1. Risk-tradeoff analysis: asymmetries and inconsistencies

In the C2G report, Tyler Felgenhauer and colleagues seek to revive and apply risk-tradeoff analysis (RTA)Footnote ²² to provide a framework for risk-risk analysis of SRM. The report sketches some of the dimensions of risk involved, while indicating where knowledge is lacking for a complete assessment. Figure 1 provides an outline of the key steps involved in RTA.

Figure 1.

Simplified process flow of RRA / RTA.

Source: Author, based on descriptions in Felgenhauer et al. (note 21) and Graham & Wiener (note 22).

RTA is based in quantification and comparison of risks. It has previously been applied where risks can be made comparable, typically using life-year metrics. RTA offers an analytic framework to assess and compare effects of interventions on both target risks and on ancillary or countervailing risks or benefits, with attention to mechanisms by which risks might be offset, transferred or transformed in terms of their nature and distribution.Footnote ²³ Through assessment of relative risks, RTA promises the possibility of identifying “risk superior” approaches. Its advocates suggest that its broad scope can help draw decision-makers’ attention to a more comprehensive set of considerations, even if some of them cannot be quantified or compared on the same metrics. In practice however, measurable risks tend to be more influential,Footnote ²⁴ and the report argues for “focus on cumulative or net present value measures of risk” (p 36) to help decide between possible interventions.

The report establishes the target risk as “climate risk,” and suggests a headline comparison of “climate risks in a world with and without SRM” (p 2). Departing from the tendency noted above – although the report invokes potential “irreversible tipping points” – climate risks are not described as existential or catastrophic, merely “multiple and interacting” (p 2). Nonetheless, the narrow specification of climate risks could be problematic and misleading. But the report extends discussion to political, social and economic consequences of climate impacts, and to “countervailing societal risks” generated by SRM. Here a “comprehensive risk-risk analysis” is seen as necessary “to assess these multiple risks thoroughly and holistically, to reduce overall risk, and to address any residual risks from SRM decisions”.Footnote ²⁵ On the other hand, no other alternative interventions that might reduce residual climate risk are discussed.

This limitation flows through into how the report establishes a benchmark for risk comparison. To measure the reduction of the target risk, it conceptualises a “level of residual climate risk that might be addressed by SRM” (p 4) as determined by the extent of mitigation and adaptation. But SRM would only influence certain effects of climate change, and cannot perfectly counter the impacts of greenhouse gas forcings, so not all residual risk is susceptible to SRM (see Box 1). The report cites multiple modelling studies that seek to quantify the climate effects of SRM on particular parameters in particular scenarios. Discussion of whether these different effects can all be achieved in the same plausible real-world scenarios is largely lacking, despite a discussion of the prospect of “risk correlation.” Nor is there any effort to determine how much of the residual risk which SRM could mitigate is otherwise avoidable through other (less risky) interventions.

Box 1.

Starting with clear definitions

The report considers a set of scenarios based on otherwise low, medium or high levels of mitigation – where the most ambitious scenario is based on the Paris Agreement (p 32ff). Although SRM is initially presented in the report purely as a supplement: an additional climate action, these scenarios treat mitigation and SRM as at least partial substitutes. And they are constructed and described in such a way that the reader is left with the impression that the level of residual risk (arising from the level of mitigation and adaptation) is entirely exogenous rather than partly reflecting decision-makers expectations regarding the availability and efficacy of SRM.Footnote ²⁶

The potential risk tradeoffs considered include both biophysical risks arising directly from the intervention; and sociopolitical risks from actors using and responding to SRM strategically. The report, however, pays somewhat asymmetric attention to these distinctive risks. The biophysical risks (and benefits) are set out in detail across ten pages with dozens of references to modelling studies, while the sociopolitical risks get three sides of brief qualitative discussion, some of which is speculative and even dismissive in tone.

