Economic development and declining vulnerability to climate-related disasters in China

This study used two types of data. First, natural disaster impacts records during 1949–2015 were retrieved from published statistical yearbooks and relevant references, which reported the direct impacts of natural disasters that measured by deaths and DELs, DELs are measured by the repair and replacement costs required to restore the affected properties (e.g. houses, roads, equipment), which does not include other indirect time-element economic losses (e.g. output losses due to business interruption) and environmental impacts (e.g. natural resource damage). The main natural disaster types in China include drought, earthquake, heavy snow, flood, mass movement wet (including rock fall, avalanche, landslide and mudslide) and storm (including typhoon, wind and hail). In this paper, we treat all these, except earthquake, as climate-related natural disasters. These disaster impacts data were from three sources:

• a.  
  The recorded annual provincial (31 provinces in the mainland China) total deaths (1978–2015) and total DELs (1989–2015) caused by natural disasters in China were obtained from China's Civil Affairs Statistical Yearbook (Ministry of Civil Affairs, various years, via data.cnki.net) published by Chinese Statistics Press.
• b.  
  Earthquake disaster event impact records (1978–2015) from China Earthquake Yearbook (China Earthquake Administration, various years), which is published by Chinese Seismological Press, and these records include information on the deaths and DELs by counties.
• c.  
  National total deaths (1949–1977) and DELs (1949–1988) data for climate-related natural disaster were complemented by statistical information from the State Scientific and Technological Commission of Major Natural Disasters (SSTCND 1994, Wu et al 2014), which systematically summarized the natural disaster impacts by main natural disaster types from 1949–1990 for China as a whole.

Second, economic and social development data of China that including: (i) historical provincial GDP and population data were collected from the China Statistical Yearbook of National Bureau of Statistics of China (www.stats.gov.cn), (ii) the consumer price index (CPI) and the GDP deflator of China provided by the World Development Indicators from the World Bank (http://data.worldbank.org/data-catalog/world-development-indicators). For the sake of spatiotemporal comparisons, original reported nominal DELs (in China Yuan) were deflated to the 2015 year price level according to the CPI. GDP is also deflated to the 2015 year price level by the GDP deflator.

The annual natural disaster impact (deaths and DELs) records by disaster types and by province are only available since 2007, thus, we resort to generate the climate-related natural disaster impact data series (Loss_climate, i.e. deaths or DELs) based on disaster impact records from all natural disasters as a whole (Loss_disasters) and earthquakes alone (Loss_earthquake) as described above, respectively (Wu et al 2014),

$$\begin{equation} {\rm{Los}}{{\rm{s}}_{{\rm{climate}}}} = {\rm{Los}}{{\rm{s}}_{{\rm{disasters}}}} - {\rm{Los}}{{\rm{s}}_{{\rm{earthquake}}}}. \end{equation} \tag{ 1 }$$

Finally, deaths (1978–2015) and DELs (1990–2015) data for climate-related disasters by year/province can be constructed and then aggregated to the West, Central and East regions.

Due to different conceptual frameworks and disciplinary views, definitions to vulnerability also differ (Adger 2006, Cutter and Finch 2008, Simelton et al 2009, IPCC 2012, Fraser et al 2013, Koks et al 2015, UNISDR 2017). In this paper, vulnerability means 'the degree of loss to each element should a hazard of a given severity occur.' (Coburn et al 1994) that is dynamic and varying across temporal and spatial scales and arising from various physical, social, economic, and environmental factors (IPCC 2012). Vulnerability can be computed as a percentage of losses as compared with total exposure (Peduzzi et al 2012, Hallegatte et al 2013, Jongman et al 2015). As such, climate-related disaster vulnerability is defined according to the mortality rate (the ratio of the reported fatalities to the exposed population) and direct economic loss rate (the percentage of the reported DELs to the exposed GDP),

