Effects of Rising Rural Labor Prices on Land Use Pattern: Evidence from Grain Production in China

Abstract

As rural labor prices have risen constantly over the last two decades, Chinese grain production that relies heavily on manual labor has been subjected to considerable challenges and has experienced profound changes in land use patterns. Using a fixed effect model and translog profit function model, this paper investigates the effects of rising rural labor prices on land use patterns in Chinese grain production. The empirical results from 2004–2022 province-level panel data showed that the rising rural labor prices provided significant incentives to adjust the land use patterns of three staple grain crops. The increase in labor prices had a negative effect on the share of the planting area of rice and maize, while wheat experienced a substantial increase in its proportion of planting area in the context of rising labor prices. A further mechanism test based on 2004–2012 farm-level panel data revealed that the factor substitutions, especially labor substitution with fertilizer and machinery, were a significant contributor to the changes in land use patterns. In the spatial–temporal analysis, changes in land use patterns were found to be more pronounced in regions with more rolling terrain conditions but remained relatively stable across years. These findings highlight the importance of the development and promotion of labor-saving technologies in grain production, especially enhanced-efficiency fertilizer and small-sized agricultural machinery. In addition, agricultural subsidies targeted at farmers in hilly and mountainous regions might be a good way to mitigate potential land abandonment in the context of rising labor prices.

1. Introduction

Due to the relative scarcity of farmland and increasing food demand caused by the large population and changing consumption structure, stable grain production growth through improvements in farmland allocation efficiency is of fundamental importance for ensuring Chinese food security. Since the comprehensive rural reform initiated in 1978, Chinese agriculture has been dominated by smallholder farms [1,2,3,4]. Despite promoting the transfer of farmland to more productive farmers through the establishment of a farmland rental market, farms in China are still much smaller than those in developed countries. According to the Third Agricultural Census of China in 2016, there are 230 million farms with 140 million hectares of arable land, and the average farm size is only about 0.6 hectares. In the context of limited farmland resources, China has still attained substantial grain harvests over the past two decades. According to the National Bureau of Statistics of China, the grain yield per mu increased by approximately 22% during the 2004–2022 period, which is partly attributed to the intensive use of factors such as fertilizer and labor in grain production [5]. However, as the Chinese agricultural labor force has migrated massively to urban and non-agricultural sectors due to rapid urbanization and industrialization in recent decades [6], the opportunity cost of agricultural labor has also constantly risen, raising the costs of grain production and discouraging farmers’ enthusiasm for grain planting. Although increases in agricultural labor prices have led farmers to adjust factor use in agricultural production, especially to increase the use of fertilizer and agricultural machinery in order to substitute for labor [7], labor costs remain the largest part of production costs, accounting for more than 30 percent of the total costs in Chinese rice, wheat, and maize production, according to the National Bureau of Statistics of China. As labor costs escalate, small-scale farming that relies heavily on manual labor is subjected to considerable challenges [8,9], calling into question the sustainability of this intensive labor-dependent production approach. Against this background, the changes that will occur in the land use pattern of Chinese grain production are worth studying.

Over the past two decades, China has witnessed rapid developments in urbanization and industrialization, generating substantial demand for labor forces. Since off-farm activities provide higher wages and more regular working hours than agricultural production, rural labor has changed from farming to off-farm employment. According to the Report of Survey of Rural Migrant Workers issued by the National Bureau of Statistics of China, the number of rural migrant workers in China has constantly risen from 225 million in 2008 to 295 million in 2022. Despite the substantial number of rural migrant workers nowadays in China, its annual growth rate has shown a declining trend. The annual growth rate of the number of rural migrant workers was approximately 5.4% in 2010, while in 2022, it was less than 1%. This suggests that as the rural labor force migrates massively to urban and non-agricultural sectors, Chinese rural labor is becoming increasingly scarce, leading to significant increases in rural labor prices. As shown in the Figure 1, the rural labor prices for all three crops exhibit an upward trend. Specifically, rural labor prices increased by around 5 times from 2004 to 2022, from approximately CNY 20 per day to about CNY 100 per day. A more labor-intensive farming approach is indicated by the relatively higher labor price in rice production.

After the comprehensive rural reforms initiated in 1978, Chinese agriculture has not only been dominated by smallholder farms, but also transformed from planned to market-oriented agriculture. This market mechanism provides farmers with a free production decision for agricultural production in the given conditions. The market-oriented agricultural reforms have formed a relatively complete market mechanism in Chinese agriculture in recent decades, and agricultural price has become a primary factor in farmers’ production decisions, including factor inputs and the planting area. The adjustment of the crop planting area induced by changes in agricultural prices can also be referred to as land use pattern change. As shown in the figure, there were significant changes in the land use pattern in Chinese grain production between 2004 and 2022. The proportion of planting area for rice and wheat decreased by 7% and 5%, respectively, over the period from 2004 to 2022. In comparison, the proportion of planting area for maize increased annually, from 0.34 in 2004 to 0.45 in 2022. It is worth noting that while labor prices increased substantially during this period, agricultural product prices, farmland prices, and the prices of fertilizer and machinery also increased to varying degrees. Therefore, it remains unclear to what extent labor prices drove these changes in land use patterns.

Previous literature has discussed the impact of rural labor shortages or labor migration on agricultural production. In terms of the impact of rising agricultural labor prices and the resulting increase in production costs on China’s crop production and food security, most studies have concluded that farmers have two main responses: product substitution with changes in the type of production (or abandonment of farms in special cases), and factor substitution without changes in the type of production [10]. Farmers will make land use decisions based on the opportunity cost of farming, and as the opportunity cost of farming rises, farmers will choose to operate roughly or even let their land fallow, thus increasing the amount of marginalized land [11]. On the one hand, higher labor prices dampen the share of grain crops grown and boost the share of cash crops grown, with the biggest boost to the share of vegetables grown [12,13]. But the negative impact of higher labor prices on grain production will be mitigated to a certain extent by the relative ease of mechanization in the plains [14]. On the other hand, there is also literature that has come to the opposite conclusion, suggesting that rising labor costs increase the production of the three major staples but decrease the production of other crops [10,15]. A few studies have suggested that the rise in non-farm income can prompt farmers to rationally allocate production factors and improve production management techniques, thereby improving production efficiency. In particular, most of them believe that rural labor migration can promote agricultural mechanization [15,16,17,18,19,20]. And the substitution of labor by machinery exists not only between grain crops and cash crops, but also among different cash crops [15].