This reflects the distribution of studies in the existing peer-reviewed literature.Footnote ²⁷ Yet for different reasons both sets of risks are presented in ways that underestimate their potential seriousness. The assessment of biophysical risks is based on simplified and idealised modelling which tends to provide an overly optimistic impression of SRM,Footnote ²⁸ in part because it rarely examines the effect of SRM interventions on “otherwise unavoidable impacts of climate change,” but rather the outcome of interventions which presume exogenously determined levels of mitigation.Footnote ²⁹ Thus the biophysical risks appear both limited in impact and scientifically well constrained, although that is a false impression based on unrealistic assumptions of effective governance baked into the models.

The sociopolitical risks are also framed as manageable – but in ways that climate risks (and countervailing risks associated with mitigation and adaptation responses) have so far proved unmanageable. The report argues that sociopolitical risks can be abated by good governance, while ignoring the shortcomings of and opposition to such governance found elsewhere in the climate sphere.Footnote ³⁰ This results in further asymmetry: the risks associated with SRM are limited by idealised implementation and governance; while the risks on the other side of the trade-off (those associated with climate change) are not similarly constrained.Footnote ³¹ RTA’s role in helping identify possible ways to constrain or mitigate countervailing risks is valuable, but building such measures into the calculus on one side only introduces a potentially seriously harmful bias.

Theoretically, in RTA, through attention to risk offsets or transfers, the distribution of risk is also considered. In this report the relevance of risk distribution is repeatedly stressed, yet also then downplayed: the analysis only once addresses the issue, with a brief statement that the distributional effects might be either positive or negative. Nor is any attention paid to the factors that might contribute to differentiated vulnerability to either biophysical or sociopolitical risks, and how these in turn might be affected by different interventions.

A further asymmetry becomes clear when we turn to how the concern of potential mitigation deterrence is treated. The risk that pursuit of SRM might distract from mitigation is presented as a major countervailing sociopolitical risk (alongside conflict generation and termination shock), although described as “moral hazard” or “abatement displacement.” But the literature cited here is highly partial and one-sided, relying primarily on the dismissive assessments of ReynoldsFootnote ³² who misinterprets the problem through a neoclassical individualist economic lens. Similar weight is given to the unsubstantiated argument that consideration of SRM might instead galvanise climate action (because of the perceived risks of SRM), which downplays the risk of mitigation deterrence.Footnote ³³ Yet however unlikely, the risk of deterrence merits serious consideration because of the potential for irreversible impacts: should deterrence arise, and SRM fail, then there is no possibility to retrospectively increase mitigation.

Notably in the C2G report no scenario is considered in which the risks generated by SRM exceed the risks mitigated, despite plausible circumstances for such an outcome. These include, first, that feasible planned or possible future mitigation is traded away in favour of SRM which then proves technically or politically impractical: with the net effect of increased overall risk.Footnote ³⁴ Secondly, if SRM is introduced but terminated while masking appreciable levels of warming, the result would be much more biophysically and sociopolitically harmful accelerated warming: again increasing overall risk. Arbitrarily ruling out such scenarios is particularly problematic in that in alternative (non-utilitarian) maximin ethical decision-making approaches, alternative strategies would be ranked not by average risk, but by maximum risk avoided.Footnote ³⁵ The possibility of this scenario of termination shock would eliminate SRM from consideration under such a rule but has no effect in this proposal for RTA.

The report acknowledges the asymmetry of evidence between sociopolitical and biophysical risks, and the limited scope for quantifying the former. The authors’ response is to suggest a qualitative assessment of such risks to make trade-offs and weightings explicit. Yet, they still pursue quantified metrics for most factors, and treat uncertainties not as inherently unpredictable or unknowable factors,Footnote ³⁶ but merely as a measure of the variability of predicted likelihoods or magnitudes. This reflects a modernist technocratic and calculative conception of risk.Footnote ³⁷

Here again, uncertainties are presented as susceptible to reduction by further research. Both this, and the structure of the sketched RTA implies that the process might be iterated, as knowledge expands. However, they are not explicit about the case for repetition, nor about the distinct decision points – regarding enhanced research, outdoor experimentation or delivery technology development, as well as deployment itself – which might be informed by such analysis (see Table 1).

