Author’s Accepted Manuscript

Using structural equation modeling to identify the

psychological factors influencing dairy farmers’
intention to diversify agricultural production

Igor Senger, João Augusto Rossi Borges, João

Armando Dessimon Machado
www.elsevier.com/locate/livsci

PII: S1871-1413(17)30217-2
DOI: http://dx.doi.org/10.1016/j.livsci.2017.07.009
Reference: LIVSCI3260
To appear in: Livestock Science
Received date: 24 September 2016
Revised date: 4 July 2017
Accepted date: 24 July 2017
Cite this article as: Igor Senger, João Augusto Rossi Borges and João Armando
Dessimon Machado, Using structural equation modeling to identify the
psychological factors influencing dairy farmers’ intention to diversify agricultural
production, Livestock Science, http://dx.doi.org/10.1016/j.livsci.2017.07.009
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
 Using structural equation modeling to identify the psychological factors influencing

dairy farmers’ intention to diversify agricultural production

Igor Sengera1, João Augusto Rossi Borgesb2, João Armando Dessimon Machadoc

a
Federal University of Santa Maria, CESNORS, Linha 7 de Setembro, BR-386 Km40, Frederico Westephalen,

Brazil
b
Federal University of Grande Dourados, Rodovia Dourados – Itahum, Km 12, Dourados, Brazil
c
Federal University of Rio Grande do Sul, Avenida João Pessoa, 31, Porto Alegre, Brazil

igorsenger@ufsm.br

joaoborges@ufgd.edu.br

joao.dessimon@ufrgs.br

Abstract

Actions and public policies have been developed to encourage Brazilian farmers to diversify

their agricultural production. However, such actions have been unable to encourage

production and economic diversification. This article uses the theory of reasoned action

(TRA) and the theory of planned behavior (TPB) to understand the intention of dairy farmers

to diversify agricultural production. Results showed that the TRA had better explanatory

power of farmers’ intention to diversify agricultural production than the TPB

model. Corroborating the theory, results revealed that attitude and subjective norm positively

influence the intention of farmers to diversify agricultural production. Implications for public

policies are discussed.

Keywords: Diversification; Dairy farmers; Attitude; Social pressure; Theory of Reasoned

Action; Theory of Planned Behavior.

1
Telephone: +55 (55) 37448964.
2
Telephone: +55 (51) 97499180.
1 Introduction

In Brazil, actions and public policies have been developed to encourage Brazilian

farmers to diversify their agricultural production. The National Program for Strengthening

Family Agriculture (Pronaf), the National School Meal Program (PNAE), the Food

Acquisition Program (PAA), the National Plan for Sustainable Rural Development and

Solidarity and the Shares for the Diversification of Production and Income in Tobacco

Cultivated Areas, are some examples designed to encourage farmers to produce food and

therefore diversify production on their farms. Although such policies have encouraged

farmers to produce food, these actions have been unable to encourage the productive and

economic diversification (Gazolla, 2004). In this context, it is relevant to understand the

intentions of farmers regarding the diversification and the factors that influence intention.

Previous literature on the main drivers of farm diversification found that there is a set

of factors influencing farmers’ decisions to diversify their agricultural production and to start

new business. For instance, previous studies showed that farm characteristics (i.e., size of the

farm, farm type, location of the farm), and farmers characteristics (i.e., educational level, age,

gender) influence farmers’ decisions to diversify (Bateman and Ray, 1994; Mcnally, 2001;

Mishra et al., 2004; Pfeifer et al., 2009). Others studies focused on the motives underlying

farmers’ decisions to diversify. This strand of literature found that farmers’ financial and

nonfinancial goals, such as generating additional income, the continuance of farming and

ranching, and the enhancement of quality of life influence farmers’ diversification decisions

(Barbieri and Mahoney, 2009). In addition, Jongeneel et al. (2008) found that farmers’ trust in

the government is an important factor influencing farmers’ decisions to diversify. Moreover,

Hansson et al. (2013) found that the decision to start new ventures depends on the situation of

the family farm (i.e., if the spouse support the creation of new venture). In summary, previous
literature found that the drivers of farmers’ decisions to diversify and to start new business are

complex and includes several factors (Hansson et al., 2013).

In this study, the theory of reasoned action (TRA) and the theory of planned

behavior (TPB) are used to understand the intention of farmers to diversify3 agricultural

production. According to the TRA and the TPB the intention is originated from two latent

constructs: attitude and subjective norm, and an additional TPB construct, perceived

behavioral control. Both theories have been used to understand different farmers’ decisions. In

a first study, which is part of the same research project, Senger et al. (2017) found, by using

correlation coefficient, that attitude, subjective norm and perceived behavioral control are

significant and positively correlated with intention. Martínez-García et al. (2013) used the

TRA to study the decisions of livestock producers and found that attitude and subjective norm

are correlated with the intention of farmers to use improved pastures. Borges et al. (2014)

correlated the intention of farmers to use pasture improvement with the three constructs of

TPB (attitude, subjective norm and perceived behavioral control). However, these three

studies used only correlations, evaluating the relation between the constructs of TRA/TPB

one at a time. Additionally, this methodology does not allow assessing the relative importance

of the TRA/TPB constructs. To overcome the limitations of the use of correlations, Structural

Equation Modeling (SEM) has been suggested (Bleakley and Hennessy, 2012). SEM allows

the simultaneous estimation of all relationships in the TRA/TPB models and also identifies

the relative importance of each construct.

The combination of TRA/TPB and SEM has been used in different agricultural

contexts to understand: adoption of precision agriculture technologies (Adrian et al, 2005),

3
In the context of this paper, diversification concerns the development of activities inside the farm,
focusing on agriculture. By this definition, diversification entails the processing and improvement of
products (e.g. making and selling cheese rather than milk), adding value to the products (e.g. creating
a cheese brand), and selling products on the market (Barbieri and Mahoney, 2009; Ilbery, 1991;
Mahoney et al., 2004; Ploeg and Roep, 2003; Turner et al., 2003).
intention to purchase genetically modified agricultural products (Chen, 2008), intention to

adopt precision farming technologies (Kurosh and Saeid, 2010), the use of agricultural

information (Sharifzadeh et al., 2012), adoption of pro-environmental agricultural practices

(Price and Leviston, 2014). However, to the best of our knowledge, this is the first study using

TRA/TPB and SEM to analyze the intention of farmers to diversify agricultural production.

