conditions under which it might be considered as a market determinant have so far only been sketchily discussed and applied incompletely to consumer behaviour under uncertainty (Misztal, 1998, p. 29).
Regardless of the renascent interest in a conceptualisation of the multifaceted element of trust in recent years, the prevailing methodological diversity mostly circumvents a distinct definition of trust. Yet, the perhaps most commonly used concept of trust – particularly in the environment of economics – implies a disposition towards trusting behaviour; i.e. behaviour accepting vulnerability based upon the personal expectation. Nooteboom (1996, p. 246) remarks that ‘X trusts Y to the extent that X chooses to cooperate with Y on the basis of a subjective probability that Y will choose not to employ opportunities for defection that X considers damaging, even if it is in the interest of Y to do so. The trust worthiness of Y depends on Y’s true propensity to employ those opportunities’.
One of the first elementary approaches to analyze trust in the perspective of a rational choice model of neoclassical economics has been presented by Coleman (1990, p. 99). His approach is based on the postulate of maximizing utility under uncertainty and requires the trustor to decide between investing trust – which would yield an expected utility of the expected value of a potential gain less the expected value of a potential loss, and not investing trust – which would not change his utility. The decision whether or not to trust the trustee is based on the probability that the trustee is trustworthy, the potential gain, and the potential loss that might occur if the trustee is not trustworthy. It appears logically consistent to consider trust as a subjective probability in the above context.
One of the first multilevel approaches to formally introduce the element of trust into decision making under uncertainty was undertaken in B?cker and Hanf’s (2000) seminal model of individual information processing. The model proposes a two step risk perception process in which differences in the reliability between single types of suppliers are captured by subjective failure probabilities. Thus, trust is understood as a subjective probability that the trustee, i.e. the supplier of a food, is reliable. Formally, consumer K distinguishes between two different types of suppliers. Whereas suppliers of type A are regarded as reliable, those of type B are assumed to be less reliable. Consequently, K judges the probability P(G│A), to purchase an unsafe item from type A to be smaller than P(G│B), the respective failure probability assigned to type B. Referring to available information and personal experience, K generally purchases from supplier J which he presumes to be of type A. Since K does not possess perfect information, however, he cannot be sure that J actually belongs to type A. His trust in J to be reliable is expressed through the subjective probability PJ , leaving a residual probability of (1 - PJ ) for J belonging to type B. Naturally, K can modify his decision to purchase a potentially unsafe item X anytime by replacing it through substitute Y which he considers to be more secure. The substitution, however, would require that the expected utility of Y exceeds the expected utility of X. The likelihood for K to purchase X depends on the subjective probability PJ . B?cker and Hanf (2000) assume that if K comes to know about the occurrence of a disconcerting incidence, caused by good X which J has sold, K will revise any prior belief PJ about J's reliability to the posterior probability PPJ . PPJ is the conditional probability of 'J being of type A' after having observed that X is unsafe.
The following paragraphs will discuss approaches that evolved as conceivable alternatives to the Expected Utility Theory. Among these are as well the Prospect Theory as the Theory of
2
Reasoned Action, which are both considered as methodological precursors to the Theory of Planned Behaviour (TPB), on which this paper will predominantly focus. The Theory of Reasoned Action, as introduced by Fishbein and Ajzen (1975) and Ajzen and Fishbein 1980), aims at predicting the volitional behaviours and at comprehensively explaining their underlying psychological determinants. In doing so, the theory combines Fishbein’s (1963) Attitude Theory and Dulany’s (1967) Theory of Propositional Control which previously did not explicitly address social behaviour. Consequently, the Theory of Reasoned Action emphasises the impact of behavioural and normative beliefs on the consumer’s intention to conduct a given behaviour (East, 1997, p. 134).
According to the Theory of Reasoned Action, intentions comprise two conceptually different determinants. The first predictor of intention is the consumer’s attitude towards the behaviour, which refers to the degree to which a consumer has an either favourable or unfavourable evaluation of the behaviour in question (Ajzen, 1991, p. 188). The second predictor of intention is a social factor termed subjective norm, and refers to the consumer’s perception of contingent social pressures to perform the behaviour in question. Subjective norms are a function of normative beliefs that indicate the likelihood that important individuals or groups in the consumer’s social environment have in his selection of behavioural patterns. The consumer will intend to perform a certain behaviour when he perceives it as being positively evaluated and as desired by the social environment – and vice versa.
The TPB differs from the Theory of Reasoned Action in its addition of a third determinant of intention; the perceived behavioural control, PBC. The perceived behavioural control refers to the consumers’ perceptions of their ability to perform a given behaviour. In analogy to the attitudinal beliefs, perceived behavioural control is determined by control beliefs, i.e. beliefs about the presence of factors that facilitate or impede the performance of the behaviour in question. Control beliefs are mostly determined through the consumer’s individual experiences, but also through information and experience of the social environment that influences the subjectively perceived difficulty of performing the behaviour in question. The more resources and opportunities individuals assume to possess, and the fewer impediments they anticipate, the greater is their perceived control over the behaviour (Ajzen, 1991, p. 196).
Accordingly, the consumer’s perceived behavioural control varies across situations and actions. The TPB is illustrated in figure 1.
Figure 1: The Theory of Planned Behaviour. Source: Ajzen (1991, p. 182).