Table 1.

Decision points at which RRA / RTA might help

This brief review has identified methodological shortcomings and asymmetries in the proposed RTA which seem to persistently favour considering the SRM option. In theory these might be corrected by including a full range of alternatives and scenarios, defining the risks involved with more care to ensure appropriate comparisons are being made, considering interaction effects in detail, paying more attention to the distribution (and causes) of risks and vulnerabilities. However, such corrections may not prove easy if the shortcomings are rooted in cognitive framing limitations, particular social imaginaries, or associated political or ideological presumptions. Nor if the motive for proposing risk-risk assessment is in fact intended to promote consideration of SRM. In the following sections, I turn to consider the politics of RTA, and subsequently the ideological frames and imaginaries of risk–risk approaches more generally.

2. The politics and social imaginaries of RTA

It’s not enough to assess the logical consistency and objectivity of the methodology. We need to ask how the introduction of such approaches might serve to influence the politics of geoengineering discourse. Such choices are not without rhetorical or ideological effects, which might be more significant in influencing policy choices than the content of a risk assessment exercise.

The history of RTA offers some insights here. RTA emerged as a challenge to perceived regulatory overreach in the 1970s, and formed part of the Reaganite de-regulatory agenda in the 1980s. It is rooted in a technocratic risk management culture dominated by statistical analysis of economic costs and impacts on health and life. In contrast to emerging ideas of a “risk society”Footnote ³⁸ which were more influential in European politics, RTA helped sustain a “modern.” social imaginary in US politics. Here I use the term social imaginary to describe widely shared presumptions about “how the world works” socially, politically and economically,Footnote ³⁹ and following Christopher Groves,Footnote ⁴⁰ understand the dominant imaginary as modernist and technocratic.

In practice RTA was mobilised to roll-back regulation, providing a method to support industry concerns about the economic harms arising from allegedly unbalanced or overly precautionary pollution and safety regulations. One of the primary developers of RTA ran the Office of Information and Regulatory Affairs for a period, implementing corporate friendly decisions. An influential critique of RTA published in 2002 found that it paid excessive attention to the countervailing risks of environmental regulation and too little attention to ancillary benefits.Footnote ⁴¹ But the critique also defended the underlying model of Cost Benefit Analysis (CBA) – despite problems of incommensurability – as a tool for making regulatory decisions. One of the authors of that critique was more recently appointed by Biden to run OIRA, a decision which might be seen as a riposte to Trump’s rejection of “rational policy.”

RTA is still promoted as “rational” policy today. The C2G report adopts the language and framing of CBA, and explicitly rejects “non-rational” public responses (p 3) to SRM and counter-poses “science” as a way to avoid “biases and heuristics” (p 5). It also repeatedly insinuates that existing policy may be “distorted” by irrational responses.Footnote ⁴² However, rationalist presumptions are not necessarily helpful in assessing options and interventions. For example, presuming rational, optimised SRM deployment in a world of irrational public opinion and policy making means risks of SRM are easily underestimated.

In other words, far from being a neutral, objective tool, RTA is highly politicised. It is not only an appeal to the politics of rationalist modernism, but one that also more than nods to the anti-regulatory political right. In a debate where the political right remains often sceptical of the case for climate action, this might, arguably, encourage such actors to embrace action. However, previous research suggests that if such an effect materialises, the action embraced would likely be to replace mitigation with geoengineering, not to supplement it.Footnote ⁴³ This likelihood is increased if RTA is understood as “normalising” SRM as a climate policy option to be traded off against mitigation according to relative costs and benefits, rather than as an exceptional measure only justified by existential risks.

In the present context, RTA is being advocated ostensibly not to resist regulation, but to add a new response in the face of inadequate regulation. However, the past record of RTA demands that we are particularly attentive to whether the structure of the methods proposed might instead imply using SRM as a substitute for other climate action. In the following section I turn to discuss this and other implications of risk–risk approaches for climate policy making.

