The Impacts of Government Support Schemes on Technological Innovation in High CO₂ Emitting Industries: The Case of Korea

Abstract

Korea’s industrial structure faces a critical challenge: the proportion of high CO₂ emitting industries in the GDP decreased during the 2000–2021 period, whereas their contribution to national CO₂ emissions increased. This study evaluates the impact of the government’s comprehensive support system—encompassing seven areas: tax incentives, general funding, financial support, human resource development, technical support, certification, and public purchase—on technological innovation in high CO₂ emitting industries. Using the Probit model on data from the Korea Innovation Survey 2022 (2019–2021) and a financial statement database of firms, we analyzed firms in high CO₂ emitting industries. The findings reveal that among the various forms of government support, human resources development ranks highest in its positive impact on innovation, followed by tax incentives and general funding. The results suggest that effective climate policy must shift to prioritize targeted technical and knowledge-based assistance to overcome the specific innovation prevalent in high CO₂ emitting industries.

1. Introduction

The Republic of Korea (hereafter Korea) has committed to the 2015 Paris Agreement and declared the 2050 carbon neutrality goal in 2020. This commitment was enacted in 2021 with the Basic Carbon Neutrality and Green Growth Act and aimed to reduce national greenhouse gas emissions by 40% by 2030 compared to 2018 levels.

Despite these policy commitments and emissions slowdown [1], Korea still faces a severe structural vulnerability. Korea’s global CO₂ emissions ranking worsened from 16th in 1990 to 8th in 2021. The structural dependence on carbon-intensive sectors remains critical: while the GDP share of high CO₂ emitting industries (HCO2I) declined from 9.70% in 2000 to 6.98% in 2021, their share of total national CO₂ emissions increased from 30.07% to 31.08% over the same period. This opposite direction emphasizes the immediate challenge of transitioning toward a low-carbon industrial structure.

This climate imperative is compounded by Korea’s demographic crisis (e.g., total fertility rate of 0.75 in 2024) and a shrinking working-age population. Given that traditional drivers of economic expansion—capital accumulation and labor force growth—are diminishing, technological innovation becomes the indispensable driver for sustaining economic development and productivity improvements, which is a principle well-established in innovation-led growth theories [2,3].

Korea thus stands at a crossroads, required to simultaneously achieve sustained economic growth while aggressively controlling CO₂ emissions. Given the dominance of HCO2I and the central role of innovation, enhancing technological innovation within these sectors has become the strategic policy imperative.

Government intervention is justified by the concept of market failure. Innovation inherently involves positive externalities (knowledge spillover) and high uncertainty [4], leading to chronic under-investment. For green innovation, this failure is compounded by environmental externalities and high switching costs. Consequently, governments use targeted support schemes to mitigate financial risk and stimulate the necessary technological transition.

While prior research tends to broadly deal with market-based climate policies (e.g., carbon tax) or aggregate effects of government R&D support, the research question for this study is to determine whether government support schemes have differentiated and measurable effects on innovation in high-carbon sectors. Based on the role of government intervention in mitigating market failures, we posit the following hypothesis: government support schemes contribute to technological innovation within HCO2I.

To complement this gap, this study offers four contributions. First, we provide evidence by focusing exclusively on technological innovation within Korea’s HCO2I. Second, we uniquely investigate the distinct impacts of seven specific government support schemes (including tax incentives, general funding, financial support, human resource development, technical support, certification, and public purchase) on innovation. Third, we employ a comprehensive measure of total innovation expenditure (R&D and non-R&D costs). Finally, we enhance empirical rigor by integrating the Korea Innovation Survey (KIS 2022) with financial statements [5].

The remainder of this paper is structured as follows: Section 2 explains the theoretical background and methodology. Section 3 presents the empirical analysis. Section 4 presents the discussion including policy implications; and Section 5 offers the conclusions.

2. Materials and Methods

2.1. Theoretical Background

Technological innovation is important in enabling industries, particularly those with high carbon dioxide (CO₂) emissions, to transition towards sustainable growth. The Schumpeterian framework emphasizes innovation as the engine of economic growth, fostering productivity improvements and competitive advantages [6]. However, innovation in high CO₂ emitting industries faces challenges due to the capital-intensive nature of these sectors, long investment cycles, and significant technological uncertainties [7].

