The effect of economic variables on natural disasters and the impact of disasters on economic variables.

This study aims to study how disaster losses are affected by unemployment and poverty, then how disasters also cause income inequality in Indonesia and the causal relationship between disaster losses and income inequality. To determine the structural relationship between economic variables and disaster losses, the Structural Equation Model-Partial Least Squares (SEM - PLS) approach is used. This approach is an approach previously not found in economic and environmental studies. This study uses secondary data consisting of 30 years 1990-2019 collected from the territory of Indonesia. The results of the study found that unemployment and poverty variables had a significant effect on the disaster loss variable. The disaster loss variable has a significant effect on the income inequality variable. The income inequality variable has no effect on the disaster loss variable. Research is expected to contribute to the study of the impact of economic development and the environment.

Introduction

The study of whether natural disasters are a real obstacle to the growth and economic development of a country is a study that is being carried out by many experts, considering that natural disasters are unpredictable events but have a real impact on the economy. A disaster in an area will have an impact on economic losses, among others, in the form of infrastructure damage in the area where the disaster occurred (Hidalgo and Baez, 2019). Studies on disasters conducted by many experts have found that disasters with the category of disasters that occur suddenly (hurricanes, earthquakes, floods) will damage productive capital and infrastructure. Different things happen to disasters categorized as disasters that occur slowly (drought and floods) where these disasters have a wider and long-term impact.

Climate change and its impacts are of concern to researchers in economics and environmental economics today. Because they are increasingly aware of the significant impacts of disasters and climate change on human and economic development. Furthermore, it is also suspected that the disaster losses that occur will also have an impact on the level of income distribution and poverty (Groeschl, 2020). The same condition is supported by the results of research conducted in Pakistan that unemployment and poverty have a significant relationship with disasters and income inequality. This study also proves the occurrence of the EKC hypothesis in the study area (Hassan et al., 2015). Recommendations in environmental economics studies state that fighting reduced production and rising unemployment targets low-carbon investment, will be able to create up to 5 million 'green collar' jobs and revive the economy (Dietz and Maddison, 2009). This proves that the green economy can reduce unemployment. Poor unemployment will result in high income inequality as well, this relationship is exacerbated by disasters. Studies that examine the effect of unemployment on disasters are still quite difficult to find, where most studies are still on the impact of disasters on economic variables, not the other way around.

Furthermore, in the study of the impact of other economic variables on disasters we can refer to research conducted by (Murad. and Musthapa, 2010), In a study conducted in Kuala Lumpur Malaysia, it was found that there was a relationship between poverty and disaster. Poverty increases the risk of the severity of disaster impacts borne by the poor (Rozenberg and Hallegatte, 2015) and (Hallegatte et al., 2020) In more detail, it is also explained that poverty occurs as a result of income inequality. However, the results of his research reject the hypothesis that proves that the poor cause environmental disasters, but poverty is the impact of the disaster. Poverty that occurs in an economy will exacerbate income inequality in society. Poverty affects disaster losses also found by (Abbas Khan et al., 2019) (Padli et al., 2019) (Markhvida et al., 2020). On the other hand, there are also research results which find that inequality will also affect poverty levels (Hallegatte et al., 2020). So, this is where the role of the environmental economy concept comes to provide a solution. Based on scientific studies conducted with the case of 92 countries in the world, it was found that environmental conditions without climate change were able to encourage a reduction in the level of poverty. The disaster will cause an increase in food prices due to a decline in agricultural production in Africa and South Asia and a deterioration in the quality of health in all regions (Martin et al., 2020).

Disaster recovery damage costs accounted for 2.44% of people's income, while their annual income was reduced by 21.49%. This causes the economic capacity of the economically vulnerable lower class to decline (Songwathana, 2018).