Prior studies have mainly treated grain crops as a single class and combined them with major cash crops to investigate the land use pattern change induced by rising labor prices. Few studies have distinguished between grain crops and focused on changes in their land use patterns. Second, most of the existing studies have employed long-term province-level panel data to investigate the changes in land use patterns. Few studies have identified the mechanisms of these changes from the micro-perspective of a farm household’s decision-making, namely the factor substitution induced by changes in labor prices.

This paper aims to analyze the effect of rising rural labor prices on land use patterns in Chinese grain production. A further mechanism test is established to identify the reasons behind the land use pattern changes. The empirical results show that the increase in labor prices significantly decreases the share of the planting area of rice and maize, while wheat experiences a pronounced expansion. Labor substitution with fertilizer and machinery was found to be a significant reason for the land use pattern changes. The more elastic the factor substitution is for a crop, the greater the cost advantage it achieves with the rapid increase in rural labor prices. Furthermore, changes in land use patterns are found to be more substantial in regions with poor terrain conditions but remain relatively stable across years.

The main contributions of this paper are as follows. First, it uses province-level data to investigate the macro-trends in land use patterns in the context of rising labor prices and combines them with farm-level data to identify the underlying micro-mechanisms, making the results more comprehensive and effective. Second, it focuses on the production of rice, wheat, and maize, providing a more disaggregated perspective on the effect of rising labor prices on land use patterns in Chinese agricultural production. Third, it identifies the role of factor substitution in the land use pattern, enriching our understanding of input portfolio adjustment in the context of rising labor prices, which provides a basis for the formulation of policies aimed at sustainable agricultural development.

The rest of the paper is organized as follows. Section 2 presents the conceptual framework and outlines the methodology used to examine the effects of rising labor prices on land use patterns. The data sources and descriptive statistics are also provided in this section. Section 3 reports the empirical results. Conclusions and policy implications are presented in Section 4.

2. Materials and Methods

2.1. Conceptual Framework

Farm households are basic decision-making units in grain production. In order to investigate changes in agricultural land use patterns, it is necessary to understand the production decisions made by farm households. As Chinese agriculture transformed from a planned to market-oriented economy after the comprehensive rural reforms initiated in 1978, agricultural market prices have become the main factors influencing grain production, and profit maximization has been viewed as the primary pursuit of rational farmers. Since grain production in China is still in the labor-intensive phase, rising labor prices are expected to reduce farm profits and induce changes in the relative prices of factors. According to the induced technical change theory [21,22], when the relative prices of factors change, farmers tend to adjust their input portfolios towards a scarce-factor-saving direction, namely, substituting scarce and expensive factors for cheap and abundant factors. In farming practice, fertilizer and machinery are commonly found as net substitutes of labor [23,24,25,26]. Nevertheless, there are significant differences in labor substitution with fertilizer and machinery among different crops, which suggests the varied capacity of farmers to mitigate rising labor costs by adjusting their input portfolio. For crops where labor can be easily substituted, farmers are likely to reduce labor inputs at the intensive margin—replacing labor with fertilizers and machinery—while keeping the planting area relatively stable. In contrast, for crops with less elastic labor substitution, farmers may need to reduce labor inputs at the extensive margin, significantly cutting back on the crop planting area while maintaining a relatively constant amount of labor per unit of land. The employment of these two strategies for coping with rising labor costs may give rise to discrepancies in changes in the crop planting area, which can be viewed as land use pattern changes.

The substitution relationships between factors are also subjected to the application conditions of factors [27]. In Chinese grain production, there are hardly any barriers to fertilizer use, which ultimately leads to severe agricultural non-point source pollution [28,29]. However, due to the pronounced regional variations in topography across the country, the utilization of machinery in agriculture exhibits substantial differences across regions [27]. Compared with plain areas, hilly and mountainous areas pose significant obstacles to mechanical operations. This suggests that as the terrain becomes more rolling, the changes in land use patterns induced by rising labor prices may be more substantial, leading to the heterogeneous effects of labor prices across regions. Furthermore, the factor substitution relationship may also be affected by technical advancements [30] and change over time, which, in turn, results in disparities in the price effects in the long run. The spatial–temporal characteristics of the price effects are worth analyzing. Figure 2 shows the analysis framework of the study. Whether and to what extent the labor prices affect land use patterns remains an empirical question to be investigated using econometric methods.

2.2. Methodology

First, this paper establishes the following regression model to investigate the effect of rising labor prices on the land use pattern in grain production at the province level:

$$Y_{it}={\beta }_0+{\beta }_1{Laborprice}_{it}+X_{it}\beta +{\partial }_i+{\lambda }_t+{ϵ}_{it}$$

(1)

where i and t represent province and time, respectively. $Y_{it}$ denotes the crop-specific proportion of planting area to that of all three crops. ${Laborprice}_{it}$ refers to the rural labor price for grain production. $X_{it}$ is a series of control variables. Following a previous study [15], the control variables consist mainly of three parts, namely market price signals, provincial production and demand conditions, and agricultural policy support. Among them, the market price signals include product price, farmland price, fertilizer price, machinery price and the price ratio of all three crops. The provincial production and demand conditions consist of the cultivated farmland per capita, the rural labor quality, and the urbanization level. Agricultural policy support is measured by the annual provincial fiscal expenditure for agriculture. ${\partial }_i$ and ${\lambda }_t$ are the fixed effects for province and year, respectively. ${\beta }_1$ is the coefficient of interest.