Market failures, including information asymmetries, externalities, and financing constraints, often inhibit private firms from investing sufficiently in green technologies and process upgrades [8]. Consequently, government support emerges as a critical mechanism for correcting these market imperfections, facilitating innovation that aligns economic and environmental objectives [9]. Schemes such as tax incentives, R&D subsidies, technical assistance, and public procurement policies can lower the cost and risk of innovation activities in energy-intensive sectors [10].

Empirical evidence demonstrates the heterogeneous effectiveness of different government supports in stimulating innovation. While R&D subsidies increase overall inventive output [11], targeted programs that address firm-specific barriers—such as knowledge spillovers and skill development—often yield higher impact in heavy industries [12]. Recent studies also highlight the importance of complementary policies integrating financial support with human capital development to accelerate the adoption of green innovation [13].

Nesta & Saviotti suggests that innovation outcomes depend on interaction effects between policy design, industry characteristics, and firm capabilities [14]. To address the complexity of Korea’s comprehensive government support for HCO2I, this study adopts a disaggregated approach, linking each of the seven support categories. Tax incentives reduce tax by R&D expenditures, which operate via a fiscal mechanism. General funding indicates governments’ direct subsidies or grants. Financial support indicates government-sponsored loans, guarantees, or indirect subsidies. Human resource development fosters human capital to enhance absorptive capacity. Technical support addresses technology development, commercialization, or transfer. Certification schemes indicate corporate certification, technical product certification, or awards, which use a mechanism to convey product quality. Finally, public purchase means that governments prioritize purchase systems with small and medium-size enterprises’ products, which acts as a demand-pull mechanism.

2.2. Data

2.2.1. High CO₂ Emitting Industries

The first step in analyzing the determinants of technological innovation in high-carbon industries is to define which sectors constitute high CO₂ emitting industries (HCO2I) in this study. Although the identification of HCO2I varies among studies and there is some debate due to the lack of a common scientific standard, the concept is used interchangeably with energy intensive industries. In 2021, energy consumption accounted for 93.6% of Korea’s total CO₂ emissions [15].

Liddle categorized energy-intensive industries into five manufacturing sectors: pulp and paper, petrochemicals, basic metals, non-ferrous metals, and non-metallic mineral products [16]. Similarly, Song & Oh classified energy-intensive industries into four different kinds if their energy consumption exceeds the average across all industries [17]. Their definition includes pulp and paper, coke and refined petroleum products, non-metallic mineral products, and primary metals.

In the Korean context, researchers defined energy-intensive industries as petrochemicals, non-metallic minerals, primary metals, and metal products [18]. Kim & Lee calculated values of direct and indirect greenhouse gas emissions per unit of value added by industry for all industries and then the study categorized them into “high”, “medium”, and “low” emission groups [19]. The high group is defined as containing high CO₂ emitters such as electricity supply, chemicals and chemical products, non-metallic mineral products, primary metals, petroleum refining products, textiles, metal processing products, other transport equipment manufacturing, and metal mining.

In this study, high-carbon industries are defined based on the research findings of [17,18,19] as well as the Korean Standard Industrial Classification (KSIC) from Statistics Korea, which serves as the foundation for the Korea Innovation Survey (KIS 2018). From Kim & Lee, the metal mining, electricity supply, and transportation industries were excluded because they are not related to manufacturing innovation [19]. This is in line with other existing studies [17,18]. Textile and other transport equipment manufacturing (including shipbuilding) were also excluded due to their negligible CO₂ shares and inconsistencies in classification by [20] and Korean Standard Industrial Classification (KSIC). Textile and shipbuilding do not belong to energy-intensive industries as in [16,17,18]. The CO₂ share of Textiles and Leather is 0.09%. Thus, the HCO2I are defined in this study as the aggregate of five sectors, which are refined petroleum products, chemical products, non-metallic mineral products, primary metals, and fabricated metal products.

2.2.2. Korea Innovation Survey 2022: The Manufacturing Sector

This study utilizes data from the Korean Innovation Survey 2022: Manufacturing Sector (KIS 2022), conducted by [20], to analyze the determinants of technological innovation in high-carbon industries. KIS 2022 is aligned with [21] and is the most recent publication of Korea’s collection of innovation data. KIS 2022 comprehensively collects detailed innovation-related information, including innovation inputs (total innovation expenditure and R&D expenditure), innovative activities, regulation, and government support utilization for firms (reflecting the 2019–2021 activity period).

