The Nexus between Economic Growth and Natural Resource Abundance in Selected ASEAN countries before Pandemic Covid-19

Abundant natural resources play an important role in boosting economic growth. However, as the Dutch disease emerged in the 1970s, innumerable researchers were attracted to investigate the negative impact of abundant natural resources on economic growth. The issue has been rigorously discussed in a large number of previous studies. However, they did not investigate the effect of economic growth on natural resource abundance. Therefore, this study embarks on an investigation into the effect of economic growth on natural resource abundance in selected ASEAN countries, namely Thailand, Singapore, Malaysia, Brunei, the Philippines, Vietnam, Cambodia and Indonesia. This study employs the static panel method to analyse data ranging from 2000 to 2016. The results show that foreign direct investment and direct investment can have negative impacts on natural resource abundance. An increase in economic growth, on the other hand, can have a positive impact of natural resource abundance. Other factors such as financial development, trade openness and governance do not influence natural resource abundance. Therefore, these findings can shed light for policymakers to formulate policies for future references as the outcomes based on before the spread of Covid-19 diseases. The countries can enhance economic growth to pave the way for more exploration of natural resources. Higher economic growth can lead to more advanced technologies and thus it is easy to extract more natural resources. Foreign direct investment and direct investment should be controlled to ensure that natural resources will not be exhausted.


INTRODUCTION
Natural resources, especially oil, gas and coal, are of utmost importance in generating economic activities. Therefore, countries with abundant natural resources can help boost economic activities and thus there is no fear of exhaustion of natural resources, such as oil, that can disrupt economic growth (Rahman et al., 2018). Brunnschweiler (2008) and Fan et al. (2012) found that abundant resources can positively affect economic growth. Countries with abundant resources have a head start over other countries exporting their resources at low prices and thus they are more competitive in the global market. Hence, higher exports ensue, leading to higher standards of living. Natural resources should be treated to be on a par with inputs such as capital and labour. Inputs act as a catalyst for economic growth. In the absence of a higher number of inputs, economic growth can be dampened. This implies that natural resource depletion can serve as a stumbling block to economic development. Nasiru et al. (2019), Shaari et al. (2012), Ighodaro (2010) stated that natural resources such as gas and oil can spur the economy to faster growth. Besides, managing effectively the country's resources would help the country to achieve sustainable economic development goal [Vija Kumaran et al. (2020); Ridzuan et al. (2019), Ridzuan et al. (2018), Ridzuan et al. (2017)].
Countries with abundant natural resources can be independent as there is no need for them to import. Due to natural resource dependence, they can avail themselves of the opportunity to export. Therefore, lower imports due to less independence and higher exports attributed to natural resource dependence can lead to higher economic growth. However, it is not always tenable as it is dependent on how they manage their resources. If they really utilise their resources, they can reap the benefit. Thus, job opportunities can be created, and higher standards of living ensue. However, if they do not manage their resources efficiently and effectively, a waste of resources occurs and thus its economic growth remains unchanged.
Alarm bells started to ring when the issue of the resource curse emerged. The Dutch disease occurred in the Netherlands during the 1970s and 1980s in the aftermath of a great decline in its exports (Zhang and Brouwer, 2019). Despite the country's dependence on natural resources, particularly gas, the country plunged into recession due to the fact that its exports slumped. The increase in the price of the resource due to its higher exchange rate caused the product to be less competitive in the global market. Due to the higher exchange rate, the prices of other Dutch products were higher and thus it affected the manufacturing industry. This suggests that abundant natural resources do not always suggest higher economic growth as it can also reduce economic growth. Unemployment will be on the rise stemming from abundant natural resources. The country's heavy reliance on exports has fallen prey to global crises that lead to a dramatic plunge in the global demand for natural resources. A slump in the price of natural resources will have a deleterious effect on exports. Countries without abundant resources such as Japan and Korea experience great economic growth while the economies of countries with abundant resources grow slowly. That is the reason why Saudi Arabia has also inevitably experienced the natural resource curse.
