Does natural capital depletion hamper sustainable development? Panel data evidence

This paper addresses growing concerns that the global decline in natural capital hurts wellbeing in the long-run. We examine empirically how natural capital depletion affects sustainable development as measured by a positive change in the United Nation’s Inclusive Wealth indicator. Drawing on panel data for 140 countries between 1990 and 2010, within-country analysis reveals that many countries convert their depleted natural capital sufficiently into humanand produced capital such that the comprehensive stock of wealth grows. However, some developed economies are struggling to harness their natural wealth for sustainable development. The policy implication is that there is no universally applicable recipe for the effective management of countries’ natural resources. Countries that are poorly endowed with humanand produced capital, in particular, may have a window of opportunity in which natural capital depletion is temporarily able to foster rather than hamper sustainable development.


Introduction
The worldwide depletion of natural capital helped trigger widespread concerns that societies do not manage their natural resources sustainably (Arrow et al., 2012). Sustainable development requires that natural resources are exploited without compromising the ability of future generations to satisfy their needs (Solow 1974). The loss in welfare from resource extraction can conceivably be mitigated by making sure that natural resource revenues are reinvested sufficiently in other capital assets (Hartwick 1977;Collier et al., 2010). Notably, the conversion of depleted natural capital into forms of human-and produced capital can maintain intergenerational welfare and be part of a sustainable development agenda.
The debate on whether countries succeed in converting natural capital sufficiently is far from settled. What limited empirical work has been done finds that the shares of natural resource rents and exports as a percentage of GDP are negatively associated with wealth growth (Atkinson and Hamilton 2003;Neumayer 2004;Dietz et al., 2007;Hess 2010;Boos and Holm-Müller 2013;Forson et al., 2017;Koirala and Pradhan 2020). Hence, these studies 1 suggest that undiversified economies not only experience lower income growth (Abdulahi et al., 2019; Pèrez and Claveria 2020; see Papyrakis 2017 for survey), but also face a reduced capacity to develop sustainably. However, new evidence shows that natural resource exploitation may foster human-or produced capital accumulation (Sun et al., 2019;Zallé 2019;Lashitew and Werker 2020;Ouoba 2020), which suggests natural resource-dependence need not be a curse for sustainable development. This calls for a comprehensive empirical analysis that evaluates if the global decline in natural capital has hindered or promoted sustainable development.
Our paper contributes to the literature by analyzing the net effect of natural capital depletion on sustainable development for a large panel of 140 countries from 1990 to 2010, employing Inclusive Wealth accounting and data (United Nations University International Human Dimension Program & UNEP, 2015). Inclusive Wealth measures the social value of all productive assets that contribute to wellbeing, comprising three categories: human, produced, and natural capital. The time-series level-data on each stock allows us to study the biophysical loss of natural capital, instead of relying on the imperfect and highly criticized proxies of resource usage (James, 2015;Stijns 2005;Brunnschweiler and Bulte 2008). Moreover, Inclusive Wealth accounting is considered by many to be a state-of-the-art measure of sustainable development (Dasgupta 2014;Engelbrecht 2016;Polasky et al., 2015;Clark and Harley 2020). It indicates a country's overall economic prosperity and its ability to meet Sustainable Development Goals, rendering it particularly useful to derive policy prescriptions (Dasgupta 2014).
Our study employs fixed-effects regressions corrected for clustering at the country-level. We examine how within-country changes of natural, produced, and human capital affect rates of per capita Inclusive Wealth growth. We focus, in particular, on inter-economy differences by evaluating if economies that depend heavily on natural capital (i.e., natural-resource dependence) exhibit a weaker ability to convert natural capital into human-and produced capital. To this end, we introduce a new empirical measure of resource dependence and perform various robustness checks with traditional but less adequate measures of resource-dependence. Results indicate that natural capital depletion is associated positively with per capita Inclusive Wealth growth, but not for all countries. While natural capital depletion fosters sustainable development in developing and resource-dependent economies, it may have become unsustainable in several developed countries.
The rest of this paper is organized as follows. Section 2 briefly discusses the theoretical links among economic performance, Inclusive Wealth, and sustainability before formulating hypotheses. Section 3 details the data and empirical approach. Section 4 presents the results and various robustness checks. Section 5 summarizes the main findings and provides some concluding remarks.

