1 Introduction

Ever since the Intergovernmental Panel on Climate Change (IPCC) released its third climate assessment report in 2001, addressing climate change has emerged as a worldwide concern that aims to achieve a harmonious equilibrium between development and sustainability on a global scale [1]. The carbon dioxide (CO2) emissions from the industrial sectors are one of the major driving forces leading to the severity of global climate change accelerated through climate warming [2]. Consequently, the environmental impacts of the industrial sector have emerged as a progressively significant subject of global public discourse. CO2 emissions from energy combustion and industrial activities increased by 6% in 2021, reaching 36.3 gigatonnes in 2021 [3]. Subsequently, the industry is the driving force behind economic growth, a catalyst for innovation and entrepreneurship, and a significant source of direct and indirect employment. Considering its priority, the United Nations (UN) has also emphasized the Sustainable Development Goal (SDG) 9 that focuses on “build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation.” Simultaneously, the UN highlights SDG 13, which focuses on immediate action to reduce climate change and its consequences [4]. Thus, achieving a low-carbon economy by balancing environmental-industrial growth has become a significant global challenge, especially for emerging countries like India. India is a rapidly expanding economy with a strong rate of economic growthFootnote 1 [5, 6]; rapid industrialization and globalization have driven the energy demand up, doubling in the last few decades [7]. This has highlighted India’s carbon emissions levels, which are contributing significantly to global warming.

India is the third-largest CO2 emitter in the world, contributing 3571 m tonnes of CO2 emissions, i.e., an increase of over threefold since 1970 [8]. Again, the World Bank addressed that “India could see $1.2tn of lost GDP, plus lower living standards for nearly half of its population by 2050, compared to a scenario with no climate change” [9]. Additionally, carbon emissions contribute to climate change, increasing the likelihood of adverse health effects on humans, particularly in growing economies like India [10]. Further, reference [11] mentioned that the rising temperatures resulting from increasing carbon emissions and climate change hurt India’s agriculture and agricultural workforce. Considering the gravity of this issue, the government of India has taken several initiatives, such as the Energy Conservation Act of 2001, to provide a legal mandate for implementing energy efficiency initiatives. Similarly, the National Mission on Enhanced Energy Efficiency (NMEEE) is a part of the National Action Plan for Climate Change (NAPCC) and consists of many projects aimed at improving energy efficiency. In addition, India has also announced the Intended Nationally Determined Contribution (INDC) to reduce the emission intensity of its GDP by 33–35% by 2030 from the 2005 level [12]. However, despite the number of flagship initiatives, India has yet to achieve its carbon-saving potential.

Energy use in Indian industries has significantly increased, accounting for 40 percent of total final energy consumption. India accounted for 5.3 percent of the worldwide primary energy consumption in 2015, making it third position in the worldwide, surpassing Japan after China and the USA. The annual increase in average energy usuage is anticipated to be 4.2 percent by 2035. As a result of its substantial energy consumption, the economy is generating carbon dioxide emissions around 2.47 billion tons in 2015, an increase of 5.1 percent compared to 2014 [13]. The trend of GDP and severity of CO2 emissions in India can be well noticed from Fig. 1. The manufacturing industry accounts for 50 to 60 percent of commercial energy use and remains the highest carbon emission intensity industry in the industry sector [14]. The trend and pattern of energy consumption and CO2 emission resulting from from the use of fuels in India can be well noticed in Fig. 2. The principal energy-intensive manufacturing sectors in India consist of paper, fertilizers, iron, steel, chemicals, textiles, aluminum, and fertilizers. More than sixty percent of the energy consumed in the industrial sector is attributed to these sectors. The importance of manufacturing output and the complexity of CO2 emission intensity in India can be seen in Fig. 3. Thus, it has become essential for driving economic growth in a productive and environmentally conscious manner. To maintain competitiveness in a globalized marketplace and surmount the obstacles presented by escalating energy costs and more exacting pollution control regulations, it is imperative that all industrial facilities, regardless of the size of their facilities, exercise strict control over their carbon dioxide (CO2) emissions. Addressing the solution to the problem of high energy consumption, which leads to a high level of CO2 emissions, many studies have advised low energy consumption and a means for a reduction in carbon emission that could simultaneously enable the adoption of advanced technologies [15,16,17]. Additionally, existing research has explored how the use of improved technology can reduce the energy intensity and CO2 emissions level of industrial sectors [18,19,20].

Fig. 1
figure 1

Source: World Development Indicators. The World Bank Group, 2021; International Energy Agency (IEA), 2021

Trend of CO2 emissions and GDP of India.

Full size image
Fig. 2
figure 2

Source: International Energy Agency (IEA), 2021

Trend of Energy consumption and CO2 emission from fuels.