A further impact of the disaster-economic relationship, namely the relationship between environmental damage, natural disasters and income inequality, has been described by (Kuznets, 1955). In the long-term study model of economic variables and climate change, it is found that climate change also has a significant effect on the distribution of opinions and the same is explained by (Piontek et al., 2019). This was also found by (Peter and Aung, Lwin, 2018) and (Hassan et al., 2015), who stated that the disaster increased economic inequality in the affected areas. Economic conditions in developing countries that are less stable and good make developing countries very vulnerable to the effects of climate change which in the end, if not properly anticipated by these countries, will lead to a disaster. However, most developing countries are often unable to implement sound adaptation strategies to reduce the negative impacts of climate change (Al-Amin et al., 2019). Disasters due to climate change create a food crisis problem in which developing countries are very vulnerable to this, due to the lack of technology in these countries. In a study conducted by (Tseliosa and Tompkins, 2018), it was found that the relationship between the total impact of disasters and income resembles an inverted 'U' letter. Furthermore, this study found that disaster impact conditions would differ between regions. The magnitude of the impact of a disaster in an area is determined by the socio-economic conditions of the area. Poor countries will suffer more from disasters than developed countries.

A disaster that occurs in a country will cause a decrease in productivity, which in turn will cause a decline in economic growth in that country. In fact, the study conducted found that flooding reduced the GDP growth rate per capita by 0.005% for every thousands of every million people affected. Then the disaster that occurs will have an irreversible impact on people's livelihoods (Shabnam, 2014). The findings of disaster research in Japan reinforce the impact of losses due to disasters, where in this study it is found that disasters will significantly damage the agricultural production sector, fisheries and loss of capital stock in the manufacturing industry. Then the disaster will cause a problem of considerable economic loss (Okiyama, 2017). Then the disaster will ultimately lead to a higher level of income inequality. The same is the case with the snow disaster that occurred in the Qinghai-Tibet Plateau which caused serious losses to the livelihoods of the rural population. In more detail, it is also found that this condition occurs due to losses in the form of loss and damage to financial capital, physical capital, human capital, social capital and natural capital (Fang et al., 2018).

Disasters that have an impact on worsening conditions of poverty and inequality were also found in the Vietnam study (Priaga, 2008) and (Bui et al., 2014). Disasters that occur will not only cause economic losses in the form of increased poverty but also change the socio-economic characteristics of the community affected by the disaster (Priaga, 2008) and (Bui et al., 2014). So that disasters can be understood to lead to higher income inequality and the level of welfare. The same thing was found in research conducted in Vietnam which found that the losses incurred were in the form of poverty and inequality of expenditure. Thus, disaster studies are something that must be considered in making economic development policies in the form of reducing poverty and equalization of income.

Indonesia as a region with a disaster-prone topography is exacerbated by the risk of climate change. As a result, routine natural disasters such as flash floods and landslides have become a serious threat to various regions in Indonesia. The Indonesian region experienced more than 1,800 disasters in 2005–2015. Of the total disasters that occurred, 78% (11,648) were hydrometeorological disasters and 22% (3,810) were geological disasters. Disasters with the category of hydrometeorological disasters were in the form of floods, extreme waves, forest and land fires, drought, and extreme weather. Types of hydrometeorological disasters are closely related to human involvement in the use of natural resources. Meanwhile, geological disasters are earthquakes, tsunamis, volcanic eruptions, and landslides. Climate change is increasing the impact of disasters that occur from hydrometeorological disasters. The high frequency of hydrometeorological disasters will have an even bigger impact on the economic sector of the community (BNPD, 2017).

Therefore, the causes and consequences of disasters must be considered as factors in designing policies to achieve development goals. However, from the literature review conducted, it can be concluded that studies that attempt to examine how disasters are both cause and effect variables of economic variables in structural relationships have not been found, especially for studies in Indonesia which are disaster-prone countries and economically have relatively unstable economic indicators. Existing literature review still looks at the relationship between the influence of the variable disaster partially, there is no study that tries to conduct a study with a structural variable relationship model. While economic variables are systemic variables, the impact on a variable will have an impact on other variables, and even this impact can be a causal relationship. So this research tries to answer the gaps in the study, so it is hoped that it will be known how unemployment and poverty affect disasters. Then how will this disaster strengthen the level of income inequality that occurs in the community and vice versa, how does the inequality affect the level of disaster losses. This model study uses a structural equation modeling (SEM) approach which has never been found in economic and environmental studies using the SEM approach. In general, economic and environmental literature uses a regression approach, it is hoped that this approach will become a new approach that will answer structural problems in economic and environmental problems.