$$\begin{equation} {\rm{Vulnerability}} = \frac{{{\rm{Los}}{{\rm{s}}_{{\rm{climate}}}}\left( {{\rm{deaths \, or \, DELs}}} \right)}}{{{{\rm{E}}_{{\rm{exposed}}}}\left( {{\rm{population\, or\, GDP}}} \right)}}, \end{equation} \tag{ 2 }$$

where E_exposed is the situation of people, infrastructure, housing and other tangible human assets located in (a particular) hazard-prone areas (UNISDR 2017). Population and GDP are used as exposure indicators in this paper. Measuring exposure is not easy (e.g. Wu et al 2017), for floods as an example, exposed population or GDP can be modelled using maximum inundation map combined with population or GDP maps (Jongman et al 2015). While for a group of climate-related disasters as in this study, this exposure estimation is extremely a difficult task due to large amount of hazard data needed. As such, we supposed that for each province it was hit by one or more climate-related disaster sub-types in each year, total population or GDP was defined as an exposure measurement indicator for the vulnerability calculation (equation 2), which will induce exposure measurement bias, and affect the vulnerability value, this uncertainty will be discussed in section 4.

This study performs a sub-national (i.e. West, Central and East) analysis of the relationship between economic development and climate-related disasters. Regional exposure is the accumulated value of GDP or population in the provinces that belong to the region. China is generally divided into East, Central and West regions according to the economic development levels and geographical locations of each province, and the East region is generally accepted as the most developed area, with the Central and West regions representing underdeveloped and less-developed regions, respectively.

Finally, these values were converted to US dollars (US$) using the exchange rates between China Yuan and US$ in 2015 (i.e. 1 US$ = 6.2284 China Yuan). For each year of analysis, provincial disaster impact and exposure data were summed on a region level. For each region (West, Central or East), the mortality and direct economic loss rates (vulnerability) can be calculated as the ratio of reported fatalities and DELs for the exposed population and GDP. The Mann–Kendall trend test is used in this study as the significance test for the disaster impact trend.

The number of climate-related disaster events in China, as reported in the Emergency Events Database (EM-DAT), has increased notably since the 1980s (figure 1(a)). Here we group the subcategories of hydrological (like floods and landslides), meteorological (like tropical cyclone) and climatological disasters together as the climate-related disasters. However, the fatalities and DELs reported before 1980 (figures 1(b) and (c)) are not consistent with the number of reported disaster events (figure 1(a)) because fewer disaster events were recorded before 1980 owing to limited data accessibility and information gathering, thus, many events were not recorded in the EM-DAT, which partially explains the seemingly drastic increase in reported disaster events since 1980.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For annual total number of reported fatalities, there is a clear declining trend (p < 0.001, Mann–Kendall test) since the founding of the People's Republic of China of 1949–2015. The decrease surpassed 5.8% per year during 1978–2015 and 4.4% per year since 1990 as indicated by the bar in figure 1(b), which means that the death toll number reduced by 224 per year since 1978. Ten-year mean fatality statistics show that the average annual total fatalities of China had decreased from over 14 000 persons before the 1970s, to approximately 7000 between the 1970s and 1980s and approximately 6000 in the 1990s, and this number further declined to approximately 2000 in the 21st century (figure 1(b) and figure 2(a)). By economic region, the West contributed approximately half (49.1%) of the accumulated total annual fatalities between 1978 and 2015, which was followed by the Central (29.6%), where the more economically developed East has the lowest share. Fatality decreasing trends are consistent (p < 0.001), but the rate of change varies between 1978 and 2015, in each of the economic regions, with absolute fatalities declining by 67.0%, 86.7% and 80.2%, in the West, Central and East, respectively, in the 2010s compared with the 1980s (figure 2(a)).

In contrast with the decreasing trend of fatalities, the reported DELs from climate-related disasters show a significant increasing trend (p < 0.001) during 1978–2015 in China (1.7% per year during 1978–2015 and 1.3% per year since 1990, bar in figure 1(c)), and the DELs increased by US$ 1 billion per year. Decadal average annual DELs (figure 1(b)) has increased significantly, from 24 billion (US$) in the 1980s, to 45~47 billion in the 1990s and an alarming 64 billion for the years in the 2010s by far (figure 2(c)). In regional distribution, the Central, East and West regions accounted for 35.9%, 35.2% and 28.9%, respectively, of the accumulated total annual DELs from 1990 to 2015.

