Scenarios for different ‘Future Indias’: sharpening energy and climate modelling tools

ABSTRACT

India is crucial to the future of the global energy system, and hence to efforts to mitigate climate change. However, projecting India’s energy future is challenging because of a number of structural uncertainties related to its still emergent development process – there are multiple ‘future Indias’. This is particularly so given India’s unusual historical development pattern and economic structure. This paper reviews more than 40 energy system scenarios for India to 2050 including scenarios of different policy stringency. This analysis suggests that the task of drawing policy insights is hampered by widely divergent, often non-transparent and insufficiently discussed, assumptions about GDP and its structural attributes as implied by sectoral energy consumption, industrial intensity of GDP, and sectoral demand patterns. It is these assumptions that crucially drive the divergence in scenario results, even more than assumptions about energy and climate policy per se. While divergent results are important for exploring the range of possible future pathways, policy insights can only be generated if scenario producers adequately convey a sense of causality, probability and desirability underpinning their scenarios. This requires linking more detailed socio-economic scenario storylines with the input assumptions and output results of energy system scenarios. To improve modelling for policy insight, the paper proposes that more attention be devoted to developing overarching socio-economic development scenarios, linking these to sector dynamics, and unpacking and interpreting model results.

Key policy insights

• India’s development pathway is characterized by a relatively low level of industrialization, precocious growth of services, and low and slow urbanization compared to similar countries.

• The unusual features of India’s development pathway have a material impact on its energy system.

• There is a wide range of uncertainty pertaining both to India’s future rate of GDP growth, but also to the persistence or reversal of the aforementioned structural characteristics.

• Existing modelling studies do not adequately detail and contextualize assumptions about India’s macro-scale development pathway, and its links to micro-scale patterns of sectoral energy demand and supply, leading to a wide divergence of results and uncertainty over how to interpret them.

• Addressing these lacunae requires enhanced transparency and discussion of overarching socio-economic development scenarios for development, linking these to sector dynamics, and unpacking and interpreting model results.

KEYWORDS:

• India
• energy scenarios
• climate mitigation
• energy and development

1. Introduction

India is the second most populous country in the world, and the third largest economy measured in purchasing power parity (PPP) (World Bank, 2020). By 2050, India will be the largest country in the world by population, and the second largest economy (OECD, 2020). India is thus projected to account for 27% of the increase in global primary energy demand over the period 2018–2040 (IEA, 2019).

As a lower-middle-income country, India is by far the poorest country in the G20 group of major economies, with a GDP per capita at PPP that is almost 50% lower than the next poorest member, Indonesia (World Bank, 2020). India is still undergoing the large macro-scale transitions that characterize the development process: urbanization, industrialization, and demographic transitions. In developed countries, and increasingly in China, patterns of urbanization and spatial development, physical infrastructure, and economic structure and specialization are already committed. However, in a fast-growing developing country like India, these emerging but still uncertain transitions are critical to shaping the future pattern of material and energy consumption (Grubb et al., 2015; Pye & Bataille, 2016).

Despite the importance of India to the future of the global energy system and to climate mitigation, there is relatively little systematic study of Indian energy and emissions scenarios in the global scientific literature (an exception is Dubash et al., 2018). Moreover, energy system scenarios and modelling are sparsely used in national policy-making processes, and often without efforts to reconcile differing results.

This paper aims to contribute to improved use of modelling analysis for policy-making, by critically reviewing the results of a recent modelling project, combining both global Integrated Assessment Models (IAM) and national bottom-up energy system models. The key argument of the paper is that the explicitly and implicitly assumed future macro-structural features of the Indian economy – GDP growth rates, industrialization, urbanization among them – account as much if not more for emissions outcomes in these scenarios than assumptions on energy and climate policy per se. Therefore, modelling studies need to better provide users with the necessary information to interpret model results for policy analysis. In other words, modelling needs to explicitly account for different ‘future Indias’.

The paper is structured as follows. Section 2 briefly reviews certain key specificities in India’s development pathway and builds the argument that India’s long-term energy scenarios need to be interpreted not just in terms of assumptions on climate policy, but also in terms of – explicit and implicit – assumptions on India’s development pathway. Sections 3.1 and 3.2 give an overview of the models and scenarios analysed in this paper. Section 4 gives the results, starting with headline scenario results (4.1), assumptions on long-term GDP growth rates (4.2); energy demand growth and the sectoral structure of demand (4.3); and concluding with the final consumption and primary supply mixes (4.4). Section 5 ties together the analysis by conducting an illustrative exercise of developing combined socio-economic and climate policy storylines that can help in interpreting scenario results analysed in Section 4. The conclusion in Section 6 gives some recommendations for future modelling exercises.

2. India’s development pathway

This section provides a brief overview of some of the key ‘puzzles and anomalies’ (Lamba and Subramanian (2020)) in India’s historical macroeconomic development pathway. The objective here is to draw on analysis of India’s economic history and performance, in order to highlight certain aspects thereof that are most material to India’s energy system future and emissions pathway. Given that a key argument of our paper is that explicit and implicit assumptions around India’s macroeconomic development pathway drive the key results of the scenarios we assess, we foreground this argument by briefly describing India’s recent development performance.

