Timber Industrial Policies and Export Competitiveness: Evidence from China’s Wood-Processing Sector in the Context of Sustainable Development

Abstract

In the era of climate change, the strategic importance of forestry products for sustainable development is increasingly recognized. Amid a global resurgence of industrial policy aimed at addressing environmental challenges, this study investigates the impact of China’s central and provincial green industrial policies on the export competitiveness of wood-processing enterprises. Utilizing firm-level data from the China Industrial Enterprise Database and China Customs Export Database (2000–2013), we apply a double machine learning (DML) approach and construct a heterogeneous competitiveness model to evaluate policy effects along two dimensions: export quantity (volume and intensity) and export quality (product complexity and consumer-perceived quality). Our findings reveal a clear dichotomy in policy outcomes. While industrial policies have significantly improved export product complexity—reflecting China’s comparative advantage in labor-intensive production—they have had limited or even negative effects on export volume, intensity, and product quality. This suggests that current policy frameworks disproportionately reward horizontal innovation (product diversification) while neglecting vertical upgrading (quality enhancement), thereby hindering comprehensive export performance gains. Those results highlight the need for more balanced and targeted policy design. By aligning industrial policy instruments with both complexity and quality objectives, policymakers can better support the sustainable transformation of China’s forestry sector and enhance its competitiveness in global value chains.

1. Introduction

According to United Nations research, over 1.6 billion people worldwide rely on forest resources for survival, development, and livelihood [1]. Amidst climate change, forestry products are pivotal for green and sustainable development [2]. Forests and forest-based industries play a crucial role in achieving the Sustainable Development Goals (SDGs), particularly in combating climate change (SDG13), promoting sustainable economic growth (SDG8), and protecting terrestrial ecosystems (SDG15). Globally, more than 1.6 billion people depend on forest resources for livelihoods, employment, and income generation. The promotion of harvested wood products (HWPs), such as construction timber and biomass energy, has emerged as a key strategy for enhancing the economic and ecological value of forest ecosystems [3,4]. However, developing countries continue to face the dual challenges of persistent deforestation and poverty. Unlike developed nations such as Finland and Sweden, which have achieved industrialized and sustainable forest sectors through long-term policy interventions, many developing countries struggle to balance forest utilization with ecological conservation. This contradiction is particularly acute in China, a country that has become one of the world’s largest producers and exporters of wood-processing products and is deeply embedded in global forest product value chains [5]. Forestry-based industries are increasingly recognized as important engines for inclusive green growth, especially in developing countries, where they can generate employment, stimulate exports, and enhance environmental sustainability [6]. The forest sector has great potential for contributing to inclusive green growth, as it is labor intensive, provides significant export opportunities, and, if managed sustainably, can help to mitigate climate change. While some forestry policies are adapting to these challenges [7], developing countries grapple with the dual challenges of deforestation and poverty [8]. In contrast to the sustainable and industrially advanced forest sectors observed in countries like Finland and Sweden, many developing nations struggle to reconcile forest resource utilization with conservation goals, resulting in ongoing deforestation and the exacerbation of poverty in forest-dependent regions. Meanwhile, government intervention is gaining traction, especially amidst financial and economic crises, leading to a resurgence of industrial policy [9]. China plays a significant role in forest trade product development, being a major producer and participant in the global value chain of the wood industry [10,11]. Deeply integrated into the global value chains (GVCs) of various forest products [12], China’s forest products industry faces challenges in maintaining its position amidst rising domestic labor costs and global sustainability demands [13].

In recent years, there has been a resurgence of industrial policy as a tool for promoting sustainable development, particularly under the influence of climate goals and green economic transitions. As the wood-processing industry encompasses economic, social, and ecological attributes simultaneously, government intervention in this industry is essential. The central Chinese government has successively issued numerous policies for the wood-processing industries. Developing countries have experimented with a range of industrial policies and policy combinations. Developing and developed countries face poverty, climate change finance crises, and other issues. Government intervention is becoming increasingly popular, and financial and economic crises have prompted a remarkable renaissance in industrial policy [14]. The effectiveness of industrial policies in emerging markets and countries is dissimilar [15]. Otherwise, the effectiveness of industrial policies in developing countries would be beneficial for other developing countries. However, some policy lessons from this look at recent economic history, which is different for economies at different stages of development [16]. The export performance of rapidly growing developing countries [17,18] suggest that export growth may play a key role in helping countries attain high income levels. Previous research on industrial policy evaluation, using the difference-in-difference method, propensity matching score [19,20], or government subsidies [21,22,23] as measurement variables, all rely on a strict hypothesis system and a priori set antecedent relationships between explanatory variables and explained variables. Moreover, if subsidies are used to define industrial policies, it is impossible to distinguish between central and local industrial policies.

This study aims to fill this gap by examining the impact of China’s central and provincial green industrial policies on the export competitiveness of firms in the wood-processing sector. Drawing on the Melitz model of heterogeneous firms, we construct a theoretical framework that links government intervention with firm-level production and export behaviors. To address the methodological limitations of traditional econometric models, we employ a double machine learning (DML) approach, which accommodates high-dimensional controls, mitigates endogeneity, and avoids the mis-specification risks of parametric regressions.