Full size image
Fig. 3
figure 3

Source: World Development Indicators. The World Bank Group (2021); International Energy Agency (IEA), 2021

Trend of Carbon intensity and manufacturing output of India.

Full size image

According to the United Nations Environment Programme [21], technological progress is a powerful instrument in achieving the Intended Nationally Determined Contribution (INDC) to tackle GHG emissions following the United Nations Framework Convention on Climate Change (UNFCC). Considering the state of the environment and technical advancements [22], maintained that energy-efficient technology has emerged as one of the most affordable solutions for addressing the problems of high energy costs, energy accessibility, climate change, and sustainable economic growth. Concentrating on technological advancements to handle the dual objectiveFootnote 2 of reducing CO2 emissions, [16] has highlighted that technological progress in the manufacturing sector is necessary to mitigate the adverse effects of energy use. Since the 1990s, the influence of technical progress on reducing CO2 emissions has been studied widely [18, 19, 23,24,25]. In the Indian context, literature has reported that technical efficiency in terms of energy-efficient technology helps to reduce industrial emission levels with a sustainable amount of growth [26,27,28]. In line with the energy-efficient technology [29], has highlighted that implementing new technology might cause a high cost of electricity consumption. Similarly, the reference [30] has argued that increasing energy efficiency does not guarantee reduced energy use. The existing studies on the relationship between technical efficiency and emissions have been limited to the influence of technical efficiency at the emission level. Very little attention has been given to the polluting effects of socio-economic activities [31, 32], and the studies are majorly focused on the cross-country level. However, industrial analysis at the regional level has not been explored much. As industry-level CO2 emissions information is not at a level, few studies have analyzed the Indian manufacturing industries’ CO2 emissions aspect. Next, hardly any study focuses on the impact of CO2 emissions on the technical efficiency of Indian manufacturing industries. As CO2 emissions are a significant by-product of industrial energy consumption, any study investigating technical efficiency and CO2 emissions intensity in India could provide a better picture of the industries’ nexus between CO2 emissions and technical efficiency. Therefore, this paper undertakes research to fill the gap related to changes in these industries’ technical efficiency that will significantly influence carbon emissions.

Thus, this paper has two broad objectives: first, to measure the technical efficiency and carbon intensity at the 3-digit industry level in Maharashtra, Odisha, and all-India manufacturing industries. Second, is to ascertain the impact of CO2 emissions intensity on the technical efficiency of the Indian manufacturing sector in these three cases. The novelty of this paper lies two-fold. First, this research computes the technical efficiency and CO2 emission intensity for 3-digit Indian manufacturing industries for Maharashtra, Odisha, and all-India manufacturing industries. The rationale for selecting the states is that following the Annual Survey of Industries (ASI) [33], these two states can be proxies for high-industrialized and low-industrialized regions, respectively. A comparative assessment of these states, along with the aggregate performance of Indian manufacturing industries, would help the industrialists and policymakers to synthesize the technical efficiency and CO2 emissions intensity performance at different regions with different levels of technology. Second, the paper examines the impact of CO2 emission intensity on technical efficiency with panel regression analysis. Thus, this study could strengthen the policymakers’ actions to control CO2 emissions and improve the technical efficiency of manufacturing industries.

The subsequent sections of this work are structured in the following manner. Section 2 offers a concise overview of both theoretical and empirical literature. Section 3 explains the data, variables, and methods. Section 4 provides a detailed account of the empirical findings and subsequent analysis. Section 5 provides the final analysis and discusses the policy implications.

2 Review of literature

This section briefly reviews theoretical and empirical literature to study the linkage between technical efficiency and environmental quality.

2.1 Technical efficiency and carbon intensity: theoretical background

The Green Solow Model (GSM) brought the theoretical linkage between technology and environmental quality [34]. It combined the Environment Kuznets Curve (EKC) and the Solow and Swan model [35, 36] and examined the empirical relationship between income, pollution, and abatement cost. Grossman and Krueger [37] created the EKC theory, suggested that economic growth and environmental quality had an "inverted U"-shaped relationship, and stated that as income rises, emissions will eventually decrease. However, the EKC hypothesis remains unanswered during its application with three major limitations [38]. First, the EKC hypothesis based on a square form of heavy reduction in ongoing emission intensities under fixed pollution abatement costs remains unexplained. Second, the hypothesis remained silent in explaining the ‘humped-shape’ picture of emission levels when plotted against per capita income across time. Third, the empirical results of the EKC hypothesis remained inconclusive with different samples and methodologies.