Background and hypotheses
There has been a growing academic interest in the sustainable exploitation of natural resources (e.g., Blanco and Grier, 2012;Mardones and del Rio, 2019;Qureshi et al., 2019;Ouoba 2020). This is fueled, in part, by resource-rich economies' persistent struggle to develop sustainably. Although many resource-rich economies (e.g., Liberia, Qatar, United Arab Emirates) are often able to achieve impressive rates of per capita GDP growth (World Bank Group 2019), they also face a shrinking capital stock and corresponding deterioration of their future productive capacity (United Nations University International Human Dimension Program & UNEP, 2015). The sustainable extraction of natural resources requires that depleted natural capital is compensated by an accumulation of produced-or human capital assets that make an equal or greater contribution to welfare (Hartwick 1977;Hamilton and Hartwick 2014).
In this light, the Inclusive Wealth approach measures the productive capacity of an economy by estimating its productive base, comprising all capital assets 2 : (i) human, (ii) produced, and (iii) natural. If the productive base grows, future generations have more resources and, therefore, an improved ability to support and raise their standard of living. Resonating with weak sustainability (Pearce et al., 1996;Arrow et al., 2012), sustainable development is thus achieved when per capita Inclusive Wealth stocks are non-declining. Fig. 1 illustrates how economies are able to expand their comprehensive stock of Inclusive Wealth. Each capital stock (i.e., human, produced, and natural) supplies a flow of resources to be used in production. The resulting output, in turn, is either reinvested to form new capital or used to satisfy present-day needs. However, natural capital exploitation comes at a cost, as natural resources depreciate in production (Daly 2020). 3 Hence, a sufficient amount of output must be reinvested to compensate for the loss of natural capital and prevent a decline in Inclusive Wealth. Countries that rely heavily on natural resources in production must be particularly cautious and safeguard a relatively high rate of capital reinvestment to prevent overconsumption in the short-run. Nevertheless, even the most vulnerable countries can extract natural resources sustainably as long as these are sufficiently transformed into produced-or human capital.
Theoretically, the ability to convert natural capital into other forms of wealth differs among countries due to diminishing returns to capital accumulation (Krugman 1994;Johnson and Papageorgiou 2020). All else constant, developing economies possess fewer produced capital assets and are sparse in human capital, as underscored by lower longevity and educational outcomes. As a result, these economies accumulate produced-and human capital more easily. For instance, the investment necessary to achieve higher educational attainment is modest when a population is undereducated. Similarly, when natural resources are extracted, the share of resource revenues required to be reinvested in produced-and human capital to compensate for the loss of natural wealth is much lower in developing economies. Conversely, diminishing returns to human-and produced capital accumulation swell as economies develop, increasing the required reinvestment of depleted natural wealth to develop sustainably. Although it may appear disadvantageous for developed economies, evidence shows most developed economies better utilize their natural capital (Kurniawan and Managi 2018). We therefore propose the following general hypothesis (H1): Hypothesis 1. Ceteris paribus, natural capital depletion increases a country's rate of Inclusive Wealth growth.
In contrast to the main hypothesis, most empirical studies find a negative association between some operationalization of natural capital and wealth growth (Atkinson and Hamilton 2003;Dietz et al., 2007;Fig. 1. Model of capital conversion and sustainable development Note: This figure illustrates how the productive base comprising natural, produced, and human capital assets drives sustainable development. Capital stocks supply resource flows to the production of goods and services, which can be reinvested in as new capital to develop sustainably or used for present-day needs such as consumption and other forms of leakages.
2 Some studies explicate social or intangible capital as a form of capital; which is an intangible source of productive capacity embedded in societal institutions. This study and others like it account for social capital implicitly via shadow prices. Calculations of social capital do not yet meet the standards for empirical analysis to warrant a separate stock value (Engelbrecht, 2016).
3 Natural resources become physically embodied in the product, whereas labor and produced capital inputs do not. Holm-Müller 2012, 2013;Kurniawan and Managi 2019). These studies propose that resource-dependent economies fail to reap the theoretical gains and instead squander natural resource revenues. Boos and Holm-Müller's (2013) theoretical overview of the literature offers the following explanation. Resource-dependence is often accompanied by or may even cause political systems that are not conducive to wealth growth. Governments in natural resource-dependent countries may become rentier states (Turan and Yanıkkaya 2020). Resource rents provide a vast share of government revenues in these countries, exemplified by Nigeria where 95% of government revenues stem from the resource sector (Gupta and Chu 2018). More importantly, these rents are used mainly for government consumption, corruption, and other forms of short-term wastage instead of being used for productive investments (Lange and Wright 2004;Papyrakis and Gerlagh 2004;Barbier 2010). Hence, the lack of proper institutions discourages investment in human capital while encouraging natural capital depletion. The perverse incentives that stifle sustainability resulting from resource-dependence may even extend beyond the natural capital sector. We therefore include the following hypothesis (H2) 4 : Hypothesis 2. Ceteris paribus, natural capital depletion decreases a country's rate of Inclusive Wealth growth in natural resource-dependent countries.
We test these hypotheses using panel data for up to 140 countries. The next section presents details on these data and the empirical approach to testing these hypotheses.

