environmental scores from the Thomson Reuters ASSET4 database, which compiles standardized environmental, social, and governance indicators from corporate disclosures.

How the empirical strategy works

For the announcement effect, the paper uses a standard event study. The idea is direct: if equity investors believe a green bond issue is value-relevant news, the issuing firm’s stock price should move around the announcement window, and the sign and magnitude of that movement summarize the market’s assessment.

For the longer-run outcomes, the paper uses a matched difference-in-differences design. Each green-bond issuer is matched to a similar firm that issued a conventional bond in the same year, using pre-period firm characteristics and outcome trends. The matching variables include size, profitability, leverage, and pre-period environmental, social, and governance indicators. The paper then compares the evolution of financial and environmental outcomes for treated firms relative to their matched controls before and after the green bond issuance.

This design matters because firms that issue green bonds are not a random sample of public firms. They may already differ on the dimensions that determine financial and environmental performance. Matching combined with difference-in-differences does not resolve every concern about selection on unobservables, but it makes the comparison substantially more credible than a raw before-after change in the treated group alone.

The role of certification

One of the most important distinctions in the paper is between certified and uncertified green bonds. In principle, any firm can label a bond green. If there is no credible external verification of that label, the designation carries little information content, and investors should reasonably worry about greenwashing.

Flammer treats third-party certification as a governance mechanism. Certification, typically through organizations such as the Climate Bonds Initiative, imposes standards on project eligibility and on the use of proceeds, and it provides ongoing reporting requirements. In economic terms, it raises the cost of misusing the proceeds and therefore makes the green label a more credible commitment device for the issuer.

That distinction turns out to be central to the results. The paper uses certification status as a split of the treated group and shows that the estimated effects load disproportionately on the certified issues.

Main findings

The first headline result is that the stock market reacts positively to green bond announcements. In a two-day event window around the announcement date, the cumulative abnormal return is approximately 0.67 percent, relative to a standard asset-pricing benchmark.

That is an economically meaningful magnitude for a bond issuance event. It suggests that equity investors do not view green-bond issues as wasteful diversions or purely symbolic commitments. They treat them as value-enhancing news about the firm.

The second set of results concerns longer-run financial performance. Relative to their matched controls, green-bond issuers show improvements in post-issuance profitability, with gains appearing in standard accounting measures such as return on assets and return on equity. The magnitudes are consistent with the interpretation that green-bond issues are associated with, or contribute to, financially productive investments rather than purely reputational activity.

The third set of results is the core of the paper for environmental policy. Firms that issue green bonds reduce their carbon emissions and improve their environmental ratings after issuance, relative to their matched controls. Both margins move in the direction the design of the instrument is intended to produce.

Crucially, these effects are concentrated among certified green bonds. The positive stock market reaction is statistically significant primarily for certified issues, and the improvements in financial and environmental performance are likewise concentrated among certified issues. That pattern is strongly consistent with the interpretation that governance and verification matter for whether the label translates into real effects.

What the results mean

The paper’s message is not that every bond labeled green is effective. The message is narrower and more useful: green bonds appear to work when the market has a credible governance mechanism that ties the label to real action on the part of the issuer.

That distinction is important for policy. A label without verification invites greenwashing, because issuers face little cost from using the designation loosely. Certification raises the credibility of the instrument, and in Flammer’s data the certified bonds are precisely the ones for which the evidence of real effects is strongest.

The financial-performance result is also interesting on its own terms. It suggests that green bonds are not purely altruistic activity by firms or investors. They may help firms finance valuable long-run projects, attract a supportive investor base, or strengthen internal commitment to environmental investments that then yield both environmental and financial returns.

Why the paper matters

The paper matters because it connects sustainable finance to real environmental outcomes, rather than stopping at the question of whether investors like green assets. Much of the early green-finance literature focused on pricing and demand. Flammer asks the more demanding question of whether green finance changes what firms do, and it is that question that ultimately determines whether these instruments are policy-relevant.

The paper also carries a clear policy lesson. If governments or regulators want green-bond markets to matter for real environmental outcomes, they cannot rely on issuer self-labeling. Standards, reporting, and third-party verification are likely to be central to whether the market generates measurable environmental gains.

For this course, the paper pairs naturally with Baker et al. Flammer focuses on real effects on firms, while Baker et al. focus on pricing and investor demand. Together, they provide a more complete picture of how green-bond markets operate: one side documents that investors are willing to pay a modest premium for the label, and the other documents that, under credible certification, the label is associated with measurable improvements in firm-level environmental performance.

What to focus on when you read

On a first read, focus on four things.