In the light of the foregoing, the objective of this study was to use TRA and TPB to

determine the effect of attitude, subjective norm and perceived behavioral control on the

intention of farmers to diversify agricultural production.

2 Theory of Reasoned Action (TRA) and the Theory of Planned Behavior (TPB)

The theory of reasoned action (TRA) and the theory of planned behavior

(TPB) assume that the intention to act is the immediate determinant of behavior (Ajzen,

2005). According to these theories, the stronger the intention to engage in a behavior, the

more likely to be their performance (Ajzen, 1991).

Both theories assume that there are independent latent constructs that influence the

intention. These constructs are: attitude and subjective norm. Attitude refers to the degree to

which a person has a favorable or unfavorable evaluation in relation to behavior (Ajzen and

Madden, 1986; Ajzen, 1991; Ajzen, 2005). Individuals form their attitudes based on their

perception of what can be true about a particular subject and this perception may or may not

be based on information, knowledge or even be an emotional reaction to the object,

sometimes supported by beliefs and values (Willock et al., 1999). Subjective norm, which is a

social factor, corresponds to the perceived social pressure of performing or not such

behavior (Ajzen and Madden, 1986).

The theory of planned behavior (TPB) is an extension of TRA. The TPB presents an

additional construct called perceived behavioral control. This construct is also assumed to

influence intention. Perceived behavioral control is equivalent to the ease or difficulty

perceived by the individual to perform the behavior. In TPB, the more favorable these three

constructs are, the stronger is the intention of an individual to express the analyzed behavior

(Ajzen, 1991).

In this study, the intention of dairy farmers to diversify agricultural production in

their farms in the next five years was measured. Therefore, farmers will have high intention to

diversify their production when they: understand the production diversification as being more

favorable (attitude); when they realize a high social pressure to diversify (subjective norm);

and when they perceive their own ability to implement this strategy in their properties as

positive (perceived behavioral control). Figure 1 shows the conceptual models to be tested in

this study, which originated three hypotheses:

H1: Attitude has a positive influence on farmers’ intention.

H2: Subjective norm has a positive influence in farmers’ intention.

H3: Perceived behavioral control has a positive influence in farmers’ intention.

TRA TPB

ATT + ATT +

+
INT SN INT

SN + PBC +

Figure 1: TRA and TPB models.

3 Methodology

3.1 Measures

The research instrument used in this study had two sections: the first containing

demographic questions and characterization of the farms, and the second with the items to
directly measure the construct intention, and also the items to measure directly and indirectly

the constructs attitude, subjective norm and perceived behavioral control. For the purpose of

this study, only direct measurements of each construct were used.

According to Fishbein and Ajzen (2010), direct measures are considered sufficient to

predict intention. Seventeen items were used to represent the TRA/TPB constructs. The

statements used to measure each item are shown in Table 1. The items were measured using a

five-point scale, with one being the value assigned to negative answers and five to positive

answers. Five-point scales have been used in agricultural studies (Bergevoet et al.,

2004; Barbieri and Mahoney, 2009; Ferguson and Hansson, 2015; Hansson et al, 2013; Zubair

and Garforth, 2006; Heong and Escalada., 1999; Sok et al, 2015), since they are considered

short enough for respondents to distinguish between the response options (Hansson et al,

2012).

Table 1: Statements used to measure each item of intention (INT), attitude (ATT), subjective

norm (SN) and perceived behavioral control (PBC) and the scales used to measure each of

them.

Item Statements Scale (1-5)

INT1 Do you intend to diversify agricultural production on Definitely Not
your farm in the next five years? Definitely Yes

INT2 Your intention to diversify agricultural production on Extremely Weak

your farm in the next five years is? Extremely Strong

INT3 Will you diversify agricultural production on your farm Certainly Not
in the next five years? Certainly Yes

INT4 I am NOT planning to diversify agricultural production Fully Agree

on my farm over the next five years. Fully Disagree

ATT1 The diversification of agricultural production on my farm Extremely Bad

in the next five years is: Extremely Good

ATT2 The diversification of agricultural production on my farm Extremely Unnecessary

in the next five years is: Extremely Necessary

ATT3 The diversification of agricultural production on my farm Extremely Disadvantageous

in the next five years is: Extremely Advantageous
ATT4 The diversification of agricultural production on my farm Extremely Impossible
in the next five years is: Extremely Possible

ATT5 The diversification of agricultural production on my farm Extremely Not Important

in the next five years is: Extremely important

SN1 Most people who are important to me think I should Strongly Disagree
diversify agricultural production on my farm in the next Strongly Agree
five years.
SN2 Most people that I hear opinions approve that I diversify Strongly Disagree
agricultural production on my farm in the next five years. Strongly Agree

SN3 Do you think that most farmers like you, will diversify Certainly Not
their agricultural production on their farms in the next Certainly Yes
five years?
PBC1 If you want to diversify agricultural production on your Definitely Not
farm in the next five years, you have enough knowledge. Definitely Yes

PBC2 If you want to diversify agricultural production on your Definitely Not

farm in the next five years, you have enough resources Definitely Yes
(machinery, financial resources, land, etc.)

PBC3 How confident are you to diversify agricultural Extremely Not Confident
production on your farm in the next five years? Extremely Confident
PBC4 The diversification of agricultural production on your Strongly Disagree
farm the next five years depends on you only. Strongly Agree
PBC5 For you, the diversification of agricultural production on Strongly Disagree
your farm the next five years is under your control. Strongly Agree

3.2 Sampling and data collection

Since the objective of this study is to determine the effect of attitude, subjective norm

and perceived behavioral control in the intention of farmers to diversify production, initially

farmers specialized in milk production were identified. The participation of rural activities in

the gross income of the property was used as a criterion to distinguish between specialized

farms from diversified ones (Hansson et al., 2010). Thus, if 50% or more of the income came

from a single activity, the farmer was considered to be specialized, and the greater this value,

the greater their specialization (Hoffmann et al., 1987).