A vast array of previous literature has delved into the effects of abundant natural resources on economic growth (Daniele, 2011;Mittal and Gupta, 2017;Amini, 2018). However, no attention has been given to the other way round. Therefore, this study attempts to be the first to ascertain whether higher economic growth can help increase or decrese natural resources. Our justification for the exploration is that countries with higher economic growth have advanced technologies (Zhou and Luo, 2018) and thus they can discover more natural resources such as oil, gas and coal. Hence the reason their exports escalate. However, at a certain point where there is higher economic growth that requires more natural resources, natural resources will be exhausted. This means that higher economic growth can contribute to reducing natural resources. We extract non-renewable resources in gas, liquid or solid forms, then we convert them to final energy to be used to generate economic activities especially in the transportation and industrial sectors. It takes billions of years for natural resources, particularly fossil fuels to form and therefore, it takes billions of years to replace the resources that have been used.
This study focuses on selected ASEAN countries, namely Malaysia, Indonesia, Thailand, Brunei, the Philippines, Vietnam, Cambodia and Singapore. It is important to examine the effect of economic growth on natural resource dependence in the countries due to their varied patterns of natural resources. For example, Malaysia and Indonesia contibuted 85% of total world palm oil production (Fathana, 2017). The advent of new technology due to higher economic growth can help these countries boost their palm oil production. Figure 1 shows total natural resources rents as a percentage of GDP over a 6-year period. Total natural resources rents in Brunei exhibited a slumping trend from 2011 to 2016. Malaysia and Cambodia recorded steady decreases in total natural resources rents as a share of GDP. Despite great economic development in Singapore, the country still ranked bottom in natural resources rents as a share of GDP. Notwithstanding a small country, Brunei ranked top in 2016 (14.72%), followed by Malaysia (5.51%), Indonesia (3.06%), Vietnam (2.62%), Cambodia (1.89%), Thailand (1.60%), and the Philippines (1.23%). Economic growth in all of the countries showed uncertain trends. Cambodia exhibited the strongest economic growth as its economic growth stood at higher than 7% over the period. Singapore is a developed country but could not beat Cambodia in terms of economic growth. Thailand's economy grew at a slow pace with average growth of 3.04.

LITERATURE REVIEW
Natural resources such as minerals, oil, forests, land, water and materials are assets for a country as they can have a significant contribution towards economic growth. Natural resources create numerous economic activities and job opportunities. Countries with an abundance of natural resource are likely to grow faster than those countries in the absence of abundant natural resources. Ramez et al. (2017). There are two keys to measuring countries' natural resource abundance: resource abundance of output (revenue from resources) and stock of resource abundance (amount of subsoil mineral, oil and gas) (Brunnschweiler and Bulte, 2008). Basically, there are some benefits of having natural resource abundance. One of the benefits is that it can generate income from the extraction of resources and this it can result in a higher standard of living. Other than that the income can be used to boost public and private consumption, to increase investment and to remove barriers of development due to lack of fiscal resources (Sachs, 2007). However, the overuse of natural resource especially non-renewable resources such as oil, gas and minerals can result in exhaustion. Sachs and Warner (1995) found there is an inverse relationship between natural resource abundance and economic growth in cross-country studies. Modern economists found the idea that there might be a strong negative relationship between natural resources, and economic and political development that prevailed in the 1980s (Sen, 1999). There are three essential reasons that contribute to the negative relationship, such as an appreciation in the real exchange rate that can lead to the Dutch disease, volatility in natural resource export revenues and the resource curse (Morisson, 2015). Meanwhile, Deng et al. (2014) provided four reasons that abundant natural resources can hinder economic development: Dutch disease effects, crowding effects (i.e., declining savings and investment, human capital under development, innovation declines), a deterioration in the institutional quality and volatility in resource trade. Ramez et al. (2017) highlighted five causal mechanisms of the natural resource curse i.e. the Dutch disease, a fluctuation in commodity prices, economic mismanagement, rent seeking, and corruption and institutional quality.