First, keep the distinction between a label and a real treatment clear. The paper is asking whether green bonds cause meaningful changes in firm outcomes after issuance, not whether they are correctly marketed.

Second, understand the two empirical pieces. The event study speaks to the stock market’s assessment of green-bond issues, and the matched difference-in-differences speaks to longer-run changes in financial and environmental outcomes.

Third, focus on the outcome measures. The paper does not rely on firm promises or disclosures about intent. It uses carbon emissions, standardized environmental ratings, and accounting-based profitability measures to track what actually changes after issuance.

Fourth, keep certification front and center. It is not a secondary result. It is the mechanism that distinguishes cases where the label appears to matter from cases where it does not, and it is the most policy-relevant finding in the paper.

Terms to know

Green bond: a bond whose proceeds are earmarked for projects with stated environmental benefits.
Event study: an empirical design that measures the movement in an asset price in a narrow window around a public announcement, typically benchmarked against a standard asset-pricing model.
Difference-in-differences: a before-after comparison between a treated group and a comparison group, which differences out time-invariant group characteristics and common time shocks.
Matching: a procedure for constructing a comparison group whose pre-period characteristics are similar to those of the treated group, in order to improve the credibility of the counterfactual.
Certification: third-party verification that a green bond meets a stated standard for project eligibility and use of proceeds.
Greenwashing: claiming environmental benefits from an activity when those benefits are weak, misleading, or unverifiable.

Giglio, Maggiori, and Stroebel (2015): Very Long-Run Discount Rates

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper anchors the discounting material because it speaks directly to the central difficulty in climate policy: how to value benefits and damages that arrive far in the future. A large share of climate policy asks society to pay costs today in exchange for benefits that arrive 50, 100, or 200 years from now. Whether those future benefits look first-order or negligible depends heavily on the discount rate applied to them, and at very long horizons there is very little market evidence that pins that rate down.

That is where Giglio, Maggiori, and Stroebel make progress. They exploit an unusual institutional feature of residential property markets in the United Kingdom and Singapore to recover the implicit market valuation of cash flows that begin more than a century in the future. The resulting estimates are among the most direct evidence we have on how markets price ownership claims at horizons long enough to be policy-relevant for climate.

The question

The paper asks a single, sharply defined question: at what rate do housing markets discount cash flows that begin in the very distant future?

The question is abstract in phrasing but first-order in climate economics. If market behavior implies that a dollar 100 years from now carries almost no present value, then long-run climate damages enter present-value calculations with very little weight, and stringent mitigation today becomes hard to justify on cost-benefit grounds. If market behavior instead implies that century-ahead cash flows still carry meaningful weight, then far-future climate damages matter substantially in current policy analysis and cost-benefit work.

The institutional setting

The identification strategy rests on a specific feature of property ownership in England, Wales, and Singapore. In these markets, residential property is sold either as a freehold or as a leasehold.

A freehold conveys effectively perpetual ownership. A leasehold conveys ownership for a finite number of years, commonly 99, 125, 250, or even 999 years, after which ownership reverts to the freeholder. A leasehold owner therefore receives the flow of housing services and the resale value only until the lease expires, while a freehold owner also owns the stream of housing services beyond that date.

That contractual difference is exactly what makes the setting useful for measuring very long-run discount rates. If a 100-year leasehold sells for less than an otherwise identical freehold on the same property type and in the same local market, the price gap reflects the present value of the housing-service stream that begins only after year 100. In other words, the cross-section of lease lengths reveals how the market prices cash flows at horizons that standard bond markets do not reach.

What the authors do

The authors assemble transaction-level data on the universe of residential property sales in England and Wales from 2004 to 2013 and in Singapore from 1995 to 2013. For each transaction, they observe the sale price, the contract type, the remaining lease length, and a rich set of property characteristics including location, size, and structural attributes.

The empirical strategy is hedonic. The authors compare prices across otherwise-similar properties that differ in remaining lease length, absorbing variation in location and structure with fine-grained controls and fixed effects. They do not compare small apartments in weak markets to mansions in strong markets; they compare properties that are close substitutes in location and structure and ask how price varies with remaining lease length.

Several features of the design are important for credibility. The authors include detailed controls for structure and location, exploit within-area comparisons at a fine geographic scale, and show that the same patterns appear in two institutionally distinct countries. They also study a broad range of lease maturities. That is essential, because it is the full schedule of discounts across 50-year, 100-year, 150-year, and longer leases that identifies the term structure of long-run discount rates rather than a single point estimate.

How the strategy works

The key economic object is the present value of the rent stream that begins after the lease expires. A freehold can be decomposed into two assets bundled together:

the right to the housing-service flow until year $T$,
the right to the housing-service flow from year $T$ onward.