While the DELs changes varied notably among the three economic regions, the West has seen the most rapid increase. Comparing the DELs (figure 2(c)) between the 1990s and 2010s, the West differed by a factor of 2.3, whereas the East only showed a minor increase with a factor of 1.3. Statistically, the West economic region showed a significant upward trend (p < 0.001), similar trend was not statistically significant for the Central (p = 0.770) and East (p = 0.359) over the same period (figure 2(c)).

Overall, the climate related disaster impact data from 1978–2015 in China has shown an increase of DELs (with average annual rate of 1.7%) and a decline of life loss (with an average annual rate of 5.8%). These overall trends reconfirm general observations around the world in relation to development, that is, with economic development, the per disaster event life loss tends to decrease while the DELs tends to increase (Hallegatte et al 2017, Hallegatte 2017). In other words, with economic development, climate-related disaster is becoming less deadly but more expensive.

Since the open and economic reform period in the late 1970s, China has seen transformative socioeconomic changes over the recent decades, with an average annual national GDP growth rate of 9.4% between 1978 and 2015 (figure 1(c)). The total disaster DELs reported in this period showed a significant (p < 0.001) correlation with the national GDP (R² = 0.664). Linking to the changing vulnerability, we next explore in detail the relationship between vulnerability reduction and economic development in China as a whole as well as among the economic regions in China.

The results presented in figures 3(a) and (b) show that the overall national vulnerability to climate-related natural disasters declined significantly (p < 0.001) over the period from 1990–2015. At the national scale, the 5 year moving average of the mortality rate (the ratio of reported fatalities to the total population) and the DELs rate (percent of reported DELs to GDP) both show significant (p < 0.001) declining trends over the period from 1978–2015 (figures 3(a) and (b)), and the decadal average mortality rate and DELs rate declined by 74.4% and 74.1%, respectively, from the 1990s–2010s (figures 2(b) and (d)). At the regional level, this declining trend is also significant (p < 0.001) for the West, Central and East regions for both the mortality rate and DELs rate. The mortality rate decreased by 59.5%, 84.2% and 84.5% from the 1980s–2010s for the West, Central and East, respectively (figure 2(b)). The mortality rate is higher in regions with a lower income level (i.e. in the West) and lower in regions with higher income levels (i.e. in the East) (figure 3(a)).

Zoom In Zoom Out Reset image size

The DELs rate is higher in the middle-income region, i.e. Central China, before 2006 (figure 3(b)), it is higher in the West region after 2006, with the DELs rate decreasing by 60.3%, 82.4% and 79.5% from the 1990s–2010s for the West, Central and East regions, respectively (figure 2(d)). Furthermore, a trend of relative convergence of vulnerability (mortality rate and DELs rate) among the West, Central and East regions is observed from 1990–2015 (figure 3). To quantitatively demonstrate how income level co-relates with vulnerability, we fitted the income and vulnerability data as shown in figure 4.

The results clearly show a strong negative non-linear (power function) relationship between income (GDP per capita) and mortality rates for the period from 1978–2015 (figure 4(a)), and between income (GDP per capita) and direct economic loss rate from 1990–2015 (figure 4(b)). To further illustrate this co-relation, we grouped the income levels into Low, Low-middle and Upper-middle according to the World Bank classification criteria in 2015 (table 1).

There are notable differences on the DELs rates between income levels. It seems that there is a possible income level threshold (around 4000US$ per capita GDP) above which we see rather significant vulnerability reduction in terms of disaster mortality and DELs rates.