3
With reference to the previously discussed determinants of consumer behaviour under uncertainty, the TPB has been extended by Mazzocchi et al. (2004) towards the inclusion of trust, T , as an additional predictor of consumer behaviour. Trust was shown to be a crucial prerequisite for consumers to engage in economic interactions under uncertainty when the obtainment of complete information can only be ascertained at prohibitively high costs. This applies particularly for the credence qualities of a good (Darby and Karni, 1973, p. 69). Since trust under certainty, however, is tantamount to knowledge, any extension of the theory needs to include the element of risk, likewise. Consequently, emphasis will be put on the consumer’s perceived risk.
The introduction of trust and perceived risk into the TPB has not affected the consumer’s nonvolitional beliefs, i.e. the perceived behavioural control and its direct influence on the consumer’s intention to perform a given behaviour. The system is expected to model the average relation among the global variables and the behavioural intention and ought to assess whether these relations vary according to other factors. In consideration of the fact that particularly information and socio-demographic variables ultimately determine the consumer’s (volitional) beliefs, another extension of the original theory in order to comprehensively explain consumer behaviour under uncertainty seems inevitable (Mazzocchi et al., 2004). This conceivable revision of the TPB is depicted in figure 2.
Figure 2: The Theory of Planned Behaviour – Extended (SPARTA model)
Due to a low correlation between certain determinants, the model was simplified as follows.
4
Figure 3: The SPARTA II Model. Source: Modified from Mazzocchi et al., 2005b, p. 23
3 Empirical Analysis of Trust as a Determinant of Consumer Behaviour 3.1 Data
The element of trust and its alleged impact on the consumer’s intention to purchase was empirically assessed through a pan-European survey comprising 2,725 thirty minute face-to-face, in-home interviews with the family member responsible for purchase and/or preparation of food (Dierks, 2005). The interviews were conducted in spring 2004 throughout the United Kingdom, France, Italy, the Netherlands, and Germany. Since a sampling frame significant at national level for those in charge of purchasing food is nearly impossible, it is obvious to maintain the household as the sampling unit and to ensure that the respondent is representative for the entire household. The sample is based on simple random sampling and probabilistic extraction which guarantees national representativeness.
Within the scope of European Commission’s research project Food Risk Communication and Consumers’ Trust in the Food Supply Chain, country-specific observations were transmitted to the University of Reading where the data was collated and processed. Subsequent to its conversion into a single data set, elementary statistical analyses and estimations were performed and then placed at the disposal of the respective cooperating institutions. This task was mostly performed by Lobb et al. (2005), Mazzocchi et al. (2005), and Cavicchi et al. (2005), whose efforts provide the data basis for the analyses conducted.
3.2 Quantifying Trust
Within the scope of the survey, respondents were asked to indicate their trust in information provided by selected sources on a seven point Likert scale. In an adjacent step, a factor analysis was performed on 451 German observations. Following a varimax rotation, the factor analysis yields five well distinguishable principal components whose loadings are depicted in table 1.
Table 1: Principal component loadings for trust in food safety information Information Source Shopkeepers TM 0.001 TF 0.823 5
Components of Trust TI 0.156 TA 0.010 TV 0.129
Supermarket Organic Shop Specialty Store Farmers /Breeders Processors Health Officials University Scientists National Food Authority Government Political Groups Environmental Groups Animal welfare Organisations Consumer Organisations European Food Safety Authority Television documentary Television news / current affairs Television advertising Newspapers Internet Radio Magazines Product Label Component Label 0.119 0.175 0.220 0.131 0.107 0.207 0.160 0.041 0.161 0.162 0.138 0.105 0.208 0.206 0.705 0.801 0.196 0.786 0.520 0.824 0.577 0.272 0.792 0.715 0.780 0.739 0.609 0.288 0.165 0.182 0.118 0.101 0.058 0.070 0.113 0.136 0.082 0.089 0.312 0.193 0.048 0.139 0.247 0.426 0.175 0.121 0.160 0.133 0.243 0.755 0.687 0.818 0.561 0.262 0.219 0.053 0.540 0.659 0.195 0.288 0.1016 0.125 -0.072 0.229 0.125 0.190 -0.059 0.368 0.168 0.035 -0.102 0.091 0.229 0.056 0.086 0.291 0.844 0.881 0.482 0.005 0.211 0.035 0.104 0.149 0.000 0.056 0.102 -0.028 0.206 -0.069 0.078 0.186 0.467 0.045 0.151 0.081 0.569 0.733 0.166 0.135 -0.056 0.282 0.113 0.007 0.695 0.047 0.203 0.124 0.431 0.445 Media Food Chain Independent Alternative Lobbies Note: A varimax rotation with Kaiser Normalisation has been conducted. The rotation converged in six iterations. Values exceeding 0.5 are printed bold.
In an adjacent step, a cluster analysis (hierarchical k-means cluster analysis) was performed on the observations. In accordance with the pan-European findings, the analysis was preset to three clusters (Dierks, 2005). Results are illustrated in table 2.
Table 2: Categorization of clusters featuring the German data set according to the k-means method Clusters Trust in media Trust in food chain actors Trust in independent sources Trust in alternative sources Trust in vested interests 1 -0.23 -0.94 0.38 0.61 -0.17 6
2 0.04 0.60 -0.34 0.20 0.39 3 0.21 -0.04 0.22 -1.22 -0.59