Socio-Economic Aspects of Drought Impact on Forest Industry in China

Abstract

Extreme drought events under the background of global warming pose a significant threat to China’s forest industry system. This study aims to analyze the impact of drought on China’s forest industry from the perspective of economics, and, through the mechanism test, to further study the communication path of drought to the forest industry. The study finds: (1) drought has been increasing in China in the last 50 years. (2) Drought has a significant restraining effect on the Chinese forest industry. Moreover, inland areas are more vulnerable to drought than coastal areas. (3) Drought will lead to an increase in the occurrence area of forest pests and diseases, forming a compound eco-economic loss effect, thus inhibiting the development of the forest industry. (4) Drought can drive an increase in irrigation facilities. Efficient irrigation may mitigate biological damages caused by drought, so as to resist part of the economic losses. (5) Drought may cause labor to migrate from the forest sector to non-forest sectors, leading to lower production in the forest industry. Labor transfer effects caused by drought are more obvious in the areas where the income gap is relatively big between rural and urban regions, e.g., in inland areas.

1. Introduction

The forest is one of the most important ecosystems on Earth and plays a crucial role in maintaining the global ecological balance and stability of the Earth’s climate system [1]. However, drought strongly restricts the development of forests. Chaudhary et al. found that tree growth declined sharply during droughts, and although it recovered significantly after the drought, it remained below pre-drought levels. Moreover, drought will directly lead to increased tree mortality and change the composition and structure of forests [2]. Global economic losses due to drought are estimated to be as high as USD 0.6–8 billion per year, far more than other climate disasters [3]. Under the background of global warming, drought has become an important ecological and economic issue threatening the forest industry [4].

In China, drought is a major natural disaster. Statistics show that China experiences an average of 10.62 million hectares of drought annually for more than 50 years [5]. According to the Ministry of Natural Resources of China, as of 2025, China’s total forest area is 283.7 million hectares, with forest coverage exceeding 25% of the country’s total land area [6]. The forestry industry is an important force in China’s economic development [7]. The output of China’s forestry industry is keeping increasing. In 2010, the total output value of China’s forestry was RMB 2.28 trillion, and by 2021, it had increased to RMB 8.68 trillion. The growth of the forestry industry is crucial to the sustainable utilization of forest resources, and is closely related to afforestation, biodiversity protection, and carbon sink functions [8,9]. All of these factors create an urgent need to understand the impact of drought on the growth of the forest industry in China.

The impact of drought on forest industry is a complex systemic issue involving multiple fields such as ecology, forestry, meteorology, and economics. Therefore, interdisciplinary research cooperation is particularly important. Most current research focuses on the negative impact of drought on forest ecosystems and tree physiology, but very little research analyses the impact on the forest industry from a social–economic perspective.

This study aims to reveal the impact mechanism of drought on the forest industry, from a social–economic perspective. Our research questions are whether drought had a negative impact on the growth of the forest industry in China in the last 50 years, and what the internal mechanisms of drought stress on the growth of forest industry in China are.

2. Literature Review

2.1. Drought Is Intensifying

As a costly natural disaster, drought has been widely studied by scholars. A large amount of research has found that the global drought phenomenon is intensifying [10,11]. Sergio M. Vicente-Serrano et al. have found that the severity of global agricultural droughts has increased over the past 120 years and made drought predictions based on the global system model [12]. Paola A. Arias et al. pointed out the increasingly severe drought in Central and South America was caused by the low precipitation [13]. Williams et al. found that the severity of drought in the southwestern part of North America increased rapidly from 2021 to 2022, and 19% of it was caused by human activities [14]. Markonis et al. have revealed the fact that the European region has suffered huge economic losses and environmental damage over the past few decades due to the increasingly arid phenomenon [15]. Some research has studied drought in China. For example, Liu et al. studied the drought trend in Sichuan Province, China based on the observation data of 44 meteorological stations [16]. Cao et al. revealed the spatio-temporal variations in drought in the northwest region of China [17]. These research all found that drought was becoming more intensified in a small part of China, but did not reveal the development of drought in the entire China. Therefore, we assume that drought has been increasing in China in the last 50 years.

Hypothesis 1.

Drought has been increasing in China in the last 50 years.

2.2. Impact of Drought on Forest Industry in China

A large number of scholars have studied the effects of drought on forests. Brodribb et al. revealed the long-term impact of drought on the structure and function of forest ecosystems through dendrochronology and remote sensing analysis [18,19]. McDowell et al. analyzed the effects of drought on forest growth, mortality, and carbon cycling through model simulation and data analysis [20]. Gao et al. comprehensively evaluated the impact of drought on forest ecosystem productivity through field experiments and data analysis [21]. Li et al. explored the impact of drought events on vegetation productivity in China by using remote sensing data and statistical analysis [22]. Schwalm et al. concluded that frequent drought events led to a weakened carbon absorption capacity of trees through the observation of remote sensing data and global monitoring network data [23]. Gessler et al. proposed the influence of nutrients on tree growth during drought through case analysis [24]. Sungmin et al. analyzed the response of forest ecosystems to sudden drought events by monitoring data such as leaf area index and solar-induced chlorophyll fluorescence [25]. Hisano et al. found that forests in drought-prone areas were extremely sensitive to environmental changes such as climate warming and drought, and their productivity generally declined due to the pressure of water shortage [26]. Li et al. used leaf area index to measure the impact of drought on vegetation growth [27].

