The theory of planned behavior

Most previous related research has been carried out in the framework of the TPB (Yanakittkul and Aungvaravong, Reference Yanakittkul and Aungvaravong2020). Several later studies extended the TPB and combined it with the diffusion of innovation theory (Doanh et al., Reference Doanh, Quynh and Pham2022b) and the norm activation model (Nguyen et al., Reference Nguyen, Doan, Nguyen and Nguyen2021). Proposed by Ajzen (Reference Ajzen1991), TPB explains behavioral intention through three factors, including attitude (ATT), subjective norm (SN) and perceived behavior control (PBC). Accordingly, ATT reflects the degree to which a person rates a behavior favorably or unfavorably. The more favorable the individual's assessment is, the more likely he or she is to perform the behavior (Borriello et al., Reference Borriello, Massey and Rose2022). Many studies have confirmed the positive influence of ATT on intention. For example, Yanakittkul and Aungvaravong (Reference Yanakittkul and Aungvaravong2020) found that ATT promotes farmers' intention to grow organic rice. Likewise, Lapple and Kelley (Reference Lapple and Kelley2014) also documented that ATT influences Irish farmers' intention to convert from traditional to organic farming. Based on the above evidence, we hypothesize that the better the farmer's assessment of the conversion to organic farming, the more likely he or she is to convert from traditional to organic farming.

SN describes an individual's perception of external pressures (e.g., pressure from society, relatives and neighbors) when performing a particular behavior (Issa and Hamm, Reference Issa and Hamm2017). According to Yanakittkul and Aungvaravong (Reference Yanakittkul and Aungvaravong2020), the more favorable an individual's perception of pressure is, the more likely he or she will carry out his or her intention. Therefore, the higher the SN, the more farmers intend to convert from traditional farming to organic farming. Nguyen et al. (Reference Nguyen, Doan, Nguyen and Nguyen2021) showed that SN positively influences farmers' intention toward organic farming. A positive correlation between SN and intention was also found in the study of Govindharaj et al. (Reference Govindharaj, Gowda, Sendhil, Adak, Raghu, Patil and Damalas2021), who showed that SN plays a vital role in shaping farmers' behavior. Therefore, this study hypothesizes that the higher the SN, the more farmers intend to convert from traditional farming to organic farming.

The PBC reflects the perceived difficulty of performing a particular behavior (Ajzen and Madden, Reference Ajzen and Madden1986). According to Shukri et al. (Reference Shukri, Jones and Conner2022), PBC is a multi-structural factor that consists of two main parts: perception of ease (e.g., the individual's perception of how difficult or easy it is when he or she performs the intention) and self-efficacy (e.g., the perception of the degree of confidence when he or she implements the intention). Yanakittkul and Aungvaravong (Reference Yanakittkul and Aungvaravong2020) indicated that PBC plays a vital role in shaping farmers' organic conversion. The positive impact of PBC on intention was also found in the study of Jones et al. (Reference Jones, Sok, Tranter, Blanco-Penedo, Fall, Fourichon and Sundrum2016) and Yanakittkul and Aungvaravong (Reference Yanakittkul and Aungvaravong2020). Therefore, this study hypothesizes that the higher the PBC, the more farmers intend to convert from traditional to organic farming.

The theory of HB and the NE

Herd behavior can be defined as the phenomenon in which the individual decides to follow others and imitate the group's behaviors rather than deciding independently based on his or her own information (Baddeley, Reference Baddeley2010). In society, philosophers consider HB the practice of individuals mimicking the behavior of crowds because they consider it the safest behavior. This imitation is reflexive and does not require the intervention of reason (Sun, Reference Sun2013).

There are two main conditions under which HB can occur: uncertainty about the decision and observing others' actions (Sun, Reference Sun2013; Youssef, Reference Youssef2020). In terms of the first condition, people tend to follow the crowd's behavior when uncertain about the decision they will make due to a lack of information or asymmetric information (Pavlović-Höck, Reference Pavlović-Höck2022). Regarding the second condition, observing many people making the same decisions is also a condition that motivates individuals to perform HB. Here, there are two important factors involved in influencing such observations. The first factor is the number of people who have performed the behavior before. According to Sun (Reference Sun2013), the number of people who have performed a specific behavior plays an essential role in forming the intention and behavior of the individual. Accordingly, the individual will tend to perform any behavior if many others have performed the same behavior before. The second factor is the identity of those who performed the previous act. An individual is likely to perform the same behavior as his predecessor if he considers his predecessor to be successful (Bandura, Reference Bandura1986) or reputable (Abrahamson, Reference Abrahamson1991).