1. A false binary?

Any claim about the nature of plausible pathways to climate safety depends on presumptions about the feasibility of rapid social change, carbon removal techniques and the acceptability of temporary temperature overshoot which merit interrogation. The risk–risk frame however, tends to imply that the only alternative to climate harms is solar geoengineering. This has two related consequences which I examine below – that we face a choice between a greater and a lesser evil, and that the two scenarios are otherwise independent of one another – but in itself establishes a false duality, despite the rapidly depleting carbon budget for 1.5°C. Even if alternative pathways might involve other risks and harms, they should not be excluded from assessment if RTA’s objective of identifying potential “risk-superior” pathways is genuinely to be pursued.

Some rehearsals of the risk–risk framingFootnote ⁴⁴ even suggest that the alternative to solar geoengineering is unmitigated climate change, as if no further emissions reduction or adaptation can be foreseen. Other risk–risk accounts create space in which existing and possible further mitigation might be accounted for. This includes the C2G report, but that still treats the residual risk within any given scenario as not only all susceptible to SRM, but also immune to other measures.

The binary framing not only establishes SRM as the only alternative to further climate harms, but it also implies that all the risks of climate change might be eliminated by SRM (albeit replaced by a new set of risks). However, this is false: ocean acidification is the most obviously unaffected impact, but there are multiple ways in which the forcings from SRM are not reciprocal with those from greenhouse gases. Some harms and risks would remain, however well designed, implemented and governed the SRM intervention is.Footnote ⁴⁵ Moreover, this binary approach risks overlooking other ways in which harm and vulnerability to it might be reduced, especially those which counter the ways in which neoliberal economies actively create inequality and vulnerability.Footnote ⁴⁶

If RRA is to be helpful to policymakers the appropriate comparison would be comparing the risks introduced by adding SRM with the otherwise unavoidable residual risks (after foreseeable mitigation and adaptation) arising from the climate change factors it could affect. Box 1 and Fig. 2 suggest ways of more clearly specifying and categorising the (climate) risks involved. This should help ensure that the potential benefits of SRM are not exaggerated. It also adopts a deliberately precautionary starting point: placing an onus on those suggesting a novel untested technology with acknowledged serious risks to demonstrate that other measures cannot deliver the same benefits at less risk.Footnote ⁴⁷

Figure 2.

Understanding residual risk.

2. The “lesser evil”

The risk–risk binary suggests a choice must be made between undesirable options, and implies that the appropriate choice would that with less risk: the “lesser evil.” If there are only two options and both are bad, the implication here is that our moral obligation is to choose the one that is “less bad” on aggregate. Such analysis is morally consequentialist (and utilitarian) – a standard element of modern “realist” ethics, in which leaders are urged to “assess rival policies against their likely consequences and to seek the ‘lesser evil’ among feasible choices.” ^{Footnote 48} Lesser-evil choices are suggested to be particularly appealing if the alternative is catastrophic or existential risk.

But such a comparison of harms is not necessarily the most ethical approach. Stephen Gardiner argues that the “lesser evil” justification would be susceptible to “moral corruption,”Footnote ⁴⁹ and choices made in line with the convenience and interest of those with the power to decide. The false binary arises here because alternative strategies are not (or have not been) seen as feasible by the powerful countries and wealthy elites shaping the policy decisions (and responsible for the existing state of the climate). A lesser evil approach based on utilitarian aggregation also risks downplaying the issue of distribution of harms. By contrast, many ethical theories suggest a need to avoid additional harm to the less well-off, or at least would weight such concerns disproportionately.Footnote ⁵⁰ This is not to argue that one school of ethics is correct and the other incorrect, but to recognise that RRA methodologies presume an ethical approach which is contested.

For RRA to be useful to decision makers the embodied ethics must be part of the debate, not just about how to interpret risk–risk analysis, but also whether to use it in the first place. In turn this implies a need for broad public participation and international deliberation.