A leasehold with $T$ years remaining conveys only the first component. The freehold-leasehold price gap therefore isolates the present value of the second component, which is a pure long-horizon cash flow.

That logic is informative because the relevant cash flow starts far in the future. For a 100-year leasehold, the missing rent stream begins only after year 100. The observed freehold-leasehold price gap therefore reveals how the market discounts a claim whose payoffs all arrive beyond a century. Standard government and corporate bonds, with maturities in the range of years to a few decades, do not let researchers isolate discounting at these horizons. The housing market creates variation in effective maturity that reaches 80 to 250 years and beyond.

Main findings

The headline result is that very long-run discount rates are low, and that markets do not treat distant future cash flows as irrelevant.

Quantitatively, 100-year leaseholds trade at roughly 10 to 15 percent discounts relative to otherwise similar freeholds. For shorter remaining maturities of 50 to 70 years, the discount grows to around 30 percent. Translated into implied discount rates, these gaps imply total long-run discount rates at horizons of 100 years or more on the order of 2 percent annually, and below approximately 2.6 percent. These rates are substantially lower than the 5 to 6 percent figures that are sometimes invoked in applied cost-benefit work on climate policy.

The evidence is inconsistent with a single, constant, high discount rate. Housing as an asset class earns relatively high average returns, and yet the market assigns meaningful value to cash flows far in the future. A constant high discount rate cannot reconcile those two facts. The estimates instead point toward a downward-sloping term structure, in which discount rates are higher at short and intermediate horizons and fall at very long horizons. The authors refer to this tension as a long-run valuation puzzle.

Why the result matters

The implications for climate policy are first-order. If century-ahead damages are discounted at 6 percent annually, they are close to zero in present-value terms, and stringent mitigation today is hard to justify from a narrow cost-benefit perspective. If long-run discount rates are closer to 2 percent, as the market evidence suggests, then far-future climate damages retain substantial weight, and the optimal scale of near-term mitigation rises accordingly.

It is important to be careful about the interpretation. The paper estimates how markets price private cash flows, not how a social planner should discount intergenerational welfare. It does not resolve debates over pure rates of time preference, intergenerational equity, or ethical parameter choices in integrated assessment models. What it does do is discipline one common claim in climate-policy debates: the assertion that markets obviously discount far-future cash flows at very high rates is inconsistent with the housing-market evidence. The far future is not priced at zero.

The paper is also consequential outside climate economics. It provides asset-pricing theory with a new moment to match. Standard consumption-based and long-run-risk models often struggle to generate both high average equity and housing returns and low discount rates on very distant cash flows. The evidence is therefore useful both as a policy input and as a test for asset-pricing theory.

What to focus on when you read

On a first pass, keep four things in mind.

First, understand the asset being priced. A freehold is a claim on the rent stream forever. A leasehold is a claim only until the lease expires.

Second, understand why the price gap is informative. The difference between freehold and leasehold prices is the market value of the rent stream that begins only after expiry, which is a pure long-horizon cash flow.

Third, focus on the headline magnitudes and their implications. A 100-year leasehold discount of 10 to 15 percent is much larger than a high constant discount rate would predict, and it implies long-run discount rates around 2 percent.

Fourth, separate the positive claim from the normative one. The paper tells us how housing markets price distant cash flows. It does not, by itself, tell us how governments should discount intergenerational welfare in social cost-benefit analysis.

Terms to know

Discount rate: the rate used to convert future dollars into present-value terms.
Present value: the value today of a cash flow or stream of cash flows that arrives in the future.
Freehold: perpetual ownership of a property.
Leasehold: ownership of a property for a fixed number of years, after which ownership reverts to the freeholder.
Term structure of discount rates: how the discount rate applied to a cash flow varies with the horizon at which that cash flow arrives.
Long-run valuation puzzle: the tension between relatively high average asset returns and the relatively low discount rates implied by market prices on very distant cash flows.

Greenstone and Gallagher (2008): Does Hazardous Waste Matter?

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper is foundational for the hedonic-pricing portion of the course. It poses a classic revealed-preference question: when a credible policy improves local environmental quality, does the local housing market respond in the way that standard hedonic theory predicts?

The paper matters for two reasons. First, it studies a major federal environmental policy, the Superfund program, whose per-site cleanup costs are large and whose benefits have been the subject of substantial policy debate. Second, it is a cautionary benchmark. Even in a setting where many observers would expect cleanup to generate large willingness-to-pay responses in nearby housing markets, the measured response is small and statistically indistinguishable from zero in the authors’ preferred specifications.

