Does Adoption of Biofortification Increase Return on Investment? Evidence from Wheat Farmers in China

Abstract

Biofortification is a new agricultural intervention to alleviate hidden hunger in China and even the world. Exploring the impact of farmers′ adoption of biofortification on return on investment (ROI), which is calculated from farmers′ production net income and profitability, can provide empirical support for the development of sustainable agriculture. This paper examines the impact of the adoption of biofortification on ROI using cross-sectional data from a survey of farmers in China. An endogenous switching regression model that accounts for selection bias was used in the research. The empirical results revealed that the adoption of biofortification has a positive and statistically significant impact on ROI. A heterogeneity analysis also revealed that comparatively high annual income adopters and female adopters tend to benefit more from the adoption of biofortification than middle- and low-income adopters and male adopters.

1. Introduction

Smallholder farmers are usually in a vulnerable position in terms of technology adoption, input use efficiency, and various other uncertainties, which makes it difficult for them to benefit from agricultural production and selling [1,2]. As an efficient agricultural intervention to improve the micronutrient deficiency in the context of global hidden hunger, biofortification (the process of breeding staple crops to have higher levels of essential nutrients, either through selective breeding or genetic modification) can not only improve the nutritional health status of residents [3,4] but can also ensure the performance of farmers in modern agricultural production. Previous studies have verified the positive impact of farmers′ adoption of biofortification on their performance using propensity score matching and found that the adoption of iron biofortified varieties increases average household yield by 22–23% and potential agricultural income by 24–25% [5]. However, there may be some limitations in using income as the sole measure of farmers′ production performance due to the differences in the production and investment capacity of farmers, caused by inconsistent government subsidy policies [6]. Therefore, it is imperative to empirically assess the profitability of investments in the adoption of biofortification by farmers to determine whether they can actually improve health and economic outcomes, especially in low- and middle-income developing countries.

A large body of scholars have conducted research on the adoption of biofortification. However, most studies have focused on the contributing factors of farmers′ adoption of biofortification, including individual and household characteristics of farmers, variety availability [7], farmers′ experience and seed delivery methods [8], farmers′ beliefs [9], yield [10], nutritional attributes [11], nutritional information intervention [12], and so on. Research on the impact of farmers′ adoption of biofortification technology on downstream variables is relatively limited.

Asare-Marfo et al. [13] analyzed the data of iron biofortified soybeans in Rwanda in eight growth cycles involving 19,575 farmers across the country and found that the planting area of iron biofortified soybeans expanded and that the planting structure is moving from traditional varieties to biofortified varieties. Vaiknoras and Larochelle [14] took the iron biofortified bean variety RWR2245 as an example to evaluate the effect of farmers′ adoption of iron biofortified beans on household consumption. The results showed that the adoption of iron biofortified beans could improve the nutritional and health consumption levels of households.

Fortunately, some scholars have explored the impact of farmers′ adoption of biofortification on production performance and believe that it can positively affect yield [15], income, and sales [14]. However, due to the differences in subsidy policies for local farmers in different countries or regions, they were limited to measuring farmers′ production performance by factors such as income or output. The ROI not only measures the net returns of farmers′ production but also considers the profitability of agricultural investment and thus becomes a more effective indicator of farmers′ adoption performance [16,17,18,19]. Previous studies have not identified in a sufficient manner how the farmers′ adoption of biofortification affects ROI.

Therefore, in order to solve the problem of selection bias caused by the systematic differences between observable and unobservable factors faced by farmers in the adoption and non-adoption of biofortified technology, our work estimates the direct marginal effect and average treatment effect of farmers′ adoption of biofortified wheat on the ROI by using an endogenous transformation regression model in order to eliminate the selection bias caused by observable variables and unobservable variables. The present study used data collected in 2020 from a survey of 606 wheat farmers (including wheat biofortification adopters and non-adopters, i.e., conventional wheat farmers) in Henan, Hebei, Gansu, and Shanxi provinces in China.

