Analysis of Influencing Factors and Income Effect of Heterogeneous Agricultural Households’ Forestland Transfer

Abstract

After the collective forest tenure reform, the forestland transfer behavior choices did not reach the policy expectations. In order to explore the factors influencing the behavior of farmers’ forestland transfer and the income effect of forestland transfer, this paper constructs a binary logistic regression model and a propensity score matching (PSM) model and uses the data obtained from a questionnaire survey of 500 farmers from 10 counties in Liaoning Province for quantitative analysis. Considering the heterogeneity, agricultural households are divided into three categories, namely, pure agricultural households, part-time agricultural households and non-agricultural households. The study found that the influencing factors of different types of agricultural households’ forestland transfer behavior choices were not exactly the same and the degree of influence by the same variable was also different. Forestland transfer can effectively promote the increase of agricultural and forestry income and total household income of pure agricultural households, part-time agricultural households and non-agricultural households, among which the promotion effect on pure agricultural households is the largest.

1. Introduction

Forestland is the land covered by woody plant communities, which belongs to an important land resource. At the same time, forestland, as the carrier of forest products and ecological benefits, is an important part of forest resource assets. To ensure the sustainable development of forestry and improve the productivity of forestland, China implemented collective forest tenure reform in 2003, establishing the dominant position of agricultural households in forestland management [1]. The Liaoning Province, together with the Fujian and Jiangxi Provinces, as the pilot areas for collective forest tenure reform, carried out a reform exploration with the main content of “clarifying property rights, liberalizing management rights, implementing disposal rights and guaranteeing income rights” [2].

Combined with the growth characteristics of forest resources, intensive management is more conducive to the development of the forestry economy [3,4,5]. In order to improve the efficiency of forestry production and realize the rational allocation of forestland resources, forestland with clear property rights can be transferred under the premise of meeting the policy regulations [6]. The transfer of forestland mainly refers to the transfer of forestland use rights and ownership of forestland attachments, including forestland inflow and forestland outflow. The transfer of forestland ownership not only affects the management mode of forestland and the business environment of forest products, but also affects people and regions that rely on this resource for their livelihoods [7,8]. Taking the forest harvesting rate as an example, the harvesting rate of forestland with ownership change is relatively higher and its economic benefits are also more obvious [9]. Forestland transfer also significantly alters the ability of natural ecosystems to provide ecosystem services that support human well-being [10]. In the first few years after the collective forest tenure reform in 2003, the forestland transfer area increased significantly and reached a high peak in 2010 and 2011 and then showed a significant downward trend [11]. After several collective forest tenure reforms, the problem of the low utilization rate of forestland still exists [12,13]. Therefore, based on the perspective of individual agricultural households, this paper studies the influencing factors of heterogeneous agricultural households’ forestland transfer and further explores the income effect of forestland transfer. The research aims to promote the more standardized development of forestland transfer and provide theoretical reference for the further development of collective forest tenure reform.

2. Literature Review

By consulting a large number of literature materials, this paper sorts out the research findings related to the two aspects of heterogeneous agricultural households and the influencing factors of forestland transfer and makes a brief comment based on the research content of this paper.

2.1. Heterogeneous Agricultural Households

With the in-depth development of industrialization and urbanization in China, agricultural households began to shift from full-time farming to part-time farming [14,15]. Differences in the degree of concurrent occupations make the differences between agricultural households gradually obvious [16,17] and the homogeneity among agricultural households also decreases. When studying the behavior of agricultural households, most scholars regard agricultural households as a unified whole without differences, ignoring the heterogeneity of agricultural households [18,19]. The results of direct research on farmer households are unreliable. It is necessary to identify the heterogeneity of agricultural households and adopt different standards to classify agricultural households. Some domestic and foreign scholars study the decision-making behaviors of heterogeneous agricultural households by identifying the heterogeneity of agricultural households’ farming skills [20], goals and attitudes [21,22], agricultural households’ productivity composition [23], risk exposure [24] and preferences [25,26,27,28,29]. Among them, in terms of farmers’ willingness to agricultural production and disposal of land, the degree of concurrent employment is widely used as a classification standard [30]. Klaus Deininger and Songqing Jin [31] used data from Vietnam to show that the proportion of non-agricultural income in the total household income will affect the disposal of land by agricultural households. Based on a field survey of agricultural households in the Jianghan Plain, Xin Yang and Yiming Sang [32] found that part-time farming had a significant positive impact on the possibility of adopting conservation agriculture (CA) technology. Under the premise of fully considering the heterogeneity of agricultural households, this paper divides agricultural households into pure agricultural households, part-time agricultural households and non-agricultural households according to the degree of part-time farming.

2.2. Influencing Factors of Forestland Transfer

Characteristics of agricultural households, such as the age of householders, education level and the number of household laborers, play an important role in the decision-making of forestland transfer behavior. The complexities of the life experiences of private forestland owners (PFLs) and these experiences will affect their decision-making process for the future planning of forestland [33,34]. Intergenerational land transfer and labor migration of family members have a periodic impact on land transfer [35,36]. Markowski-Lindsay et al. [37] found that forestland transfer decisions of family forest owners (FFOs) were positively correlated with older age, women, owning more forestland and agreeing to sell land at reasonable prices and were negatively correlated with higher education levels. Stone and Tyrrell [38] surveyed forestland owners in the Catskill-Delaware basin and found that they cited property taxes, aging, family members’ lack of interest in forest ownership and personal circumstances as reasons for land segmentation and sale.

