How does CEO green experience affect green innovation of energy firms? Evidence from China

Literature review

Carbon-based energy adversely impacts environmental sustainability. ¹⁹ Therefore, research on energy transition and green technological innovation is burgeoning. This study draws on two main strands of literature. The first strand concerns GI concepts and its driving factors. The second discusses the impact of CEO characteristics.

First, GI refers to technological innovation, such as renewable energy development and waste recycling, aimed at reducing pollution and saving energy. ²⁰ GI improves energy efficiency and develops efficient and low-carbon alternatives. ²¹ Many studies have investigated objective factors influencing GI, such as government subsidies, energy policies, and environmental management.^22–24 For example, Adebayo and Ullah ²⁵ argue that to improve environmental quality, governments should encourage renewable energy development and technological innovation through green subsidies and improved financial stability. Yu et al. ²⁶ investigate the effect of financial constraints on GI and find that green financial policy alleviates financial constraints and promotes corporate GI. Yan et al. ²⁷ argue that city development policy promotes GI, upgrading industrial structure and green efficiency. Yang et al. ²³ suggest that renewable energy policies promote GI; this effect is heterogeneous in countries with differing GI capacities. However, these objective perspectives ignore humanistic factors. Considering the upper echelon theory, ²⁸ some studies identify CEO characteristics that influence corporate GI strategies, such as foreign experience, ²⁹ staff turnover, ³⁰ gender, ³¹ and hometown identity. ³² Moreover, some literature explores the effect of GreCEO on corporate environmental performance ¹⁰ ; however, its direct effect on energy firms’ GI, its influencing mechanisms, and boundary conditions remain unclear.

Second, based on the upper echelon theory, scholars state that CEOs strongly impact corporate strategy. Relevant studies examine firm performance, ³³ cash management behavior, ³⁴ and corporate risk-taking. ³⁵ For example, Lee et al. ³⁶ establish a negative correlation between CEO overconfidence and bank recognition of loan-loss provisions. Chen and Huang ³⁵ find a positive correlation between CEO reputation and corporate risk-taking, particularly when the reputation is positive and external governance is restricted. Chen ³⁷ argues that CEO polychronicity promotes innovation in firms with dynamic environments, larger sizes, and inferior performance, but reduces it in firms with less dynamic environments, smaller sizes, and superior performance. However, existing research has not paid enough attention to the GI effect of CEO characteristics. Therefore, there is a need to explain a different story, i.e., GI effect of GreCEO in Chinese energy firms, which has been ignored in previous literature and motivates our research.

In summary, existing studies have overlooked the mediating channels between GreCEO and GI, as well as the moderating impact of external policy on them. Therefore, this study employs a moderated mediation model to examine the mediating effect of GM and DAR, and the moderating role of REP. It will provide a reference for green development and energy transition from the perspective of CEO initiative and policy coordination.

Research hypotheses

GI is an important measure in the transition from fossil to renewable energy sources. ³⁸ However, high risk and long-term return of R&D investments, as well as path dependency on traditional innovation, present significant challenges to GI. ³⁹ Unlike traditional innovation, GI creates negative externalities in the R&D phase and positive externalities in the adoption and diffusion phases. ¹³ This increases the importance of CEOs’ awareness of green transition for energy firms. ²⁹

According to upper echelon theory, ²⁸ CEO early experiences influence their judgment and strategic choices. ³⁰ CEOs’ environmental awareness is shaped through their environmental education and work experience. CEOs with green experience are concerned about environmental sustainability. They tend to combine commercial goals with social responsibility. Their actions steer energy firms’ innovation direction away from reliance on traditional technology. ⁴⁰ In addition, CEOs with green experience are more likely to be informed about environment policies and to accurately predict policy directions. ⁴¹ These tendencies will reduce policy uncertainty risks, mitigate information asymmetry and create a first-mover advantage. Furthermore, green experience improves CEOs’ risk identification on GI projects, enhances investor confidence, and protects stakeholder interests. As a result, green R&D investment in energy firms will increase. Those energy firms that employ CEOs with green experience tend to engage in green R&D activities, thereby increasing GI level. Therefore, we present the first research hypothesis as follows.

H1: GreCEO helps promote energy firms’ GI.

