How Climate Economics Got the Risks Wrong

The publication of a new study by researchers associated with the University of Exeter and Carbon Tracker has reopened a debate that many policymakers and economists falsely assumed was settled. The study argues that widely used economic models underestimate the risks of climate change because they smooth impacts over time, rely on average temperature changes, and ignore shocks, tipping points, and cascading failures. That critique lands uncomfortably close to the core of how climate risk has been framed in economic terms for more than three decades, largely shaped by the work of William Nordhaus and his Dynamic Integrated model of Climate and the Economy, better known as DICE.

Nordhaus is widely respected, and his work was recognised with a Nobel Prize in Economics in 2018. That recognition, however, was for methodology, not for the accuracy of long term forecasts or the adequacy of policy prescriptions derived from them. Before Nordhaus, climate change sat largely outside macroeconomic analysis. Climate scientists described physical risks, while economists focused on growth, capital accumulation, and productivity as if the climate system were stable background conditions. Nordhaus built a bridge between those domains. He created a framework that linked emissions, temperature, economic output, and policy choices into a single, computable model. Finance ministries and regulators could now ask what level of mitigation made sense in economic terms. That was a genuine advance, and it permanently changed how climate policy was discussed.

Understanding why the model became so influential requires being clear about what it actually claimed. In DICE, climate change is treated as an intertemporal optimisation problem. Economic output grows over time, emissions rise with that growth, temperatures increase in response to emissions, and damages from warming reduce output as a percentage of global GDP. Society then chooses an emissions pathway that balances near term mitigation costs against discounted future damages. The central claim is not that climate change is harmless, but that its economic impacts are gradual, manageable, and largely compensable through continued growth and adaptation. Even at 3C or 4C of warming, early versions of the model suggested global GDP losses of only a few percent relative to a much larger baseline economy.

The first major critique of this approach concerns the damage functions at the heart of the model. The relationship between temperature rise and GDP loss is not derived from historical data at comparable conditions, because no such data exist. Instead, it is calibrated using a mix of sectoral studies, limited empirical observations, and expert judgment. Small changes in the curvature of that function can produce very different outcomes. A quadratic function might show a 3% GDP loss at 3C, while a slightly steeper function might show 10% or more. Because the damage function drives estimates of the social cost of carbon and optimal mitigation pathways, this sensitivity matters. It means that outputs that appear precise are resting on assumptions that are not well constrained by evidence.

Closely related is the issue of discounting. DICE applies a positive pure rate of time preference, which means future damages are valued less simply because they occur in the future. In practical terms, a damage of $1 trillion in 2100 counts for far less in today’s decision making than a $1 trillion cost incurred now. Discounting makes sense when individuals choose between consuming today or saving for their own future consumption. It is much harder to justify when the costs are imposed on future people who have no ability to consent to the tradeoffs being made. In the context of climate change, discounting future harm assumes that future generations would accept higher risks and irreversible losses in exchange for marginally higher consumption enjoyed by people alive today. That is not an empirical assumption. It is a normative one embedded deep inside a model that is often treated as purely technical.

Another well established critique comes from risk theory. Climate change does not present a neat, well behaved probability distribution with thin tails. Physical scientists warn of tipping points such as ice sheet collapse or major changes in ocean circulation that could trigger large, abrupt shifts. When risks have fat tails, expected value optimisation becomes unreliable, because low probability outcomes dominate welfare. DICE, by construction, smooths damages over time and space, effectively averaging away catastrophic outcomes. This leads to policy recommendations that tolerate higher levels of warming because the worst cases are treated as unlikely and heavily discounted. The problem is not that catastrophe is certain. It is that the decision framework is poorly suited to situations where irreversible losses are possible.

Even before considering those deeper issues, the choice of GDP as the primary measure of damage is problematic. GDP measures market activity, not welfare. After disasters, GDP can rise because reconstruction spending increases economic activity, even as lives are lost, communities are fractured, and ecosystems are degraded. Political instability, declining health, and loss of institutional capacity are poorly reflected in national accounts. When climate damages are expressed as a percentage reduction in GDP, much of what actually matters to societies is pushed out of view.

All of these critiques have circulated in academic literature for years. What has been missing from most integrated assessment models, including DICE, is a realistic treatment of how climate stress translates into displacement and migration. Climate impacts do not only reduce output where they occur. They often undermine livelihoods, particularly in agriculture and informal economies, and trigger movement when survival thresholds are crossed. Migration is a behavioural response, not a marginal economic loss, and it operates through nonlinear dynamics that GDP based damage functions cannot represent.