Adequate studies on disasters and socio-economic indicators have been carried out, but there is a lack of research on how the disaster loss variable is able to mediate the effect of macroeconomic indicator variables on other macroeconomic variables. So, this study aims to fill the gap in existing environmental economics research.

The model in this study is an extension of the green Solow model. This study wanted to determine the effect of unemployment and poverty variables on income inequality through the intervening variable of economic losses due to disasters. From this model it can also be seen how unemployment and poverty affect each other. The analysis approach is carried out using a structural equation modeling (SEM) model. This approach is a new way of studying that includes the disaster loss variable as the intervening variable of unemployment and poverty.

This study aims to build a model of the influence of macroeconomics on other macroeconomics mediated by economic losses due to disasters. This model is expected to be the basis for developing a theory of the impact of disasters on the economic condition of the community. At the application level, this model is expected to be an evaluation model to see the effectiveness of disaster management and formulate strategic policies for communities affected by disasters.

Literature review

The study of economic and environmental growth models has been developed by Solow in the economic growth model. This model is built with the assumption that each production process also has negative externalities due to environmental damage that occurs during the production process carried out in economic development. This condition results in part of the production output must also be allocated for environmental improvement, so that not all resources or inputs can be allocated for production, this causes a decrease in economic growth (Brock and Taylor, 2005).

Every production process which produces output also produces environmental impacts. This condition can be quantitatively described by the following equation:where, in Eq. (1) E is a net degradation, which is the difference in the value of pollution resulting from the process of creating output in the economy and the effort to neutralize pollution. Every output causes environmental degradation. It means pollution is a function of output, so

In Eq. (2) pollution is a production function in an economy. The assumption in Eq. (2) is that there is a positive relationship between the level of production and the impact of a disaster. E is net degradation or a proxy for the level of the disaster that occurred. From the value of E, it can be understood that pollution is part of a disaster which will result in a decrease in the output received by the community because it must be allocated for environmental improvement. A decrease in resource capacity and a decrease in productivity can occur as a result of a disaster. Disasters will result in economic losses in the form of assets and community income. The severity of the disaster will differ based on the socio-economic conditions of the affected community. The poor will experience higher income losses. Poor households have low disaster preparedness. The poor who live in disaster-prone areas with income derived from the traditional agricultural sector which depends on the weather result in the poor people experiencing higher income losses. This condition will result in higher levels of poverty and income inequality.

The disaster that occurs has a strong relationship with the economic conditions of the affected community. Disasters have a negative impact on people's incomes and decrease the level of macroeconomic indicators (Vale and Campanella, 2005). The severity of the economic impact experienced by the community will be determined by the economic condition of the community before a disaster occurs (Songwathana, 2018). The damage to productive capacity that occurs as a result of a disaster will of course have a negative impact on regional economic growth, employment, poverty and income distribution.

The productive capacity damage could negatively lead to regional economic growth, employment, poverty, and income distribution In (Peek, 2010) reports, the results of several previous studies found that the risk of disaster impact was also determined by the income of the affected community, poorer people and lower income were considered more at risk. The research results are in line with research by Adeagbo (Adeagbo et al., 2016) which examines the impact of natural disasters on household livelihoods, their assets and other aspects of welfare in Nigeria. This study found that natural disasters affect household economic conditions and welfare.

Environmental damage is a result of the socio-economic conditions of the community. It was found that the level of disaster losses that occurred was related to the social and economic conditions of the community, the better the socio-economic conditions of the community the lower the level of economic losses that occurred (C.E Haque, 2003) the same thing is also confirmed by the findings (Padli et al., 2018). The same study also found that (Izevbuwa and Adeolu, 2015), stated that half of the population of Nigerians who experienced floods had lost up to 79% of their main income and what was worse were farmers who had lost all their income, while they only received compensation from government about 13% of all losses. As a result of the flood disaster, there was a very significant reduction in the rate of economic growth and of course in the end it led to income inequality.

However, it was also found that the severity of this disaster was relatively lower if the economic conditions of the community were worse, but if the economic conditions were better, the damage caused by the disaster would be even higher (Padli, J., & Habibullah, 2009). and (Fournier Gabela and Sarmiento, 2020). This model is a development of the basic EKC model, the Environmental Kuznet Curve, where economic growth has a potential relationship with environmental damage, while the EKC model can be seen in the following equation:where I and t are the dimensions of place and time, and ED is environmental damage and GDP is the level of regional income, GDP2 is the optimum point in GDP which has the maximum impact on environmental damage.

