Decarbonizing Energy-Intensive Steel Production: Dynamic Analysis of CO₂ Emission Persistence in Poland’s Basic Oxygen Furnace Sector

Abstract

This paper analyses the factors that affect CO₂ emissions in the BF-BOF steelmaking process using a dynamic econometric approach based on annual data from the Polish steel industry. The analysis commences with the estimation of a baseline dynamic model that describes the relationship between CO₂ emissions in the industry and investment allocations, crude steel production, and lagged CO₂ emissions. The baseline analysis illustrates the dominant feature of strong emission level persistence and poor tracking of selected conventional production-related factors. The analysis proceeds by extending the baseline results through additional consideration of technological factors, material composition factors, and resource use factors in the generation of CO₂ emissions. The additional factors include the use of coke, electricity consumption, fixed asset value, and the scrap ratio. The analysis indicates that these additional factors are essential in improving the accuracy of the modeling process and in clarifying the significance of material composition in CO₂ emissions in particular. The analysis further illustrates the critical result that increased use of electricity leads to high CO₂ emissions in the BF-BOF process. Further analysis indicates that increasing the use of steel scrap leads to substantial CO₂ reductions in the BF-BOF route and other steelmaking technologies. The results also show that CO₂ emissions in the BF-BOF process depend not only on production volume, but also on material composition and the technological structure of the process. In the context of the WFESF project, these findings provide evidence-based guidance for metal industry research by identifying priority levers for mitigation, particularly through improvements in process technology and scrap-based material substitution.

1. Introduction

Over the past two decades, the steel industry in Poland has undergone a series of profound transformations. Steel mills have implemented new technologies in compliance with increasingly stringent EU directives, decommissioned obsolete installations, and modified numerous production processes to significantly reduce their negative environmental impact. In line with European climate policy, the Polish steel sector aims to achieve carbon neutrality by 2050, as outlined in the European Steel and Metals Action Plan (ESMAP) and the Clean Industrial Deal [1,2].

Decarbonization represents a major challenge for the steel industry across Europe. For this process to succeed, several key conditions must be met: access to large-scale renewable energy at competitive prices; the adoption of modern low-emission technologies; the development of electric arc furnace (EAF) technologies and hydrogen-based production processes; as well as the deployment of carbon capture and storage (CCS) and carbon capture, utilization, and storage (CCU/CCUS) technologies to either store or reuse captured CO₂.

It is well understood that the transformation process will be long-term; therefore, intensive work is required to develop a coherent strategy for change at multiple levels of the economy—governmental, local, and industrial—to mitigate the negative environmental impact of industries, including the steel sector, under current production processes. On 19 March 2025, the European Union introduced the document titled “A European Steel and Metals Action Plan” (ESMAP) [1]. In this document, the European Commission presented the current state of the steel sector within the EU and identified key actions to be taken in economic, market, social, legal, and financial areas that enable the European steel industry to carry out decarbonization and energy transition projects while maintaining its competitiveness in the global steel market.

Decarbonization and energy transition constitute the primary directions of change for many industrial sectors, including steel production. These two strategic pillars of industrial and climate policy require substantial financial investment in new projects, particularly in low-emission and energy-efficient technologies. The purpose of this publication was to present the results of an analysis of how technological investments implemented in the Polish steel sector affected its emission intensity. The author employed a dynamic emission model (with lagged CO₂ emissions). The use of such a model in emission research is particularly valuable because it captures the temporal dependence of industrial activity on CO₂ emissions. This type of approach is poorly represented in stationary models, which the author had previously examined [3,4,5,6].

The author found that emissions in the steel industry are driven not only by current levels of production and investment but also by past operational practices, technological inertia, and long-term structural or strategic frameworks for change. By incorporating the lagged value of emissions, the dynamic model accounts for the persistence of environmental impacts over time and provides a more realistic depiction of how technological investments influence emission trajectories. Such an approach enhances understanding of the gradual nature of decarbonization processes and underscores the importance of long-term planning compared to short-term interventions or isolated projects. The research gap addressed in this paper refers to the lack of adequate understanding of temporal dynamics and the persistence of emissions in the decarbonization process of the Polish steel industry. Most current analyses of CO₂ emissions are based on static econometric models that capture only direct and short-run relationships between technological investments, production volume, and emissions. The directions mentioned above have ignored the presence of inertia within industrial processes, especially in industries characterized by a long lifespan of technology (such as BOF steelmaking). Hence, previous studies did not focus on how patterns of production and investments in the past continue to affect current levels of emissions, which has resulted in incomplete estimation of the long-term efficiency of modernization policies. This study fills that gap by applying a dynamic econometric model with lagged CO₂ emissions to capture persistence effects and delayed responses of emissions to technological and economic variables.

A further research gap arises from the lack of empirical evidence specific to Poland’s metallurgical sector within the broader European decarbonization discourse. Though several international studies and industrial roadmaps have been developed under the IEA, OECD, and the Mission Possible Partnership, they set out technological pathways for emissions reduction in steelmaking but often do not take into account regional and structural constraints that are usual for post-industrial economies with coal-dependent energy systems. Typical in this respect is the case of the Polish steel sector, whose modernization faces significant constraints due to high energy prices, dependence on fossil fuels, and limited access to low-carbon technologies. The contribution of this paper to narrowing this empirical gap lies in providing a country-specific dynamic analysis based on national data for 2005–2022. Because the model includes lagged variables and is estimated only up to 2021, the post-2022 data are incomplete and lack several required parameters. In this framework, we present new insights into structural inertia and time-lagged effects that shape the decarbonization trajectory of Polish BOF technology.

The main objective of this study is to analyse the dynamic determinants of CO₂ emissions in the Polish BF-BOF steel sector, with particular emphasis on the role of investment-induced structural and technological changes.

In this paper, we formulated three research questions (RQs):

• RQ1: What are the key dynamic determinants of CO₂ emissions in the Polish BF-BOF steel sector, and to what extent are current emission levels shaped by persistence and technological inertia?
• RQ2: Do investment expenditures in BOF modernization exert a direct short-run effect on CO₂ emissions once emission persistence is taken into account?
• RQ3: Through which technological and structural channels—such as energy intensity, material composition (scrap ratio), and capital stock—do investments influence long-term CO₂ emission dynamics in the steelmaking process?

2. Background of Analysis

The metallurgical industry, along with other energy-intensive sectors, is subject to the European Union’s long-term strategic goals under the framework of the European Green Deal (European Commission, 2019) and the broader ambition to make Europe the first climate-neutral continent by 2050. As a major contributor to greenhouse gas emissions, the steel industry accounted for about 4% of total EU emissions in 2017 and 23% of manufacturing-sector emissions [7]. According to WiseEuropa, the steel industry is responsible for around 2.5% of CO₂ emissions in Poland, 5% in the European Union, and approximately 7–8% globally [8]. The European steel industry is a key sector within the New Industrial Strategy for Europe, published in March 2020, the primary directions of which are the green and digital transitions [9].

The development of the steel sector requires the decarbonization of production processes in accordance with the key challenges arising from European climate policy, market trends, and the strategic directions set in numerous legislative documents. In June 2025, the Polish Ministry of Industry prepared a draft entitled “Action Plan for the Sustainable Development of the Steel Industry in Poland.” This document reflects the structure and priorities of EU actions, which is understandable given Poland’s EU membership. It identifies five strategic areas for the transformation of the steel sector: (i) competitiveness, (ii) environment, (iii) energy, (iv) decarbonization, and (v) circular economy. These domains are closely interrelated and collectively define the framework for the sustainable development of the Polish steel industry [10].

The decarbonization of the steel industry is a strategic challenge for many steel and steel-product manufacturers. Reducing emissions in this high-emission sector requires extensive investment to implement low-emission technologies. Steelmaking involves complex industrial processes that demand optimization and the capacity to minimize emissions. The ongoing transition toward low-carbon industrial technologies is particularly difficult in Poland due to the historical legacy of industrialization based on inexpensive coal energy. The position of Polish steel producers has deteriorated as a result of high electricity prices and a considerable carbon footprint. Electricity prices in the EU are two to three times higher than in the United States (EUR 0.16 per kWh in the EU versus EUR 0.07 per kWh in the U.S.) during the first half of 2024—despite having decreased slightly in the EU (by EUR 0.04) and remaining nearly stable in the U.S. (+EUR 0.01) compared with the first half of 2023, and despite falling overall energy demand [11].