In Section 2, we conduct a literature review about forestry industrial policy and enterprise competitiveness, China’s forest industrial policy change and theoretical mechanism analysis, and the proposed research hypotheses. In Section 3, we construct the econometric model based on DML methods, and specific variables and data sources are elaborated. In Section 4, the results of the econometric models are presented, as well as a series of robustness testing, heterogeneity analysis. In Section 5, the results of the empirical analysis are discussed. In Section 6, we summarize the findings and describe the limitations of the study.

2. Literature Review and Theoretical Analysis

2.1. Literature Review

This study primarily investigates the impact of the forestry industrial policy on a company’s export performance, namely competitiveness. Therefore, it is closely associated with research on industrial policies, company export behavior, and the international competitiveness of forest products.

Some studies focus on the relationship between industrial policy effectiveness and causal effects. The eligibility of policy subsidies to identify the causal effects of industrial policies [24,25]. Governments must intervene because market mechanisms such as prices alone fail to bring about drastic and rapid changes to the fabric of the economies required to protect our planet [26]. Competitiveness became crucial to regional and national industrial policy in the 1990s. Industrial policy is a neglected research area that might seem to be in abeyance with “sectoral” to “horizontal” measures. Competitiveness is not a formal economic concept but has developed as part of the policymaking process in an ill-defined way [27]. The effect of trade policy uncertainty on firms’ export behavior using panel data of Chinese listed companies in the industrial sector based on the new structural economic theory and heterogeneity of firm viability [28]. The empirical results show that high trade policy uncertainty significantly inhibits the extensive and intensive margins of firm exports. Industrial policy aims to explain how companies can occupy a position in global competition when resource factors are not circulated because of government intervention in developing countries, whereas corporate competitiveness theory focuses on how companies can maintain their positions in the long term [29]. This provides a competitive advantage. In other words, there is an excess profit. In recent years, scholars have researched how industrial policies under the influence of government intervention affect industry development [30,31,32,33]. Additionally, the firm’s export competitiveness is impacted by credit constraints and some financial problems [34]. Some studies measured competitiveness using productivity and efficiency [35]. China’s environmental policy, its Two Control Zones, and its export effectiveness [36]. Reputation comparative advantage model and conducts research was construct on export industrial policies [37]. Some research is about the relationship between industrial and competitive policies [38]. The green industrial policy frameworks is to improve and counter both informational and political risks [39]. “Efforts to build a green industry have raised the economic and geopolitical stakes of environmental issues, as states seek to position their firms in global value chains and reshore strategic industries.” Research studies industrial policy and its effectiveness using a model of both market and government failures [40]. He introduces a public agency responsible for industrial policy into self-discovery model [41,42]. When public agencies are poorly informed, industrial policies are ineffective but not necessarily highly politically motivated. Given a politically motivated public agency, industrial policies are effective if and only if the institutional setting ensures that such policies are modes, for example, restricting the agency’s budget [43].

Moreover, some research highlights the Chinese forest export trade, focusing on forest products international trade and competitiveness, particularly from an industry-level perspective. Export competitiveness is a new interdisciplinary subject of strategic management and international trade. It has good practicality and applicability and therefore has attracted wide attention from academia and policy circles [44]. Previous studies on the export competitiveness of the wood processing industry have mainly focused on the industry level, focusing on the international competitiveness of the industry and studying the factors that affect this factor from aspects such as labor and marketization [45], or studying the international competitiveness of a specific type of forest products in the wood processing industry, such as wood-based panels and papermaking [46], mainly using Michael Porter’s diamond model and many models derived from it. With the development of big data, “export product quality”, “export product technical complexity”, “export product diversity”, and other structural and quality competitiveness issues and theories have been proposed [47,48].

Overall, this study addresses this research gap. First, it uses microdata to measure export competitiveness from the quantity and quality dimensions using enterprise export volume, intensity, technical complexity, and quality as dependent variables, which can better reflect enterprise heterogeneity characteristics. Second, it uses double machine learning to correct bias and high-level data problems.

2.2. China’s Forestry Industrial Policy Background

Industrial policy is a systematic industrial plan and a specific policy formulated by a country to achieve its economic growth goals and industrial development. As representatives of local interests and competitors, local governments have an independent tendency to formulate and implement industrial policies based on their own local interests. Factors such as the interests of local governments and their departments, personal promotion opportunities, and social responsiveness of policies affect the formulation and implementation of local government industrial policies.