Between 1980 and 2017, India’s real per capita GDP growth rate was the fifth fastest among large economies accounting for more than 0.5% of world GDP in 2017, behind South Korea, Egypt, China, and Vietnam (Feenstra et al., 2015). The poverty rate, having been essentially stagnant for the decades since independence, dropped from above 50% in the 1970s to below 20% in 2011 (Datt et al., 2016). This represents one of the fastest and largest processes of economic growth and poverty reduction in history. Nonetheless, India’s growth performance has been characterized by the ‘contrast between India’s growth dynamism – notably rapid, long, and consistent – and its social and structural transformations, which … have not matched its overall growth’ (Lamba & Subramanian, 2020).

A key aspect of this lagging structural transformation has been the relatively slow process of transition of employment out of agriculture and the stagnant share of industry in both GDP and employment (Timmer et al., 2015). The flipside of this lagging structural change has been the precocious development of high-productivity, export-oriented business services. Business Services (including the high-profile Business Process Outsourcing sector) and Financial Services accounted for 15% of total value added in 2016, up from 6% in 2000 and comparable with the level in some developed economies (Reserve Bank of India, 2019). Likewise, India’s manufacturing structure and manufacturing exports are skewed towards capital-intensive sectors, and away from the labour-intensive sectors one would expect given India’s wage level and large working-age population(Chatterjee & Subramanian, 2020). The structure of India’s rapid growth has thus been unusual, driven by knowledge- and capital-intensive sectors such as petroleum refining, and leaving huge parts of India’s low-skilled workforce excluded from the fruits of structural change and economic growth (Panagariya, 2020).

The other side of India’s atypical development process has been the relatively slow process of urbanization. Relative to its GDP per capital level, India’s urbanization level is low, and the growth rate of the share of the population in urban areas is slow (Tumbe, 2016). India’s large urban slum population, estimated at 24% of all urban population in 2014 (World Bank, 2020), has also led to India’s urbanization process being characterized as ‘messy and hidden’ (Ellis & Roberts, 2016). India’s relatively slow process of urbanization is generally projected to continue in the future, with the total urbanization rate by 2050 projected to touch only slightly above 50% (UN DESA, 2018). While part of India’s low measured level of urbanization may be due to its strict definition of ‘urban’ areas, substantial evidence at the subnational level shows that urban and rural definitions correlate relatively well with socio-economic indicators. At the subnational level, higher levels of urbanization are correlated with higher levels of development (proxied by household consumption expenditure); higher levels of energy intensity (proxied by household ownership of energy-consuming equipment); and higher levels of material intensity (proxied by the material of dwelling construction) (NSSO, 2013; ORGCC, 2011). A recent analysis by IMF economists shows that urban–rural location by itself accounts for around a third of the differences in household consumption expenditure (Balasubramanian et al., 2021).

The lack of low-skill manufacturing, the dominance of high-productivity, high-skill services and the ‘marooning’ of huge swathes of the population in unproductive agriculture create problems of exclusion from the fruits of growth (Panagariya, 2020). Likewise, the slow and messy process of urbanization has inhibited the rural to urban structural transformation that is at the heart of the development process, and contributed to the observed increase in urban poverty (Datt et al., 2016). Thus, two deep structural specificities of India’s development pathway are mutually reinforcing: suboptimal urbanization inhibits non-farm job creation, and the lack of low-skill non-farm job creation drives the persistent ruralization of Indian society and the Indian economy.

These idiosyncrasies lead to uncertainties about both the extent and the quality of India’s future growth. There are multiple possible ‘future Indias’. This is not an abstract concern for long-term energy modellers: long-term scenarios must take a view on India’s long-term growth prospects. Which side to take? With the pessimists, for whom ‘India will be hard put to achieve high-quality and enduring per capita growth’ in the coming decades (Joshi, 2016, p. 6). Or with the optimists, for whom India is poised to transform into a prosperous high-income country by 2050? This debate gained extra salience even before the historic shock of COVID-19, as India’s growth exhibited a marked slow-down in the 2010s compared to the 2000s (Subramanian & Felman, 2019).

But the challenge for energy modellers runs deeper: as discussed here, the forms of growth and its macro-structural attributes – particularly patterns of industrialization and patterns of urbanization – carry implications for future material consumption and energy needs. This is not to suggest that India’s future is over-determined; India could well break out of past patterns. For this reason, a range of macro-scale assumptions, beyond the relatively narrow confines of climate or energy policy, need to be examined when analysing Indian energy and emissions pathways. In the following Section 3, we show that these macroeconomic assumptions, both explicit in model input assumptions and implicit in model structures, crucially determine results. This creates the need for transparent socio-economic storylines and input assumptions, in order to enable scenario users to interpret and generate insights for policy from divergent results.