Data source and sample
The main data source for this paper is the 2014 Inclusive Wealth Report developed by the United Nations Environment Programme and the United Nations University -International Human Dimensions Program (United Nations University International Human Dimension Program & UNEP, 2015). The Inclusive Wealth indicator measures countries' total wealth by aggregating the natural, produced, and human capital stocks. Inclusive Wealth is considered to be a superior indicator of sustainable development. Moreover, it fits this research's aim best because it offers the most comprehensive array of natural resources that comprise the natural capital stock (Engelbrecht 2016). The data provides real dollar values of each capital stock for 140 countries at five moments in time (1990, 1995, 2000, 2005, and 2010). Together, these countries cover 95% of the world population. Table 1 provides key summary statistics for the variables discussed below.

Dependent variable
Sustainable development, the dependent variable, is measured as the percentage change in per capita Inclusive Wealth over 5-year intervals. Inclusive Wealth growth rates are slightly positive on average (Table 1; see also Table 5 in Appendix A) and distributed approximately normally across countries. The best-performing countries show an average growth rate of some 10% (0.1), whereas poorly performing countries achieve a growth rate of around − 10%. Several countries experience large, idiosyncratic drops in Inclusive Wealth, oftentimes due to events like wars or conflicts, mostly via sharp declines in produced capital. We control for these as a robustness check (Appendix B.4.).

Main independent variables
The main independent variables concern countries' per capita stocks of human, produced, and natural capital. 5 Data on each stock at each time comes from the Inclusive Wealth report's data Appendix. We calculate the natural logarithms of each stock as is common in growth accounting (Mankiw et al., 1992). Per capita human capital is the largest capital stock, whereas natural capital shows the most cross-country variation. Mostly oil-states (e.g., Qatar, Kuwait, United Arab Emirates) have vast natural capital wealth, while others (e.g., Singapore, Maldives, Haiti, Lesotho) are natural resource-scarce. More importantly, we identify several countries that have experienced significant drops in human capital (i.e., Moldova), produced capital (e.g., Afghanistan, Liberia, United Arab Emirates, etc.), or natural capital (e.g., Bahrein, Qatar, United Kingdom, Greece) in at least one period. Robustness checks confirm that these potential outliers do not influence any results.

