Ecological Modernisation Theory (EMT) has been using mainstream economic development theory to include how and when environmental factors are brought into the mainstream of development. EMT was developed by Joseph Huber [32] and the central tenet is to explore institutional reforms and technological advancement to mitigate ecological crises of modernity [33] [34]. EMT uses the Environmental Kuznets Curve (EKC) as the main tool to explain the evolving relationship between environmental quality and the economic development for a country or countries. The EKC tool was first developed by Panayotou [35]. The EKC suggests that an inverted U-shaped curve best explains how economic growth relates to the environment with a certain minimum in growth being required before environmental outcomes start to peak and then decline [35] [36]. The EKC literature has been growing dramatically [37] [38] suggesting the importance of and the interest to understand the environmental impact of economic development with which the world is still grappling.

Studies to test EKC theory were analysed using regression analysis on time series and cross-sectional data. For instance, Song et al. [39] used an Ordinary Least Square estimation for a cubic log-log model to assess EKC in the provinces of China between GDP per capita and the three environmental degradation parameters of waste water, solid waste and waste gas and found the EKC could explain all three. Similarly, Riti et al. [40] also found evidence of EKC for CO₂ emissions and economic growth in China using various regression techniques. Al-mulali et al. [41] conducted multi-variant regression to examine EKC in 93 countries using ecological footprint as environmental variable and showed that the EKC holds for high incomes countries but not for low income countries. Similarly EKC was found for 10 Middle East and North African countries [42] and for a group of 132 countries [43]. EKC was also confirmed by using ecological footprint in place of carbon emissions as the environmental impact parameter [44].

Other studies have demonstrated six different relationships besides the inverted U-shaped curve between economic growth and environmental pollution [45]. Pérez-Suárez and López-Menéndez [28] examined the EKC puzzle and the forecasting ability of the EKC equation for 175 countries (including Bhutan). They added a third degree polynomial term of GDP per capita to extend the existing EKC equation and they found different patterns of EKC, while that for Bhutan along with other 10 countries were classified under “other cubic pattern” of EKC, a pattern other than the typical U-shaped, inverted U-shaped, N-shaped, or inverted N-shaped curves. The present study will explore whether Bhutan exhibits this pattern when examining the nexus between CIWB and economic growth.

The EKC has been used to support the need for economic growth in emerging nations as the basis not just of poverty reduction but also environmental improvement, especially if it is connected to urbanization [46]. The evidence from Asia is very supportive of this, particularly China and has been used to generate support for such urbanization-related economic growth. The Asian Development Bank [47] has shown that Asian countries are significantly earlier in their inverted U-shaped curve and this can be attributed mostly to faster and denser urbanisation; in later years Asia also has higher levels of renewable energy. Similarly, Jamel and Derbali [48] have shown that urbanisation in selected Asian countries had a statistically significant negative relation to carbon emissions. Zhang et al. [49] found that the EKC could be established between urbanisation and carbon emissions for 141 countries; this was attributed to urban agglomeration economies that improve the economy and environmental outcomes through economies of scale and density. Similarly, Lin et al. [50] have shown that economic growth in non-high income countries will decrease carbon emissions and even accelerating urbanisation will lead to just small increases in carbon emissions. Wang et al. [51] have shown that urbanisation, income and energy efficiency can allow China and India to pursue economic development without growth in emissions.

The TPT was first proposed by Schnaiberg in 1980 [52] , based on political economy whereby a capitalistic system becomes like a treadmill creating social and environmental impacts without integrating ways to deal with such issues. Thus, TPT calls for fundamental changes to the economic system rather than just reform and mostly people in this tradition see GDP as a short hand for capitalism [53]. The TPT also now incorporates a more biological approach developed by a number of scientists pioneered by Ehrlich and Holdren (1992); they began to try and explain environmental impact using the formula I = PAT. This is a combination of Population, Affluence measured as GDP per capita and Technology measured as impact per unit of GDP. GDP is used in this approach to be short for consumption. Quantitative discussions in this approach have been used to show that global impact is likely to continue to increase as long as population and GDP continue to grow [54] [55]. This has been criticised as not understanding that GDP covers much more than consumption, not being able to distinguish the multiple interactions between PAT, not including social change or political change as factors in environmental impact and not doing justice to spatial factors and agglomeration economies as in the EKC and the EMT approach [56] [57] [58].