Dutch disease effects are a common term when discussing the relationship between natural resources and economic growth. The Dutch disease arose when abundant natural resources increased domestic income and the demand for goods, resulting in inflation and an appreciation in the real exchange rate (Sachs and Warner (1995), Gylfason, (2001),, Papyrakis andGerlagh (2004), Frankel (2010). The prices of non-resource commodities especially in the manufacturing sector become less competitive in the global market. Investment and human capital may be drawn away from the manufacturing sector to the natural resource-based sectors. However, if natural resources are not properly managed, it may affect economic performance, causing people to be worse off (Ahmad and Armida, 2006). Thus, the Dutch disease effect leads to the resource curse. According to Muhammad et al. (2019), the resource curse hypothesis refers to a situation where countries with a large amount of natural resources achieve a lower rate of economic growth, compared to those countries with a small amount of natural resources. Thus, the resource curse refers to an inverse relationship between natural resource and economic growth. However, the resource curse involves non-renewable natural resources (i.e., oil, mineral, gas).
Empirical evidence regarding to the resource curse can be categorised into three groups; resource abundance or resource dependence, various economic factors affected by natural resource wealth and the validity of the resource curse hypothesis (Ramez et al., 2017). Besides, the institutional quality and government intervention are important factors that can contribute to the resource curse. There are negative effects of natural resources on economic growth in countries with poor institutional quality (Mehlum et al., 2006). Governments misspend money from natural resource revenues to gain supporters can be detrimental to the economy. The lack of accountability can result in a shortage of public funds (Mehlum et al., 2006;Robinson et al., 2006). However, there are several ways to avoid the resource curse. curse: institutional quality, governance quality, effectiveness of governance, membership of extractive industries transparency initiative (EITI), natural resource-sustainability and accounting, accounting and petroleum revenue management, and accountability and corruption control mechanisms.
Recently, there are numerous studies that investigated the existence of the resource curse among countries. Table 1 summarises past studies that validated the resource curse hypothesis across countries. Most researchers used GDP or real GDP per capita as an indicator of economic growth and used almost similar explanatory variables. For explanatory variables, there are a few common variables to measure the resource curse such as resource rent, investment, international trade, financial development, government expenditure, employment or unemployment rate, education, natural resource abundance, population, institution quality and government quality.
Based on our observation, this paper is expected to generate new findings and contribute to the body of literature by investigating the presence of natural curse hypothesis in the case of selected ASEAN countries. The selection of the variables has been meticulously made by referring to previous studies. Therefore, we attempt to explore the determinants of natural resource abundance in selected ASEAN countries as it has yet to be explored by previous studies. Thus, it will help readers gain new knowledge.

Model Specification
In this study, a static panel analysis is used to deal with a small sample size for selected ASEAN countries over a period of 17 years ranging from 2000 to 2016. Our models are specified in linear and nonlinear forms. To estimate the non-linear relationship between independent and dependent variables, our model specification is extended by using the non-dynamic panel threshold model. The formulation of the models are briefly explained in this section.
Model 1: The impact of natural resource abundance on economic growth.
To investigate the effect of natural resource abundance on economic growth and to validate the natural resource curse hypothesis, a linear model in Model 1 is formulated as follows: where GDP represents GDP per capita or economic growth and NA represents total natural resource rents. It is the sum of rents from oil, natural gas, coal (hard and soft), minerals, and forests. Eq. (1) is transformed into an econometric model to estimate the relationship between natural resource abundance and economic growth. In order to achieve the consistency and reliability of the estimation by standardising the data scale, these variables are transformed into the logarithms (indicated by ln in Eq. [2]) to explain the results of long-run elasticity (Shahbaz, 2010). Eq., (1) is re-written as follows: Where δ 0 denotes the intercept term; β 1 is the undetermined coefficient; μ it is the error term; i represents countries (i = 1, 2,…, N) and t represents time (t = 1, 2,…, T). The model is extended by including a set of parameters for control variables such as domestic investment (DI), foreign direct investment (FDI), financial development (FD), openness to trade (TO), and governance (GOV). Some of these variables were introduced by previous studies such as Apergis and Payne (2014), , and . A new form of the equation is expressed as follows: lnGDP it =δ 0 +β 11 lnNA it +β 12 lnDI it +β 13 lnFDI it +β 14 lnFD it +β 15 lnTO it +β 16 GOV it +μ it (3) The coefficient of natural resource abundance,  11 ( ) β may be positive or negative depending on the existence of resources. The expected sign of the coefficient  12 ( ) β for domestic investment should yield a positive sign as founded by Barro (1991), and Levine and Renelt (1992). Meanwhile, the coefficient  13 ( ) β of FDI as estimated by Mankiw et al. (1992) and Obstfeld (1994) is positive. Both FDI as shares of GDP and domestic investment capture the capital accumulation requisite for economic growth. The coefficient  14 ( ) β of financial development is also positive as found by Beck and Levine (2004), and Bertocco (2008) β of openness to trade, (TO) reflects the impact of globalisation and it is expected to have a positive sign as addressed by Lee et al. (2004), Freund and Bolaky (2008) and Chang et al (2009). The expected sign related to governance (β 16 ) is positive as addressed by Costantini and Monni (2008), and Abou-Ali and Abdelfattah (2013).