The survey population comprises manufacturing firms with 10 or more employees active in the years 2019 to 2021. Stratified random sampling uses the Korean Standard Industrial Classification revision 10 and size categories to ensure representative coverage. Firms with over 500 employees were sampled exhaustively, while smaller firms were selected using the Neyman allocation. Our sample is specifically restricted to firms classified within HCO2I based on the Korea Standard Industry Classification 10th codes to analyze the differential impact of aggregate and seven specific government support schemes on innovation. The KIS 2022 population comprises 4000 firms, with 1008 identified as HCO2I. After excluding 575 firms lacking matching financial statements, a final sample of 433 firms was utilized for estimation.

Unlike OECD, which divided innovation into four categories (product innovation, process innovation, organizational innovation, and marketing innovation) [22], KIS 2022 classifies innovation primarily into two types: product innovation and business process innovation. The business process innovation consists of seven categories: (i) production and development of goods or services, (ii) logistics, shipping, or distribution methods, (iii) information processing or communication methods, (iv) accounting or other administrative task procedures, (v) business practices for establishing procedures or external relations, (vi) methods of job assignment, decision-making, or human resource management, and (vii) marketing methods including promotion, packaging, pricing, product placement, and after-sales service.

In this study, technological innovation is defined as the occurrence of either product innovation or the production and development of goods or services (the first category of business process innovation), or both. Consistently with [22], which groups product and process innovation under technological innovation, firms are considered to have exhibited innovation if they report at least one of these two types.

It is worth noting that this study uses total innovation expenditure instead of the conventional R&D expenditure as a measure of firms’ innovation input. While the KIS 2022 dataset is based on the framework of [21], the conceptual foundation for distinguishing innovation from research and development spending originated in [22]. This earlier edition established the crucial necessity to account for all non-research and development spending, such as the acquisition of capital machinery and external knowledge. By adopting this approach—which is important in analyzing high-carbon industries where process changes demand substantial investment in deployment and adoption rather than only R&D—we ensure that our analysis captures the full, practical scope of firms’ decarbonization efforts.

Key variables used in this study include firm size measured by sales, squared firm size to assess nonlinear effects, firm age, market concentration ratio of the top three firms (CR3), total innovation expenditure intensity (total innovation costs relative to sales), export intensity, operating profit margin, stock market listing status, and use of government support schemes. Total government support is the sum of seven types of government support schemes, as detailed in

Table 1. Detailed types of government support schemes.

No.	Type of Support Scheme	Description
1	Tax incentives	11 tax incentives, such as a research and development expenses tax deduction
2	General funding	15 programs where firms benefit from investment, direct subsidies, or grants provided by the government
3	Financial support	55 programs where firms receive government-sponsored support such as loans, guarantees, or indirect subsidies through financial institutions (e.g., banks or financial organizations)
4	Human resources development	13 programs for human resources development
5	Technical support	16 programs for technology development, technology commercialization, technology transfer, patent strategy, etc.
6	Certification support	15 programs for corporate certification, technical product certification, awards, etc.
7	Public purchase	5 programs for public purchase systems with small and medium-sized enterprises

Source: [23].

.

Seven types of support schemes are used by individual firms arbitrarily, from none to all 7 types or any combination of the 7 types.

Table 2. Number of firms using government support schemes in HCO2I.

	Number of Government Support Schemes
	0	1	2	3	4	5	6	7	Sum
Number of Firms	379	227	169	109	33	30	22	39	1008
Share (%)	37.6	22.5	16.8	10.8	3.3	3.0	2.2	3.9	100.0

shows the number of government support schemes used by firms in HCO2I: 60.1% use zero (37.6%) or 1 case (22.5%) of a government support scheme, while 12.4% use four to seven cases of government support. In theory, seven combinations (5040 cases) are possible. However, if zero cases (379 firms) are excluded from 1008 firms, which is out of the scope, 629 firms remain. Using one type (227 firms) of government support scheme reaches up to 36.1% of the 629 firms. This is a representative case. To avoid multicollinearity among different types of support schemes, this study is adopted to estimate only one type of government support scheme used by the firms.

2.3. Methodology

This study analyzes a technological innovation-dependent variable. The variable is a binary indicator, taking the value of 1 if innovation occurs and 0 otherwise. To explore the factors influencing technological innovation, we employ a Probit regression model, which has been widely used in prior studies [24,25,26].