The EMT suggests that rising economic opportunities will enhance technological advancement and their combination leads to reduced environmental impact once a certain minimum economic growth level is reached. The TPT, however, suggests that economic growth and technological advancement inevitably accelerate resource consumption—the treadmill—leading to further environmental degradation. The two theories are said to be opposing and orthogonal to each other [59].

Research on the effect of population on carbon emissions found it was not simply population but was dependent on household size and age structure [60] ; others have found more than a proportional impact of population on carbon emissions [55] [61] , though Dietz and Rosa [61] found that carbon emissions tended to decline at a per capita GDP of US $10,000 as would fit with the EKC approach and thus EMT. At national level, Zhang et al. [62] interpreted EMT for china and its relevance to the rapid socio-economic growth experienced by China in the past decade. At a household level, modest support for EMT was found but strong support for TPT was observed also [63]. For Bhutan, it was found that income and education can lead to choosing cleaner energy [64] , which in turn can reduce emissions. This supports EMT especially from consumer behaviour and technological choices, though the study was not analysed under the EMT and TPT perspectives. Renewable energy sources make up cleaner forms of energy and it was found to suit the logics of EMT [65].

At the microeconomic level such as business and firms, the Porter hypothesis [66] [67] argues that environmental regulation can trigger technological innovation to reduce impact; this approach has led to the phenomenon of decoupling. The decoupling agenda and green growth are closely associated with Ecological Modernisation Theory. Szigeti et al. [68] found decoupling between ecological footprint and GDP in 90% of the 131 countries examined. Evidence of economic growth decoupling from greenhouse gas emissions were observed globally [13] [69]. However, Ewing [70] and Jackson [11] argue that green growth postulated by EMT should be rejected as a way to restore environmental degradation evident from climate change and biodiversity loss due to rebound effects.

York et al. [29] employed an Ordinary Least Square analysis under the STIRPAT framework (a flexible version of IPAT), and found a positive effect of both population and economic growth on environmental degradation—supporting TPT and contradicting the expectations of the EMT. They also provided data on the eco-efficiency for 142 nations ranging between 2.76 to 0.52 with smaller values being more efficient and that for Bhutan was reported as 0.75. Similarly, Dietz and Rosa [61] reported the technology multiplier for 111 nations ranging between 0.24 to 5.75 and that for Bhutan as 0.47, the smaller value indicating a more environmentally benign socio-economic system. Jorgenson and Clark [71] also found a positive effect of economic growth and population on total carbon emissions and per capita carbon emissions in their cross-national study of 86 countries. Interestingly, they found relative decoupling between economic growth and total carbon emissions but no relative decoupling was observed between economic growth and per capita carbon emissions. Thus, partly supporting both EMT and TPT. This suggests that the unit of measurement can affect the analysis and thereby the outcome of the analysis.

The analyses so far have all been about the nexus between economic growth and environmental outcomes. Now there is an emerging research strand that has applied the EKC hypothesis to incorporate variables that can capture human well being. These are discussed in the next section.

The type of variables used for evaluating the environmental intensity of human wellbeing vary depending on the data availability and scope of the study. For instance, some studies used CO₂ emission per capita [18] [72] [73] , others have used ecological footprint per capita [16] [30]. Similarly, for human well being, life expectancy was used as a proxy variable in some studies [18] [30] [31] , while others have used a life satisfaction index [16]. The choice of variables for the present study was limited by the availability of longer time series data. A longer life expectancy represents healthy living which arguably indicates physical fitness, health services, good food and all other social embedding that impacts longevity. While GDP per capita as a measure of economic growth is well known, the per capita carbon emission is of interest for a carbon-constrained world and for carbon neutral Bhutan. In the present study, per capita CO₂eq emissions includes GHG emissions from the energy sector, cement production, and from farming and agriculture, thereby capturing most anthropogenic sources of CO₂eq emissions compared to most of the existing research that has accounted for emissions from energy and cement production only [71] [72] [77]. The time series data used in the present study were sourced from the World Bank [78].