The remainder of the study is structured as follows: The next section gives an overview of the wheat sector and biofortification in China. The ‘Data and descriptive statistics′ section shows the data and descriptive statistics in our research. The ‘Conceptual basis and empirical specification′ section outlines the theoretical framework and the empirical models employed in the analysis. The ‘Results′ section presents the empirical results and discussions. The final section provides a conclusion and policy implications.

2. Literature Review

2.1. Biofortification and Its Effect

Biofortification, also known as field fortification, is defined as the process of breeding staple crops to have higher levels of essential nutrients, either through selective breeding or genetic modification [3]. Biofortification is considered the most economical and effective method to combat micronutrient deficiency. The nutritional efficacy of biofortification has been proven by scholars. Orange sweet potato rich in vitamin A [20,21], orange corn rich in vitamin A [22], yellow cassava rich in vitamin A [23,24], iron-rich pearl millet [25], and iron-rich soybean [26] have improved the micronutrient intake and deficiency of the target population.

Biofortification also ensures the performance of farmers in modern agricultural production. The existing research on the performance of farmers′ adoption of biofortification mainly focuses on health benefits and economic benefits. For example, Vaiknoras and Larochelle [14] took the iron-fortified bean variety rwr2245 as an example and used the control function method to evaluate the impact of the adoption of iron-fortified beans by farmers on household consumption, output, and soybean planting area. The results showed that, compared with traditional soybean varieties, rwr2245 increased the yield by 20–49% and increased the sales probability of beans by 12%, which provided an opportunity to improve the nutritional status of families purchasing iron-rich soybeans. Mottaleb et al. [15] found that the yield of zinc-rich wheat varieties was 5.2% higher than that of other wheat varieties by applying the ex ante impact assessment framework. Mahboob et al. [27] conducted a randomized controlled trial in remote rural areas of Pakistan to test the knowledge of rural residents on zinc-rich flour and explore the knowledge and attitude of local residents on biofortification. The results showed that the subjects expounded the motivation of obtaining and consuming zinc-rich flour, believed that zinc-rich flour could bring nutrition and health, and promoted the local willingness to adopt zinc-rich wheat.

2.2. Overview of the Wheat Sector and Biofortification in China

China is the biggest country for wheat production, recording a total of 134.25 million MT (17.64% of the world′s total) in 2020, followed by India, Russia, the United States and Canada, who produced 107.59, 85.90, 49.69, and 35.18 million MT, respectively (FAOSTAT). Wheat is one of staple foods in China, mainly in North China (Hebei, Shanxi, and Inner Mongolia), East China (Jiangsu, Anhui, and Shandong), Central China (Henan and Hubei), and Northwestern China (Xinjiang, Shaanxi, and Gansu). In particular, Henan, Shandong, Anhui, Jiangsu, and Hebei covered 72.70% of the total planting area and accounted for 80.20% of the total production in 2020.

With the concern of malnutrition in China, the Chinese government took some actions to fight hidden hunger in 1990s, including supplementation and food fortification [28]. Since 2004, China began to introduce and support biofortification research and development projects, aiming to alleviate malnutrition in rural areas of China [29]. Several biofortified crops have been developed and launched into the market, including high-zinc wheat in North China, high-iron rice in South China, high-beta-carotene sweet potato in Sichuan Province, and high-folate corn in Hebei Province [30].

In our research, we chose a biofortified wheat variety (Zhongmai 175, which was developed through natural screening breeding) to investigate the effect of farmers′ adoption of biofortification on ROI. The biofortified wheat variety (Zhongmai 175) was developed by the Chinese Academy of Agricultural Sciences (CAAS) and was tested on irrigated land and dry land in 2008 and 2011. In addition, Zhongmai 175 has been approved by the Ministry of Agriculture of China and become the main cultivated variety in Henan, Shanxi, Shaanxi, and Gansu.