A significant surge in energy prices was observed during the early months of the energy crisis, which intensified after the COVID-19 pandemic. The energy crisis was influenced by multiple factors beyond supply chain disruptions [12]. One major consequence for the steel sector was the dramatic increase in energy costs—from 17% of total steel production costs to as much as 80% in 2022 [11,13]. Overall energy expenses for large-scale consumers are the highest in Poland among all industrialized European nations. Costs in Poland are approximately 98% higher than in Spain, about 72% higher than in France, 32% higher than in Italy, and roughly 23% higher than in Germany [10].

Energy conservation thus remains one of the greatest challenges for the steel sector, which is among the most energy-intensive industries and the largest energy consumer within the entire manufacturing system. It should also be emphasized that, in the context of the ongoing energy transition increasingly based on renewable energy sources (RES), Poland’s electricity mix—still heavily dependent on coal—translates into high operational costs. Consequently, Polish steelworks face significantly more difficult conditions than their foreign competitors. Furthermore, Poland lacks nuclear power plants that could quickly provide the energy capacity required by the steel industry. In 2024, Poland’s energy mix continued to be dominated by fossil fuels, particularly coal, which accounted for approximately 70% of electricity generation. The positive aspect of this mix, however, was the record-high share of renewable energy, which reached 29.6% of total electricity generation—the highest level to date [14].

The energy challenges are deepened by the necessity for substantial investments in decarbonization to meet climate targets, which require significant capital resources as well as stable and affordable renewable energy sources. The steel sector, both in the EU and in Poland, faces not only rising energy prices but also declining global competitiveness due to a worldwide overcapacity of steel production and strong international competition from countries with lower production costs and weaker environmental regulations. According to the Report of the European Policy Centre [15], global steel overcapacity—primarily in China—may become five times greater than the total steel production of the European Union by 2027. Moreover, tariffs introduced by the President of the United States and the redirection of cheap Chinese steel to the EU market have raised additional concerns about economic security. Since 4 June 2025, the U.S. has imposed a 50% tariff on imports of crude steel and derivative products from most countries (with some exceptions, such as the United Kingdom). The previous tariff level was 25%, meaning that the current 50% rate could seriously disrupt the global steel market balance.

Excess production capacity has resulted in an influx of cheap export goods, especially from China, South Asia, the Middle East, India, and Japan, which have been redirected to the EU market and have caused a decline in domestic production. In 2024, China produced 1005 million tons of steel, India 150 million tons, and the EU 129.6 million tons [16]. EU steel production has declined significantly over the past decade, reaching a low of 126 million tons in 2023—25 million tons below the decade’s average [17]. In Poland, the average steel production volume over the last decade was approximately 8.5 million tons. Only twice in the period 2014–2024 did Poland’s steel output exceed 10 million tons: 10.33 million tons in 2017 and 10.157 million tons in 2018, both before the COVID-19 pandemic [18,19,20]. In 2024, the Polish steel industry produced 7.1 million tons of steel with an installed production capacity of 10 million tons, representing a 73% capacity utilization rate [21].

Domestic steel consumption in Poland was relatively high at 13 million tons, but the supply structure was unfavorable: Imports accounted for 79% of consumption, while domestic deliveries made up only 21% [21]. In recent years, the steel sector has also faced weak demand from key industries such as construction and automotive manufacturing. Apparent steel consumption in the EU fell by 1.1% to 129 million tons in 2024—the third consecutive annual decline and the fifth in the last six years. Except for 2021, apparent consumption has continuously decreased since 2019, clearly reflecting the challenging condition not only of the European steel industry but of the broader EU economy.

Even global steel product consumption declined in 2024 by 0.9% to 1751 million tons (although one could note that global crude steel production increased by 10.7% year-on-year in 2024) [22]. Meanwhile, overall global production fell by 0.8%, with China remaining the dominant producer. Consequently, global production capacity increased from 2439 million tons in 2023 to 2455 million tons in 2024 [16,22]. This means a further rise in global surplus capacity to 572.4 million tons (5.8% more than the previous year) [16,22]. Experts in the steel market agree that there is a persistent imbalance between supply and demand. According to the OECD, global steel demand between 2019 and 2024 amounted to 1890, 1898, 1962, 1895, 1891, and 1870 million tons, respectively, while expected production capacity during the same period was 2416, 2424, 2427, 2453, 2456, and 2472 million tons [23].

The ongoing problem of global overcapacity reduces the profitability of the steel industry and limits the capital available for investments in new technologies, thereby hindering the sector’s decarbonization efforts. In the Polish steel consumer market, conditions have also worsened. In 2024, the Polish construction sector entered a slowdown phase—the number of completed housing units fell by 9.6%, totaling fewer than 200,000. In non-residential construction, activity decreased by 5% compared to the previous year, while the specialized construction segment recorded a decline of 9.8%. The metal industry also experienced a 2% drop, and after three years of growth, the production of machinery and equipment fell by 2.8% in 2024. The automotive sector declined as well: the value of sold production of motor vehicles, trailers, and semi-trailers decreased by 0.7% [21].

Another destabilizing factor for the steel market is geopolitical instability, particularly the ongoing wars in Ukraine and the Middle East. Global supply chains remain disrupted, leading to higher investment costs and necessitating continuous adaptation of business operations. The persistent economic uncertainty, intensified by these armed conflicts, has further discouraged investment decisions in both the private and public sectors. Additionally, the war in Ukraine has reshaped Ukraine’s steel market [24], which remains an important partner for the Polish metallurgical industry.

Finally, the steel sector faces a serious challenge of decarbonization and industrial transformation, as it remains one of the most carbon-intensive industries in Europe. In the global steel sector, the dominant production technology is the BF-BOF route. The essence of BF-BOF (blast furnace–basic oxygen furnace) technology in the steel industry is an integrated, two-stage process for producing steel from iron ore and coke. First, the ore is smelted into liquid pig iron in a blast furnace (BF), and then this pig iron, along with scrap, is converted into steel in a basic oxygen furnace (BOF) by blowing oxygen, which removes undesirable elements and improves steel quality. This is the traditional and dominant method of steel production, underpinning most global production. Currently, 73% of the world’s steel is manufactured through the coal- and coke-based blast furnace–basic oxygen furnace (BF-BOF) process, which emits approximately two tons of CO₂ per ton of steel produced [25]. In the European Union, the share of steel production by technology in 2024 was as follows: BF-BOF and other routes—55.4%, and EAF—44.6% [17]. In Poland, the share of integrated steel production was slightly higher than that of electric steelmaking. In 2024, 3.9 million tons of converter steel (54%) and 3.3 million tons of electric steel (46%) were produced. Compared to 2023, electric steel production decreased by 1%, while converter steel production increased by 23% [21].

In a steelworks with a full BF-BOF production cycle (coking plant → blast furnace → basic oxygen converter), the main emissions result from the reduction of ore with coal and the preparation of the charge. The BF-BOF technology is characterized by high CO₂ emissions. The steel industry accounts for around 8% of total global carbon dioxide emissions, with an average of 1.9 tons of CO₂ emitted per ton of steel produced [23]. Production in blast furnaces (BF-BOF), based primarily on coking coal and iron ore, generates approximately 2.3 tons of CO₂ per ton of steel, while electric arc furnace (EAF) production using scrap emits on average only 0.7 tons [23]. Reducing emissions is therefore a fundamental structural challenge the industry must address. However, the transformation of steel assets toward low-emission production methods takes place in a context where the sector already faces other severe structural difficulties, including overcapacity and related market distortions. According to the OECD, over 40% of the 165 million tons of new steel production capacity expected to enter the market between 2025 and 2027 will rely on relatively emission-intensive blast furnace–basic oxygen furnace (BF-BOF) processes, potentially increasing global CO₂ emissions [23].

Reducing emissions in the steel industry requires deep and costly transformations of production processes. These include the following: (1) improving efficiency through enhanced energy performance; (2) shifting away from coal and gas as fuels; (3) developing and implementing new steel production technologies such as DRI-EAF (natural gas direct-reduced iron combined with electric arc furnaces); and (4) expanding activities in carbon capture, utilization, and storage (CCUS). Considering the long operational lifespan of steel production facilities, investing in new technologies requires a high degree of certainty that such investments will remain economically viable in the long term. A necessary precondition is a healthy market environment characterized by fair competition and the absence of structural overcapacity.