Second, this paper investigates the channels through which rising labor prices in agriculture affect the land use pattern in grain production. From the farm-level perspective, rising labor prices are expected to reduce farm profits and thus lead to the adjustment of the input portfolio towards a labor-saving direction. This indicates that factor substitution, which determines the capacity of farmers to mitigate the rising labor costs by adjusting factor inputs, might be a mechanism involved in changes in land use patterns. To comprehensively examine the effect of rising labor prices on the production decisions made by farmers who pursue profit maximization, this paper establishes a profit function model. In general, the profit function can be written as follows:

$$Max \pi \left(p,\mathit{w},\mathit{z}\right)=pQ\left(p,\mathit{w},\mathit{z}\right)-{\mathit{w}}^{\mathit{\prime }}\mathit{x}\left(p,\mathit{w},\mathit{z}\right)$$

(2)

where $\pi$ denotes the profits from production, $p$ is the output price, $\mathit{w}$ is a vector of factor prices, $\mathit{x}$ is a vector of variable inputs, $Q$ refers to the yield, and $\mathit{z}$ is a set of fixed inputs. To derive the supply and input demand functions, the normalized Cobb–Douglas, generalized Leontief, and normalized translog functions are some of the approaches used to analyze the profit function. This paper applies the translog profit function to describe the production decisions, not only due to its flexibility and freedom from some restrictive assumptions, such as the unitary elasticity of substitution and constant return to scale required by the Cobb–Douglas approach, but also because it can estimate the output supply and input demand simultaneously. It is noteworthy that the translog profit function model operates under the assumption of constant substitution elasticities between input factors over time, which means that it circumvents the potential effects of technological advancements and policy changes on factor inputs and enables the study to better focus on the factor substitution induced by changes in labor prices [31,32,33,34]. The resultant profit function is specified as follows:

$$\begin{gathered} ln\pi ={\alpha }_0+{\alpha }_{p}lnp+\sum _{i=1}^n{\alpha }_{w_i}ln{w}_i+\sum _{j=1}^m{\alpha }_{z_j}ln{z}_j+{\displaystyle \frac{1}{2}}{h}_{pp}lnplnp \\ +{\displaystyle \frac{1}{2}}\sum _{k=1}^n\sum _{i=1}^n{h}_{w_i{w}_k}ln{w}_{i}ln{w}_k+\sum _{i=1}^n{h}_{p{w}_i}ln{w}_{i}lnp+\epsilon \end{gathered}$$

(3)

Applying the Hotelling’s Lemma principle, the output and input share equations can be derived using Equation (3), and are as follows:

(1)

Output share equation

$$S_p={\displaystyle \frac{\partial ln\pi }{\partial lnp}}={\alpha }_p+h_{pp}lnp+\sum _{i=1}^n{h}_{p{w}_i}ln{w}_i>0$$

(4)

(2)

Input share equations

$$S_{w_i}={\displaystyle \frac{\partial ln\pi }{\partial ln{w}_i}}={\alpha }_{w_i}+h_{w_i{w}_i}ln{w}_i+\sum _{j=1 j\ne i}^n{h}_{w_i{w}_j}ln{w}_j+h_{p{w}_i}lnp<0$$

(5)

where $S_p$ denotes the share of output revenue in profits, $S_{w_i}$ denotes the share of input costs of $x_i$ in profits. ${\alpha }_p$, ${\alpha }_{w_i}$, $h_{pp}$, $h_{p{w}_i}$, $h_{w_i{w}_i}$ and $h_{w_i{w}_j}$ are the parameters to be estimated. Since, by definition, $S_p+\sum S_{w_i}=1$, to avoid the generation of singular matrices in the process of estimation, a common practice is to drop one of the input share equations and normalize the output and input prices by the selected input price, and then estimate the remaining equations as simultaneous systems. To control for the differences between years and regions, this paper also incorporates year and province dummies in the regression. The partial elasticities of the respective output supply and input demand with respect to the output and input prices are computed using Equations (6)–(10). Furthermore, the substitution elasticities between factors can be computed based on Equation (11).

(1)

Partial elasticities of the output supply

$${\displaystyle \frac{\partial lnQ}{\partial lnp}}=S_p+{\displaystyle \frac{h_{pp}}{s_p}}-1$$

(6)

$${\displaystyle \frac{\partial lnQ}{\partial ln{w}_i}}=S_{w_i}+{\displaystyle \frac{h_{p{w}_i}}{s_p}}$$

(7)

(2)

Partial elasticities of the input demand

$${\displaystyle \frac{\partial ln{Q}_i^d}{\partial lnp}}=S_p+{\displaystyle \frac{h_{p{w}_i}}{s_{w_i}}}$$

(8)

$${\displaystyle \frac{\partial ln{Q}_i^d}{\partial ln{w}_i}}=S_{w_i}+{\displaystyle \frac{h_{w_i{w}_i}}{s_{w_i}}}-1$$

(9)

$${\displaystyle \frac{\partial ln{Q}_i^d}{\partial ln{w}_j}}=S_{w_j}+{\displaystyle \frac{h_{w_i{w}_j}}{s_{w_i}}}$$

(10)

(3)

Substitution elasticities between factors

$${\delta }_{ij}={\displaystyle \frac{2h_{w_i{w}_j}}{S_{w_i}+S_{w_j}}}-{\displaystyle \frac{S_{w_i}{h}_{w_j{w}_j}}{(S_{w_i}+S_{w_j})S_{w_j}}}-{\displaystyle \frac{S_{w_j}{h}_{w_i{w}_i}}{(S_{w_i}+S_{w_j})S_{w_i}}}+1$$

(11)

The partial elasticities of the input demand with respect to the labor price computed from Equations (9) and (10) can be interpreted as the price effects of labor on factor use.

2.3. Data and Variables

For the empirical analysis, this paper establishes two panel data sets, namely province-level panel data of 921 observations from 2004 to 2022 for the baseline regression and heterogeneity analysis, and farm-level panel data of 84,798 observations from 2004 to 2012 for the mechanism test. The province-level data were mainly collected from the China Statistical Yearbook, the Compiled Materials of Costs and Returns of Agricultural Products of China and the China Rural Statistical Yearbook. The farm-level data were obtained from the annual Survey of Costs and Returns of Agricultural Products conducted by China’s National Development and Reform Commission. Covering 30 provinces in China and more than 20 kinds of agricultural products, this survey provides detailed crop-specific indicators regarding the costs and returns of agricultural production, such as farm size, yield, product price, and the quantity and price of main input factors, making it well-designed for the translog profit function analysis. As this paper mainly concentrates on labor price effects, it is sensible to apply data from 2004 onwards, after which time China witnessed a sustained and rapid increase in the cost of rural labor. Due to data availability, nine rounds (2004–2012) of the survey data are used in this study.