3. Data and Descriptive Statistics

3.1. Data Sources

The data employed in the present study came from a farmer survey that was conducted from July to August 2020 in four different provinces (Henan, Hebei, Gansu, and Shanxi). Stratified random sampling in five stages was used. First, the main planting provinces with biofortified wheat were purposefully selected using information from the Chinese Academy of Agricultural Sciences (CAAS). In the second stage, four county-level districts where biofortified wheat is intensively produced at the provincial level were chosen. These include Sanmenxia city in Henan, Gaobeidian city in Hebei, Pingliang, Qingyang and Tianshui cities in Gansu, and Pinglu and Jiexiu counties in Shanxi. Third, around 2–4 towns were randomly selected from those districts using information provided by the local agricultural bureau, the planting corporation, and dealers of biofortified wheat. Fourth, around 2–6 villages in each selected town were randomly selected. Finally, around 6–20 wheat farmers, including both biofortified wheat farmers and conventional wheat farmers in each village, were randomly selected, resulting in a total of 672 samples. In total, 352 biofortified wheat farmers and 254 conventional wheat farmers were interviewed. The collected data include information on wheat inputs and outputs (e.g., input use, costs, yields, and output price), and legitimacy perception as well as household and farm-level characteristics (e.g., age, gender, education, and household labor force).

3.2. Descriptive Statistics

Table 1. Definition and summary statistics of selected variables.

Variable	Description	Mean	S.D.
Adoption	1 if farmer grew biofortified wheat, 0 otherwise	0.5809	0.4938
ROI	ROI of growing wheat (%)	1.1050	1.2016
Age	Age of farmer	54.2574	10.1791
Gender	Gender of farmer: 1 if male, 0 otherwise	1.2327	0.4229
Planting area	Total growing wheat (mu a)	6.5693	9.7808
Education	Farmer′s maximal education level (years)	7.9455	3.2497
Household labor force	Number of members with working capacity in household	2.8366	1.1845
Government service extension programs	1 if farmer recently participated in the government service extension programs for wheat, 0 otherwise	0.4191	0.6106
Information availability	1 if farmer had access to information about biofortified wheat, 0 otherwise	0.2145	0.4108
Membership	1 if farmer belonged to agricultural cooperatives, 0 otherwise	0.6930	0.2542
Risk attitude	1 if farmer can accept the risks of planting new varieties, 0 otherwise	0.4389	0.4967
Experience	1 if farmer used to plant biofortified wheat, 0 otherwise	0.7607	0.4270
Legitimacy	The government should formulate laws and regulations to ensure the extension of biofortified wheat. The government should provide support in the extension of biofortified wheat. 1 = Strongly disagree; 5 = Strongly agree).	4.3597	0.8875

^a 1 mu = 1/15 hectare.

presents the definition and summary statistics of the variables used in the analysis. The independent variable used in the study is a dummy variable that takes the value of one if the farmer adopted biofortification technology and the value zero if the farmer did not adopt biofortification technology. The outcome variable used in the study is ROI. It can be observed from Table 1 that about 58% of farmers had grown biofortified wheat and that about 76% of farmers had experience in growing biofortified wheat. The average age of farmers was almost 54 years. The average education year was about 7.95 years. The average planting area of wheat was 6.57 mu. The proportions of farmers who had participated in government service extension programs or belonged to agricultural cooperatives were 41.91% and 69.30%, respectively. Only 21.45% of farmers believed that they had access to information about biofortified wheat. Most farmers believed that government and legal support should be provided in the adoption of wheat biofortification.

The mean differences in the characteristics of biofortified wheat farmers and conventional wheat farmers are presented in

Table 2. Mean differences in characteristics between biofortified wheat farmers and conventional wheat farmers.