Investment is structured under Keynesian economics as a determinant of economic growth. Looking at the domestic economy, domestic investment may depend on natural resources in a country to be utilised in order to generate economic growth. Suppose that the level of natural abundance may influence the performance of domestic investment on economic growth, the baseline model in Eq.
(3) is further extended by including the interaction term between natural abundance (NA) and domestic investment (DI) as follows: Where β 22 * denotes the interaction term between domestic investment and natural abundance. If the interaction term coefficient is significant, it means that natural abundance is a catalyst to influence the effect of domestic investment on economic growth. Therefore, a non-linear relationship exists in the model as specified in Eq. (4).
Model 2: The impact of economic growth on natural abundance.
In comparison with Model 1, Model 2 is introduced by specifying natural resource abundance as a dependent variable. Meanwhile economic growth is being treated as an independent variable and the others are control variables as incorporated in Model 1. The baseline model in a linear form is shown as follows:   lnNA it =α 0 +γ 11 lnGDP it +γ 12 lnDI it +γ 13 lnFDI it +γ 14 lnFD it +γ 15 lnTO it +γ 16 GOV it +ε it (5) The purpose of introducing the baseline model in Model 2 is to examine the nexus between natural resource abundance and economic growth in two directions. Besides, the consistency of the sign in the coefficients of all control variables (DI, FDI, FD, TO and GOV) in both Models 1 and 2 are used for a robustness check.
Assuming that the relationship between economic growth and natural resource abundance is related to the environmental Kuznets curve hypothesis, thus the relationship between these variables is expected not to be linear. The linear model as specified in Eq. (5) is therefore extended by incorporating the square term of GDP. The quadratic model is expressed as below:

Non-dynamic Panel Threshold Regression
To examine whether there is a structural change in the relationship between the dependent and independent variables, the static panel threshold regression model as proposed by Hansen (1999) is employed. If the interaction term coefficient in Eq. (4) for Model 1 is significant, it means that the effect of domestic investment on GDP is contingent with the level of natural resource abundance.
On the other hand, if the p-value of the square term in Eq. (5) is significant, the effect of GDP on natural resource abundance is dependent on the level of GDP. Thus, Models 1 and 2 need a further investigation by using the non-dynamic panel threshold regression model and it can be written in Eq. (7) and Eq. (8) Where lnNA it is the threshold variable for Model 1 representing natural resource abundance in selected ASEAN countries. I(.) denotes a Heaviside function indicating I(lnNA it ≤λ)=1 when I(lnNA it ≤λ)=1 or otherwise, I(lnNA it ≥λ)=0, for β 31 , and vice-versa for β 31 * . If the sign of the coefficient of β 22 *  in Eq. (4) is positive and statistically significant, then the coefficient of β 31 *  , in Eq. (7) should be positive or higher than β 31  in the same equation. Lastly, η i denotes the country specific effect to address the heterogeneity.
