Climate Adaptation: Issue 2

Spatial linkages will be central in mediating the impacts of climate change, both via the role they may play in the transmission of climate shocks, and by facilitating adaptation. A growing literature has examined such effects, leveraging the increasing availability of global, spatially-explicit micro-data (see, for example, Auffhammer et al. 2013, Donaldson and Storeygard 2016, Balboni and Shapiro 2025), as well as recent advances in quantitative models of the spatial distribution of economic activity (reviewed in, for instance, Redding and Rossi-Hansberg 2017, Allen and Donaldson 2022, Allen and Arkolakis 2023, 2025) to estimate aggregate and distributional implications for climate damages and adaptation policy.

Spatial linkages and the transmission of climate shocks 

The effects of climate shocks may be transmitted across regions, via general equilibrium adjustments, along a range of margins which have been investigated in recent empirical studies. Accounting for these responses has been found to have quantitatively meaningful implications for estimates of climate change damages and our understanding of adaptation requirements and mechanisms.

Trade channels have been found to be important in propagating and amplifying natural disaster shocks both within and across countries using firm-level micro-data (for a recent review, see Farrokhi et al. 2025). Barrot and Sauvagnat (2016) find that natural disaster shocks in the US over the last 30 years that affect supplier firms in production networks also impose large downstream output losses on their customers, which result in substantial losses in the latter’s market value and spill over to other suppliers. These effects are especially pronounced for suppliers producing specific inputs where few substitutable inputs are likely to be available. Consistent with this, two papers studying the 2011 Great East Japan Earthquake document evidence for substantial propagation effects both within Japan (Carvalho et al. 2021) and across countries (Boehm et al. 2019). Carvalho et al. (2021) find evidence for the propagation of the earthquake’s effects both upstream and downstream along supply chains, with impacts felt on indirect and direct suppliers and customers of disaster-affected firms. Boehm et al. (2019) consider cross-country transmission of disruption induced by the earthquake via inelastic production linkages, primarily of multinational firms. Zappalà (2023) highlights that the effects of extreme heat shocks can also be propagated through sectoral input linkages to other countries. Their results suggest that the propagation of agricultural heat shocks to non-agricultural sectors through intermediate input linkages amplifies aggregate economic losses by 31% beyond direct effects on agricultural sectors.

Migration is another channel via which spatial linkages may be important in transmitting the effects of climatic disasters across regions. Empirical studies spanning a range of contexts find evidence for out-migration responses following temperature extremes and natural disasters, the severity and frequency of which are projected to rise as the global climate changes (IPCC 2021). Missirian and Schlenker (2017) consider how weather variations in 103 source countries impacted asylum applications to the European Union from 2000 to 2014, finding that temperatures deviating from the moderate optimum increased asylum applications. Cai et al. (2016) find a positive relationship between temperature and international out-migration only in the most agriculture-dependent countries, and that this relationship follows a non-linear pattern consistent with non-linearities in the relationship between temperatures and agricultural yields. Similarly, Jessoe et al. (2016) find evidence for increased migration from Mexico’s rural areas to the US (especially from regions close to the border) and to other domestic areas in years with high occurrences of hot days, causing labour supply shortages in the origin locations. Albert et al. (2024) find increased local dryness over a decade induces out-migration from drying municipalities in Brazil. A number of studies suggest that effects may be heterogeneous across disaster types, with evidence that migration responses to flooding may be more muted than those to heat stress in Pakistan (Mueller et al. 2014) or tornados in historical data from the US (Boustan et al. 2012).

There is emerging evidence suggesting that financial capital flows may also be an important channel underlying the spatial transmission of climate disasters. Albert et al. (2024) find that persistent increases in dryness in Brazil over a decade reduce credit supply to all sectors in both drying areas and other financially integrated regions, highlighting that credit markets may propagate negative local economic impacts of drought shocks across space. Similarly, Aguilar-Gómez et al. (2024) find that extreme heat increases firms’ credit delinquency rates in Mexico and reduces their access to credit.