The GSM model provides a simple solution that, with technological progress, emission intensity can be pushed descending steadily [34, 35]. This explains that with technological progress, the economy can reduce the emissions per unit of output. However, the aggregate emissions rise and approach towards balanced growth, Subsequently, the overall emissions begin to decrease, while the emissions associated with each unit of production continue to decrease as well [38]. It is comparable to the new growth theory model of “pollution augmenting technological progress” to use this technological progress abatement model [39]. However, the GSM is not free from limitations [40]. First, it does not clearly express the role of technology in the production process and its effect on aggregate emissions. Second, it does not differentiate the different forms of technology and the differentiated stages of its development. Thus, the theoretical linkages in this relation between technology or technical efficiency and emission level or CO2 emission intensity are complex and demand empirical support from methodology and its findings.

2.2 Empirical literature

This section gives two sub-sections. The first part reports a brief review of technical efficiency and carbon intensity measurement. At the same time, the following section explains the Indian literature on the relationship between technical efficiency and carbon emission intensity, which is mainly centered on energy efficiency literature.

2.2.1 Measurement of technical efficiency and CO2 emissions intensity

The two principal methodologies used to estimate the technical efficiency of any economic production unit are the stochastic production function approach (SFPFA) and the data envelopment analysis (DEA). The measurement of SFPFA estimation has been based on the distance from the frontier and to differentiate the effects of noise from inefficiency and simultaneously developed by [41] and [42]. Thus, stochastic frontier analysis is a parametric approach that conjures three components: the deterministic production function, the idiosyncratic error, and the inefficiency components [43]. This approach has multiple benefits, including the ability to account for the unpredictable variations among organizations that are not directly observed, and the ability to verify the statistical significance of the variables that determine efficiency through statistical tests. However, the methodology has drawbacks as it necessitates making assumptions about the distribution of the two error components and assuming independence between the error terms and the regressors. Additionally, the optimal outcome of this approach depends on selecting a specific functional form, the suitability of which raises concernsFootnote 3 [44].

Secondly, the framework for the non-parametric frontiers is derived from the original work of [45]. It has been extended by Charnes, Cooper, and Rhodes [46] and further generalized by Banker, Charnes, and Cooper [47], giving rise to what is known as data envelopment analysis (DEA). The DEA is a non-parametric approach derived from a mathematical programming approach and helps estimate the distance function. This approach gives the efficiency score of units and helps to find the threshold efficiency, allowing the target values of input and output for inefficient units [48]. The DEA method has several advantages. First, it removes the necessity for arbitrary assumptions about the frontier’s functional form and the error term’s distributional assumption. Secondly, it allows multiple-output technologies to be examined very quickly without aggregation. However, the DEA does not account for the possible influence of measurement error and other noise in the data, thus producing biased estimates when any measurement error or statistical noise exists [49]. Considering both the merits and demerits of the DEA technical efficiency measurement, many literature have examined the technical efficiency of the Indian manufacturing sector. For instance, in the case of the sugar industries [50], the pharmaceutical firms [51], and the steel manufacturing industries [52, 53] have also examined the TE for various manufacturing industries in India.

The primary method for calculating the carbon dioxide emission intensity is to compare the amount of carbon emissions to the gross output at different levels (e.g., firm, plant, state, or aggregate) [54]. However, in calculating CO2 emission intensity for India, the crucial task is to compute the carbon emission at different disaggregate levels, such as firms and industry. This is because India has no direct information on CO2 emission available at disaggregating levels such as industry or firm level. However, data is available at the national level [55]. However, indirect measurement is available using a scientific approach produced by the IPCC from fossil fuels. The IPCC reference method uses a bottom-up approach using disaggregate-level information on fossil fuel consumed [14]. However, the limitation of this methodology is that it cannot quantify the quality of the disaggregated fossil fuel data (for instance, the quality of coal consumed) during the computation [55]. Despite its limitation, a set of existing studies have used this methodology to get the CO2 emission computed at industrial units in India (see references [10, 14, 26, 55, 56]).

2.2.2 Indian literature on technical efficiency and carbon intensity

The existing literature reported that the Indian manufacturing sector is the prime energy user and thus has wider policy implications than other sectors in concerns about climate change. However, studies conducted in this context mostly centered on energy intensity and efficiency. They were very limited in technical efficiency and energy intensity due to the unavailability of emission data at disaggregate. As per reference [57], developing countries should adopt advanced technology to increase their efficacy. They further advised governments to incentivize the private sector to create low-carbon technology. Much literature has emphasized that the economies of scale might have influenced the firms’ size with decreasing returns on energy use [58]. Similarly, [59] has highlighted a negative relationship between R&D dummy and energy intensity. However, [60] revealed no significant relationship between these two.

