Choosing a Stimuli Presentation Method
In order to get reliable and accurate data, it is necessary to present the combination of factors (stimuli) to respondents in the natural and realistic way through verbal or visual methods (Brascamp, 1996; Green & Srinivasan, 1990; Gustafsson et al., 2007; Hair et al., 2010). Although, a few researchers (Brascamp, 1996; Green & Srinivasan, 1978, 1990) have mentioned two methods for the presentation of stimuli but according to Hair et al. (2010), the three methods include two factors-at-a-time or trade off, the full profile and the pair-wise comparison methods.
The two factor-at-a-time also called trade-off method (Johnson, 1974) is simple method in which respondents rank all combinations of each pair of factor levels at one time (Brascamp, 1996; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007; Hair et al., 2010). It is easy to apply and reduces information overload on the part of respondents and frequently used for mail questionnaire form of data collection procedure (Brascamp, 1996; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007). The major disadvantages with this method are the loss of realism by using only two factors at a time, large number of evaluations required and only employ verbal descriptions of factor’s combination rather than pictorial or non-writing forms (Brascamp, 1996; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007).
The full-profile method or concept evaluation method or multiple factor evaluation method makes use of the complete set of factors including all levels in the form of a profile card (Brascamp, 1996; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007; Hair et al., 2010). It gives more realistic picture of the preference and allows to make
101
decision as occurs in reality by considering all determinant factors together on one profile card, therefore, commonly used on large scales (Brascamp, 1996; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007; Hair et al., 2010; Wittink & Cattin, 1989). The major disadvantage includes information overload on the part of respondents that makes the decision/preference task difficult (Brascamp, 1996; Cattin & Wittink, 1982; Green & Srinivasan, 1990; Hair et al., 2010; Payne et al., 1992; Wittink & Cattin, 1989).
The pair-wise comparison method is a combination of trade-off and full profile methods and used in specialized and modern conjoint techniques for a large number of factors, such as adaptive conjoint analysis (ACA) (Gustafsson et al., 2007; Hair et al., 2010). Contrary to the full profile method, it includes only few factors in the construction of profiles and respondents have to rate two profiles according to their preference at a time (Hair et al., 2010). Similarly, in pair-wise comparison method profiles are rated whereas in trade-off method pairs of factors are evaluated (Hair et al., 2010).
After identification and selection of determinant factors and deciding presentation method, the next important task is the creation of combination of factors (stimuli) (Brascamp, 1996; Green & Devita, 1974; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007; Hair et al., 2010). It is worth considering keeping the number of stimuli to minimum in order to avoid information overload at the end of respondent and to increase the validity of the results (Brascamp, 1996; Green & Devita, 1974; Green & Srinivasan, 1978, 1990; Gustafsson et al., 2007; Hair et al., 2010). According to Hair et al. (2010), the number of stimuli depends upon the number of factors in trade-off method and can be calculated as:
Number of trade-off stimuli
2 ) 1 (
N N
Where N is the number of factors.
In case of small number of factors (6 or few), all possible combinations of factors can be used for evaluation for the full profile and pair-wise comparison methods, hence, called factorial or complete design (Green & Devita, 1974; Green & Srinivasan, 1990; Gustafsson et al., 2007; Hair et al., 2010). For large number of factors (7 to 10), a reduced design is preferred to avoid inconsistency in the evaluation and overload of information at the end of respondents (Green & Devita, 1974; Green & Srinivasan,
102
1990; Gustafsson et al., 2007; Hair et al., 2010). There are two procedures in conjoint analysis to reduce (fractionate) the number of combination of factors namely: simple random sampling of complete set of stimuli and fractional factorial design (Green & Srinivasan, 1978; Gustafsson et al., 2007; Hair et al., 2010). Simple random sampling is the quickest and easiest way to reduce the complete set of stimuli to desired number by randomly selecting stimuli but this approach is not common for marketing research (Green & Devita, 1974; Green & Srinivasan, 1978; Hair et al., 2010). A number of researchers (Assmus & Key, 1994; Green & Devita, 1974; Green & Srinivasan, 1978; Gustafsson et al., 2007; Hair et al., 2010) advocate that fractional factorial design is a systematic and frequently used procedure to reduce the number of combinations of factors and achieve two goals at the same time:
a) reduce the number of stimuli to desired level
b) maintain the orthogonality (independence of the factors)
In designing combination of factors with realistic combination of factors for this study, an orthogonal stimulus was created through fractional factorial procedure using the computer software, SPSS. This orthogonal design helped creating the orthogonal main- effects design (combination of factors without inter-attribute correlation) that allowed the statistical testing of several factors without testing every combination of factor levels.
In conjoint analysis, data is usually collected on a non-metric scale (ranking) (Brascamp, 1996) but a number of researchers (Brascamp, 1996; Green & Devita, 1974; Green & Srinivasan, 1978; Gustafsson et al., 2007; Hair et al., 2010) point out that both non-metric and metric (rating) scale can be used in conjoint analysis. To be more specific, trade-off method uses non-metric (ranking) scale while full-profile and pair- wise comparison employ both type of non-metric and metric scales (Gustafsson et al., 2007; Hair et al., 2010). For this study with full-profile combination of factors, a ranking scale from most preferred to least preferred was used for data collection. According to Hair et al., (2010), profiles can be presented in written description, physical or pictorial models to be effectively used in data collection procedure. In order to make profiles attractive, understandable and easy to evaluate, respondents were presented with cards showing factor levels with appropriate signs and pictures. For example, different levels of factor ‘price’ were supported with different sizes of the sign ‘Rs’ (which is commonly used for price in Pakistan) in this study. In addition,
103
considering the target audience these cards were also presented in local language (Urdu), where necessary. All profiles or decision cards for citrus growers and pre- harvest citrus contractors used for data collection are attached in appendix (Appendix D-1, D-2, E-1 & E-2).
Survey Administration
There are three common procedures that are followed for data collection in conjoint analysis namely; phone surveys, mail (with pencil and paper questionnaire or computer- based surveys) and personal interviews (Akaah, 1991; Bisgaard, 1992; Cattin & Wittink, 1982; Green & Srinivasan, 1990; Gustafsson et al., 2007; Hair et al., 2010; Witt & Bernstein, 1992; Wittink & Cattin, 1989). Phone and mail surveys are not used frequently and are usually conducted to ensure geographical representation, reduce the cost of conducting survey and increase the return rate of questionnaires (Brascamp, 1996; Gustafsson et al., 2007; Vriens, 1995; Wittink & Cattin, 1989). However, a number of limitations are associated with phone and mail surveys that include respondents training, advance call for cooperation, efficient delivery of questionnaires, giving call back number and follow up calls (Gustafsson et al., 2007). A personal interview is best suited in complex and difficult situations or when respondents need some assistance to complete the questionnaires (Gustafsson et al., 2007). This method is common with the marketing research on a large scale (Brascamp, 1996; Gustafsson et al., 2007). Therefore, personal interview method was adopted for the collection of data from citrus growers and contractors from Pakistan. During personal interviews, respondents were presented with cards or profiles with complete set of relevant factors including different levels of each factor and were asked to rank order these cards or profiles on a 1-10 preference scale.