Moderating variable: natural resource-dependence
We include an interaction variable for natural resource-dependence, which moderates the effect of natural capital accumulation on per capita Inclusive Wealth growth. We introduce a new indicator for natural resource-dependence to address some of the many issues troubling conventional indicators (Bulte et al., 2005;Stijns 2005;Dietz et al., 2007;Brunnschweiler and Bulte 2008;Ploeg 2011). 6 We consider countries to be natural resource-dependent when natural capital is overrepresented in the total wealth stock. Hence, we label countries as such when (1) human-and produced capital are scarce, and (2) natural capital is more abundant relative to countries with comparable human and produced capital scarcity. The second criterion distinguishes developing economies from resource-dependent ones. Our indicator is robust to short-term fluctuations in economic activity because it is based on (the composition of) wealth, which changes gradually over time.
We operationalize this indicator following Ahmad et al. (2018) network-based frequency analysis that groups countries based on the relative abundance of types of capital. Briefly put, the method compares every country with every other and links them when values are close. The country with the most links becomes the trend and the country-distance from this trend (orbital distance) determines relative abundance. Based on this analysis, countries have high-or low values of each type of capital, corresponding to a relative abundance or scarcity of that type of capital. For our purposes, we identify three groups of countries based on the orbital distance at the beginning of the sample (1990).
The first groupcomprising 57 developing countriesscores low on all capital types. The natural resource-dependent group contains 30 countries that score low on produced-and human capital but score high on natural capital. The 31 countries that score high on produced and human capital can be considered developed economies independent of natural capital abundance. Although we include the group of developed economies in the main analysis we do not evaluate specific hypotheses by doing so. The 22 remaining uncategorized countries are very diverse and omitted from the main analysis. A time-invariant categorical variable indicates group membership, which takes a value of 1 for the developing countries (reference category), a value of 2 for resourcedependent, and a value of 3 for the developed countries. Note that due to the fixed-effects approach (subsection 3.2.), the time-invariant dummy itself is omitted from the analysis due to collinearity, but the 4 In addition to Hypothesis 2, a leakage beyond the resource sector is the crowding-out of human capital accumulation by resource dependence (Cockx and Francken 2016). To be complete, our empirical strategy also considers this potential transmission channel of resource-dependence-induced wastage. 5 We use a fixed-effects regression analysis and thus effectively study the changes in these stocks. See subsection 3.2. 6 The conventional indicator of natural resource dependence is the share of the sum of raw material exports in GDP (Sachs and Warner, 1995;Atkinson and Hamilton, 2003;Dietz et al., 2007;Brunnschweiler and Bulte, 2008;Boos and Holm-Müller, 2013;Lashitew and Werker, 2020). This indicator is inadequate because it varies with (whimsical) short-term fluctuations in countries' market activity, international trade, and demand for raw materials.
moderating effect is retained. Appendix A lists the final sample of 140 countries.
To be sure, for the reasons highlighted above, we prefer this indicator over common indicators of natural resource-dependence. However, to be complete, we also repeat our main analysis using the following, less adequate indicators as a robustness check. The first robustness check measures natural resource-dependence as the share of natural in total wealth (e.g., Gylfason 2001;Hodler 2006). The share of natural capital of total capital takes a value between 0 and 1, where 1 indicates the absence of produced-and human capital. This (standardized) variable interacts with each type of capital. The sample for this analysis contains 140 countries and 700 observations. The second robustness check follows seminal papers (e.g., Sachs and Warner 1995;Atkinson and Hamilton 2003;Dietz et al., 2007) in which natural resource-dependence is measured as the sum of raw mineral and ore exports over GDP. World Development Indicators online database provides the data (World Bank Group, 2019). Some data are missing, reducing the sample to 133 countries and 501 observations (Appendix 1, Table 5).