In this regard, [61] demonstrated that new investment and scale economies were helpful in reduction in the intensity of the energy at 3-digit industries among the Indian states. Similarly [27], and [62] have argued that R&D and technology can reduce CO2 emissions in Indian manufacturing firms. In addition [63], emphasized that the Indian iron and steel firms possess the potential to achieve a 50% increase in their energy efficiency. Therefore, companies must allocate more resources towards R&D, patenting efforts, and disembodied technology in order to enhance energy efficiency. Similarly [64], suggested that Indian manufacturing firms be sustainable and adopt cleaner energy to reduce the CO2 emissions from fossil fuel consumption. Over the long term, firms need to increase their investments in research and development as well as technology. In contrast, reference [30] has argued that there is no guarantee that increasing energy efficiency will result in a decrease in energy usage. However, these studies have not emphasized the direct relation between CO2 emission and technical efficiency. Extending further, there is hardly any study available in the Indian context to investigate the impact of CO2 emission intensity on the technical efficiency of manufacturing units. Therefore, this paper tries to minimize this gap threefold. First, the study follows a scientific methodology to measure technical efficiency. Second, the research computes the 3-digit industry-level CO2 emission from the physical units of fossil fuel consumption. Third, this study examines the effect of CO2 emission intensity on technical efficiency in the Indian manufacturing sectors.

3 Data, variable, and methodology

The purpose of this section is to analyze the data and variables, as well as the econometric methods utilized in this study.

3.1 Data description and data source

The first objective of this paper is to draw the trends and patterns of carbon intensity and technical efficiency of the 3-digit manufacturing in India, Maharashtra, and Odisha from 2001 to 2018. In the later part, this paper investigates the relationship between technical efficiency and carbon intensity across the manufacturing industries in the above three destinations during 2001–2018. For the empirical analysis, the data period is limited to 2001–2018 for two reasons: first, the lower limit is to 2001 as in this year, India took several energy efficiency measures (such as the Energy Conservation Act, 2001) to control the carbon dioxide emissions in the form of controlling energy consumptions. Second, the upper limit of the data set is restricted by the availability of data. The study uses data from the Annual Survey of Industries for this empirical analysis. The unit of analysis is at the aggregate all-India level, state level, and industry level. This study considers Maharashtra and Odisha for state-level measures because these two states have been taken as a proxy for high-industrialized and low-industrialized states, respectively. Maharashtra represents a highly industrialized state with 11.10 percent of the total number of factories and contributed 14.86 percent to output in 2017–18. Odisha represented a low industrialized state with 1.3 percent of total factories and 2.8 percent of output in 2017–18 [33]. Concerning industrial importance, the data covers 28 NIC 3-digit manufacturing industries for Odisha, 47 NIC 3-digit manufacturing industries for Maharashtra, and 54 NIC 3-digit manufacturing industries for the all-India level.

To examine the objectives of this paper, the study has two major variables: technical efficiency (TE) and carbon intensity (CI) of the Indian manufacturing industries. The research uses output, labor, capital, energy, and material in the empirical set-up for estimating technical efficiency. Output is measured as gross output. In the input variables, the study uses the number of workers for labor, fixed capital for capital, fuels consumed for energy, and materials consumed for the material are all examples of consumption. Except for labor, these output and factor inputs are stated in a manner consistent with reality by deflating them with their respective price indexes using the Wholesale Price Index (WPI) in the base year of 2011–12. The WPI series, used to deflate variables to constant prices, is provided by the Office of Economic Advisor within the Ministry of Commerce and Industry of the Government of India.

3.1.1 Measurement of technical efficiency

The Data Envelopment Analysis (DEA) approach is utilized to obtain the technical efficiency score. The DEA approach is a flexible and non-parametric tool that may be easily employed to evaluate industrial performance. The analysis of efficiency gains across different units that use resources in an economy or multiple economies [65]. In literature, conventional methods such as regression analysis and simple ratio analysis are considered [66, 67]. It is employed to assess the effectiveness of DMUs (Decision-Making Units) that utilize the input and output values. DEA forms a production possibility space by taking the observed input and output values, against which the individual units are compared to determine their efficiencies. Charnes, Cooper, and Rhodes (CCR) [46] model was the first model of DEA developed in 1978. CCR assumes a constant return to scale, i.e., an increase in input results in a proportionate increase in output. This model assumes perfect competition for DMUs to operate at the optimal level. However, the real world is subject to imperfect competition, financial constraints, regulatory measures, and other various other issues. This can cause in suboptimal performance of the DMUs. Hence, the assumption of variable returns to scale is better suited. It helps in determining efficiencies, by assessing whether changes in input or output lead to corresponding changes in outputs or inputs, respectively [68]. An inherent benefit of this strategy is its ability to operate without making any prior assumptions about the specific form of the production function. Therefore, the study can avoid the model setting problem. Simultaneously, the DEA can quantify several dimension indicators and employ optimization techniques to ensure that the evaluation results exhibit strong objectivity. The DEA model is extensively used for assessing technical efficiency and has emerged as the predominant and effective non-parametric method [69]. Therefore, for the current study, the DEA has been utilized to determine a standard measure of technical efficiency.