The Probit model relates the binary dependent variable to a set of explanatory variables through the following formulation:

P(y_i = 1) = F(x_iß)

where P(y_i = 1) denotes the probability of innovation by firm i, F represents the cumulative distribution function (CDF) of the standard normal distribution, y_i is the binary outcome variable indicating innovation, x_i symbolizes explanatory variables, and ß denotes the parameters.

In this context, the probability that y_i = 1 reflects the occurrence of innovation is modeled as a function of x_iß, with the Probit model specifically using the normal CDF for F(x_iß). The parameters ß capture the influence of each explanatory variable on the likelihood of innovation.

Formally, the probability can be expressed as

$$P\left(y_i=1\right)=\mathsf{\Phi }\left(x_i\beta \right)={\int }_{-\mathrm{\infty }}^{x_i\beta }\varphi \left(t\right)dt$$

where ϕ(x_iβ) and Φ(t) indicate the probability density function (PDF) and CDF of the standard normal distribution, respectively. The estimation of β proceeds through maximum likelihood methods based on this equation.

Unlike linear regression models, where coefficients directly represent marginal effects, Probit model coefficients ß indicate only the direction (positive or negative) of the impact on innovation probability, since the relationship is nonlinear. Therefore, marginal effects—quantifying how average changes in explanatory variables affect the probability of innovation—must be explicitly calculated.

The marginal effect of an explanatory variable is given by the derivative of the expected value of y with respect to x_j:

$${\displaystyle \frac{\partial E\left[y_i∣x_i\right]}{\partial x_{ij}}}=\varphi \left(x_i\beta \right)\cdot {\beta }_j$$

Since ϕ(x_iβ) depends on individual x_i, marginal effects vary across observations. To summarize these varying effects, researchers often calculate the average marginal effect—also called the partial effect at the average (PEA)—defined as

$$PE{A}_j={\displaystyle \frac{1}{N}}\sum _{i=1}^N\varphi \left(x_i\beta \right)\cdot {\beta }_j$$

This average provides a single interpretable estimate for the effect of each variable on the probability of innovation across the sample.

Following the primary estimation, multicollinearity was diagnosed by calculating the Variance Inflation Factor (VIF) and robustness checks were conducted using a Logit model.

3. Results

Table 3. Descriptive statistics.

Variable	Mean	Std. Dev.	Maximum	Minimum
INNO	0.48	0.50	1.00	0.00
SALE	1211	4469	69,923	6
SALEQ	21,388,607,113	244,495,922,996	4,889,165,140,926	37,252
AGE	30.75	15.61	76.00	5.00
CR3	30.43	15.59	63.49	8.62
TIEINT	1.54	2.98	29.11	0.00
EXPINT	17.87	24.05	100.00	0.00
OPM	0.05	0.12	1.53	−0.74
IPO	0.25	0.43	1.00	0.00
GOVT	0.73	0.45	1.00	0.00
GOV1	0.48	0.50	1.00	0.00
GOV2	0.25	0.43	1.00	0.00
GOV3	0.21	0.41	1.00	0.00
GOV4	0.37	0.48	1.00	0.00
GOV5	0.30	0.46	1.00	0.00
GOV6	0.38	0.49	1.00	0.00
GOV7	0.14	0.35	1.00	0.00

Note: INNO = technological innovation, SALE = sales, SALEQ = Squared sales, AGE = firm’s age, CR3 = the market shares of the three largest firms, TIEINT = total innovation expenditure intensity = total innovation expenditure/sales, GOVT = total government support system, EXPINT = export intensity = export/sales, OPM = operating profit margin = operating income/sales, IPO = initial public offering, GOV1 = tax incentives, GOV2 = general funding, GOV3 = financial support, GOV4 = human resource development, GOV5 = technical support, GOV6 = certification support, GOV7 = public purchase.

details descriptive statistics for the variables. Of the firms in high-carbon industries, 48% are engaged in technological innovation. Sales revenue exhibited wide variation, with an average of KRW 1.211 trillion. The mean market concentration ratio (CR3) stood at 30.43%. Firms spent an average of 1.54% of sales on innovation, and approximately 73% benefited from at least one form of a government support scheme.

Table 4. Estimation results.