In the model specification, the proportion of crop planting area is measured by the ratio of crop planting area to the total area of all three crops in the specific province. In Chinese grain production, farms’ labor inputs consist mainly of household labor, with a relatively limited proportion of hired labor. To better reflect the opportunity cost of agricultural labor working, the rural labor price is defined as the weighted average price of household and hired labor in crop production. In this paper, the product price is measured by the current price rather than the previous price. The main reason for this treatment is that China launched the minimum purchasing price policy in 2005, which effectively reduces the fluctuation of domestic grain prices and leads to a more expectable product price. Therefore, this paper argues that farmers’ production decisions are more likely to be affected by the current product price, and that the profit function should be a function of the current price. The farmland price is calculated by the weighted average price of contracted and transferred land. Due to the unavailability of the machinery price in the dataset, this paper constructs a province-level machinery price, which is measured as the aggregated price of three machinery services, including farming, sowing, and harvesting per mu, by combining the agricultural machinery service prices from a Special Issue of Agricultural Mechanization released by China’s Management Division of Agricultural Mechanization and the machinery price index from China Statistical Yearbook (2004–2022). The rural labor quality is measured by the average years of schooling of the rural population. The urbanization level, which reflects the demand for food in a specific province, is calculated as the ratio of the urban population to the total population of the province. Agricultural policy support is defined as the annual provincial fiscal expenditure for agriculture. In the mechanism test, farmland, fertilizer, machinery, and labor are chosen as the main input factors of the translog profit function model. Factors other than the aforementioned are collectively referred to as other inputs, which is also the normalizing factor in the model.

Table 1. Descriptive statistics.

Variable	Definition	Rice		Wheat		Maize
		Mean	S.D.	Mean	S.D.	Mean	S.D.
Panel A 2004–2022 province-level panel data
Proportion of planting area	Ratio of crop planting area to the total planting area of all three crops	0.37	0.35	0.27	0.25	0.36	0.27
Product price	Current weighted average of market price and purchase price (CNY/kg)	2.46	0.94	2.19	1.53	2.03	1.08
Farmland price	Weighted average price of contracted and transferred land (CNY/mu)	234.17	219.21	148.61	108.75	158.30	97.00
Fertilizer price	Weighted average price of N, P and K fertilizer (CNY/kg)	5.30	1.26	4.99	0.99	5.41	3.00
Machinery price	Aggregated price of agricultural machinery services, including farming, sowing and harvesting (CNY/mu)	167.08	167.15	97.62	59.14	69.67	46.66
Labor price	Weighted average price of household and hired labor price (CNY/day)	59.52	32.61	55.75	33.96	56.58	30.90
Cultivated farmland per capita	Arable farmland area divided by total population in a specific province (Mu/person)	1.61	1.44	1.61	1.44	1.61	1.44
Rural labor quality	Average years of schooling of the rural population (Year)	7.47	0.88	7.47	0.88	7.47	0.88
Urbanization level	Ratio of urban population to the total population in a specific province	0.55	0.15	0.55	0.15	0.55	0.15
Agricultural policy support	Annual provincial fiscal expenditure for agriculture (CNY billion)	398.05	321.43	398.05	321.43	398.05	321.43
Observations		301		301		319
Panel B 2004–2012 farm-level panel data
Planting area	Area of working arable land during the year (Mu/household)	6.76	28.05	6.88	24.00	7.87	19.10
Yield	Crop output per unit area (Kg/mu)	401.65	70.20	339.38	102.37	445.05	121.71
Revenue	Product of crop price and 50 kg crops (CNY/50 kg)	80.66	11.64	70.26	7.84	66.17	10.32
Cost	Costs of all five input factors per 50 kg crops (CNY/50 kg)	61.83	18.88	62.61	27.24	48.57	21.34
Profit	Difference between revenues and costs (CNY/50 kg)	18.83	16.82	7.66	26.43	17.60	19.87
Fertilizer input	Fertilizer used in pure volume per mu, including N, P and K (Kg/mu)	20.18	6.19	21.20	8.74	21.31	8.46
Machinery input	Machinery used per mu, calculated as the machinery costs divided by the aggregated price including farming, sowing and harvesting (${\mathrm{M}\mathrm{u}}^{-1}$)	0.49	0.36	0.88	0.48	0.37	0.31
Labor input	Aggregated working days of household and hired labor per mu (Days/mu)	8.55	3.30	7.14	4.04	9.51	4.91
Observations		24,596		27,018		33,184

presents detailed variable definitions and descriptive statistics. As shown in the table, the average proportion of planting area is 0.37 for rice, 0.27 for wheat, and 0.36 for maize. The standard deviation of the proportion of rice planting area is bigger than that of wheat and maize, suggesting a significant change in rice production from 2004 to 2022. The average price of rice is approximately CNY 2.5 per kilogram, while that of wheat and maize is slightly lower, at CNY 2.2 and 2.0 per kilogram, respectively. The relatively low standard deviations of the product prices for the three staple crops reflect the stability of product prices in China and the effectiveness of the minimum purchase price policy. Since the establishment of the farmland rental market in 2005, farmland prices in Chinese grain production have experienced sustained increases. As shown, wheat and maize share similar farmland prices in production, while the price of farmland for rice production is higher, at about CNY 240/mu. The average prices of fertilizer for the three crops are similar, at around CNY 5/kg. There are significant differences in the price of the agricultural machinery service used for the three crops. The agricultural machinery price is about CNY 170 per mu for rice, CNY 100 per mu for wheat, and CNY 70 per mu for maize. In terms of labor, the labor prices for the three crops average around CNY 60, with a standard deviation of almost half the average, suggesting significant price changes. In addition, a more intensive labor-dependent production approach is indicated by the relatively higher labor prices for rice and maize. Because of obvious regional differences in natural resource endowment in the country, the conditions for grain production in China have significant regional differences. As presented in the table, although the average cultivated farmland per capita is only 1.6 mu, the standard deviation of the cultivated farmland per capita reaches 1.4, which implies an uneven distribution of the arable land resources across regions. The human capital of the rural labor force is still low, with an average of 7.5 years of schooling. The urbanization level averages 0.55. It is worth noting that the standard deviation of agricultural policy support reaches 320, close to its average level, suggesting a rapid increase in the fiscal expenditure for agriculture and a sustained effort by the Chinese government to maintain food security. Table 1 also presents summary statistics of variables regarding input and output in the grain production. As shown, the average planting area for three crops ranges from 6 to 8 mu, with the standard deviations larger than the average, suggesting that most of the farms plant crops with an area less than 6 mu. The yield of maize is the largest among the three crops, with an average of 445 kg per mu, followed by rice and wheat at 400 and 340 kg, respectively. Compared with wheat and maize, rice production is characterized as high revenue, high cost and high profit. In Table 1, excessive fertilizer use is found in crop production, exceeding 20 kg per mu. Wheat production uses mechanization throughout the whole process, while the mechanization levels of rice and maize are relatively low. Accordingly, the labor use in wheat production is the lowest, with an average of 7 days per mu, followed by rice and maize, indicating a potential substitution relationship between labor and machinery.

