CHAPTER FOUR: DECISION MODELS
4.5. Decision Models
4.5.2. Multi-equation Models
Multi-equation models may help explain a sequential and/or dynamic process of adoption-decision making. They may stem from a single equation model which is extended to cover interrelationships among the variables involved. One example is the study by Negatu and Parikh (1999), in which binary probit and ordered probit models were employed to explain the sequential learning process among farmers relating to improved wheat varieties. The learning process was described as a two- way effect between farmers’ perceptions and adoption behaviour. The ordered probit
models were used for testing the effect of perceptions given the predicted adoption behaviour; while the binary probit models were used for estimating adoption behaviour given the predicted perception on market or yield. The high explanatory power of these models (between 79 and 94 percent) shows a good estimation and explanation of the farmers’ adoption behaviour.
The other approach is using structural equation modelling (SEM), which is a multivariate statistical modelling technique that usually combines several methods such as factor analysis, path (causal) analysis and regression analysis (Austin et al., 1998; Rigdon). The combination is targeted at more thoroughly explaining the interrelationships among the variables involved. The structure of SEM may be constructed, for example, based on the Theory of Planned Behaviour (TPB). This topic is presented next.
Figure 4.3 Theory of Planned Behaviour (Ajzen, 2002a)
Behavioural Beliefs Normative Beliefs Control Beliefs Attitude toward the Behaviour Subjective Norms Perceived Behavioural Control Intention Behaviour Actual Control
4.5.2.1 Theory of Planned Behaviour (TPB)
The TPB was created in order to incorporate socioeconomic, socio-cultural,
psychological and economic approaches in behavioural analysis (Burton, 2004). This theory assumes that person’s behaviour is affected by the variation in the person’s attitudes, social pressures (“subjective norms”) and “perceived behavioural control” (Ajzen, 1991; 2002a, see Figure 4.3). Here, attitudes are defined as “a person’s overall evaluation of performing the behaviour in question” (Ajzen, 2002a, p. 5). They reflect the person’s positive and negative judgment about the effects brought by the behaviour. The subjective norms reflect an individual’s “perceived social
pressure” that may emerge in the form of her/his “beliefs about the normative expectations of others and motivation to comply with these expectations” (Ajzen, 2002a, p. 1). The perceived behavioural control represents an individual’s belief in their own capacity to deal with an event (Burton, 2004). The relationships between behaviour and its precursors are mediated by the person’s intention. These
relationships are considered to help delineate the person’s learning and mental process, the latent determinants of one’s behaviour (Ajzen, 1991, 2002a). This theory is in accord with the concept of mental process outlined by Antonides (1996), in which an objective environment (stimuli, e.g. socioeconomic and personal
conditions) induces mental processes (attitudes, expectations, social norms) and, in the end, results in an action (outcome, behaviour). Thus, the TPB may provide a clearer explanation of differences in how farmers behave.
The procedures for quantifying a TPB model involve interviews and statistical analyses. The interview involves a questionnaire that is constructed based on a standardized procedure (see Ajzen, 2002b). The questions reflect the complexity associated with the behaviour, and contain scaling techniques to quantify different degree of responses (Ajzen, 2002b; Beedell & Rehman, 2000). The most common scale used is 7-point contrasting adjective scale, e.g. from strongly agree to strongly disagree, but fewer points and/or other scaling procedures may also be used (Ajzen, 2002c). The responses are then tested for internal consistency (Ajzen, 2002b) in order to make sure that the responses can be added up into reliable indexes for a regression analysis. This test can be based on Cronbach’s alpha (Ajzen, 2002b; Hair, Anderson,
Tatham, & Black, 1998), where the acceptable scale for high consistency is 0.7 or above, or 0.6 or above for exploratory analyses (Hair et al., 1998).
Structural equation models (SEMs), then, will be constructed based on the interview results and the consistency test. The SEMs are constructed to identify latent
interrelationships among variables (Austin et al., 1998; Babbie, 2004; Hair et al., 1998; Rauniyar & Goode, 1992). This kind of model is termed a structural model, in which the “path” (Hair et al., 1998, p. 17) of multi-correlations between dependent and independent variables is explained. The variables may be determined based on, for example, the results and recommendation from previous studies (Hair et al., 1998).
Another type of SEMs is a “measurement” model (Hair et al., 1998, p. 17) which provides the basis for categorizing farmers’ responses to a number of adoption- decision variables and assesses the correlation among these variables. This model is often replaced by factor analysis because both have the same function. According to Hair et al. (1998, p. 14), factor analysis is targeted to “find a way of condensing the information contained in a number of original variables into a smaller set of varieties (factors) with a minimum loss of information”. Babbie (2004, p. 455) also asserts that factor analysis can also help identify “patterns among the variations in values of several variables”. In the case of technology adoption, factor analysis can help determine the adoption patterns, whether independent, sequential, or simultaneous, and create a single factor accordingly (Rauniyar & Goode, 1992). Therefore, the results from the factor analysis can be used to determine the optimal combination of the farmers’ adoption-decision variables, including the heuristic patterns.
The applications of a TPB model in agricultural related areas are still limited, but so far they have shown promising results. The applications also appear to vary
according to the case situation under study (see Beedell & Rehman, 2000; Bergevoet, Ondersteijn, Saatkamp, Van Woerkum, & Huirne, 2004; Chetsumon, 2005; Coleman, McGregor, Hemsworth, Boyce, & Dowling, 2003; Hrubes, Ajzen, & Daigle, 2001; Zubair & Garforth, 2006). Some introduce new components to the original TPB concept, while others prefer to use a simpler version of the concept. Beedell &
Rehman (2000), for example, used a TPB model to explain the motives behind different conservation behaviours among different groups of farmers and found the relationships between the farmers’ behaviour, beliefs, attitudes and motivations (intentions), and social pressures. Bergevoet et al. (2004) have also successfully applied a model derived from the TPB concept for examining the relationships between Dutch dairy farmers’ entrepreneurship and their goals, objectives and attitudes. Hrubes et al. (2001) also introduced some personal characteristics, i.e. self- transcendence, self-enhancement, openness and conservation to the TPB concept, and reported a strong applicability of the model for predicting the rate of hunting
behaviour among outdoor recreationists. Using SEM, Chetsumon (2005) analyzed the combination of the TPB components and some personal factors (a measure of
intelligence, openness and extraversion) to explain extension agents’ intention to adopt an expert system. In contrast, Coleman et al. (2003) used a simpler TPB model to estimate the behaviour of abattoir stockpeople, which was affected mainly by attitudes and the “tough-mindedness” character. A study by Zubair and Garforth (2006) also limited the measures of perceived behavioural control to include only the perceptions on the impediments relevant to the Pakistani farmers’ intention and behaviour towards growing trees. The latter appears to be the only study that has applied the TPB concept in the case of farmers in developing countries.
Some argue, however, that there are pitfalls in the TPB. Burton (2004), for example, argues that the TPB may not be effective as a predictive model, rather it is best used for quantifying some qualitative determinants of behaviour (e.g. socio-cultural or socio-psychological factors). Such an argument is reasonable, since according to Nuthall (2005)2, the TPB questionnaire “merely asks superficial questions for putting in the model” and, hence, it can only provide a general sense of the latent variables (behavioural beliefs, normative beliefs and control beliefs). This calls for other models and methods that can help reveal a more detailed description of the basic structure of the latent variables behind one’s behaviour. The Ethnographic Decision Tree Modelling (EDTM) and the Personal Construct Theory (PCT), for example, may
provide more definite personal constructs in decision making and, thus, their results may be complementary to the TPB model. These will be discussed next.