At the global level, there is no turning back from the adopted directions of climate policy, since the goal is to preserve the environment for present and future generations—a principle consistent with the concept of sustainable development. Therefore, energy transformation and decarbonization represent the strategic directions of change for the steel sector in the coming years.

Table 1. CO₂ emissions for global steel production in various combinations.

Routes	CO2 Emissions (t-CO2/t Steel)	Energy Use (GJ/T Crude Steel)
Blast furnace–basic oxygen furnace (BF-BOF)	2.3	24.0
Scrap-based electric arc furnace (Scrap-EAF)	0.7	10.2
Natural gas direct reduction iron (DRI-EAF)	1.4	22.4
Routes	CO2 emissions (t-CO2/t steel)	Emissions intensity relative to the Paris limit by 2030
HDRI-EAF: Hydrogen direct reduction using electric arc furnace (or H-DR/EAF)	0.4	well below the Paris limit by 2030
DRI-EAF with CCS: Direct reduction using electric arc furnace (DR/EAF) with carbon capture and storage	0.7	below the Paris limit by 2030
DRI-EAF without CCS: Direct reduction using electric arc furnace (DR/EAF)	0.9	below the Paris limit by 2030
DRI-BF-BOF with CCS: Direct reduction using BF-BOF with carbon capture and storage	1.3	above limit
DRI-BF-BOF without CCS: Direct reduction using BF-BOF	1.4	above limit
Routes	CO2 emissions (t-CO2/t steel)	Capital expenses EUR/t and TRL status
Top gas recycling blast furnace (TGRBF/BOF)	1.44–1.98	632 TRL 7
Direct reduction using electric arc furnace (DR/EAF)	0.63–1.15	414 Commercial TRL 9
Hydrogen direct reduction using electric arc furnace (H-DR/EAF)	0.025	550–900 TRL 1–4
Electric arc furnace/biomass (EAF/biomass)	0.005	169–184 TRL 6–8

presents energy and emission intensity for global steel production and for specific production technologies. The data presented in Table 1 come from many sources; therefore, there may be discrepancies in the reported CO₂ emission levels, and the data provided are approximate or averaged for periods of analysis used by particular authors of papers [26,27,28,29,30,31,32,33]. As a guide, global levels are assumed to be as follows: BF-BOF ~2.3 tCO₂/t, EAF (scrap) ~0.7 tCO₂/t, and DRI-EAF (gas/hydrogen) usually in the middle. These are comparative values—individual smelters may significantly improve or worsen them [26,27,28,29,30,31,32,33].

3. Materials and Methods

The methodological framework is designed to address the three research questions by combining a dynamic econometric specification with a stepwise model extension. The approach allows for the identification of emission persistence (RQ1), the assessment of direct short-run investment effects (RQ2), and the examination of structural and technological channels through which investments influence long-term emission dynamics (RQ3).

The empirical base of the study consisted of time series data, covering the years 2005–2021, with annual information on three key variables: (i) total CO₂ emissions from the iron and steel sector; (ii) the value of investments in BOF modernization; and (iii) the volume of crude steel production in Poland. Data were taken from national industrial reports, mainly from the Polish Steel Association, the National Centre for Emissions Management, and the Central Statistical Office. All variables were transformed into natural logarithmic forms to stabilize variance and reduce skewness in order to make the model coefficients interpretable as elasticities. The data table was presented in the authors’ own paper [34] https://doi.org/10.3390/su17094045 (in

Table 2. The dataset used for the estimation of the dynamic model.

Year	ln(CO2 Emissions)	ln(BOF Investments)	ln(Crude Steel Production)	ln(CO2 Emissions Lagged)	ln(Coke Consumption)	ln(Energy Consumption)
2006	2.4849	6.8081	9.2106	2.4510	7.8430	7.9924
2007	2.3609	7.6172	9.2702	2.4849	8.0240	7.9997
2008	2.1041	7.6972	9.1817	2.3609	7.8330	7.9470
2009	1.6094	6.8376	8.8718	2.1041	7.3520	7.7620
2010	1.9741	5.4889	8.9872	1.6094	7.5100	7.8260
2011	2.1001	5.8545	9.0790	1.9741	7.5380	7.9010
2012	2.1072	5.7071	9.0302	2.1001	7.5370	7.8400
2013	2.1401	5.6595	8.9802	2.1072	7.5960	7.7400
2014	2.1518	5.6168	9.0514	2.1401	7.8400	7.7270
2015	1.9459	5.7838	9.1282	2.1518	7.7330	7.7190
2016	1.9021	6.0283	9.1043	1.9459	7.6950	7.7960
2017	1.9879	5.5797	9.2415	1.9021	7.6930	7.9440
2018	1.9169	6.0753	9.2271	1.9879	7.7120	7.9170
2019	1.8405	6.1862	9.1047	1.9169	7.6330	7.8160
2020	1.6094	5.1648	8.9708	1.8405	7.4010	7.7020
2021	1.7047	5.9108	9.0418	1.6094	7.4590	7.8460

). The dataset used for the estimation of the dynamic model consisted of the ln of each parameter used in the model.

The inclusion of the lagged variable $\mathrm{l}\mathrm{n}\left({{\mathrm{CO}}_2}_{\mathrm{t}-1}\right)$ enables the model to account for temporal persistence and path dependence, both characteristic of capital-intensive industrial processes with long technological lifecycles. The inclusion of lagged CO₂ emissions directly addresses RQ1 by capturing technological inertia and path dependence characteristic of capital-intensive steelmaking processes. Further, estimates of model parameters are obtained using the Ordinary Least Squares method, while robust standard errors alleviate potential heteroskedasticity. The coefficient of determination, R², was applied to test the goodness of fit of the model. To check for the presence of autocorrelation in residuals, the Durbin–Watson statistic was applied. Residual normality, checked with the Shapiro–Wilk test, confirms that statistical inference based on t and F statistics is appropriate.

Sensitivity analysis was then conducted by permitting hypothetical percentage changes in the key explanatory variables to have their calculated effects on predicted CO₂ emissions. Since the model is specified in log–log form, the estimated coefficients can be considered elasticities; hence, it becomes an easy task to determine how a given percentage change in investment in BOF or in the volume of production would be translated into changes in the emission level. This meant that an evaluation of the relative magnitude of both short-term and long-term effects of investment policies in the steel sector could be carried out.

The methodological approach adopted here is deductive and empirical and combines econometric modelling, comparative analysis, and quantitative inference. In this framework, identification of the determinants of CO₂ emissions and confirmation of a hypothesis on the dominant role of technological inertia and historical path dependence in shaping the decarbonization capacity of the Polish steel industry are possible.

In this specification, investment expenditure is treated as a flow variable, allowing the assessment of whether investments exert a direct short-run effect on emissions once persistence is controlled for (RQ2).

To ensure econometric robustness, several diagnostic and specification procedures were incorporated into the modelling framework. All variables were transformed using natural logarithms to stabilize variance, reduce skewness, and allow for elasticity-based interpretation of coefficients, which is standard practice in dynamic models with multiplicative relationships. The model was estimated using Ordinary Least Squares with heteroskedasticity-robust HC3 standard errors, which provide reliable inference in small samples and under potential heteroskedasticity. Residual diagnostics were conducted to assess normality (Shapiro–Wilk test), serial correlation (Durbin–Watson statistic), and overall model adequacy. Furthermore, multicollinearity was evaluated through inspection of correlation structures and condition numbers to verify the stability of parameter estimates. These additional steps strengthen methodological transparency and demonstrate that the extended dynamic model satisfies the core econometric requirements for valid inference and empirical interpretation.

To test the determinants of CO₂ emissions from the Polish steel industry, the dynamic econometric model was employed that included contemporaneous explanatory variables, together with the lagged dependent variable. Specification of the model is intended to pick up persistence in the run-up in emissions, or structural inertia typical of industrial processes. The dynamic model is mathematically defined as follows:

ln(CO₂_emissions_t) = α + β₁ ln(CO₂_emissions_t−1) + β₂ ln(BOF_investments_t) + β₃ ln(crude_steel_production_t) + ε_t

where

• ln(CO₂_emissions_t) is the natural logarithm of total CO₂ emissions in year t,
• ln(CO₂_emissions_t−1) is the lagged value of the dependent variable (previous year’s emissions),
• ln(BOF_investments_t) represents the natural logarithm of investments in basic oxygen furnace modernization in year t,
• ln(crude_steel_production_t) denotes the natural logarithm of crude steel production volume in year t,
• α is the model intercept,
• ε_ is the error term.