In fact, China’s wood industry policies grew significantly after 2003, and the role of government intervention continued to increase. With the central Chinese government prioritizing the ecological environment in the late 1990s and focusing on industrial development in rural mountainous areas, the conflict between “ecological needs and economic needs” in the wood-processing industry intensified. The central level not only meets the ecological needs of the wood-processing industry through the implementation of the “natural forest protection” project but also meets the economic needs of the industry through industrial policy measures such as the “2003 Decision of the Central Committee of the Communist Party of China on Accelerating Forestry Development.” Since World War II, developing countries have adopted industrial policies to protect nascent industries and enhance the competitiveness of domestic industries against developed countries. In the first 30 years following the war, these policies promoted industrial development. Seeing Figure 1, since 1978, there has been a gradual increase in the issuance of industrial policies, peaking in 1989. This surge coincided with a pivotal period in China’s economic paradigm marked by the adoption of selective industrial policies reminiscent of those employed in Japan. In 1989, the State Council promulgated a landmark policy directive on the key points of current industrial policy. Notably, this document represents the inaugural formal industrial policy since the inception of the People’s Republic of China (PRC). Emphasizing the imperative of comprehensive wood utilization, production optimization, and advancement of plywood within the realm of artificial boards, it sets a strategic course for the wood-processing sector. Subsequent policy initiatives have underscored the sustained focus on timber-related industries, particularly after the issuance of the Decision on Accelerating Forestry Development by the Central Committee of the Communist Party of China in 2003. This pivotal decree catalyzed a notable surge in policies geared towards fostering the growth and sustainability of the wood-processing industry.

2.3. Theoretical Analysis

2.3.1. Defining Export Competitiveness: Quantity vs. Quality

Competitiveness became a key part of regional and national industrial policy in the 1990s. Competitiveness refers to the degree of advantage a country, region, enterprise or product has over its competitors in the international market or a specific market and is an indicator of its ability to successfully compete [49]. At the national level, competitiveness may include macroeconomic stability, infrastructure construction, education level, technological innovation, and other aspects; at the enterprise level, competitiveness may involve product quality, brand image, marketing strategy, cost control, and other aspects. In the field of strategic management, Professor Michael Porter answered this question from a new perspective and proposed a concept corresponding to comparative advantage, namely competitive advantage [50]. Porter believes that the productivity of a country ultimately depends on the production capacity of its enterprises, and that competition between countries is actually between enterprises; a country cannot have international competitive advantages in all industries, but rather concentrates on a few limited industries, and industries should be the basic unit for studying national competitive advantages; the source of international competitive advantages of industries is not only the “factor conditions” in trade theory, but the formation of the “diamond system” on which industrial development depends—the interrelationship of factor conditions, demand conditions, competitive background of corporate strategy and structure, related supporting industries, opportunities, and government [51]. At the micro level, Melitz (2003) [50] introduced firm-level heterogeneity into a model of trade by adapting a Ricardian model to firm-specific comparative advantage. Firm export competitiveness refers to the ability of enterprises to obtain the best market share and profit in foreign markets with lower costs (cost strategy) and differentiated products (product innovation). In the international environment, firms have higher export performance and excess profits than other competitors under scarce resources (i.e., competition and non-monopoly conditions). Competitiveness of enterprises in the context of globalization is actually the international competitiveness of enterprises [52]. Figure 2 illustrates the conceptual model that guides this study. Government industrial policies influence firm-level export competitiveness through two primary channels—quantity and quality. These effects are mediated by firm heterogeneity and shaped by the broader institutional and market environment.

2.3.2. The Impact of Industrial Policies on the “Quantity” of Firms’ Export Competitiveness

Melitz develops a dynamic industry model dynamic industry model monopolistic competition with heterogeneous firms. The model shows how the exposure to trade induces only the more productive firms to export while simultaneously forcing the least productive firms to exit [53]. However, in China, the improvement of enterprise productivity depends on the influence of government, not spontaneous improvement. Taking the labor-intensive industry of wood processing as an example, the development of forest product trade is mainly due to low labor costs and entrepreneurial spirit, which promote corporate export trade. Although a series of industrial policies introduced by the government have improved productivity, they have suppressed the self-selective export behavior of enterprises, resulting in China’s “corporate export rate paradox” of low productivity and high exports or high productivity and low exports. Industrial policy, defined as government intervention aimed at influencing the structure and performance of industries, can affect firm competitiveness through multiple channels. From the neoclassical perspective, such policies correct market failures and enhance productivity. However, scholars such as Porter emphasize that the effectiveness of government intervention depends on the ability to foster a competitive business environment, rather than distorting market signals [54]. In developing economies, excessive intervention may lead to unintended consequences such as rent-seeking, misallocation of resources, or collusion between governments and firms. This is particularly relevant for the wood-processing industry in China, where labor-intensive production has historically benefited from low costs and entrepreneurial dynamism. Although government policies have enhanced productivity in some areas, they may suppress firms’ self-selection into export markets—resulting in a paradox of low-productivity firms exporting, or high-productivity firms not exporting. Therefore, we propose the following hypotheses:

Hypothesis 1. 

Industrial policy is negative for wood-processing company “quantity” competitiveness.

2.3.3. The Impact of Industrial Policies on the “Quality” of Firms’ Export Competitiveness

From the perspective of innovation economics, product quality and complexity are central to long-term competitiveness. However, investing in high-quality, technologically advanced products requires significant R&D expenditure, incurring high fixed costs and uncertain returns. In such contexts, government support through subsidies, tax incentives, and trade facilitation can alleviate the cost burden and reduce market entry risks [54,55]. Nevertheless, in practice, the effectiveness of industrial policies in stimulating innovation is mixed. Firms may engage in short-term opportunism, using subsidies for arbitrage rather than genuine innovation. Moreover, government utilitarianism—emphasizing measurable short-term performance over long-term innovation—may distort firm incentives, encouraging them to focus on increasing complexity (e.g., product variety or technology layers) while neglecting quality improvements valued by end consumers (e.g., durability, safety, brand reputation). In order to reflect this duality, we propose the following hypotheses:

Hypothesis 2. 