The method based on the STIRPAT framework used in earlier studies [29] [77] [79] was employed to formulate a functional form specified in Equation (1). STIRPAT is flexible and it is being used for hypothesis testing and also the elasticity between the variables can take any value—not restricted to unity as in the IPAT.

ln C I W B = C + β 1 ln G D P p c + β 2 ( ln G D P p c ) 2 + β 3 ( ln G D P p c ) 3 + β 4 ln P O P + β 5 ln C I W B ( − 1 ) + ε t (1)

where, CIWB is the ratio variable between CO₂pc and LE adjusted for coefficient of variation to reduce the dominance of one variable driving the ratio [72] [73]. C is the intercept. lnGDPpc is the per capita GDP in natural logarithmic form and the polynomial terms of GDPpc are centred by subtracting the mean of lnGDPpc to reduce collinearity [16] [18] [29]. It is also logical to add the lag term of the dependent variable as one of the regressors in a time series data analysis-lnCIWB(−1) for the present study. POP is the population and ε is the error term that captures those unobserved explanatory variables. Finally, β₁ to β₅ are the coefficients of the corresponding explanatory variables that are of primary interest in this study, which are to be estimated by regressing Equation (1) using EViews.

On regressing Equation (1) it was found to have serial correlation and very high variance inflation factor (VIF) and also the linear term and the quadratic term of the GDPpc were found to be less significant than the cubic term despite being jointly significant as per the Wald test. This led to constructing two reduced form models: without the quadratic term (Equation (2)) and without the linear term (Equation (3)).

ln C I W B = C + β 1 ln G D P p c + β 2 ( ln G D P p c ) 3 + β 3 ln P O P + β 4 ln C I W B ( − 1 ) + ε t (2)

ln C I W B = C + β 1 ( ln G D P p c ) 2 + β 2 ( ln G D P p c ) 3 + β 3 ln P O P + β 4 ln C I W B ( − 1 ) + ε t (3)

Small models with fewer parameters were preferred over large models to reduce parameter uncertainty [80]. Between 2) and 3), we chose model 3) which scores relatively better in most of the model selection criteria. It has higher adjusted R² and lower Akaike info criterion (AIC), Hannan-Quinn criteria (HC); lower VIFs, and lower SE of regression and standard error for the estimated coefficients. In Equation (3), the quadratic- and the cubic-term were jointly significant at 5% significance level as per Wald test though the quadratic GDPpc was significant at 25% significance level only. The joint significance is important here as argued by Brown and McDonough [81]. Moreover, this is within the significance level up to 25% argued by Agung [80] for cases where a theoretically important variable became statistically insignificant. Furthermore, statistical analysis is considered partly a science and partly an art [82] requiring trial and error methods. In keeping with this, interpretation of the results in the next section requires caution and it is valid within the data set used in this study.

For regression analysis, stationarity of data is essential to avoid any spurious regression. In this study, any presence of unit root for all the variables were tested by using the Augmented Dickey Fuller unit root test and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) stationarity test. They were found to be non-stationary at levels but stationary in first difference. However, if the residual from their regression is stationary, then their combination can be considered stationary [81] [82] [83]. This condition was fulfilled in this study—that is the regression results were not spurious.

Statistical tests were conducted on Equation (3). The VIFs were below 16 [84] and it is not reasonable to remove variables just to improve the VIF [82]. The tests for the presence of serial correlation and heteroskedasticity were conducted using the Breusch-Godfrey (BG) LM test and Breusch-Pagan-Godfrey (BPG) test respectively [75] [85]. The p-values failed to reject the null hypothesis of no serial correlation and homoskedasticity. The robustness of Equation (3) was also checked through a robust least squares MM estimation confirming the signs of the explanatory variables with small differences for the values of the estimated coefficients and their associated standard errors.

The result of GDPpc causing greater carbon intensity per unit of well being suggests that GDP is not (yet) helping with the social and environmental issues associated with development in Bhutan as outlined in the TPT. The coefficient of the quadratic GDPpc was comparable with that reported in a cross national data [29]. Thus GNH or at least the existing socio-economic pattern appears to be leading to greater carbon intensity in Bhutan because enhancing human well being follows a significantly positive and curvilinear relationship with economic growth. And the economic growth has not yet reached a level where the EKC curve would suggest that economic growth will start to reduce carbon intensity. Thus, at this juncture in Bhutan’s growth trajectory, the current per capita GDP for Bhutan is far below the various threshold levels reported in the literature suggested before the EKC turns around on the inverted U curve and begins to decline². The way ahead for Bhutan which is seeking to resolve the three factors of wealth, happiness and the environment, is therefore quite complex.