Empirical model
We use panel data to estimate the effect of changes to natural capital, produced, and human capital on per capita Inclusive Wealth growth. All regressions use fixed-effects models. 7 This approach evaluates the hypotheses by observing the within-country changes to each capital stock and rooting out the time-invariant country specific-properties. The country-fixed effects capture all heterogeneity among countries unaccounted for by the explanatory variables, including mean values of each capital stock and group membership. Moreover, fixed-effects measure the time-invariant country-specific properties that influence economic performance, such as cultural heritage and geography (Mankiw et al., 1992;Quah 1996;Temple 1999). Hence, the empirical model isolates the effect of changes to each capital stock (i.e., capital accumulation or depreciation) on the rate of Inclusive Wealth growth over time within economies. The empirical model that we estimate is as follows: where G it is the rate of per capita Inclusive Wealth growth for country i at time t, N it is the log of natural capital per capita, H it is the log of human capital per capita, and P it is the log of produced capital per capita. R i captures the moderating effect of natural resource-dependence, which interacts with each aforementioned explanatory variable. We estimate this model for country i at time t, where δ i is the country fixed-effect and ε it is the error term. We add time dummies T t for each period to control for global periodic fluctuations in per capita Inclusive Wealth growth that are unaccounted for by the explanatory variables.
The coefficient of N is expected to be negative thus supporting Hypothesis 1, which would indicate that natural resource depletion affects rates of Inclusive Wealth growth positively. We expect a positive coefficient for the interaction term N*R, supporting Hypothesis 2. 8 This would indicate that natural capital depletion has a less positive effect on the rate of Inclusive Wealth growth in resource-dependent countries relative to the reference category (developing economies). If the coefficient for N*R exceeds |β 1 | then the net effect of resource depletion in resource-dependent countries is negative and demonstrates the resource curse. Further, we expect negative coefficients for H and P, indicating diminishing returns to human-and produced capital accumulation respectively. Table 2 presents the results for different specifications of the model described by Equation (1). Akin to a standard neoclassical production function, Model (1) considers only human-and produced capital as sources of Inclusive Wealth growth. Model (2) considers the effect of all three types of capital accumulation, while not yet accounting for crosscountry differences. The results show that natural capital depletion fosters Inclusive Wealth growth, supporting Hypothesis 1. A one percent decrease in per capita natural capital increases the rate of per capita Inclusive Wealth growth by approximately 0.06 percentage points, meaning that natural resource usage creates a net wealth gain via conversion to produced-and human capital. The underlying mechanism is that, although some of the depleted natural capital will have been converted to present-day consumption, a sufficient part is reinvested to create new human and produced capital. 9 Human capital accumulation lowers the rate of Inclusive Wealth growth. A one percent increase in human capital per capita decreases a country's Inclusive Wealth per capita growth by approximately 0.26 percent, conditional on the specification of the model. This does not mean that human capital hampers sustainable development, but rather that each subsequent increase in human capital results in a decreasing Inclusive Wealth gain. Produced capital's effect is not statistically dependence. It could therefore also moderate any of the other independent variables negatively. Our regression models consider all possible configurations. 9 In our sample, only 10 out of 118 countries experienced an increase rather than a decrease in their per capita natural capital stock. These countries include Latvia, Belgium, Estonia, Cuba, and some Eastern European economies.

Baseline results
significant. The lack of statistical significance persists throughout the main analysis and robustness checks. A possible explanation is that diminishing returns to produced capital accumulation are typically too small to be observed within 20 years. Models (3) to (6) in Table 2 include a moderating effect indicating country-type (developed, developing, or natural resource-dependent). The interaction effects mean that we evaluate whether each capitaltype's accumulation affects Inclusive Wealth growth differently among country-types. The results indicate no discernible differences between developing countries (the reference category) and natural resourcedependent countries. In particular, the coefficient of natural capital for resource-dependent countries is insignificant in Model (3) and Model (6). This means that even natural resource-dependent economies, which are largely associated with unsustainability, manage to convert their natural resources sufficiently into other types of capital. Hence, we fail to find support for Hypothesis 2. Similarly, there is no statistically significant difference in the way produced-or human capital accumulation affects rates of sustainable development in resource-dependent economies.
Thus, there is no evidence that resource-dependence hampers Inclusive Wealth growth in the baseline results. In fact, natural capital depletion fosters hamper Inclusive Wealth growth in most economies. Interestingly, developed ones are the exception. We cannot confirm that natural capital depletion affects rates of Inclusive Wealth growth in these economies with statistical confidence. Hence, the conjecture that the lack of sustainable development by resource-dependent and developing economies is caused by a mismanagement of natural capital is unfounded. The observed gap in rates of sustainable development between rich and poor societies cannot be attributed to squandering activities by the latter group. Instead, our analysis reveals that the effect of natural capital utilization on sustainable development is dominated by the potential gains from natural capital depletion, which follow from diminishing returns to human-and produced capital. This is underlined by produced-and human capital's lower coefficients for developed economies. We dedicate the remainder of this paper to the robustness of our findings.