Meanwhile, the extended model from Eq. (6)  Both Eq. (7) and (8) are estimated according to Hansen (1999). When is known, the ordinary fixed effect regression model is employed to estimate the values of lnDI in Model 1 and lnGDP in Model 2, and the corresponding sum of squared errors is S i (λ). However, if λ is unknown, Hansen (1999) introduced an estimate of λ by using the least-squares based on the non-linear specification in Eq. (7) and (8). This is the easiest way to achieve by minimising the concentrated sum of squared errors. Hence, the least squares estimator of λ is shown as follows: The minimum sum of squared errors from Eq. (9) is S i (λ) with a variance estimate as follows: > @

S nT
Next we test whether there is a threshold effect as proposed by Hansen (1996Hansen ( , 1999 included a hypothesis test for both models, Thus, the approximate likelihood ratio test of zero versus single threshold can be based on the following statistic formulae: Where S 0 represents the sum of squared errors when the null hypothesis is accepted. However, if the null hypothesis of no threshold is rejected, the single threshold is chosen and if F 1 is large. As addressed by Hansen (1999), since the null asymptotic distribution of the likelihood ratio test is not pivotal, the bootstrap procedure is used to estimate the sampling distribution. If the bootstrap is asymptotic, the efficient of the p-value is derived in accordance with the F-value under H 0 . The null hypothesis of no threshold effect is rejected under this condition, where the probability value (P-value) is smaller than the critical value. Furthermore, Hansen (1996) showed that for a large sample size, the statistic of the P-value is based on the uniform distribution, and thus the Bootstrap method can be used to get the value. Finally, we consider the construction of confidence intervals for the threshold parameters to test whether the value of the estimated threshold is reliable. Due to the nuisance parameters, the traditional statistics will be non-standard estimation. In order to overcome this problem, Hansen (1999) built a no-rejection region of asymptotic and efficient confidence interval using the maximum likelihood ratio LR statistics. Thus, the confidence interval is constructed as follows: Our asymptotic of (1-α)% confidence interval for 1 is a set of values of and thus LR O d f 2 1 1 ln .One of the strong features of this confidence region is that it is a natural model estimation. The likelihood ratio sequence LR is a simple renormalisation of numbers and it requires no further computation. The model above only considers a single threshold. In some cases, it can be a multiple threshold. However, in our case, we can identify two or more threshold values, as pointed out and discussed by Hansen (1999).

Data description and source of data
A series of data for all variables used in this analysis are shown in Table 2. All variables are transformed into the natural logarithms except for GOV due to its small data scale ranging from 1 to 4.5. The data are described as a preliminary analysis to be used as a reference in the main analysis. The highest value of the variable is GDP per capita (GDP) and the smallest number is natural resource abundance (NA). NA has the biggest gap in the data, followed by GDP as indicated by the standard deviation. It suggests that NA and GDP among selected ASEAN countries are relatively different. Thus, the heterogeneity of NA and GDP motivates us to further investigate the nexus between GDP and NA. Meanwhile, the portion of domestic investment (DI) is higher than foreign direct investment (FDI), indicating that DI is more important than FDI as investment is one of the core determinants of economic growth in Keynes's theory. However, the high kurtosis of DI suggests that any changes in DI are uncertain. Hence, these variables are considered as focal variables in our investigation. The units of measurement and sources of data for all variables are shown in Table 3. Real GDP per capita is used as a proxy for economic growth. NA that refers to total natural resource rents is the sum of rents from oil, natural gas, coal (hard and soft), minerals, and forests.

EMPIRICAL FINDING
Our empirical findings are discussed based on the estimation results for the baseline, non-linear and non-dynamic panel threshold models.
Model 1: The impact of natural abundance on economic growth.
The relationship between natural resource abundance and economic growth as specified in Eq.