The evidence from Bhutan is that the transition to reduced carbon intensity of well being is not yet under way. In other words, the result supports the Treadmill of Production Theory. The result reflects the transitioning of the Bhutanese economy from agriculture-based economy to manufacturing and industrial-based economy [86] [87]. The latter being more polluting than the former. To avoid environmental degradation in the early stage of development, Panayotou [35] suggests that environmental protection should be an integral part of development plans especially in developing countries. For Bhutan with its GNH strategy this had always been the objective and it cannot be stressed more especially now with its carbon neutral pledge that Bhutan will need to bend the emission curve before other nations have managed to do this.

One way to examine if the EKC is perhaps beginning in a developing country is to examine if relative decoupling of wealth and environmental impact has begun. Developing countries with low GDP have been found to exhibit relative decoupling [88]. Bhutan’s GDP seems to exhibit a relative decoupling from CO₂eq emissions as shown in Figure 2, although this was not the case for carbon intensity of well being. This suggests that there is a need to enhance life expectancy (the denominator of the CIWB) considering that the LE ranges from 45 years in 1980 to just 68 years in 2012. Moreover, Bhutan is still in its early stages of socio-economic development thus struggling to enhance liveability conditions for its citizenry.

This calls for greening Bhutan’s socio-economic activities in line with the aspiration of GNH and carbon neutral emphasised in their 12^th five-year plan [89] and their economic development policy [90] [91]. The need to further green growth in Bhutan was also emphasised by the World Bank [92]. It can be expected that economic growth in middle-income countries will decrease carbon emissions [50] but it needs action sooner to be consistent with its global commitments to happiness and carbon neutrality.

If the TPT approach had been the dominant story with Bhutan then the coefficient of population being negative in Equation (4) would not have been predicted. The result suggests that growing population will help carbon intensity per unit of well being. This finding may seem spurious to the IPAT identity that considers population as the main driver of environmental degradation [5] [29] [61]. Furthermore, the finding contradicts the adverse environmental impact of population growth speculated in the national strategy of Bhutan [93]. One way to explain this in the TPT model is to suggest that it is consumption (GDP) that is over-riding the population factor as Toth and Szigeti [94] found in their studies. However, it is also possible to see within the EMT approach that population growth, if it is in cities, can be an important factor in more rapidly leading to the EKC turn-around that would help Bhutan achieve its multiple goals. Lamb et al. [79] found this to be the case for population growth in their global study, and for the case of China by Zheng et al. [95] and Meng and Huang [96]. The plausibility of this is that in Bhutan farming and agriculture activities, which are concentrated in rural areas, are the major contributors of CO₂eq emissions for the time period prior to 2000 [97]. Rural Bhutan accounts for 66% of the total population [98].

Thus as wealth and population grow there is a faster move to the city and then a series of other factors follow that may indeed reduce consumption as economies of scale and density create smaller houses, better public transport and other potential impacts on consumption like greater recycling [51] [99] [100] [101]. Bhutan expects its urbanisation rate to increase from 10% to 51% in just 55 years compared to 95 years for Asia and 150 years for Europe [47]. This trend can be a positive phenomenon as Bhutan aspires to be a knowledge-based society [91] while also sustaining its carbon neutral goal. Wang [102] argues that rather than focusing on reducing population, the aim should be to educate and create awareness of how to reduce carbon emissions through behaviour changes. Similarly, Costantini and Monni [103] emphasised the need to invest in human and social capital to simultaneously achieve a higher living standard and environmental protection.

From Sections 5.1 and 5.2, it is clear that the case of Bhutan provides support for TPT with regard to GDP as the driver and EMT approach for its population growth that is showing positive outcomes as would be suggested with its rapid urbanization. Furthermore, urban areas were found to be happier [22] and wealthier [98] [104] compared to their rural counterparts. In the interim as it moves towards the top of the inverted U-shaped curve there are appropriate policy interventions that can be put in place so well being and environmental quality are not degraded while the economy grows [105].