One possible mechanism through which GreCEO influences energy firms’ GI is corporate GM. GM refers to the process of allocating and utilizing the enterprises’ internal resources, ⁴² such as using eco-friendly materials, establishing an ecological governance system, introducing environmental education, and increasing green R&D investment, in order to achieve sustainability goals. ¹¹ Corporate internal resources are central to generating and sustaining competitiveness. Enterprises primarily rely on internal resources to develop strategies and enhance competitiveness. CEOs with green experience will allocate more resources to GM to support green transformation.^43,44 Therefore, GM facilitates green resources integration and improves GI. Thus, CEOs with green experience drive energy firms’ GI through GM. Therefore, we propose the following hypotheses:

H2: GreCEO positively impacts energy firms’ GM.

DAR is another mechanism through which GreCEO influences GI adoption. GI projects require long-term investments with high initial input, lagging returns, and high risks. According to the risk theory, R&D investments are associated with untested, unproven technological innovation. ⁴⁵ Since a considerable portion of R&D investments are spent on human capital, energy firms face challenges in reducing their R&D expenditure without significant layoffs in skilled personnel. In addition, owing to asymmetric financial information and asset liquidity constraints, financing green R&D through debt is challenging. Since the issuance of the Green Credit Guidelines by the Chinese government, banks are restricted from lending to firms with higher environmental risks. High polluting energy firms face financing constraints. These issues create difficulties for energy firms to obtain debt financing that requires servicing at maturity. ⁴⁶ Therefore, equity is better than debt financing for high-risk green R&D investments. ⁴⁷ CEOs with green expertise will reduce DAR to mitigate green R&D debt risk. GreCEOs may lead energy firms to reduce debt and increase equity financing. Therefore, we propose the following hypotheses:

H4: GreCEO negatively impacts the energy firms’ DAR.

REP is a government-proposed policy to increase investment in low-carbon technologies. CEOs’ innovation strategy choice depends on the trade-off between the “compliance cost” and the “innovation compensation” effects. REP guides CEOs with green experience to increase green R&D investment, thereby contributing to firms’ GI outputs. ⁴⁸

The policy environment will inevitably affect GM and subsequent innovation decisions. ⁴⁹ Energy firms are encouraged to monitor GM by REP benefits such as green subsidies and tax incentives. Therefore, under the guidance of REP, energy firms tend to allocate resources towards green R&D projects, thereby increasing the GI effect of GM.

Due to the imperfect financial systems in developing countries, Chinese energy firms are exposed to increased innovation risk, such as information asymmetry, moral hazard, and adverse selection, thereby making it difficult to raise debt. In addition, GI's positive externalities exist in renewable energy sector, which may cause the underinvestment of green R&D. However, REP sends positive signals to markets regarding renewable energy, thereby encouraging financial institutions to support GI projects. ⁷ To mitigate financial constraints, CEOs with green experience will make rational lending decisions and raise bank loans, leading to an increase in energy firms’ DAR. Hence, we propose the following hypotheses:

H6: REP positively moderates the relationship between GreCEO and energy firms’ GI.

Figure 1 shows the research models based on above hypotheses.

OPEN IN VIEWER

Figure 1.

The influencing mechanism of CEO green experience on energy firms’ GI.

Variables selection

Sample and data

According to the 2012 Industry Classification Guidelines for Listed Firms issued by China Securities Regulatory Commission, this study classifies all A-share listed firms in the coal, petroleum, and natural gas extraction and processing industries and electricity, heat, and gas industries as energy firms. We obtained the unbalanced panel data for these firms in China from 2004 to 2021, with a sample of 10,079 energy firm observations, after excluding ST and *ST firms and winsorizing all continuous variables at the 1% quantile. The original firm-level and CEO profile data are obtained from the CSMAR database and the CNRDS database, respectively. The regional-level data are obtained from the China Regional Economic Statistical Yearbook. Table 1 presents the sample statistics.

OPEN IN VIEWER

Table 1.

Descriptive statistics of the variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
GI_IP	10079	0.330	0.773	0.000	7.228
GI_UMP	10079	0.293	0.679	0.000	5.814
GreCEO	10079	0.026	0.160	0.000	1.000
GM	8761	0.220	0.253	0.000	1.000
RAT	10079	0.454	0.202	0.007	3.625
Policy	10079	0.620	0.485	0.000	1.000
Size	10079	22.227	1.347	19.916	26.272
Age	10079	2.770	0.399	1.609	3.466
ATO	10079	0.726	0.474	0.094	2.731
Growth	10079	0.190	0.375	−0.490	2.255
Fixed	10079	0.298	0.177	0.020	0.776
Indep	10079	0.366	0.051	0.000	0.714
TobinQ	10079	1.783	0.975	0.863	6.382
Inst	10079	0.352	0.248	0.000	0.894

Model specification

To test the effect of GreCEO on GI, we specify the following basic model:

G I i j k t = β 0 + β 1 G r e C E O i j k t + β 2 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(1)

What transmission channel exists between GreCEO and energy firms’ GI? To illuminate the influencing mechanism, we used Wen et al.'s ⁵⁷ Bootstrap method to measure the mediating effect and construct models (2)–(3) as follows.