Beyond specific transmission channels, the spatial structure of climate shocks has important implications for aggregate damage estimates. Dingel et al. (2019) develop a framework to estimate climate damages, accounting for increased spatial correlation of productivity shocks from climate change. They find that accounting for spatial correlation increases projected inequality in the welfare impacts of climate change. 

Spatial linkages and adaptation to climate change

While spatial linkages play a role in transmitting climate shocks, they may also facilitate adaptation to their consequences. Climate change is projected to result in highly heterogeneous impacts across regions, sectors, and crops. In light of this, recent literature has considered the possibility that changes in the spatial distribution of economic activity may attenuate losses, and that trade and other linkages may mitigate the costs of reallocating consumption and production in response to climatic changes. These studies lean on quantitative spatial models, estimated using fine-grained data and empirical estimates of climate-induced changes, to project the implications of such general equilibrium adaptive adjustments for projected climate damages.

Costinot et al. (2016) estimate the macro-level consequences of micro-level climate change-induced shocks to crop yields, incorporating general equilibrium adjustments of trade and production. They use a spatial general equilibrium model together with micro-data on the productivity of each of the ten crops, before and after accounting for the future effects of climate change, for 1.7 million fields across the earth. The results suggest that climate change would result in a 0.26% reduction in global GDP through its effects on crop yields when incorporating adjustments in both trade and crop production choices across fields. While estimated declines are three times larger when crop production adjustments are shut down – highlighting the importance of this adaptive margin, in response to changing comparative advantage, in mitigating climate change impacts – only modestly larger declines are estimated when trade adjustments are shut down.

Gouel and Laborde (2021) similarly consider the livestock sector and find that trade adjustments may play a more substantial role in reducing projected losses via reallocation between import sources, rather than adjustments to the exported share of crops. Nath (2023) considers adjustments between agricultural and non-agricultural sectors and finds that sectoral reallocation may result in increases in estimated climate change losses due to the dominant effect of the so-called ‘food problem’ – whereby poorer countries specialise in low-productivity agriculture to meet subsistence needs – as climate change exacerbates this by pulling more labour into agriculture as productivity declines. Across these studies, the evidence suggests that there will be significant heterogeneity across regions: for example, Nath (2023) estimates that global losses are on average 17% higher when accounting for sectoral reallocation, but this rises to 49% in the poorest quartile of countries.

Recent work also highlights the importance of these spatial mechanisms in determining how climate shocks propagate through food markets, with consequences for food security that depend critically on spatial integration. Baptista et al. (2025) combine household panel data with a quantitative spatial model calibrated to 51 districts in Nepal, showing that the effects of floods on food prices are twice as large in remote regions. Their estimates suggest that poorer and more food-insecure households draw down assets to maintain food consumption, generating negative effects on capital accumulation. Their counterfactual simulations suggest that lowering trade or migration costs by 10% would reduce the welfare impact of climate shocks by roughly one-fifth and the rise in undernourishment by nearly one-third, though with important distributional consequences across districts. Janssens (2024) similarly finds that extreme dryness in sub-Saharan Africa has smaller effects on food insecurity in regions with shorter travel times to cities and ports, with trade integration acting as a form of consumption smoothing against local climate shocks. Crews et al. (2025) consider the international propagation of domestic heat shocks on food prices in export destinations through segmented international agricultural markets, and find that countries respond by increasing imports from existing trade partners rather than diversifying to new sources. Hsiao et al. (2024) find that governments respond to domestic agricultural heat shocks by tilting policy toward protecting consumers through export taxes or import subsidies, while responding to foreign shocks by assisting producers.

Beyond heterogeneity in mean welfare losses, Bearpark et al. (2025b) build a dynamic spatial general equilibrium model linking countries through trade and migration to consider how these channels influence welfare risks from uncertainty over future temperature realisations. Using CMIP-6 projections and an empirically estimated nonlinear damage function, they find that the standard deviation of country-level welfare losses averages over 8%, compared to a mean loss of 13%. Their model shows that trade and migration can reduce this welfare risk by up to 40% in low-income, globally integrated nations by enabling diversification of exposure to local climate shocks.