The question

How much are nearby residents willing to pay for the cleanup of a hazardous waste site, and can that willingness to pay be recovered from housing-market behavior in a quasi-experimental research design?

The standard hedonic logic is direct. If a cleanup reduces local disamenity and risk, then properties near the site should become more attractive to potential buyers and renters. In a flexible housing market, that increase in demand can show up on several margins: higher sale prices, higher rents, more construction, or compositional changes in who chooses to live there.

Institutional background

The Superfund program identifies hazardous waste sites across the United States and places the most dangerous ones on the National Priorities List (NPL) for federally funded cleanup. Sites are scored using the Hazard Ranking System (HRS), which assigns a composite risk score based on indicators of exposure pathways, toxicity, and waste quantity. During the period the paper studies, cleanup resources were limited, and in practice a score above approximately 28.5 served as the effective threshold for NPL inclusion.

That administrative threshold is the key institutional feature the paper exploits. Sites scoring just above and just below the cutoff are likely to be similar on the dimensions that matter for local housing outcomes, including underlying site hazard, neighborhood characteristics, and local trends in housing demand. The main difference between them is that sites above the cutoff received federally funded cleanup and sites below it did not.

What the authors do

Greenstone and Gallagher compare neighborhoods around the first 400 hazardous waste sites selected for Superfund cleanup to neighborhoods around 290 sites that narrowly missed selection. They study multiple outcomes at the census-tract level:

housing prices,
rental rates,
housing supply,
population, and
neighborhood compositional characteristics such as income and demographic composition.

The breadth of the outcome set is important. In a standard supply-and-demand picture of local housing markets, an environmental improvement does not have to be capitalized entirely into sale prices. If a neighborhood becomes more attractive, equilibrium adjustment can also occur through new housing construction, adjustments in rents, and compositional changes in the resident population driven by differences in willingness to pay for environmental quality. Looking only at a single outcome, such as sale price, would miss these alternative margins and understate the equilibrium response to cleanup.

How the empirical strategy works

The paper uses the HRS selection rule as a quasi-experiment and embeds it in a regression-discontinuity-style comparison. The intuition is direct: sites that score just above the cutoff received NPL designation and federally funded cleanup, while sites that score just below the cutoff did not.

A simple comparison of cleaned and uncleaned sites across the full distribution would be biased because the most hazardous sites are systematically more likely to receive cleanup, and those sites differ from less hazardous ones on many other dimensions that affect housing demand. Among sites near the cutoff, however, the difference in cleanup status is driven primarily by an administrative scoring rule rather than by large differences in underlying site quality. That is the source of variation the paper uses.

The quasi-experimental design is embedded in a hedonic framework. If consumers value cleanup, then cleanup should raise housing demand in nearby neighborhoods. In equilibrium, that demand shift can be accommodated through higher sale prices, higher rents, increased housing supply through new construction, or inflows of residents with higher willingness to pay for environmental quality. The paper therefore treats the local housing market as an equilibrium system rather than as a single-price outcome, which is one of its most valuable features.

Main findings

The headline result is a near-zero one. Superfund cleanup is associated with economically small and statistically indistinguishable from zero changes in residential property values, rental rates, housing supply, population, and neighborhood composition, as measured at the census-tract level.

In the course notes, the price estimates are summarized as approximately 0.7 to 2.7 percent across specifications, generally not statistically distinguishable from zero. The paper’s preferred interpretation is that the local benefits of these cleanups, measured through nearby housing markets, are small on the margins the research design can detect.

The cost comparison is what makes the result especially striking. The authors estimate average cleanup costs of approximately $43 million per site, and their preferred housing-market estimates imply local housing-market benefits well below that number. Taken at face value and under the research design’s assumptions, the local revealed-preference benefits do not come close to covering average cleanup costs.

What this does and does not mean

It is tempting to read the paper as a verdict that Superfund cleanup was not worth its cost. That reading is too strong, and the paper itself is careful not to make it.

The result is that local benefits, as measured through the nearby housing market, appear small. That statement leaves open several alternative possibilities consistent with the data:

true local benefits may in fact be small at the census-tract scale studied,
benefits may have been capitalized earlier, including at the time of NPL listing rather than at cleanup completion, which would attenuate the estimated effect, and
a large share of the relevant benefits may occur outside the local housing market, including improvements in health, ecological outcomes, and broader social welfare that are not captured in local property values.

The paper is therefore best understood as a disciplined revealed-preference estimate on one particular margin, not as a complete social cost-benefit analysis of the Superfund program.

Why the paper matters

The paper matters because it illustrates both the promise and the limits of hedonic valuation in environmental economics.