M e d i u m i j k t = ξ 0 + ξ 1 G r e C E O i j k t + ξ 2 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(2)

G I i j k t = π 0 + π 1 G r e C E O i j k t + π 2 M e d i u m i j k t + π 3 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(3)

Furthermore, we introduce the moderating role of REP and construct the mixed model as follows.

G I i j k t = ρ 0 + ρ 1 G r e C E O i j k t + ρ 2 R E P i j k t + ρ 3 G r e C E O i j k t × P o l i c y i j k t + ρ 4 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(4)

M e d i u m i j k t = λ 0 + λ 1 G r e C E O i j k t + λ 2 R E P i j k t + λ 3 G r e C E O i j k t × P o l i c y i j k t + λ 4 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(5)

G I i j k t = σ 0 + τ 1 G r e C E O i j k t + σ 2 R E P i j k t + σ 3 G r e C E O i j k t × R E P i j k t + σ 4 M e d i u m i j k t + σ 5 M e d i u m i j k t × R E P i j k t + σ 6 C o n t r o l s i j k t + γ j + δ k + φ t + ε i j k t

(6)

Main results

Table 2 reports the baseline estimate results for model (1). Columns (1) and (3) demonstrate the regression results without control variables using GI_IP and GI_UMP as explanatory variables, respectively. The results show that GreCEO significantly promotes GI. We again estimate to control for the variable series; Columns (2) and (4) present the results, which are similar to those without control variables. Compared with CEOs with no green experience, those with it improve GI_IP and GI_UMP by 61.3% and 37.9%, respectively. These findings confirm H1.

OPEN IN VIEWER

Table 2.

Benchmark results.

Variables	GI_IP	GI_IP	GI_UMP	GI_UMP
	(1)	(2)	(3)	(4)
GreCEO	0.704***	0.613***	0.448***	0.379***
	(0.059)	(0.055)	(0.053)	(0.052)
Size		0.199***		0.146***
		(0.011)		(0.010)
Age		−0.172***		−0.228***
		(0.026)		(0.024)
ATO		0.093***		0.056***
		(0.019)		(0.016)
Growth		−0.005***		−0.004***
		(0.001)		(0.000)
Fixed		−0.038		0.153***
		(0.050)		(0.045)
Indep		0.247*		0.365***
		(0.144)		(0.131)
TobinQ		0.047***		0.040***
		(0.009)		(0.008)
Inst		0.029		0.084***
		(0.036)		(0.033)
Constant	0.311***	−3.884***	0.281***	−2.646***
	(0.007)	(0.264)	(0.006)	(0.235)
Year FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes
Adj. R2	0.249	0.315	0.234	0.290
Observations	10075	10075	10075	10075

Robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, * p < 0.1.

The coefficients of control variables Age and Growth in Table 2 are significantly negative, inconsistent with our expectations. This finding may corroborate the results of Jovanovic ⁵⁸ and Pellegrino. ⁵⁹ Upon entering an industry, firms must choose to emulate the operations and techniques of established firms to survive. Thus, entrants have higher GI motivation and fewer barriers. This trajectory depresses older firms’ innovative vigor. When growth is too rapid, firms spend considerable resources on marketing while scaling back R&D investments.

Endogenous analysis: instrument variable (IV) approach

Through social networks, CEOs in the same city in a given year learn from and compete with each other, fostering strong peer effects. Intuitively, it is difficult for one person to influence the surrounding environment in a society, which provides an opportunity to construct a proper instrument. ⁶⁰ We measured the average value of GreCEO of neighboring listed firms within a city in a given year as an instrumental variable (IV), denoted IV.GreCEO.

Columns (1)–(3) in Table 3 report the IV method estimation results. Column (1) displays the first stage estimation of the IV method, indicating that GreCEO significantly positively correlates with IV.GreCEO. The Cragg-Donald Wald F and Kleibergen-Paap Wald F statistics are 1149.52 and 176.41, respectively. Both robust F-values are well above the 10% significance level of the Stock-Yogo weak ID test (i.e., 16.38), negating the possibility of a weak IV. Columns (2) and (3) report the IV estimations for the effects of GreCEO on GI_IP and GI_UMP, respectively. The estimated coefficients are significantly positive, further enhancing the robustness of the main results.