All variables were transformed into natural logarithm form to stabilize variance, reduce skewness, and allow the coefficients to be interpreted as elasticities.

Estimation was performed utilizing Ordinary Least Squares (OLS), with robust standard errors to avoid heteroskedasticity in the residuals. Durbin–Watson diagnostic statistics were utilized to check for the occurrence of autocorrelation in the residuals. Dynamic model estimates were contrasted with reduced-form specifications (log–log model, lagged investment model, and classical linear model) to ascertain excellence in explanation, as well as emission dynamics in the steel industry in the real world. The dataset period is 2005–2021 and shows the year-by-year dynamics of the Polish steel industry. Through logarithmic transformations, data structure allows the estimated coefficients to be interpreted as elasticities, providing a natural relative interpretation of the variables’ relationships. Application of the lagged emissions variable directly accommodates the persistence in observed emissions behavior and, therefore, enables the model to capture lagged, as well as direct, effects of technological and economic variables. Overall, the structured dataset forms a good foundation for dynamic analysis and showcases the importance of temporal dependencies in decarbonization procedures in the sector.

The estimation approach of the extended model originates from a dynamic perspective of CO₂ emissions, in which the level of CO₂ emissions in each year is described as a function of CO₂ emissions in preceding years and selected technological and energy-related determinants in the BF-BOF steelmaking process. Different from the baseline model, in which only the level of output and investment costs are accounted for, selected variables in the extended model are combined to reflect process-related sources of CO₂ emissions, material efficiency, and technical conditions of production infrastructure. The list of selected independent variables includes the natural logarithm of coke consumption measured in terms of main process fuel use, the natural logarithm of electricity use measured in terms of energy intensity, natural logarithm of the value of fixed assets measured in terms of technical level of installations, and the natural logarithm of scrap ratio measured in terms of material efficiency. All variables are transformed into natural logarithms, allowing straightforward interpretation in terms of elasticities and reduction of variance in the time series.

The estimation of the model was performed employing the classical Ordinary Least Squares approach, with standard errors corrected for heteroskedasticity (HC3), allowing valid inferences even in the context of the relatively small number of observations. The lagged variable was included in the equation in the form of CO₂ emissions from the preceding period, thereby allowing the model to take into consideration the inertia in steelmaking technology and the process persistency caused by gradual structural changes. The period under consideration in this analysis spanned from 2006 through 2021, after which the entire data series was converted into the logarithmic scale and arranged in a manner allowing direct estimation of parameters in the dynamic equation.

To address RQ3, the baseline dynamic model is extended to explicitly capture the structural and technological channels through which investments affect emission dynamics.

The additional equation can be represented in the following way:

ln(CO₂_emissions_t) = α + β1·ln(CO₂_emissions_(t−1)) + β2·ln(Coke_consumption_t) + β3·ln(Energy_consumption_t) + β4·ln(Fixed_assets_t) + β5·ln(Scrap_ratio_t) + ε_t

Variable descriptions

• CO₂_t—Annual CO₂ emissions generated by steel production in year t (in million tons). The logarithmic transformation allows the coefficients to be interpreted as elasticities and stabilizes the variance of the series.
• CO₂_{t−1}—One-period lag of CO₂ emissions, capturing the persistence and inertia of emission processes inherent to BF-BOF technology. A positive coefficient reflects slow adjustment dynamics.
• $\mathrm{C}\mathrm{o}\mathrm{k}{\mathrm{e}}_{\mathrm{t}}$—Amount of coke consumed in steel production (in thousand tons). Coke is the primary reductant and energy carrier in the BF-BOF route, and its consumption directly influences process emissions.
• $\mathrm{E}\mathrm{n}\mathrm{e}\mathrm{r}\mathrm{g}{\mathrm{y}}_{\mathrm{t}}$—Electrical energy consumption in steelmaking processes (in GWh). This variable represents the energy intensity of technological operations beyond coke-based reduction.
• $\mathrm{F}\mathrm{i}\mathrm{x}\mathrm{e}\mathrm{d}\mathrm{A}\mathrm{s}\mathrm{s}\mathrm{e}\mathrm{t}{\mathrm{s}}_{\mathrm{t}}$—Value of fixed assets in the steel sector (in million PLN). This variable acts as a proxy for technological advancement, capital modernization, and infrastructure quality.
• $\mathrm{S}\mathrm{c}\mathrm{r}\mathrm{a}\mathrm{p}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{i}{\mathrm{o}}_{\mathrm{t}}$—Share of steel scrap in total furnace charge. A higher scrap ratio reduces the need for primary ore reduction and therefore is associated with significantly lower CO₂ emissions.
• $\mathsf{\alpha }$—Intercept term capturing emission determinants not explicitly included in the model.
• ${\mathsf{\epsilon }}_{\mathrm{t}}$—Random disturbance term representing unobserved factors affecting emissions in year t.

This formulation makes it possible to analyze fuel-related, energy-related, and structural effects simultaneously while preserving the dynamic nature of emission developments over time. The model was subjected to diagnostic testing, including assessments of residual normality, autocorrelation, and the degree of multicollinearity. These results form the basis for interpreting the relationships between technological variables and the dynamics of emissions in the steel sector.

4. Results: Dynamic Model (With Lagged CO₂ Emissions)

The results are presented in a sequence that directly reflects the research questions of the study. First, the dynamic properties of CO₂ emissions are analyzed in order to identify the degree of emission persistence and technological inertia (RQ1). Second, the baseline dynamic model is used to assess whether investment expenditure in BOF modernization exerts a direct short-run effect on emissions once persistence is controlled for (RQ2). Finally, the extended model examines the structural and technological channels through which investments influence long-term emission dynamics, including energy intensity, material composition, and capital stock (RQ3).

The application of a dynamic model in this case is especially significant because it captures the time dependence underlying industrial CO₂ emissions, which are poorly described with stationary specifications. Steel industry emissions are induced by existing production and investment levels and by past operating practices, technological inertia, and long-run structure. By incorporating the lagged value for emissions, the dynamic model accounts for persistence in the overtime dimension of environmental effects and offers a more realistic picture of the kind of pressure technology spending and policy measures impose on emission paths. This approach offers a better understanding of decarbonization processes that take place over time in a gradual fashion and highlights the importance of long-run planning as compared to short-run interventions alone.

Table 3. Model parameters.

Variable	Coefficient	Std. Error	t-Statistic	p-Value
Constant	−5.5001	3.8317	−1.436	0.171
Lagged CO2 emissions (ln)	0.7020	0.2465	2.848	0.016
BOF Investments (ln)	−0.1217	0.1088	−1.118	0.273
Crude Steel Production (ln)	0.7476	0.4375	1.709	0.111

and

Table 4. Dynamic model estimates.

Statistic	Value
R-squared	0.610
Adjusted R-squared	0.505
Durbin–Watson Statistic	1.519
Number of Observations	15

present, respectively, relative results of different model specifications and the complete estimation output of the ultimate dynamic model. Table 3 presents incremental model fit improvement, as provided by the R-squared statistic, with the maximum explanatory power provided by the dynamic specification (61%). It further indicates that statistically significant predictors in the dynamic and lagged models, i.e., crude steel output and historic CO₂ emissions, do exist. The dynamic model estimates are given in Table 4, where last year’s emissions of CO₂ are a statistically significant and powerful determinant of current emissions, with BOF investments and crude steel output not being significant in isolation after accounting for the persistence effect. Collectively, the tables indicate the general predominance of emissions inertia in shaping current conditions and the lack of any short-term effect from modernization investment during the period observed.

The Shapiro–Wilk test was used to test for normality of model residuals. The test’s p-value was above 0.05, and so the null hypothesis of normally distributed residuals is not rejected. This result justifies the use of t-statistic-based inference and F-tests-based inference.

The Durbin–Watson statistic for the model was 1.519, slightly below the ideal value of 2. This provides an indication of no chance of first-order autocorrelation in the residuals.

The dynamic specification, adding the lagged CO₂ emission variable with contemporaneous-year BOF investment and crude steel production, demonstrates an astonishing increase in fit compared to previous specifications. For the dynamic model, an estimate of R-squared of 0.610 indicates that approximately 61%, or slightly more of the variation in the log of CO₂ emissions is explained by the entered variables. The robust growth of the model is evidence of the importance of inertia in emissions over time in the steel sector.