The promise is that when a credible source of quasi-experimental variation is available, housing markets can recover economically meaningful estimates of willingness to pay for environmental quality. The HRS cutoff provides exactly that kind of variation, and the paper shows how to embed it inside a hedonic framework that takes equilibrium adjustment on multiple margins seriously.

The limit is that even with a well-designed research strategy, interpretation is not mechanical. A small house-price response does not imply a small total social welfare gain. It may reflect timing, supply responses, information frictions, or the fact that the largest benefits are not local and therefore cannot be recovered from local housing-market outcomes.

The paper is also consequential for subsequent empirical work. Later studies have argued that the geographic aggregation used here, at the census-tract level, may be too coarse to capture the full effect of cleanups that are localized to a small area around the site itself. Spatially more granular follow-up work has found larger effects very close to the site. Those extensions do not make Greenstone and Gallagher less important; they make the paper more important, because it established the key empirical question and motivated the spatial precision that later work has pursued.

What to focus on when you read

On a first read, focus on four ideas.

First, understand the Superfund institutional rule. The 28.5 HRS cutoff is what makes the empirical design credible, and without it the comparison of cleaned to uncleaned sites would be confounded by underlying hazard differences.

Second, understand why the authors examine more than house prices. In a hedonic setting, equilibrium adjustment can occur on several margins simultaneously, and a single-outcome analysis would miss rents, construction, and compositional change.

Third, understand the headline result. The paper finds little evidence of large local housing-market gains from NPL cleanup at the census-tract level, and the estimates are generally not statistically distinguishable from zero.

Fourth, keep the interpretation disciplined. The paper documents local revealed-preference effects through housing markets. It is not a moral or comprehensive verdict on every cleanup, and it is not a full social cost-benefit analysis of the Superfund program.

Terms to know

Superfund: the federal program, formally the Comprehensive Environmental Response, Compensation, and Liability Act, that funds the cleanup of hazardous waste sites.
Hazard Ranking System (HRS): the scoring system used to prioritize hazardous sites for inclusion on the National Priorities List for federal cleanup.
Hedonic pricing: an empirical framework that uses housing-market outcomes to infer the implicit price of amenities, disamenities, and other location-specific attributes.
Regression discontinuity intuition: the idea of identifying a policy effect by comparing observations just above and just below a policy cutoff, where the cutoff determines treatment but is only weakly related to outcomes through any other channel.
Local welfare effect: the benefit to nearby residents, as distinct from broader social benefits that may accrue outside the local housing market.

Painter (2020): An Inconvenient Cost

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper anchors the municipal-bond portion of the lecture because it establishes that climate risk can raise borrowing costs well before the underlying physical damages arrive. That is a central lesson in climate finance. Forward-looking capital markets do not need to wait for realized flooding to reprice the debt of exposed jurisdictions.

The setting is also pedagogically useful because the identification idea is direct. If sea level rise threatens a county’s future tax base, public infrastructure, and service delivery, then long-maturity municipal bonds issued by that county should command higher yields than short-maturity bonds issued by the same or comparable counties. Painter tests exactly that implication and shows that the maturity-based pricing gradient lines up with the economics.

The question

Do investors demand higher yields and underwriting fees from municipalities whose future economies are more exposed to climate risk, and do these effects concentrate at long horizons?

The paper focuses on physical climate risk arising from sea level rise. The economic logic is that local governments are geographically fixed. If projected inundation is expected to damage infrastructure, lower property values, shrink the property and sales tax base, or force costly adaptation investments, then the probability that long-term debt is serviced on its original terms falls. Investors who price that probability should demand compensation today in the form of higher yields, higher issuance costs, or both.

Why municipal bonds are a useful setting

Municipal bonds are well suited to this question for two reasons.

First, municipalities are immobile. A corporation can relocate a plant away from a coastline. A county cannot relocate its roads, schools, water systems, and property tax base nearly so easily. That immobility means the exposure measure is tied closely to the issuer, not to an adjustable production decision.

Second, municipal debt is issued across a wide maturity spectrum, from short-dated notes to multi-decade general obligation bonds. That variation provides a built-in test of the climate-pricing hypothesis. Sea level rise is a slow-moving risk whose realizations are concentrated in the distant future, so it should matter much more for debt with long duration than for debt that matures within a few years. If the data show an effect concentrated in long-term bonds, that pattern is much more consistent with pricing of long-horizon climate risk than with a generic coastal penalty.

Data and measurement

Painter studies new issues in the U.S. municipal bond market. The main outcome variable is the annualized issuance cost, measured as the sum of two components:

the initial yield that primary-market investors require, and
the gross spread paid to underwriters.