(3) is shown in Table 4. The P-value of the Hausman test in both of the baseline and interaction models are higher than 0.05, hence, the alternative hypothesis is rejected indicating that the interpretation of the results for both of the baseline and interaction models must be based on the RE model. In the baseline model, the impact of natural resource abundance (NA) on economic growth is positive (0.141) and statistically significant at the 1% level. Similarly, the coefficient of domestic investment (DI) is also positive (0.359) and statistically significant at the 1% level. Our assumption that NA may influence DI and thus affects GDP is proven when the coefficient ( ) * β 22  of the interaction term is statistically significant at the 1% level. The sign of the coefficient for the interaction term, ( ) * β 22  , is however negative (−0.024), indicating that natural resource abundance can dampen the positive effect of DI on GDP. Thus, natural resource abundance plays a moderating role in influencing the impact of DI on GDP. All coefficients of the control variables are statistically significant at various significance levels (1%, 5% and 10%) and have positive signs in line with the theory, except for TO that has a negative sign.
Next, the non-dynamic panel threshold method is employed to confirm the effect of NA as a catalyst to dampen the effect of DI on GDP. The role of NA as a moderating variable in influencing the relationship between DI and GDP can be tested by using a threshold test as shown in Table 5. The statistics of F 1 and F 2 as shown in Table 5 are to test the number of thresholds, either single or double. Both statistics are based on Model 1 in Eq. (7) that is estimated by using the least squares. Table 5 shows that the test of the single threshold F 1 is significant with a bootstrap P = 0.033, however, the bootstrap P-value for F 2 is 0.379 and it is >0.05, suggesting that the model has no double threshold. We conclude that there is evidence of one threshold in the regression relationship. The estimated value of the single threshold within the range of 95% confidence intervals is shown in Table 5.1. The estimated value is 0.004 and it is very small in the empirical distribution of the threshold variable of NA. Hence, the countries are categorized into two groups. One is those with low natural resource abundance and the other is those with high natural resource abundance. Table 5.2 shows the core analysis of this research paper. The relationship between DI and GDP is dependent on the level of countries' natural resource abundance. A 1% increase in DI will cause GDP to increase by 0.368% at a level of NA below 0.004. Other than that, GDP increases by 0.296% for every 1% increase in DI when the level of NA is >0.004. It means that the effect of DI on GDP becomes smaller after NA is larger than the threshold value. The results of the threshold test confirm that NA will diminish the effect on GDP through DI, suggesting the presence of the natural curse hypothesis in the selected ASEAN countries. A huge amount of natural resources might be exploited and wasted by domestic investors and thus boosts GDP through DI. Besides, there are positive effects of FD, FDI, GOV on GDP in the countries. Statistically, a 1% increase in FD, FDI and GOV can increase GDP by 0.28%, 0.04% and 0.12%, respectively. With better financial markets and institutions, loans can be easily accessible, especially for small and medium enterprises (SMEs) and thus it generates more economic activities and output. Besides, higher inflows of FDI especially from developed countries will help enhance productivity and output as more advanced technologies are brought in the countries. These findings are consistent with the results of Omri et al. (2015), and . Governance can also contribute to higher economic growth as most resources are more efficiently and effectively used as the governments are  monitoring corruption and controlling the rent-seeking behavior that can cost a fortune to these countries. These findings support the findings of Siddiqui and Ahmed (2013). TO, on the other hand, has a negative and significant relationship with GDP which is in line with the preliminary results in Table 4. A 1 % increase in TO can reduce GDP of ASEAN countries by 0.03%. The sign of coefficient for all control variables have a similar direction as shown in the previous results in Table 4.
Model 2: The impact of economic growth on natural abundance.
The relationship between economic growth and natural abundance for the baseline and interaction models as specified in Eq. (5) and Eq. (6), respectively, are shown in Table 6. Similarly, the Hausman test for both models suggested that the RE model produce better results as the P-values are higher than 0.05 (0.974 and 0.914, respectively) and we fail to reject the null hypotheses.