In addition, forestland resources of agricultural households, forestland behaviors of agricultural households, forestry compensation policies and details of forestland transfer all have different degrees of influence on agricultural households’ forestland transfer behavior decisions. According to data provided by the United States Department of Agriculture in 2013, family forestland owners with 4 hectares or more of land are less likely to sell their forestland in the next 5 years [37]. Chen Liming’s research results show that the number of forestland plots and the area of forestland significantly affect the transfer of forest resources [39]. In the study of forestland behaviors of agricultural households, participation in community forestry development [40], purchase of forest insurance [41] and forest tenure mortgage loans [42] are closely related to the inflow and outflow of agricultural households’ forestland. Forestry compensation policies, such as afforestation subsidy policy, forest tending subsidy policy and ecological benefit subsidy policy, are effective measures to encourage agricultural households to develop and manage forestland, improve forestland use rate and promote forest economic development [43,44,45]. The transaction price of forestland transfer is the core of the forestland transaction market and it is also an important signal for the allocation of forestland resources and an effective measure for evaluating the use value of forestland. Most of the forestland transfer price is the result of mutual agreement, so the forestland transfer transaction between acquaintances is easier to achieve [46].

Although scholars’ research has achieved great results, there are still some problems. The conclusions drawn from the theoretical studies are not detailed enough, so the recommendations made are more general. In addition, compared with empirical research, they lack data support and accuracy. Whether data from empirical study can accurately express the agricultural households’ real choices about the forestland transfer behavior needs further demonstration and the digital results may not show the dynamics of agricultural households’ choice behavior. Therefore, this paper conducts a field survey in Liaoning Province and the survey samples cover 500 households in 50 villages in 10 counties, to ensure the authenticity and validity of the data as much as possible. At the same time, based on the research of scholars at home and abroad, this paper cautiously selects five types of influencing factors (namely, characteristics of agricultural households, forestland resources of agricultural households, forestland behaviors of agricultural households, forestry compensation policies and details of forestland transfer), and adopts a combination of descriptive analysis and empirical analysis to explore the behavior choice of heterogeneous agricultural households’ forestland transfer. Moreover, the propensity score matching (PSM) model is further established to analyze the income effect of heterogeneous agricultural households’ forestland transfer. Based on the above research significance and research purpose, this paper puts forward the following research hypotheses:

Hypothesis 1.

There are differences in the factors influencing the choice of forestland transfer behavior of heterogeneous agricultural households.

Hypothesis 2.

The transfer of forestland can promote the improvement of agricultural households’ income level and different types of agricultural households receive different degrees of promotion.

Hypothesis 3.

The effect of forestland transfer on the income increase of pure agricultural households is greater than that of part-time agricultural households and non-agricultural households.

3. Data Source, Variable Selection, Model Construction

3.1. Samples and Data

The data used in this paper come from a field survey in the Liaoning Province in 2014. The Liaoning Province, with diverse terrain and rich forest resources, is one of the pilot provinces for the collective forest tenure reform.

The percentage of forest coverage in the eastern part of Liaoning Province amounts to 70% and it is a key area for water conservation and public welfare forest protection. The central and southern parts of the Liaoning Province are located in the Liaohe Plain and are mostly plain shelterbelts. Bordering the Horqin Sandy Land, the northwest part is an important part of the Three-North Shelter Forest Area. The Liaoning Province boasts shelter forests, special-purpose forests, timber forests, fuelwood forests and other forest species structures.

Considering the research purpose, research maneuverability, research convenience and other factors, we selected ten counties in the Liaoning Province, namely Xinbin, Qingyuan, Benxi, Huanren, Kuandian, Liaoyang, Tieling, Kaiyuan, Beipiao and Jianchang and selected five villages from each of the ten counties and ten agricultural households were randomly selected from each village. After confirming the sample, the questionnaire survey was undertaken in the form of interviews and 500 questionnaires were collected. The principle of household head priority was adopted in the interview. In general, we preferred to conduct interviews with household heads. In the case that the head of the household could not be contacted, we selected the most important agricultural and forestry laborer in the household for interview. All respondents were informed that the results of the survey were only for scientific research and would not reveal their identity. The study did not involve violations of social ethics and other issues. The interviews were conducted only when the interviewee was allowed to interview.

3.2. Variable Selection

Combining the field survey results and literature review, this paper selects five types of influencing factors (namely, characteristics of agricultural households, forestland resources of agricultural households, forestland behaviors of agricultural households, forestry compensation policies and details of forestland transfer), and selects a total of 12 covariates from the five types of influencing factors. In the next part of the model construction, this paper explores the effect of 12 covariates on the forestland transfer of agricultural households through logistic regression. At the same time, in the PSM model, the logarithm of total household income and total agricultural and forestry income is used as the explanatory variable and whether forestland transfer is carried out is used as the core explanatory variable. The detailed variable descriptions are shown in

Table 1. Description of all variables in the text.