Since it has been established that drought has a negative impact on the productivity of forests, it is reasonable to believe that drought also has a negative influence on the growth of the forest industry in China.

Hypothesis 2.

Drought is negatively related to the growth of China’s forest industry.

2.3. Internal Mechanism of Drought Stress in Forest Industry in China

For the past few decades, a wide range of studies have found that drought hinders plant growth and induces plant death by increasing their susceptibility to wildfire, insect pests, and disease. Yu et al. revealed the different response mechanisms of drought in different grassland ecosystems, providing theoretical support for addressing climate change and adaptive management [28]. Ernandez et al. found that during droughts, highly correlated hydraulic constraints are often the underlying mechanism leading to an increase in forest mortality [29]. Chitra-Tarak et al. discovered that when droughts occur, deep-water channels are an important mechanism for plants to survive [30]. Bose et al. found that one of the main reasons for the accelerated withering of Scottish pines in the past 20 years is self-thinning caused by drought [31].

The internal mechanisms of drought stress in the forest industry involve complex interactions across multiple levels. This study discusses the internal mechanisms from ecological, economic, and social perspectives.

2.3.1. Ecological Mechanism

Disease and pest outbreaks are more likely to occur during drought stress due to the decrease in physiological resistance of trees [32]. If the drought persists, the damage area of coniferous dry pests can increase by 2–3 times [33]. Occurrence area of forest pests and diseases is selected to reflect the direct destruction mechanism of biological stress. Kotlarz and Bejger mentioned increasing climate change effects like pets and diseases may reduce the timber supply in the forest industry, thus cause significant economic impacts [34]. The outbreak of insects would lead to a large number of tree deaths, and have a negative impact on the forest industry [35]. Drought stress can reduce tree resistance and make forest ecosystems more vulnerable to pests and diseases, negatively affecting the forest industry (Keenan et al., 2019) [36]. Thus, it is reasonable to assume that forest pests and diseases are more likely to break out under drought stress in China.

Hypothesis 3.

Drought can curb the development of the forestry industry by increasing the area where forest pests and diseases occur.

2.3.2. Economic Mechanism

In the face of the challenges of climate change and water scarcity, irrigation facilities as adaptive measures can help maintain the stability and productivity of forest ecosystems. Through reasonable irrigation management, the loss of forest industry caused by drought and other natural disasters can be reduced and the sustainable use of forest resources can be ensured [37]. The number of forestry irrigation facilities is selected to reflect the regulation mechanism of economic resources [38]. It has been established that a moderate irrigation strategy could optimize forest growth [39]. Fertilized and dry-season irrigated plantations promote timber production and carbon storage [40]. Therefore, it is reasonable to assume that drought will increase irrigation facilities in China.

Hypothesis 4.

Drought can offset some of the losses in the forestry industry by increasing the number of forestry irrigation facilities.

2.3.3. Social Mechanism

The drought-induced drop in income will drive some labor force from the forest to the non-forest sector [41], This phenomenon is consistent with the dual economy theory proposed by Lewis (1954) [42]. Labor transfer rate is selected as a mechanism variable to show the influence of drought on social life. Labor force transfer would lead to the reduction in rural forestry production labor force, increase the price of forestry labor force, and thus increase the production cost of forestry [43]. Meanwhile, the migration of the labor force will also lead the left-behind farmers to switch to low-labor-intensity grain cultivation to ensure their basic food and clothing needs, ultimately suppressing the development of the forestry industry [44]. Therefore, droughts could constrain the development of forestry by increasing the transfer rate of the labor force away from forest industry. This leads to our fifth hypothesis.

Hypothesis 5.

Drought will intensify the transfer of labor force and thereby restrict the development of the forestry industry.

3. Materials and Methods

3.1. Variable Selection and Data Collection

This study analyses the influence of drought on the growth of forest industry with a regression model. The growth of the forest industry is described by a composite index of the forest industry as the explained variable, while drought is described by a drought index as an explanatory variable. In order to eliminate other potential interfering factors and ensure that the relationship between the core explanatory variable and the explained variable is not distorted, control variables are selected from three perspectives:, i.e., economic, social and ecological, based on the theory of social–ecological systems [45,46].

We also conducted mechanism tests to evaluate internal mechanisms of drought stress in forest industry in China. Three mechanism variables reflect economic, social, and ecological factors.

The definitions of all variables are shown in

Table 1. Variables definition.

Type	Variable	Symbol	Unit	Meaning	Sources
Explanatory variable	SPEI	SPEI	\	The standardized precipitation–evapotranspiration index	Global SPEI database (https://spei.csic.es/ (accessed on 28 April 2025))
Explained variable	FICI	Score	\	The forest industry comprehensive index	China Forest and Grassland Yearbook
Control variables	Level of economic development	PerGDP	10 k CNY	Per capita GDP	China Statistical Yearbook
	Urban-rural income gap	Incomegap	\	Per capita disposable income of urban residents/per capita disposable income of rural residents	China Population and Employment Statistical Yearbook
	Urbanization rate	Urbanization	\	Urban permanent residents/total permanent residents	The China Statistical Yearbook
	Average years of education in rural areas	Education	yr	The average years of education in rural areas	The China Statistical Yearbook
	Number of employees in the forestry system at the end of the year	Employees	10 k persons	The number of employees in the forestry system at the end of the year	The China Forest and Grassland Yearbook
	Forest ecological environment	Coveragerate	\	Forest coverage rate	The China Forest and Grassland Yearbook
	Forest area	Forest	10 k ha	Forest area of each province	The China Forest and Grassland Yearbook
Mechanism variables	Occurrence area of forest pests and diseases	Acreage	ha	The area where forest pests occur	The China Forest and Grassland Yearbook
	Number of forestry irrigation facilities	Quantity	device	The total number of all available irrigation facilities for forestry	The China Forest and Grassland Yearbook
	Labor transfer rate	Rate	\	The quantity of transferred labor force/the total labor force	The China Statistical Yearbook

.