3. Results

3.1. Baseline Regression

Table 2. The effect of rising labor prices on the proportion of rice planting area.

Dep. Var.	The Proportion of Rice Planting Area
	(1)	(2)	(3)	(4)
Labor price	0.133 ***	0.109 ***	−0.032 ***	−0.028 ***
	(0.031)	(0.040)	(0.007)	(0.009)
Constant	1.239 ***	1.439 ***	0.351 ***	0.368 ***
	(0.160)	(0.190)	(0.053)	(0.103)
Control variables	YES	YES	YES	YES
Year FE	NO	YES	NO	YES
Province FE	NO	NO	YES	YES
Observations	302	302	301	301
R2	0.425	0.478	0.985	0.986

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

,

Table 3. The effect of rising labor prices on the proportion of wheat planting area.

Dep. Var.	The Proportion of Wheat Planting Area
	(1)	(2)	(3)	(4)
Labor price	−0.027	0.001	0.044 ***	0.042 ***
	(0.019)	(0.025)	(0.011)	(0.013)
Constant	−0.116	−0.367 *	0.702 ***	0.594 ***
	(0.130)	(0.222)	(0.095)	(0.171)
Control variables	YES	YES	YES	YES
Year FE	NO	YES	NO	YES
Province FE	NO	NO	YES	YES
Observations	301	301	301	301
R2	0.604	0.620	0.940	0.948

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table 4. The effect of rising labor prices on the proportion of maize planting area.

Dep. Var.	The Proportion of Maize Planting Area
	(1)	(2)	(3)	(4)
Labor price	−0.054 ***	−0.065 ***	−0.024 ***	−0.020 **
	(0.019)	(0.021)	(0.007)	(0.008)
Constant	0.174	−0.079	0.043	0.131
	(0.131)	(0.175)	(0.085)	(0.168)
Control variables	YES	YES	YES	YES
Year FE	NO	YES	NO	YES
Province FE	NO	NO	YES	YES
Observations	319	319	319	319
R2	0.464	0.497	0.939	0.941

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

report the estimated results of the baseline regression model for the three crops, with all models controlling for a series of agricultural market price signals, provincial production and demand conditions, and agricultural policy support. Columns (1) to (4) of the tables present the estimates, with the stepwise inclusion of year and province fixed effects. The estimated coefficients of labor prices change significantly with the inclusion of fixed effects, especially province-fixed effects, which implies that province-level unobservable characteristics have pronounced effects on land use patterns. The R-square of the models with year and province fixed effects ranges from 0.95 to 0.99, suggesting the outstanding goodness of fit of the model. The results presented in Column (4) are used as the measured effect of rising labor prices on land use patterns.

As shown in the table, the estimated coefficients of labor price for the share of the planting area of three crops are all statistically significant but different in sign. Specifically, a 1% increase in the rural labor price decreases the share of the planting area of rice and maize by 0.028% and 0.020%, respectively. The effect of rising labor prices on the share of the wheat planting area is positive and more substantial, with the coefficient being 0.042. This suggests that the increase in rural labor prices leads to changes in the land use pattern in grain production, namely the expansion of wheat production and the shrinkage of rice and maize.

For the robustness of the estimated results, mitigating the potential bias caused by extreme samples, this paper further restricts the sample with a crop planting area greater than 5%. As shown in

Table 5. The effect of rising labor prices on the proportion of crop planting area with restricted samples.

Dep. Var.	The Proportion of Crop Planting Area
	Rice	Wheat	Maize
Labor price	−0.033 ***	0.041 ***	−0.020 **
	(0.012)	(0.013)	(0.008)
Constant	0.509 ***	0.609 ***	0.131
	(0.144)	(0.171)	(0.168)
Control variables	YES	YES	YES
Year FE	YES	YES	YES
Province FE	YES	YES	YES
Observations	210	298	319
R2	0.976	0.947	0.941

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the sample sizes of wheat and maize change slightly, but the sample of rice drops by nearly 30%. The estimated results are in line with Table 2, Table 3 and Table 4, indicating the good robustness of the baseline results.

3.2. Mechanism Test

Farmers tend to substitute expensive scarce factors with cheap abundant ones in response to market price signals. Rising rural labor prices are thus expected to lead to the adjustment of the input portfolio towards the labor-saving direction and a change in the crop planting area. Based on this logic, this subsection examines the effect of rising labor prices on farming decisions to identify the mechanisms through which labor prices change the patterns of land use.

In the estimation of the translog profit function model, the crop output share equation and factor input share equations have to be estimated as simultaneous systems. However, the sum of the crop output share and factor input shares should equal 1 by definition, which results in a correlation between the disturbance terms of these equations. In this case, ordinary least squares estimation cannot produce unbiased results. This paper, therefore, employs seemingly unrelated regression. The estimated coefficients are presented in

Table 6. Seemingly unrelated regression results of crop production.