Recent work also uses quantitative spatial models to examine the consequences of changing water availability and the potential role of spatial linkages in adaptation to these effects. Carleton et al. (2025) combine microdata on water resources and agricultural production with a dynamic quantitative spatial equilibrium model to study whether the international trade in crops reduces agricultural water use through specialisation. They find that trade reduces long-run water depletion due to agricultural production, relative to autarky, even when over 93% of agricultural production happens in contexts without formal property rights for water extraction. Despite this finding, certain historical trade liberalisation reforms such as the Uruguay Round worsened water depletion by shifting agriculture towards water-scarce regions. Jedwab et al. (2025) examine factor reallocation and welfare impacts of a permanent change in local water availability, in the context of Lake Chad’s 90% shrinkage due to weather fluctuations between 1963 and 1990. Through a quantitative spatial model, they estimate overall welfare losses from shrinkage of the lake of 2.7%, with losses exceeding 10% for residents in Lake Chad areas, but averaging 1.5% in more distant regions as the shock propagates through trade and migration linkages, affecting wages and prices. Counterfactual exercises suggest that restricting labour mobility would have increased overall losses by a third and nearly tripled losses for lake residents, highlighting that spatial reallocation partially attenuates the underlying productivity shock.

Firm-level adaptation to climate risk has also been examined through the lens of quantitative equilibrium models incorporating spatial linkages in recent literature. Balboni et al. (2026) combine data on the flood exposure of firms and transportation links in Pakistan with a model of endogenous production network formation. They find that flood-affected firms relocate to safer ground, and shift purchases towards suppliers in less flood-prone regions and reached via less flood-prone roads, suggesting that firms’ learning will shape the impacts of increasingly frequent climate disasters. Jia et al. (2022) characterise how flood risk influences firms’ location choices and workers’ employment, and use this to estimate the aggregate impact of increased flood risk on the economy as the climate changes. Castro-Vincenzi (2023) uses the example of the global car industry to consider how firms redesign the organisation of plant networks and use their multi-plant structure to hedge when one of their plants faces an extreme weather event. Pankratz and Schiller (2024) use firm-level data across 74 countries to show that customers are more likely to terminate supplier relationships when heat exposure exceeds historical expectations, and switch to less-exposed replacements, consistent with experience-based learning about gradual climate change. Castro-Vincenzi et al. (2025) study supply chain diversification as an adaptation margin, showing that Indian firms source inputs from multiple locations to hedge against climate-related disruptions. While this adaptation reduces aggregate output volatility, it amplifies distributional consequences: demand shifts away from high-risk regions, depressing wages there even before shocks occur.

Dynamic considerations

Given that many of the impacts of climate change are projected to be realised over several decades, or even centuries, dynamic considerations are central in estimating long-run changes and adaptation. Recent examinations of climate change adaptation have incorporated insights from frontier innovations using quantitative models of the spatial distribution of economic activity to combine spatial linkages with intertemporal decisions across locations (see Desmet and Rossi-Hansberg 2024 for a recent review).

One branch of this literature has used models accounting for individuals' expectations and the cost of trade and migration (see, for example, Artuç et al. 2010, Caliendo et al. 2019). Balboni (2025) uses such a setup together with district-level data from Vietnam to demonstrate that accounting for future sea level rise renders contemporary investments in road infrastructure significantly less valuable than when returns are estimated without considering the impacts of future inundation. Rudik et al. (2022) use such a framework to quantify the economic effects of climate change accounting for market-based adaptation through trade and labour migration, with a focus on the US. Their estimates suggest that adaptation via these two channels is complementary, together increasing welfare by 50% more than the effects from each channel individually.