Thus, the NE may affect tea farmers' intention to convert to organic practice. Specifically, the traditional tea farmer (the individual) observes other people (other farmers) who have converted to organic tea production. Based on these observations, traditional tea farmers decide to convert to organic tea production even if the information obtained suggests that they should not convert to organic tea production. In other words, if more neighbors and people who have a close relationship with traditional tea farmers have converted to organic tea production, the more traditional tea farmers intend to convert to organic tea production.

Here it is essential to distinguish SN and NE because they have similarities and differences. SN and NE are similar in that they both refer to the influence of intimate relationships (such as family, neighbors and close friends) on the subject's intentions. However, SN and NE differ in the following points. First, SN focuses on the influence of the thoughts of intimate people on the subject's intentions (e.g., my closest family members think that I should develop organic farming, my closest friends think that I should develop organic farming and people who are important to me think that I should develop organic farming). In this case, the subject's intentions appear to be passive. In contrast, in the case of NE, the subject's intention comes from the subject's self-observation of its intimate people. In this case, the subject's intention seems proactive (the subject actively observes and follows its intimate people). Secondly, in the case of SN, the subjects' intimate people are not necessarily those who have converted to organic tea or have succeeded in organic tea farming. In contrast, in the case of NE, the subjects' intimate people are those who have intended or have successfully converted from traditional to organic tea farming.

Several previous studies have shown that individuals can change their beliefs by observing the actions of others. Accordingly, individuals can ignore their initial beliefs and form new beliefs about usefulness after observing the actions of their predecessors (Sun, Reference Sun2013). Basically, all three factors (ATT, SN and PBC) are formed based on the perception and beliefs of the individual (Issa and Hamm, Reference Issa and Hamm2017; Romero-Colmenares and Reyes-Rodríguez, Reference Romero-Colmenares and Reyes-Rodríguez2022). Therefore, this study hypothesizes that the NE can positively affect ATT, SN and PBC.

Besides such factors as ATT, SN, PBC and NE, we also incorporate the demographic characteristics of tea farmers, including gender, age, education level and culture.

Model of estimation

According to Figure 1, our research model consists of two main structures. The first structure is the direct impact of ATT, SN, PBC and NE on INT. The second structure is the indirect impact of NE on INT through ATT, SN and PBC. Previous studies often used structural equation modeling (SEM) in this case. The strength of this model is its powerful flexibility to jointly test a variety of hypotheses that involve different types of complicated cause–effect relationships (Katebi et al., Reference Katebi, Homami and Najmeddin2022). However, the SEM's assumption (all responses are continuous) exposes users to severe biases, especially if the dependent variable is a dummy variable (Hieu et al., Reference Hieu, Van and Doanh2021). Therefore, following Hieu et al. (Reference Hieu, Van and Doanh2021) and Katebi et al. (Reference Katebi, Homami and Najmeddin2022), we apply generalized SEM (GSEM), which is a combination of the generalized linear model and SEM (Zhang and Zhang, Reference Zhang and Zhang2018). Combining the two models makes it possible to handle multiple structures with different types of response variables (ATT, SN and PBC are continuous response variables, while INT is a non-continuous response variable). Basically, the GSEM model has the following form:

(1)$$g\;\{ {{\rm E\;}( {y{\rm \vert }X} ) } \} = {\rm x}\beta $$

$$y{\rm \;}\sim {\rm F}$$

where y is the dependent variable, X is the explanatory variable(s), β is the coefficient of X, g(.) is called the link and F is called the family.

Fig. 1. Factors affecting farmers' intention to convert to organic tea farming.