3.1.1. Explanatory Variable

The explanatory variable of this study is drought data based on the standardized precipitation–evapotranspiration index (SPEI).

Drought is usually divided into four categories: meteorological drought, agricultural drought, hydrological drought, and socioeconomic drought [47]. Meteorological drought is usually defined as the lack of precipitation in an area over a period of time [48]. Agricultural drought refers to a type of drought where the soil moisture supply is insufficient to meet the normal growth needs of crops or vegetation, resulting in reduced crop yields, restricted growth, or damage to the ecosystem [49]. Hydrological drought refers to a type of drought where the reserves of surface water or groundwater remain persistently below normal levels, resulting in insufficient water supply and affecting activities such as human water use, ecological water demand, and hydropower generation [50]. Socioeconomic drought refers to a type of drought where water shortage leads to the inability to meet social production, economic activities, or living demands, thereby causing economic losses, difficulties in people’s livelihood, or social conflicts [47,51]. Among these droughts, meteorological drought is usually regarded as the origin of other droughts [50,52]. In light of the research questions, this study should choose a drought index that is both meteorologically and agriculturally oriented.

There are many drought monitoring indexes; the most commonly used are the Palmer drought index (PDSI), the standard precipitation index (SPI), and the SPEI. Among them, the PDSI lacks the validity of time and space comparison, and has strong subjectivity in drought classification [53]. SPI is based on precipitation only, and does not consider the influence of temperature and evapotranspiration on drought conditions [54]. Vicente-Serrano et al. proposed the SPEI in 2010, which not only retains the advantage of PDSI’s sensitivity to potential evapotranspiration, but also has the characteristics of SPI’s multiple time scales, which is suitable for drought characteristics analysis under the background of global warming [55]. Moreover, the SPEI may reflect both meteorological and agricultural droughts. It is well suited for the research purposes of this study [56].

The SPEI dataset for this study was provided by the global SPEI database (https://spei.csic.es/ accessed on 28 April 2025)) with a spatial resolution of 0.5° × 0.5°. The precipitation and potential evapotranspiration data used to calculate the SPEI data were provided by the Climatic Research Center of the University of East Anglia [57,58].

3.1.2. Explained Variable

The explained variable of this study is the forest industry comprehensive index (FICI), and its construction is based on the existing literature [59,60,61]. The specific contents of each indicator are shown in

Table 2. Construction of an FICI.

First-Level Indicator	Secondary Indicators	Tertiary Indicators
Forest industry composite index	Fixed asset investment (infrastructure investment)	Planned investment (state, self-financing) (CNY ten thousand)
		Completed investment (national, self-funded) (CNY ten thousand)
	Industrial distribution	The proportion of the primary industry
		The proportion of the secondary industry
		The proportion of the tertiary industry
	Economic benefits	Output value per unit of forest area (CNY/hectare)
		Per capita forestry output value (CNY/person)

. After comprehensive consideration of the validity and availability of data from various statistical yearbooks, empirical analysis was conducted on panel data from 31 provinces (excluding Hong Kong, Macao, and Taiwan) from 2001 to 2020. Data were collected from the ERA5-Land dataset, the Guotai’an database, the China Statistical Yearbook, the National Economic Statistical Bulletin, the China Forest and Grassland Yearbook, and the China Population and Employment Statistical Yearbook. In addition, missing data were supplemented by interpolation.

Because FICI, the explained variable, is composed of multiple sub-indicators, in order to provide a more accurate and comprehensive reflection of the overall situation of the variable, an entropy method is used to assign weights to sub-indicators to eliminate the cross-overlap of variables between sub-indicators and the influence of subjective assumptions in sub-indicator design [62].

Firstly, the tertiary indicators such as planned investment, completed investment, proportion of primary industry, proportion of secondary industry, proportion of the tertiary industry, output value per unit forest area, and per capita forestry output value are normalized by min–max scaling.

For positive indicators:

$$X_{hijt}=\left(X_{ijt}-X_{jtmin}\right)∕\left(X_{jtmax}-X_{jtmin}\right)$$

For negative indicators:

$$X_{hijt}=\left(X_{jtmax}-X_{ijt}\right)∕\left(X_{jtmax}-X_{jtmin}\right)$$

where $X_{hijt}$ represents the scaling result of the j indicator of the i province in the t year; $X_{ijt}$ represents the score of the i province in the j indicator in the t year; $X_{jtmin}$ represents the lowest score of indicator j in all provinces; and $X_{jtmax}$ represents the highest score of indicator j in all provinces.