Variable	Symbol	Type of Variable	Definition
Total agricultural and forestry income	$\mathrm{Ln}{Y}_1$	continuous	Annual income of agricultural households in agriculture and Forestry (USD), logarithm
Total household income	$\mathrm{Ln}{Y}_2$	continuous	Annual income of agricultural households (USD), logarithm
Forestland transfer	Transfer	binary	Transfer = 1, no transfer = 0
The age of householders	$X_1$	continuous	The age self-reported by the interviewee
Education level	$X_2$	binary	Primary school and below = 1, junior middle school = 2, senior high school = 3, junior college or bachelor degree and above = 4
The number of household laborers	$X_3$	continuous	Number of those who are elder than 16 years old and less than or equal to 60 years old
Forestland area	$X_4$	continuous	Total forestland area of households (hectare)
Number of forestland blocks	$X_5$	continuous	Number of total forestland blocks of households
Forestry production cooperative	$X_6$	binary	Participated forestry production cooperative = 1, otherwise = 0
Forest insurance	$X_7$	binary	Purchased forest insurance = 1, otherwise = 0
Afforestation subsidy policy	$X_8$	binary	Understand afforestation subsidy policy = 1, Otherwise = 0
Forest tending subsidy policy	$X_9$	binary	Understand forest tending subsidy policy = 1, Otherwise = 0
Ecological benefit subsidy policy	$X_{10}$	binary	Understand ecological benefit subsidy policy = 1, otherwise = 0
Transfer price	$X_{11}$	binary	Understand transfer price = 1, otherwise = 0
Transfer procedure	$X_{12}$	binary	Understand transfer procedure = 1, otherwise = 0

.

3.3. Model Construction

This paper first constructs a binary logistic regression model to study the impact of 12 covariates on different types of agricultural households’ forestland transfer and then further constructs the PSM model to measure the impact of forestland transfer on the total household income and total agricultural and forestry income of heterogeneous agricultural households.

3.3.1. Binary Logistic Model

In this paper, there are only two results of the agricultural households’ forestland transfer (“transfer” and “no transfer”), so this paper adopts a two-category discrete choice model for analysis. The two-category discrete choice model describes that regardless of the influence of several factors, the decision maker has only two choices of “positive” and “negative” results for a decision. The commonly used models of two-category discrete models are the linear probability model, logistic model and Probit model. Since the probability of forestland transfer in the collected data is less than 15%, the logistic model is selected in this paper.

Based on the above analysis, the function $Y_i=f\left(X_i\right)$ between agricultural households’ forestland transfer behavior and influencing factors is established, where $Y_i$ represents forestland transfer, $i=1$ represents the behavior choice of pure agricultural household’s forestland transfer, $i=2$ represents the behavior choice of part-time agricultural household’s forestland transfer and $i=3$ represents the behavior choice of non-agricultural household’s forestland transfer. $X_i$ represents each influencing factor. $Y_i=1$ means transfer, $Y_i=0$ means no transfer. Set the probability $P_i$ when $Y_i=1$, the value range of $P_i$ is between 0 and 1 and the following binary logistic regression Equation (1) is established:

$$P_i=\frac{1}{1+\exp \left[-\left(\alpha +\beta X_i\right)\right]}$$

(1)

3.3.2. Propensity Score Matching

Agricultural households, as rational persons, will decide whether or not to transfer forestland according to their conditions; the choice of transfer/non-transfer behaviors is not random. Therefore, when studying the impact of forestland transfer on agricultural households’ income levels, there are confounding variables between the experimental group and the control group, which can easily produce systematic deviation. The PSM model can effectively reduce the influence of systematic bias and confounding variables, so as to make a more reasonable comparison between the experimental group and the control group. By constructing a counterfactual framework, PSM finds samples in the control group (non-transferred households) that are as similar as possible to the observable variables of the experimental group (transferred households) to ensure comparability between variables. This can effectively reduce the estimation error caused by sample selection and improve the credibility of the empirical research results. The specific steps are as follows:

First, the “psestimate” command was run to select the covariates that achieved the best fitting effect for different types of agricultural households from 12 covariates by comparing the maximum likelihood values. Then, logistic regression was performed on the selected covariates with the core explanatory variable (Transfer) and propensity score values (p-values) were calculated.

Second, the experimental and control groups were matched by propensity score values. Due to the different matching values and weights applied by different matching methods, the matching results are different. This paper attempts to use three matching methods, namely nearest neighbor matching, radius matching and kernel matching and compare the matching results to find differences. If the results of different matching methods are similar, the matching results are robust and independent of the specific method.

Third, average treatment effect for the treated (ATT) is used to measure the average treatment effect of the sample under the intervention state, that is, the difference between the observation result of sample $i$ under the intervention state and its counterfactual. The formula for calculating the ATT value is shown in Formula (2).

$$ATT=E\left(Y_{1i}|D_i=1\right)-E\left(Y_{0i}|D_i=1\right)$$

(2)

Among them, $D_i=1$ means farmer, $i$ is forestland transfer, $D_i=0$ means farmer $i$ does not have forestland transfer, $Y_{1i}$ means the income of farmer $i$ when forestland transfer occurs, $Y_{0i}$ represents the income of farmer $i$ with forestland transfer if the forestland is not transferred, $E\left(Y_{1i}|D_i=1\right)$ is the mean value of the income of farmer $i$ when the forestland transfer occurs and $E\left(Y_{0i}|D_i=1\right)$ is the mean value of the income of farmer $i$ who actually has forestland transfer if there is no forestland transfer.

