Research on economic damage caused by drought in Central Vietnam
Short-term and long-term effects of drought on manufacturing industries and domestic trade
Contributing to proposing solutions for drought management in the context of climate change in Vietnam
Climate change is projected to create new hazards such as glacier melting, sea level rise and extreme weather events in proportions never seen before (Bosello
To mitigate the damage caused by natural disasters, scientists have developed many methods and models to predict natural disasters before they occur and help people to be more proactive in coping with natural disasters (Okuyama
This study applies the Synthetic Control Method (SCM) to estimate the economic damage caused by the extreme drought in 2013 in Phu Yen Province, Central Vietnam. SCM is a statistical method used to evaluate the effect of an intervention in comparative case studies by constructing a weighted combination of control groups, to which the treatment group is compared. This comparison is used to estimate what would have happened to the treatment group if it had not received the treatment. The drought in Phu Yen 2013 is chosen for the case study because it caused great damage in the Central region in the past 10 years. Additionally, it is appropriate time to apply SCM since it requires a pre-drought data series and a postdrought data series. The results of the damage valuation can inform decision makers in the design and implementation of solutions to prevent and increase community resilience to disasters.
Phu Yen Province is located in the lower Ba River basin. It is the largest river in the central coastal region, flowing through four central provinces of Viet Nam including Kon Tum, Gia Lai, Dak Lak and Phu Yen with an area 13,900 km2 (Fig. 1a). In 2013, it was the heaviest drought event in the period 2005-2016 in Phu Yen Province. Many areas in Phu Yen Province did not have rainfall in the first 6 months of the year. According to the Provincial Irrigation Sub-Department, in nearly 25,000 hectares of summerautumn rice, over 1,300 hectares in total had been lost. There were also 5,000 hectares of rice waiting to die (http://thongkephuyen.gov.vn/). All irrigation reservoirs in this province were below the dead water level. Two reservoir systems of Ba Ha and Song Hinh were used to supply water for the downstream area but water level in the rivers and streams in Phu Yen were dry, leading to a serious shortage of domestic water. According to Department of Agriculture and Rural Development (DARD) Phu Yen’s statistics, over 35,000 people in more than 10,200 households suffered from severe water shortage for their daily life. Dong Xuan, Son Hoa, Tuy Hoa, and Song Cau districts were severely affected. This province’s historic drought has also led to an unprecedented wildfire disaster. Only in the first three months of the year, there were 66 forest fires in the area, burning 1,200 hectares (http://sonongnghieppy.gov.vn/Home/List/29). In addition, there were over 2,400 hectares of burned plantation forest.
According to the Center for Research on the Epidemiology of Disasters (CRED) (2020) (https://www.cred.be/), a natural disaster is defined as a natural event that is out of local control and requires external assistance. Also, according to CRED, a natural disaster is recorded when one of the following 4 conditions is established: (1) More than 10 deaths are reported; (2) More than 100 people are affected; (3) A state of emergency is declared; and (4) Call for external assistance is settled.
This study adopted the SCM model which have been applied in previous studies for disaster damage valuation (Abadie 2021; Albalate and Padró 2018; Cavallo
Suppose that J is the number of provinces in control group. Thus, there are J+1 provinces in the dataset (including 1 province in treatment group). Call YitN is per capita income of province i at the time t if not being affected by the drought and Yit is per capita income of province i at the time t in case being affected by the drought 2013. For provinces
For years after the occurrence of the drought, let αit=Yit-YitN representing impacts of the drought on province i at time t, (t = T0 + 1, T0 + 2, …T). The objective of the study is to compare the per capita income of the province affected by the drought with other unaffected provinces. Therefore, we have to estimate:
Data about Yit is available, thus estimation of
with wj≤0 (wj is the weight for the control provinces).
The estimation of wj as follow (Abadie 2010):
subject to the restriction that w2,…, wJ+1 are nonnegative and sum to one.
where W is the matrix (J × 1) that contains the weights of the provinces in the control group, X1 is the matrix (K × 1) contains explanatory variables for per capita income of Phu Yen. X0 is the matrix (K × J) contains explanatory variables for per capita income of provinces in the control group. In the treatment province, there are similar explanatory variables.
Hence, (X1-X0W) is the difference between the treatment and control groups for explanatory variables. The SCM estimates the weights by minimizing the difference between Phu Yen province and the control group at the period of time before the drought occurs. The estimation of wj will help to identify provinces in the control group that are similar to Phu Yen in terms of income and other social and economic factors (Table 1).
Once the weighting of the SCM has been determined for the pre - drought period, it is used to construct a counterfactual trend for the economic variables in the post-drought period. The difference between this counterfactual trend and the actual trend for the treated unit (Phu Yen) represents the estimated drought effect.
In this case study, if the drought has an impact on reducing per capita income, it is able to estimate that the value of
where p_value is the probability that the impact of the extreme drought in Phu Yen is less than that on the provinces in control group. In which
In the comparison, it is necessary to determine a control group to compare with a treatment group. The results are only meaningful when the control group is not affected by natural disasters and the treatment group is affected. However, among 63 provinces and cities (total in Vietnam), none of the provinces was not affected by the drought 2013. Furthermore, the impact of natural disasters on provinces varies in the extent of damage (number of deaths, destroyed houses). Therefore, the control group is defined again as a group that is not affected by the “major disaster”. So, to what extent of damage that a natural disaster is called “major disaster”.