There is a reciprocal relationship between environmental damage and income inequality (Stiglitz, 2013). Disasters will result in the loss of economic capacity and employment, this will exacerbate the income inequality of the community (Saliminezhad et al., 2021). Environmental damage can cause inequality, on the other hand, inequality can cause environmental damage. Where the poor tend not to care about the environment, because they are more focused on meeting the needs of life, as a result their consumption does not pay attention to environmental damage

More specifically, research on a vicious circle between poverty and disaster loss finds that: poverty is the main driver of people's vulnerability to natural disasters (Hallegatte et al., 2017). Disasters are not only caused by socio-economic factors., but also socio-economic conditions that depend heavily on regional economic conditions (Li et al., 2020), then based on by model (Padli et al., 2018), shows socio-economic impact of disaster variables as follows:where.

i indicates countries 1, 2, 3, … n,

j shows the type of natural disaster and ejit is error term.

ND is a catastrophic loss.

As for the regressors, RGDPCit is GDP per capita;

POPit is total population,

POPDENit is population density,

UNEMPit is the unemployment rate,

RINVit is the ratio of real investment to GDP,

RGCONit is real government consumption as a percentage of GDP,

OPENit is openness,

EDUit is education level.

CORit is corruption.

So, this research model is built with the hypothesis that disasters have a direct impact on poverty, in addition, income will also determine the magnitude of the impact of disasters on poverty that occurs.

Method

The model in this study is an extension of the green solow model. The model is made from production theory and tests it with the impact of economic variables in the form of unemployment and poverty on the level of disaster losses and the impact of disaster losses on income inequality, using the structural equation modeling (SEM) model. This approach is a new perspective in looking at the causal relationship of disaster losses with economic variables. Through the SEM method used, it can be seen that the direct and indirect relationships of unemployment and poverty to inequality with disaster losses as an intervening variable. This model is a new study in disaster loss studies.

This study aims to build a model of the economic impact variable on disasters and the disaster impact variable on the economic variable. This model is expected to be the basis for developing a theory of the impact of disasters on the economic conditions of the community. The results of this study are expected to be useful as study material to develop a theory of the impact of disasters on the socio-economic impacts of society. At the application level, this model can become an evaluation model to see the effectiveness of disaster management and formulate strategic policies for disaster-affected communities.

This research uses secondary data and literature studies from related agencies such as the National Disaster Management Agency (BNPB), the Environment Agency (KLH) and the Central Statistics Agency (BPS). The sample in this study includes Indonesian disaster data as a unit of analysis for 30 years, 1990–2019. Disaster loss data used in this study is the total economic loss in Indonesia, which includes disasters in units of Rp/year. In this study, the data used is limited to disasters that occur due to human interference in activities utilizing natural resources. The disasters analyzed are limited to flood, fire, and landslides. Meanwhile, disasters due to natural processes such as earthquakes and volcanic eruptions are not discussed in this study. Unemployment, poverty and income inequality data used in this study are the concepts of poverty used by the Indonesian statistical center (BPS, 2020). The unemployment data used is the number of people categorized as open unemployment, in units of people/year. Open unemployment is defined as the large number of the workforce who are not working and are actively looking for work. Poor people are people who have an average monthly expenditure per capita below the poverty line. The poverty data referred to is the number of poor people in people/year (BPS, 2020). Next, the income inequality variable used is the Gini ratio coefficient, the Gini coefficient is based on the Lorenz curve, which is a cumulative expenditure curve that compares the distribution of a certain variable (for example income) with a uniform distribution that represents the cumulative percentage of the population (BPS, 2020).