OPEN IN VIEWER

Table 3.

Robustness results: lag one year and IV method.

Variables	Lag One Year		IV Method
	GI_IP	GI_UMP	First-Stage	GI_IP	GI_UMP
	(1)	(2)	(3)	(4)	(5)
L. GreCEO	0.693***	0.400***
	(0.080)	(0.073)
IV.GreCEO			0.010***
			(0.001)
GreCEO				0.487***	0.248**
				(0.104)	(0.102)
Controls	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes
Adj. R2	0.325	0.299	0.174	0.098	0.069
Observations	8668	8668	8792	8792	8792

Robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, * p < 0.1.

Robustness tests

Propensity score matching (PSM)

Unobservable factors may influence whether firms choose CEOs with green experience. Therefore, self-selection behavior leads to biased results. The propensity score matching (PSM) approach can mitigate self-selection bias by matching a sample with similar characteristics that influence GreCEO. Figure 2 illustrates the narrowed covariate standard deviations after kernel and nearest neighbor matching with caliper implying that the post-matching individual characteristics are less volatile. Table 4 reports the PSM method estimation results. Columns (1) and (2) display kernel matching. Columns (3) and (4) present nearest neighbor matching with caliper. The coefficients of GreCEO after two matches are all significantly positive, consistent with the main results.

OPEN IN VIEWER

Figure 2.

Standardized bias across covariates before and after matching.

OPEN IN VIEWER

Table 4.

Robustness results: PSM.

Variables	Kernel Matching		Nearest Neighbour Matching
	GI_IP	GI_UMP	GI_IP	GI_UMP
	(1)	(2)	(3)	(4)
GreCEO	0.502***	0.331***	0.552***	0.368***
	(0.049)	(0.046)	(0.062)	(0.054)
Controls	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes
Adj. R2	0.618	0.528	0.600	0.511
Observations	5500	5500	1862	1862

Robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, * p < 0.1.

Marginal treatment effect (MTE) framework

The IV and PSM methods primarily capture average treatment effects, i.e., local average treatment effect (LATE). However, multiple unobservable endowments exist among enterprises, commonly associated with individual treatment effects, resulting in heterogeneous self-selection. Individual treatment effects vary with a firm's willingness to choose CEOs with green experience. More importantly, a firm's unobservable endowments usually depend on whether or not the firm is treated (essential heterogeneity), ⁶¹ leading to the heterogeneity of hidden bias. To address hidden heterogeneous selection motivations, the MTE approach was used to illustrate the treatment effect distribution across firms alongside hidden heterogeneity by measuring the propensity score of firms to select CEOs with green experience.

MTEs are founded on the generalized Roy model as follows.

Y j = μ j ( X ) + U j , for j = 0 , 1

(7)

Y = D Y 1 + ( 1 − D ) Y 0

(8)

D = 1 { μ D ( Z ) > V } , where Z = ( X , I V )

(9)

Y₁ and Y₀ are the potential outcomes in the treated and untreated states, such as the GI of enterprises with and without CEOs with green experience. They are denoted as a function of the observable variables, which include year, city, and industry fixed effects. Model (9) is the discrete choice model of Carneiro et al., ⁶² where Z and V are matrices of observable instrumental and unobservable variables, respectively.

When μ_D (Z)>V, firms choose CEOs with green experience. Otherwise, firms choose CEOs without it. Hence, unobservable factors determine the extent of resistance to treatment. That is, firms’ higher unobservable gains will foster a higher level of treatment resistance. Firms with low resistance will be more prone to choose CEOs with green experience than those with high resistance. With conditional independence ((U₀, U₁, V) ┴ IV | X) and separability (E U_j | V, X .=E U_j | V.), the MTE is defined as follows.

MTE ( x , u ) ≡ E [ Y 1 − Y 0 | X = x , U D = u ] = x ( β 1 − β 0 ) heterogeneity in observables + E [ U 1 − U 0 | U D = u ] κ ( u ) : heterogeneity in unobservables ,

(10)

implying that given the observable variable X, the impact of selecting CEOs with green experiences varies with the unobservable resistance to treatment. Following Brinch et al., ⁶³ we employ a separate approach to estimate MTEs.