The Syrian refugee crisis provides a useful illustration. Between 2006 and 2010, Syria experienced one of the worst droughts in its recorded history, affecting up to 60% of its agricultural land. Crop failures and livestock losses forced around 1.5 million people to move from rural areas into cities. Climate change did not cause the subsequent civil war, but peer reviewed research suggests it increased the likelihood and severity of the drought that acted as a stress multiplier in an already fragile political system. The conflict ultimately displaced around 14 million Syrians, roughly 7 million internally and 7 million externally. Around 1 million reached Europe over several years.

From a GDP perspective, the numbers involved were small relative to the European Union’s population of over 500 million. From a political and economic perspective, the impacts were large. The sudden increase in asylum seekers strained administrative systems, reshaped electoral politics in multiple countries, and contributed to the rise of anti immigration parties. The costs were not proportional to the number of arrivals or to Syria’s GDP loss. They emerged from institutional and political feedbacks that are invisible in aggregate economic output measures.

Central America offers a second, slower moving example. In the so called Dry Corridor spanning parts of Guatemala, Honduras, and El Salvador, repeated droughts and heat stress have reduced crop yields for smallholder farmers over the past decade. Combined with poverty, limited access to credit, and weak state capacity, these climate stresses have contributed to sustained migration toward the United States. Many migrants cite economic survival rather than conflict as the primary driver. The numbers involved are significant. U.S. border encounters exceeded 2 million in several recent years. Climate change is not the sole cause, but it interacts with existing vulnerabilities in ways that make migration a rational response.

In both Syria and Central America, the most consequential impacts of climate linked displacement were not local output losses. They were cross border spillovers that affected governance, fiscal priorities, and political stability in destination regions. These effects arrive early, at 1.2° C to 1.5° C of warming, not at the higher temperature levels where DICE projects substantial GDP damages. Discounting pushes these impacts out of relevance just as they are becoming more frequent.

This is not a matter of adding another variable to an existing model. Displacement breaks the underlying logic of the optimisation framework. Damages are not local, not smooth, and not easily compensable. They propagate through social and political systems that shape long run economic trajectories. Once large scale migration is acknowledged as a core transmission mechanism of climate risk, the idea of an economically optimal level of warming becomes difficult to defend.

The reliance on models like DICE has had real world consequences. Integrated assessment models have been used to estimate the social cost of carbon, which in turn informs regulatory impact analyses and carbon pricing discussions. When damages are understated and heavily discounted, recommended carbon prices are lower, and mitigation is delayed. For many years, values in the range of $30 to $50 per ton of CO2 were treated as economically justified, even as physical science pointed to rapidly shrinking carbon budgets. Underinvestment in mitigation and adaptation followed, not because policymakers denied climate change, but because economic models suggested there was time.

There is also a subtler impact. Models that present climate change as a manageable drag on growth encourage confidence in gradual adjustment. They underweight the risk of surprise, whether in the form of sudden migration surges, infrastructure failure, or political backlash. When those surprises arrive, governments scramble, often at much higher cost than earlier preventive action would have required.

There is also an uncomfortable political economy dimension to this story that deserves explicit attention. The authority of Nordhaus’s work, reinforced by a Nobel Prize that many non-economists understandably read as an endorsement of conclusions rather than methodology, created a convenient shield for those seeking to delay climate action. Selective citation of DICE outputs, especially its low near-term damage estimates and high optimal discounting of future harms, became a way to present inaction as economically responsible rather than politically motivated. In regulatory impact assessments, industry submissions, and some government analyses, the model’s apparent precision was used to argue that aggressive mitigation was premature, too costly, or unnecessary until later decades.

This was not a misunderstanding in all cases. It was, at times, a deliberate strategy to wrap short-term interests in the language of Nobel-backed economic rigor, even as the model’s own limitations were well known in academic circles. The result was not neutral delay. It was lost time during which emissions rose, infrastructure lock-in deepened, and the very nonlinear impacts the models struggled to represent, including displacement, institutional stress, and political backlash, moved from theoretical risk to observed reality.

None of this means that Nordhaus’s contribution should be dismissed. He created the first coherent framework that forced economists to take climate change seriously. That achievement deserved recognition. The problem is that the framework has persisted long after its limitations became clear, and its outputs have been treated with a level of confidence they do not warrant.

Climate change is no longer a distant externality that can be smoothed into a growth model. It is an active stressor reshaping societies through mechanisms we can already observe, including displacement, institutional strain, and political instability. Economic models that focus on average outcomes and marginal damages miss what matters most. As the new critiques make clear, the challenge now is not to refine damage functions at the margin, but to rethink how climate risk is framed in economic terms, shifting from optimisation toward risk management and from elegant abstraction toward messy reality.