3. Independency or interaction

Modernist technocratic social imaginaries are firmly rooted in scientific reductionism, which struggles to understand co-evolving and interconnected systems. In some applications of risk–risk analysis the risks are clearly independent: for instance the likelihood of the earth being hit by an asteroid does not change if we consider building a nuclear armed asteroid destruction capability.Footnote ⁵¹ But the risk of climate catastrophe is a product of human actions, not astrophysics. Its likelihood does change when we consider developing SRM: shrinking if the “galvanisation” hypothesis is correct or dominant, growing if the mitigation deterrence effect is dominant. This socially constructed impact is separate from considering the material risk-mitigation possibility of developing these techniques in practice.

But RTA engages poorly with such systemic interaction effects between the development and deployment of SRM and residual climate risk. Mitigation deterrence is a major ethical concern for SRM – whether understood as an emergent effect, or a consequence of vested interests; not only because of potential addition to climate harm if SRM proves ineffective, but also because of additional side-effects (e.g., millions of deaths from continued particulate pollution from fossil fuel combustion) even if SRM is effective. The challenge of mitigation deterrence has to be included and strongly weighted. While the C2G report recognises interdependence of risks, and the prospect of “unjust applications of SRM,” they imply these problems can be easily overcome by design and governance despite the persistent failures of efforts to establish effective international climate governance.

Moreover, as noted already, the report downplays deterrence risks, citing the “galvanisation” claim. But there is a contradiction here in that SRM is being considered because there are (implicitly) no other ways to increase mitigation (or otherwise reduce risks). Uniquely granting SRM the power to galvanise additional mitigation is entirely unjustified.Footnote ⁵²

Thus, for RRA to be useful for decision makers it has to encompass reflexive analysis of the prospect of mitigation deterrence, and to extend to other possible ways of stimulating mitigation in excess of the limited scenarios included by C2G. Figure 3 suggests one way this might be conceptualised.

Figure 3.

Incorporating interaction into RRA.

4. Exceptionality

Most advocacy for a risk–risk framing justifies consideration of SRM in the context of the potentially existential risks it might abate. This positions SRM as an exceptional measure (a view also widely held by policy makers).Footnote ⁵³ If “exceptional measures” are legitimately acceptable responses to existential climate risks, then “radical” or transformative proposals to phase out fossil fuels, overturn consumerist values and behaviours, or even to embrace “degrowth” should be considered alongside the dominant model of technological innovation and market-based policies.Footnote ⁵⁴ But such measures are generally rejected as impractical within the modernist social imaginary in which technological innovation both drives economic growth and tames risk.

And if assessments inspired by “existential risk” rhetoric then fail to consider other “exceptional responses” to climate risk, or apply risk management techniques that are not designed for existential risks, this creates what might be called an “exceptionality paradox.” This might even be described as a “bait and switch” move, if the assessment is justified by existential risks, but then adopts conventional CBA style methods, founded on fungibility and aggregation. Such methods may be suited to everyday issues such as road safety regulation, but cannot quantify the most serious risks involved here.

Managing existential risks with exceptional measures is not simple and raises further political concerns. Notably, exceptional responses are often justified in emergencies – with rapid decision making which might exclude normal political and democratic processes. The idea that a “climate emergency” might justify SRM deployment is not new, but has been widely criticised on ethical grounds – that exceptional measures undermine democratic politics and the international collaboration needed to deliver mitigation (and that would also be essential to deliver just and ethical solar geoengineering).Footnote ⁵⁵

For RRA to be useful to policy makers, it should not retreat into technocratic methodologies, but examine a wider range of potential responses, with methods that can both reflexively evaluate uncertain and extreme risks, and enable democratic consideration of radical responses (see Fig. 4).

Figure 4.

Reflexive and participatory RRA.