Secondly, proportion $Y_{ijt}$ of the indicator j in the i province is calculated in the t year, as follows:

$$Y_{ijt}=\left(X_{ijt}+1\right)/\sum _{i=1}^n(X_{ijt}+1)$$

Thirdly, the information entropy $E_{jt}$ and information utility $D_{jt}$ of indicator j are calculated as follows:

$$E_{jt}=(-1)/\ln \left(m\right)\ast \sum _{i=1}^n(Y_{ijt}\ast \mathrm{l}\mathrm{n}\left(Y_{ijt}\right))$$

$$D_{jt}=1-E_{jt}$$

Then, the weight $M_{jt}$ of indicator j can be calculated as follows:

$$M_{jt}=D_{jt}/\sum _{i=1}^n{D}_{jt}$$

Finally, the comprehensive score $S_{it}$ of the i province in the t year is calculated as follows:

$$S_{it}=\sum _{j=1}^{\mathrm{m}}(M_{jt}\ast Y_{ijt})$$

3.1.3. Control Variable

Control variables are selected from three perspectives, i.e., economic, social, and ecological, based on the theory of social–ecological systems [45,46]. The control variables included in this study’s list of potentially affecting factors are as follows.

On the economic aspect, the level of economic development and the urban–rural income gap are set as control variables, because regional economic development may affect forestry resource management through capital formation ability, and the income gap between urban and rural areas may affect the decision for rural labor transfer through push–pull effect, resulting in the loss of special human capital from the forest industry [63,64]. Secondly, at the level of social system adjustment, the urbanization rate, the average years of education in rural areas, and the number of employees in the forest industry at the end of a year were set as control variables. Urbanization may adjust the forestry structure through the change in land use and wood demand. The level of rural education may improve knowledge-based ecological management through technology adoption and ecological consciousness. As a direct representation of the industry scale, the number of employees in the forestry system will directly affect the development of the forest industry [65,66,67]. Finally, for ecosystems, the forest ecological environment and forest coverage are set as control variables. Forest ecological environment (such as forest coverage rate) can lower drought stress intensity by regulating microclimate and conserving soil and water, so as to maintain forest productivity. Forest area constitutes the basis of industrial resilience through scale effect and resource reserve [68,69]. Therefore, the control variables of this study are the level of economic development (PerGDP), the urban–rural income gap (Incomegap), the urbanization rate (Urbanization), the average years of education in rural areas (Education), the number of employees in the forestry system at the end of the year (Employees), the forest ecological environment (Coveragerate), and forest area (Forest).

3.1.4. Mechanism Variable

Mechanism variables are also selected from economic, social and ecological perspectives [60,70]. The occurrence area of forest pests and diseases (Acreage) is selected to reflect the direct destruction mechanism of biological stress. The number of forestry irrigation facilities (Quantity) is selected to reflect the regulation mechanism of economic resource [38]. The labor transfer rate (Rate) is selected as a mechanism variable to show the influence of drought on social life. These three variables construct a mechanism chain, which conforms to the multi-scale analysis framework of complex systems theory [7,71].

3.2. Model

Before the empirical analysis, in order to ensure the rationality of model selection, we carried out a series of relevant tests. First of all, through the multicollinearity test, since all variables and the average VIF value are less than 10, it can be considered that there is no multicollinearity problem, as shown in

Table 3. Multicollinearity test.

	VIF	1/VIF
Urbanization	4.237	0.236
PerGDP	3.478	0.288
Education	2.476	0.404
Incomegap	2.162	0.462
Forest	2.036	0.491
Employees	1.843	0.542
Coveragerate	1.304	0.767
SPEI	1.107	0.903
Mean VIF	2.331	.

. Secondly, through the Hausmann test, it can be seen from the results that for the explained variable forest industry composite index (score), the Chi-square test value is 75.05 and the p-value is 0, and the null hypothesis is rejected at the 1% level, as shown in

Table 4. Hausmann test.

	Coef.
Chi-square test value	75.05
p-value	0

. This result supports our decision to choose the two-way fixed effect model.

The following model was constructed to clarify the impact of drought on the forest industry and its mechanism:

$${Sco}_{it}=\alpha +{\beta Index}_{i\mathrm{t}}+\gamma X_{it}+{\theta }_i+{\pi }_t+{\epsilon }_{it}$$

where ${Sco}_{it}$ is the explained variable, representing the FICI of province $i$ in period $t$; ${Index}_{i,t}$ is the explanatory variable, representing the SPEI value of province $i$ in period $t$; $X_{it}$ is the control variable of province $i$ in period $t$; ${\theta }_i$ is the fixed effect at the provincial level; ${\pi }_t$ is the time fixed effect; and ${\epsilon }_{it}$ is the random disturbance term.

Secondly, this study also reveals the intermediate mechanism of drought on the forest industry from the perspective of eco-economic coupling [60]. The model is set as follows:

$${Mechanism}_{it}={\gamma }_0+{\gamma }_1{Index}_{it}+{\theta }_i{Control}_{it}+{\mu }_i+v_t+{\epsilon }_{it}$$

where ${Mechanism}_{it}$ is the explained variable, that is, the transmission channel to be verified in this study, and represents the mechanism variable of province $i$ during period $t$. ${Index}_{it}$ is the explanatory variable, representing the SPEI value of province $i$ in period $t$. ${Control}_{it}$ is the control variable of province $i$ in period $t$. ${\mu }_i$ is the fixed effect at the provincial level. $v_t$ is the time fixed effect, and ${\epsilon }_{it}$ is the random disturbance term.

3.3. Endogenous Processing, Robustness Test, and Heterogeneity Analysis

3.3.1. Endogenous Processing

This study contains three types of endogenous problems—reverse causality, measurement error, and missing variables. In order to overcome endogeneity, precipitation and the normalized vegetation index were selected as instrumental variables [72,73,74] for a two-stage instrumental variable regression.

