World - The macroeconomic effects of adapting to high-end sea-level rise via protection and migration
Climate change-induced sea level rise (SLR) is projected to be substantial, triggering human adaptation responses, including increasing protection and out-migration from coastlines. Yet, in macroeconomic assessments of SLR the latter option has been given little attention. We fill this gap by providing a global analysis of the macroeconomic effects of adaptation to SLR, including coastal migration, focusing on the higher end of SLR projections until 2050.
We find that when adapting simultaneously via protection and coastal migration, macroeconomic costs can be lower than with protection alone. For some developing regions coastal migration is even less costly (in GDP) than protection. Additionally, we find that future macroeconomic costs are dominated by accumulated macroeconomic effects over time, rather than by future direct damages, implying the need for immediate adaptation. Finally, we demonstrate the importance of including autonomous adaptation in the reference scenario of economic assessment studies to avoid overestimation of adaptation benefits.
Climate change is already visible via bio-physical impacts around the world, which are expected to increase, depending on future greenhouse gas emissions1,2. This will trigger a range of socio-economic consequences and risks. To adapt effectively, and as a motivation for climate change mitigation, it is key that policy makers understand these potential consequences as well as the socio-economic effects of adaptation.
The literature has identified sea-level rise (SLR) as one of the major risks from climate change2,3. Even when meeting the target of the Paris agreement of staying well below 2 °C of global warming, mean and extreme sea levels are projected to rise substantially during the 21st century1. It has also been shown that SLR will continue for further centuries and can only be slowed down but not avoided4. The regions facing highest risk from SLR are small island states, delta regions, and often regions of the global south5,6,7, but SLR also poses significant risks for developed regions8,9,10.
There are various adaptation options to combat SLR, which can be categorized into: advance (creating new land seawards), protection (blocking inland propagation of mean and extreme sea levels), retreat (giving up land and out-migration of people), accommodation (reducing vulnerability to flooding through, e.g., floodproofing buildings or early warning systems) and ecosystem-based solutions (supporting advance, protection and/or retreat through restoring and maintaining coastal ecosystems) (cf. Oppenheimer et al.7). It has been shown that adaptation in the form of coastal protection can be very effective at reducing impacts11,12,13, and also that protection and advance is cost-efficient for densely populated and urbanized areas, but inefficient and very expensive relative to local GDP for rural and less densely populated areas11,14,15,16. Accommodation can be particularly effective for small rises in sea levels, but ceases to be so for higher SLR7. As a result, high SLR may trigger coastal retreat and the associated out-migration of people, a possibility that has received increasing attention in the literature. Accelerated SLR and increased coastal flooding has the potential to displace millions of people17,18,19,20,21. As an example, even under cost-benefit optimal protection decisions, it is estimated that 17 to 72 million people globally will have to migrate from coastal areas during 21st century22.
Given the potential major impacts of SLR, decision makers need to understand its macroeconomic effects, i.e., how SLR and different adaption responses affect whole societies and economies. On the one hand, this need has been addressed through fully integrated models (often called integrated assessment models, IAMs) which include simple representations of the economy (e.g., a simple growth model), the climate (e.g., the MAGICC23model) and impact systems (i.e., damage functions) in order to solve the climate and the economic systems’ equations simultaneously (e.g., the DICE24,25 or the FUND13 model). The downside of this approach is the lack of detail and in particular the non-consideration or only very stylized consideration of adaptation, which constitutes a major limitation, as adaptation is the dimension to which SLR impacts are by far the most sensitive26. On the other hand, the need for macroeconomic assessments is addressed by connecting climate, impact and economic models in a soft-linked manner, i.e., a sequential passing-on of information from climate models to detailed sectoral models and typically to macroeconomic models at the end of the modelling chain (e.g., a computable general equilibrium (CGE) model as for example in Parrado et al.27). As opposed to fully-integrated IAMs, which are rather coarse and highly aggregated28, the soft-link approach allows for a more detailed impact- and sector-wise assessment, enabling us to look into the distribution of effects of impacts and different types of adaptation across regions, sectors and even households29. In addition, it allows for differentiation between direct and indirect effects via market interconnections and thus for drawing conclusions as to how much localized climate shocks are amplified or absorbed within the economic system (see e.g., Hallegatte et al.30).
There are, however, two major limitations in the macroeconomic literature that we address in this paper: First, the existing literature typically assumes a no-adaptation reference scenario27,31,32,33,34, which means that coastal societies neither raise coastal protection nor retreat from the floodplain as sea levels rise. Clearly, this is an unrealistic assumption, because people have been upgrading coastal protection as response to local SLR and other coastal hazards for centuries in the past and if this fails, people will not just simply stay in the floodplain experiencing higher and higher floods every year7,35,36. Second, the adaptation scenarios considered in the macroeconomic literature so far predominantly focus on coastal protection, thus disregarding retreat, the other major adaptation response to be expected, as argued above. We acknowledge that a few macroeconomic studies do include migration; some of the IAM literature based on the FUND model has considered retreat12,37, but using a stylized national-level damage function in which extremes are only included implicitly, which according to the empirical evidence are the main drivers of retreat20. Further, FUND solves for an optimal outcome under perfect-foresight (resembling managed retreat by a perfectly informed social planner) and thus also captures migration in a highly stylized way and not in the form of reactive migration. Yet these studies indicate that the costs of SLR-induced displacement are substantial, even in optimal outcomes. Further, Pycroft et al.33 use a CGE model and include migration in their analysis. However, migration is modelled as forced consumption, thus ignoring productivity losses from losing and moving capital, and—as they use a static model—endogenous dynamics over time are not accounted for (see Tol et al.38 for a critique). A key finding of Pycroft et al. is, that the derived welfare losses from SLR increase substantially when using a broader set of impacts, e.g., by also including migration costs. Uniquely, the work of Desmet et al.39 considers migration, but without considering protection, which results in unrealistically high numbers of migrants. It also does not include sea flood cost. One interesting aspect of the study is that it considers the additional losses in economic performance due to the dispersion of spatial economic agglomerations, which is beyond the scope of models that are not spatially explicit, such as CGE models and IAMs.
In the analysis presented here we address both of these limitations and contribute to the soft-linked assessment literature. Specifically, we include—alongside a no-adaptation scenario—combinations of protection and retreat as adaptation options, allowing us to test the importance of including autonomous retreat in the reference scenario (instead of assuming no adaptation). We use a model compound that connects the detailed bottom-up coastal impact and adaptation model DIVA with the multi-sectoral and multi-regional global CGE model COIN-INT. Protection is typically a form of anticipatory, publicly planned, and capital-intensive adaptation, and hence we model it as such. Conversely, out-migration is interpreted as an ad-hoc individual retreat and thus is an example of reactive, private and autonomous adaptation40,41, with costs arising due to moving and rebuilding capital stocks. We regard these two possibilities as contrasting cases and thus select them as key adaptation assumptions in our scenarios.