Cruz (2024) develops a dynamic framework that accounts for forward-looking migration and sectoral reallocation, with six sectors and 287 countries and subnational units, to study the labour market effects of global warming, finding that agricultural workers face welfare losses three times larger than the average worker and that employment in agriculture increases in the hottest regions as falling incomes shift consumption toward subsistence goods. This dynamic manifestation of the 'food problem' echoes the findings of Nath (2023). Shin (2025) extends this by developing a dynamic framework covering 60 countries and 10 crops, that incorporates crop switching, international trade, and forward-looking migration, both within and across countries. The results suggest that negative agricultural income shocks can accelerate structural transformation in regions including Northern and Western Africa, while positive shocks in Central Asia slow it. Counterfactual exercises restricting labour mobility reveal that domestic sectoral reallocation accounts for most of the gains from adaptation, with agricultural workers experiencing welfare gains of over 14% from labour mobility under current barriers relative to a scenario with no mobility.

A second approach applied in a number of recent studies of adaptation to climate change instead models dynamics as arising from investments in local technology (Desmet et al. 2018). Firms can invest in improving technology, and regions are linked via trade, migration, and the diffusion of technology over time. This setup has been applied to the structural estimation of the effects of dynamic adaptation to sea level rise and global warming. Desmet et al. (2021) estimate the global consequences of sea level rise for 64,800 cells across the world accounting for dynamics of migration, trade, and innovation. Their results suggest a hugely important role for adaptation: the welfare costs of sea level rise are projected to be reduced fivefold in a model incorporating all adaptation channels compared to a naïve scenario with no adaptation. The significantly larger losses from sea level rise in the scenario ignoring dynamic spatial adjustments of the economy result from individuals being unable to migrate out of flooded locations, and foregone innovation in newly emerging clusters of economic activity. Cruz and Rossi-Hansberg (2023) extend the framework to incorporate energy as a factor of production and climate change impacts manifested via impacts of temperature on local productivities, amenities, and natality rates. The results suggest highly heterogeneous losses from global warming – with losses reaching 20% in areas of Africa and Latin America while some northern latitudes gain – but that migration and innovation can serve as important adaptation channels. Conte et al. (2021) add consideration of the impacts of changes in local specialisation, finding that trade costs make adaptation via changes in sectoral specialisation more costly, which reduces geographic concentration in agriculture and drives larger flows of climate migrants.

Recent work has drawn on theoretical advances combining forward-looking dynamic migration models with local capitalists who make dynamic consumption-investment decisions (Kleinman et al. 2023). Bilal and Rossi-Hansberg (2025) build on such a framework to demonstrate that climate impacts on capital depreciation and mobility magnify the aggregate welfare costs of climate change in the US, and that adaptation via migration and investment are central in mitigating these damages. Another strand of this work integrates homeownership, wealth heterogeneity, and realistic geography into dynamic spatial models (such as Greaney et al. 2025, Donald et al. 2025, Fajgelbaum and Gaubert 2025).

Implications for spatial policy and investment decisions

The structural modeling frameworks described above open up the possibility to simulate policy counterfactuals in order to assess how alternative adaptive policies and investments may influence the projected dynamic evolution of both the economy and aggregate welfare in the face of a changing climate. Desmet et al. (2021), for instance, highlight that estimated damages from sea level rise depend crucially on the magnitude of migration restrictions, estimating that prohibitive migration restrictions would lead to projected real GDP losses of 4.5% from sea level rise in 2200, relative to the baseline case of 0.11% in the benchmark model incorporating migration adjustments. Similarly, Conte (2022) uses a similar setup to estimate that reducing migration barriers from the standards in sub-Saharan Africa to those in the European Union eliminates the aggregate economic losses of climate change in sub-Saharan Africa, but at the expense of greater climate migration and regional inequality. 

Government investment decisions can also be evaluated through the lens of these frameworks. Balboni (2025) finds that accounting for the projected impacts of future sea level rise leads to meaningful differences in assessments of where infrastructure should be allocated today. Hsiao (2025) considers how government intervention may complicate long-run adaptation to climate change by inducing coastal moral hazard, using the example of a proposed sea wall in Jakarta.

For full reference list see the end of the conclusion chapter.