3.3.2. Robustness Test

A robustness test was conducted to further verify the robustness of the regression results, prevent pseudo-regression problems, and increase the authenticity of the regression results. On the basis of the previous regression, we replaced the explanatory variables, replaced the models, and eliminated the impact of the financial crisis and the COVID-19 pandemic to cross-verify the regression results. First, the explanatory variables were replaced by delaying them for one period. Second, the system GMM model was used for regression [75]. Third, the financial crisis and the COVID-19 pandemic seriously affected local economies and the macro social environment, resulting in sample data of the current year that did not truly reflect the actual situation; that is, the data were distorted due to changes in the external environment. Therefore, a robustness test was conducted by eliminating the samples from relevant years. In addition, the regression of data samples was also conducted by removing outliers and 1% tail shrinkage.

3.3.3. Heterogeneity Analysis

Considering the differences in natural endowment, industrial structure, and policy support in different regions, it is particularly important to explore the differences in SPEI’s impact on forest industry in different regions. The samples were divided into two sub-samples by region: coastal area and inland area.

4. Results

4.1. Descriptive Statistics

Table 5. Descriptive statistics of variables.

	Unit	N	Mean	SD	Min	Median	Max
Score	\	620	0.187	0.134	0.045	0.157	0.812
SPEI	\	620	−0.266	0.676	−1.725	−0.288	1.483
Forest	10 k ha	620	734.849	622.931	5.97	578.82	2614.85
Employees	10 k persons	620	3.848	5.776	0.002	2.517	40.456
Coveragerate	\	620	32.148	17.948	4	34.3	66.8
Urbanization	\	620	0.518	0.157	0.196	0.51	0.896
Incomegap	\	620	2.797	0.543	1.845	2.679	5.646
Education	yr	620	7.304	0.901	3.238	7.489	9.66
PerGDP	10 k CNY	620	3.571	2.761	0.3	3.055	16.416

reports descriptive statistical results for the main variables selected. These data provide a wealth of information. First, in the forest industry, the data show significant differences in the value of the industry between provinces. Specifically, the minimum value of the forest industry composite index is 0.045, the maximum value is 0.812, and the standard deviation is 0.134, which indicates that there are obvious individual differences among provinces in the forest industry. Secondly, from the perspective of drought, the minimum value of drought is −1.725, the maximum is 1.483, and the standard deviation is 0.676. These data reflect some differences in the degree of drought among provinces.

4.2. Analysis of SPEI at Different Time Scales

As depicted in Figure 1, notable differences exist in the variation trends of the SPEI time series across diverse time scales. The monthly SPEI (Figure 1a) has the most frequent fluctuation, with obvious alternation between dry and wet. The seasonal SPEI in May, August, November, and February (Figure 1b) was followed by continuous dry periods and wet periods, while the annual SPEI (Figure 1c) had the least fluctuating frequency, showing an obvious switch between dry and wet periods. Therefore, when the time scale is extended, the SPEI changes gradually trend to flatline, reflecting that water differences over multiple months are more average than the water difference in a single month, and overall periodicity is more obvious. SPEI values for different time scales have different sensitivities to precipitation and potential evapotranspiration. The larger the time scale, the less obvious the response of SPEI to short-term precipitation and potential evapotranspiration, and only long-term precipitation and potential evapotranspiration can cause the SPEI to fluctuate, which is more suitable for detecting long-term water status [76,77]. On the whole, SPEI linear tendency rates at monthly, seasonal, and annual scale were −0.004·10-a, −0.006·10-a and −0.012·10-a, respectively, showing a decreasing trend. Clearly, drought in China has increased significantly in the past 50 years. Hence, Hypothesis 1 is supported.

4.3. Results of Baseline Regression Analyses

Referring to accepted practice, the benchmark regression gradually adds control variables at both the forestry level and the provincial economic level for regression [78,79].

Table 6. Regression results of SPEI on the forest industry composite index.

	(1)	(2)	(3)
	Score	Score	Score
SPEI	0.006	0.007	0.010 *
	(0.006)	(0.006)	(0.006)
Forest		0.001	0.001
		(0.001)	(0.001)
Employees		0.004 **	0.004 **
		(0.002)	(0.002)
Coveragerate		0.006 **	0.010 ***
		(0.003)	(0.003)
Urbanization			−0.035
			(0.148)
Incomegap			−0.058 ***
			(0.018)
Education			−0.015
			(0.017)
PerGDP			−0.010 *
			(0.005)
_cons	0.251 ***	0.052	0.351 **
	(0.014)	(0.085)	(0.142)
Province fixed effect	YES	YES	YES
Year fixed effect	YES	YES	YES
Observed values	620	620	620
R square	0.471	0.485	0.505

Note: *, **, *** stands for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

details the regression results of the bidirectional fixed-effect model of SPEI for the composite index of the forest industry, which includes provincial and year fixed effects. In Column (1), we only controlled provincial fixed effects and year fixed effects. The regression results show that the SPEI had a positive impact on the resilience of the forest industry. In Columns (2) and (3), on the basis of controlling the fixed effects of provinces and years, we gradually introduced control variables, including forest area, number of employees in the forestry system at the end of the year, forest ecological environment, urbanization rate, urban–rural income gap, average years of education in rural areas, and economic development level in each province. After gradually adding these control variables, the SPEI had an increasingly significant positive promoting effect on the comprehensive index of the forest industry. In Column (3), in particular, when all control variables were controlled, the SPEI increased by 1%, and the forest industry composite index increased by 0.01 units. The results showed that SPEI had a positive effect on the comprehensive index of forest industry, that is, drought would significantly inhibit the development of forest industry. Therefore, Hypothesis 2 is supported.