The impact of industrial policy on quality competitiveness is heterogeneous across firms and products.

Hypothesis 3a. 

Industrial policy positively influences horizontal innovation, enhancing product complexity.

Hypothesis 3b. 

Industrial policy negatively influences vertical innovation, diminishing perceived product quality.

3. Materials and Methods

3.1. Materials and Data

Enterprise-level data were obtained from the China Industrial Enterprise Database and the Customs Trade Database. The National Bureau of Statistics publishes the China Industrial Enterprise Database, which primarily comprises quarterly and annual reports submitted by the sample enterprises to the local statistical bureau. The database includes the “database of all state-owned and non-state-owned industrial enterprises above designated size” (The annual main business income (i.e., sales revenue) of enterprises is CNY 5 million or more. After 2011, it is CNY 20 million or more). This study conducts data screening of the database to screen out data from wood-processing enterprises and remove unreasonable data, such as missing enterprise data on total assets and current assets, as well as data that do not comply with accounting calculation standards and have obvious statistical errors. Moreover, based on the standard field information of the two databases, including the enterprise name, contact person, phone number, postal code, and other information, multiple iterative matches were performed to obtain matched export enterprise data.

China’s forestry industrial policy was based on the Peaking University Law Database (http://www.pkulaw.cn). We collected data on the Chinese forestry industrial policy in 1949. In particular, due to the 1994 decentralization reform in China, industrial policy was concluded between the central and local governments. Therefore, this study divides industrial policies into central and local levels. The process of collecting China’s forestry industry policies for this study is as follows: We collected the policy documents using the following keywords: “wood”, “wood processing”, “sawn wood processing”, “woodchip processing”, “artificial board”, “plywood”, “fiberboard”, “particle board”, “wood products”, “wood containers”, “bamboo products”, and “forestry residues.” The latter was added because China introduced the forestry residues reform to increase the efficiency of forest products.

3.2. Models and Methods

Traditional parametric regression methods require preset functional forms between variables, such as linear regression, which can lead to the problem of mis-specified functions. However, as a non-parametric regression, the double machine learning model does not require a preset function and avoids misconfiguration bias. In addition, traditional linear regression is unsuitable for large-sample multi-variable scenarios, which will bring about problems such as the curse of dimensionality, multicollinearity, and limited control of critical variables, resulting in biased variable estimation [56].

This study used double machine learning to estimate the effect of industrial policies. Here is the double machine learning calculation process. We used the Stata 17 software for the calculations. Equation (1) is the basic model. IP is the policy/treatment variable of interest and represents industrial policies. $y_{i\mathrm{j}t}$ the firm’s export competitiveness and is the main regression coefficient that we would like to infer. $Z_{ijt}$ is the control variable. In double machine learning, if IP is exogenous conditional on controls Z, ${\theta }_0$ has the interpretation of the treatment effect parameter. Equation (2) keeps track of confounding, namely the dependence of treatment variable on controls. This equation is not of interest per se, but it is important for characterizing and removing regularization bias. The confounding factors X affect the policy variable IP via the function $m_0(Z)$ and the outcome variable via the function $g_0(Z)$.

$$y_{ijt}={\theta }_0*I{P}_{jt}+g_0(Z_{ijt})+U_{it},E[U_{it}|Z_{ijt},I{P}_{jt}]=0$$

(1)

$$I{P}_{\mathrm{jt}}=m_0(Z_{ijt})+V_{ijt},E(V_{ijt}|I{P}_{jt})=0$$

(2)

3.3. Description of Variables

3.3.1. Dependent Variables

$y_{i\mathrm{j}t}$ includes the export value ($lnexpor{t}_{ijt}$), since the enterprise production function usually uses the Douglas production function, the enterprise’s production value is exponentially related to other production factors. Therefore, it is more appropriate to take the logarithm of the export value. This study uses $lnexpor{t}_{ijt}$ to represent it, which can intuitively express the enterprise’s export willingness and ability, and is an indicator of the enterprise’s export behavior. Export density ($exintt{y}_{ijt}$), in the new trade theory and the theory of heterogeneous enterprises; the export efficiency of enterprises is affected not only by the export volume but also by the total industrial output value. Therefore, this study borrows the export intensity ($exintt{y}_{ijt}$) to express how much of the total output value of the enterprise is used for export, specifically by dividing the enterprise’s export value by the total industrial output value. To construct export product complexity ($ES{I}_{ijt}$), this study refers to the approach of Wang (2022) [55]. To construct export product quality ($expqu{a}_{ijt}$), this study refers to the approach of Shi Bingzhan (2014) [56] and Khandelwal et al. (2013) [57] (the calculation process is shown in Appendix A).