Develop SEM for Based on the literature study conducted, it can be derived a research model that will be analyzed in this study by using the Structural Equation Model The approach using SEM is appropriate to be used in this study. Through the SEM method, it can be obtained the relationship between the independent variable and the dependent variable that is modeled simultaneously, resulting in a single, systematic, and comprehensive analysis that can evaluate the model (Gefen D, Rigdon E, StraubDGefenD, Rigdon E, 2011). The SEM method is superior to linear regression and multivariate regression methods (C, 1987). In addition, SEM method has greater flexibility for researchers to relate theory to data. SEM method can determine the validity of a theory even though it is supported by new and minimal theoretical concepts by examining existing empirical data (EWL, 2001). This study predicts a direct relationship between disaster loss and poverty variables and examines the indirect relationship between disaster losses and the income variable as an intervening. The SEM method provides output in the form of a prediction of the relationship between the analyzed variables. SEM is considered suitable in this study, because it aims to build a new model in assessing the causes and effects of disaster losses by economic variable.

SEM modeling is a statistical technique that allows testing a series of relatively complex relationships simultaneously. Complex relationships can be built between one or more dependent variables with one or more independent variables. The stages in SEM modeling and structural equation analysis include the following stages: (see Figures 1, 2, 3).

Figure 1

The path diagram of the variable causality relationship.

Figure 2

Results of the structural equation research model analysis.

Figure 3

Modified SEM model.

Development of a path diagram to show the causality relationship can be seen in the image below:

The figure above shows the conceptual relationship between the variables built in the research to be carried out, where it appears that unemployment and poverty directly affect disaster losses and are indirectly related to income inequality through the intervening variable disaster losses.

Convert flowcharts into a series of structural equations and model specifications measurement.

By elaborating on Eqs. (4) and (5) above, the research models to be analyzed in this study are:

Structural model

where

loss of disaster = unemployment + .

Income inequality = income + .

This research model consists of two research models, where the first model shows a variable model that affects loss of disaster consisting of unemployment and poverty. Meanwhile, the second model shows the effect of income inequality on loss of disaster.

Based on the theoretical studies that have been carried out, the research hypotheses in this study were built as follows:

H1: Unemployment affects disaster losses

H2: Poverty affects disaster losses

H3: Disaster losses affect income inequality

H4: Income inequality affects disaster losses

Assess the identification problem

At this stage, an analysis of the assumptions that must be fulfilled in the data collection and processing procedure is analyzed using SEM modeling as follows:

Normality and Linearity

The data distribution must be analyzed to see whether the normality assumption is fulfilled so that the data can be processed further for this SEM modeling.

Outliers

Outliers are observations that arise with extreme values both univariate and multivariate, namely those that arise because of a combination of unique characteristics that they have and look very much different from other observations. The detection of multivariate outliers is carried out by considering the value of mahalonobis distance. An observation is stated as an outlier if it has a significant distance from the center of observation at a significance level of p < 0.001 with the degrees of freedom of a number of constructs used in the study. In AMOS tools, calculating the value of the mahalanobis distance produces p1 and p2 values. A data includes outliers if the p1 and p2 values are less than 0.05. To perform SEM analysis, data outliers must be removed first (Ghozali, 2014).

Multicollinearity and Singularity

Multicollinearity can be detected from the determinant of the covariance matrix. The determinant value of the covariance matrix is very small (extremely small), indicating a multicollinearity or singularity problem. In general, SEM computer programs provide a "warning" facility whenever there is an indication of multicollinearity or singularity.

After looking at the SEM assumptions, the next thing is to determine the criteria that will be used to evaluate the model and the effects shown in the model. Some of the suitability indices and their cut-off values are used in testing whether a model can be accepted or rejected as described below.

Chi-Square Statistic (λ2)

Chi-square statistic is the most fundamental test tool for measuring overall fit. This chi-square is very sensitive to the size of the sample used. The model being tested will be considered good or satisfactory if the chi-square value is low. The smaller the λ2 value the better the model because in the chi-square difference test, λ2 = 0, it means that there is really no difference (Ho is accepted) based on probability with a cut off value of p > 0.05 or> 0.10 (Ferdinand, 2002).

RMSEA (The Root Mean Square Error of Approximation)

The RMSEA is an index that can be used to compensate for the chi-square statistic in a large sample. The RMSEA value indicates the expected goodness-of-fit if the model is estimated in the population. The RMSEA value which is less than or equal to 0.08 is an index for the acceptance of the model which shows a close fit of the model based on the degrees of freedom.