3. Methods

3.1. Models

The analysis in the rest of this paper is based on the results of a multi-model project called CD-LINKS, which developed energy and emissions pathways out to 2050 for several large economies, including India (CD-LINKS, 2020), and which have been made available in a public database (IIASA, 2020) forming the basis for several scientific articles (Mathur & Shekhar, 2020; Schaeffer et al., 2020; Vishwanathan & Garg, 2020). The CD-LINKS database contained India-specific results for seven models, described in Table 1. The models are classified as ‘global’ or ‘national’ depending on their geographical scope. All global models represent India as an individual region. Table 1 also gives a brief description of the model type.

Table 1. Models in the scenario database.

Model	Number of Scenarios	Model Scope	Model type
AIM/CGE 2.1	6	Global	Inter-temporal general equilibrium optimization
AIM/Enduse 3.0	7	National	Bottom-up energy system optimization model
IMAGE 3.0.1	6	Global	Energy system optimization model
India MARKAL	7	National	Bottom-up energy system optimization model
MESSAGEix-GLOBIOM 1.0	6	Global	Energy system optimization model
WITCH-GLOBIOM 4.4	6	Global	Inter-temporal general equilibrium optimization
REMIND-MAgPIE 1.7-3.0	6	Global	Inter-temporal general equilibrium optimization
Total	44	n.a.

3.2. Scenarios

Scenarios were grouped according to climate policy stringency, based on the descriptions given in the documentation available (IIASA, 2020, see Supplementary Material for further details). Four broad categories of climate policy stringency were identified:

• No policy scenarios (NPS)

• Moderate policy scenarios (MPS), entailing an implementation of India’s National Determined Contribution (NDC) to 2030 and a continuation of a comparable level of effort thereafter.

• Strong policy scenarios. For global models, this group equated to global emissions budgets consistent with limiting warming to 2°C. These are grouped within the ‘G_2C’ group. In the case of the national models, following the bottom-up logic of the Paris Agreement, national modelling teams selected their own levels of policy stringency, and hence emissions budget, consistent with a global storyline of policy stringency under a global 2°C scenario. These scenarios are grouped under the ‘N_SPS’ label, for ‘National Strong Policies Scenarios’.

• Very strong policy scenarios. For global models, this group equates to scenarios with a global emissions budget consistent with limiting warming to 1.5°C. Only one national model implemented a scenario in this category. The group label is ‘N_VSPS’, for ‘National Very Strong Policy Scenarios’.

4. Interpreting the scenario results

4.1. Headline results: energy emissions pathways

Figure 1 shows the headline results in terms of energy CO₂ emissions from 2005 to 2050 for two scenario groups in the scenario ensemble. We group scenarios into moderate policy scenarios and strong policy scenarios. Clearly, results diverge substantially, with scenarios of apparently similar policy stringency diverging by an order of magnitude in several cases. There is nothing wrong with divergence per se: modelling should aim to provide insights into inherently uncertain futures (Huntington et al., 1982). But if scenarios are to provide insights, policy-makers need to be given the tools to interpret them. If it is not differences in policy stringency alone that drive scenario results, what additional factors contribute to this divergence? In the following sections, we aim to unpack the scenario results in a step-by-step fashion, in order to answer this question.

Figure 1. Energy CO₂ emissions 2005–2050, all models, moderate and strong policy scenarios.

4.2. GDP growth rate assumptions

Figure 2 allows us to compare and interpret the GDP growth rate assumption of the models. The left panel shows historical growth rates for a selection of emerging countries as a function of GDP per capita. The right panel shows the same information for India in the models from 2010 to 2050. The model assumptions for GDP per capita growth are usefully discussed in terms of three categories. The first, comprising only the national India MARKAL model, assumes an extremely rapid and sustained growth in GDP per capita, well above India’s historical achievement. The projected rate also exceeds the historical growth performance of South Korea, which achieved the fastest sustained growth rate of any large economy from 1980 to 2017. Assuming that India can exceed this rate across the forty years from 2010 to 2050 is thus a strong assumption that merits discussion and explanation, particularly given the differences in the observed development pathway of India versus the East Asian industrialized countries (Section 2).

Figure 2. Scenario forecast GDP per capita growth rates, benchmarked against historical performance in emerging countries.

Note: MER = Market exchange rates. IND = India, CHN = China, KOR = South Korea, ZAF = South Africa, THA = Thailand, BRA = Brazil.

The second group of models assume a sustained high rate of per capita GDP growth, between 4-6% per year across the forecast period, and comprise MESSAGE, IMAGE, and AIM/Enduse. This assumption implies a continued exceptional economic performance, with India escaping the so-called middle-income trap of substantial growth slowdowns at middle-income, seen in the historical experience of Thailand, South Africa, and Brazil in the left panel of Figure 2. These countries typically achieved per capita GDP growth rates of less than 4% per year during the development transition between 5000 and 10,000 USD2010 MER per capita (which India is projected to traverse between 2020 and 2050). According to this group of scenarios, by 2050, India’s per capita GDP would be comparable with those of major Latin American or poorer Eastern European countries today. The third group of models envisages a lower per capita GDP growth rate of 3-5% per year across the projection period, and consist of two CGE and one integrated assessment model (WITCH, Aim/CGE, and REMIND). These result in a GDP per capita that attains slightly more than 5000 USD2010 MER by 2050, up from about 2000 USD2010 MER today (World Bank, 2020). India’s level of development would then be broadly comparable to that of Thailand today (World Bank, 2020).