As for the baseline model, the impact of GDP on NA is positive (0.910) and statistically significant at the 1% level. The relationship between GDP and NA is not linear and it exhibits an inverted U-shaped curve. The coefficient of GDP ( ) γ 21  is positive and then the coefficient of GDP square ( ) * γ 21  is negative and both are statistically significant at the 1% level. This suggests that in the early stage of higher economic growth, NA will simultaneously increases. However, in the final stage of higher economic growth, NA starts to decrease. Even though the results in the quadratic model show an inverted U-shaped curve, but we still do not rely on these results since the turning point from this estimation results is out of the range of the maximum GDP. Therefore, these findings need a further analysis by using the threshold method for nondynamic panel. Table 7 shows that there is only a single threshold as the bootstrap P-value of F1 is 0.030 and statistically significant at the 5% level. However, the value of F 2 is not significant. The point estimate of the single threshold reported in Table 7.1 is USD1, 818.378 and it is within the range of the 95% confidence intervals. Table 7.2 shows the threshold regression results that reveal the effect of GDP on NA. The coefficients of GDP before and after the threshold levels are positive and statistically significant at the 5% and 1% levels, respectively. The results show that a 1% increase in GDP will cause NA to increase by 0.469% at a level of GDP below $1,818.378. Interestingly, NA will increase by 0.612 % when there is a 1% rise in GDP and its level is higher than $1,818.378. It means that the effect of GDP on NA becomes greater after GDP is larger than the threshold value. Thus it can have a potential to boost their industries due to advanced technologies to utilise natural resources. Higher GDP can lead to better financial institutions, high-skilled labour and advanced technologies that can improve natural resources.
However, there are no significant effects of FDI, TO and GOV on NA in the countries. Meanwhile, the coefficient signs of DI and FD indicate there are negative effects of DI and FD on NA. Statistically, a 1 % increase in DI can reduce NA by 0.22%. This implies that with higher DI, new areas will be developed to build roads, houses, factories and other infrastructures. These new areas can potentially lead to discovery of more natural resources such       as earth minerals. Similar to DI, FDI can also have a negative and significant relationship with NA. Statistically, a 1% increase in FDI can result in a decrease of 0.20% in NA. This will lead to better financial institutions and thus more loans can be provided for the service industry compared to the heavy industry that deals with natural resource extraction as it becomes a growing industry in most emerging economies in ASEAN.

CONCLUSION AND POLICY RECOMMENDATION
The findings contribute to the existing literature that investigated the relationship between abundant natural resources and economic growth to validate the natural resources curse hypothesis in ASEAN countries. We introduce two different models in our study. First, we find that there is a relationship between domestic investment and economic growth depending on the level of natural resources abundance. When the threshold level of natural abundance is below 0.004, GDP will increase 0.368%. GDP will increase 0.296 % for an additional percentage of domestic investment if the level of natural resource abundance beyond 0.004. The empirical results suggest that the impact of domestic investment on GDP become slower after the natural abundance in the countries reaches beyond its threshold value. Thus, these findings support the natural resources curse hypothesis as a whole. Second, the estimated results in Model 2 show the relationship between GDP and natural resource abundance depends on the level of GDP. We find that when the GDP level reaches $1,818,378, there is a significant abundance of natural resources, indicating that after GDP reaches the threshold value, the effect of GDP on natural resources surplus is increasing.
Therefore, the findings are important for policymakers to formulate policies. The selected ASEAN countries can diversify and generate economic growth in various sectors such as manufacturing, agriculture and services. Economic diversification may spur higher productivity and can lead to more job opportunities. It can also stimulate the economy and thus sustainability can be achieved especially during a plunge in global demand for natural resources. Besides, the governments in the selected ASEAN countries need to attract more domestic investment which may lead to adopt the latest technologies and gain expertise. This will help enhance the exploration and extraction of new resources especially in the mining and quarrying sector. In addition, improving the upstream activities of natural resources, such as crude oil, gas and coal, is of utmost importance in boosting economic growth. Therefore, measures such as comprehensive trade policies and governance which involve natural resource management can be appropriately addressed to increase economic activities not only within the regions but also other trading partners such as the US, Japan and European countries. This is crucial to ensure the credibility of the governments in managing their resources and thus contributes to economic sustainability.

ACKNOWLEDGMENT
This research is funded by the international matching grant, TEJA (GSAT2020-12), from two universities, namely Universiti Teknologi MARA, Melaka Campus, Malaysia and Universitas Binus, Indonesia. This research team would like to express gratitude to Dr Elya Nabila Abdul Bahri from Universiti Malaya for sharing her expertise in panel data analyses.