Variable	Biofortified Wheat Farmers (352)	Conventional Wheat Farmers (254)	Diff.	t-Statistic
Age	53.2443 (10.4697)	55.6614 (9.6067)	2.4171 **	2.9019
Gender	1.2585 (0.4384)	1.1969 (0.3984)	−0.0617 *	−1.7746
Planting area	6.1835 (6.2837)	7.1039 (13.1716)	0.9204	1.1433
Education	8.1534 (3.2563)	7.6575 (3.2249)	−0.4959 *	−1.8574
Household labor force	2.8892 (1.1605)	2.7638 (1.2156)	−0.1254	−1.2869
Government service extension programs	0.4915 (0.6493)	0.3189 (0.5378)	−0.1726 ***	−3.4644
Information availability	0.2358 (0.4251)	0.1850 (0.3891)	-0.0508	−1.5022
Membership	0.0909 (0.2879)	0.0394 (0.1949)	−0.0515 **	−2.4732
Risk attitude	0.4915 (0.5006)	0.3661 (0.4827)	−0.1253 **	−3.0867
Experience	0.9716 (0.1664)	0.4685 (0.5000)	−0.5031 ***	−17.5816
Legitimacy	4.3509 (0.8497)	4.3720 (0.9390)	0.0212	0.2898

Notes: Standard errors in parentheses; *** p < 0.01; ** p < 0.05; * p < 0.1.

. There were significant differences in the characteristics of age, gender, years of education, participation in government service extension programs, belonging to agricultural cooperatives, risk attitude, and experience between farmers who grew biofortified wheat and conventional wheat. Other household and farm-level characteristics, such as planting area, information availability, household labor force, and legitimacy perception, hardly differed between farmers who adopted biofortified wheat and those who grew conventional wheat.

The mean differences in ROI and its components between biofortified wheat farmers and conventional wheat farmers are presented in the

Table 3. Mean differences in production costs and profit between biofortified wheat farmers and conventional wheat farmers.

Variable	Mean (606)	Mean of Biofortified Wheat Farmers (352)	Mean of Conventional Wheat Farmers (254)	Diff.
Total income (CNY a/mu)	821.7286	900.5898	712.4408	−188.149 ***
Sales (kg)	2596.6130	2514.8295	2709.9510	195.1215
Sale price (CNY/kg)	2.1610	2.2644	2.0178	−0.2466 ***
Seed costs (CNY/mu)	68.2259	69.3076	66.7268	−2.5809
Pesticide costs (CNY/mu)	26.6471	27.1986	25.8830	−1.3156
Fertilizer costs (CNY/mu)	138.7529	140.6114	136.1774	−4.4339
Machinery costs (CNY/mu)	132.1953	134.1986	129.419	−4.7797
Employment costs (CNY/mu)	3.5068	3.4745	3.5516	0.0771
Irrigation costs (CNY/mu)	45.1174	37.0887	56.2437	19.1550 ***
Other costs (CNY/mu)	2.4651	2.0540	3.0349	0.9809
Total costs (CNY/mu)	416.9105	413.9334	421.0363	7.1029
Net returns (CNY/mu)	404.8181	396.5507	416.2753	19.7246
ROI	1.1050	1.3475	0.7690	−0.5784 ***

^a CNY is a Chinese currency unit (USD 1 = CNY 6.37). Notes: Total income is actual sales income; *** p < 0.01.

. As evident from the Table, the average income, sales price, and ROI of biofortified wheat farmers were significantly higher than those of conventional farmers, and the excess ratios were 26.41%, 12.22%, and 75.23%, respectively. The irrigation costs of biofortified wheat were significantly lower than those of conventional wheat at the proportion of 34.06%. In addition, the average values of seed costs, pesticide costs, fertilizer costs, and machinery costs of biofortified wheat were slightly higher than those of conventional wheat. However, the differences were not significant, and the total costs of biofortified wheat were slightly lower than those of conventional wheat. In addition, there was no significant difference in net returns. The potential reasons for these differences are as follows: First, biofortified wheat farmers enjoyed the policy benefits (e.g., subsidies for production materials of biofortified crops, mainly referring to the deduction of the cost of seeds for biofortified agricultural products), which led to this situation where parts of growing biofortified wheat cost more than growing conventional wheat. However, the total cost of growing biofortified wheat was lower. Second, farmers′ production behavior (e.g., fertilization) was basically not affected by wheat varieties. Therefore the difference in net returns between biofortified wheat farmers and conventional wheat farmers was small.