Here, the paper used the input-oriented BCC variant of DEA for a radial measure of technical efficiency. To formally define the model, the study assume a typical firm produces a single output y by employing m inputs \(X= {x}_{1}, {x}_{2}, \dots ., {x}_{m}\). Let there be n numbers of firms (j = 1, 2, …., n) to be evaluated, \({y}_{j}\) be the output and \({x}_{j}\) be the input bundle of the jth firms. The research can now specify the production possibility set:

For DEA, the study uses the following model to denote the technical efficiency of individual DMUs,

$$t\left(x\right)=\{\left(x, y\right)\in R:x \, can \, produce \, y\}$$
(1)

where T includes all feasible inputs and output bundle, the efficiency of jth DMU with the input–output bundle \({(x}_{0} {y}_{0})\) can be estimated using the following model.

$${\theta }^{*}=min \, \theta$$
(2)

Subject to

$$\sum_{j=1}^{n}{x}_{ij}{\lambda }_{j}\le {\theta x}_{j0}$$
(3)
$$\sum_{j=1}^{n}{y}_{j}{\lambda }_{j}\le {\theta y}_{0}$$
(4)
$$\sum_{j=1}^{n}{\lambda }_{j}=1$$
(5)
$${\lambda }_{j}\ge 0, j=1, 2, \dots .., n$$

The linear programming model above seeks to systematically decrease the quantities of all inputs in proportion to achieve a specific level of output. The major goal here is to create a production frontier that characterizes the production function in mainstream economics. Where x and y are the vectors of inputs and outputs, respectively. i is an index of inputs (capital, labor, material, and energy); j is an index of firms under consideration. λ is a scaler magnitude at which the firm’s inputs under evaluation must be scaled down to reach the frontier point. In contrast, θ measures the firm’s technical efficiency through the proportionate reduction in all the inputs.

3.1.2 Measurement of carbon intensity

Carbon dioxide emissions have been used as a proxy indicator for the undesirable outcome of the manufacturing activity. While there are numerous GHGs, the by-product of manufacturing activity, CO2 is used as a metric measure to calculate the Global Warming Potential (GWP) of other GHGs. Nevertheless, considering CO2 intensity as a more reliable indicator of the environmental impact of industries are preferable to solely focusing on the absolute level of CO2 emissions or its per capita value [70]. Thus, in this study, the research take CO2 emissions intensity (CI) measured through the share of CO2 emissions of total output as a proxy for environmental quality.

For this paper, the study have resorted to the Intergovernmental Panel on Climate Change (IPCC) 2006 methodology to estimate carbon emissions. This paper employed the Tier 1 method of the 2006 IPCC Guidelines, for which fuel-based and emission factors are available for all relevant direct greenhouse gases. Thus, all combustion sources can be estimated based on the quantities of fuel combusted collected from national energy statistics and average emission factors. The research uses coal, electricity, petroleum products, and other fuels as inputs and gross output as variables for calculating CI. For the other fuels, wood products are taken as a proxy. For coal and electricity, quantity is available. The values are given for petroleum products and other fuels, and they have to be converted to quantity by deflating it with the Wholesale Price Index (WPI) at the 2011–12 base. Each industry’s CO2 emission is calculated as each fuel’s aggregate carbon dioxide emission. The CO2 emission of each fuel is calculated by multiplying its quantity with the carbon emission factor. The carbon intensity of each industry is calculated as the ratio of carbon emission of each industry and the corresponding output. Therefore, for this study, CI is calculated as:

$${\text{CI}} = \frac{{CO_{2} }}{Total\;Output}$$
(6)

The data for output, labor, capital, energy, and material, used for estimating technical efficiency, and the activity data for fuels and output data of industries, used for calculating carbon intensity, are collected from ASI Volume I report based on the National Industrial Classification-2008. The petroleum, coal, and wood carbon emission factors are taken from the IPCC (International Panel for Climate Change) National Greenhouse Gas Inventory, 2006. The carbon emission factor for electricity is taken from the CEA (Central Electricity Authority), All India Electricity Statistics General Review report. The Wholesale Price Index (WPI) series, used to deflate variables to constant prices, is obtained from the Office of Economic Advisor, Ministry of Commerce and Industry, Government of India. The study takes three control variables: profit share (PS), the number of factories (NF), and time dummies in the empirical estimation to remain safe from the omission bias indicated by [71]. To tackle the sharpness of data and for better distribution of the dataset, the natural logarithmic transformation has been taken for all the variables; however, to deal with zeros and retain the negative values, the paper follows the semi-logarithmic form for the carbon dioxide intensity variable following [72]. The variables utilized in the empirical estimation, together with their corresponding measurement, definition, source, and expected signs are presented in Table 1.