According to a report provided by Frederico Westphalen's local government, there

were 460 farmers that sold milk during 2013. Agricultural extension technicians that work in

the region indicated 120 farmers specialized in milk production to be part of the sample of
farmers. If a farmer was not found or unwilling to participate in the survey, then another

farmer was selected. In some cases, a farmer indicated another farmer, often someone in his

own community. The final sample consisted of 101 farmers which were personally surveyed

by an interviewer, representing 22% of milk farms in the region, or 84% of farm that have

50% or more of income from milk production. The survey was conducted during November

and December 2014.

3.3 Data analysis

The method used in this study was Structural Equation Modeling (SEM) with latent

constructs. To test the proposed model, the two-step approach proposed by Anderson and

Gerbing (1988) was used. First, Confirmatory Factor Analysis (CFA) was used to obtain a

satisfactory Measurement Model (MM). Second, a structural model was developed and tested.

3.4 Measurement Models (MM)

This research tests and compares two Measurement Models (MM), the MM/TRA and

the MM/TPB, which are shown in Figure 2. The MM/TRA model had three latent constructs:

intention (INT), attitude (ATT), and subjective norm (SN). The MM/TPB model adds up the

construct perceived behavioral control (PBC).

In both measurement models all latent constructs were allowed to freely inter-relate

to each other. However, all the items were allowed to load on only one latent construct. For

instance, the item INT1 only loaded on the latent construct INT (intention). Moreover, the

errors terms were not allowed to relate to any other item.

To assess the construct validity of the measurement models (MM), we examined the

convergent and discriminant validity. Convergent validity was verified by checking the

magnitude, direction and significance of standardized factor loads on each latent construct. In

addition, average variance extracted (AVE) and construct reliability (CR) were used to

examine convergent validity. A measurement model presents an acceptable convergent

validity when AVE exceeds the minimum level of 50 percent, and when CR is above the

minimum value of 0.7 (Hair et al., 2010). Discriminant validity was assessed by comparing

the average variance extracted (AVE) for each latent construct with the squared inter-

construct correlations associated with that latent construct. To ensure discriminant validity the

AVE for each construct should be greater than the squared inter-construct correlations

associated with that latent construct (Borges and Lansink, 2016; Hair et al., 2010).

To check the validity of the measurement models a set of indicators was

used. Initially it was analyzed the validity of the measurement models by statistical qui-

square (x2), together with the degrees of freedom (df), and the probability value (p

value). Byrne (2001) recommends that the rate x 2 / df should not be greater than five. It was

also checked the root mean square error of approximation (RMSEA), the 90 percent

confidence interval for RMSEA, the comparative fit index (CFI), the Tucker-Lewis index

(TLI) and the standardized root mean squared residual (SRMR). It was also checked the

diagnostics model, as Hair et al. (2010) consider that this may indicate potential

improvements to the model or specify issues that were not previously identified. The

diagnostic measures used were the standardized residuals and modification indices. In this

study, all guidelines and threshold values used to evaluate the validity of the construct, the

measurement model validity and diagnostic measures were based on Hair et al. (2010).

MM TRA MM TPB
Figure 1: Measurement models used for TRA and TPB.

3.5 Structural Model (SM)

After obtaining a satisfactory measurement model, we run the structural model (SM).

Hair et al. (2010) point out that in the structural model a set of multiple regressions is

estimated and the emphasis is on the nature and magnitude of the relation between the latent

constructs. According to Borges and Lansink (2016) structural modeling is considered an

appropriate method to understand the causal relation between the constructs of TRA/TPB and

to test the underlying assumptions.

3.6 Socioeconomic characteristics of the sample

Among the 101 farmers in the final sample, 75.2% were male and 24.8% female

(woman also work on household chores). Most farmers (69.3%) work full time on the farm.

95% of the farmers have no other income source than agriculture. 37.6% of the farmers have

completed elementary school, 32.7% incomplete high school, 17.8% completed high school

and 5% higher education (complete and incomplete). No farmer mentioned having

postgraduate studies. 84.2 % of farmers rely on technical assistance. Among these farmers,

44.7% have private technical assistance, 20% are mainly using government technical

assistance, and 35.3% have both. Other socioeconomic characteristics of the sample are

presented in Table 2.

Table 2: Socioeconomic characteristics of the sample, minimum values (Min), maximum

(Max), mean ( ) and coefficient of variation (CV).

Variable Min. Max. CV (%)

Age (years) 20 81 48.8 33.8

Experience (years) 2 62 36.5 40.8

Estimated monthly gross income (R$) a 794 39.708 9.134 68.3

Percentage of income from milk production 50 100 75 24.1

Monthly milk production (liters) 800 28.000 6701 67.7

Number of dairy cows 3 55 16.3 51

Number of people on the farm 1 10 3.5 36.6

Number of children on the farm 0 8 1.1 102.7

Number of agricultural activities on the farm 1 4 1.9 36.8

Size of farms (hectares) 0 96 20.5 66.3

Arable land not used for agriculture 0 21 1.8 181.6

a
calculated based on monthly milk production data, milk participation and other activities in the gross
income of the farm. Also, it used the average of the nominal value of the milk price paid to farmers in
Rio Grande do Sul state in the year 2014 (CEPEA, 2015).

4. Results

4.1 Constructs of TRA/TPB

In general, farmers’ intention to diversify agricultural production in at least some of

the activities on their farms in the next five years was low. All items used to measure

intention presented mean below 3. See Table A1 in the appendix. Intra-construct correlations

for intention were high, ranging from 0.58 to 0.75. In Table A1 in the appendix, the mean,

standard deviation and correlations among all the items used to measure each construct are

presented.

Farmers showed predominantly a positive attitude towards diversification. All items

used to measure attitude presented mean above 3. The attitude intra-construct correlations

ranged from 0.10 to 0.59. The correlations between the items of intention with the items of the

construct attitude ranged from 0.23 to 0.51.