Since $E\left(Y_{0i}|D_i=1\right)$ cannot be directly observed in the actual survey, we can select a non-transferred household in the control group with similar characteristic variables to the farmer households with forestland transfer through PSM and observe its income status. The counterfactual $E\left(Y_{0i}|D_i=1\right)$ can be replaced by $E\left(Y_{0i}|D_i=0\right)$ after the PSM matches. Therefore, the calculation formula of the ATT value is further expressed as Formula (3):

$$ATT=E\left(Y_{1i}|D_i=1\right)-E\left(Y_{0i}|D_i=0\right)$$

(3)

4. Analysis of Influencing Factors

4.1. Descriptive Statistics

During the survey, it was found that most of the agricultural households in the surveyed areas were not solely engaged in agricultural production. They were engaged in both agricultural production and other non-agricultural labor or other industries. Therefore, agricultural households could be classified according to their degree of concurrent employment. Combining the literature with relevant studies and characteristics of survey data, this paper classifies agricultural households into three categories, namely pure agricultural households, part-time agricultural households and non-agricultural households. Pure agricultural households are those whose agricultural and forestry income accounts for more than 80% of their total household income, non-agricultural households are those whose agricultural and forestry income accounts for less than 20% and other agricultural households are called part-time agricultural households. The details are shown in

Table 2. Statistics on the types of agricultural households.

Types of Agricultural Households	Number of Households	Percentage
pure agricultural households	197	39.40%
part-time agricultural households	192	38.40%
non-agricultural households	111	22.20%
Total	500	100.00%

.

Among the 500 households surveyed, the number of pure agricultural households and part-time agricultural households was similar, accounting for 39.40% and 38.40% of the total number of households, respectively. The number of non-agricultural households was relatively small, only 111 households, accounting for 22.20%.

As shown in

Table 3. Situation of different types of agricultural households’ forestland transfer.

	Transfer Behavior	Transfer		No Transfer
Types of Agricultural Households		Household	Percentage	Household	Percentage
pure agricultural households		14	7.11%	183	92.89%
part-time agricultural households		13	6.77%	179	93.23%
non-agricultural households		18	16.22%	93	83.78%
Total		45	9.00%	455	91.00%

, non-agricultural households had the highest proportion of forestland transfer, accounting for 16.22% of the total number of non-agricultural households. There were 93 households that had not transferred, accounting for 83.78%. A total of 14 forestland transfers occurred in pure agricultural households, accounting for 7.11% of the total number of pure agricultural households. There were 183 agricultural households that did not transfer, accounting for 92.89%. Among the types of part-time agricultural households, 13 households had conducted forestland transfer and 179 households had not conducted forestland transfer, accounting for 6.77% and 93.23%, respectively.

4.1.1. Characteristics of Agricultural Households

As shown in

Table 4. Forestland transfer and the characteristics of agricultural householders.

		Transfer			No Transfer
		Pure	Part-Time	Non-	Pure	Part-Time	Non-
Age of householders	Under 25 years old	0	0	0	1	0	1
	26 to 45 years old	3	6	3	33	37	18
	46 to 60 years old	9	5	13	83	85	45
	60 years old or above	2	2	2	66	57	29
Number of household laborers	Mean	2.96	3.39	3.76	2.95	3.43	3.73
Education level of householders	Primary School and below	4	3	7	81	72	24
	Junior high school	7	6	10	73	92	53
	High School	2	2	1	21	12	14
	College or undergraduate and above	1	2	0	8	3	2

, the majority of household heads were over 46 years old and the proportion of household heads under 25 years old was very small. The age of householders of the pure agricultural households and the non-agricultural households participating in the forestland transfer was concentrated between 46 and 60 years old, while the age of householders participating in the forestland transfer among the part-time agricultural households was concentrated between 26 and 45 years old. There was no forestland transfer among household heads under the age of 25. There was no significant difference in the number of household laborers between agricultural households who participated in forestland transfer and those who did not. The average value of the laborer’s number of pure agricultural households was the smallest among the three types of households, while the average value of the laborer’s number of non-agricultural households was the largest. In terms of education level, the education level of householders was generally not high, mainly concentrated in junior high school, primary school and below. The education level of householders involved in forestland transfer was mostly junior high school, followed by primary school and below. The education level of householders of pure agricultural households and non-agricultural households without forestland transfer was mostly primary school and below, while the education level of householders of part-time agricultural households without forestland transfer was concentrated in junior high school.

4.1.2. Forestland Resources of Agricultural Households

As shown in

Table 5. Forestland transfer and forestland resources of agricultural households.

	Transfer		No Transfer
	Average Forestland Area (Hectare)	Average Number of Forest Blocks (Blocks)	Average Forestland Area (Hectare)	Average Number of Forest Blocks (Blocks)
Pure agricultural households	13.774	3.71	6.132	2.75
Part-time agricultural households	7.904	3.46	3.163	3.1
Non-agricultural households	7.159	2.89	2.819	2.47

, the average forestland area and the average number of forest blocks of the households with forestland transfer were more than those of the households without forestland transfer. Among the pure agricultural households, the average forestland area with forestland transfer was the largest, at 13.774 hectares, and the average number of forestland blocks was also the largest, at 3.71.