GFI (Goodness of Fit Index)

This fit index will calculate the weighted proportion of the variance in the sample covariance matrix described by the estimated population covariance matrix (Ferdinand, 2002). GFI is a non-statistical measure that ranges from 0 (poor fit) to 1.0 (perfect fit). High scores on this index indicate a "better fit".

AGFI (Adjusted Goodness-of-Fit Index)

GFI is the analogue of R2 in multiple regression. The Fit Index is adjusted to the available degrees of freedom to test whether the model is accepted or not.

A value of 0.95 can be interpreted as a good level of good overall model fit, while the value of 0.90–0.95 indicates an adequate fit

CMIN/DF

This fit index is the minimum sample discrepancy function (CMIN) divided by the degree of freedom which will produce the CMIN/DF index. Generally, researchers report it as an indicator to measure the fit level of a model. In this case CMIN/DF is nothing but the chi-square statistic, c2 divided by its DF so that it is called the relative chi square. The relative value of c2 is less than 2.0 or sometimes even less than 3.0, indicating that the model and data fit according to Arbuckle.

TLI (Tucker Lewis 60 Index)

TLI is an alternative incremental fit index that compares a tested model against a baseline model. The recommended value as a reference for acceptance of a model is acceptance> 0.95 and a value very close to 1 indicates a very good fit.

CFI (Comparative Fit Index)

This index has a range of values between 0 and 1. Getting closer to 1, indicating a very good fit. The recommended value is CFI> 0.94. This index is not affected by the sample size, because it is very good for measuring the level of acceptance of a model.

The result

Model evaluation

The model feasibility test consists of two stages of testing. The two stages of testing the model are as follows:

Test the validity of the measurement model

Several types of model feasibility tests that can be done to test the goodness of fit (GOF) and the indices used for the model feasibility test and the cut-off values that must be met include Chi Square, significancy probability, RMSEA, GFI, AGFI, CFI, TLI and Cmin/Df.

Test the validity of the structural model

The significance of a relationship between constructs can be seen from the critical value (Critical Ratio/CR) obtained from the estimation results. If the CR value is greater than 1.96, the hypothesis is accepted (Ghozali, 2014).

Normality

The normality test is performed using a critical ratio value of ±2.58 at a significance level of 0.01% (Ghozali, 2014). The results of the normality and linearity test of the research data are presented in the following table (Tables 1, 2, 3, 4, 5, 6).

Table 1

Research data normality test results.

Variable	min	max	skew	c.r.	kurtosis	c.r.
poverty	24.8	49.5	0.921	2.06	0.004	0.004
unemployment	1,951,664	10,932,000	-0.415	-0.928	-0.714	-0.799
Income_inequality	0.3	0.6	2.435	2.444	8.452	0.945
Lots_of_disasters	1	1920	0.832	1.861	-0.534	-0.597
Multivariate					14.8	0.585

Table 2

Outliers test results.

Observation number	Mahalanobis d-squared	p1	p2
6	25.794	0	0.001

Table 3

Research model test results.

Goodness of Fit (GOF) Index	Cut Off Value	Research result	Model Evaluation
Chi Square	Small	5.354	Small
Probability	≥0.05	0.069	Not good
GFI	≥0.9	0.922	Well
AGFI	≥0.9	0.611	Not good
TLI	≥0.95	0.521	Not good
CFI	≥0.95	0.840	Not good
RMSEA	≤0.080	0.240	Not good

Table 4

Research model test results.

Goodness of Fit (GOF) Index	Cut Off Value	Modified Research Model	Model Evaluation
Chi Square	small	0.265	small
Probability	≥0.05	0.607	well
GFI	≥0.9	0.995	well
AGFI	≥0.9	0.955	well
TLI	≥0.95	1.210	well
CFI	≥0.95	1.000	well
RMSEA	≤0.080	0.000	well

Table 5

Influence test results. (Regression weights-maximum likelihood estimates).

	C.R.	P
Lots_of_disasters ← Unemployment	3.148	0.002
Lots_of_disasters ← Poverty	-3.557	0.000
income_inequality ← Lots_of_disasters	2.688	0.007
Lots_of_disasters ← income_inequality	-1.762	0.078

Note: hypothesis is accepted: If P < 0.05 and CR value > 1.96.