The social pressure perceived by farmers towards diversification was moderately

high. The three items used to measure subjective norm presented mean of at least 3. The

subjective norm intra-construct correlations ranged from 0.11 to 0.55. The correlations

between the items of the constructs intention and subjective norm were 0.22 to 0.53.
 Farmers did not perceive that they have capability to diversify their production since

all items used to measure perceived behavioral control presented a low mean. The highest

mean of the five items used to measure this construct was 3.2, and the lowest mean was

2.2. The perceived behavioral control intra-construct correlations were also low. The items

PBC3 and PBC4 showed negative correlation (0.22). The other correlations between the items

in this construct ranged from 0.05 to 0.44.

In general, it was observed that the correlations between the constructs intention and

perceived behavioral control are lower when compared to the correlation between intention

and attitude, and intention and subjective norm. The exception was the PBC3 item, which

correlated with the four items of intention, and these correlations ranged from 0.33 to 0.49.

4.2 Measurement Model of the theory of reasoned action (MM/TRA)

According to the guidelines for GOF statistics (Hair et al., 2010), the results for the

validity of the measurement model (MM/TRA) showed that the model did not provide a good

fit to the data (x2 =76.12; df=54; p=0.0128; and x 2/df =1.4; RMSEA=0.07; 90 percent

confidence interval for RMSEA=0,033-0,101; CFI=0.95; TLI=0.94; SRMR=0.07). Results of

the factors loadings, average variance extracted (AVE) and construct reliability (CR) are

shown in Table A2 in the appendix. The correlation matrix is presented in Table A3 in the

Appendix. These results suggested that the MM/TRA should be re-specified, particularly

focusing on the subjective norm construct, since the item SN3 had a low factor loading.

The results of the re-specified MM (rMM/TRA) without the item SN3 showed a

satisfactory fit to the data (x2=56.82; df =44; p=0.0512; and x2/df =1.3; RMSEA=0.062; 90

percent confidence interval for RMSEA=0.000-0.098; CFI=0.97; TLI=0.96;

SRMR=0.06). The insignificant x2 demonstrated that the model is well adjusted. The AVE

was recalculated for all the items, which improved mainly the convergent validity of the
subjective norm construct. In addition, the subjective norm construct reliability (CR) also

presented better fit (Table 3).

The results of the standardized factor loadings of rMM/TRA are shown in Table 3.

All items showed the expected sign and were significant to the critical level of 5%. The factor

loadings for all items of the constructs intention, attitude and subjective norm were above the

minimum value of 0.4. The AVE of the construct attitude was 46.3%, being just below the

50% recommended. For the constructs intention and subjective norm, the AVE was above

50%. The reliability of the construct (CR) for all the analyzed constructs was above the

threshold value of 0.7.

Therefore, the results of factors loadings, AVE and CR, analyzed together, indicate

convergent validity of rMM/TRA. Moreover, all the correlations between the TRA constructs

were significant at the 5% critical level and greater than 0.5. Furthermore, the AVE for all

constructs was greater than the square of the inter-construct correlations associated with that

construct (Table A3).

Table 3: Standardized factor loadings for each item of the respective constructs of TRA, with

standard errors between brackets, and the average variance extracted (AVE) and construct

reliability (CR) for each construct of the rMM / TRA.

SN
INT ATT
(Subjective
(Intention) (Attitude)
norm)
INT1 0.89 (0.03) ATT1 0.79 (0.05) SN1 0.95 (0.08)
INT2 0.77 (0.04) ATT2 0.58 (0.07) SN2 0,61 (0.08)
INT3 0.86 (0.03) ATT3 0.66 (0.06)
INT4 0.78 (0.04) ATT4 0.48 (0.09)
ATT5 0.83 (0.04)
AVE (%) 68.4 46.3 66.0
CR 0.90 0.81 0.79

These results indicate the discriminant validity of the model. The analysis of

standardized residuals among the items of the constructs did not identify major
problems. Therefore, it is considered the rMM/TRA satisfactory and that it could then be used

to explain the determinants of the intention of farmers to diversify agricultural production.

4.3 Measurement Model of the theory of planned behavior (MM/TPB)

The results of the GOF statistics for MM/TPB showed a poor fit of the model to the

data (x 2 =191.31; df =117; p=0.0000; and x2/df =1.6; RMSEA=0.08; 90 percent confidence

interval for RMSEA=0062-0103; CFI=0.877; TLI=0.852; SRMR=0.086). Moreover, the

significance given by x 2 demonstrates that the model needs to be adjusted. Results of the

factors loadings, average variance extracted (AVE) and construct reliability (CR) are shown

in Table A4 in the appendix. The correlation matrix is presented in Table A5 in the Appendix.

These results suggested that the MM/TPB should be re-specified, particularly focusing on the

subjective norm and perceived behavioral control constructs, since items SN3, PBC1, PBC4

and PBC5 had low factor loadings. Therefore, it was decided to re-specify and re-estimate the

MM/TPB eliminating the items SN3, PBC1, PBC4 and PBC5.

The results of the rMM/TPB still showed limitations (x2=78.02, df=63, p=0.0493 e

x2/df=1.2, RMSEA = 0.056, 90 percent confidence interval for RMSEA=0.003 – 0.088, CFI =

0.97, TLI = 0.96, e SRMR = 0.065). Results of the factors loadings, average variance

extracted (AVE) and construct reliability (CR) are shown in Table A6 in the appendix. The

correlation matrix is presented in Table A7 in the Appendix. Particularly, the limitations of

the perceived behavioral control construct persisted. Therefore, both the MM/TBP and the

rMM/TPB were insufficient models to explain the intention of farmers to diversify

agricultural production.

4.4 Comparing TRA and TPB measurement models

Before examining the structural model, the measurement models of the TRA and

TPB are compared. Using SEM, it was possible to independently test and compare the
models. The results presented in Table 4 showed that rMM/TRA showed better fit to the data

compared to the other models. Therefore, the rMM/TRA was used to run a structural model.