4. Conceptual Basis and Empirical Specification

4.1. Theoretical Framework

The conceptual framework employed here is based on the assumption that farmers derive the maximum net returns from wheat production and sales [2]. In order to analyze conveniently, $Y_i$ represents the total household wheat yield, $P_i$ represents the price of wheat$,I_{ik}$ represents the vector of the input variable, $R_{ik}$ represents the corresponding price of the input variable, $k$ represents input type $m$, such as the fertilizer, pesticide, employment, irrigation, and machinery costs required for wheat production, $L_i$ represents the planting area for wheat, $X_i$ represents the vector of variable individual characteristics (such as age, education, and experience) that may affect the net returns of wheat production, and $Z_i$ represents the adoption of wheat biofortification. Therefore, the net returns function can be expressed as:

$$\pi \left(P,R,L;X,Z\right)=P_i{Y}_i-{{\displaystyle \sum }}_{k=1}^{k=m}{I}_{ik}{R}_{ik},k=1,2,\dots ,m$$

(1)

Considering that $L_i$ is a fixed input with homogeneity, Equation (1) can be expressed as:

$$\pi \left(P,R,L;X,Z\right)=L_i·\tilde{\pi }\left(P,R,L;X,Z\right)$$

(2)

where $\tilde{\pi }\left(P,R,L;X,Z\right)$ represents the net returns per unit of the planting area, which is defined as the difference between the unit output value and the input costs of wheat. Therefore, the wheat yield and the input per unit of the planting area can be expressed respectively as (3) and (4), which can directly be expressed as (5) and (6), respectively:

$$\widetilde{Y_i}=\frac{Y_i}{L_i}$$

(3)

$$\widetilde{I_i}=\frac{I_i}{L_i}$$

(4)

$$\widetilde{Y_i}=\frac{\partial \tilde{\pi }\left(P,R,L;X,Z\right)}{\partial P_i}$$

(5)

$$\widetilde{I_i}=\frac{\partial \tilde{\pi }\left(P,R,L;X,Z\right)}{\partial R_i}$$

(6)

Equations (2), (4) and (5) show that the net returns of wheat, calculated by yield, yield per mu, and input costs per mu of wheat, are affected by the input and output prices of wheat, household and farm endowment, and farmers′ adoption. Specifically, the adoption of biofortification (Z) has an influence on the net returns of wheat ($\tilde{\pi }$) by directly affecting the output value ($\tilde{Y}$) and input costs ($\tilde{I}$) per unit of planting areas.

4.2. Empirical Model

The purpose of our work is to examine the impact of the adoption of biofortification on ROI. ROI is the preferred index to measure farmer performance because it links the net returns with the input costs of wheat [15,16]. The empirical study on the impact of the adoption of wheat biofortification on ROI is based on the premise that ROI is affected by the individual and household characteristic vector ($X_i$) and the adoption of biofortification ($Z_i$). The model is specified as:

$$V_i=\beta X_i+\delta Z_i+{\epsilon }_i$$

(7)

where $V_i$ is the ROI of farmer $i$, $\beta$ and $\delta$ represent the estimated parameter, ${\epsilon }_i$ is the random disturbance term, and $Z_i$ represents the adoption of wheat biofortification. The ROI is:

$$ROI=\frac{Profit}{Investment}$$

(8)

Based on the previous research [2], our work adopts a random utility framework to model the adoption of wheat biofortification by farmers. If the expected incomes obtained by adopting biofortification ($U_i^m$) are greater than the expected incomes obtained without adopting biofortification ($U_i^n$), that is, $Z_i^{*}=U_i^m-U_i^n>0$, then $Z_i^{*}$ represents the difference between the expected incomes from adopting or not adopting wheat biofortification. However, $Z_i^{*}$ cannot be observed but can be expressed as a function of observable elements in the following latent variable model:

$$Z_i^{*}=\lambda D_i+u_i,Z_i=\{\begin{array}{c}1,Z_i^{*}>0\\ 0,Z_i^{*}\le 0\end{array}$$

(9)

where $\lambda$ is a vector of the parameters to be estimated, $u_i$ is an error term, and $D_i$ represent the factors affecting the adoption of wheat biofortification.