Alternative indicators of natural resource-dependence: natural capital share
The first robustness check considers a different measurement of the moderating variable natural resource-dependence: the share of natural capital in total wealth. Table 3 presents the results.
The main analysis' conclusion proves robust to this operationalization of resource-dependence. Natural capital has a negative coefficient in Model 8 meaning that resource depletion increases Inclusive Wealth growth, supporting Hypothesis 1. However, the effect size has changed slightly relative to the baseline analysis (Model (2); Table 2). The difference comes from the uncategorized 22 countries that are included in this robustness check but omitted in the main analysis. Human capital accumulation lowers the rate of per capita Inclusive Wealth growth and produced capital accumulation again has no statistically significant effect.
Models (9) to (12) in Table 3 introduce the moderating effects of natural resource-dependence. These models include a direct-and indirect (moderating) effect because the share of natural capital varies over time. The direct effect shows that a decreasing share of natural capital increases the rate of Inclusive Wealth growth. We do not find a statistically significant moderating effect of resource-dependence on natural capital depletion, meaning we find no support for Hypothesis 2. However, we find a positive indirect effect via human capital accumulation. This interaction has two interpretations. The first is that natural resource-dependence fosters Inclusive Wealth growth more in countries with higher rates of human capital accumulation. The second is that human capital accumulation fosters Inclusive Wealth growth more in Table 2 The effect of capital accumulation on sustainable development: moderation by country-type. countries that become more natural resource-dependent. Regardless of which interpretation one adheres to, the results suggest that natural resource-dependence is a blessing rather than a curse. Although this conclusion is tempting, we believe it is biased. Adding this particular operationalization of natural resource-dependence removes the statistical significance of the natural capital (i.e., we cannot confidently claim that resource depletion itself affects development positively). This is likely due to multicollinearity. Natural capital depletion decreases both the stock value of natural capital and its share in total wealth, despite standardization. The ostensible resource blessing, reported by the moderating effect in Table 3, should not be taken at face value.

Alternative indicators of natural resource-dependence: resource exports and resource rents
The second robustness check operationalizes natural resourcedependence as the sum of raw mineral and ore exports over GDP. This particular indicator of natural resource-dependence is not expected to affect rates of Inclusive Wealth growth. 10 Nevertheless, the goal is to rule out biases due to our novel operationalization of resourcedependence in the main analysis. Table 4 presents the results. In contrast to extant studies using this metric (e.g., Dietz et al., 2007;Boos and Holm-Müller 2012), we find no evidence that higher resource exports hamper sustainable development. The coefficient for natural resource-dependence is not statistically significant, neither as an explanatory variable nor as a moderator. We cannot claim that when countries increase resource exports their wealth growth decreases. Mineral and ore exports as a percentage of GDP yields no other meaningful results as a moderating variable.
We consider the possibility that the lack of meaningful results follows from our specific operationalization of resource trade intensity. We therefore add agricultural exports to mineral exports as a percentage of GDP, which is a common variation of this measure (Sachs and Warner 1995;Mehlum et al., 2006). Nevertheless, all coefficients for natural resource-dependence remain statistically insignificant. 11 All aforementioned findings are robust to restricting the sample to only countries with 3% or more resource exports as a share of GDP. We conclude that our main analysis' findings are robust to alternative ways to measure natural resource-dependence. 12 Finally, we conduct the same exercise while using natural resource rents as a percentage of GDP as our moderating variable for natural resource dependence. The results show that natural capital depletion increases rates of Inclusive Wealth per capita growth. We do not find a negative relationship between natural resource rents and sustainable development. In fact, we find some evidence that higher natural resource rents increase the effect of human capital accumulation on Inclusive Wealth growth, indicating that resource rents are converted sufficiently.