5. Excluding other logics

Insofar as the risk–risk approach frames assessment in terms of climate risks, it excludes other logics for undertaking solar geoengineering. It ignores the prospect that SRM might be deployed (or avoided) for reasons other than seeking to reduce climate change itself. To research advocates, steeped in climate science, reducing climate impacts might seem the only possible rational reason for considering such a risky option. But to security experts, and students of political science, geoengineering appears as a hybrid, dual-use security technology.Footnote ⁵⁶ Its deployment might involve climate-related goals, but that could merely mean masking impacts enough to justify continued exploitation of fossil fuels for geopolitical reasons.Footnote ⁵⁷ The risk equation in a world of high continued emissions, masked by solar geoengineering, would probably look very different from one where an idealised intervention helps “shave the peak” of emissions related temperature rises.Footnote ⁵⁸ The implications of geoengineering initiated, or manipulated as a security intervention for relative national advantage might look very different again.Footnote ⁵⁹

For RRA to be useful to policy makers, the implications of deployments following non-climate logics,Footnote ⁶⁰ as well as those arising in non-ideal, but still climate-motivated, deploymentsFootnote ⁶¹ must be assessed too.

6. Governance and risk repertoires

Despite the justificatory rhetoric of climate risk as an existential threat to some valued referent object – such as civilisation, thus meriting consideration of the otherwise unthinkable, the risk–risk approach, and RTA in particular, invokes specific concepts of risk that are largely embedded in a conventional technocratic risk management repertoire closely wedded to the modernist social imaginary. Risk is understood as undesirable: “any chance of an adverse outcome to human health, the quality of life, or the quality of environment;”Footnote ⁶² yet tame-able.Footnote ⁶³ Risk here is quantifiable, objective, subject to aggregative calculus and technically manageable. Uncertainties and the unknown are spaces colonisable by research, advancing a frontier of knowledge and certainty. Decisions about acceptability of risk are treated as matters for expert estimates rather than public deliberation. Public perceptions of risk – especially regarding novel technologies – are irrational and should be corrected by improved knowledge.Footnote ⁶⁴

The heritage of CBA and risk-benefit approaches is clear.Footnote ⁶⁵ But this risk repertoire is inappropriate for addressing potentially existential risks and exceptional responses to them, not least in how it treats uncertainty. There is good reason to consider that the risks of climate change have been underestimated by quantifying modelling techniques in the technocratic paradigm. This might well justify the addition of SRM to the basket of responses we should consider. But we should beware of then using the same methods that underestimate the risks of climate change to assess the risks of SRM.

This leads to the same conclusion as the exceptionality paradox: if the reason for assessing SRM is the threat of existential risk, then the appropriate risk assessment and management repertoire is more like that of security: worst case scenarios, focus on low probability but high impact events, and not treating incommensurate risks as fungible. But a securitised risk management repertoire would also face serious problems, not least regarding justice and ethics.

There would seem no technical reason to prevent RRA being reconfigured to form part of a novel, genuinely precautionary, risk management repertoire that helps humanity live with uncertainty, supports meaningful action to avoid worst-case outcomes, and reflects an ethics of care.

Such a redesign would mean multiple changes, including the following: Assessing both the risks and ethics of a full range of exceptional responses without trying to wedge them into a calculative framework. Ensuring balanced treatment of sociopolitical risks as well as biophysical, and of possible risk mitigation measures. Fully acknowledging uncertainties and unknowables and developing non-quantitative ways to assess them. Building in systemic analysis of interaction effects, including those triggered by early decisions on research or experimentation. More generally avoiding misleading aggregation and not treating distinctive risks as commensurable. Examining worst cases, and non-ideal scenarios – on both sides – and giving them ethically appropriate weight. Incorporating the risks of misuse, or suboptimal deployment, not just presuming good faith efforts to optimise climate outcomes and manage side effects. Putting the distribution of risk and vulnerability at the centre. Embedding public participation and deliberation at every step. And enabling deliberation and debate over appropriate ethical bases for assessment and decisions.

Providing anything more than a sketch of such an alternative risk management repertoire is beyond the scope of this article, but given the proliferation of “existential threats” and a state of “polycrisis” such development would seem urgent.

Whether RRA’s advocates would support such transformative methodological and epistemological changes is at best an open question – even laying aside the possibility that such advocacy might be serving a predetermined view regarding the value of SRM – given the stickiness of social imaginaries and ideological presumptions. But without them, RRA seems unlikely to provide practical and meaningful information to decision-makers wrestling with multiple and intersecting crises.