The differences in growth rates, and therefore per capita GDP levels, are vast, with total 2050 GDP varying by an order of magnitude between models. Clearly, a large part of the divergence in emissions highlighted in Figure 1 is driven by factors that are completely exogenous to climate policy per se. Without understanding and contextualizing this divergence in macroeconomic assumptions, it is challenging to develop policy insights for energy and climate policy.

4.3. Growth rate and sectoral structure of energy consumption

4.3.1. Aggregate final energy consumption

Drilling deeper to explore energy sector results, Figure 3 shows the total level of final energy consumption (FEC) in 2050 versus total GDP. The figure shows the average value for each model for each scenario grouping. In addition to showing that there is a strong relationship between the assumed level of GDP in 2050 and total FEC, as would be expected, the figure illustrates some differences in the assumed relationships between GDP and FEC across models, which merit exploration and discussion.

Figure 3. 2050 Final energy demand versus 2050 total GDP.

To begin with, the India Markal model is clearly an outlier in terms of FEC, with a projected FEC an order of magnitude above that of any other model. This is due to the high GDP growth rate assumed to 2050 as seen in Figure 2. However, other models stand out as outliers for the implied relationship between GDP and FEC. The REMIND model assumes a significantly higher level of final energy intensity of GDP compared to the other global models, resulting in total FEC that is higher than other models with comparable levels of GDP, by more than a factor of two. Conversely, the national scope AIM/Enduse model projects levels of FEC that are substantially lower than other models with similar levels of GDP in 2050.

Different models also clearly make different assumptions regarding the elasticity of FEC to the stringency of climate policy, seen in the distance on the y-axis between the series for each model. The national scope AIM/Enduse model assumes almost no elasticity of demand to more stringent climate policy. Global models tend to assume very little reduction in energy demand between no policy and moderate policy scenarios, followed by the substantial reduction in demand between moderate policy and strong policy scenarios. In absolute terms, between moderate and strong policy scenarios, energy savings range from just 3.3% in AIM/Enduse to almost 40% in the WITCH model.¹

4.3.2. Industrial final energy consumption

Figure 4 displays results for industrial FEC, showing scenario results for per capita industrial final energy consumption in 2050 as a function of 2050 GDP per capita. It also includes historical data for comparator countries to serve as points of reference. The 2018 position of India on these two indicators, industrial FEC per capita and GDP per capita, is highlighted in the red series. Normalizing industrial FEC per capita allows one to compare the modelled pathway for India with that of other countries, in order to gain insights as to the kind of socio-economic trajectory (implicitly) assumed in the modelled scenarios for India.

Figure 4. Industrial FEC per Capita, India scenarios for 2050 versus historical cross-country benchmarks.

The various projections range from high industrial FEC, comparable to China’s trajectory, to extremely modest increases, more in line with Indonesia. Understanding these assumptions, and the storylines they represent, are important for the purpose of drawing policy insights. To begin with, the results for the REMIND and MARKAL models are outliers, with a high level of industrial FEC per capita relative to GDP per capita comparable to China’s trajectory (although China’s is still higher). Clearly, this would imply a large structural break in India’s development trajectory, which as detailed above has not so far been as industry-driven as other East Asian countries. Although not displayed on Figure 4, at a comparable level of GDP per capita as projected for India in 2050 (roughly 20,000 USD2010 MER), South Korea was consuming about 44 GJ of industrial FEC per capita, compared to 35 GJ projected in the India MARKAL no policy scenario in 2050. At that time, South Korea was running a net manufactures export surplus equivalent to 10-13% of GDP, compared to India’s historical structural deficit in manufactures (World Bank, 2020). Thus, such a high projected level of industrial FEC would be consistent only with a substantial break in India’s historical pattern of lower industrialization.

At the other end of the spectrum, the results for the WITCH and AIM/CGE models see only a marginal increase in industrial FEC per capita from the present level, despite a more than doubling of GDP per capita. While the historical dataset does contain an example of a similarly ‘frugal’ country with respect to industrial FEC per capita (Indonesia), this low growth of industrial FEC must be seen in the context of the projected growth in urbanization in India, from about 33% to more than 50% by 2050 (UN DESA, 2018). Urbanization will necessitate substantial increases in the consumption of energy-intensive materials like cement and steel (Hall et al., 2020). Moreover, the results for WITCH in the moderate and strong policy scenarios assume a more than halving of the level of 2050 industrial FEC per capita relative to the historical level of 2018, resulting in an absolute decline in industrial FEC by 2050 relative to the historical level. Given the high technical efficiencies of India’s industrial facilities (IEA, 2021) and the challenge of electrifying energy-intensive industries, this is a significant assumption.