	M1	M2	M3	M4	M5	M6	M7	M8
SALE	0.0000 *	0.0000 *	0.0000	0.0000	0.0000	0.0000	0.0000 **	0.0000
	(0.0721)	(0.0993)	(0.1242)	(0.1520)	(0.3826)	(0.2455)	(0.0409)	(0.1484)
SALEQ	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
	(0.5743)	(0.5781)	(0.7510)	(0.7706)	(0.9857)	(0.9004)	(0.4153)	(0.7202)
AGE	−0.0084 **	−0.0049	−0.0017	−0.0029	−0.0059	−0.0031	−0.0010	−0.0020
	(0.0251)	(0.1704)	(0.6246)	(0.4079)	(0.1012)	(0.3803)	(0.7635)	(0.5723)
CR3	−0.0311 ***	−0.0294 ***	−0.0242 ***	−0.0212 ***	−0.0251 ***	−0.0221 ***	−0.0216 ***	−0.0215 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)
TIEINT	0.1252 ***	0.1523 ***	0.1526 ***	0.1541 ***	0.1414 ***	0.1398 ***	0.1610 ***	0.1575 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)
EXPINT	0.0079 ***	0.0079 ***	0.0083 ***	0.0083 ***	0.0086 ***	0.0092 ***	0.0090 ***	0.0090 ***
	(0.0061)	(0.0063)	(0.0038)	(0.0033)	(0.0038)	(0.0012)	(0.0014)	(0.0014)
OPM	1.2270 **	1.2950 **	1.2169 **	1.1878 **	0.9039	1.1459 **	1.2093 **	1.2042 **
	(0.0243)	(0.0162)	(0.0216)	(0.0250)	(0.1034)	(0.0334)	(0.0220)	(0.0226)
IPO	0.3767 **	0.3962 **	0.4307 **	0.4585 ***	0.4053 **	0.4330 **	0.4689 ***	0.4683 ***
	(0.0351)	(0.0268)	(0.0139)	(0.0085)	(0.0286)	(0.0136)	(0.0067)	(0.0069)
GOVT	0.8681 ***
	(0.0000)
GOV1		0.8157 ***
		(0.0000)
GOV2			0.6513 ***
			(0.0000)
GOV3				0.4862 ***
				(0.0043)
GOV4					1.0173 ***
					(0.0000)
GOV5						0.5113 ***
						(0.0009)
GOV6							0.0107
							(0.9376)
GOV7								0.3556 *
								(0.0760)

Note: *, **, *** denote a significance level at 10%, 5%, and 1%, respectively. M indicates Model.

presents eight Probit models, including estimates for total (Model 1) and specific government support schemes (Model 2–Model 8), which are separated to control for multicollinearity. In the Korean context, sales has a minimally positive effect on technological innovation in Models 1, 2, 7, offering negligible support for the Schumpeter hypothesis that larger firms innovate more. The squared sales is insignificant and rejects the inverted U-shaped hypothesis. Firm age (AGE) is negatively related to innovation only in Model 1 (coefficient −0.0084, p < 0.05), which indicates that older firms are less likely to innovate in this initial specification. While the coefficient loses statistical significance in the subsequent models (M2–M8), this negative relationship is generally reported to be more robust among startups and SMEs in the relevant literature.

Market concentration (CR3) consistently inhibits technological innovation, meaning that stronger competition contributes to innovation rather than a monopoly on power, which challenges the Schumpeterian notion. Total innovation expenditure intensity (TIEINT) and export intensity (EXPINT) both significantly enhance technological innovation, with which Korea has achieved rapid economic growth [27,28]. Exporting firms demonstrate increased productivity and innovation, which help them compete in international markets. Harris & Moffat also supports a positive relationship between exports and innovation in the UK’s manufacturing sector [29]. Operating profit margin and IPO status are positive drivers, revealing that profitability and capital market access spur innovation.

Government support—both total and in the form of human resources development (GOV4), tax incentives (GOV1), general funding (GOV2), technical support (GOV5), and financial support (GOV3)—is a robust and positive force for technological innovation. Certification support is not significant, and public purchase support is only marginally significant. These results confirm that government intervention functions as a critical factor for overcoming market failures and encouraging R&D [30].

Additionally, studies of corporate green innovation in China demonstrate that government environmental subsidies significantly promote enterprise innovation, emphasizing direct funding and guidance as vital tools for overcoming financing constraints and market failures [31,32,33].