Regionally, there are substantial differences of overall vulnerability among the economic regions—West, Central and East (figure 4). Take the 2010s as example, with a national average –about 1.3 for morality rate and 0.7% for DELs rate (figures 2(b) and (d)), the West and East regions differ by a factor of 6 (3.1/0.5, mortality rate) and 3 (1.3/0.4, DELs rate) (figures 2(b) and (d)). Such difference is highly correlated to the overall income levels of the different economic regions with the East's per capita GDP is approximately two times of the West (figure 4).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Previous studies have observed an inverted U-shaped relationship between economic development (per capita GDP) and DELs rate on a national scale for natural disasters as a whole (Kellenberg and Mobarak 2008, Zhou et al 2014). However, clear evidence of such a pattern for climate-related natural disaster vulnerability is not observed in any of the aggregated economic regions in this study, which is consistent with other climate-related natural disaster studies described above (Kahn 2005, Jongman et al 2015, Georgeson et al 2017, Ward and Shively 2017). Our results show a general pattern of decreasing vulnerability over time within all economic regions and a clear vulnerability gap between them. Importantly, the vulnerability declined non-linearly with per capita GDP.

Economic development-induced security demand should be an important driver for vulnerability reduction activities (figure 5). With rapid urbanization and economic growth since 1978, the value of assets and population at risk of climate-related natural disaster has escalated (Qin et al 2015), major climate-related disasters and their impacts have forced the government to pursue effective new strategies to reduce disaster risk and to reduce vulnerability. Taking Chinese flood disaster management as an example, the government started to focus on essential engineering measures in the 1990s, including the construction of both the Three Gorges Dam in the upper Yangtze River and the Xiaolangdi Water Control Project in the Yellow River. In 1994, the Standard for Flood Control (GB 50201–94) was approved, and a Flood Control Law was passed in 1997. Finally, Chinese average annual flood control investment had increased from 2.6 billion US$ in the 1980s, to 15.0 billion US$ in the 1990s, and to 83.5 billion US$ in the 2000s and further to 203.2 billion US$ in the 2010s (2010–2015) (figure 6), especially in the 2000s, this investment accounted for 46.4% of the annual gross investment for water project construction on average. For non-engineering measures, China issued a national emergency plan for natural disasters in 2006, and the government began to seek a balanced approach using both structural and nonstructural measures to reduce disaster risk.

Overall, from the supply-demand relationship perspective, increases in income increase the demand for safety, and a higher income enables individuals to employ costly precautionary engineering and non-engineering measures in response to this demand (Toya and Skidmore 2007). Evidently, richer regions are better able to agree upon, invest in, and implement zoning regulations, building codes, early warning systems, emergency response systems and other measures that reduce natural disaster risk and mitigate the consequence of natural disasters (Ferreira et al 2013, Park et al 2015, Hallegatte et al 2017). Which should be the causal link between economic growth and vulnerability reduction, but it is extremely hard to quantify such relationship due to the difficulty in attribution since many other factors involved in shaping the vulnerability that is highly context specific (Cutter and Finch 2008).

Generally, studies that statistically analyze the relationship between natural disasters and economic development have focused on the worldwide scale and largely depend on the available global disaster loss datasets, i.e. the EM-DAT from CRED (Anbarci et al 2005, Kahn 2005, Rashky 2008, Kellenberg and Mobarak 2008, Schumacher and Strobl 2011, Tanoue et al 2016). However, these worldwide or cross-country analyses were based on natural disaster records combining a large number of countries, country clustering reduces the bias of the tests of statistical significance on standard errors but tend to overstate the significance of explanatory variables (Ferreira et al 2013). Therefore, these analyses do not represent a true dynamic study of the relationship between economic development and disaster vulnerability in any particular country. Meanwhile, although EM-DAT is widely used in these studies, EM-DAT has the shortcoming of underreporting smaller natural disaster events, especially, in developing countries (Jonkman 2005). Because there is high spatial disparity in economic growth from the west to the east in China, a sub-national level analysis of the complex dynamic relationships between economic development and climate-related disasters in China is helpful to obtain insights into the necessity of adaption strategies, especially in undeveloped regions.