3.3.2. Core Explanatory Variables

$I{P}_{jt}$ represents the central and local government’s introduced forestry industrial policies. According to the company database, the China economic industry classification (CIC) code was from the Forest Product Classification Catalog of the National Bureau of Statistics of China. If the product type is referred to as a policy, $I{P}_{jt}$ is 1. This study used $I{P}_{jt}>0$ as the experimental group and $I{P}_{jt}<0$ as the control group.

3.3.3. Control Variables

Regarding productivity calculation, Solow introduced the total factor productivity into the neoclassical growth theory, and the total factor productivity of enterprises is one of the main sources of a country’s economic growth. Schultz also mentioned in his “Transforming Traditional Agriculture” that the source of economic growth is technological progress, improving total factor productivity, and the methods for calculating total factor productivity of enterprises include the OP method, LP method, and OLS method. Among Research believed that the existence of OLS and OP methods and endogenous problems affect the final estimation effect [58]. This article refers to the research methods of Bai (2012) [58] and others, and it uses LP to calculate enterprise productivity. This study used the LP model to measure TFP ($tfpl{p}_{ijt}$); number of employees ($employeemea{n}_{ijt}$); $L{P}_{ijt}$ represents the share of labor income. This paper adopts the concept of value added based on factor cost method to estimate, that is, the share of labor income of an enterprise is the cash paid to and for employees/(operating income − operating costs + cash paid to and for employees + depreciation of fixed assets); capital from foreign countries and Hong Kong, Macao, and Taiwan ($capitalfo{r}_{ijt}$; $capitalhm{t}_{ijt}$); and research and development investment ($feer{d}_{ijt}$). At the industrial level, this study used the industrial intensity Herfindahl–Hirschman index ($hh{i}_{jt}$) calculated at the industry level (4-digit code) based on corporate sales revenue.

4. Empirical Results

4.1. Summary Statistics

Table 1. Summary statistics of the selected variables.

Variable	Mean	Std. Dev.	Min	Max
lnexport	8594.696	63,400.886	0.000	4,665,786.000
ex_intty	0.144	0.323	0.000	1.000
ESI	285.605	215.195	0.000	4011.774
expqua	0.578	0.176	0.007	1.000
ip central	0.900	0.301	0.000	1.000
ip local	0.624	0.484	0.000	1.000
employ mean	148.136	319.023	0.000	14,105.000
LP	321.397	401.032	0.000	10,815.846
capital for	1393.520	15,275.569	0.000	1,481,180.000
capital_hmt	1306.556	15,298.706	0.000	1,481,180.000
fee tax	115.298	727.441	−559.000	87,429.000

presents the statistical data of the main variables in the model. The mean export value and export density of a company’s export behavior were 8594.696, with a standard deviation of 63,400.886, indicating significant differences in export values among different wood-processing companies. Therefore, the numerical form of a company’s export value should be logarithmic. In the export innovation behavior of enterprises, the difference in product quality (expqua) is the smallest, with a standard deviation of 0.176, whereas the variance in product technical complexity (ESI) is relatively large, with a value of 215.195. Second, export intensity and export product quality are both continuous variables between 0 and 1, so the selection of regression models cannot be limited to general OLS regression models and panel models. In subsequent studies, a fractional logit regression model is added to the regression of export intensity and export product quality. The fractional logit model is based on the discrete 0–1 value logit model which can handle special cases in which export intensity is a continuous variable between [0, 1] wells.

4.2. Baseline Results

This study utilized double machine learning to estimate the policy effectiveness of a Chinese wood-processing company. The sample segmentation ratio was 1:4 and a random forest algorithm was used to predict and solve the primary and auxiliary regressions. The regression results are shown in

Table 2. Baseline results.

Variables	Quantity Competitiveness		Quality Competitiveness		Quantity Competitiveness		Quality Competitiveness
	lnexport	Ex_intty	ESI	Expqua	lnexport	Ex_intty	ESI	Expqua
ip_central	−0.0846 **	−0.0692 ***	45.9371 ***	−0.0095
	(−2.0032)	(−9.4381)	(4.4378)	(−0.9239)
ip_local					−0.1856 ***	−0.0351 ***	18.8725 *	−0.0230 **
					(−3.6605)	(−7.0496)	(1.8845)	(−2.3775)
ID fixed	YES	YES	YES	YES	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES	YES	YES	YES	YES
Constant	0.0524 ***	−0.0095 ***	−6.0723	−0.0004	0.0440 **	−0.0110 ***	−6.3594	−0.0007
	(2.9690)	(−4.1518)	(−1.4278)	(−0.1008)	(2.4056)	(−4.8058)	(−1.6143)	(−0.1575)
Observations	4057	17,884	2320	1723	4057	17,884	2717	1723

Notes: robust z-statistics in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

.