The lagged CO₂ emissions variable is significant and positive (coefficient = 0.7020, p = 0.016), which confirms the presence of high temporal persistence in emission patterns. The BOF investment coefficient is not statistically significant (coefficient = −0.1217, p = 0.273) and negative, which implies that investments in modernization do not have any significant or direct impact on the level of emissions after controlling for dynamic effects. Likewise, the crude steel production factor is positively related to emissions (coefficient = 0.7476), though its p-value (0.111) shows that the relationship is not significant at the 5% level.

The intercept of the model is negative and statistically insignificant, as is expected to be in dynamic specifications where the main interest is in the explanatory variables but not the intercept. There are no severe problems with the residual distribution based on diagnostic tests, and the value of the Durbin–Watson statistic of 1.519 is a good sign of the absence of autocorrelation.

Overall, the results show that CO₂ emissions in Polish steel manufacturing are characterized by high persistence across time, while short-run investment or level-of-production variations have relatively weak independent explanatory power after controlling for previous emissions. The results highlight the need to control for dynamic processes when studying industrial decarbonization trajectories.

The results provide strong empirical evidence of emission persistence in the Polish BF-BOF steel sector. The coefficient of lagged CO₂ emissions is positive and statistically significant, indicating that current emission levels are strongly influenced by their historical values. This finding confirms the presence of technological inertia and path dependence, characteristic of capital-intensive industrial systems with long asset lifetimes and slow adjustment processes. As a result, short-term changes in production conditions or investment activity are insufficient to generate immediate reductions in emissions.

Once emission persistence is explicitly controlled for in the dynamic specification, investment expenditure in BOF modernization does not exhibit a statistically significant direct short-run effect on CO₂ emissions. This result indicates that investment impacts are not immediate and cannot be fully captured through contemporaneous financial flows. Rather, the absence of a short-run direct effect suggests that the environmental impact of investments materializes gradually and operates through structural and technological changes within the production system.

The dynamic model estimated here significantly enhances understanding of the determinants of CO₂ emissions of the Polish steel industry. By including the lagged emission level as an explanatory variable in the model, the model can reflect the empirical reality that there is persistence in the emission level over time. The significant and positive value of the lagged emissions confirms the strong temporal dependence, which signals the reality that past trends greatly affect present environmental impacts. This is an effect that is dynamic in its sense and refers to the structural inertia of the production processes of industry, whereby changes in the behavior of emissions are slow but accumulate over extremely long time spans.

Interestingly, once the persistence effect has been accounted for, the model reveals that Basic Oxygen Furnace modernization investments have no statistically significant independent influence on CO₂ emissions now. Whereas modernization is otherwise held to reduce energy intensity and environmental load, what is found instead is that, during the study period, such investment either has long-term effects or is too small to create short-term and measurable effects. Similarly, whereas the effect of crude steel production level on emissions is positive, it is below statistical significance after controlling for dynamic determinants. The findings here suggest that simple contemporaneous relations have the ability to exaggerate the impact of investment and production activity on emissions unless the timing behavior of emissions is very well-described.

The dynamic model indicates the intricacy of attempts to curb emissions in heavy industry. Policy interventions and firm investment policy must be adapted to the lagged response of industrial systems and the necessity for long-term frames. Short-term changes in production or investments need not necessarily result in immediate changes in emission levels. Deep decarbonization strategies hence need to do more than highlight inducing technological revolutions; they must also defeat system inertial resistance by ensuring policy support, regulatory consistency, and disrupt innovation incentives. Future studies should also investigate dynamic modeling methods, potentially with other structural variables like technological changes in industries, product mix changes, and changes in energy source.

In order to test the sensitivity and interpretability of the dynamic model, sensitivity analysis was performed, simulating the impact of hypothetical changes in the main explanatory variable of interest on forecasted CO₂ emissions. Because the model is log-log specified, one can directly use the estimated coefficients as elasticities to assess percentage changes.

• First, theoretically, a 10% increase in BOF investments (heteroskedastically controlling for other variables) would lower CO₂ emissions by approximately 1.22%. However, because the coefficient on BOF investments (−0.1217) is not statistically significant, such a simulated effect needs to be interpreted with caution and cannot be considered credibly causal under the specified model.
• Second, a 10% increase in crude steel production would lead to an estimated increase of 7.48% in CO₂ emissions based on the production coefficient (0.7476). Although the p-value for this variable (0.111) is at the margin for typical significance levels, the magnitude of the coefficient suggests that fluctuations in levels of production can have a significant effect on patterns of emissions over time.
• Third, the lagged variable of emissions (coefficient = 0.7020) indicates that any increase in emissions in a previous year would still persist in the following year at a rate of approximately 70%. Consistently, tiny increases or decreases in CO₂ emissions in any given year have multiplier effects in subsequent years, highlighting the crucial significance of the early and sustained mitigation measures.

The sensitivity analysis emphasizes that while short-run, direct technology investments will make a modestly perceptible impact on emissions in the short term, changes in production levels and historical trends in emissions have significantly more powerful and longer-term impacts. This finding emphasizes the significance of long-term, systemic solutions to achieve significant and lasting industrial CO₂ emissions reductions.

The sensitivity analysis (

Table 5. The sensitivity analysis.

Scenario	Change in BOF Investments	Change in Crude Steel Production	Predicted Change in CO2 Emissions (%)
1	+5%	0%	−0.61%
2	+10%	0%	−1.22%
3	+20%	0%	−2.43%
4	0%	+5%	+3.74%
5	0%	+10%	+7.48%
6	0%	+20%	+14.95%
7	+10%	+10%	+6.26%
8	+20%	+20%	+12.52%

) reveals that changes in the production of crude steel have a much larger and direct effect on CO₂ emissions than changes in investments in BOFs. A 10% increase in levels of production would be equivalent to an estimated 7.48% increase in emissions, demonstrating the high elasticity of emissions with regard to levels of output. Conversely, a 10% boost in BOF investment is associated with a drop in emissions of just 1.22%, and this is not statistically significant under the dynamic model specification. These results validate that the size of production remains a prime factor influencing emission dynamics in the steel industry, even when controlling for the history of emission levels. Moreover, combined scenarios suggest that the positive impact of expanded production would easily dominate any gain in emissions reduction from expanded investment. For instance, simultaneous 10% rises in investment and production would still yield a net rise in emissions of approximately 6.26%. Such empirical data highlight the structural significance of decarbonizing the steel sector: Independently, overhauls of the technology may not be enough to allow for major emission reductions. Policies for decarbonization with an expectation of success thus need to go hand in hand with technological upgrades and intervention in overall energy and production policies to address short- and long-term dynamic spans of industrial pollution drivers.

The sensitivity analysis results and the dynamic model confirm that policy measures directed solely towards triggering technological modernization, such as investment in basic oxygen furnace modernization, may not be sufficient to guarantee large-scale and timely CO₂ reductions in the steel sector. Although modernization is essential, its impact is small and delayed when considered in isolation. Since emission trends exhibit high inertia, policymakers must determine, beyond a set of stand-alone capital investments, the necessity of long-term and coordinated strategies. Planning for stimulating long-term technology transitions, carbon pricing, the phased elimination of fossil fuel-based production technologies, and deployment of innovative low-carbon technologies (like hydrogen-based steel production) is critical to achieve meaningful emissions reductions over many years.

Sensitivity analysis shows that level of production emerges as the key driver of emissions even after constant holding of dynamic drivers. Climate policy interventions focusing on the steel industry must incorporate measures that decouple production growth from emissions growth. These can include encouraging higher-value and lower-carbon-content steel production, augmenting regulation of emission intensity, and improving circularity such as increased recycling of scrap. By eschewing the consideration of the embedded correlation between the production scales and emissions, decarbonization in industry would be susceptible to increased production scales even in the face of continued technological upgrading. Technology, production patterns, and energy usage must all be addressed simultaneously in order to ensure industrial expansion is in line with climate targets.

To further investigate how investment effects translate into emission dynamics, the baseline dynamic model is extended to explicitly capture the technological and structural channels through which investments operate. This specification moves beyond financial investment flows and focuses on the material and energetic configuration of the steelmaking process.