According to Le (2013), “major disasters” are defined as 25% of the biggest recorded natural disasters or those with more than 5 deaths per million people and over 275 houses destroyed per million people. Therefore, the control group is the provinces that were neither affected by this drought nor any “major disaster” during the study period. Twenty-nine (29) provinces met the conditions of being the control group (Table 1, Fig. 1b).
The research collected natural disaster and economic data during 2000-2019. Natural disaster data is applied to build the control group while economic data is to estimate the impact of disasters on per capita income. Disaster data was collected from the database “Disaster Information Management System” (www.desinventar.net) of the United Nations Office of Disaster Risk Reduction. This database provides information regarding disasters for all nations and regions including Vietnam. To test whether the control group is economically affected by major natural disasters after 2013, the dataset from CRED (2020) recording the losses caused by natural disasters from 1953 to the present is used. This dataset contains information in a provincial level.
Data on per capita income was collected from the General Statistics Office (GSO) for the period of 2009-2019 and variables are denoted in Table 2. The explanatory variables for per capita income were chosen based on recent empirical research on income (Vu and Im 2011). They are Sales Retail per capita (SALES) as a proxy for domestic trade, goods transported (INFR) representing infrastructure, capital for provinces (CAP), number of doctors per capita (DOCTOR) representing health care, the number of pupils per capita (STUDENT) representing education, and land per capita (LAND).
From the available dataset, SCM was applied to assess the impact of drought on per capita income. The independent variables were selected including SALES, INFR, CAP, DOCTOR, STUDENT, and LAND. To minimize the gap between the treatment and control groups, the weights of provinces are estimated in details in Table 3. It is able to estimate the per capita income of Phu Yen province if there was no occurrence of drought 2013 by comparing the estimated income value and the actual income value of Phu Yen.
Fig. 2 shows the actual income trend of Phu Yen (dashed line) and the control group (solid line). The detailed analysis shows that the drought reduced the per capita income of Phu Yen in the period 2013-2019 by VND 160,000. These results had statistical significance of less than 10% for the period 2013-2015 while other years had no statistical significance. In other words, in the short term, the 2013 drought caused a decrease in per capita income in Phu Yen, but in the long term, the drought had no impact on per capita income.
The reliability of the research results is very important. Based on research by Abadie
The GSO report on per capita income for provinces and cities was divided into smaller categories: income from salary, income from agriculture, forestry and fishery, income from industry, construction, trade, services, and income from other sources. In order to get more specific policy recommendations, the impact of drought on these income components was analyzed in more detailed.
When analyzing the impact of drought 2013 on income from agriculture, forestry, and fishery (AFF-income) and in order to increase the reliability of the research results, the study included the following explanatory variables: Annual rice yield (RICE), annual grain yield (NUTTREE), annual timber production (WOOD) and annual aquaculture production (FISH) (Table 2). The analysis showed that the drought reduced the per capita income from agriculture, forestry, and fishery by an average of VND 220,000 per month, accounting for 11% of the average income of local people in Phu Yen. Fig. 5a shows the trends of income from agriculture, forestry and fishery of Phu Yen and the control group.
Permutation test shows that the above results are statistically significant at 10% in 2013 and 2014, 5% in 2015, 2016, 2017, and 2018 (Fig. 4). From the above results, it is confirmed that in the long term, the drought impacts reduced incomes from agriculture, forestry, and fishery of people in Phu Yen.
For income from salary, the selection of explanatory variables (SALES, INFR, CAP, DOCTOR and STUDENT) was similar to the income analysis. After analysis using synthetic control methods, the results show that this drought reduced income from salary. However, the results were not statistically significant after permutation tests. Fig. 5c shows the trend of Phu Yen’s income from salary compared to the control group.
When analyzing the impact of this drought on income from industry, construction, trade, and services, the same control variables of the income from salary was selected (SALES, INFR, CAP, DOCTOR and STUDENT). After applying SCM, the trend analysis shows that drought increased the income from industry, construction, trade, and services (Fig. 5d). However, the results were also not statistically significant after permutation tests. This can be explained by the fact that reconstruction has been done after the disaster. Therefore, people’s income from these activities would be increased but only existed a few years after the drought occurred.
In this study, the SCM is applied for comparison of the control and the treatment groups to understand the effect of an extreme drought on per capita income in Vietnam. The treatment group was Phu Yen Province and 29 provinces selected from 63 provinces in Vietnam was included in the control group. The results showed that the drought reduced short-term per capita income. The decrease in per capita income caused by drought was estimated at VND 160,000 per month. This impact only existed within 3 years after the disaster occurred. When analyzing the components of income in more details, this extreme drought had no impact on income from wages nor income from industry, construction, trade, and services. However, it reduced the income from agriculture, forestry, and fishery in the long term. Per capita income from agriculture, forestry, and fishery due to the drought 2013 was reduced at VND 220,000 per month, equivalent to 11% of the per capita income of Phu Yen and lasted up to 6 years after tit occurred.
Therefore, the drought has a major impact on people whose income are mainly from agriculture, forests, and aquaculture. Therefore, policy makers and relief organizations should prioritize the affected people from agriculture, forestry, and fishery to avoid inefficient and unfair relief operations. Relief activities should not stop at the year when a disaster occurs but should be extended up to several years after that. Communities do not prepare to prevent for a drought as they do not think the event will last longer than expected. Therefore, to reduce the damage caused by drought and other natural disasters, localities need to raise peoplés awareness by propagating the impacts of natural disasters and the measures to mitigate those impacts. Authorities need to provide early and accurate warnings to help people to be more proactive in disaster prevention.
The authors receive support from the National Economics University, Vietnam.
The data presented in this study are available in article.