Table 5. Marginal effects.

	M1	M2	M3	M4	M5	M6	M7	M8
SALE	0.0000 *	0.0000 *	0.0000	0.0000	0.0000	0.0000	0.0000 **	0.0000
	(0.0721)	(0.0993)	(0.1242)	(0.1520)	(0.3826)	(0.2455)	(0.0409)	(0.1484)
SALEQ	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
	(0.5743)	(0.5781)	(0.7510)	(0.7706)	(0.9857)	(0.9004)	(0.4153)	(0.7202)
AGE	−0.0032 **	−0.0021	−0.0008	−0.0013	−0.0025	−0.0014	−0.0005	−0.0009
	(0.0251)	(0.1704)	(0.6246)	(0.4079)	(0.1012)	(0.3803)	(0.7635)	(0.5723)
CR3	−0.0054 ***	−0.0055 ***	−0.0056 ***	−0.0055 ***	−0.0056 ***	−0.0056 ***	−0.0056 ***	−0.0056 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)
TIEINT	0.0627 ***	0.0763 ***	0.0765 ***	0.0772 ***	0.0708 ***	0.0701 ***	0.0807 ***	0.0789 ***
	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)	(0.0000)
EXPINT	0.0040 ***	0.0040 ***	0.0042 ***	0.0042 ***	0.0043 ***	0.0046 ***	0.0045 ***	0.0045 ***
	(0.0061)	(0.0063)	(0.0038)	(0.0033)	(0.0038)	(0.0012)	(0.0014)	(0.0014)
OPM	0.6440 **	0.6814 **	0.6385 **	0.6225 **	0.4685	0.5996 **	0.6343 **	0.6315 **
	(0.0243)	(0.0162)	(0.0216)	(0.0250)	(0.1034)	(0.0334)	(0.0220)	(0.0226)
IPO	0.2076 **	0.2193 **	0.2404 **	0.2577 ***	0.2249 **	0.2419 **	0.2642 ***	0.2638 ***
	(0.0351)	(0.0268)	(0.0139)	(0.0085)	(0.0286)	(0.0136)	(0.0067)	(0.0069)
GOVT	0.6672 ***
	(0.0000)
GOV1		0.5674 ***
		(0.0000)
GOV2			0.3740 ***
			(0.0000)
GOV3				0.2681 ***
				(0.0043)
GOV4					0.6988 ***
					(0.0000)
GOV5						0.2948 ***
						(0.0009)
GOV6							0.0054
							(0.9376)
GOV7								0.1875 *
								(0.0760)

Note: *, **, *** denote a significance level at 10%, 5%, and 1%, respectively. M indicates Model.

presents marginal effects. In estimating regression models such as Probit, coefficients do not represent the direct effect of a predictor on the probability of an outcome but instead reflect changes in the latent variable’s index. This makes it inappropriate to compare their relative magnitudes across variables. Thus, marginal effects are calculated to interpret the direct influence of each predictor on the probability of the outcome, which is the probability of technological innovation in this study.

The estimates in Table 5 quantify the change in probability for each variable, holding all other factors constant. The results from Model 1, which employs the aggregate government support variable (GOVT), demonstrate that total government support (0.6672, p < 0.01) yields the greatest marginal effect across all independent variables. The next most influential variables are operating profit margin (OPM, 0.6440, p < 0.05) and initial public offering status (IPO, 0.2076, p < 0.05). Total innovation expenditure intensity (TIEINT, 0.0627, p < 0.01) and export intensity (EXPINT, 0.0040, p < 0.01) are also statistically significant positive factors. Meanwhile, sales (SALE) have almost no influence with a marginal effect of 0.0000, which is significant at the 10% level (p < 0.10). Finally, market concentration (CR3) is consistently and negatively significant across all models (ranging from −0.0054 to −0.0056, p < 0.01), which implies that higher concentration can hinder innovation.

Among the seven types of specific government support schemes, human resource development (GOV4, 0.6988), tax incentives (GOV1, 0.5674), and general funding (GOV2, 0.3470) exhibit the greatest positive marginal effects, followed by technical support (GOV5, 0.2948) and financial support (GOV3, 0.2681). This pattern highlights the relative efficacy of schemes that address workforce skills, fiscal incentives, and direct funding in stimulating innovation within high CO₂ emitting industries. Public purchase support (GOV7) and certification (GOV6) are not statistically significant at the 5% significance level. These results reinforce the policy recommendation to prioritize these types of schemes in determining government schemes for HCO2I.