More in line with historical trends in Latin American comparator countries (represented by Argentina) and non-East Asian emerging countries (represented by Turkey) are a group of models consisting of IMAGE, MESSAGE and AIM/Enduse. These scenarios see industrial FEC growing more slowly than GDP: relative to the historical 2018 India data in red, the 2050 scenario data for these models shifts more to the right than it does upwards.

4.3.3. Transport final energy consumption

From Figure 5, we see that all models, except REMIND and India MARKAL, project a level of transport FEC per capita in 2050 that is substantially below the level of countries at comparable levels of GDP. For example, the projected level of GDP per capita in AIM/Enduse and MESSAGE is comparable with that of Argentina in 2018, while the projected level of transport FEC is less than half. Even more striking are the results of WITCH and AIM/CGE, which project more than a doubling of GDP per capita by 2050, but essentially no increase in transport FEC per capita even in a no policy scenario. It is possible that India’s population density and messy urbanization process could lead to a more ‘frugal’ pattern of growth in transport FEC than that seen in less densely populated emerging countries (see the historical data for Argentina, for example). However, projecting zero growth is a strong assumption. In comparison to the other models, the higher energy intensity assumed in REMIND and India MARKAL leads to projected levels of transport FEC consumption per capita broadly consistent with that seen in historical comparison countries (see, for example, Indonesia, China and Turkey).

Figure 5. Transport FEC per Capita, India scenarios for 2050 versus historical cross-country benchmarks.

4.3.4. Buildings final energy consumption

Turning to the buildings sector (Figure 6), all the models project a very ‘frugal’ path for buildings energy consumption, with projected per capita levels of buildings FEC generally substantially below the historical cross-country benchmarks. For example, in 2018 Indonesia consumed about 12 GJ in buildings FEC per capita, which is above the projected 2050 level for India in five out of the seven models, even as these models project levels of GDP per capita above the level of Indonesia today, in some cases by as much as a factor two (see AIM/Enduse, for example). The example of Indonesia is relevant given climatic similarities with India.

Figure 6. Buildings FEC per Capita, India scenarios for 2050 versus historical cross-country benchmarks.

It is important to understand the current extremely low levels of buildings service demand in India: according to the most recent sample survey, just 24% of urban and 16% of rural households possessed a fan; 6% and 1.6% an air conditioning unit; 5% and 0.7% a washing machine; and 10% and 2.5% a fridge (NSSO, 2013). The International Energy Agency projects around 1350 TWh of cooling demand by 2050, which equates to about 3 GJ per capita just for cooling (IEA, 2018). It is thus an open question as to what extent the countervailing forces of biomass transition and improved energy efficiency would mitigate income-driven increases in services demand to subdue growth in buildings energy demand. The only way to really answer this question is with an explicit representation of equipment stocks and efficiencies (de la Rue du Can et al., 2019).

4.4. Energy mix

4.4.1. Final energy consumption

Showing the carbon intensity of FEC across the projection period, Figure 7 allows us to examine the process of ‘carbonization’ of FEC, that is, the endogenous energy transition seen historically as countries shift away from traditional biomass and towards higher carbon energy carriers, notably coal and oil.

Figure 7. Carbon intensity of FEC in global versus national models and FEC Mix in 2010 and 2050 in moderate and strong policy scenarios.

Note: CI = carbon intensity. In the top two panels, solid lines represent moderate policy scenarios, dashed lines represent strong policy scenarios.

Global models project a much lower propensity to ‘carbonize’ FEC than national models. They have a substantially lower level of the carbon intensity of FEC than national models in the base year, which we discuss further below, and a much lower level of increase in the carbon intensity of FEC, even in no policy scenarios. Moreover, global models are more optimistic as to the potential to decarbonize FEC in strong policy scenarios, displaying a far greater and earlier decline in FEC carbon intensity across the projection period in strong policy scenarios versus moderate or no policy scenarios.

The very different base year values and projected trajectories envisaged for the carbon intensity of FEC warrant further investigation. Figure 7 also displays the final energy consumption mix in the most recent base year 2010², and for 2050 in moderate and strong policy scenarios. There are surprising differences between models in 2010 data, even for commercial energy sources for which energy balance data based on actuals is clearly available. The different 2010 data for biomass may be more justifiable, given that this is a modelled not an observed datapoint, as no comprehensive statistics are gathered on traditional biomass use. Given that residential biomass enters into the denominator in the carbon intensity of FEC, model differences in historical data for residential biomass may explain some of the differences in historical values seen in Figure 7.

In 2050, in the moderate policy scenarios, with the exception of MESSAGE and to a lesser extent IMAGE, no global model displays substantial consumption of final coal. This is odd, given India’s still large process of urbanization and expected demand growth for cement and steel. This lack of final coal consumption combines with an apparent ‘pessimism’ regarding the transition away from biomass in FEC in most global models: these two factors drive the much lower projected carbon intensity of FEC seen above. In comparison, national models display proportionally larger shares of final coal and in the case of India MARKAL of final liquids too. It is noteworthy as well that in strong policy scenarios, consistent in global models with limiting warming to less than 2°C, global models essentially eliminate final coal from the consumption mix.