Observe the other control variables. First of all, looking at the situation at the economic level, the improvement of economic development level has a slight inhibitory effect on the forest industry, which may be due to the phenomenon of resources being tilted to other industries in the process of economic development. The income gap between urban and rural areas is significant at the 1% level, indicating that the widening of the income gap between urban and rural areas significantly inhibits the forest industry. Secondly, looking at the situation at the social level, the inhibitory effect of urbanization rate on the forest industry has not passed the significance test. The effect of rural education level on the forest industry is not significant; however, it is worth noting that the increase in the number of employees in the forestry system has a significant positive driving effect on the development of the forest industry. Finally, at the ecological level, the forest area has no significant effect on the forest industry, but the improvement of forest ecological environment has a strong and significant effect on the development of the forest industry. In conclusion, drought has an inhibitory effect on the forest industry. The increase in the number of employees in the forestry system and the improvement in the ecological environment in the forest region can effectively promote the development of the forest industry, while the widening of the urban–rural income gap and the improvement of the economic development level have an inhibitory effect on the forest industry. For specific inhibitory regions, see the heterogeneity test section below.

4.4. Results of Endogenetic Processing and Robustness Test

4.4.1. Results of Endogenous Processing

The endogenous processing results obtained from the two-stage instrumental variable approach are presented in

Table 7. Regression results of instrumental variable I (precipitation).

	(1)	(2)
	SPEI	Score
Precipitation	634.676 ***
	(52.138)
Forest	0.001	0.001
	(0.001)	(0.001)
Employees	−0.029 **	0.004 **
	(0.011)	(0.002)
Coveragerate	0.021	0.010 ***
	(0.018)	(0.003)
Urbanization	0.167	−0.035
	(0.969)	(0.148)
Incomegap	0.219 *	−0.059 ***
	(0.118)	(0.018)
Education	0.044	−0.015
	(0.109)	(0.017)
PerGDP	0.017	−0.010 *
	(0.034)	(0.005)
SPEI		0.011
		(0.013)
_cons	−3.510 ***	0.353 **
	(0.945)	(0.143)
Province fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	620	620
R square	0.364	0.5046
One stage F value	11.903

Note: *, **, *** stands for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

and

Table 8. Regression results of instrumental variable II (NDVI).

	(1)	(2)
	SPEI	Score
NDVI	11.323 ***
	(1.537)
Forest	0.001	0.001
	(0.001)	(0.001)
Employees	−0.036 ***	0.003
	(0.012)	(0.002)
Coveragerate	0.003	0.011 ***
	(0.020)	(0.003)
Urbanization	−1.391	−0.062
	(1.044)	(0.156)
Incomegap	0.129	−0.047 **
	(0.128)	(0.020)
Education	0.002	−0.013
	(0.117)	(0.018)
PerGDP	0.097 **	−0.009
	(0.038)	(0.006)
SPEI		0.036 *
		(0.020)
_cons	−7.212 ***	0.289 *
	(1.276)	(0.152)
Province fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	620	620
R square	0.267	0.4485
One stage F value	20.49

Note: *, **, *** stand for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

.

In the first-stage regression of Instrumental Variable I (precipitation), the results demonstrate a significant positive correlation between precipitation and the standardized precipitation–evapotranspiration index (SPEI). Specifically, an increase in precipitation corresponds to a higher SPEI score, which aligns with real-world observations. The second-stage regression results are congruent with the aforementioned baseline regression findings, thereby validating the robustness of the model. Regarding Instrumental Variable II (the normalized difference vegetation index; NDVI), the first-stage regression reveals a significant positive relationship between the normalized vegetation index and the SPEI, in accordance with practical observations. The second-stage regression results further corroborate the robustness of the baseline regression results, highlighting the stability and reliability of the overall analytical framework.

4.4.2. Results of Robustness Test

The robustness test results are shown in

Table 9. Robustness Test I (replacement of explanatory variables).

	(1)	(2)
	Score	Score
L.SPEI	0.006	0.008
	(0.006)	(0.005)
Forest		0.001 *
		(0.000)
Employees		0.003 *
		(0.002)
Coveragerate		0.010 ***
		(0.003)
Urbanization		−0.107
		(0.147)
Incomegap		−0.076 ***
		(0.019)
Education		−0.015
		(0.016)
PerGDP		−0.013 ***
		(0.005)
_cons	0.268 ***	0.473 ***
	(0.014)	(0.141)
Province fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	589	589
R square	0.509	0.549

Note: *, *** stand for significant at 10% and 1% levels, respectively, with robust standard error in brackets.

,

Table 10. Robustness Test II (regression with system GMM model).

	(1)	(2)
	GMM Model	GMM Model
	Score	Score
SPEI	0.022	0.032
	(0.021)	(0.022)
Forest		0.002 **
		(0.001)
Employees		0.004 ***
		(0.001)
Coveragerate		0.010 ***
		(0.003)
Urbanization		−0.093
		(0.155)
Incomegap		−0.080 ***
		(0.014)
Education		−0.017
		(0.016)
PerGDP		−0.014 **
		(0.006)
_cons	0.191 ***	0.445 *
	(0.022)	(0.231)
Province fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	589	589
R square	0.714	0.734

Note: *, **, *** stand for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

,

Table 11. Robustness Test III (excluding effects of the financial crisis and the COVID–19 pandemic).