4.1.3. Forestland Behaviors of Agricultural Households

As shown in

Table 6. Forestland transfer and agricultural household’s behaviors.

	Transfer		No Transfer
	Proportion of Participating in Forestry Production Cooperatives	Proportion of Purchasing Forest Insurance	Proportion of Participating in Forestry Production Cooperatives	Proportion of Purchasing Forest Insurance
Pure agricultural households	35.71%	35.71%	13.66%	47.54%
Part-time agricultural households	23.08%	53.85%	11.73%	57.54%
Non-agricultural households	33.33%	50.00%	11.83%	37.63%

, the proportion of agricultural households participating in forestry production cooperatives among the agricultural households with forestland transfer was higher than that of agricultural households without forestland transfer. However, the proportion of the transfer households participating in forestry production cooperatives was not high and the lowest was 23.08% of concurrent agricultural households. The proportion of pure agricultural households and part-time agricultural households with forestland transfer that purchased forest insurance was lower than that of agricultural households without forestland transfer. Overall, the recognition of forest insurance among agricultural households was higher than that of forestry production cooperatives.

4.1.4. Forestry Compensation Policy

Table 7. Forestland transfer and forestry compensation policies.

	Afforestation Subsidy Policy		Forest Tending Subsidy Policy		Ecological Benefit Subsidy Policy
	Know	Don’t Know	Know	Don’t Know	Know	Don’t Know
Pure agricultural households	5	9	7	7	7	7
Part-time agricultural households	10	3	9	4	6	7
Non-agricultural households	12	6	5	13	9	9

lists the understanding of the afforestation subsidy policy, forest tending subsidy policy and ecological benefit subsidy policy for different types of agriculture with forestland transfer. Few of the pure agricultural households with forestland transfer knew about the afforestation subsidy policy, but the part-time agricultural households and non-agricultural households with the forestland transfer were far more aware of the afforestation subsidy policy than those who did not. Among the part-time agricultural households with forestland transfer, more agricultural households understood the forest-tending subsidy policy, while among the non-agricultural households, there were fewer agricultural households who understand the policy. Among the agricultural households with forestland transfer, the number of agricultural households who understood and did not understand the ecological benefit compensation policy was similar.

4.1.5. Details of Forestland Transfer

Table 8. Forestland transfer and transfer price and procedure.

	Transfer		No Transfer
	Proportion of Know Transfer Price	Proportion of Know Transfer Procedure	Proportion of Know Transfer Price	Proportion of Know Transfer Procedure
Pure agricultural households	50.00%	35.71%	28.96%	24.59%
Part-time agricultural households	53.85%	61.54%	30.73%	25.14%
Non-agricultural households	88.89%	88.89%	31.18%	31.18%

lists the proportion of agricultural households who knew the transfer price and transfer procedure to the total agricultural households of different types. Compared with the non-transferred households, the agricultural households who had undergone forestland transfer had a better understanding of the transfer price and transfer procedures of the forestland. Among the transfer households, the non-agricultural households accounted for the largest proportion of them knowing the transfer price, accounting for 88.89%, while among the non-transfer households, the three types of agricultural households who knew the transfer price accounted for a similar proportion, all around 30%. Regardless of whether there was forestland transfer, there were more agricultural households who knew the transfer price than those who understand the transfer procedure, indicating that the agricultural households are more concerned about the transfer price.

4.2. Empirical Analysis

This paper used STATA 15.0 for binary logistic regression and the regression results are shown in

Table 9. The results of logistic regression analysis (n = 500).

Variable	Forestland Transfer
	Pure Agricultural Households	Part-Time Agricultural Households	Non-Agricultural Households
$X_1$	1.008	0.946	0.861 ***
	(0.819)	(0.148)	(0.006)
$X_2$	1.333	4.083 **	0.457
	(0.422)	(0.011)	(0.280)
$X_3$	0.981	0.712	0.476 *
	(0.930)	(0.322)	(0.097)
$X_4$	1.002 *	1.018 **	1.022 **
	(0.067)	(0.004)	(0.017)
$X_5$	1.160	0.863	0.680
	(0.299)	(0.513)	(0.150)
$X_6$	3.830 *	5.219	5.391
	(0.059)	(0.118)	(0.107)
$X_7$	0.505	0.506	1.207
	(0.299)	(0.379)	(0.832)
$X_8$	0.096 **	3.620	0.560
	(0.011)	(0.184)	(0.555)
$X_9$	2.006	2.491	1.500
	(0.339)	(0.344)	(0.676)
$X_{10}$	1.441	0.066 ***	0.109 **
	(0.612)	(0.007)	(0.024)
$X_{11}$	3.240	0.662	4.990
	(0.149)	(0.699)	(0.158)
$X_{12}$	0.852	8.601 *	27.768 **
	(0.846)	(0.055)	(0.018)
Constant term	0.015	0.049	1486.381
−2 Log likelihood	80.616	57.963	44.3
Cox and Snell R2	0.098	0.176	0.386
Nagelkerke R2	0.245	0.45	0.656
Hosmer–Lemeshow test	0.1	0.055	0.982

p-Values in parentheses. *, **, *** indicate significance at the 10%, 5% and 1% levels, respectively.

. From the regression results, it can be seen that the influencing factors of the choice of different types of agricultural households’ forestland transfer behavior are not exactly the same and the degree of influence by the same variable is also different.