Regarding the first structure, we will use the Bernoulli family with a logit link function to evaluate the direct impact of ATT, SN, PBC and NE on INT. Specifically, the mathematical equation is as follows:

Bernoulli family:

(2)$$\log f\;( {INT{\rm \vert }\mu } ) = INT\log \mu + ( {1-INT} ) \log ( {1-\mu } ) $$

and Logit link:

$$g( \mu ) = \log \mu -{\rm log}( {1-\mu } ) $$

$$\mu = g^{{-}1}( z ) = \displaystyle{1 \over {1 + e^z}}$$

where INT takes the value of 1 if the farmer intends to convert to organic farming and 0 if he or she does not. μ is the probability that a farmer intends to convert from traditional to organic farming.

Our second structural model estimates the impact of NE on ATT, SN and PBC. We use the Gaussian family with an identity link for this structure. Specifically, the mathematical equations are as follows:

Gaussian family and identity link of response are:

ATT

(3a)$$\eqalign{& \log f( {ATT{\rm \vert }\mu_{ATT}, \;{\rm \Sigma }_{ATT}} ) = {-}\displaystyle{1 \over 2}\{ {d_{ATT}\log 2\pi + \log \vert {\;{\rm \Sigma }_{ATT}} \vert }\cr & + ( {ATT-\mu_{ATT}} )^{\prime}{\rm \Sigma _{ATT}^{{-}1} ( {ATT-\mu_{ATT}} ) } \} \; g\;( {\mu_{ATT}} ) = \mu _{ATT}} $$

SN

(3b)$$\eqalign{& \log f\;( {SN{\rm \vert }\mu_{SN}, \;\;{\rm \Sigma }_{SN}} ) = {-}\displaystyle{1 \over 2}\{ {d_{SN}\log 2\pi + \log \vert {\;{\rm \Sigma }_{SN}} \vert }\cr & + ( {SN-\mu_{SN}} )^{\prime}{\rm \Sigma _{SN}^{{-}1} ( {SN-\mu_{SN}} ) } \} \;g\;( {\mu_{SN}} ) = \mu _{SN}} $$

PBC

(3c)$$\eqalign{& \log f( {PBC{\rm \vert }\mu_{PBC}, \;{\rm \Sigma }_{PBC}} ) = {-}\displaystyle{1 \over 2}\{ {d_{PBC}\log 2\pi + \log \vert {\;{\rm \Sigma }_{PBC}} \vert }\cr & + ( {PBC-\mu_{PBC}} )^{\prime}{\rm \Sigma _{PBC}^{{-}1} \;( {PBC-\mu_{PBC}} ) } \} \; g\;( {\mu_{PBC}} ) = \mu _{PBC}} $$

where μ _ATT, μ _SN and μ _PBC are the means of ATT, SN and PBC, respectively. d _ATT, d _SN and d _PBC are the dimensions of the observed response vectors ATT, SN and PBC, respectively. Σ indicates the variance matrix of unexplained errors. In which, Σ_ATT is the variance matrix of ATT's unexplained errors, Σ_SN is the variance matrix of SN's unexplained errors and Σ_PBC is the variance matrix of PBC's unexplained errors.

Following Hieu et al. (Reference Hieu, Van and Doanh2021) and Katebi et al. (Reference Katebi, Homami and Najmeddin2022), we use Akaike's information criterion (AIC) and the Bayesian information criterion (BIC) to confirm the effectiveness of the GSEM. Two indexes are calculated as follows:

$$AIC = -2\ln ( L ) + 2p$$

$$BIC = -2\ln ( L ) + p\ln ( n ) $$

where L is the likelihood function, p is the number of parameters and n is the size of the sample.

In summary, we take the following steps to estimate the impact of variables on farmers' intention to convert to organic farming. First, following Doanh et al. (Reference Doanh, Do Dinh and Quynh2022a), we use exploratory factor analysis (EFA) to check the factor structure and confirmatory factor analysis (CFA) to check the reliability, validity and measurement model fit. Here, we use the goodness of fit (GOF) indices to check the efficiency of the model, including model fit chi-squared test (χ²), comparative fit index (CFI), Tucker–Lewis index (TLI) and root mean squared error of approximation (RMSEA). Secondly, we use STATA-17 software to simultaneously estimate the structures in the model (see Fig. 1) with the GSEM command. GSEM estimates the impact of control variables on INT, and hypotheses H₁, H₂, H₃ and H₄ with Bernoulli family and logit link; and hypotheses H₅, H₆ and H₇ with Gaussian family with an identity link.