	(1)	(2)	(3)	(4)
	Excluding Financial Crisis	Take Out the COVID-19 Pandemic	Strip Out the Financial Crisis	Take Out the COVID-19 Pandemic
	Score	Score	Score	Score
SPEI	0.007	0.009	0.011 *	0.012 **
	(0.006)	(0.006)	(0.006)	(0.006)
Forest			0.001	0.001
			(0.001)	(0.001)
Employees			0.005 **	0.005 **
			(0.002)	(0.002)
Coveragerate			0.010 ***	0.009 ***
			(0.003)	(0.003)
Urbanization			−0.042	−0.030
			(0.153)	(0.158)
Incomegap			−0.059 ***	−0.055 ***
			(0.019)	(0.019)
Education			−0.017	−0.011
			(0.017)	(0.017)
PerGDP			−0.010 *	−0.012 **
			(0.005)	(0.006)
_cons	0.251 ***	0.252 ***	0.360 **	0.320 **
	(0.015)	(0.015)	(0.146)	(0.145)
Province fixed effect	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES
Observed value	589	589	589	589
R square	0.469	0.479	0.504	0.511

Note: *, **, *** stand for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

and

Table 12. Robustness Test IV (excluding outliers and 1% tailing treatment before and after).

	(1)	(2)
	Score	Score
SPEI	0.006	0.010 *
	(0.006)	(0.006)
Forest		0.001 *
		(0.000)
Employees		0.005 ***
		(0.002)
Coveragerate		0.010 ***
		(0.003)
Urbanization		−0.043
		(0.138)
Incomegap		−0.067 ***
		(0.023)
Education		−0.018
		(0.017)
PerGDP		−0.011 **
		(0.005)
_cons	0.251 ***	0.414 ***
	(0.014)	(0.151)
Province fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	620	620
R square	0.476	0.509

Note: *, **, *** stand for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

.

In Table 9, the comprehensive index of the forest industry is delayed by one stage for the robustness test. In Column (1), with only the province fixed effect and year fixed effect controlled, the regression results show that the SPEI has a positive impact on the forest industry. In Column (2), after controlling for province fixed effects and year fixed effects variables, and controlling for all control variables, the results still showed that the SPEI had a positive influence on the forest industry, with the effect becoming increasingly pronounced. Table 10 uses the GMM model for regression. With or without control variables, the results show the positive impact of SPEI on the forest industry. After excluding the effects of the financial crisis and COVID-19 pandemic, Table 11 Columns (1) and (2) show that SPEI has a positive effect on the forest industry when only the fixed effects of provinces and years are controlled. The regression results of Table 11 Columns (3) and (4) show that, on the basis of the fixed effects of controlled provinces and years, when all control variables were controlled, the SPEI showed a more significant positive effect on the forest industry. The regression results in Column (4), in particular, were significant at the 5% level after excluding the COVID-19 pandemic. Table 12 reports the regression results after 1% tail reduction treatment of the samples and further removing the sample outliers. The findings indicate that in Column (1), where no control variables are incorporated, the SPEI exerts a positive influence on the forest industry. Subsequently, when all control variables are introduced in Column (2), the coefficient gradually increases to 0.010. This suggests that, at the 10% significance level, the SPEI demonstrates a significant positive association with the forest industry. These outcomes are consistent with the baseline regression results, thereby validating the robustness of the baseline regression findings.

4.5. Results of Heterogeneity Analysis

The results of the heterogeneity analysis are shown in

Table 13. Heterogeneity analysis.

	(1)	(2)
	Coastal Areas	Inland Areas
	Score	Score
SPEI	0.004	0.016 **
	(0.007)	(0.008)
Forest	0.001	0.001
	(0.001)	(0.001)
Employees	−0.008	0.005 **
	(0.008)	(0.002)
Coveragerate	0.004	0.013 ***
	(0.004)	(0.004)
Urbanization	−0.061	0.110
	(0.172)	(0.287)
Incomegap	−0.026	−0.062 ***
	(0.053)	(0.023)
Education	0.023	−0.043 *
	(0.021)	(0.023)
PerGDP	−0.002	−0.046 ***
	(0.005)	(0.012)
_cons	0.062	0.508 ***
	(0.218)	(0.187)
Provincial fixed effect	YES	YES
Year fixed effect	YES	YES
Observed values	220	400
R square	0.753	0.463

Note: *, **, *** stand for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

.

When all control variables are added, it is concluded that in inland areas, SPEI is significantly positive at the 5% level, but in coastal areas, SPEI does not show a significant impact on its forest industry. It indicates that from coastal to inland, the promoting effect of SPEI on the forest industry gradually increases.

In addition, the increase in the number of employees in the inland forestry system at the end of the year and the improvement of the forest ecology environment have a significant role in promoting the forest industry. The widening of the income gap between urban and rural areas, the average years of education in rural areas, and the level of economic development have a significant inhibitory effect on the inland forest industry. However, the effect of these control variables in coastal areas is not obvious, and even the increase in average years in education in rural areas will promote the development of forest industry in coastal areas. This may be because the subsidy measures in inland areas are not enough, resulting in the liberation of specialized talents, which in turn inhibits the development of the forest industry.

4.6. Mechanism Testing

This study tested three mechanism variables, namely the occurrence area of forest pests and diseases, number of forestry irrigation facilities, and labor transfer rate, to explain the intermediate mechanism of drought on forest industry.

As shown in

Table 14. Mechanism inspection.