Factors that have a significant impact on the choice of pure agricultural households’ forestland transfer behavior include family forestland area, forestry production cooperatives and afforestation subsidy policy. Agricultural and forestry production activities are the sources of income for more than 80% of pure agricultural households and the area of family forestland, as the direct carrier of agricultural and forestry activities, is an inevitable factor for agricultural households’ behavior choices. The pure agricultural households participating in the forestry production cooperatives are 3.830 times more likely to carry out forestland transfer than those that do not participate. Agricultural households participating in forestry production cooperatives can cooperate in the process of agricultural and forestry production, exchange production experience and then improve the economic benefits of forestry and agricultural production. Therefore, agricultural households will conduct more forestland transfers, especially forestland inflows. The afforestation subsidy policy has a significant impact on the forestland transfer at the level of 5%. The afforestation subsidy policy can directly bring an additional economic subsidy to agricultural households and increase their income. Therefore, more agricultural households who understand the application conditions and application process of the afforestation subsidy policy will transfer forestland.

The factors that have a significant impact on the choice of the forestland transfer behavior of part-time agricultural households include education level, family forestland area, ecological benefit compensation policy and transfer procedure. The education level of householders has a significant positive impact on the forestland transfer behavior of part-time agricultural households. Among the part-time agricultural households, the higher the level of education, the higher the understanding and acceptance of the new policies and systems, so they will conduct more forestland transfers. The area of family forestland also has a significant impact on the part-time agricultural households. In order to obtain more income, part-time agricultural households who own a large area of forestland will flow into forestland to expand the scale of operation. The ecological benefit compensation policy is significant at the level of 1% for the behavior choice of part-time agricultural households’ forestland transfer. The ecological benefit compensation mainly includes two parts, namely, the value of ecological service functions and the materialized cost of environmental governance and ecological restoration. The part-time agricultural households who understand the transfer procedure have the willingness to transfer forestland, on the one hand, and on the other hand, they understand the transfer process, so it is easier to transfer the forestland.

Factors that have a significant impact on the choice of non-agricultural households’ forestland transfer behavior include the age of householders, number of family laborers, family forestland area, ecological benefit compensation policy and transfer procedure. For non-agricultural households, young agricultural households have more energy and enthusiasm to take into account agricultural and forestry production labor in addition to other business activities. Therefore, young agricultural households will choose to inflow more forestland to expand the scale of operation and obtain multiple benefits. Less than 20% of the income of non-agricultural households comes from agricultural and forestry income. Therefore, when the number of household laborers increases, the income obtained from other business activities will increase accordingly and the attention to forestland transfer will also decrease. Similar to part-time agricultural households, family forestland area, ecological benefit compensation policy and transfer procedures have the same significant effect on non-agricultural households.

5. Income Effect of Forestland Transfer

5.1. Descriptive Statistics

Table 10. Income of different types of agricultural households.

		Pure Agricultural Households	Part-Time Agricultural Households	Non-Agricultural Households
Transfer households	Annual agriculture and forestry Income	115,304.84	45,792.61	4689.67
	Annual household income	116,437.31	72,398.95	40,651.59
No transfer households	Annual agriculture and forestry Income	28,766.91	6353.51	1207.81
	Annual household income	29,725.45	12,001.37	27,492.40

The currency unit of measure: USD.

shows the average annual agricultural and forestry income and annual household income of different types of agricultural households. On the whole, the average income of agricultural households with forestland transfer is higher than that of agricultural households without forestland transfer. Among them, the income level of pure agricultural households with forestland transfer is higher than that of other households.

5.2. Empirical Analysis of PSM

5.2.1. Average Treatment Effect

In this paper, Stata15.0 statistical software was used to run the “psestimate” command and based on the maximum likelihood value, the covariate that can achieve the best-fitting effect was selected from 12 covariates for different types of agricultural households. The average treatment effect expression was estimated by 1-to-1 nearest-neighbor matching, radius matching and kernel matching of PSM to analyze the impact of forestland transfer on the family agriculture and forestry income and annual family income of different types of agricultural households. The specific estimation results are shown in

Table 11. Average treatment effect for the treated.

		Pure Agricultural Households		Part-Time Agricultural Households		Non-Agricultural Households
		$\mathbf{Ln}{\mathit{Y}}_1$	$\mathbf{Ln}{\mathit{Y}}_2$	$\mathbf{Ln}{\mathit{Y}}_1$	$\mathbf{Ln}{\mathit{Y}}_2$	$\mathbf{Ln}{\mathit{Y}}_1$	$\mathbf{Ln}{\mathit{Y}}_2$
Nearest-neighbor matching	ATT	2.549 ***	2.535 ***	0.914	0.961	3.209 ***	2.093 ***
	S.E.	0.820	0.823	0.801	0.654	0.923	0.389
	t	3.110	3.080	1.140	1.470	3.470	5.380
Radius matching	ATT	1.489 **	1.489 **	1.706 ***	1.521 ***	1.684	0.683
	S.E.	0.632	0.632	0.379	0.327	0.955	0.566
	t	2.360	2.360	4.500	4.650	1.760	1.210
Kernel matching	ATT	1.690 ***	1.690 ***	1.646 ***	1.578 ***	2.344 *	0.811
	S.E.	0.654	0.654	0.426	0.350	1.227	0.652
	t	2.580	2.580	3.870	4.510	1.910	1.240
ATT mean		1.909	1.905	1.422	1.353	2.412	1.196

*, **, *** indicate significance at the 10%, 5% and 1% levels, respectively.