The presented dynamic model aligns with the strategy of decarbonizing the steel industry in Poland. It demonstrates both the rationale for investing in new steel production technologies and the importance of improving existing ones. A key limitation in constructing this model was the restricted availability of data referring specifically to the steel sector and steel producers. The data used by the author were obtained from reports published by industry organizations. The Central Statistical Office of Poland (in Polish: GUS) aggregates its industrial data under broader categories, combining metal producers and manufacturers of metal products. The steelmaking industry is classified under the general category of metal production. The model presented does not include many potential determinants, and the analysis was concluded in 2021. Meanwhile, by 2024, investment expenditures in the Polish steel industry amounted to approximately PLN 1.5 billion, which was lower than in the previous year [21].

The new dynamic model describing CO₂ emissions has been expanded to include variables reflecting the energy-related, fuel-related, and technological conditions of steel production in the BF-BOF route. In addition to lagged emissions, which capture the inertial nature of industrial processes, the analysis incorporates the logarithms of coke consumption, electricity consumption, the value of fixed assets, and the logarithm of the scrap ratio. This specification reflects the multidimensional nature of emission formation, integrating both process-level and structural determinants. Including these variables makes it possible to capture mechanisms related to fuel intensity, material efficiency, and the technological level of the production infrastructure (

Table 6. Dynamic model—extended specification.

Variable	Coefficient	Std. Error (Robust)	t-Statistic	p-Value
const	−0.7963	9.8265	−0.081	0.935
ln_CO2_lag	0.1554	0.6106	0.255	0.799
ln_Coke_k_t	−0.5295	0.9445	−0.561	0.575
ln_Energy	0.9828	0.8522	1.153	0.249
ln_FixedAssets_mln_PLN	−0.2695	0.5197	−0.519	0.604
ln_Scrap_ratio	−3.3028	2.8014	−1.179	0.238

) Model fit statistics (extended model) Number of observations: 15; R² = 0.681.

The parameters of the new model indicate a differentiated response of CO₂ emissions to technological and energy-related variables. The coefficient for energy consumption (β_Energy = 0.9828) is positive, suggesting that increasing the energy intensity of the process leads to an almost proportional rise in emissions. The scrap ratio exhibits a strongly negative coefficient (β_Scrap = −3.3028), meaning that even a small increase in the share of scrap results in a noticeable reduction in emissions by lowering the need for ore-based reduction. The coefficient for fixed assets (β_FixedAssets = −0.2695) is also negative, which may indicate that technological modernization and capital accumulation contribute to reducing the emissions intensity of production processes. Coke consumption has a coefficient of β_Coke = −0.5295, which in this short time series reflects co-movements with production and energy variables rather than a direct process effect. The coefficient on lagged emissions (β_CO₂, lag = 0.1554) is relatively small, indicating weaker emission inertia compared with the baseline model. Taken together, these coefficients form a pattern in which higher energy use increases emissions, whereas a greater scrap ratio and higher levels of technical capital act to reduce them.

The extended model reveals that CO₂ emissions are primarily shaped by structural and technological variables rather than by investment expenditure alone. Energy consumption exhibits a positive relationship with emissions, reflecting the energy-intensity channel through which technological configurations translate into higher emission levels. In contrast, the scrap ratio shows a strong negative association with emissions, highlighting the role of material efficiency as a key decarbonization mechanism.

The negative coefficient of fixed assets suggests that accumulated capital stock, representing the technological embodiment of past investments, contributes to lower emission intensity over time. The results demonstrate that investment impacts are realized indirectly, through changes in energy use, material composition, and technological structure, rather than through immediate expenditure effects.

The estimated coefficients show that the directions of influence are consistent with the technological characteristics of BF-BOF steelmaking. Energy consumption exhibits a positive relationship with emissions, reflecting the increasing load of energy-intensive operations and their contribution to the overall emissions profile. The negative coefficient on the scrap ratio confirms that higher use of secondary material reduces the need for primary ore reduction and thereby lowers greenhouse gas emissions. Coke consumption and fixed assets also operate in directions consistent with technological intuition, although their individual statistical significance is weakened by multicollinearity and the short time span of the dataset. Despite this, the model allows for the identification of clear qualitative relationships that describe the formation of emissions in the steel sector.

The extended specification displays a higher level of fit compared with models based solely on production volume or investment expenditure, demonstrating that the energy and material structure of production plays a crucial role in emission dynamics. The residuals exhibit an appropriate distribution, and their variability remains moderate, confirming the stability of the estimates despite sample limitations. This model provides a more realistic and multifaceted picture of CO₂ emission formation, enabling the interpretation of emission processes in the context of their fundamental technological determinants. As a result, it offers a solid foundation for analyses concerning energy efficiency, technological modernization, and structural changes in steel production.

The results of the extended model indicate that CO₂ emissions in the steel sector are shaped primarily by energy-related and material-related variables, which better capture the structure of steelmaking processes than production volumes or investment levels alone. The positive coefficient on energy consumption suggests that increasing energy intensity leads to a proportional increase in emissions, consistent with the BF-BOF route where a portion of emissions arises from energy-consuming secondary processes. The negative coefficient on the scrap ratio confirms that higher recycling reduces emissions by lowering the demand for energy-intensive and carbon-intensive ore reduction. The coefficients for coke consumption and fixed assets remain negative, indicating that changes in the fuel structure and capital modernization have the potential to reduce emissions, although their individual statistical significance is limited by multicollinearity and the small sample size.

The extended model achieves a higher R² than the baseline model, indicating improved ability to capture the temporal variability of emissions. The residuals display a proper distribution, and the parameter structure aligns with the technological logic of BF-BOF operations, enhancing the qualitative credibility of the results. Although individual coefficients do not reach full statistical significance, their directions and relationships create a coherent picture of the technological determinants of emissions, with energy intensity and material structure playing central roles. Thus, the model shows that emissions in the steel sector arise from the complex interaction of energy, fuel, and infrastructure factors, and that incorporating these dimensions allows for a substantially more complete explanation of emission dynamics than traditional models based solely on production variables.

Short-run elasticities indicate that CO₂ emissions respond most strongly to adjustments in the material composition of the furnace burden. A 10% increase in the scrap ratio yields, on average, a decline in emissions of roughly 9% in the short run, with the effect strengthening to approximately 11% in the long run. This pattern underscores that enhanced steel recycling constitutes one of the most effective levers for emission mitigation in the analyzed production system. Variations in electricity use also exert a pronounced influence on emissions: a 10% increase in energy intensity raises CO₂ output by around 5% in the short run and nearly 6% in the long run. These figures highlight the sector’s deep dependence on the energetic efficiency of metallurgical processes and the overall load imposed on electrical systems.

The model exhibits noticeably lower sensitivity to coke consumption and fixed-asset value, although the direction of the effects is consistent with established technological mechanisms (

Table 7. The sensitivity analysis extended model.

Variable	Short-Run Elasticity	Long-Run Elasticity	CO2 Change for +10% in X (Short-Run, %)	CO2 Change for +10% in X (Long-Run, %)
Coke consumption	0.2102	0.2494	2.00	2.38
Energy consumption	0.5287	0.6273	5.04	5.98
Fixed assets	–0.1707	–0.2025	–1.63	–1.93
Scrap ratio	–0.9783	–1.1606	–9.32	–11.06

). A 10% rise in coke use generates an increase in emissions of approximately 2–2.4%, reflecting the direct role of coke as a reducing agent and energy carrier. By contrast, a comparable increase in fixed-asset value produces a reduction of around 1.6–1.9%, which can be interpreted as the result of modernization, equipment renewal, and cumulative improvements in process efficiency. Overall, the sensitivity analysis indicates that material structure (scrap ratio) and energy intensity remain the principal drivers of emission variability, whereas fuel-related and capital-related determinants operate as complementary factors. The extended model therefore not only clarifies the directional influence of key variables but also quantifies the magnitude of emissions’ response to technologically realistic parameter shifts.

The sensitivity assessment of the extended dynamic model confirms that technological inputs exert heterogeneous impacts on CO₂ emissions. The scrap ratio emerges as the most influential determinant: a 10% increase leads to a reduction of about 9% in the short term and more than 11% in the long term, marking recycling as the dominant decarbonization pathway within the BF-BOF route. This outcome reflects the fundamental discrepancy in emission intensity between ore reduction and scrap remelting, thereby substantiating the strategic importance of secondary feedstock in emissions-oriented policy design.

Electricity consumption displays similarly significant elasticity, with a 10% increase generating an emission rise of roughly 5% in the short run and close to 6% in the long run. This indicates that the sector’s emission profile is shaped heavily by the energy burden of steelmaking processes, independent of the specific mix of fossil inputs. Coke consumption shows weaker elasticity—10% higher use results in an emission increase of approximately 2%—a pattern attributable to the tight coupling between coke demand and overall production structure. Such changes exert less direct influence on emissions than modifications in scrap share or electrical load.