Dep. Var.	Output Revenue Share to Profits	Input Cost Share to Profits
		Farmland	Fertilizer	Machinery	Labor
Panel A Rice Production
Product price	−1.685 ***	0.219 **	0.405 ***	0.293 ***	0.418
	(0.597)	(0.098)	(0.102)	(0.093)	(0.260)
Farmland price	0.219 **	−0.339 ***	0.020	−0.001	0.097 **
	(0.098)	(0.023)	(0.019)	(0.017)	(0.042)
Fertilizer price	0.405 ***	0.020	−0.268 ***	−0.052 *	0.012
	(0.102)	(0.019)	(0.030)	(0.030)	(0.039)
Machinery price	0.293 ***	−0.001	−0.052 *	−0.161 **	0.005
	(0.093)	(0.017)	(0.030)	(0.070)	(0.039)
Labor price	0.418	0.097 **	0.012	0.005	−0.597 ***
	(0.260)	(0.042)	(0.039)	(0.039)	(0.131)
Observations	24,596	24,596	24,596	24,596	24,596
Panel B Wheat Production
Product price	0.871	−0.043	0.024	−0.014	−0.678 **
	(0.920)	(0.174)	(0.179)	(0.130)	(0.329)
Farmland price	−0.043	−0.191 ***	0.070*	0.005	0.107 *
	(0.174)	(0.052)	(0.038)	(0.030)	(0.063)
Fertilizer price	0.024	0.070 *	−0.111 **	0.002	0.041
	(0.179)	(0.038)	(0.048)	(0.038)	(0.067)
Machinery price	−0.014	0.005	0.002	−0.152 **	−0.012
	(0.130)	(0.030)	(0.038)	(0.073)	(0.049)
Labor price	−0.678 **	0.107 *	0.041	−0.012	0.377 **
	(0.329)	(0.063)	(0.067)	(0.049)	(0.171)
Observations	27,018	27,018	27,018	27,018	27,018
Panel C Maize Production
Product price	−1.869 ***	0.354 ***	0.436 ***	0.153 ***	0.672 **
	(0.591)	(0.102)	(0.112)	(0.047)	(0.294)
Farmland price	0.354 ***	−0.353 ***	−0.019	−0.009	0.039
	(0.102)	(0.029)	(0.023)	(0.010)	(0.053)
Fertilizer price	0.436 ***	−0.019	−0.270 ***	−0.049 ***	−0.029
	(0.112)	(0.023)	(0.033)	(0.015)	(0.057)
Machinery price	0.153 ***	−0.009	−0.049 ***	−0.133 ***	−0.093 ***
	(0.047)	(0.010)	(0.015)	(0.034)	(0.024)
Labor price	0.672**	0.039	−0.029	−0.093 ***	−0.566 ***
	(0.294)	(0.053)	(0.057)	(0.024)	(0.160)
Observations	33,184	33,184	33,184	33,184	33,184

Notes: Product and factor prices are all in log form. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

and used in Equations (6)–(10) to derive the price elasticities of factors, which can also be referred to as the effects of price change on factor use. Since this study mainly focuses on the effects of rising labor prices on agriculture, only elasticities with respect to labor prices are presented in

Table 7. The effect of rising labor prices on yield, planting area and factor use.

Dep. Var.	Yield	Planting Area	Factor Use Per Mu
			Fertilizer	Machinery	Labor
Rice
Labor price	0.459	−1.474	0.330	0.332	0.083
Wheat
Labor price	0.109	−0.562	0.139	0.229	−1.344
Maize
Labor price	0.314	−1.016	0.206	0.802	−0.014

Source: Author’s calculation.

.

As shown in Table 7, the steady rise in rural labor prices exerts pronounced effects on the factors used in grain production. Rises in the labor price significantly increase the fertilizer and machinery use per mu, suggesting substitution relationships between fertilizer, machinery, and labor, which is in line with previous studies [23,24,25,26]. Given the evidence of excessive use of fertilizer in grain production from descriptive statistics, the increasing intensity of fertilizer use caused by rising labor prices will inevitably lead to more severe environmental pollution. The labor price effects on machinery are more pronounced for rice and maize due to their relatively low-level use of machinery in production. Each 1% increase in labor price will increase the use of machinery per mu by 0.33%, 0.23%, and 0.80% for rice, wheat, and maize, respectively. When it comes to the effects on labor use, the price effects show two different patterns among the three crops, with the labor intensity remaining almost unchanged for rice and maize but decreasing sharply for wheat, which suggests the relatively higher importance of labor in rice and maize production. The divergence in labor use change also leads to differences in the adjustment of the planting area for the three crops. For labor-intensive crops such as rice and maize, farmers tend to reduce their planting area significantly to mitigate the rising costs caused by higher labor prices, with each 1% increase in the rural labor price decreasing their planting area by more than 1; meanwhile, for wheat, the decline in area was relatively small, at only 0.6%. The uneven decline in the crop planting area ultimately results in changes in the share of the crop planting area, which is consistent with the baseline results.

A further exploration of the reasons for the distinct patterns of labor use change sheds light on the impact of rising labor prices on changes in the land use pattern. To this end, this paper calculated the substitution elasticities between fertilizer, machinery, and labor by combining the estimated results in Table 6 and Equation (11). As shown in

Table 8. The substitution elasticities between fertilizer, machinery and labor.

	Rice	Wheat	Maize
Substitution Elasticity
Fertilizer-Labor	0.269	1.195	0.237
Machinery-Labor	0.235	1.172	0.176

Source: Author’s calculation.

, both fertilizer and machinery are substitutes for labor. Consistent with the aforementioned discussion, the substitution elasticities are moderate for rice and maize. Each 1% increase in the relative price of labor increases the use of fertilizer and machinery by approximately 0.2% relative to labor use. The estimated substitution elasticities are more pronounced for wheat, with coefficients of more than 1.0, suggesting elastic labor substitution with fertilizer and machinery. From this perspective, it is the factor substitution that determines the capability of farmers to mitigate the rising labor costs for the three crops by adjusting the input portfolio, which ultimately results in different degrees of decline in the crop planting area.