4.3. Treatment Effect Model

Assuming that $u_i$ follows a binary normal distribution, the ROI of the adoption decision can be calculated as follows:

$$E\left(V_i|Z=1\right)=\beta X_i+\delta +E\left({\epsilon }_i|Z=1\right)=\beta X_i+\delta +{\rho }_{\epsilon u}{\sigma }_{\epsilon u}\frac{\phi \left(\lambda X_i\right)}{\varnothing \left(\lambda X_i\right)}$$

(10)

$$E\left(V_i|Z=0\right)=\beta X_i+E\left({\epsilon }_i|Z=0\right)=\beta X_i-{\rho }_{\epsilon u}{\sigma }_{\epsilon u}\frac{\phi \left(\lambda X_i\right)}{1-\varnothing \left(\lambda X_i\right)}$$

(11)

Then, the average treatment effect (ATE) of the adoption of biofortification on ROI can be expressed as:

$$ATE=\frac{1}{N}{\displaystyle \sum }_{i=1}^N\left[E\left(V_i|Z=1\right)-E\left(V_i|Z=0\right)\right]$$

(12)

5. Results

5.1. Effect of Adoption of Wheat Biofortification on ROI

The results of the endogenous switching regression (ESR) model are presented in

Table 4. Estimation results of ESR for adoption of biofortification and impact of adoption on ROI.

Variable	Selection Equation		ROI
			Biofortified Wheat Farmers		Conventional Wheat Farmers
	Coefficient (1)	Std. Err. (2)	Coefficient (3)	Std. Err. (4)	Coefficient (5)	Std. Err. (6)
Age	−0.0078	0.0062	−0.0030	0.0070	−0.0002	0.0076
Gender	−0.0653	0.1407	−0.0372	0.1611	−0.3242 *	0.1728
Planting area	−0.0005	0.0074	0.0155	0.0107	0.0112 **	0.0053
Education	0.0002	0.0188	−0.0184	0.0223	−0.0134	0.0219
Household labor force	0.0334	0.0510	0.0366	0.0613	0.0929	0.0576
Government service extension program	0.0904	0.1022	0.1225	0.1102	−0.0721	0.1344
Information availability	−0.0664	0.1449	0.0638	0.1650	0.5442 **	0.1899
Membership	0.2635	0.2345	0.7866 **	0.2470	0.1320	0.3599
Risk attitude	0.3416 **	0.1195	0.3067 **	0.1408	−0.0513	0.1536
Experience	2.1391 ***	0.1770	2.5075 ***	0.4250	0.3493	0.3641
Legitimacy	−0.1436 **	0.0656
Constant	−0.6339	0.5686	−1.7301 **	0.7277	0.7421	0.5776
${\mathit{\rho }}_1$			0.7971 ***	0.0761
$\ln {\mathit{\sigma }}_1$			0.3136 ***	0.0605
${\mathit{\rho }}_2$					0.0482	0.2961
$\ln {\mathit{\sigma }}_2$					0.0742 *	0.0449
LR	7.00 **
Log likelihood	−1214.1387
Sample size	606		352		254

Notes: *** p < 0.01; ** p < 0.05; * p < 0.1.

. Columns (1) and (2) present the estimated coefficients and standard errors of the selection equation, while the outcome equations are presented in columns (3) to (6). ${\rho }_1$ and ${\rho }_2$, respectively, represent the correlation coefficients of the error terms between the adoption decision and the impact of adoption on ROI. The results show that ${\rho }_1$ is significant at the 1% level, but ${\rho }_2$ is not, which indicates that there is a self-selection problem with the samples. Whether farmers adopt wheat biofortification is the result of the combined effect of observable and unobserved factors, namely, farmers′ adoption of biofortification is not randomized, which is a self-selection bias based on the farmers′ own utility before and after adoption. If it is not corrected, the estimation results will be biased. According to the positive and negative values of ${\rho }_1$ and ${\rho }_2$ as well as their significance levels, farmers′ adoption of wheat biofortification has negative selection bias, namely, farmers whose ROI values are lower than the average level are more likely to adopt wheat biofortification.