.

The ATT mean values of forestland transfer for pure agricultural households on agricultural and forestry income and total household income were 1.909 and 1.905, respectively; that is, the agricultural and forestry income and total household income of pure agricultural households with forestland transfer had an average increase of 190.9% and 190.5% compared with those of pure agricultural households without forestland transfer, indicating that forestland transfer can effectively improve the income level of pure agricultural households. The results of nearest-neighbor matching for part-time agricultural households were not significant, but the results of radius matching and kernel matching were significant at the 5% and 1% levels, respectively. The average ATT values of forestland transfer for part-time agricultural households on agricultural and forestry income and total household income are 1.422 and 1.353, respectively, indicating that forestland transfer has a certain promoting impact on the income increase of part-time agricultural households and its effect is slightly less than that of pure agricultural households. The impact of forestland transfer on the agricultural and forestry income of non-agricultural households is significant, but the impact on the total income of their families is limited. The main reason is that the forestland transfer directly affects the income of agriculture and forestry, while the total income of non-agricultural households has a diverse income composition and the income from agriculture and forestry is less than 20%, so its promoting effect is not obvious.

5.2.2. Balance Test

To ensure the effect of PSM matching, the balance test was carried out. The balance test results are listed in

Table 12. The results of balance test before and after PSM.

		Pseudo-${\mathit{R}}^2$	LR	Mean Bias	Med Bias	B
Pure agricultural households	before matching	0.17	17.2 ***	48.1	51.4	115.0 +
	after matching
	nearest neighbor matching	0.134	4.47	27.8	19.7	84.7 +
	radius matching	0.026	0.93	14.9	11.2	36.8 +
	kernel matching	0.034	1.21	16.2	12.1	42.2 +
Part-time agricultural households	before matching	0.143	13.61 **	44.5	47	97.8 +
	after matching
	nearest neighbor matching	0.01	0.26	16.9	12.3	22.6
	radius matching	0.005	0.15	6.4	9.2	15.9
	kernel matching	0.002	0.05	3.4	4.2	8.8
Non-agricultural households	before matching	0.551	54.03 ***	65.8	49.8	231.7 +
	after matching
	nearest neighbor matching	-	-	-	-	-
	radius matching	0.155	5.58	22	17.7	95.9 +
	kernel matching	0.207	5.74	23.8	23.5	110.9 +

*** means significant at the 1% level, ** means significant at the 5% level, ⁺ means the B value is greater than the 25% critical value.

. It can be seen from Table 12 that before and after nearest-neighbor matching, radius matching and kernel matching, pure agricultural households had a significant decline in Pseudo-$R^2$ value, LR statistic, standardized deviation, median deviation and B value, indicating that the sample-matching effect of pure agricultural households was good and the best effect was radius matching. After matching, the Pseudo-$R^2$ value of the part-time agricultural households decreased significantly, from 0.143 before matching to 0.002~0.01, almost close to zero; the LR statistic decreased from 13.61 before matching to 0.05~0.26; the mean standardized deviation was less than 17%, which significantly reduces the overall bias; the median deviance drops from 47 before matching to 4.2~12.3; the B value also drops significantly, from 97.8% before matching to below 23%. The above indicators all show that the sample matching results of part-time agricultural households were good. The Pseudo-$R^2$ value, LR statistic, standardized deviation, median deviation and B value of non-agricultural households all decreased after matching. In general, the results of pure agricultural households, part-time agricultural households and non-agricultural households through PSM matching all passed the balance test.

6. Discussion

This paper answers the question of the impact of forestland transfer on the income of heterogeneous agricultural households, filling the gap that the current research does not analyze the income changes of heterogeneous agricultural households after forestland transfer. Through the analysis of influencing factors, this paper also provides a more accurate development direction for the heterogeneous agricultural households’ forestland transfer.

When studying the income effect of agricultural households’ forestland transfer, it is inevitable to consider the expenditure of agricultural households. In general, the agricultural households with high incomes also put in more expenditures accordingly. However, due to the limitation of data acquisition, the relevant data on agricultural households’ expenditures could not be collected. To reduce the error caused by the lack of expenditure data, this paper adopts the PSM model in empirical research. Through the PSM model, each agricultural household with forestland transfer is matched with the most similar sample, which effectively reduces the impact of factors other than forestland transfer on income. The results of this paper confirm that forestland transfer can promote the income increase of agricultural households, which is consistent with Kaili Peng’s [47] empirical results through the China Family Panel Studies. Due to the widening differences among agricultural households, it is a breakthrough to divide agricultural households into pure agricultural households, part-time agricultural households and non-agricultural households according to the proportion of non-agricultural income. The empirical results show that the effect of forestland transfer on increasing the income of different types of agricultural households is indeed different. Among them, the promotion of pure agricultural households is the largest, followed by part-time agricultural households, and non-agricultural households is the smallest. The mean ATT values of forestland transfer for pure agricultural households on agricultural and forestry income and total household income are 1.909 and 1.905, respectively and for part-time agricultural households are 1.422 and 1.353, respectively (Table 11). From the numerical value, it can be seen that the effect of forestland transfer on the income increase of pure agricultural households is slightly higher than that of part-time agricultural households. In addition, the inflow and outflow of forestland have different effects on income [48]. In the future, we will more fully consider the various situations of forestland transfer and conduct more in-depth research. Since China implemented the Returning Farmland to Forest Program in 1999, the government has successively issued several compensation policies to provide economic compensation to agricultural households [49]. Driven by these compensation policies, the enthusiasm of agricultural households to participate in forestry production activities has been greatly improved [50]. This paper also confirms that the forestry compensation policies have a certain degree of positive impact on the forestland transfer of agricultural households. In existing studies, payments for ecosystem services [51,52] and the contingent valuation method [53] are widely used in estimating agricultural households’ compensation.