This outcome is informative because it captures the total cost of raising funds for the issuer, not just the coupon that investors observe. The two components also reflect somewhat different margins: the yield reflects investor compensation for risk and liquidity, while the gross spread reflects the cost of placing the bond through intermediaries.

The climate risk variable is a county-level measure of expected mean annual loss from sea level rise, expressed as a share of county GDP. The construction is forward-looking: the measure is built from projected sea level rise exposure rather than from realized flood history. That distinction is important, because the economic question is about pricing of future risk, not about reactions to past realizations.

What the paper does

The empirical design compares issuance costs of bonds from counties with different levels of projected climate risk, conditional on the bond and issue-level characteristics that ordinarily determine municipal pricing. These controls include maturity, issue size, credit rating, call provisions, and other standard issue-level features, as well as broader market conditions at the time of issue.

The core comparison is between long-term and short-term bonds. That maturity split functions as an internal test. If climate risk is being priced because it represents genuine long-run economic damage, then long-maturity debt from exposed counties should carry a premium and short-maturity debt from the same counties should not. If instead the risk variable were proxying for some time-invariant coastal feature, we would expect to see effects across the maturity spectrum.

The paper also pushes on the result in two directions. First, it examines heterogeneity by credit rating, asking whether climate risk interacts with underlying credit fundamentals. Second, it exploits a shift in investor attention by comparing the period before and after the 2006 Stern Review, which sharply raised the salience of climate economics in policy and financial markets.

How the empirical strategy works

The identification logic is not that climate-exposed counties are identical to unexposed counties on every observable or unobservable dimension. They clearly are not. The logic is that climate exposure should predict issuance costs specifically where theory predicts: at long maturities, for lower-rated issuers, and after investor attention rises.

That layered approach is stronger than a single cross-sectional comparison. If climate risk were capturing an unrelated coastal feature, such as a regional business cycle or a time-invariant amenity, we would not expect the effect to load sharply on long maturities, interact with credit quality in the predicted direction, and strengthen after 2006. The fact that all three patterns appear simultaneously, in the same direction the economics predicts, is what makes the pricing interpretation credible.

The paper also runs a placebo-style check using noncoastal counties adjacent to exposed ones. If the climate-risk variable were simply capturing broad regional characteristics, we would expect similar patterns in those neighboring jurisdictions. That pattern does not appear.

Main findings

The headline estimate is that a one-percentage-point increase in county-level climate risk, as measured by projected mean annual loss from sea level rise as a share of GDP, is associated with a 23.4 basis point increase in the annualized issuance cost of long-term municipal bonds. Scaled to the average county’s debt issuance, this translates into roughly $1.7 million in additional annualized borrowing cost.

A basis point is one one-hundredth of a percentage point, so 23.4 basis points is not large relative to typical bond-yield movements over a business cycle. It is, however, economically meaningful on large outstanding principal balances issued over multi-decade horizons, and it represents a persistent wedge between the financing costs of exposed and less-exposed counties.

The effect does not appear for short-term bonds. This is one of the cleanest results in the paper, because it directly distinguishes long-horizon risk pricing from unrelated coastal penalties. The maturity gradient is exactly what the economic logic predicts.

The paper also decomposes the effect across the two components of issuance cost. Both the initial yield and the gross spread rise with climate risk. In other words, climate risk makes it more expensive both to attract investors and to place the bond through intermediaries.

Heterogeneity by credit rating reinforces the interpretation. The effect is larger for lower-rated bonds. That pattern is consistent with the economics: counties with weaker underlying credit fundamentals have less slack to absorb additional future shocks, so an additional long-horizon risk weighs more heavily on their pricing.

Finally, the paper documents a timing pattern that lines up with investor attention. Before the 2006 Stern Review, there is little evidence that climate-exposed counties paid more than comparable less-exposed counties for long-term debt. After the Stern Review, the gap widens significantly at long maturities. The natural interpretation is that markets did not fully price long-run climate risk until a salient information event raised attention.

Why the paper matters

The broad lesson is that climate change can affect public finance substantially before major physical damages are realized. A county does not need to flood today for climate risk to raise its borrowing costs today. If bond investors expect higher future repayment risk, they demand higher yields now, and this compensation is capitalized into the cost of public capital.

The implications for local governments are direct. Higher borrowing costs make it more expensive to build and maintain public infrastructure, and they tighten the budget constraint exactly in the jurisdictions that will later need to finance adaptation. Climate risk can therefore weaken an exposed municipality on two margins: once through expected future physical damages and again through higher current financing costs for the infrastructure needed to prepare for those damages.

