The impacts of natural disasters, in particular those related to climate change, have been attracting growing attention. The results show that floods hamper GDP, with the extent depending on how financial intermediary activities respond to the floods. The study is related to a growing literature that quantifies the impact on natural disasters such as climate change and the role of and implications for the financial system.
Quantifying the impacts of natural disasters, in particular those related to climate change, is important. This column assesses the impacts of floods on financial intermediation and GDP in Japan over the past 50 years using a DSGE model. The results show that floods hamper GDP, with the extent depending on how financial intermediation activities respond. In addition, projections regarding how floods will affect GDP from 2020 to 2100 are made using scenarios provided by the Network for Greening the Financial System.
The impacts of natural disasters, in particular those related to climate change, have been attracting growing attention (e.g. Noy 2015). In a new paper (Hashimoto and Sudo 2022), we quantify how floods affect GDP, focusing on the role of and implications for the financial system using data from Japan. Japan has suffered from floods over the course of its history and there are rich data on the direct effects of floods, including the amount of physical damage to private assets and public infrastructure, collected by the government. Figure 1 shows the time path of flood damage during the last 50 years in Japan. 1 Over the past few years, large-scale floods occurred one after another, such as the heavy rain in July 2018 in Western Japan and Typhoon Hagibis in October 2019 . In the case of the latter, the flood damage in economic terms totaled 2.1 trillion yen – the largest since the statistics were first released in 1961 – amounting to 0.54% of the national income.
Figure 1 Time series of flood damage in Japan
Image: Ministry of Land, Infrastructure, Transport, and Tourism
Using the flood damage data, we quantify the indirect effect, i.e. changes in financial intermediation activity and/or GDP that occur as the secondary effect of the direct impact of floods, by constructing and estimating a DSGE model. DSGE models are the standard analytical toolkits for studies on business cycles including monetary policy analysis; as of the time of writing, only a few applications have been made for the studies on the impacts of natural disasters.
Figure 2 illustrates three types of channels through which the direct effect of floods impacts the economy in our model. The first is the depreciation of capital stock. This corresponds to a case when floods render factory equipment inoperable, leading to a drop in the total amount of capital stock available for production. The second is a decline in total factor productivity. When floods disrupt public infrastructure, such as roads and bridges, firms have to use inefficient routes that they would not otherwise employ in the absence of such floods. Consequently, GDP declines even when the amount of production inputs is unchanged. The third channel is impairments of private firms’ balance sheets due to physical damage to their assets. Existing studies, such as Bernanke et al. (1999), stress that firms’ balance sheet conditions play an important role in shaping the external finance premium they face. For example, when assets pledged as collateral lose their value as a result of floods, lenders may tighten their lending standards for borrowers. Such tightening may lead to a decline in investment, reducing GDP.
Figure 2 Transmission mechanism of flood shocks
The first and second channels reduce GDP straightforwardly from the supply side. The third channel works from the demand side. All things being equal, firms with impaired balance sheets reduce borrowing from lenders due to flood-induced rises in the external finance premium, and thus reduce investments. Without much investment following floods, a decline in GDP persists, which in turn endogenously damages firms’ balance sheets, resulting in a further decline in GDP.
Our study is related to a growing literature that quantifies the impacts of natural disasters. For example, Hsiang and Jina (2015) construct the depreciation rate of capital stock due to cyclones by country/region and show that the rate is negatively related to the long-run growth rate of an economy. Noy (2012), using a synthetic control, shows that the Kobe earthquake in Japan in 1995 resulted in a persistent adverse impact on the local economy. Our study is also related to work that sheds light on the financial aspect. Hilbert (2021), using data on firms in the EU, shows how risks of natural disasters are located and how these risks translate into risks faced by financial intermediaries. Hosono and Miyakawa (2014) argue that the direct effects of the Kobe earthquake on banks reduced firms’ investment and exports. von Peter et al. (2012) find that, while large natural disasters have large and significant negative indirect effects on GDP, it is mainly uninsured losses that drive the costs, and sufficiently insured events are inconsequential in terms of foregone GDP.
Figure 3 shows the response of the capital stock, lending, and GDP when flooding equivalent to that created by Typhoon Hagibis occurs in period zero. To see the role of an insurance scheme and financial intermediation, we show the response of the variables in a hypothetical economy, which we refer to as the ‘model with insurance’, in which flood-induced damage to firms’ balance sheets at the impact period is fully insured by the households.
Figure 3 Impulse response functions for flood shocks (K, L, Y baseline+insurance in the same figure)
In the figure, a decline in capital stock in period zero represents the size of the direct effect of floods. The development in the capital stock in period one and beyond, or that of lending and GDP throughout the entire simulation period, represents the endogenous reactions of the economy, i.e. indirect effects. GDP falls by about 0.1% in period zero and sluggishly recovers to the pre-flood level. The 40-quarter cumulative sum of the decline is about 0.7%. The role of the demand factor can be seen from a comparison between the two models. With initial disruptions to firms’ balance sheets, GDP in the former economy declines to a greater degree than does the latter – by 0.4 percentage points in terms of the 40-quarter cumulative sum. Note that even in the latter model, the second-round effect on firms’ balance sheets remains at play. That is, as a result of GDP falling due to the supply factors, firms earn less and their balance sheets are endogenously impaired, which increases the external finance premium, thereby depressing the investment.
A growing number of studies predict that the scale of natural disasters will increase going forward. If so, the indirect effect may also become larger. To see this, we borrow from scenarios on damage due to river floods constructed by the NGFS, feed those scenarios into our model, and compute GDP from now up to 2100. We specifically use two of their scenarios: the ‘Current Policies’ scenario and the ‘Net Zero 2050’ scenario. 2
Figure 4 shows the time path of floods from 2020 up to 2100 constructed from the scenarios. In the ‘Current Policies’ scenario, the average flood shock starts to increase at a rapid pace from around 2080. In 2100, it will become 5.4 times higher than the average of the 2000s. Under the ‘Net Zero 2050’ scenario, the increase in flood shocks is limited to around 1.4 times.
Figure 4 Flood-induced capital depreciation rates in climate change scenarios
Figure 5 shows the GDP projections for the baseline model and the ‘model with insurance’. As of 2100, under ‘Current Policies’, reflecting the rapid increase in the flood-induced capital depreciation rate, the deviation of GDP is largely negative, reaching -0.18% compared with the case of no floods. By contrast, under ‘Net Zero 2050’, due to the limited size and frequency of floods, the two variables remain roughly the same as in the current years. The role of the insurance scheme and financial intermediation is seen from the comparison of the two models. In the ‘model with insurance’, within ‘Current Policies’, the negative deviation of GDP is limited to -0.03% in 2100.
Figure 5 Economic projections (Y-only baseline, model with insurance in the separate figures)
We quantitatively assess the impact of floods on financial intermediation and GDP using a DSGE model. In our model, floods affect GDP through three channels: depreciation of capital stock, declines in productivity due to public infrastructure destruction, and impairment of firms’ balance sheets. We estimate the model using Japan’s data from 1980 to 2019 and show that floods reduce GDP at a statistically significant level and that financial intermediation activity plays an important role in the transmission.
In addition, using scenarios published by the NGFS, we compute how floods will affect GDP from 2020 to 2100. We show that flood-induced GDP declines will become considerably larger under ‘Current Policies’ compared to the current level, although the declines may be alleviated when other alternative scenarios are considered or if the effect of the insurance scheme is incorporated.