Panels A and B of Figure 3 show the estimated MTE curves and their confidence intervals for GI_IP and GI_UMP, respectively. The downward sloping curve indicates that the higher the likelihood of a firm choosing CEOs with green experience (i.e., lower unobservable resistance to treatment), the more favorable it is for GI. Figure 4 illustrates the MTE compilers for those with LATE weights for GI_IP and GI_UMP, respectively. Figure 4 shows that the MTE curve adjusted by the LATE weights follows the same trend as the original MTE curve. It exhibits a smooth decrease with increasing unobservable treatment resistance. In addition, Figure 5 shows our segmented MTEs curves and potential outputs, which elaborate on the MTEs estimation and the difference between Y₁ and Y₀.

OPEN IN VIEWER

Figure 3.

Estimated MTE curve.

OPEN IN VIEWER

Figure 4.

MTE for compliers with LATE weights.

OPEN IN VIEWER

Figure 5.

MTEs and potential outcomes.

Heterogeneity analysis

We have discussed GreCEO's robustness and positive impact on GI; however, these may vary with changes in the characteristics of energy firms. Therefore, we investigate the heterogeneity of GreCEO's effect on GI of two types of firms. Table 5 shows the results of heterogeneity induced by two factors: whether it is a state-owned enterprise (SOE) or a high-tech enterprise (HT). Columns (1)–(4) of Panel A in Table 5 show that GreCEO can substantially improve GI_IP and GI_UMP in SOEs and non-SOEs. However, the coefficients of GreCEO in SOEs are smaller than in non-SOEs. It can be explained that, CEOs of non-SOEs may have greater economic agency power than those of SOEs. They can appropriately utilize their personal attributes in firms’ innovation activities. ⁶⁴ SOEs show a strong path dependence on production, lack innovation motivation, and tend to be satisfied with only the current practical technology, ignoring green invention patent innovation. ⁶⁵ Specifically, innovation strategies in SOEs are more likely to be influenced by the government's green development philosophy and suffer from fewer financing constraints than non-SOEs. ⁶⁶ Most SOEs can increase their investment for GI regardless of whether the CEOs have green experience. In contrast, CEOs with green experience in non-SOEs gain support from their stakeholders and significantly reduce financing constraints. These firms devote more resources to innovative activities. Therefore, there is not much potential room exists for CEOs with green experience to improve GI in SOEs. The GI effect of GreCEO is stronger in non-SOEs than in SOEs.

OPEN IN VIEWER

Table 5.

Results for different internal factors of firms.

	GI_IP		GI_UMP
Panel A: Heterogeneity of SOE and non-SOE	SOE	Non-SOE	SOE	Non-SOE
	(1)	(2)	(3)	(4)
GreCEO	0.486***	0.559***	0.272***	0.448***
	(0.072)	(0.082)	(0.069)	(0.075)
Controls	Yes	Yes	Yes	Yes
Adj. R2	0.479	0.268	0.396	0.274
Observations	4333	5738	4333	5738
Panel B: Heterogeneity of HT and non-HT	HT	Non-HT	HT	Non-HT
	(1)	(2)	(3)	(4)
GreCEO	0.619***	0.582***	0.375***	0.363***
	(0.096)	(0.065)	(0.075)	(0.068)
Controls	Yes	Yes	Yes	Yes
Adj. R 2	0.230	0.356	0.267	0.313
Observations	2583	7490	2583	7490

The above regressions all control for year, city, and industry fixed effects, and the results of the control variables are not exhibited; the subsequent tables are the same; robust standard errors are reported in parentheses; ***p < 0.01, **p < 0.05, * p < 0.1.

In addition, Columns (1)–(4) of Panel B in Table 5 show significant positive effects of GreCEO on GI_IP and GI_UMP for HT and non-HT firms. In contrast, the coefficients of GreCEO in HT firms are larger than in non-HT. HT firms, which are more technology-dependent, may also be highly dependent on innovation resources, including CEOs’ GI experience. Faced with fierce market competition, CEOs of HT firms focus on technology R&D in their own expertise area, with a shorter overall cycle of iterative upgrading from product design to specialized production to marketing. ⁶⁷ In particular, HT firms’ main strategic goal is R&D, and the GreCEO of HT energy firms promotes the rapid transformation of R&D results into GI. In contrast, CEOs of non-HT firms are less motivated to innovate in green technologies. Therefore, the effect of GreCEO on GI in HT firms is more prominent than in non-HT firms.