The paper is also conceptually important. It shows how expectations about slow-moving environmental change are capitalized into forward-looking asset prices, in the same spirit as the housing and weather-derivative papers in the same lecture. The object being priced here is public debt rather than land or weather contracts, but the underlying economics is closely related.

What to focus on when you read

On a first read, focus on three pieces.

First, understand why municipal bonds are a useful setting. Local governments are geographically fixed, and the maturity structure of municipal debt maps directly onto the horizon structure of climate risk.

Second, understand the maturity comparison. It is the clearest piece of the identification strategy and the feature that distinguishes long-run climate-risk pricing from unrelated coastal penalties.

Third, pay attention to the interpretation of the coefficient. A 23.4 basis point effect may sound small relative to day-to-day bond-yield movements, but on the large principal balances and long maturities characteristic of municipal debt it translates into real resources and real constraints on local investment.

Terms to know

Municipal bond: debt issued by a state or local government, typically to finance public infrastructure.
Yield: the annualized return that investors receive from holding the bond to maturity, conditional on the scheduled cash flows.
Gross spread: the underwriting compensation paid to the intermediaries who place the bond with investors.
Basis point: one one-hundredth of a percentage point.
Term structure: how yields and borrowing costs vary across maturities for otherwise comparable issuers.
Investor attention: the idea that markets may price a risk more strongly once that risk becomes salient to investors, even if the underlying physical risk has not changed.

Schlenker and Taylor (2021): Market Expectations of a Warming Climate

Sun, 19 Apr 2026 00:00:00 +0000

Why this paper is on the syllabus

This paper anchors the weather-markets portion of the lecture because it asks a sharper question than the standard “what do people believe?” survey: it asks whether a liquid financial market internalizes climate information when dollars are at stake. Prices in such a market aggregate the beliefs of participants who pay a cost when they are wrong, which makes them a particularly demanding test of revealed climate expectations.

The result matters for how we think about climate-information disclosure and market-based policy. If weather-derivative prices respond both to short-run forecast information and to long-run warming trends, then these prices are not only trading instruments. They also serve as a public signal that utilities, farmers, city governments, and researchers can read off a screen without needing to operate a forecasting system of their own.

The question

The central question is whether weather-derivative prices reflect only short-run weather noise and the seasonal climatology embedded in their pricing models, or whether they also reflect scientifically grounded expectations about long-run warming.

This is an expectations paper. The object of interest is not realized temperature, but the probability distribution over future temperature that traders appear to hold when they transact. That is why the market setting is essential: the traded price is a market-implied moment of that distribution, revealed under the discipline of real money.

The market

The authors study weather derivatives traded in the United States, primarily on the Chicago Mercantile Exchange. These are contracts whose payoffs depend on temperature-based indices such as heating degree days (HDDs) and cooling degree days (CDDs) measured at specific reference weather stations in major U.S. cities. A hot summer raises CDD totals and the payoffs of CDD-long positions; a cold winter raises HDDs and the payoffs of HDD-long positions. Firms with temperature-sensitive revenues, particularly in the energy, agricultural, and retail sectors, use these contracts to hedge weather exposure, while speculators and financial traders take the other side when they believe the market has mispriced the expected outcome.

The market is useful precisely because the contract price is a summary statistic for expected future weather. If a trader has a forecast that is more informative than the market’s, that trader can profit by trading on the difference, and in the process push the price toward the underlying expectation. Over many such transactions, the price incorporates the information held by the set of traders active in the market.

Data and setup

The paper combines several data sources across eight U.S. cities. First, it uses daily settlement prices on weather derivatives covering roughly two decades. Second, it uses observed temperatures from the reference weather stations that underlie each contract. Third, it compares market-implied trends to the output of climate models that project future warming.

This setup supports two related but distinct tests. The first is a short-run market-efficiency test: do daily changes in contract prices respond to information that is predictive of near-term weather? The second is a long-run expectations test: do the trends embedded in derivative-implied expectations line up with the trends in realized warming and in climate-model projections?

What the authors do in the short run

The short-run analysis asks a direct question: when contract prices move today, are they responding to information about future weather?

To answer it, the authors relate daily changes in contract prices to weather anomalies at different leads and lags. A weather anomaly is the deviation of realized temperature from its local climatological average, after accounting for a slowly evolving warming trend.

The identification logic is straightforward. Past weather should not move prices today, because past realizations are already public information and should already be embedded in the previous day’s settlement. Future weather can affect prices today only if traders have forecasts that are informative about that future weather. A statistical relationship between today’s price change and realized weather several days ahead is therefore evidence that traders are using forecasts, and the lead-lag structure of that relationship reveals which horizons carry the most new information.