	(1)	(2)	(3)	(4)
	Score	Acreage	Quantity	Rate
SPEI	0.010 *	−0.145 **	−0.068 **	−0.039 **
	(0.006)	(0.06)	(0.034)	(0.016)
Forest	0.001	−0.001 ***	−0.001	0.001
	(0.001)	(0.000)	(0.001)	(0.001)
Employees	0.004 **	0.004	0.001	0.001
	(0.002)	(0.005)	(0.009)	(0.005)
Coveragerate	0.010 ***	−0.011 ***	−0.007	0.005
	(0.003)	(0.002)	(0.017)	(0.007)
Urbanization	−0.035	−3.777 ***	−0.225	0.199
	(0.148)	(0.436)	(0.601)	(0.400)
Incomegap	−0.058 ***	−0.197 ***	0.062	−0.055
	(0.018)	(0.073)	(0.096)	(0.049)
Education	−0.015	0.431 ***	0.031	−0.024
	(0.017)	(0.054)	(0.084)	(0.045)
PerGDP	−0.010 *	−0.142 ***	0.003	−0.009
	(0.005)	(0.028)	(0.021)	(0.014)
_cons	0.351 **	11.216 ***	0.132	0.890 **
	(0.142)	(0.506)	(0.743)	(0.384)
Province fixed effect	YES	YES	YES	YES
Year fixed effect	YES	YES	YES	YES
Observed values	620	620	620	620
R square	0.505	0.541	0.512	0.568

Note: *, **, *** stands for significant at 10%, 5%, and 1% levels, respectively, with robust standard error in brackets.

, the coefficient of SPEI in Column (2) is significantly negative at the 5% level, indicating that drought will significantly increase the occurrence area of forest diseases and insect pests. Hence, Hypothesis 3 is supported.

In Column (3), the coefficient of SPEI is significantly negative at the 5% level, implying that drought stress will lead to a significant increase in the number of forestry irrigation facilities. So, Hypothesis 4 is supported.

In Column (4), the coefficient of SPEI is significantly negative at the 5% level, showing that the risk of labor transfer will be intensified after drought. We can see that Hypothesis 5 is supported.

5. Discussion

Our study reveals the influence process of drought on forest industry. Five hypotheses are all held to be true in this study.

It is evident from SPEI data that drought in China has increased significantly in the past 50 years, and that drought has a significant restraining effect on Chinese forest industry. Moreover, inland areas are more vulnerable to drought than coastal areas. This difference between coastal and inland areas may be closely related to the economic imbalance between the two regions [80,81].

Drought will lead to an increase in the occurrence area of forest pests and diseases, forming a compound eco-economic loss effect, thus inhibiting the development of forest industry. Kotlarz and Bejger mentioned increasing climate change effects, like pets and diseases, may reduce the timber supply in the forest industry, thus cause significant economic impacts [34]. The outbreak of insects would lead to a large number of tree deaths, and have a negative impact on the forest industry [35]. Drought stress can reduce tree resistance and make forest ecosystems more vulnerable to pests and diseases, negatively affecting the forest industry [36]. Therefore, drought will inhibit the development of the forest industry by increasing pests and diseases.

Drought can drive the increase in irrigation facilities. Efficient irrigation may mitigate biological damage caused by drought, so as to resist part of the economic losses. It has been established that a moderate irrigation strategy could optimize forest growth [39]. Fertilized and dry-season irrigated plantations promote timber production and carbon storage [40]. Therefore, drought will reduce the loss of the forest industry by demanding more irrigation facilities.

Drought may cause labor to migrate from forest sector to non-forest sectors, leading to lower production in the forest industry. Labor transfer effects caused by drought are more obvious in the areas where the income gap is relatively big between rural and urban regions, e.g., inland areas. Labor force transfer would lead to the reduction in the rural forestry production labor force, increases in the price of forestry labor force, and thus increases in the production cost of forestry [43]. Meanwhile, the migration of the labor force will also lead the left-behind farmers to switch to low-labor-intensity grain cultivation to ensure their basic food and clothing needs, ultimately suppressing the development of the forestry industry [44]. Therefore, droughts could constrain the development of forestry by increasing the transfer rate of the labor force away from forest industry.

6. Conclusions

Our study reveals the internal mechanisms of drought event transmission to the forest industry through economic, social, and ecological dimensions. It adds to the forest resilience literature and expands the research dimension of climate change economics, providing theoretical support and practical paths for improving forest resilience and the sustainable management of forest resources.

In the long term, drought may alter the distribution of China’s forest resources. The forest industry will need to adapt to these changes by adjusting its operations and management practices. The findings of this study provide two important lessons for policy-making related to resilience reconstruction of forest industry under climate risk. First, as drought tends to occur more frequently, policy-makers should make further efforts to strengthen irrigation facilities and water-saving technology. Second, since drought intensifies labor force transfer, there is a need to develop tailored social protection schemes that retains forest workers in the forest sector during and after drought.

This study has potential limitations. Firstly, the economic and social data are collected at the provincial level. The findings are relevant at provincial level, rather than at a smaller scale. It has been found in this study that the impacts of drought vary between different provinces. Further studies on the effects of drought within one or more nearby provinces may investigate the mechanism underlying the effect of drought on the forest industry with more relevant variables. Secondly, this study focuses on only one type of extreme weather events, which might affect forest resilience. Future research may explore, e.g., how other extreme weather events could affect forest industry, and how two or more extreme weather events together affect forest resilience.