Assessing the climate change exposure of foreign direct investment

The paper investigates how physical climate risks influence foreign direct investment (FDI) location and exposure. Physical climate risks—arising from extreme weather, sea-level rise, heat and water stress—are increasingly affecting facilities globally across industries, including foreign assets. While prior FDI–environment research has centered on regulatory externalities, technology spillovers, firm environmental performance, and the link between FDI and carbon emissions, little is known about how multinationals account for physical climate risks in overseas investment decisions. The study asks whether foreign facilities face different levels of physical climate risks than local facilities across and within host countries, how risk exposure varies by industry and geography, and whether Chinese FDI differs from non-Chinese FDI in physical climate risk exposure. This inquiry is important for investors, firms, and policymakers integrating climate risk into international investment strategy and risk management.

Existing literature on FDI and the environment emphasizes: (a) location choices relative to environmental regulation stringency (pollution haven and related theories); (b) diffusion of cleaner technologies and practices via foreign firms; and (c) comparative environmental performance of foreign versus local firms. Climate-related FDI studies predominantly examine carbon emissions and transition risks. Parallel work on physical climate risks documents financial impacts on firm performance, cost of capital, and asset pricing. However, the intersection of FDI and physical climate risk is underexplored. This study fills that gap by quantifying facility-level physical risk exposure across global FDI and comparing foreign to local facilities, with a special focus on Chinese FDI.

Data: Facility-level physical climate risk scores from Four Twenty Seven (Moody’s ESG Solutions), 2019 vintage, covering five drivers: heat stress, water stress, floods, sea-level rise, and hurricanes/typhoons. Scores derive from geospatial, historical, and projection models (mid-term projection 2030–2040), adjusted by sector-specific sensitivity factors reflecting facility activities. Scores are standardized to mean 0 and standard deviation 1. Sample: 2,233 public companies headquartered in 47 jurisdictions, >1 million facilities across ~200 jurisdictions and 10 SIC groups; ~28.8% of facilities are overseas. Firm financials from Compustat (size, ROA, leverage, cash). Country controls: HQ GDP per capita (World Bank) and CO2 per capita (Our World in Data). Unit of analysis: firm–host country–industry. Estimation: Analytical weighted least squares, weights equal to a firm’s facility count in a given host country–industry cell; standard errors clustered at the industry level. Models: (1a) across host countries with industry fixed effects, outcome = climate risk driver score, key regressor = Foreign (1 if foreign-owned/operated facility). (1b) within host countries adds host-country fixed effects. (2a) across host countries among overseas facilities, key regressor = ChineseFDI (1 if owned/operated by Chinese companies, including HK). (2b) within host countries adds host-country fixed effects. Controls include firm size, cash, ROA, leverage; HQ GDP per capita; HQ CO2 per capita; and firm local experience of climate risk (average risk in HQ country). Additional analysis relates HQ-country climate risk drivers (HQHeat, HQWater, etc.) to overseas exposures (Table 3). Robustness checks include excluding resource-intensive sectors, restricting to top FDI exporters, alternative controls, and aggregating risk at firm level.

• Across host countries, foreign-owned/operated facilities exhibit significantly lower exposure to several physical risks than local facilities: floods (-0.244 SD, p<0.01), sea-level rise (-0.238 SD, p<0.01), and hurricanes/typhoons (-0.569 SD, p<0.01). Differences in heat and water stress across countries are not statistically significant. - Within host countries, differences between foreign and local facilities are small. Some are statistically significant but economically minor (e.g., heat stress differences under ~2% of a SD). Within-country patterns vary by driver: foreign facilities tend to have slightly higher water stress and lower heat and flood risks relative to local facilities. - Industry variation: Agriculture and mining show the highest aggregate physical climate risks. By driver: agriculture/forestry/fishing has the highest heat stress; manufacturing the highest water stress; mining the highest flood risk; wholesale trade the highest sea-level rise and hurricanes/typhoons risks. - Geographic variation among overseas facilities: highest aggregate risks in the Caribbean (e.g., Trinidad and Tobago), Middle East (e.g., Bahrain), and Southeast Asia (e.g., the Philippines). Specific drivers peak in regions such as Middle East/Central Asia for water stress; Southeast/Central Asia for floods; certain islands for sea-level rise; and East Asia for hurricanes/typhoons. - Chinese FDI vs. other FDI (across countries): Chinese-owned/operated overseas facilities face higher water stress, flood, and hurricane/typhoon risks (p<0.05) than non-Chinese overseas facilities; heat stress and sea-level rise do not differ significantly. - Chinese FDI vs. other FDI (within host countries): Risk exposures are not significantly different from non-Chinese FDI, except water stress where Chinese facilities show a 0.09 SD lower risk (p<0.05). This implies Chinese FDI’s higher cross-country exposure is driven by country selection rather than within-country siting. - Headquarters-country risk correlation (Table 3): Overseas exposures correlate positively with HQ-country risk by driver: HQHeat→Overseas heat (β=0.241***), HQWater→Overseas water (β=0.158***), HQFloods→Overseas floods (β=0.062***), HQSealevel→Overseas sea-level (β=0.102***), HQHurricanes→Overseas hurricanes (β=0.222***). - Robustness: Results hold when excluding resource-intensive sectors, restricting to top FDI exporters, altering control sets, and aggregating risk at firm level.

Findings indicate foreign firms tend to avoid high physical climate risk countries when choosing where to invest, resulting in lower cross-country exposures relative to local firms. However, once operating within a given host country, foreign and local facilities face similar risk levels, consistent with the primacy of location-specific hazards and activity profiles in determining physical risk. Chinese FDI’s higher cross-country exposure appears driven by location choices—potentially reflecting strategic and political considerations (e.g., Belt and Road Initiative) and latecomer dynamics that channel investment to higher-risk regions—rather than riskier siting within countries or industries compared to other foreign firms. Sectoral and regional patterns underscore where FDI portfolios are particularly vulnerable (e.g., agriculture, mining; Caribbean, Middle East, Southeast Asia). The positive linkage between HQ-country and overseas risk suggests firms’ domestic risk experience may shape international exposure profiles. These insights inform investors, firms, and policymakers seeking to integrate physical climate risk into FDI strategy, due diligence, and resilience planning.

This paper provides one of the first systematic, facility-level assessments of physical climate risk exposure in global FDI. It shows foreign facilities are, on average, less exposed across countries than local facilities, but face similar risks within host countries. Chinese FDI exhibits higher cross-country exposure to water stress, floods, and hurricanes/typhoons relative to other FDI, driven by country selection rather than within-country siting. The study highlights industries and regions with elevated risks, offering guidance for portfolio risk management and policy. Future research could: (a) extend analysis beyond public firms and incorporate private FDI; (b) examine temporal dynamics as climate hazards evolve; (c) integrate adaptation investments and supply-chain spillovers; and (d) link physical risk exposure to financial outcomes and insurance/adaptation strategies at the facility and firm levels.

• Climate risk data limitations: Four Twenty Seven’s projections may not capture the most severe events, rely on multi-model means that may under-sample tail risks, and involve uncertainty in modeling average climate shifts despite validation. - Data access: Facility-level data are proprietary and under NDA, limiting public replicability (code available upon request but relies on restricted data). - Cross-sectional design: Results reflect a snapshot (2019) with mid-term projections (2030–2040); limited inference on dynamics or adaptation over time. - Measurement sensitivity: Risk scores incorporate activity-based sensitivity factors; residual heterogeneity in facility operations may influence exposure estimates. - Generalizability: Sample includes public companies; exposures for private firms or sectors not well represented may differ.