Unmitigated Disasters JIE 2024
Description
This is the Replication Package for “Unmitigated Disasters? Risk Sharing and Macroeconomic Recovery in a Large International Panel”, by G von Peter et al (2024), Journal of International Economics. The paper examines the patterns of macroeconomic recovery following natural disasters. In a panel with global coverage from 1960 to 2011, the study makes use of insurer-assessed losses to estimate growth responses conditional on risk transfer. The empirical results show that major disasters reduce growth by 1 to 2 percentage points on impact, and over time produce an output cost of 2% to 4% of GDP, on top of the initial damage to property and infrastructure. Akin to wars and financial crises, natural disasters have permanent effects, in the sense that output losses are not fully recovered over time. But it is the uninsured losses that drive the macroeconomic cost; insured losses are less consequential in the aggregate, and can even stimulate growth. By helping to finance the recovery, insurance mitigates the macroeconomic cost of disasters. Many countries lack the capacity to (re)insure themselves and would stand to benefit from international risk sharing. The Stata code referenced in this replication package produced the original findings of the article cited. They mainly consist of estimated growth responses to scaled disaster impulses in a dynamic panel setting, allowing for four types of natural disasters and distinguishing between insured and uninsured losses. The code enables the exact replication of our results when using the Munich Re NatCat statistics presented in the JIE paper. For reasons of confidentiality, this replication package contains no data from Munich Re. The code can be run using public data, such as the data included in this replication package based on the EM-DAT database (publicly available at https://public.emdat.be). The datasets included in this package are sufficient for running the code, but the quantitative results do not replicate the estimates in the paper and should not be relied upon for other purposes. For background and detailed guidance, see the READ_ME file and other materials included in this replication package.
Files
Steps to reproduce
The READ_ME.pdf details the sources and uses of the material shared in this replication package. Most of the code is written in Stata.