Table 6

Total effect test.

	poverty	unemployment	income_inequality	Lots_of_disasters
Income_inequality	-0.463	0.378	-0.332	0.534
Lots_of_disasters	-0.579	0.473	-0.415	-0.332

Based on the results of the data normality test, it appears that all indicators have a c.r. value not exceeding ±2.58 so that the requirements for the normality of the research data are met.

Outliers

The research data is said to have outliers if the p1 and p2 values are less than 5% and the data containing outliers can affect the normality of a data. The results of the outlier test on the research data found that there were 2 data containing outliers, namely the 6th data because all the p1 values of the data were <0.05. To obtain normal research data, all data containing outliers must be excluded for further SEM analysis. The p1 and p2 values from data containing outliers can be seen in the following table:

Multicollinearity and Singularity

The existence of multicollinearity and singularity can be seen through the determinant value of the covariance matrix which is really small or close to zero. For the analyzed research data, it is found that the determinant value of the covariance matrix is 1.28263E + 13 which indicates that the determinant value of the covariance matrix is far from zero and it can be said that the research data used does not have multicollinearity and singularity so it is feasible to use.

Model interpretation and modification.

After conducting the SEM analysis prerequisite test, the results of the analysis of the data used in this study produce a SEM model which can be seen in the following figure.

Based on the research model and the results of previous analyzes, the comparison of research results based on the Goodness of Fit (GOF) criteria required for model suitability analysis includes the following:

Some indexes still do not meet the GOF (Probability, GFI, AGFI, TLI, CFI and RMSEA, so the model must be modified so that all indices meet the specified criteria.

Model modification was carried out according to software modification suggestions (modification indices), namely connecting several constructs with covariance lines so that there was a relationship between constructs. After modification, the research model can be seen in the following figure:

Based on the modified research model, the comparison of research results based on the Goodness of Fit (GOF) criteria required for the suitability analysis of the modified model is as follows:

Hypothesis test

Influence Test

After the research model has met the GOF criteria, the next step is to test whether the proposed hypothesis is accepted or rejected. Hypothesis testing is carried out by observing the CR and Sig values of the studied variables based on the maximum likelihood estimates by looking at the regression weights table, which is said to have a significant effect if the CR value of the variable is ≥ 1.96 and the probability is < 0.001. The results of the influence test between the variables tested based on the research model are shown in the following table:

Based on the table above, the hypothesis test results are as follows

The Unemployment variable has a significant effect on the Lots of Disaster variable because the value of CR (3.148) ≥ 1.96 and probability (0.002) <0.05 so hypothesis H1; is accepted.

The Poverty variable has a significant influence on the Lots of Disaster variable because the CR value (3.557) ≥ 1.96 and the probability (0.000) <0.05 so hypothesis H2; is accepted.

The Lots of Disaster variable has a significant effect on the Income inequality variable because the CR value (2.688) is ≥1.96 and the probability (0.007) <0.05 so hypothesis H3; is accepted.

Income inequality variable does not have a significant effect on the Lots of disaster variable because the CR value (1.762) <1.96 and probability (0.078)> 0.05 so hypothesis H4; is reject

Total Securities

To find out the magnitude of the effect caused by the variables on other variables, the total efficiency test is carried out with the results according to the following table:

Based on the results of the total effect test above, the following can be stated:

The Lots of Disaster variable is the variable with the largest contribution in influencing the Income inequality variable, compared to other variables such as poverty and unemployment.

Unemployment variable is the variable with the largest contribution in giving effect to the Lots disaster variable, compared to other variables such as poverty and income unemployment.

Discussion

From the results above, it can be seen that the unemployment variable has a significant effect on the Lots of Disasters variable. Policies related to employment must be a comprehensive part of sustainable economic development. The development of the creative economy sector that provides space for the community to participate with low capital capacity and expertise can be a solution to reduce unemployment and dependence on natural resources.

The Poverty variable has a significant effect on the Lots of Disasters variable. The community must have better economic resilience, so that the effects of disasters that occur in disaster-prone areas can be reduced. The government needs to build a mechanism for strengthening the community's economy in disaster-prone areas. Good disaster mitigation is a good alternative in reducing the impact of poverty on disasters. Communities must be aware of the risks that make them economically worse off in the event of a disaster and must have the ability to anticipate it. This can be done with the disaster mitigation process programmed by the government.

