After more than 10,000 years of relative stability—the full span of human civilization—the Earth’s climate is changing. As average temperatures rise, climate science finds that acute hazards such as heat waves and floods grow in frequency and severity, and chronic hazards, such as drought and rising sea levels, intensify. In this report, we focus on understanding the nature and extent of physical risk from a changing climate over the next one to three decades, exploring physical risk as it is the basis of both transition and liability risks.
In this report, we measure the impact of climate change by the extent to which it could affect human beings, human-made physical assets, and the natural world. While many scientists, including climate scientists, are employed at McKinsey & Company, we are not a climate modeling institution. Our focus in this report has been on translating the climate science data into an assessment of physical risk and its implications for stakeholders. Most of the climatological analysis performed for this report was done by Woods Hole Research Center (WHRC), and in other instances, we relied on publicly available climate science data, for example from institutions like the World Resources Institute. WHRC’s work draws on the most widely used and thoroughly peer-reviewed ensemble of climate models to estimate the probabilities of relevant climate events occurring. Here, we highlight key methodological choices:
Choice of climate scenario. We draw on climate model forecasts to showcase how the climate has changed and could continue to change, how a changing climate creates new risks and uncertainties, and what steps can be taken to best manage them. Four “Representative Concentration Pathways” (RCPs) act as standardized inputs to climate models. They outline different atmospheric greenhouse gas concentration trajectories between 2005 and 2100. During their inception, RCPs were designed to collectively sample the range of then-probable future emission pathways, ranging from lower (RCP 2.6) to higher (RCP 8.5) CO2 concentrations. Each RCP was created by an independent modeling team and there is no consistent design of the socio-economic parameter assumptions used in the derivation of the RCPs. By 2100, the four RCPs lead to very different levels of warming, but the divergence is moderate out to 2050 and small to 2030. Since the research in this report is most concerned with understanding inherent physical risks, we have chosen to focus on the higher-emission scenario, i.e. RCP 8.5, because of the higher-emissions, lower-mitigation scenario it portrays, in order to assess physical risk in the absence of further decarbonization.
Case studies. In order to link physical climate risk to socioeconomic impact, we investigate nine specific cases where climate change extremes are measurable. These cover a range of sectors and geographies and provide the basis of a “micro-to-macro” approach that is a characteristic of MGI research. To inform our selection of cases, we considered over 30 potential combinations of climate hazards, sectors, and geographies based on a review of the literature and expert interviews on the potential direct impacts of physical climate hazards. We find these hazards affect five different key socioeconomic systems: livability and workability, food systems, physical assets, infrastructure services, and natural capital.
We ultimately chose nine cases to reflect these systems and based on their exposure to the extremes of climate change and their proximity today to key physiological, human-made, and ecological thresholds. As such, these cases represent leading-edge examples of climate change risk. They show that the direct risk from climate hazards is determined by the severity of the hazard and its likelihood, the exposure of various “stocks” of capital (people, physical capital, and natural capital) to these hazards, and the resilience of these stocks to the hazards (for example, the ability of physical assets to withstand flooding). Through our case studies, we also assess the knock-on effects that could occur, for example to downstream sectors or consumers. We primarily rely on past examples and empirical estimates for this assessment of knock-on effects, which is likely not exhaustive given the complexities associated with socioeconomic systems. Through this “micro” approach, we offer decision makers a methodology by which to assess direct physical climate risk, its characteristics, and its potential knock-on impacts.
Global geospatial analysis. In a separate analysis, we use geospatial data to provide a perspective on climate change across 105 countries over the next 30 years. This geospatial analysis relies on the same five-systems framework of direct impacts that we used for the case studies. For each of these systems, we identify a measure, or measures, of the impact of climate change, using indicators where possible as identified in our cases.
Similar to the approach discussed above for our cases, our analyses are conducted at a grid-cell level, overlaying data on a hazard (for example, floods of different depths, with their associated likelihoods), with exposure to that hazard (for example, capital stock exposed to flooding), and a damage function that assesses resilience (for example, what share of capital stock is damaged when exposed to floods of different depths). We then combine these grid-cell values to country and global numbers. While the goal of this analysis is to measure direct impact, due to data availability issues, we have used five measures of socioeconomic impact and one measure of climate hazards themselves—drought. Our set of 105 countries represents 90 percent of the world’s population and 90 percent of global GDP. While we seek to include a wide range of risks and as many countries as possible, there are some we could not cover due to data limitations (for example, the impact of forest fires and storm surges).