Variables and measurement

Under the framework of the TPB, farmers' intention to convert from traditional to organic farming was measured by the question: ‘Do you want to convert to organic farming?’ (1 = I want to convert; 0 = I don't want to convert to organic farming). This study measures the structures of the TPB based on the work of Hieu et al. (Reference Hieu, Van and Doanh2021) and Solesvik (Reference Solesvik2013). ATT is measured through five questions, SN is measured through six questions and PBC is measured through four questions. Details of items are given in Table A1Footnote ⁶. Items are accompanied by a response option ranging from 1 (strongly disagree) to 5 (strongly agree). Higher scores reflect higher levels in the measured constructs.

Concerning NE, this study measures it in two ways, including (1) using a 6-item questionnaire (Shen et al., Reference Shen, Zhang and Zhao2016) and (2) measuring by cases. According to Manski (Reference Manski1993) and McCartney and Shah (Reference McCartney and Shah2022), there are two problems in measuring NE: endogenous and correlated. In this study, these two problems stem from the following reasons. First, farming households often choose to live and participate in activities near households with similar lifestyles, incomes or bloodlines. Because of these similarities, they may have made the same decision to convert to organic farming. Secondly, even though they do not live close to each other, farmers can still communicate with their relatives (e.g., blood relatives and friends) in other areas, thanks to the development of technology. Therefore, they may face similar farming problems.

Based on the study by McCartney and Shah (Reference McCartney and Shah2022), we randomly selected farm households adjacent to the subject under study to solve the above problem. In addition, this study determines the NE across three cases, including:

1. (1) The converting status of farming households living close to the subject under study (the individual) within a radius of 500 m;

2. (2) The converting status of farming households that have close relationships (relatives, close friends) with the subject under study (the individual);

3. (3) The converting status of farming households living close to the subject under study (the individual) within a radius of 500 m and having close relationships (relatives, close friends) with the subject under study (the individual).

Data collection

Our data are collected through questionnaires and provided by local authorities in Thai Nguyen province, Vietnam (see Fig. 2). We chose this place because it is known as the tea land of Vietnam. According to the General Statistics Office of Vietnam, in 2020, the total tea growing area in this area was 22,399 hectares, of which the harvested area was 19,754 hectares (reaching 88.19%). Currently, Thai Nguyen province supports and encourages farmers to convert to organic tea farming. In 2020, Thai Nguyen province implemented an organic tea production model according to Vietnamese standards on a scale of 60 hectares with 186 households (Thanh, Reference Thanh2020).

Fig. 2. Study location.

Basically, after completing the process of organic tea cultivation, the yield of organic tea equals 75–85% when compared with traditional tea (approximately 12.2 tons per ha)Footnote ⁷. However, when growing organic tea, farmers need 2–3 years to improve the soil and plant trees. As a result, in the early years of conversion, the yield of organic tea decreases (equivalent to 60% of that of traditional tea). This problem makes farmers afraid to convert to organic tea farming. Some other farmers have converted to organic tea production but abandoned it due to concerns about the economic development potential of this farming method.

From the list of farming households provided by the local government, we randomly selected 300 households to participate in the survey. The only requirement for survey participants is that the farmer is growing tea using traditional methods. To ensure that all data collected from farmers was accurate and unbiased, we selected interviewers from another locality. Before conducting the interviews, the interviewers were briefed on the survey content, equipped with data collection skills and gained informed consent from the interviewees. In addition, following Tseng et al. (Reference Tseng, Pham, Cheng and Teng2018), we test farmers' attention during the interview process by using one item and asking them to check the ‘very disagreeable’ response option. If farmers are not focused, they will skip this question, and their answer sheets will be discarded. At the end of the interview, 263 farmers met our criteria. In total, 21 respondents were not eligible because they did not provide enough information, and 16 respondents did not pass the attention checking.