As the farm-level data used in this section only cover the period 2004–2012, it is possible that the evidence from the mechanism test is not sufficient to explain the results of the baseline regression. By incorporating the interaction term of labor price and a dummy variable for whether the year is after 2012 into the baseline specification, this paper provides extra evidence in response to the aforementioned concern. It is worth noting that this interaction term essentially reflects a time trend in labor price effects, so the year-fixed effects should not be controlled for in this regression. The estimated coefficients of labor price for the three crops presented in

Table 9. The heterogeneous effects of rising labor prices on the proportion of the crop planting area for different time periods.

Dep. Var.	The Proportion of Crop Planting Area
	Rice	Wheat	Maize
Labor price	−0.034 ***	0.047 ***	−0.027 ***
	(0.007)	(0.011)	(0.008)
Labor price × I (Year > 2012)	0.001	−0.006 **	0.003
	(0.001)	(0.003)	(0.002)
Constant	0.356 ***	0.674 ***	0.063
	(0.054)	(0.095)	(0.087)
Control variables	YES	YES	YES
Year FE	NO	NO	NO
Province FE	YES	YES	YES
Observations	301	301	319
R2	0.985	0.941	0.939

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

are in line with the baseline results. The interaction term shows a statistically insignificant effect for rice and maize. Although its estimated effect for wheat is significant, its coefficient is relatively small, at −0.006, which does not fundamentally change the conclusion obtained from the baseline regression. As there is no significant difference in the price effect of labor before and after 2012, the evidence from the mechanism test based on data from 2004 to 2012 is highly explanatory of the baseline results covering 2004–2022.

3.3. Spatial–Temporal Analysis

The mechanism effect of factor substitution was found in the previous subsection. It is thus reasonable to speculate that, even for the same crop, the application condition of substitute factors will directly affect the price effects of labor. Since there are hardly any restrictions on fertilizer use, this paper mainly investigates the heterogeneous price effects across regions with different machinery application conditions. Compared with plain areas, hilly and mountainous areas pose significant obstacles to mechanical operations. To precisely capture the difference in terrain conditions, following a previous study [35], this paper constructs a province-level index of the relief degree of land surface (RDLS) based on Equation (12).

$$RDLS=\{Max\left(H\right)-Min\left(H\right)\}\ast (1-{\displaystyle \frac{P\left(A\right)}{A}})/500$$

(12)

where $Max\left(H\right)$ and $Min\left(H\right)$ denote the highest and lowest elevation in the given region, respectively. $P\left(A\right)$ is the area of flat land within the region. $A$ is the total area of the region. In this paper, China’s DEM (Digital Elevation Model) data from GTOPO30, obtained by the U.S. Geological Survey, are used to calculate the RDLS. The RDLS is first calculated based on a 10 km × 10 km grid and then averaged within each province to obtain the provincial RDLS index. The RDLS averages 1.17, with a standard deviation of 1.29.

A regression model incorporating the interaction term of labor price and RDLS is estimated. As presented in

Table 10. The heterogenous effects of rising labor prices among provinces with different terrain conditions.

Dep. Var.	The Proportion of Crop Planting Area
	Rice	Wheat	Maize
Labor price	−0.023 **	0.065 ***	−0.022 ***
	(0.009)	(0.013)	(0.008)
Labor price × RDLS	−0.007 **	−0.026 ***	−0.017 ***
	(0.003)	(0.004)	(0.004)
Constant	0.313 ***	0.257	0.298 *
	(0.105)	(0.170)	(0.166)
Control variables
Year FE	YES	YES	YES
Province FE	YES	YES	YES
Observations	301	301	319
R2	0.986	0.955	0.945

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

, the estimated coefficients of the interaction term are negatively statistically significant for all three crops, indicating the negative marginal effect of the RDLS, which is consistent with the results of previous studies [27]. Each unit increase in the RDLS decreases the price effect of labor by 0.007 for rice, which suggests that the terrain condition has a moderate impact on the substitution relationship between machinery and labor in rice production. Meanwhile, for maize, mechanical operations are more demanding in terms of the terrain conditions. The estimated marginal effect of RDLS is substantial, at −0.026. The estimated coefficient of the interaction term is −0.017 for wheat. This suggests that the comparative advantage of substituting machinery for labor in wheat production diminishes rapidly as the terrain becomes more rolling.

As shown, there are apparent differences in the price effects of labor in regions with various terrain conditions. It is worth noting that poor terrain conditions not only make the use of machinery more difficult but may also induce more fertilizer use to substitute for labor, leading to severe environmental pollution. Moreover, against the backdrop of rapidly rising rural labor prices, weak substitution relationships between machinery and labor caused by poor terrains may further induce land abandonment [36]. The development of agricultural machinery that can cope with various terrain conditions is therefore of great practical importance in mitigating the pressure of rising labor costs and environmental pollution.

Factor substitution relationships may be affected by technical advancements and change over time; this, in turn, results in disparities in the labor price effects in the long run. To explore the potential temporal characteristics of the effects of rising labor prices on Chinese grain production, the interaction terms of labor price and all-year dummies are incorporated into the regression model. The estimated results are presented in

Table 11. The heterogeneous effects of rising labor prices on the proportion of crop planting area across different years.