The empirical research in this paper proves that the forestland transfer has obvious economic benefits, but the forestland transfer also has ecological benefits that cannot be ignored. After the transfer, the forestland that was idle or used for non-forest production can be reforested. Reforestation is considered an important intervention to mitigate global climate change and improve soil conditions. Imam Basuki et al. [54] found that identifying unproductive lands (shrubs, open land) with highly degraded conditions as potential areas for reforestation could effectively reduce greenhouse gas (GHG) emissions in Indonesia. Well site development associated with oil sands exploration is common in boreal mixed wood forests of northern Alberta, Canada and often necessitates reforestation to accommodate other land uses [55]. After transferring in the forestland, agricultural households can choose two reforestation methods: reforestation through assisted natural regeneration or reforestation through planting trees. The former can reduce the consumption of soil organic carbon via soil respiration and is more conducive to the accumulation of biomass carbon [56].

Although this paper provides some valuable findings for forestland transfer, there are inevitably some limitations. Due to the limitations of the research scope and data acquisition, our research is limited to the data of the Liaoning Province and does not involve other regions of China. Therefore, whether the research conclusions can be extended to these regions requires further empirical testing. Further research can attempt to use data from other provinces of China to analyze and compare their connections and differences and to provide a theoretical basis for the development of forestland transfer in different regions.

7. Conclusions and Recommendations

7.1. Conclusions

The main conclusions of this paper are as follows. First, the influencing factors of the choice of different types of agricultural households’ forestland transfer behavior are not exactly the same and the degree of influence by the same variable is also different. Second, forestland transfer can effectively improve the agricultural and forestry income and total household income of pure agricultural households, part-time agricultural households and non-agricultural households and the average treatment effect is greater than 1. Third, there is obvious heterogeneity in the promotion effect of forestland transfer on the income level of agricultural households among different groups of agricultural households. The promotion effect of forestland transfer on the income of pure agricultural households is greater than that of part-time agricultural households. Fourth, because the agricultural and forestry income of non-agricultural households accounts for less than 20%, the average treatment effect of forestland transfer on the agricultural and forestry income of non-agricultural households is significant, but the average treatment effect on the total household income is not significant.

7.2. Recommendations

In order to better revitalize the forestland transfer market and improve the utilization rate of forestland, according to the conclusions of the empirical analysis, this paper proposes targeted forestland transfer development recommendations for pure agricultural households, part-time agricultural households and non-agricultural households.

Agriculture and forestry income are the main sources of income for pure agricultural households and forestland transfer has a significant role in promoting the increase of agricultural and forestry income. Therefore, establishing a stable and efficient forestland transfer market is the top priority to increase the income of pure agricultural households. The establishment and improvement of the forestland transfer market can standardize the transfer process, make the forestland transfer safer, more reasonable and more convenient and provide a guarantee for the agricultural households who carry out the forestland transfer. The establishment of a forestland transfer market firstly requires scientific forestland price assessment agencies. The scientific and reasonable transfer price of forestland can reduce transaction costs and provide convenience for both parties in the transfer of forestland. Secondly, the establishment of forestland transfer information platforms is an important guarantee for the normal operation of the forestland transfer market. Agricultural households who are willing to transfer can obtain transfer information on the forestland transfer information platforms and can also publish transfer information.

The main reason for the emergence of part-time agricultural households is that the input-output ratio of agricultural and forestry management is low and it is easily affected by natural conditions, so it is difficult for agricultural households to obtain stable and high income from agricultural and forestry management. According to the characteristics of part-time agricultural households, there are two improvement measures as follows. First, encourage concurrent agricultural households to join forestry production cooperatives. Forestry cooperatives can provide financial, technical or business channel support for concurrent agricultural households to improve the predicament of agricultural and forestry production. By connecting agricultural households and the market, forestry production cooperatives can not only improve the utilization rate of forestland, but also improve the economic benefits of agricultural households, so as to achieve the common development of ecological and economic benefits. Second, the village collective should increase the publicity of forestry compensation policies and the popularization of forestland transfer procedures, so as to increase the enthusiasm of part-time agricultural households for agricultural and forestry production.

Non-agricultural households focus on non-agricultural industries and pay less attention to forestry production, which often results in idle forestland. Non-agricultural households can take the initiative to flow out of forestland to realize the optimal allocation and rational utilization of forestland resources in a wider range. In the process of outflowing forestland, the village collective can act as a forestland transfer trading platform, providing transfer information for non-agricultural households and, at the same time, can act as an intermediary between the two parties, providing convenience in terms of transfer prices and procedures.