The demographic characteristics of 263 farming households participating in the study are presented in Table 1. In total, 106 farmers intend to convert to organic farming (40.30%), and 157 do not intend to convert to organic farming (59.70%). We noticed a few similarities between the converting and non-converting groups. The majority of farmers in both groups are from the Kinh ethnic group, have a high school education and have 5 up to 10 years of tea growing experience. Besides, there are a few differences between these two groups of farmers. Most of the farmers interviewed were male and aged 36 up to 45, which implies that farmers who intended to convert are younger than those who did not intend to convert. The majority of farmers in the group intending to convert are 36 up to 45 years old (accounting for 53.77%), while the age group in the non-converting group only accounts for about 38.85%.

Table 1. Summary of statistics

Note: Switcher means the group of farmers who intend to switch from traditional to organic farming. Non-switcher indicates the group of farmers who do not intend to switch from traditional to organic farming.

Psychometric properties

According to the results of EFA in Table A2, the factor loadings of items in each factor are higher than 0.720. It implies that the structure of the factors is appropriate for this study. The results of CFA are presented in Table A3. Items evaluated for each construct have composite reliability values exceeding 0.921 and average variance extracted values exceeding 0.693. The entire results indicate adequate reliability. In addition, the indices of GOF, including CFI = 0.960, TLI = 0.947 and RMSEA = 0.075, imply that our data show a very good fit to the hypothetical model, and all the constructs can be used for further analysis.

The impacts of HB on farmers' converting intention

This study uses GSEM to analyze the structural relationship between the factors of the TPB and the NE on tea farmers' intention to convert from traditional to organic farming. In terms of control variables, the estimated coefficients of age and gender show a negative effect on INT but are insignificant. In contrast, the estimated coefficients of education and experience show a positive and statistically significant effect on INT. Meanwhile, culture is a factor that has a negative effect on INT but is not statistically significant. Our research results agree with Doanh et al. (Reference Doanh, Quynh and Pham2022b), who indicate that young, highly educated and experienced farmers tend to convert to organic farming.

In terms of TPB, all three structural variables in TPB impact farmers' converting intention positively. Specifically, the estimated coefficient of ATT (β _ATT = 0.612*) show a positiveeffect on INT and is statistically significant at 0.05, supporting H₁. SN (β _SN = 0.712**) and PBC (β _PBC = 0.786**) have positive and statistically significant effects on INT, supporting H₂ and H₃, respectively (see Fig. 3 and Table 2).

Fig. 3. Impact of HB on farmers' intention to convert to organic tea farming.

Table 2. NE and farmer's switching intention

Regarding the NE, NE was found to have a direct positive effect on INT but was not statistically significant. However, NE has a positive and statistically significant indirect effect on INT through ATT, SN and PBC. According to our estimation, a 1 one-unit increase in NE leads to an increase of 0.736, 0.633 and 0.404 units in ATT, SN and PBC, respectively. These results support H₅, H₆ and H₇.

To ascertain the impact of the NE on tea farmers' intention to convert from conventional to organic farming, we performed a three-case assessment: the effects of the neighborhood (Nei), relatives (Rel) and neighborhood and relatives (Nei&Rel). The estimated results are displayed in Table 3.

Table 3. Impact of the NE on farmers' switching intention according to three groups of the neighborhood (Nei), relatives (Rel) and neighborhood and relatives (Nei&Rel)

Regarding direct effects, Nei (β _Nei = 0.132) and Rel (β _Rel = 0.299) positively affect INT but are not statistically significant. Meanwhile, the Nei&Rel (β _Nei&Rel = 1090*) positively affects INT and is statistically significant at 0.05.

Regarding indirect effects, all three effects, Nei, Rel and Nei&Rel, have an indirect and statistically significant positive impact on INT through ATT, SN and PBC.

In short, empirical results indicate that the factors of TPB (ATT, SN and PBC) positively and significantly affect farmers' intention to switch to organic tea farming. Among these factors, PBC has the most impact on INT, followed by SN and ATT. These findings imply that farmers' perceptions play an important role in shaping their conversion intentions. Besides, the direct and indirect effects of NE on tea farmers' intention to convert from traditional to organic farming confirm that our hypothesis is grounded and consistent with reality.