Table 1 Names, definitions, sources and expected signs of variables
Full size table

3.2 Econometric strategy

This research employs two-panel regressions, specifically random effect (RE) and fixed effect (FE) models. The random effect mechanisim offers a benefit by taking into account both he time series and the cross-sectional aspect of the dataset. The RE method incorporates the random variables of interception through common indentities, allowing for accurate estimations even when there is a little fluctuation in time. Neverthless, the estimated effects of the models become distorted and unreliable when they are associated with certain explanatory factors [73, 74]. Therefore, it is imperative to do the Hausman test to identify the essential bias in the estimations. Suppose the test statistics indicates a rejection of the null hypothesis, the analysis favors the FE method over the RE, where the advantage of the FE estimator is that it allows each group to study different intersetions specific to unit [74]. However, references [75, 76], and [77] have addressed that second-stage regression analysis with OLS, Tobit, or dynamic panel regression could be misspecified and produce bias estimates. In contrast, a study has addressed that the correlation among the variables in the absolute forms between the independent variables and the DEA components does not exceed 0.6. This study can further support the second-stage regression, including these variables [78]. Additionally, a recent study has highlighted that if the correlation variables are between − 0.2 and + 0.4, there is no adverse impact on the estimated parameters of the second-stage models using OLS [79]. In addition to static panel regression analysis, the study uses Papke and Wooldridge’s fractional response model to control the bounded nature of the dependent variable (as the dependent variable lies between 0 and 1) [80]. Following the work of Ru and Si [81], Das et al. [82], and Pérez-Reyes and Tovar [79], the research applies the Fractional logit model as a robustness check for the second-stage regression analysis. The model is developed for panel data, follows the quasi-maximum likelihood estimator (QLME), and obtains a robust estimation method without an ad hoc transformation of the boundary values [82]. The result table also presents the average marginal effects of the fractional logit model, which can be interpreted as the probability of occurrence in percentage. However, the discussion mainly highlights the coefficients’ direction and significance level, which are crucial for policy measures [83].

For the second-stage regression, the empirical framework of this study has taken the following model;

$${TE}_{it}={{\alpha }_{0}+{\beta }_{1}CI}_{it}+{{\beta }_{2}NF}_{it}+{{\beta }_{3}PS}_{it}+{\mu }_{t}+{\varepsilon }_{it}$$
(7)

where TE represents technical efficiency, CI is carbon intensity, NF and PS are the number of factories and profit share, respectively, \({\mu }_{t}\) stands for the firm fixed effect, i = 1, … … n industries, t = 1, … … T years, \(\alpha and \beta\) are the intercept and slope coefficients, respectively, and \({\varepsilon }_{it}\) is the standard error with zero mean and constant variance.

4 Empirical results and discussion

This section shows the trends and patterns of carbon intensity and technical efficiency and their empirical relation in Maharashtra, Odisha, and India.

4.1 Trends in carbon intensity

To capture the pattern of carbon intensity across the manufacturing industries at the state level (for Maharashtra and Odisha) and the all-India level, the study depict the carbon intensity values of a sample industry, each representing high technological activity, medium technology industry, and low technology industry for the period of 2000–2001 to 2017–2018. For categorizing high, medium, and low technological industries, the study have followed the ISIC (International Standard Industrial Classification) Rev.3, categorizing industries according to their technology intensity based on research and development (R&D) activity. Accordingly, the high technology industry, ‘Manufacture of optical instruments and equipment for medium technology industry ‘Manufacture of rubber,’ and the low technology industry ‘Manufacture of paper and paper products have been taken as the sample industries (See Fig. 4a–c). In all three cases, carbon intensity has a downward trend in different sections for Maharashtra, Odisha, and the national level. However, the long-run plot shows a high spike during 2003–2005 in groups at the aggregate level. The carbon intensity shows a similar trend in all three cases, with Maharashtra lying above Odisha. Still, in the low-technology industry, Odisha’s emission level has increased from 2010–11 to 2017–18. It shows an increasing pollution trend even though the national emission level poses a slow decline over time.

Fig. 4
figure 4

a Carbon intensity pattern of the high-tech industry: manufacture of optical instruments and equipment. Odisha does not have a high-tech industry in the Manufacture of optical instruments and equipment. Source: Author's representation using data from ASI, Government of India. b Carbon intensity pattern of the medium-tech industry: manufacture of rubber. Source: Author's representation using data from ASI, Government of India. c Carbon intensity pattern of the low-tech industry: manufacture of paper and paper product. Source: Author's representation using data from ASI, Government of India

Full size image

However, this implies that the carbon intensity is declining overall. There might be two reasons for this: the carbon emission from manufacturing industries is decreasing. The second reason for the decline is the increase in output of these industries. The second reason is, however, more plausible. Since manufacturing industries have not significantly changed their fuel mix and continue to rely on carbon-based fossil fuels such as coal and petroleum products to meet their energy needs, the argument for a decline in carbon intensity due to decreased carbon emission is not strong enough. The manufacturing industries have seen an increase in output over the years. Hence, the observed decline in carbon intensity has been due to a rise in output, which is more significant than the decrease in carbon emission in the manufacturing industries.