Fixed-asset value exhibits a modest but consistent negative elasticity: a 10% increase leads to a reduction in emissions of about 1.6% in the short term and nearly 2% in the long term. This suggests that capital accumulation and modernization processes gradually reduce the carbon intensity of installations, even though the magnitude of the effect remains smaller than that associated with material or energy parameters. Taken together, the results demonstrate that energy-related and material-related variables—most prominently the scrap ratio—serve as the dominant levers of emission reduction in the BF-BOF system, while infrastructural investment provides essential long-term support. The extended model thus delineates a clear hierarchy of determinants and identifies the technological domains in which targeted interventions yield the most substantial environmental benefits.

Taken together, the results indicate that CO₂ emission dynamics in the Polish BF-BOF steel sector are dominated by persistence effects and structurally mediated technological factors. While investment expenditure alone does not produce immediate emission reductions, its long-term environmental impact operates through gradual changes in energy intensity, material composition, and capital stock. These findings provide a coherent empirical answer to the research questions and form the basis for the subsequent discussion.

5. Discussion

The findings from the dynamic econometric test for CO₂ emission series in Polish BF and BOF steel industry are useful for understanding the phenomenon of CO₂ emissions in the Polish steel industry and contribute to the discussion regarding the de-carbonization of the energy-intensive industry. The empirical findings confirm and indicate that CO₂ emission over time should be measured through existing structures centered on technological features rather than depending on changes in investment expenditure [35].

Indeed, the statistically significant coefficient of the lagged level of CO₂ emissions corroborates the existence of strong emission path dependence, illustrating the technological inertia embedded in integrated steelmaking technology. The technological structure of BF-BOF plants, being capital-intensive, rigid, and path-dependent with long technological lives, hampers the ability to rapidly adjust emission trajectories. Consequently, the current level of CO₂ emissions continues to be strongly associated with past structure and technology. This result matches the general theoretical discussion surrounding the topics of path dependence and carbon lock-in, postulating that once carbon-efficient technologies lock into industry structures, quick reductions in CO₂ emissions become impossible in the absence of radical technological advancements.

Within this ever-changing setting, a non-statistically significant direct short-run impact of investment expenditure on environmental emissions would not indicate any inefficiency in investment expenditure; rather, its implication is deeply rooted in the temporal nature of investment impacts in heavy industry. Financial investments are in financial terms, but their environmental impacts occur only when they are transformed from a financial proposition to actual operational shifts, technological advancements, or use of materials and energy. The above-mentioned findings indicate that the inclusion of investment expenditure as a contemporaneous regressor for environmental emissions may mis-specify the intricate transmission channels of investments and environmental outcomes [26,32,33,36,37,38,39,40,41,42,43,44,45].

The extended model sheds additional light on these relationships, whereby the technological and structural channels of investment effects on emissions are made explicit. The positive relationship between energy use and CO₂ emissions highlights the fundamental importance of energy intensity in influencing the emissions pro-file of BF-BOF steel production. An increase in energy use, especially in systems with a carbon-intensive electricity mix, translates almost proportionally into an increase in emissions even if production volumes are kept constant, highlighting the fact that the steel industry’s decarbonization cannot be realized by process optimizations alone but requires overall changes in the energy system [46].

The very strong negative correlation between the scrap ratio and CO₂ emissions highlights the significance of material efficiency in achieving a reduction in CO₂ emissions. An increase in the scrap proportion in a furnace charge decreases the necessity for the primary reduction of ore, which in turn affects the process-related CO₂ emissions negatively. The magnitude of this finding, as established in the sensitivity tests, indicates that even without revolutionary technological developments, there are promising CO₂ emission reduction possibilities through adjustments in material composition. On a structural level, this finding confirms the importance of a circular economy approach in the decarbonization of conventional steelmaking processes.

The negative sign of the coefficient related to fixed assets further shows that accumulated capital stock, as measured by the technological representation of past investment, is related to a decrease in emissions intensity. Compared to the investment expenditure, which represents a financial outlay at a certain point, fixed assets are related to the cumulative effects of modernization, new equipment, and improving efficiency. The size of this coefficient is quite small, as a certain technological upgrading is a time-consuming process, especially in the integrated steel industry, as shown by the coexistence of new and old facilities. However, it confirms that investment has a certain impact on the environment, although it is indirect and lagged [47,48,49,50,51,52].

In combination, the results indicate that the divergence between declared emission reduction targets within industrial roadmaps and the small time-bound effects of such investments revealed within this study is due to time and scope considerations. Most announced low-carbon projects regarding hydrogen-based direct reduction and other sector-wide electrification plans follow time horizons extending significantly beyond those revealed within our dataset and pertain to ‘transformational change’ rather than ‘incremental progress’ [53,54,55,56,57,58,59,60,61]. In contrast, most observed investments scrutinized over this time horizon pertain to a modernization process following a traditional BF-BOF system, hence experiencing a limited scope regarding emission reduction.

In terms of policies, these findings have significant implications. They imply that policies which solely focus on increasing the amount of investments made could possibly fail to provide optimal outcomes in emission reductions in the short to medium term. Instead, the focus of policies should lie in the direction, rather than merely the amount, of investments made, thereby actively promoting investments in directions that can help reduce energy intensity, modify material composition, or optimize process patterns. Such support policies, which provide incentives for increasing scrap use, increasing energy efficiency, or incorporating lower carbon energy sources, are bound to provide better outcomes in the short term than unconditional investment subsidies [62,63,64,65,66,67,68,69,70,71].

The dominant role of emission persistence suggests that ahead-of-curve actions are needed. This is because abatement actions accumulate slowly over time, and as such, procrastination in structural policy actions can result in entrenching higher emissions paths for protracted periods of time. This adds further emphasis to long-term policy predictability necessary for supporting long-payback investments as well as investments in new technology that have uncertain payoffs. In such conditions, firms can sufficiently emphasize short-term competitiveness rather than structural changes [72,73,74,75,76,77].

The results demonstrate the benefits of dynamic modelling techniques to analyze the topic of industrial decarbonization. Static models failing to account for the features of time dependence may result in a policy or investment intervention being overestimated in the short term with respect to its actual capacity to effectively address the problem [78,79,80,81,82,83,84,85,86]. This study’s dynamic model may be improved in the future to include an additional dynamic factor reflecting the evolution of the energy mix or the spread of game-changing technologies for the reduction of greenhouse gas emissions.

The paper points out that decarbonization in the BF-BOF steel industry is not mainly hampered by a lack of investment, but by the production structure and the speed at which innovations materialize. An investment is still a precondition for a reduction in emissions, but it is effective depending upon the way in which it influences the energy use and the technological structure. These processes and mechanisms are highly important for an accurate analysis and for the development of a strategy for the decarbonization of the industry.

The models presented in the publication were developed using real data that reflect the situation of the Polish steel sector. The Polish steel industry, and its main production plants, operate either blast furnace–basic oxygen furnace (BF-BOF) technology or electric arc furnace (EAF) technology. Investments made in the sector in recent years have been directed toward the modernization of steel production technologies; however, in the case of BF-BOF technology, these have not been new (radical) investments that would have transformed the existing coke-based BF-BOF technology into direct reduced iron (DRI) technology, where hydrogen would serve as the reducing agent.

The models presented therefore confirm the rationale for introducing radical investments—specifically the already mentioned DRI technology, which is classified as a clean (low-emission) technology—in order to reduce CO₂ emissions. The historical data used in the models thus confirmed the need for substantial recapitalization of integrated steel plants operating with BF-BOF technology. The drivers of this transformation are not only environmental but often also economic. The production of low-carbon steel in Poland is more expensive within the EU than, for example, in China. The models presented align with the industrial modernization policy for the steel sector that has been implemented in Poland for several years.

Poland is a country with coal resources, and the heart of the traditional steelmaking process (BF-BOF) is coke. Its reaction with iron ore generates enormous amounts of CO₂. Instead of coal, hydrogen should be used as the main reducing agent, with water vapor as the by-product. This fundamental change in the approach to metallurgy should be implemented both in Poland and in other EU countries, in line with climate policy, in order to move from the coal era to the era of chemistry.