From Table 4, we can confirm that farmers′ risk attitude, experience, and perception of legitimacy significantly affect the adoption of wheat biofortification. Farmers who are willing to take the risks of wheat biofortification, have the experience of planting biofortified wheat, and have low requirements on the standardization and legitimacy of biofortified wheat are more willing to adopt wheat biofortification. In addition, our results also reveal that risk attitude, experience, and belonging to agricultural cooperatives significantly increase the ROI of adopting wheat biofortification, while gender, planting area, and information availability significantly affected the ROI of growing conventional wheat. These results indicate that farmers who are willing to take the risks of biofortified wheat varieties, have the experience of planting biofortified wheat, and have belonged to agricultural cooperatives are more likely to adopt biofortified wheat, and their ROI is higher.

5.2. Treatment Effect Analysis of Adoption of Wheat Biofortification on ROI

The estimates for the average treatment effects on the treated (ATT), which show the effects of the adoption of wheat biofortification on ROI, are presented in

Table 5. The average treatment effects of the adoption of wheat biofortification on ROI.

Farmer Category	Adoption of Biofortified Wheat	Adoption of Conventional Wheat	ATT	ATU
Biofortified wheat farmers	(a) 1.3541	(c) 0.7772	0.5769 ***
Conventional wheat farmers	(b) 0.7745	(d) 0.9991		0.2245

Notes: ATT and ATU, respectively, represent the average treatment effect of the adoption of wheat biofortification or not. *** p < 0.01.

. (a) and (b), respectively, represent the expected ROI values of growing biofortified wheat for farmers who actually adopt biofortified wheat or not. (c) and (d) are the expected ROI values under the counterfactual circumstance. Specifically, (c) represents the expected ROI of farmers who grew biofortified wheat originally and intended not to grow biofortified wheat, while (d) represents the expected ROI of farmers who did not grow biofortified wheat and intend to grow biofortified wheat. The findings in Table 5 reveal that the average treatment effect of the adoption of wheat biofortification on ROI is significant, which implies that the adoption of wheat biofortification has a positive treatment effect on ROI. The ATT estimates show that the adoption of wheat biofortification significantly increases ROI by 42.60% when it is against the adoption of conventional wheat, which mainly comes from the policy support such as subsidies and incentives expanding the income–cost advantage of biofortified wheat farmers compared with conventional wheat farmers. This is similar to the research conclusions of Ma [2] and other scholars, which all indicate that the sample has self-selection bias.

The robustness of the average treatment effects of the adoption of wheat biofortification on ROI was checked by using a non-parametric technique, i.e., propensity score matching (PSM), reported in detail in

Table 6. The average treatment effects of the adoption of wheat biofortification on ROI for the PSM.

PSM	Adoption of Biofortified Wheat	Adoption of Conventional Wheat	ATT	t-Statistic	Change (%)
Matching (1 vs. 1)	1.3541	0.7772	0.5769	3.03 ***	42.6039
Matching (1 vs. 4)	1.3541	0.8935	0.4606	3.10 ***	34.0152
Radius	1.3559	0.8243	0.5316	3.67 ***	39.2064

Notes: *** p < 0.01.

. The results reveal that the treatment effect of the adoption of wheat biofortification on the ROI is robust.

5.3. Heterogeneity Analysis of ROI for Adopting Wheat Biofortification

To further explore the difference in the average treatment effect of adopting wheat biofortification on ROI, our work analyzed the different ROI values of adopting wheat biofortification under different individual characteristics. The results in

Table 7. The average treatment effects of the adoption of wheat biofortification on ROI.