Broadly, industrial policies indicative of government intervention adversely affect the enhancement of enterprises’ export competitiveness. Specifically, both the central and provincial industrial policies have deleterious effects on export volume, intensity, and quality. Conversely, such policies are conspicuously beneficial for augmenting the complexity of export products. From this analytical perspective, it becomes evident that government influence plays a pivotal role in enriching the complexity of exported goods. The timber industry policies enacted by the PRC have been instrumental in elevating the complexity of export products, thereby facilitating notable advancements in industrial product upgrades. In this context, the notion of product quality implies that when two entities concurrently export artificial panels, the quality of the panels exported by Company A surpasses that exported by Company B. This differential underscores the necessity for independent innovation capabilities and process innovations within enterprises, yet industrial policies inadvertently impede exports. The concept of “implicit innovation” within product quality demonstrates a negative and significant repercussion, suggesting that industrial policy has inadvertently fostered “rent-seeking” behaviors and “profit dilution” phenomena, thereby undermining authentic technological innovation within China’s wood-processing sector.

Regarding the export behavior of companies, industrial policies at both the central and provincial government levels were found to negatively influence export volume and intensity. This finding aligns with the analysis presented by Lin (2019) [47], who scrutinized the effects of government intervention on the export volumes of paper companies.

In the realm of company innovation behavior, the negative impact of provincial industrial policies on export product quality is markedly more significant than that of central industrial policies. Nevertheless, central industrial policies have a more favorable influence on the complexity of export products. A plausible explanation for this discrepancy is that government subsidies and financial support are predominantly sourced from the central government and banking institutions.

4.3. Robustness Testing

Based on these estimation results, the robustness of the model was tested to obtain more reliable estimation results. After the empirical tests, the core conclusions of this study were confirmed. However, this regression analysis may also face an endogeneity bias caused by bidirectional causality and sample selection issues.

4.3.1. Industrial Policy Variables Lag for One Period

First, there may be a reverse causal relationship between the outcomes. For reverse causality, this study uses the lagged period of the explanatory variable as a weak instrumental variable for testing. The robustness test in

Table 3. Lag one period.

Variables	lnexport	Ex_intty	ESI	Expqua	lnexport	Ex_intty	ESI	Expqua
ip_cenlag	−0.0959 *	−0.0783 ***	−0.0959 *	−0.0206
	(−1.8313)	(−8.0525)	(−1.8313)	(−1.5478)
ip_localag					−0.1271 **	−0.0274 ***	44.5668 ***	−0.0303 ***
					(−2.4801)	(−4.8732)	(3.7884)	(−2.9288)
ID fixed	YES	YES	YES	YES	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES	YES	YES	YES	YES
Constant	0.0415 **	0.0009	−0.0004	0.3760	0.0376 *	−0.0017	1.9783	−0.0015
	(1.9930)	(0.3457)	(−0.1414)	(0.0824)	(1.7777)	(−0.6121)	(0.4245)	(−0.3178)
Observations	3131	13,725	12,305	2093	3131	12,305	2093	1354

Notes: robust z-statistics in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1.

uses the lag period of the central government industrial policy and provincial industrial policy as the core explanatory variables to regress the dependent variables. The regression results show that the sign and significance of the core variable remained largely unchanged.

4.3.2. Change the Machine Learning Models

The baseline results of the machine learning models selected in this study were all random forest models. To avoid the influence of biases in the setting of double machine learning models on the conclusions, this study selected the Lasso and Nnet models as replacements. The results in

Table 4. Lasso regression model.

Variables	lnexport	Ex_intty	ESI	Expqua	lnexport	Ex_intty	ESI	Expqua
ip_central	−0.1539 ***	−0.0837 ***	61.3172 ***	−0.0029
	(−3.2998)	(−9.8360)	(5.8078)	(−0.2445)
ip_local					−0.0582	−0.0488 ***	34.9551 ***	−0.0203 **
					(−1.4877)	(−9.5609)	(3.8318)	(−2.1945)
ID fixed	YES	YES	YES	YES	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES	YES	YES	YES	YES
Constant	−0.0063	−0.0001	1.4768	−0.0004	−0.0047	−0.0002	2.0224	−0.0010
	(−0.3320)	(−0.0225)	(0.3398)	(−0.0793)	(−0.2482)	(−0.0659)	(0.4510)	(−0.2102)
Observations	4057	17,884	2320	1506	4057	17,884	2320	1506

Notes: robust z-statistics in parentheses *** p < 0.01, ** p < 0.05.

and

Table 5. Nnet regression model.

Variables	lnexport	Ex_intty	ESI	Expqua	lnexport	Ex_intty	ESI	Expqua
ip_local	−0.1690 **	−0.8562 ***	0.5616 **	−0.0525
	(−2.1776)	(−10.5661)	(2.1918)	(−1.2444)
ip_central					−0.0160 **	−1.7696 ***	0.4090 ***	−0.1063 ***
					(−2.2591)	(−7.3756)	(3.4351)	(−3.5604)
ID fixed	YES	YES	YES	YES	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES	YES	YES	YES	YES
Constant	4.2759 ***	1.9547 ***	75.7350 ***	−8.2537 ***	5.2437 ***	5.8252 ***	93.8935 ***	−2.8649 ***
	(5.3982)	(3.1615)	(13.9088)	(−6.0051)	(12.3892)	(3.3282)	(19.9830)	(−4.7313)
Observations	4057	17,884	2320	1506	4579	20,126	2717	1723

Notes: robust z-statistics in parentheses *** p < 0.01, ** p < 0.05.

show that the signs and significances of the core variables remained largely unchanged.