What this report does not do
Since the purpose of this report is to understand the physical risks and disruptive impacts of climate change, there are many areas which we do not address in this report:
- We do not assess the efficacy of climate models but instead draw on best practice approaches from climate science literature and highlight key uncertainties.
- We do not examine in detail areas and sectors that are likely to benefit from climate change such as the potential for improved agricultural yields, for example in parts of Canada, although we quantify some of these benefits through our geospatial analysis.
- As the consequences of physical risk are realized, there will likely be acts of adaptation, with a feedback effect on the physical risk. For each of our cases, we identify possible adaptation responses. We have not conducted a detailed bottom-up cost-benefit analysis of adaptation but have built on existing literature and expert interviews to understand the most important measures and their indicative cost, effectiveness, and implementation challenges, and to estimate the expected global adaptation spending required.
- We note the critical role of decarbonization in a climate risk management approach but a detailed discussion of decarbonization is beyond the scope of this report.
- While we attempt to draw out qualitatively (and, to the extent possible, quantitatively) the knock-on effects from direct physical impacts of climate change, we recognize the limitations of this exercise given the complexity of socioeconomic systems. There are likely knock-on effects that could occur which our analysis has not taken into account. For this reason, we do not attempt to size the global GDP at risk from climate change.
- We do not provide projections or deterministic forecasts, but rather assess risk. The climate is the statistical summary of weather patterns over time and is therefore probabilistic in nature. Following standard practice, our findings are therefore framed as “statistically expected values”—the statistically expected average impact across a range of probabilities of higher or lower climate outcomes.
We estimate inherent physical risk, absent adaptation and mitigation, to dimension the magnitude of the challenge and highlight the case for action. Climate science makes extensive use of scenarios ranging from lower (Representative Concentration Pathway 2.6) to higher (RCP 8.5) CO2 concentrations. We have chosen to focus on RCP 8.5, because the higher-emission scenario it portrays enables us to assess physical risk in the absence of further decarbonization. (For more details click on “Our research methodology”). In this report, we link climate models with economic projections to examine nine cases that illustrate exposure to climate change extremes and proximity to physical thresholds. A separate geospatial assessment examines six indicators to assess potential socioeconomic impact in 105 countries. We also provide decision makers with a new framework and methodology to estimate risks in their own specific context.
Seven characteristics of physical climate risk stand out
We find that physical risk from a changing climate is already present and growing. Seven characteristics stand out. Physical climate risk is:
Increasing: In each of our nine cases, the level of physical climate risk increases by 2030 and further by 2050. Across our cases, we find increases in socioeconomic impact of between roughly two and 20 times by 2050 versus today’s levels. We also find physical climate risks are increasing across our global country analysis even as some countries find some benefits (such as expected increase in agricultural yields in countries such as Canada).
Spatial: Climate hazards manifest locally. The direct impacts of physical climate risk thus need to be understood in the context of a geographically defined area. There are variations between countries and within countries.
Warming is “locked in” for the next decade because of physical inertia in the geophysical system.
Non-stationary: As the Earth continues to warm, physical climate risk is ever-changing or non-stationary. Further warming is “locked in” for the next decade because of physical inertia in the geophysical system. Climate science tells us that further warming and risk increase can only be stopped by achieving zero net greenhouse gas emissions. Furthermore, given the thermal inertia of the earth system, some amount of warming will also likely occur after net-zero emissions are reached.
Nonlinear: Socioeconomic impacts are likely to propagate in a nonlinear way as hazards reach thresholds beyond which the affected physiological, human-made, or ecological systems work less well or break down and stop working altogether. This is because such systems have evolved or been optimized over time for historical climates (Exhibit 2).
Systemic: While the direct impact from climate change is local, it can have knock-on effects across regions and sectors, through interconnected socioeconomic and financial systems.
Regressive: The poorest communities and populations within each of our cases typically are the most vulnerable. Climate risk creates spatial inequality, as it may simultaneously benefit some regions while hurting others.
Under-prepared: While companies and communities have been adapting to reduce climate risk, the pace and scale of adaptation are likely to need to significantly increase to manage rising levels of physical climate risk. Adaptation is likely to entail rising costs and tough choices that may include whether to invest in hardening or relocate people and assets.