Dep. Var.	The Proportion of Crop Planting Area
	Rice	Wheat	Maize
Labor price	−0.028 **	0.031 **	−0.021 *
	(0.014)	(0.016)	(0.013)
Labor price × I (Year = 2005)	−0.003	0.005	−0.003
	(0.004)	(0.007)	(0.004)
Labor price × I (Year = 2006)	−0.001	0.018 ***	−0.002
	(0.004)	(0.007)	(0.004)
Labor price × I (Year = 2007)	−0.004	0.014 *	−0.004
	(0.004)	(0.007)	(0.005)
Labor price × I (Year = 2008)	−0.005	0.014	−0.004
	(0.005)	(0.008)	(0.006)
Labor price × I (Year = 2009)	−0.004	0.010	−0.003
	(0.006)	(0.009)	(0.006)
Labor price × I (Year = 2010)	−0.003	0.011	−0.003
	(0.006)	(0.010)	(0.007)
Labor price × I (Year = 2011)	−0.002	0.003	−0.06
	(0.007)	(0.010)	(0.007)
Labor price × I (Year = 2012)	−0.002	0.011	−0.007
	(0.007)	(0.011)	(0.008)
Labor price × I (Year = 2013)	−0.002	0.006	−0.010
	(0.008)	(0.011)	(0.009)
Labor price × I (Year = 2014)	−0.002	0.004	−0.006
	(0.008)	(0.012)	(0.009)
Labor price × I (Year = 2015)	−0.001	0.000	−0.009
	(0.008)	(0.012)	(0.009)
Labor price × I (Year = 2016)	−0.000	−0.005	−0.004
	(0.008)	(0.012)	(0.009)
Labor price × I (Year = 2017)	−0.001	−0.006	−0.008
	(0.009)	(0.013)	(0.009)
Labor price × I (Year = 2018)	−0.001	−0.006	−0.009
	(0.009)	(0.013)	(0.010)
Labor price × I (Year = 2019)	0.003	−0.005	−0.005
	(0.009)	(0.013)	(0.010)
Labor price × I (Year = 2020)	0.002	−0.009	−0.004
	(0.009)	(0.013)	(0.010)
Labor price × I (Year = 2021)	0.001	−0.005	−0.007
	(0.009)	(0.014)	(0.011)
Labor price × I (Year = 2022)	−0.001	−0.003	−0.008
	(0.010)	(0.014)	(0.011)
Constant	0.355 ***	0.760 ***	−0.030
	(0.093)	(0.153)	(0.162)
Control variables	YES	YES	YES
Year FE	NO	NO	NO
Province FE	YES	YES	YES
Observations	301	301	319
R2	0.986	0.948	0.941

Notes: Labor price is in log form. Control variables include the product price, farmland price, fertilizer price, machinery price, the ratio of the prices of the three crops, the cultivated land area per capita, the rural labor quality, the urbanization level, and the provincial fiscal expenditure for agriculture. The standard errors listed in parentheses are clustered at the provincial level. *** p < 0.01, ** p < 0.05, * p < 0.1.

. The interaction terms of labor price and 2004-year dummy are dropped due to multicollinearity. The estimated coefficients of the interaction terms represent the difference in labor price effects between 2004 and the specific year.

As shown in the table, the estimated coefficients of labor price remain statistically significant, while most of those for the interaction terms are insignificant. The robust effects of rising labor prices imply a relatively stable factor substitution relationship across years. The estimated results also, to some extent, verify the plausibility of the assumption of the translog profit function model, which is that the substitution elasticities between factors remain constant over time. It is indicated by this finding that there is still scope for advancements in key technologies within the Chinese agricultural sector. Meanwhile, the Chinese government should vigorously promote novel agricultural technologies, especially to facilitate the adoption of labor-saving technologies in grain production. This would be of great practical significance for achieving cost reduction and efficiency promotion in Chinese grain production.

4. Conclusions and Policy Implications

This paper analyzes the effect of rising rural labor prices on the land use pattern in Chinese grain production based on province-level panel data from 2004 to 2022 and farm-level panel data from 2004 to 2012. Baseline regression is conducted using a fixed effects model to investigate the labor price effect on the provincial land use pattern. A translog profit function model is employed to identify the micro-mechanism of the price effect. The empirical results show that an increase in rural labor prices leads to significant changes in the share of the planting area of rice, wheat, and maize. As the rural labor price increases, the share of rice and maize decreases significantly, while that of wheat increases substantially. The main reason for this land use pattern change is the factor substitution, especially labor substitution with fertilizer and machinery induced by the relative change in the labor price. Compared with rice and maize, the substitution of fertilizer and machinery for labor is more elastic for wheat, providing greater space to mitigate the rising labor costs by adjusting the input portfolio, which ultimately leads to an expansion in the proportion of the planting area in the context of rising rural labor prices. Furthermore, the increase in labor prices has significant heterogeneous effects on the land use patterns among regions with various terrain conditions. As the terrain becomes more rolling, the application of machinery in substitution for labor becomes more difficult, and the negative price effect of labor becomes more pronounced, especially for wheat. No evidence of significant differences in labor price effects across years was found for all three crops.

These empirical findings have the following policy implications.

First, compared with developed countries, the yields of rice and maize in China are still at a relatively low level. The continuous decline in the share of the planting area of rice and maize may lead to a contradiction between the supply and demand structure of agricultural products. This study suggests promoting land use efficiency and grain production capacity by reducing the wheat acreage with low yields and increasing investment in soil fertility, which provides more space for the adjustment of land use patterns in the context of rising labor prices.

Second, the substitution of fertilizer and machinery for labor is still weak in rice and maize production, resulting in the limited capacity of farmers to mitigate rising labor costs by adjusting their input portfolio. Enhancing the research and development of labor-saving technologies, especially enhanced-efficiency fertilizer and agricultural machinery suitable for rice and maize farming, is of great practical importance for reducing costs and increasing efficiency in grain production.

Third, because of obvious regional differences in the natural resource endowment of the country, the conditions for grain production in China have significant regional differences, which may affect the factor substitution for labor. Compared with plain areas, hilly and mountainous areas pose significant obstacles to mechanical operations, leading to a greater reduction in the share of grain acreage and, in some cases, to the abandonment of farmland. Therefore, it is necessary not only to enhance the development of small-sized agricultural machinery but also to provide some policy support to maintain stable grain production in hilly and mountainous areas in the context of rising labor costs. At last, the Chinese government should vigorously promote novel agricultural technologies, especially to facilitate the adoption of labor-saving technologies in grain production, which would be of great practical significance for achieving cost reduction and efficiency promotion in Chinese grain production.

This paper investigates the effect of rising labor prices on land use patterns in Chinese grain production and explores the role of factor substitution on the price effect, which provides useful information for policy making. However, there are still several limitations to our study. First, the farm-level data used in the mechanism test cover only 2004–2012, which may not fully capture the factor substitution relationship during the 2004–2022 period. Second, as the objective of the mechanism test is to present a picture of factor substitution elasticity, this paper employs the translog profit function model and assumes that all the elasticities are constant over time, which might not align with the dynamic effects of technological advancements and policy changes on factor combinations. Further research should be cautious in analyzing the results. In addition, this paper only takes grain production into account. If other agricultural activities, such as cash crop production, animal husbandry, and forestry, were incorporated into the model, the estimated elasticities may change; this should be studied in future research.