On the other hand, hazard frequency is an important driver that affects the annual disaster losses and increases the variance in human and direct economic loss rates for each province, therefore, this study performing the analysis at the economic region (or zone) scale helps eliminate this variance generated by differences in hazard frequency as climate-related hazard frequencies are more uniform inter-annually in the economic region scale rather than in province scale, thus, the relationship among factors is more robust in economic region scale than in province scale theoretically.

The current paper attempts to add a subnational empirical study that maps the co-relationship and pattern between economic development and the vulnerability to climate-related disasters. While we have identified a statistically significant co-relationship between the economic development and vulnerability to climate related disasters, this co-relationship should not be taken as or confused with any claim on the causal relationship between the two. While qualitatively there are surely causal linkages between the two, quantifying such causal 'attribution' is challenging both conceptually and methodologically. First, the economic growth itself can be affected by disasters, yet quantifying those effects are difficult. Using the global disaster database, Cavallo and colleagues (Cavallo et al 2013) examined the relationship between catastrophic natural disasters and economic growth. They have concluded that only extremely large disasters have a negative effect on output in both the short and the long runs. At the same time, no significant effect on economic growth has been detected once political changes were controlled.

On the other hand, the impact of climate change on both economic growth as well as the pattern of climate-related hazards would also have profound impacts on the relationship between economic growth and vulnerability to climate-related disasters. Burke et al (2015) have found a representative concave-type non-linear effect of temperature on economic production in both poor and rich countries, and technological advances or the accumulation of wealth and experience since 1960 did not fundamentally altered this non-linear relationship between productivity and temperature. Hsiang et al (2017) further developed a new integrated architecture to report updated climate change damage based on a notable improvement in econometrically derived dose-response functions that reflect the nonlinear effects of temperature to climate cost via Bayesian meta-analysis in the United States.

Furthermore, the causal relationship between economic development and vulnerability is a complex balance of increasing exposure (more to lose) and enhanced adaptive capacity (more to invest). Yet, our knowledge on estimating the feasibility and cost of adaptive responses to climatological processes remains limited (Hsiang and Narita 2012, Fraser et al 2013).

Some limitations exist in our study. First, improved disaster reporting arising out of better information and communications technology can affect data on disaster impacts records. Disaster impacts usually could be systematically underreported in poorer regions and in early decades, as Zhang et al (2009) indicated that natural disaster impacts records prior to the early 1990s in China are considered incomplete for the description of the spatiotemporal patterns at the provincial or a lower scale, which is also one of the reasons for economic region scale analysis (in this study) but not province scale. Even after the reporting biases are adjusted (i.e. the yearly deaths were rising in poor provinces before 1990s), the declining trend of deaths and mortality rate with per capita GDP are still not change. While for DELs, we used the disaster impact data series from 1990–2015 directly to avoid this reporting bias. Anyway, compared with lots of missing disaster impact records of EM-DAT for Chinese regions, especially for early years, the published statistical disaster impacts data from Chinese government agencies (as we used above) should be a beneficial supplement to better reflect the regional disparity of disaster impacts.

Second, because we did not distinguish between sub-type climate-related disasters, there is a gap between the real exposure and the administrative population or GDP used in this paper. Extracting the real exposure to each disaster type as performed in other related studies (e.g. Jongman et al 2015) is difficult because of the absence of information on the exact location of the reported fatalities and DELs and the corresponding disaster types, therefore, using accumulated provincial total population or GDP to represent economic zone exposure will inevitably exaggerate the actual exposed population or GDP if the occurrence of multiple climate-related disaster events at the same province in the same year is not accounted for. As such, the non-linear relationship between per capita GDP and vulnerability should be taken with care. Further studies using a certain disaster type and high-resolution modelling of exposure via asset value map (Wu et al 2018) to explore this non-linear relationship is of great value for reducing this uncertainty (e.g. Jongman et al 2015, Geiger et al 2016).

Finally, as a developing country, economic development in most of China's provinces still does not reach a high-income level, therefore, how vulnerability will change with economic development in the future will depend on how effective disaster risk reduction measures are implemented now.