Conclusion
Despite widespread concerns regarding the sustainable exploitation of natural resources, our study shows that natural capital depletion may be a blessing in disguise. Studying a large panel of 140 countries between 1990 and 2010, we find that natural capital depletion largely promotes sustainable development, as measured by an increase in per capita Inclusive Wealth. The underlying mechanism is that resource revenues are reinvested sufficiently into other types of capital assets. As such, human-and produced capital accumulation compensate for depleted natural capital, meaning that future generations have more 10 Flows of mineral exports as a percentage of GDP is an indicator of trade intensity by the natural resource-sector but not an indicator of the structural importance of natural capital for long-run development. This is underlined by a weak correlation between resource exports and other metrics of resource dependence.
rather than less productive assets to satisfy their economic needs. Contrary to extant findings, even natural resource-dependent economies succeed in harnessing their natural wealth. However, developed economies may struggle to convert natural capital sufficiently. Although this may seem paradoxical at first, this outcome is quite sensible when considering that developed economies experience stronger diminishing returns to human-and produced capital. The policy implication is that natural capital management is not onesize-fits-all. Instead, countries sparse in human and produced capital have a window of opportunity where reinvesting some of the depleted natural resources can potentially create a net wealth gain. Due to its scarcity, the marginal benefits of human-and produced capital accumulation are still high and the marginal cost of natural capital depletion is relatively low. As such, resource conversion could be a viable strategy for initial development. However, countries should be cautious in designing policies that rely on prolonged natural capital depletion. The potential gains become smaller and may even turn negative as human capital accumulates.
Several limitations apply. First, our study considers weak sustainable development and does not purport to comment on whether resource depletion is ecologically sustainable. Indeed, the depreciation of natural capital may foster economic health at the detriment of ecological health. Second, Inclusive Wealth accounting, our main data source and theoretical framework, remains a work-in-progress despite cutting-edge valuation methods. Although Inclusive Wealth offers the most comprehensive and advanced dataset on natural capital assets, the lack of point-source resource data in some countries may have obscured the efficacy of our conclusions. We highly encourage other researchers to quantify and collate point-source natural resource level-data to further our understanding of the link between natural resources and sustainability.

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

Declaration of competing interest
None.

Acknowledgements
The author is grateful for the helpful comments by Esther-Mirjam Sent and André van Hoorn and insightful feedback provided by the referee. Notes: Table reports results for fixed-effects model. Standard errors (in parentheses) are robust standard errors that are clustered at the country level. P-values are reported in square brackets. The sample comprises 133 countries and 427 observations. The moderating effect is the flow of mineral and ore exports as a percentage of GDP, showing how resource trade intensity affects sustainable development directly and indirectly by moderating capital accumulation/depletion. Alternative operationalizations used but not reported are: the flow of natural resource exports as a percentage of GDP, natural resource rents as percentage of GDP, and oil rents as a percentage of GDP. *p < 0.05, **p < 0.01, ***p < 0.001.   . An asterisk denotes missing exhaustible resource data either due to the inexistence of such resources in the country or unavailability of the data in the source. A caret (^) indicates that the country has been marked as a possible outlier due to strongly deviating changes in one or more capital stocks in one or more periods.