Multicollinearity, which is the phenomenon where independent variables are highly correlated with one another, can inflate standard errors and lead to unreliable statistical inference.

Table 6. Variance inflation factor (VIF) for multicollinearity diagnostics.

Model	Abbreviation	Type of Support Scheme	Pseudo R2 (McFadden)	VIF
M1	GOVT	Total government support system	0.62	2.64
M2	GOV1	Tax incentives	0.62	2.64
M3	GOV2	General funding	0.61	2.57
M4	GOV3	Financial support	0.60	2.52
M5	GOV4	Human resource development	0.64	2.76
M6	GOV5	Technical support	0.60	2.53
M7	GOV6	Certification	0.60	2.47
M8	GOV7	Public purchase	0.60	2.49

presented the variance inflation factor (VIF) for multicollinearity diagnostics for all models. The maximum VIF is 2.76. This value is substantially below the commonly accepted threshold of 5 used in econometric analysis, so the results confirm that multicollinearity is not a significant concern in our model.

To assess the robustness of our conclusions regarding the statistical significance of government support schemes, we re-estimated the model with the Logit model. The Logit model uses a different cumulative distribution function (the logistic function) for its error term compared to the Probit model (which uses the standard normal cumulative distribution function). The results of the Logit model confirmed the findings of the primary Probit analysis. This strong consistency across specifications confirms the robustness of the primary conclusion.

4. Discussion

Regarding the relationship of economic development and CO₂ emissions, this study indicates Korea’s structural change from 2000 to 2021: while the GDP share of HCO2I fell substantially (−2.72%p), their proportional CO₂ emissions increased (1.01%p). These opposite directions emphasize both the positional importance of HCO2I and the urgent need for a sustained transition towards a lower-carbon industrial structure.

In line with this observation, this study aims to investigate the effects of both the total and seven specific government support schemes on a firm’s innovation performance in HCO2I. To simultaneously achieve sustained economic growth and CO₂ emissions reduction, the Korean government has provided government support schemes for boosting technological innovation to firms.

This study combined the KIS 2022 (2019–2021) and financial statement databases to examine the effects of financial data on technological innovation. The results of this study demonstrate the critical role of total governmental support schemes in stimulating technological innovation within Korea’s HCO2I. Several studies have addressed that government support schemes consistently enhance innovation capacity in both effect direction and magnitude [30,31].

It is worth noting that this study took a closer look at sorting government support schemes into seven specific categories. This approach is different from existing studies that usually deal with government support as a whole. Sorting support into disaggregated areas to examine diagnosis and improvement is useful to systematically assess different government support schemes. The results demonstrate that human resource development (GOV4), tax incentives (GOV1), general funding (GOV2), technical support (GOV5), and financial support (GOV3) are particularly influential in encouraging firm-level innovation activities. This result confirms that targeted government support is essential to stimulate innovation in HCO2I with a more tailored support strategy.

Second, most notably, human resource development (HRD) exerts the greatest influence among the seven government supports tested. This reinforces that human capital is the primary engine of both firm productivity and technological advancement according to Romer [34]. Especially for HCO2I, human capital influences a firm’s absorptive capacity—the ability to identify, assimilate, and apply external knowledge. In the Korean context, HCO2I are characterized as capital-intensive sectors that require substantial R&D for technological innovation, which necessitates the mobilization of highly skilled research personnel [35]. This aspect is particularly necessary for developing industrial decarbonization. Furthermore, given that innovation in HCO2I often necessitates significant process changes, success is heavily dependent on highly experienced human capital’s tacit knowledge (know-how and insights) rather than coding skills. HRD programs are uniquely effective at transferring and accumulating this experiential, tacit knowledge within the firm, making them a superior and more targeted policy tool than general financial subsidies. Government support schemes targeting HRD—such as training grants, labor subsidies, or specialized skill development programs—are frequently deployed to overcome market failures related to skill underinvestment.