4.2 Classification of industries based on technical efficiency and carbon intensity scores

The technical efficiency score of each industry is generated using DEA. The industries are then divided into high technical and low technical efficiency based on their technical efficiency scores. Based on their carbon intensity score, The industries are divided into high and low-carbon-intensity groups. Whichever industry has high carbon intensity has low energy efficiency since the energy required to produce a unit of output increases due to the high carbon intensity level. Therefore, the industries with high carbon intensity have low energy efficiency, and those with low carbon intensity have high energy efficiency. On this basis, industries are categorized into four groups: technical efficiency and energy efficiency (having low carbon intensity). The industries have been divided into four categories for each state: high technical efficiency and low carbon intensity; high technical efficiency and high carbon intensity; low technical efficiency and low carbon intensity; and low technical efficiency and high carbon intensity. The average score of four groups for each 3-digit industry at the state and national levels can be found in the supplementary material (See Table S1 in the supplementary material). The trend of Technical Efficiency and Carbon Intensity in Maharashtra, Odisha, and India can be well noticed in Fig. 5a, b, and c, respectively. The summary statistics of the technical efficiency and carbon intensity scores of four groups for Maharashtra, Odisha, and India are given in Table 2.

Fig. 5
figure 5

a Trend of Technical Efficiency and Carbon Intensity of Odisha. Source: Authors’ Calculation. b Trend of Technical Efficiency and Carbon Intensity of Maharashtra. Source: Authors’ Calculation. c Trend of Technical Efficiency and Carbon Intensity of India. Source: Authors’ Calculation

Full size image
Table 2 Summary statistics of technical efficiency and carbon intensity of Maharashtra, Odisha and India
Full size table

The average of high technical efficiency remains almost the same in Maharashtra and India (81.253 and 81.125, respectively), but Odisha’s mean performance of a high TE score is comparatively high (88.928). Similarly, the average performance of low CI, i.e., shows higher energy efficiency manufacturing industries, remains low in Odisha (910.736) compared to Maharashtra (839.819), but both remain low to the national average (5687.203). The average score for low technical efficiency is high at Odisha (72.539), followed by Maharashtra (61.562), and at the national level (59.951). Similarly, the average score of high carbon intensity is increased in Maharashtra (4131.78) compared to Odisha (2628.509), while both remain lower than the national average (18456.22). Table 2 also reports the standard deviation, skewness, kurtosis, and minimum and maximum values for the four groups of 3-digit manufacturing industries in Maharashtra, Odisha, and India.

4.3 Determinants of technical efficiency: role of carbon intensity

Before analyzing the results, the study reported the descriptive statistics and the coefficient of the correlation matrix of the variables used in the estimation. This will help to examine the behavior of variables in the analysis. Table 3 reports the summary statistics to show the mean, standard deviation, minimum, and maximum values of the variables used in the study and reports the correlation coefficient among the variables. The result demonstrates a statistically significant negative correlation between technical efficiency and carbon intensity in all three destinations. The study reports that the correlation of variables is between − 0.2 and + 0.4; thus, the estimated parameters have no adverse impact in the second-stage models used in the OLS.

Table 3 Descriptive statistics and coefficient of correlation between TE, CI, NF, PS
Full size table

Table 4 reports the study’s primary objective, which is to examine the effect of carbon intensity on technical efficiency. The table demonstrates the fixed effect estimates as the Hausman tests below are lower than 1% (but Model 1 reports it at 5%), which rejects the null and indicates that the fixed effect estimators are preferable to random effect coefficients. The table reports three models: Model 1 reports results for Maharashtra, Model 2 for Odisha, and Model 3 reports the results at the all-India level.

Table 4 Determinants of technical efficiency: role of carbon intensity
Full size table

Model 1 in Table 4 reports the result of the highly industrialized state of Maharashtra. The coefficient of carbon intensity has a negative effect on technical efficiency, but the coefficient remains statistically significant. The number of factories’ coefficient remains positive but has no significant influence on Maharashtra’s 3-digit manufacturing industries. However, the coefficient of share of profit shows a positive and significant effect on Maharashtra. Thus, the technical efficiency industries could be encouraged through high profits. Similar to the fixed effect model, the fractional logit model also shows a negative and significant coefficient of carbon intensity. This indicates that CO2 emissions reduce the technical efficiency of manufacturing industries in Maharashtra. However, the result reports that size adversely and substantially affects industries’ technical performance. Similar to the panel regression results, the coefficient of profit remains favorable and substantial in the fractional model. This implies that profit could enhance the technological performance of Maharashtra’s manufacturing industries.