Although DRI technology is already used worldwide, its economic viability on a mass scale—especially for countries such as Poland, where annual steel production does not exceed 10 million tons, of which about half is produced using BF-BOF technology, and where the largest steel plants are owned by foreign capital that globally owns plants with much larger production capacities than those in Poland—becomes a strategic decision. We do not know the profitability calculations of foreign capital owners of BF-BOF plants in Poland regarding investment in new coke-free technologies, because on a global scale it may turn out that current plant owners choose a transformation pathway different from the one assumed for Poland.

The models merely demonstrated the necessity of change, because the current technology is no longer competitive in light of stringent requirements for clean steel production. At the same time, we also confirmed that increasing the use of scrap in steel production is possible, which reduces hot metal production and thus CO₂ emissions; however, such an approach is a short-term strategy that will not revolutionize coal-based metallurgical technology.

6. Conclusions

The paper adds a dynamic and mechanism-based interpretation of CO₂ emission generation in the Polish BF-BOF steel industry to the literature, answering three related research issues regarding emission persistence, the direct effect of investment expenditure, and the structural pathways of influence of investments on emission paths. By merging a dynamic econometric model with a step-by-step model development approach, this paper advances beyond static analyses of industry emission and proposes a more realistic approach to studying decarbonization processes in capital-intensive industries.

To answer research question RQ1, the data reveals a high level of persistence in CO₂ emissions in the Polish BF-BOF industry. As shown through a statistically significant positive coefficient on the lagged emissions variable, this level of persistence overall substantiates that CO₂ emissions along this sector can be explained by a degree of technology inertia, where historical production processes continue to have a strong impact on, and define, overall emissions at a given point in time. This evidence overall lends a degree of credence to theoretical knowledge that heavy industry has a slow rate of emission processes that move along a certain path, with emission levels taking a long time to adjust to dynamic changes, such as a reduction of emissions through certain forms of policy change.

With regard to RQ2, it can be seen from the findings that investment expenditure in BOF modernization in the model of CO₂ emission did not display a statistically significant direct effect on CO₂ emission in the first stage after accounting for CO₂ emission persistence. Instead of viewing the outcome from the lens of investment ineffectiveness, it should be noted from these findings that there is a pivotal methodological and conceptual implication regarding financial investment flows and their instantaneous translation into outcomes in capital-intensive industry and production structures. These findings therefore add a new dimension to the literature with regard to making it clear that the significance of investment and emission relationships lies in their time-dependent nature and cannot be effectively measured through instantaneous relationships.

The analysis pursuing an answer to RQ3 above further refines the understanding of the structural and technological mediators of the relationship between investments and the long-term behavior of emissions. The expanded model confirms the dominant impact on emissions of variables capturing the material and energy structure of the product-ion system. Energy use appears to have a positive relationship with emissions, thus emphasizing the energy intensity mediator of the impact of technological and system-level energy ties on environmental performance. By contrast, the scrap ratio has a very strong negative relationship with emissions, thus verifying the paramount importance of material efficiency and secondary resource use as effective decarbonization technologies for the BF-BOF process. The negative coefficient on fixed assets indicates the impact of accumulated capital stocks, understood to represent the technological realization of past investment, on a decrease in the rate of emissions with time.

Together, the results shown above highlight the structural, in-direct impacts of investment, rather than those which arise from immediate expenditure. From a scientific point of view, it is clearly one of the major contributions of this paper to conceptually decouple investment flows from their technological representation, which revealed that emission cuts occur via changes in material composition, energy use, and capital structure. Moreover, this work resolves the apparent contradiction concerning the absence of short-term investment impacts on the one hand, and the typical long-term decarbonization stories often encountered in industry road maps on the other hand.

The scientific interest of the paper is not only represented by the empirical findings but rather by methodological advancement. The dynamic approach used in this paper enables the analysis of emission persistence, direct investment effects, and structural mechanisms in a united model. The proposed approach advances static studies by allowing for delayed responses and technological inertia, providing a more accurate basis for scientific interpretation and policy inference. The results imply that the focus of the analysis has to shift from the volumes to the outcomes of investment in order to achieve an effective decarbonization of energy-intensive sectors.

The provided research contributes to literature on the decarbonization of the industrial sector in the following way: it has been demonstrated that the emission processes in the BF-BOF steel sector are driven by slow dynamic structural processes, rather than investment choices. By empirically confirming the relevance of emission persistence and the role of structurally mediated investment effects, the research has created a strong analytical framework for future studies on dynamic emission models.

Polish steel plants must strive for transformation toward hydrogen-based technologies in order to meet environmental challenges, EU regulations, and global competitiveness. This implies moving away from coal toward hydrogen (green hydrogen or biomethane) in the steel production process. Although this requires significant investments and the development of hydrogen infrastructure—which is still lacking in Poland—the models have shown that investments to date merely improve the existing BF-BOF technology, whereas it already needs to be replaced with a new one.

Among EU countries, the development of low-emission steel production technology is most advanced in Swedish steel plants. Similar facilities are being built across Europe and are at various stages of implementation, including in Germany, France, Belgium, the Netherlands, and Spain. In fact, throughout Western Europe such steel plants are expected to be operational in the second half of this decade. New plants will be built because the technology itself is entirely new, but they are most often located where steel plants already exist, in order to maintain employment. These locations already have infrastructure for steel product manufacturing and are prepared for industrial investment.

This raises the question: are Polish steel plants operating with BF-BOF technology already being sidelined? A key condition for zero-emission steel production is access to green hydrogen, and access to hydrogen depends on access to large amounts of cheap green energy (while in Poland around 70% of energy is produced from coal). Therefore, such a plant may not be built in Poland in the foreseeable future. If the owners of steel plants in Poland (operating with BF-BOF technology) do not invest in radical emission reductions, these plants will likely be closed. In Kraków, the blast furnace that produced steel from iron has already been shut down, and today the only blast furnace—representing primary steel production technology—operates in Dąbrowa Górnicza and accounts for approximately half of Poland’s steel production.

The steel plant in Dąbrowa Górnicza is no longer technologically young and must be continuously modernized, which we confirmed in our models. Secondly, from 2026 onward, the number of free allowances for CO₂ emissions in the European Emissions Trading System (ETS) will be reduced for, among others, the steel, cement, and chemical sectors. All these plants will have to pay for an increasing share of their emissions each year, which reduces the profitability of steel production using BF-BOF technology. From 2034, free allowances for high-emission sectors will disappear entirely, meaning that from that point onward they will have to pay for every ton of CO₂ emitted.

7. Limitations

Some limitations apply to the study, and they should be considered while evaluating the findings of the study. To start with, the scope of the empirical study is defined by the available data regarding investments in the Polish BF-BOF steel production process. During the period of available data, from 2005 to 2021, there were no transformational investments in the blast furnace process in the country, and investments were incremental in nature, including maintenance, efficiency enhancement, and compliance with environmental regulations. Consequently, the data used in the study do not allow the direct determination of the impact of transformational investments in the blast furnace process on CO₂ emissions through cause/effect relationships in the short term.

This is borne out in the econometric model, wherein the lack of statistical significance for the investment factors, while controlling for emission persistence, is not an indicator of model misspecification but an accurate depiction of the empirical setting in which the data exist, wherein the forces of path dependence and inertia drive the emissions process, and the influence of marginal increments is not immediate but rather is manifest over a horizon that is not captured in the time series.

The inclusion of the scrap ratio in the extended model mainly verifies a known technological correlation—the fact that increased scrap usage reduces primary iron production needs and thereby CO₂ emissions. Though this correlation has been long known in the literature on steelmaking technology, its inclusion still has analytical merit in this analysis because it validates the internal consistency of the dynamic system and provides a basis for model verification against known process-level mechanisms. Yet these findings regarding scrap usage must be seen in a confirmatory rather than a generative context. Limitations arise from a relatively short series of observations per year because this input series has a low frequency. This inherently decreases the statistical potency of estimation in extended models that involve a variety of technological and energy-related regressors. Multicollinearity among process-technology regressors and the current level of aggregation in national statistics further reduces the accuracy in point estimates in this analysis because there are limitations in discerning the short-term from long-term changes in structural conditions due to these data limitations.

The analysis is based on historical dynamics and thus does not respond to the effects of recent or future transformational schemes like hydrogen-based DRI production or CCUS deployment, which lie beyond the time scope of the database. Analysis for future horizons or after the passage of investment cycles in 2022 will help evaluate future transformational changes for disruptive technological transitions in the Polish steel industry and their impact on emitted GHG concentrations.