4.4.2. Electricity generation

Global models are more optimistic on the decarbonization of electricity generation than national models, as well as differing in terms of the base year carbon intensity of electricity generation (Figure 8). This holds for endogenous transition in low or moderate policy scenarios, as well as for strong policy scenarios. In the strong scenarios of global models, India would reach essentially zero carbon or even carbon negative electricity generation by the 2040s, presumably through the application of Bio-Energy with Carbon Capture and Storage (BECCS). In comparison, in the no policy scenario of India MARKAL, the carbon intensity of electricity supply is essentially flat across the projection period. Both national models project a residual carbon intensity of about 50 tCO2 / TJ of electricity (about 180 gCO₂/kWh) in strong policy scenarios by 2050.

Figure 8. Carbon intensity of electricity generation.

Note: solid lines represent moderate policy scenarios, dashed lines represent strong policy scenarios.

4.4.3. Primary energy supply

Figure 9 shows the projected carbon intensity of total primary energy supply (TPES) across the projection period. Logically, a similar pattern is evident as in Figures 7 and 8: global models are much more optimistic on the potential to decarbonize TPES than national models. The carbon intensity of TPES generally increases less across the projection period in no policy scenarios in global models, and falls faster and further in strong policy scenarios, compared with national models. This is a consequence of the results of global models both on the demand and supply sides, with a lower ‘carbonization’ of final energy consumption across the projection period in no policy scenarios (Figure 7), and stronger endogenous decarbonization of electricity supply (Figure 8).

Figure 9. Carbon intensity of TPES in global versus national models and FEC Mix in 2010 and 2050 in moderate and strong policy scenarios.

Note: in the top two panels, solid lines represent moderate policy scenarios, dashed lines represent strong policy scenarios.

Figure 9 also shows the primary energy supply mix for 2010 and 2050 in the moderate and strong policy scenarios by model. The same differences in historical data noted above are evident again, with unexplained differences, for example, in the large presence of ‘other’ energy in the national models. Global models have a higher share of biomass, both in FEC seen in Figure 8 and in TPES seen in Figure 9. Some global models also envisage a substantial role for nuclear (notably IMAGE and MESSAGE). WITCH, IMAGE and AIM/CGE envisage very little role for wind and solar, which by contrast play a substantial role in REMIND, India MARKAL and AIM/Enduse. Global models appear much more optimistic on the transition away from coal in strong policy scenarios, through a combination of zero carbon supply and demand reduction.

5. Scenario summary

Following the discussion of the preceding sections, Figure 10 provides an attempt to unpack the key drivers of the results in the moderate policy scenarios. The left panel displays the assumption on the year-on-year GDP growth rate in the model from 2010 to 2050 (x-axis) and final energy intensity of GDP (y-axis). The right panel shows the carbon intensity of electricity supply and final energy consumption. The bubble sizes and colours represent the resulting energy CO₂ emissions (the WITCH model is missing as it does not report energy emissions separately from energy and industry emissions).

Figure 10. Scenario summary, 2050 energy and CO₂ indicators, moderate policy scenarios.

The preceding sections have highlighted that divergent scenario results are determined by explicit and implicit assumptions on alternative underlying socio-economic pathways, and that these are insufficiently captured by the simple metric of climate policy stringency. To further the use of scenarios for policy-making, scenario producers need to relate disparate bodies of knowledge – notably those related to the nexus of development economics and energy systems – to give the user sufficient information to assess the scenarios within a causal and normative framework. To do so, broader socio-economic scenario storylines can help to relate different knowledge domains, theoretical frameworks and historical experience, and normative policy objectives into a consistent heuristic framework for scenario assessment (Caetano et al., 2020; Schmidt-Scheele, 2020).

To illustrate the potential of this approach, we attempt to develop socio-economic storylines below, for the scenarios discussed here. These storylines are not those provided by the scenario producers, but rather they are the result of our own illustrative exercise. To do so, Table 2 groups models according to elements of the Kaya identity,³ and suggests an overarching ‘storyline’ for each.

• Energy profligacy: REMIND is consistent with a storyline wherein very high energy intensity drives high emissions growth, despite low GDP growth and moderate carbon intensity on both the supply (electricity) and demand side (FEC). Its plausibility depends on the likelihood of future technical change and policy effort that simultaneously result in the low carbon intensity of energy and high energy intensity of GDP. Since India has historically been energy frugal due to its services-driven growth, this scenario also appears to imply a structural break towards an energy-intensive economic structure.

• Two-speed India: AIM/CGE is indicative of a storyline wherein low GDP growth and very low carbon intensity of final energy consumption drive low levels of emissions. Based on a deeper dive into the model results (Figure 7), low carbon intensity appears to come from an assumption of substantial levels of both biomass and electricity in final energy in 2050. This suggests the label ‘two-speed India’ – an energy system combining substantial persistence of rural biomass along with urban electricity demand, driven by continued divergence in rural-urban incomes and leading to relatively low GDP growth. The plausibility of this scenario rests on the long-term prevalence of rural stagnation and hence continued biomass use. This scenario implies continuity in terms of the rural-urban divide, but a dramatic slow-down from India’s recent growth performance.