Table 4: Fit indices of theory of reasoned action (TRA) and theory of planned behavior (TPB)

measurement models (MM) and its explanatory power.

Fit Recommended Value to N<250 e

MM/TPB rMM/TPB MM/TRA rMM/TRA
Indices 12<m<30a
x2 191.31 78.02 76.12 56.82
(p=0.0000) (p=0.0493) (p=0.0128) (p=0.0512)
Df 117 63 54 44

x2/df ≤5 1.6 1.2 1,4 1,3

CFI Between 0 and 1 0.88 0.97 0.95 0.97

0,95 or better
TLI Below 0 or above 1 0.85 0.96 0.94 0.96
0,95 or better

RMSEA < 0,08 with CFI of 0.95 or higher 0.08 0.05 0.07 0.06
SRMR 0.08 or less (with CFI of 0.95 or higher) 0.08 0.06 0.07 0.06
a
based on Hair et al. (2010).

4.5 Structural Model (SM)

After a satisfactory measurement model (MM) was obtained, a structural model (SM)

was estimated to test the assumptions underlying the TRA. The SM has the same GOF

statistics of those of the rMM/TRA, as the SM has the same number of correlations between

structural relations as the constructs in MM.

Results of the structural model are shown in Table 5. The regression coefficient of

attitude on intention was significant and positive, indicating that hypothesis H1 (Attitude has a

positive influence on the intention of farmers) was not rejected. Furthermore, the regression

coefficient of subjective norm on intention was also positive and significant, suggesting that

H2 (Subjective norm has a positive influence on the intention of farmers) was not rejected. The

hypothesis H3 (perceived behavioral control has a positive influence on the intention of

farmers) was rejected, since the TRA model, considered satisfactory in this study, ignores this

construct.
 Together, the constructs attitude and subjective norm explained 49.3% of the

variance in intention. The regression coefficient indicated that attitude was the main

determinant of intention. However, subjective norm, despite having lower coefficient of

regression, also influences the decision of diversification in rural areas.

Table 5: Results of the impact of attitude (ATT) and subjective norm (SN) on intention (INT)

Structural relations Standardized parameter p (value)

ATT → INT 0.44 0.000
SN → INT 0.35 0.002
ATT correlated SN 0.57 0.000

5 Discussion and Conclusions

The comparative analysis of the four measurement models showed that the re-

specified measurement model based on the theory of reasoned action (rMM/TRA) was more

parsimonious to our data compared to the measurement models based on the theory of

planned behavior. Particularly, in the rMM/TRA the items reliably represent the latent

constructs intention, attitude, and subjective norm. Indeed, to obtain a valid measurement

model based on the theory of reasoned action only one item was excluded (SN3) from the

analysis because it had a low factor loading. On the other hand, the measurement models

based on the theory of planned behavior presented some limitations, particularly in the items

used to measure the perceived behavioral control construct. Indeed, the perceived behavioral

control construct presented low internal reliability in our study, which also occurred in

previous studies (Saba and Vassallo, 2002). A possible explanation for the low internal

reliability of the perceived behavioral control is that individuals who answered the

questionnaire based on the theory of planned behavior interpret the terms ‘control’ and

‘difficulty’ in different ways. In other words, the items used in the questionnaire to measure

the perceived behavioral control construct aim to capture whether people think they are under

control of the behavior and whether they think it is difficult to perform the behavior. Clearly,
if people think that they have control of the behavior, but still think that they would have

difficulty to perform the behavior, the items used to measure these dimensions will not be

strong correlated, and therefore, will not reliably represent the perceived behavioral control

construct. For instance, in the context of this paper, farmers may have found that

diversification of agricultural production is under their control, but at the same time consider

difficult to diversify their agricultural production.

Based on the results of the rMM/TRA, we run a structural model (SM). In SM, the

two TRA constructs attitude and subjective norm explained 49.3% of the variance in farmers’

intention to diversify, which is a common fit in similar studies in agricultural contexts (Adrian

et al., 2005; Borges and Lansink, 2016; Price and Leviston, 2014) and in behavioral sciences

(Armitage and Conner, 2001).

In the structural model, results of the impact of attitude and subjective norm on

intention are in line with Hansson et al. (2012) and Senger et al. (2017), who analyze the

influence of TRA and TPB constructs on farmers’ intention to diversify. However, Senger et

al. (2017) found that the construct perceived behavioral control positively influence farmers’

intention to diversify. As previously explained, we could not check the impact of perceived

behavioral control on intention, because the items used to measure did not reliably represent

this construct. These contradictory results might be explained because Senger et al. (2017) did

not use a confirmatory factor analysis to check whether the items used to measure the

perceived behavioral control construct reliably represent this construct, and based their

analysis on the correlation between perceived behavioral control and intention.

The regression coefficients of the structural model indicated that the effects of

attitude, and subjective norm on farmers’ intention to diversify were asymmetric. In

particular, the findings showed that attitude had a larger influence than subjective norm of

farmers’ intention to diversify. This result is in line with Hansson et al. (2012), who also
found that attitude had a higher impact than subjective norm on Swedish farmers’ intention to

diversify. Studies that used the theory of reasoned action and the theory of planned behavior

in agricultural contexts presented mixed results of the relatively impact of attitude, subjective

norm, and perceived behavioral control on intention and behavior (Borges and Oude Lansink,

2016; Borges et al., 2016; Lalani et al., 2016; Sok et al., 2016; van Dijk et al., 2016). The

mixed result of the relative impact of attitude, subjective norm, and perceived behavioral

control is expected, because the prediction of intention varies, for instance, across behaviors,

situations and cultures (Ajzen, 1991).

The high impact of attitude showed that the positive evaluation of diversifying

agricultural production was the main determinant of farmers’ intention to diversify. Therefore,

a policy intervention target to emphasize the benefits of diversification to farmers could

increase farmers’ intention to diversify. For instance, a possible intervention could be the

demonstration of cases where others farmers successfully diversified their agricultural

production. This strategy has been suggested by other studies in agricultural context to

increase farmers’ intention to adopt innovations and to even diversify their agricultural

production (Borges and Oude Lansink, 2016; Martínez-Garcia et al., 2013; Senger et al.,

2017).