4.3.3. Replacing Industrial Policy Variables

Considering the variable measurement errors, this study changed the industrial policy variables. We used the number of industrial policies that represent IP (policy_n).

Table 6. Changes in the measurement of industrial policy.

Variables	lnexport	Ex_intty	ESI	Expqua
policy_n	−0.0258 ***	−0.0116 ***	7.1847 ***	−0.0040 ***
	(−5.9071)	(−25.4604)	(7.0978)	(−3.2288)
ID fixed	YES	YES	YES	YES
Year fixed	YES	YES	YES	YES
Constant	0.0172	−0.0108 ***	3.3776	−0.0056
	(0.9117)	(−4.8283)	(0.7378)	(−1.1036)
Observations	4057	17,884	2320	1506

Notes: robust z-statistics in parentheses *** p < 0.01.

shows that there was no change in the direction of the correlation coefficients.

4.3.4. Replacing Econometric Methods for Export Intensity and Export Quality

OLS is usually sufficient due to the fact that the dependent variable indicates the probability (such as the probability of raining on a certain day) is between 0 and 1, and that there is no accumulation at both ends (for example, there are too many observations with a value of 0 or a value of 1). The dependent variable was a ratio value of 0–1, and there was no accumulation. Fractional regression is similar to logistic regression in that it can be used to model variables that take values within a bounded range. A key difference is that Y can be measured continuously and does not need to be converted into categories. The statistical significance of quality increased (see

Table 7. Fractional logit model.

Variables	Ex_intty	Expqua
IP	−0.115 ***	−0.0354 ***
Control variable	YES	YES
Constant	−1.144 ***	0.459 ***
Observations	17,609	1555

Notes: robust z-statistics in parentheses *** p < 0.01.

).

4.4. Heterogeneity Analysis

4.4.1. Policy Type Heterogeneity

We divided industrial policies into environmental regulations and economic incentive policies. First, environmental regulation was found positive for export volume and export product complexity, although it was not significant. Environmental regulations had a significantly negative effect on export intensity. An incentive policy was found significant for innovation behavior. They were significantly positive for export product complexity and passive for export product quality. Environmental regulations and economic policies were not significant for export volume. Regarding innovation effects, economic industrial policy was more significant for product complexity and quality. Regarding the export effect, the regulatory policy had a positive effect on export volume, although it was not significant. Environmental regulation and economic policies were similar in terms of export intensity (

Table 8. Environmental regulation policy and export competitiveness.

Variables	lnexport	Ex_intty	ESI	Expqua
local_gz	0.0152	−0.0146 ***	11.6791	−0.0004
	(0.3876)	(−2.9402)	(1.1147)	(−0.0501)
Constant	0.0522 ***	−0.0096 ***	−4.4480	0.0018
	(3.1373)	(−4.3770)	(−1.0976)	(0.4255)
Observations	4579	20,126	2717	1723

Notes: robust z-statistics in parentheses *** p < 0.01.

and

Table 9. Incentive policy and export competitiveness.

Variables	lnexport	Ex_intty	ESI	Expqua
local_fc	−0.0363	−0.0259 ***	42.4947 ***	−0.0464 ***
	(−0.9638)	(−5.9213)	(5.0004)	(−5.2532)
Constant	0.0505 ***	−0.0089 ***	−4.1899	0.0012
	(2.9944)	(−4.0704)	(−1.1046)	(0.2883)
Observations	4579	20,126	2717	1723

Notes: robust z-statistics in parentheses *** p < 0.01.

).

4.4.2. Productivity Heterogeneity

We compared the exports and non-exports of wood-processing companies in terms of factor productivity. Thus, the factor productivity of exporting companies was higher than that of non-exporting companies. Melitz introduced a heterogeneous model that considers productivity to be an important factor in the differences in enterprises’ export behaviors. Industrial policy is significantly beneficial for factor productivity; however, it negatively affects export volume and intensity. In heterogeneous model, only productive companies incur export costs. Additionally, Chinese factor productivity has also been influenced by government intervention. However, the effect of the cost was not significant. In addition, this study provides evidence for China’s export puzzle. Despite this, industrial policies significantly promoted factor productivity. Company productivity in China is based on government intervention. For the different policy types, the coefficient of the environmental regulation policy was larger than that of the incentive policy (

Table 10. Summary statistics of export and non-export company productivity.

Export Company	N	Mean_tfp	SD	Min	Max
non export	32,866	744.477	1039.250	0.202	20,612.867
export	14,916	913.597	1915.041	0.219	89,469.094

and

Table 11. Industrial policy and factor productivity.

Variables	lntfp_lp	lntfp_lp	lntfp_lp	lntfp_lp
ip_local	0.1409 ***
	(16.4658)
ip_central		0.0359 ***
		(3.1094)
local_gz			0.1234 ***
			(14.2012)
local_fc				0.0752 ***
				(9.2232)
Constant	0.0054	0.0028	0.0034	0.0029
	(1.3633)	(0.6949)	(0.8550)	(0.7295)
Observations	17,887	17,887	17,887	17,887

Notes: robust z-statistics in parentheses *** p < 0.01.

).

