Climate-Vulnerable Firms, Credit Supply, and Cascading Failures: An Agent-Based Model

We study the joint dynamics of firm and bank fragility under climate risk, decomposed into its physical and transition components, within a Greenwald-Stiglitz-type agent-based economy in which a population of heterogeneous firms interacts with a single aggregate bank through a credit market. The model embeds three climate-related channels: (i) a temperature-dependent damage function on output (physical risk ); (ii) a carbon levy/permit mechanism that prices greenhouse-gas emissions (transition risk ); and (iii) R&D investment that jointly raises productivity and lowers carbon intensity, so that the two risks interact endogenously with growth. We derive closed-form expressions for the bankruptcy probability, the optimal capital stock, and the Walrasian equilibrium interest rate, and we run agent-based simulations on the NGFS Net Zero 2050 pathway. Without a bank, physical risk dominates the output-growth losses while endogenous R&D pulls carbon intensity below its deterministic trend. With an active bank, the credit channel roughly doubles the macroeconomic cost of each climate channel relative to the accommodative-credit benchmark, and climate stress propagates into bank equity, bad debt, and cascading defaults rather than into the price of credit. Our aim is to reproduce stylised facts and identify emergent macroeconomic patterns rather than to forecast any specific economy: parameter values are empirically plausible (calibrated on French data where possible), and we transparently document the régime in which the endogenous default rate and interest rate settle at economically high values, leaving a finer quantitative calibration to subsequent work.