Although attitude was the main determinant, subjective norm also influenced

farmers’ intention to diversify their agricultural production. This result is in line with previous

studies conducted with Brazilian farmers, and emphasizes the important role of social

pressure and the opinion of others in influencing farmers’ decision in Brazil (Borges et al.,

2016; Senger et al., 2017). According to Burton (2004), there are important referent groups to

which people often refer their behavior, because individuals do not act independently from

cultural and social influences. Specifically in Brazil, previous research found that family,

friends, rural extension agents, and government could be used to increase social pressure and
influence farmers’ behavior (Borges et al., 2016; Senger et al., 2017). We believe that an

intervention target to involve family members in the decision to diversify would increase

social pressure upon farmers, and this could result in a higher intention to diversify. We

believe that because family members seem to play an important role on farmers’ decisions.

Because our research focused on a specific Brazilian region and on milk farmers, the

implications for policy makers do not necessarily generalize to other regions and to farmers

working with different products. However, our results showed that the combination of the

theory of reasoned action, the theory of planned behavior with structural equation modeling is

appropriate for understanding farmers’ intentions to diversify agricultural production,

suggesting that future research could use this approach to study diversification in other

contexts.

Appendix

Table A1: Mean (x), Standard Deviation (SD) and correlation among all items of intention (INT), attitude
(ATT), subjective norm (SN), and perceived behavioral control (PBC)

PBC PBC PBC PBC PBC

INT1 INT2 INT3 INT4 ATI1 ATI2 ATI3 ATI4 ATI5 SN1 SN2 SN3
1 2 3 4 5
1.00
INT1
0
.708* 1.00
INT2 *
0
.752* .668* 1.00
INT3 * *
0
.704 .589 .674* 1.00
* *
INT4 * * *
0
.395* .464* .404* .375* 1,00
ATI1 * * * *
0
* * *
.311 .417 .428 .485* 1.00
ATI2 * * * .230* *
0
.349* .293* .359* .280* .592* .447* 1.00
ATI3 * * * * * *
0
* * * * * *
.394 .425 .426 .338 .323 .293 1.00
ATI4 * * * * * * .107
0
.401* .349* .515* .428* .598* .466* .496* .373* 1.00
ATI5 * * * * * * * *
0
 .537* .421* .528* .446* .382* .320* .272* .426* .469* 1.00
SN1 * * * * * * * * *
0
* *
.269 .388 .283 .402 .558* 1.00
* *
SN2 * .227* * .238* .226* .249* .105 * * *
0
*
.317 .288* 1.00
SN3 .228* * .235* .234* .142 .235* .150 .161 .172 .117 *
0
*
PBC .261 1.00
.166 .237* * .154 .051 .114 -.102 -.011 .178 .011 .090 .203*
1 0
PBC .357* .291* .280* 1.00
.255* .197* * * -.056 .032 -.011 .210* .113 .215* .175 .043 *
2 0
PBC .469* .357* .496* .338* .331* .317* .271* .428* .359* .411* .282* .287* 1.00
* * * * * * * * * * * .055 .143 *
3 0
*
PBC .360 1.00
.045 .053 .071 .030 -.002 .027 -.089 -.132 .088 -.047 .020 .176 * .057 -.022
4 0
PBC .442* 1.00
.061 .122 .183 .050 .034 .092 -.038 .094 .087 .149 .137 -.044 .210* .144 .174 *
5 0
2.8 2.7 2.7 2.8 3.7 3.4 3.7 3.2 3.9 3.2 3.4 3.0 3.2 2.2 3.0 2.5 3.0
SD 1.47 1.28 1.38 1.53 1.0 1.26 1.11 1.23 1.01 1.3 1.3 1.3 1.19 1.13 1.23 1.32 1.31

Table A2: Standardized factors loadings for each item of the theory of reasoned action (TRA) constructs
intention (INT), attitude (ATT) and subjective norm (SN), with standard errors between brackets, and the
average variance extracted (AVE) and construct reliabily (CR) for each construct of the MM/TRA.

TRA INT ATT SN

INT1 0.88 (0.03) ATT1 0.78 (0.04) SN1 0.90 (0.07)
INT2 0.77 (0.04) ATT2 0.58 (0.07) SN2 0.63 (0.07)
INT3 0.85 (0.03) ATT3 0.66 (0.06) SN3 0.21 (0.11)
INT4 0.78 (0.04) ATT4 0.47 (0.08)
ATT5 0.83 (0.04)
AVE (%) 68.4 46.3 43.0
CR 0.90 0.81 0.64

Table A3: Correlation matrix of theory of reasoned action (TRA) latent constructs intention (INT), attitude
(ATT), and subjective norm (SN)
INT ATT SN
INT 1 0.41 0.42
ATT 0.64 1 0.39
SN 0.64 0.62 1
Diagonal elements are the constructs variances; values below the diagonal are the correlations between the latent
constructs; values above the diagonal are the square of the correlation among constructs.

Table A4: Standardized factor loadings for each item of the theory of planned behavior (TPB) constructs
intention (INT), attitude (ATT), subjective norm (SN), and perceived behavioral control (PBC), with standard
errors between brackets, and the average variance extracted (AVE) and construct reliability (CR) for each
construct of the MM/TPB.
TPB INT ATT SN PBC
INT1 0.88 (0.03) ATT1 0.78 (0.05) SN1 0.88 (0.06) PBC1 0.33 (0.12)
INT2 0.77 (0.04) ATT2 0.58 (0.07) SN2 0.65 (0.07) PBC2 0.42 (0.10)
INT3 0.86 (0.03) ATT3 0.65 (0.06) SN3 0.22 (0.11) PBC3 0.69 (0.09)
INT4 0.77 (0.04) ATT4 0.47 (0.08) PBC4 0.09 (0.13)
 ATT5 0.83 (0.04) PBC5 0.28 (0.11)
AVE (%) 68.2 46.3 42.2 17.5
CR 0.89 0.81 0.64 0.45

Table A5: Correlation matrix of theory of planned behavior (TPB) latent constructs intention (INT), attitude
(ATT), and subjective norm (SN), and perceived behavioral control (PBC)
INT ATT SN PBC
INT 1 0.41 0.42 0.50
ATT 0.64 1 0.39 0.27
SN 0.65 0.63 1 0.35
PBC 0.71 0.52 0.59 1
Diagonal elements are the constructs variances; values below the diagonal are the correlations between the latent
constructs; values above the diagonal are the square of the correlation among constructs.