Furthermore, the variable of disasters has a significant effect on the Income inequality variable. However, the income inequality variable does not have a significant effect on the lots of disaster variable. This impact is more pronounced in groups with income levels that were previously at a lower level or already below the poverty line, thus worsening their economic capacity and exacerbating their poverty level. This finding contributes to a component that needs to be taken into account in formulating strategic policies for society affected by the disaster.

Damage to resources occurs, a decrease in resource productivity will result in a decrease in the level of income and this in turn will increase the income gap.

The government must pay more attention and preventive measures to vulnerable areas by maintaining community resilience to disasters through the soft skills of the population, education and skills as well as mastery of technology. This is expected to be able to make people who are vulnerable to disasters more prepared to face the impacts of these disasters.

The government must have a program with an initial indicator that every rehabilitation program for disaster victims has been able to return to having economic capabilities such as conditions. This program becomes a measuring tool for evaluating the rehabilitation of post-disaster communities.

The results of this study can be used as material for study and evaluation to see the effectiveness of disaster management in assessing policies for empowering communities affected by disasters. Environmental impact analysis is needed in the decision-making process.

Conclusion and policy implications

Natural disasters have a huge impact on the economy and people's lives. The loss of natural disasters is increasingly felt by the poor. Because of natural disasters they do not have the ability and strength to survive that are economically good and will suffer more economically, lose their jobs and lose their source of income. For this reason, the government needs to make careful planning how to deal with the impact of disaster losses. The government is expected to have a clear and measurable program so that the impact on society can be minimized. For example, a policy in the form of an insurance scheme with government funds is needed to ensure rapid recovery from the impact of natural disasters for all categories of society.

It is clear that the need for disaster prediction models is very important for life safety and environmental protection and the regional economy. The amount of disaster losses that occur is generally determined by the initial prediction and disaster mitigation. The absence of a disaster prediction and warning system has generated panic and concern among residents as well as economic and property losses.

In reducing losses due to disasters, policies for empowering local cultural wisdom in the form of how the community can survive the impact of a disaster based on the cultural values of the local community that already exist are recommended policy alternatives. So it is necessary to develop local wisdom into something that is institutionalized and sustainable. Local cultural wealth in the form of local wisdom is expected to be a disaster mitigation force to reduce losses due to disasters.

The culture that is owned by the local community, such as the culture of resource use accompanied by natural conservation measures and the ability to read natural signs as signs of a disaster are capital for disaster mitigation An empirical study found that households that have adopted a climate risk management strategy have a higher income level and experience a lower impact on the risk of loss due to disasters. Disaster mitigation needs to adapt from traditional knowledge into planned and measurable and sustainable actions. In disaster mitigation management, poverty alleviation should also be part of disaster risk management, and disaster risk management can be considered as poverty alleviation.

Indonesian people have many local cultural instruments that can be economic resilience. The concept of a prohibited area that limits each resident to exploit a certain forest or area. The exploitation of these natural resources can be carried out if a permit has been obtained and agreed to by the indigenous community concerned.

Likewise, the ownership of land attached to the indigenous community of the local community which contains land cannot be transferred to ownership and transfer of function as desired by the land lord. Because land is the right of indigenous communities which will eventually become inheritance rights for the next generation.

The strength of local culture like this should be institutionalized and become a force in disaster mitigation.

Likewise, the effectiveness of the direct cash transfer program must have the ability not only to maintain people's purchasing power but also to increase the community's economic resilience to disasters.

Declarations

Author contribution statement

Evi Susanti tasri: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Kasman Karimi and Irwan Muslim: Performed the experiments; Contributed reagents, materials, analysis tools or data.

Funding statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement

Data associated with this study has been deposited at BPS (2020) Statistics, Central Bureau of, BPS - (Statistics Indonesia). Available at: https://www.bps.go.id/istilah/index.html?Istilah%5Bkatacarian%5D=pengangguran&yt0=Tampilkan (Accessed: 5 October 2020). BNPD (2017) Disaster Events in the Last 10 Years, BNPD. Available at: http://dibi.bnpb.go.id/(Accessed: 20 September 2019).

Declaration of interests statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.