This is an event-study-style design over forecast horizons. It lets the authors identify which lead times are most informative for traders. The economic intuition is that very short-horizon forecasts are already priced in because they were anticipated in earlier days, while very long-horizon forecasts have little skill and therefore little market impact. The middle horizons, on the order of several days to about two weeks, tend to carry the largest marginal information.

What the authors do in the long run

The long-run analysis asks whether contract prices trend over time in a way that is consistent with a warming climate.

The authors extract a time series of market-implied temperature expectations from the derivative prices and compare its trend to two benchmarks:

observed warming in historical weather-station data, and
warming projected by general-circulation climate models.

If weather-market prices were driven only by recent local weather realizations, the market-implied trend would be noisy and only loosely related to climate-model projections. If instead traders internalize climate information from the scientific literature, the market-implied trend should look similar to the trends observed in the data and predicted by climate models.

Main results on short-run forecasting

The short-run evidence is strongly supportive of market informativeness. Daily changes in contract prices respond to future weather anomalies but not to past anomalies. That asymmetry is precisely the pattern an efficient forecast-using market should display.

The horizon profile is also economically sensible. The market capitalizes information about weather roughly up to two weeks ahead. Past realizations have essentially no effect on current price changes, consistent with prior pricing of that information, and very distant future realizations have little effect, consistent with low forecast skill at long horizons.

One of the most important results is the cumulative one. When the authors aggregate the price response across forecast horizons, the total capitalization is close to one-for-one. In plain language, the market appears to incorporate forecastable short-run weather information in an approximately complete way, rather than partially.

Main results on long-run climate expectations

The long-run results are the core contribution for this class. The trends in derivative-implied expectations are statistically significant and broadly comparable in magnitude to trends in observed temperatures and to trends in climate-model projections. In other words, traders appear to anticipate warming at a pace that lines up with the scientific consensus.

This is an important result because most discussions of climate beliefs rely on surveys, which are cheap to answer dishonestly, or on indirect revealed-preference measures such as migration or insurance demand. Here, the beliefs are extracted from market prices that are tied directly to future weather outcomes, and that adjust under the trading incentives of participants with exposure to those outcomes. The evidence suggests that when money is at stake, the traders active in these contracts do not ignore climate change.

The paper also documents nuance rather than mechanical extrapolation. In some northeastern winter contracts, prices reflect a localized cooling effect associated with polar-vortex dynamics. That matters for interpretation because it shows the market is not simply drawing a linear extension of older warming averages. It is responding to updated scientific and meteorological information, including information about regional climate dynamics that would not appear in a simple global trend line.

Why the paper matters

The paper matters for two reasons. First, it provides unusually direct evidence on market expectations about climate, grounded in an information-aggregating market with real money at stake. Second, it illustrates how private information becomes public information through the price system.

The second point is the broader economics lesson. If informed traders use forecasts and climate-model output when they take positions, then the market price itself becomes informative. A utility, a farmer, a city government, or a researcher can read the current price and learn something about expected future weather without personally maintaining the best forecasting model. The market is functioning as an information-aggregation mechanism.

More generally, the paper is a clean example of how prices in a liquid, forecast-sensitive market reflect not just existing public information, but that public information filtered through the incentives of traders who gain when they are right and lose when they are wrong. That is a useful reference point for thinking about the informational content of other environmental asset prices, including housing, municipal debt, and insurance contracts.

What to focus on when you read

On a first pass, keep the structure of the paper in mind.

First, understand the product. A weather derivative is a contract whose payoff depends on future temperature-based indices such as HDDs or CDDs measured at a specific reference station.

Second, understand the short-run test. Price changes today should be related to forecastable future weather but not to past weather, and the horizon profile should peak at the leads where forecasts carry the most skill.

Third, understand the long-run test. If the market internalizes climate information, the trend in derivative-implied expectations should look like a warming trend, in line with both realized temperatures and climate-model projections, rather than like short-run local noise.

If those three ideas are clear, the econometric detail in the paper is much easier to follow.

Terms to know

Weather derivative: a financial contract whose payoff depends on temperature-based indices or other weather outcomes.
Heating degree days / cooling degree days: measures of how cold or hot a day is relative to a benchmark temperature, aggregated across a contract period.
Weather anomaly: the difference between realized weather and the location-specific climatological average, after accounting for a slowly evolving trend.
Capitalization: the extent to which information is incorporated into market prices.
Climate-model output: scientific projections of future climate conditions produced by general-circulation or Earth-system models.