Table A6: Standardized factor loadings for each item of the theory of planned behavior (TPB) constructs
intention (INT), attitude (ATT), subjective norm (SN), and perceived behavioral control (PBC), with standard
errors between brackets, and the average variance extracted (AVE) and construct reliability (CR) for each
construct of the rMM/TPB.

rMM/TPB INT ATT SN PBC

INT1 0.88 (0.03) ATT1 0.79 (0.04) SN1 0.95 (0.07) PBC2 0.38 (0.10)
INT2 0.77 (0.04) ATT2 0.58 (0.07) SN2 0.61 (0.07) PBC3 0.72 (0.13)
INT3 0.86 (0.03) ATT3 0.66 (0.06)
INT4 0.78 (0.04) ATT4 0.48 (0.08)
ATT5 0.83 (0.04)
AVE (%) 68.3 46.3 64.3 33.0
CR 0.89 0.81 0.77 0.47

Table A7: Correlation matrix of theory of planned behavior (TPB) latent constructs intention (INT), attitude
(ATT), and subjective norm (SN), and perceived behavioral control (PBC).
INT ATT SN PBC
INT 1 0.41 0.38 0.53
ATT 0.64 1 0,35 0,32
SN 0.62 0.59 1 0,38
PBC 0.73 0.56 0.61 1
Diagonal elements are the constructs variances; values below the diagonal are the correlations between the latent
constructs; values above the diagonal are the square of the correlation among constructs.

REFERENCES

21
Adrian, A. M., Norwood, S. H., Mask, P. L., 2005. Producers’ perceptions and attitudes

toward precision agriculture technologies. Computers and Electronics in Agriculture 48, 256-

271.

Ajzen, I., 1991. The theory of planned behavior. Organizational behavior and human decision

processes. Amsterdan 50, 179-211.

Ajzen, I., 2005. Attitudes, personality and behavior, 2 ed. Open University Press,

Maidenhead.

Ajzen, I., Madden, T. J., 1986. Prediction of Goal-Directed Behavior: Attitudes, Intentions,

and Perceived Behavioral Control. Journal of Experimental Social Psychology 22, 453-474.

Anderson, J. C., Gerbing, D. W., 1988. Structural equation modeling in practice: A review

and recommended two-step approach. Psychological Bulletin 103, 411-423.

Armitage, C. J., Conner, M., 2001. Efficacy of the theory of planned behaviour: A meta-

analytic review. British Journal of Social Psychology 40, 471-499.

Barbieri, C., Mahoney, E., 2009. Why is diversification an attractive farm adjustment

strategy? Insights from Texas farmers and ranchers. Journal of Rural Studies 25, 58 - 66.

Bateman, D., Ray, C., 1994. Farm pluriactivity and rural policy: some evidence from Wales.

Journal of Rural Studies 10, 11 – 13.

Bergevoet, R. H. M., Ondersteijn, C. J. M., Saatkam, H. W., Woerkum, C. M. J. Van, Huirne,

R. B. M., 2004. Entrepreneurial behaviour of dutch dairy armers under a milk quota system:

goals, objectives and attitudes. Agricultural Systems 80, 1-21.

22
Bleakley, A., Hennessy, M., 2012. The Quantitative Analysis of Reasoned Action Theory.

The ANNALS of the American Academy of Political and Social Science, 640, 28-41.

Borges, J.A.R.; Tauer, L.W.; Lansink, A.G.J.M.O., 2016. Using the theory of planned

behavior to identify key beliefs underlying Brazilian cattle farmers’ intention to use improved

natural grassland: A MIMIC modelling approach. Land Use Policy 55, 193 – 203.

Borges, J. A. R., Lansink, A. G. J. M. O., 2016. Identifying psychological factors that

determine cattle farmers' intention to use improved natural grassland. Journal of

Environmental Psychology 45, 89-96.

Borges, J. A. R., Lansink, A. G. J. M. O., Ribeiro, C. M., Lutke, V., 2014. Understanding

farmers' intention to adopt improved natural grassland using the Theory of Planned Behavior.

Livestock Science 169, 163-174.

Burton, R. J. F., 2004. Reconceptualising the ‘behavioural approach’ in agricultural studies: a

socio-psychological perspective. Journal of Rural Studies 20, 359-371.

Byrne, B. M., 2001.Structural equation modeling with AMOS : basic concepts, applications,

and programming. Lawrence Erlbaum Associates, New Jersey.

CEPEA (Centro de Estudos Avançados em Economia Aplicada), Escola Superior de

Agricultura Luiz de Queiroz – Universidade de São Paulo (ESALQ/USP). Preços ao Produtor

- Valores nominais do leite (R$/litro). São Paulo, 20,15. Disponível em:

<http://cepea.esalq.usp.br/leite/?page=155>. Acesso em: 04 abr. 2015.

Chen, M., 2008. An integrated research framework to understand consumer attitudes and

purchase intentions toward genetically modified foods. British Food Journal 110, 559-579.

23
Ferguson, R., Hansson, H., 2015. Measuring Embeddedness and Its Effect on New Venture

Creation - A Study of Farm Diversification. Managerial and Decision Economics 36, 314-

325.

Fishbein, M., Ajzen, I., 2010. Predicting and Changing Behavior: The Reasoned Action

Approach. Psychology Press, New York.

Gazolla, M., 2004. Agricultura familiar, segurança alimentar e políticas públicas: uma análise

a partir da produção para o autoconsumo no território do Alto Uruguai/RS. Dissertação de

Mestrado em Desenvolvimento Rural. Porto Alegre, PGDR/UFRGS, 306.

Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., 2010. Multivariate Data Analysis, 7

ed. Prentice Hall, New Jersey.

Hansson, H., Ferguson, R., Olofsson, C., 2010. Understanding the diversification and

specialization of farm businesses. Agricultural and Food Science 19, 269-283.

Hansson, H., Ferguson, R., Olofsson, C., 2012. Psychological Constructs Underlying

Farmers’ Decisions to Diversify or Specialise their Businesses – An Application of Theory of

Planned Behaviour. Journal of Agricultural Economics 63, 465-482.

Hansson, H., Ferguson, R., Olofsson, C., Rantamäki-Lahtinen, L., 2013. Farmers’ motives for

diversifying their farm business: the influence of family. Journal of Rural Studies 32, 240-

250.

Heong, K. L., Escalada, M. M., 1999. Quantifying rice farmers' pest management decisions:

beliefs and subjective norms in stem borer control. Crop Protection 18, 315-322.

Hoffmann, R., Serrano, O., Neves, E. M., Thame, A. C. de M., Engler J. J. de C., 1987.

Administração da empresa agrícola. 7 ed. São Paulo: Pioneira.

24
Ilbery, B. W., 1991. Farm diversification as an adjustment strategy on the urban fringe of the

West Midlands. Journal of Rural Studies 7, pp. 207-218.

Jongeneel, R.A., Polman, N.B.P., Slangen, L.H.G., 2008. Why are Dutch farmers going

multifunctional?. Land Use Policy 25, 81 – 94.

Kurosh, R., Saeid, S., 2010. Agricultural specialists' intention toward precision agriculture

technologies: Integrating innovation characteristics to technology acceptance model. African

Journal of Agricultural Research 5, 1191-1199.

Mahoney, E., Armstrong, J., Barbieri, C., 2004. Farm and ranch diversification: one of the

futures of Canadian agriculture. In: Managing Excellence in Agriculture – 2004 Conference,

Niagara on the Lake, Ontario, Canada, pp. 26-28.

Martínez-García, C.G., Dorward, P., Rehman, T., 2013. Factors influencing adoption of

improved grassland management by small-scale dairy farmers in central Mexico and the

implications for future research on smallholder adoption in developing countries. Livestock

Science 152, 228-238.

McNally, S., 2001. Farm diversification in England and Wales – what can we learn from the

farm business survey?. Journal of Rural Studies 17, 247 – 257.

Mishra, A.K., El-Osta, H.S., Sandretto, C.S., 2004. Factors affecting farm enterprise

diversification. Agricultural Finance Review 64, 151 – 166.

Lalani, B., Dorward, P., Holloway, G., Wauters, E., 2016. Smallholder farmers' motivations

for using Conservation Agriculture and the roles of yield, labour and soil fertility in decision

making. Agricultural Systems 146, 80 – 90.

25
Pfeifer, C., Jongeneel, R.A., Sonneveld, M.P.W., Stoorvogel, J.J., 2009. Landscape properties

as drivers for farm diversification: a Dutch case study. Land Use Policy 26, 1106 – 1115.

Ploeg, J. D., Roep, D., 2003. Multifunctionality and rural development the actual situation in

Europe. In: Huylenbroeck, G. van, Durand, G. (Org.), Multifunctional Agriculture. A New

Paradigm for European Agriculture and Rural Development.Ashgate, Aldershot, Hampshire,

England, pp. 37-54.

Price, J. C., Leviston, Z., 2014. Predicting pro-environmental agricultural practices: The

social, psychological and contextual influences on land management. Journal of Rural Studies

34, 65-78.

Saba, A., Vassallo, M. 2002. Consumer attitudes toward the use of gene technology in

tomato production. Food Quality and Preference 13, 13-21.

Senger, I., Borges, J.A.R., Machado, J.A.D., 2017. Using the theory of planned behavior to

understand the intention of small farmers in diversifying their agricultural production. Journal

of Rural Studies 49, 32 – 40.

Sharifzadeh, M., Zamani, G. H., Khalili, D., Karami, E., 2012. Agricultural Climate

Information Use: An Application of the Planned Behaviour Theory. Journal of Agricultural

Science and Technology 14, 479-492.

Sok, J., Hogeveen, H., Elbers, A.R.W., Oude Lansink, A.G.J.M., 2016. Using farmers’

attitude and social pressures to design voluntary Bluetongue vaccination strategies. Preventive

Veterinary Medicine 133, 114 – 119.

Sok, J., Hogeveen, H., Elbers, A. R. W., Oude Lansink, A. G. J. M., 2015. Farmers’ beliefs

and voluntary vaccination schemes: Bluetongue in Dutch dairy cattle. Food Policy 57, 40-49.

26
Turner, M., Whitehead, D., Barr, D., Fogerty, M., Errington, A., Lobley, M., Reed, M., 2003.

Farm Diversification Activities: Benchmarking Study 2002. Final Report by the Universities

of Exeter and Plymouth to Defra. CRR Research Report 4. Exeter, University of Exeter. 2003.

Retrieved online on August 20, 2015. Available from: https://socialsciences.exeter.ac.uk/.

van Dijk, W.F.A., Lokhorst, A.M., Berendse, F., Snoo, G.R., 2016. Factors underlying

farmers’ intentions to perform unsubsidised agri-environmental measures. Land Use Policy

59, 207 – 216.

Willock, J., Deary, I. J., Edwards-Jones, G., Gibson, G. J., Mcgregor, M. J., Sutherland, A.,

Dent, J. B., Morgan, O., Grieve, R., 1999. The Role of Attitudes and Objectives in Farmer

Decision Making: Business and Environmentally-Oriented Behaviour in Scotland. Journal of

Agrincural Economics 50, 286-303.

Zubair, M., Garforth, C., 2006. Farm level tree planting in Pakistan: the role of farmers’

perceptions and attitudes. Agroforestry Systems 66, 217-229.

Highlights:

 Farmers’ intention to diversify is determined by social pressure

 Farmers’ intention to diversify is influenced by their attitudes towards diversification
 TRA performs better than TPB to explain farmers' intention to diversify

27