• Middle-of-the-road: Image and Message are placed in this ‘middle–of-the-road’ storyline in that low/moderate GDP growth, moderate energy intensity and moderate carbon intensity of energy drive moderate levels of baseline CO₂ emissions in 2050. The slowing of GDP growth compared to India’s historical performance but consistent with the performance of non-East Asian middle-income countries, suggests no structural break with historical cross-country trends, but hints at the tightening grip of a middle-income trap.

• Energy frugality: AIM/Enduse appears driven by energy frugality – low levels of energy intensity, with low energy demand growth particularly in transport and buildings. However, this combines with a high carbon intensity of energy to drive a low-to-moderate level of emissions. The plausibility of this scenario depends on understanding how and why very low energy intensity coexists over time with relatively high carbon intensity. This scenario implies continuity in terms of India’s recent relatively frugal development pattern, combined with assumptions on continued relatively rapid economic growth.

• India becomes South Korea: the results of India Markal are driven by massive GDP growth that overwhelms a very low level of energy intensity, and combines with high carbon intensity, to produce very high energy CO₂ emissions. This storyline requires assuming a combination of a structural break in India’s development pathway towards an industry-intensive model, an extremely rapid pattern of economic growth (indeed world-beating in a historical cross-country comparison), as well as divergent pathways in the energy transition in terms of energy intensity (rapid improvement) and carbon intensity (slow improvement).

Table 2. Scenario grouping and storylines.

Model	GDP	Energy Intensity	Carbon Intensity	Storyline
REMIND	Low	Very high	Low	Energy profligacy
AIM CGE	Low	Moderate	Low	Two-speed India
Image	Low	Moderate	Moderate	Middle-of-the-road
AIM Enduse	Moderate	Very low	High	Energy frugality
Message	Moderate	Low	Moderate	Middle-of-the-road
Markal	Very high	Very low	High	India becomes South Korea

6. Conclusions: interpreting the scenario results

This paper has attempted to bring together two strands of analysis: a discussion of India’s recent socio-economic development pathway and its peculiarities (Section 2), and a discussion of recent modelling studies of India’s emissions pathway (Section 3). The reason for combining these strands is that the models seem to be envisaging very ‘different Indias’ by 2050, with important relevance for energy and climate policy-making. As the discussion above shows, aggregate scenario emissions differ by an order of magnitude among scenarios of apparently similar policy stringency. We show that independent of climate policy, implicit and explicit assumptions on the broader socio-economic pathway are substantially driving these results. Interpreting and generating policy-relevant insights from these results therefore requires an analysis of the assumed socio-economic development pathways underpinning the energy and emissions trajectories.

From a policy-maker’s perspective, it is one challenge to implement stringent climate policy in a 2050 India that has an economy and energy system structure similar to that of South Korea and China today, suggestive of a massive, industry-intensive rise in energy demand (e.g. India MARKAL). It is quite another challenge to implement stringent climate policy in a 2050 India with barely any energy demand growth (outside of population growth) in either industry or transport (e.g. WITCH and AIM/CGE). The technology and innovation options, and the policy effort required, to shift these two contrasting socio-economic pathways towards lower emissions would be quite different.

It is misleading to group such contrasting scenarios together under a common scenario nomenclature, differentiated only by the level of climate and energy policy stringency. Moreover, it limits understanding from a policy perspective. The literature arising out of these scenarios fails to adequately explain the socio-economic assumptions that drive scenarios. For example, in Schaeffer et al. (2020), the driver of the order of magnitude difference in 2050 FEC between the two India models, AIM/Enduse and India MARKAL, is described briefly as follows: ‘models differ mostly due to different GDP growth assumptions’. We are not, however, provided with any description of what these GDP assumptions were, what socio-economic storyline underpinned them, and how they are positioned relative to macroeconomic theory and history. The policy-maker is thus at a loss as to how to derive insights from these scenarios.

The objective of model analysis should indeed be to explore divergent and alternative futures, rather than trying to predict a single, most likely future. But for a policy-maker, facing decisions under uncertainty, it is not sufficient to present the full range of ‘possible’ outcomes, without also providing a sense of causality, probability and desirability: i.e. what would it actually take for the envisaged socio-economic pathway to unfold, both in terms of exogenous policy and the endogenous dynamics of the system (Schmidt-Scheele, 2020)? Part of the answer rests in transparency, but it is useful to go beyond this broad statement to specify just how modelling can best help generate insights and not just numbers (Huntington et al., 1982). Below, we offer a few suggestions.

6.1. The need for scenarios of socio-economic development pathways

As a developing country, India’s future is still being determined and is uncertain. Assumptions made about the rate and structure of growth, pace and material intensity of urbanization, divergence in urban and rural welfare and energy service demands have a huge impact on the future energy pathway, prior to climate and energy policy. Too often, these assumptions are made implicitly, and revealed only in variables like the structure of FEC and the elasticity of sectoral FEC to GDP growth. How a country’s development pathway unfolds is critical to understanding the challenge of reaching climate targets, including for India, both in terms of the quantitative level of effort required, as well as the qualitative kinds of effort. Contrasting and clearly discussed storylines of socio-economic development pathways are required to explore these possible different futures, and their consequences for the energy system and emissions. We have attempted to ‘add-on’ some socio-economic storylines into the existing energy scenario ensemble for India as an illustration of this process (Section 5), but ultimately this task of providing a heuristic and multi-domain scenario storyline should be undertaken ex ante by scenario developers, and ideally in collaboration with scenario users such as policy-makers (Schmidt-Scheele, 2020).