5. Discussion

This study finds that both central and local industrial policies significantly influence the export competitiveness of China’s wood-processing firms. While these policies positively affect product complexity, their impact on export volume, intensity, and quality is either neutral or negative. These findings suggest that existing policies tend to encourage horizontal product innovation—the diversification and technological layering of products—rather than vertical upgrading related to intrinsic quality improvement.

Our findings are consistent with Martincus (2008) [59], who observed that industrial policy in developing countries, such as Peru, often fails to stimulate export expansion. Similarly, Su et al. (2020) [60] noted that China’s forest products industry has shifted from a “large but weak” status to a more competitive position within the global value chain, with increasing export complexity. Schott (2008) [53] and Amiti and Freund (2010) [61] also found that China’s exports have become more sophisticated over time, entering product categories traditionally dominated by developed economies. However, this study diverges from the optimistic view of industrial policy proposed by Lin (2014) [18], which posits that latecomer countries can emulate advanced economies’ innovation paths through government-led industrial upgrading. Our results suggest a more nuanced reality: while industrial policies may enhance productivity and complexity, they do not necessarily translate into improved product quality, indicating a potential misalignment between policy tools and long-term innovation goals.

A key explanation for these findings lies in the recognition and evaluation mechanisms embedded in industrial policy. Governments are more capable of identifying and supporting complexity-related innovation—such as the number or types of exported products—than they are at assessing improvements in product quality, which is often subjective and dependent on consumer perception. Moreover, short-term performance metrics and rent-seeking behavior may incentivize firms to pursue extensive innovation (e.g., expanding product lines or adopting intermediate technologies) rather than investing in brand value, functional performance, or certification systems. This behavior may contribute to the observed “productivity–export paradox”, where firms demonstrate higher productivity without corresponding improvements in export quality.

Our results have important implications for policymakers aiming to foster sustainable forestry development. First, industrial policies should move beyond capacity expansion and support vertical innovation, including high-quality product development, branding, and market certification. Second, a stage-based policy design is necessary: in early stages, policies should emphasize cost reduction and scale; in later stages, they should encourage market-based innovation and quality upgrading. In the context of sustainable development, governments must balance quantitative growth with qualitative improvement. Otherwise, reliance on complexity alone may hinder the long-term competitiveness and ecological resilience of the forestry sector in global markets.

6. Conclusions

This study offers new empirical insights into the effects of industrial policy on export competitiveness in China’s wood-processing industry. Our findings reveal a nuanced picture.

First, China’s timber industrial policy has historically been shaped by limited forest resources and strong domestic demand, leading to an emphasis on the “comprehensive utilization of timber” since the founding of the PRC. Although China has emerged as a global leader in artificial board production, it remains primarily focused on mid- to low-end products such as plywood, where its comparative advantages in labor costs are more pronounced. In contrast, higher-complexity products such as particleboard and fiberboard show limited international competitiveness. Meanwhile, advanced forestry countries like Finland have leveraged technological superiority to achieve higher value-added production.

Second, our analysis reveals a paradox in the role of industrial policy. While such policies have significantly boosted firm-level productivity and product complexity, they have not effectively enhanced export volume, intensity, or product quality. This paradox likely stems from a policy orientation that prioritizes production value over export competitiveness. In the context of rapid domestic economic expansion and rising internal demand—particularly from the real estate sector—industrial policy has tended to incentivize output growth rather than targeted export upgrading.

Third, although industrial policy has encouraged firms to increase export product complexity—an explicit, easily measured form of innovation—it has been less effective in stimulating improvements in product quality, which depends more on intangible attributes such as consumer satisfaction, brand perception, and long-term supply chain optimization. These findings suggest that while complexity-based innovation aligns well with government subsidy mechanisms, quality upgrading requires a more nuanced policy toolkit that incorporates branding, certification, and international market positioning.

From a policy perspective, these results underscore the need for differentiated industrial strategies that align with comparative advantages across product categories. In the early stages of industrial development, selective support for infant industries may facilitate technology imitation and catch-up. However, persistent deviation from comparative advantage through large-scale subsidies can lead to resource misallocation and long-term inefficiency. Policymakers should therefore ensure that industrial policy supports both horizontal and vertical innovation while maintaining efficiency and sustainability.

Finally, we acknowledge several limitations in this study. First, due to data constraints, we could not perform a fine-grained semantic analysis of policy documents to distinguish between different government policy instruments or attitudes. Second, the theoretical framework could be further strengthened by incorporating formal mathematical modeling to distinguish the mechanisms through which industrial policies influence export behavior and innovation outcomes. Future’s research could benefit from integrating policy text mining techniques and firm-level case studies to trace how specific policy tools—such as R&D subsidies, export rebates, or green certifications—affect firms’ strategic decisions. Such extensions would provide deeper insights into how to design more adaptive, innovation-driven, and sustainability-oriented industrial policies for forestry development in China and other emerging economies. Finally, this article only provides a preliminary perspective on industrial policy and export competitiveness and provides a perspective on the heterogeneity of industrial policy and export competitiveness for the sustainable development of developing countries. In the future, researchers can conduct an in-depth analysis of the mechanism between the two.