Robustness check B.1.: controlling for missing exhaustible resource data
Natural capital comprises of renewable and exhaustible natural resources. The Inclusive Wealth Report, the data source, lacks data on exhaustible resources for several countries (marked in Appendix A; Table 5). Some countries possess significant unreported fossil fuel or mineral deposits (e.g., Sierra Leone, Uganda, Rwanda), which may give a false positive indication of sustainability when converted into produced-and human capital (Hamilton 2012). This would inflate the coefficient of one or more predictors. We control for this issue by including a dummy variable indicating missing data that interacts with produced capital stock for countries. 13 Further, we distinguish between the stock of forest and agricultural resources, for which data are complete, and fossil fuels and minerals, for which data may be incomplete.  13 Alternatively, the coefficient of human capital accumulation can be inflated. To be complete, we interacted the dummy with human capital, and with both stocks in two separate estimations. No specification of the model resulted in different assessments of the hypotheses. We opted for reporting and discussing only one specification. . The main effect is omitted due to collinearity with country-fixed effects. Missing variable dummy (com) indicates the group of countries where exhaustible natural resources are reported, (mis) indicates the group of countries where they are missing. *p < 0.05, **p < 0.01, ***p < 0.001. Table 6 presents the results. Agricultural-and forest resource depletion increases the rate of Inclusive Wealth growth, supporting Hypothesis 1. Fossil fuel and mineral depletion has no statistically significant effect, except in developed countries. Model (23) shows that exhaustible resource depletion lowers Inclusive Wealth growth in developed economies. Furthermore, produced capital's predictor is indeed inflated. Although produced capital appears to increase the rate of Inclusive Wealth growth in countries where mineral and fossil duel data are incomplete, this is likely the result of a bias. Produced capital neither has a statistically significant coefficient in any of the previous models nor for countries whose data are complete in this estimation.
Overall, this exercise strengthens the general conclusion that natural capital depletion increases rates of Inclusive Wealth growth. The lack of consistent fossil fuel and mineral data prompts caution regarding claims about its effect on sustainable development.

Robustness check B.2.: grouping countries based on institutional quality
To inspect whether grouping based on country-type (developing, resource-dependent, developed) in the main analysis did not inadvertently capture some other unobserved property, we group countries based on several other dimensions. The first approach considers institutional quality as measured by government effectiveness and rule of law from the Worldwide Governance Indicators (World Bank Group 2019). Specifically, we assess whether it provides a better moderating effect than country-type in the original analysis. Some argue that natural resource-dependence is merely the result of deeply-rooted institutional heterogeneity across countries (Bulte et al., 2005;Brunnschweiler and Bulte 2008;Boos and Holm-Muller 2012;Lashitew and Werker 2020).
The main findings prevail. 14 Natural capital depletion fosters per capita Inclusive Wealth growth more in economies with low-quality institutions than ones with high-quality institutions. While being cautious in exclaiming a resource blessing, this finding is certainly in stark contrast with the literature. For instance, whereas Mehlum et al. (2006) find that low institutional quality exacerbates the resource curse, this finding suggests that natural capital depletion fosters net wealth growth. The idea that resource-dependence leads to inferior institutions that, in turn, lower wealth growth is not supported by the results.

Robustness check B.3.: grouping countries based on income and geography
Furthermore, we group countries based on geography and the World Bank income classification system to identify low-, middle-low-, middle-high-, high-, and high-income OECD countries. The results show that produced, human, and natural capital accumulation do not affect rates of Inclusive Wealth growth differently among income classes, except for high-income and OECD countries. Results are comparable to repeating the exercise based on regions. However, a construct of three groups based on a particular mixture of income and location (i.e., high-income, Middle-East and Africa, and rest of the world) finds noteworthy results, reported in Table 7.   High-income denotes all high-income countries in the World Bank's income classification system per 1990. *p < 0.05, **p < 0.01, ***p < 0.001.
Model (29) shows that natural capital depletion has a slightly less positive effect in Africa and the Middle East compared to other non-high-income countries. This regional disparity in the extent to which natural capital depletion is beneficial is the only glimpse our efforts yield that not all developing economies benefit equally from natural capital conversion. As with all preceding analyses, the results demonstrate that human capital accumulation lowers the rate of Inclusive Wealth growth and all high-income countries report lower coefficients across the board.