Model 2 demonstrates the fixed effect estimates for the low-industrialized state of Odisha. The coefficient of carbon intensity has a substantial adverse effect on technical efficiency. This implies to increase technical efficiency, carbon intensity should be reduced. The coefficient of the number of industries remains insignificant and has no statistical effect on technical efficiency in Odisha. Regarding technological efficiency, the coefficient of profit has a strong and favorable effect. Thus, similar to Maharashtra, the manufacturing industries, in Odisha’s case, are positively motivated by the share of profit. After the panel regression analysis, the study verified the results using the fractional logit model. The result also reported that carbon dioxide emissions intensity negatively affects technical performance; however, the result is not statistically significant. Unlike the panel regression, the fractional response model reveals that the size of the industries negatively affects technical performance. However, the coefficient of profit share is also positive and significant in the case of the fractional model. This implies that higher profits could be a motivator for more technically sound Indian manufacturing industries.

Model 3 reports the aggregate level fixed effect estimates. In this model, technical efficiency is significantly affected by three independent factors. The coefficient of carbon intensity is negative and significant. Therefore, to increase technical efficiency, the carbon intensity must be reduced by increasing output or decreasing carbon emissions. The coefficient of the number of factories also negatively and significantly affects technical efficiency. This shows that the size of industries is negatively related to technical efficiency. This implies that technical efficiency decreases due to an increase in the number of factories or a decline in the competence of industries. Similar to the state level results, profit has a favorable and substantial effect on technical efficiency. Consequently, a rise in the proportion of profit results in a corresponding improvement in the technological efficiency of Indian industrial sectors.

The estimation results are also consistent with the fractional logit model. The result supports that technical efficiency is highly sensitive to carbon intensity. Hence, a decreasing carbon intensity would increase technical efficiency in highly industrialized states and benefit low industrialized states. Profit, irrespective of high or low industrial state, significantly contributes to technical efficiency, so profitable industries tend to be technically efficient in the case of Maharashtra, Odisha, and India. As the industry’s size negatively affects manufacturing industries’ technical performance, large-scale industries need to take more initiatives to sustain their technical performance with many factories.

5 Conclusion and policy implication

In the last few decades, research on environmental quality, energy intensity, and technological development has remained a central area. However, minimal studies have computed the technical efficiency and carbon intensity and examined their relation at 3-digit manufacturing industries in Maharashtra, Odisha, and India during the sample period of 2001–2018. This paper uses three-panel data at 3-digit industries by using fixed effect regressions. The paper computes the technical efficiency scores with the DEA non-parametric methodology for three cases. The study uses the IPCC-2006 Tier I methodology to compute carbon intensity. Based on the technical efficiency and carbon intensity scores, industries are divided into four groups: high technical efficiency and low carbon intensity, high technical efficiency and high carbon intensity, low technical efficiency and high carbon intensity, low technical efficiency and low carbon intensity for Maharashtra, Odisha, and India. The summary statistics, comparative technical efficiency, and low carbon intensity are highlighted for each state and India. The panel regressions and fractional logit model results show that carbon intensity influences the technical efficiency in all the cases. Additionally, the results demonstrated that the profit share is a positive instrument for increasing technical efficiency in all industries. However, the study has a few limitations; first, the study has ignored the other industrial characteristics while computing the relationship between technical efficiency and carbon intensity. Second, the research is limited to Maharashtra and Odisha and does not cover other states. Despite its shortcomings, this study can extend to future research by considering regional heterogeneity. It could be expanded beyond three locations and take into account more states. Studies can be extended to a more disaggregated level (such as 4-digit industry or firm-level) employing different econometric specifications. A comprehensive analysis of the impacts of different regional heterogeneities will help region-specific policy-making. Future studies should cover more factors affecting technical efficiency.

This paper suggests several policy implications for Indian manufacturing industries to increase technical efficiency by reducing carbon intensity. While the effect of individual fuels on carbon emissions is well-recognized by academicians and policymakers, an in-depth understanding of the emission estimation due to fuel activity is highly demanded. Thus, policymakers should consider the fuel mix structure of manufacturing industries while drafting carbon intensity reduction policies. Policymakers should consider the manufacturing industries’ geographical heterogeneity during the policy framing to improve technical efficiency. As the share of profit can increase industrial technical efficiency, an incentivized policy may be proposed at the industry level with a special focus on low-carbon technology. As reducing carbon intensity could be effective in increasing technical efficiency, policymakers should emphasize the role of states in making different regulations and initiatives to foster technical efficiency and to substitute a renewable and sustainable technology for a green and emission-free economy.