6.2. Linking socio-economic storylines to sectoral dynamics

This paper has shown that just as much as the quantity of GDP growth, the underlying macroeconomic structure behind a trajectory determines the dimensions of the energy system. This thus structures the challenge of decarbonizing the energy system. The scenario ensemble includes scenarios with similar levels of aggregate GDP but very different sectoral structures, FEC and fuel mixes; these different features of the scenarios lead to different implementation challenges for similarly ‘stringent’ climate policy pathways. Thus, it is necessary not just to develop socio-economic scenarios in terms of aggregate indicators like GDP, but also to link those scenarios better to the sector dynamics of energy demand and supply.

As discussed in this paper, a large part of the uncertainty in India’s energy system pathway relates to the demand side. How much steel and cement will its projected low level of urbanization require in 2050? How will India’s pattern of urbanization impact transport energy demand? How will the projected continuation of a large rural population influence urban–rural divergence in economic welfare and hence energy service demand? Should one expect a convergence based on ‘economic urbanization’ even in ‘nominally’ rural areas? These questions are of macro-scale importance in determining India’s energy future, yet they receive too little attention in the modelling literature.

Ultimately, understanding the policy implications of these scenarios requires shifting from discussion of abstract units of energy used and CO₂ emitted, into details of the physical world. How many passengers and ton kilometres? How many tons of steel and cement? How many air conditioning units per rural and urban household? It is only at this level of detail that one can make an informed and informative link between macro-scale socio-economic scenarios and the actual drivers of energy service demand.

6.3. A critical unpacking of model results

To develop insights for policy-makers, not only must scenario storylines and numerical assumptions be transparent – as input to the modelling, but results – or output from the modelling – should also be critically analysed by scenario producers to elucidate assumptions that may be of interest to the policy-maker. For example, some models studied here show the possibility of a low-industry, low-transport demand development pathway, largely due to assumptions about the sectoral elasticities of energy demand to income and price. Implicit in this result is a ‘vision’ of a country that is growing its GDP without growing demand in sectors that have, historically and in cross-country comparative perspective, seen strong links between GDP and demand growth.

India could theoretically achieve this development future by importing its industrial material requirements (although this would seem undesirable from a policy-maker’s perspective, and challenging for a bulky commodity like cement). Or India’s messy urbanization could restrict future private transport demand growth. Whatever the causal mechanism in the model, the driver behind the model result should ideally be interpreted for the policy-maker and its implications for climate policy and other socio-economic goals explained (Huntington et al., 1982).

Moreover, a more rigorous unpacking of model results could also encourage modellers to be more discriminating with their own scenarios and results. Several of the scenario results discussed here invite questions about their likelihood, given empirical and theoretical knowledge of the system dynamics in play (Schmidt-Scheele, 2020). Modelling studies that focus on climate policy scenarios per se, and do not pay sufficient attention to the underlying dynamics of the broader socio-economic system lose an opportunity for richer, more useful and possibly more robust policy insights. An integrated framework, encompassing careful definition of the socio-economic scenario, soft-linked macro-scale and micro-scale numerical assumptions, and a structured analysis of model outputs can help to contextualize results, and help to catch and adjust for less plausible outcomes.

6.4. This is an important agenda for Indian and global modelling

India is a perfect case study of the need to integrate energy system and climate modelling with careful scenario-building of socio-economic development pathways. This is because India’s pathway has been historically unusual and because India still has so much development ahead of it. This paper shows that how India develops is as critical to climate mitigation as is the stringency of domestic and global policies. And because India is so critical to the future of global energy demand and emissions growth, understanding its potential future development pathways, and their implications for climate mitigation, is also of global importance. This suggests an exciting interdisciplinary future research agenda that brings together macro- and development economics, energy system research and modelling, and national and global perspectives.

Data availability statement

The scenario data used in this paper are available at https://data.ene.iiasa.ac.at/cd-links/.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Navroz K. Dubash’s contribution to this paper was supported by a grant from the Oak Foundation (OCAY-19-247).

Notes

1 WITCH also endogenizes GDP growth, resulting in a small reduction in GDP and hence energy demand in strong policy scenarios compared to moderate or no policy scenarios.

2 The CD-LINKS project kicked off in 2015, and hence taking 2015 as the most recent historical year may not be reasonable, therefore 2010 was chosen. Although this may not be the model base year, depending on which year the model in question has been calibrated to.

3 The Kaya identity describes total emissions as the product of: Population * GDP per capita * energy intensity of GDP * carbon intensity of energy supply. See, for example, Yamaji et al. (1993)