	Average Treatment Effects		t-Statistic	ATT	Change (%)
	Adoption of Biofortified Wheat	Adoption of Conventional Wheat
The average treatment effect on ROI for annual household income
ROI (Annual household income ≤ Mean)	1.2540	0.8615	1.97 *	0.3925	31.2998
ROI (Annual household income > Mean)	1.4898	0.8121	2.84 ***	0.6777	45.4893
The average treatment effect on ROI for gender
ROI (Male)	1.3542	0.9227	2.27 **	0.4315	31.8638
ROI (Female)	1.3287	0.7417	1.71 *	0.5870	44.1785
The average treatment effect on ROI for belonging to agricultural cooperatives
ROI (Belong to agricultural cooperatives)	1.9486	0.7240	2.28 **	1.2246	62.8451
ROI (Do not belong to agricultural cooperatives)	1.2881	0.8180	2.36 **	0.4701	36.4956
The average treatment effect on ROI for perceived risk
ROI (Presence of perceived risk)	1.4130	0.6748	2.91 ***	0.7382	52.2435
ROI (Absence of perceived risk)	1.2799	0.8838	1.92 *	0.3961	30.9477

Notes: *** p < 0.01; ** p < 0.05; * p < 0.1.

suggest that the ROI of biofortified wheat farmers was always higher than that of conventional wheat farmers, considering farmers′ annual household income, gender, and risk perception as well as whether farmers belong to agricultural cooperatives. Specifically, compared with conventional wheat farmers, the ROI of biofortified wheat farmers with annual household incomes higher than the sample mean were higher than those with annual household incomes lower than (including equal to) the sample mean by 14.19%. Female adopters had a higher ROI than male adopters by 12.31%. The biofortified wheat farmers who belonged to agricultural cooperatives had higher ROI values than those who did not belong to agricultural cooperatives by 26.35%. Our results also reveal that, compared with conventional wheat farmers, the ROI of biofortified wheat farmers who perceived risks was higher than that of those who do not perceive risk by 21.30%. This is consistent with the conclusion of Ma [2], and participation in agricultural cooperatives is expected to improve the return on investment of farmers. A potential reason is that farmers who perceived risks may take some measures (such as improving their planting capability and standardizing pesticides inputs) to avoid or reduce the potential risks from adopting wheat biofortification, which may also contribute to their higher ROI.

6. Conclusions and Policy Implications

Whether the adoption of biofortification can improve farmers′ production performance in developing countries remains a contentious issue. Our research contributes to the empirical literature on the issue by finding the determinants of the adoption of biofortification and revealing the positive effect of the adoption on ROI. The study utilized farm-level data of Chinese wheat farmers collected from Henan, Hebei, Shanxi, and Gansu provinces in 2020 from a randomly selected sample of 606 households.

The empirical results showed that there was a selection bias in the influence of the adoption of wheat biofortification on farmers′ ROI, namely, the farmers whose ROI were below the average ROI were more willing to adopt wheat biofortification, and the adoption could increase wheat farmers′ ROI by 42.60% when it was against the adoption of conventional wheat. Second, we found that farmers′ experience of planting biofortified wheat, the risk perception of biofortification, and the perception of legitimacy were the determinants of adoption and affected farmers′ adoption of wheat biofortification significantly.

The heterogeneity analysis in our research found that the ROI from biofortified wheat farmers was consistently higher than that of conventional wheat farmers under different individual characteristics (such as annual income of household, gender, engagement in agricultural cooperatives, and the risk perception of biofortification). Specifically, compared to conventional wheat farmers, adopters whose annual income of household was above the average annual income, female adopters, adopters who engaged in agricultural cooperatives, and adopters who had a risk perception of wheat biofortification had a comparatively higher ROI values by 14.19%, 12.31%, 26.35%, and 21.30%, respectively.

The results of our research have several implications for wheat farmers and policy makers in China. First, the finding that the adoption of biofortification can not only improve the nutritional health of the population in rural areas but also improve wheat farmers′ production performance (ROI) suggests that it is necessary to launch biofortification research and development and promotion projects. Second, our findings about the determinants of the adoption of biofortification suggest that government ought to take some measures to reduce farmers′ risk perception of biofortification, strengthen farmers′ perception of legitimacy, and guide farmers to plant biofortified crops in order to further increase farmers′ adoption of biofortification. Third, the results of the heterogeneity analysis suggest that government should actively encourage the farmers who have a comparatively high annual income or have engaged in agricultural cooperatives to grow biofortified crops and increase the ROI of adoption to lead by example.