Third, it is an important finding that tax incentives and general funding have significantly positive impacts (as the next strongest values). At the same time, it is significant that technical support, which is often assumed to have a strong effect on technological innovation, is a relatively lower value in this study. Targeted government support is useful to boost government support schemes in a disaggregated way. Compared to earlier aggregate analyses, this disaggregated approach provides more insights into which government schemes yield the greatest innovation outcomes in HCO2I, offering policymakers evidence that can help them prioritize support schemes that will directly address firm-specific needs. Additionally, these empirical results reflect that the Korean government is faced with a challenge to fulfill the dual objectives of Korea’s green growth agenda: sustaining economic growth and achieving carbon reduction.

Fourth, there is a statistically insignificant relationship between firm age and innovation in seven models. This aspect implies that after controlling total innovation expenditure intensity, firm age is not a determinant factor of technological innovation in this industrial context, likely due to internal offsetting effects of the dominant factors like total innovation expenditure intensity and CR3.

The findings of this study carry policy implications for government support to boost technological innovation in HCO2I. First, governments must re-evaluate and enhance the overall effectiveness and long-term sustainability of support schemes. Second, based on the empirical evidence showing the HRD scheme as the most effective driver of performance, policy priorities must center on cultivating a high-caliber R&D workforce and institutionalizing specialized training programs, thereby equipping researchers and professionals to spearhead the technological innovation and decarbonization efforts for Korea’s HCO2I sectors.

Third, centered on a structural budget shift, the innovation support budget for high-carbon firms should be systematically prioritized toward HRD and Technical Support initiatives. Implementation for HRD should focus on establishing Public–Private Skill Renewal Programs for deep decarbonization technologies and certification programs co-developed with public research institutes. Furthermore, financial support must be tightly coupled with measurable decarbonization milestones via “Green Performance Contracts” with private sectors to ensure accountability and verified results.

Fourth, to mitigate high market risks, the government should strategically utilize public purchase instruments, acting as a guaranteed customer for low-carbon materials in state-led infrastructure projects, thereby providing scale-up capital for firms making the initial high-cost transition. Ultimately, targeted government support measures, including tax incentives, general funding, technical support, and direct financial subsidies, should be implemented in an integrated manner to effectively address firm-specific innovation barriers and ensure the efficient use of public resources.

5. Conclusions

This study employed a Probit model on the Korea Innovation Survey (KIS 2022) data to evaluate the differentiated impact of seven government support schemes on innovation within Korea’s High CO₂ Emitting Industries (HCO2I) from 2019 to 2021. The empirical analysis reveals that not all policy instruments are equally effective, underscoring the necessity for targeted intervention over generic R&D subsidies. HRD demonstrated the highest positive statistical significance and suggested that specialized human capital and knowledge transfer is necessary for industrial transition rather than capital availability. Broad financial incentives, including tax incentives and general funding, also confirmed their significant role in mitigating risk and cost. Consequently, effective climate policy should prioritize HRD and targeted technical assistance to maximize public spending efficiency.

This study is subject to limitations. First, using binary measures of technological innovation is a limitation of the current study. These measures do not capture the intensity, quality, or economic significance of innovation, although they are common in [22,24,25]. Future studies may use continuous dependent variables such as patent counts or citations, innovation revenue, and R&D output efficiency. Second, the data collection period (2019–2021) coincides with the global COVID-19 pandemic. This major external shock might likely have a systematic impact on business operations and innovation activities and influence the accurate assessment of policy effects.

Third, a potential issue of reverse causality may exist between government support and firm innovation: more innovative firms might also be more proactive in securing government support. This potential endogeneity could affect the reliability of the findings.

Fourth, the impact of innovation on actual CO₂ emissions levels is omitted due to data availability. Successful technological innovation in HCO2I directly contributes to the reduction in CO₂ intensity (CO₂ per unit of value added). This reduction in intensity can occur through three ways: reducing absolute CO2 emissions, increasing value added while maintaining CO₂ emissions, or a combination of both. Future research, therefore, should develop a comprehensive methodology and measurement framework to effectively link policy-induced firm-level innovation to quantifiable changes in absolute CO₂ emission levels and environmental performance.

Future research can be expanded by employing panel data to investigate the persistence and lag effects of government support schemes on innovation outputs. When extended data becomes available, it will be feasible to estimate the impacts of the COVID-19 pandemic. A separate analysis of product innovation versus process innovation can be conducted to differentiate the impact of government support since process improvements are key for decarbonization in HCO2I.