• Keine Ergebnisse gefunden

The Meaning Structure Method

4.2 Original Means-End Approach

4.2.2 Measurement Procedure

The measurement of means-end chains takes place in five phases:

1. Elicitation of the attributes;

2. Selection of the attributes;

3. Elicitation of the attribute levels;

4. Performing laddering interviews;

5. Determination and coding of means-end chains.

These phases are discussed below. The data used to illustrate several aspects of the measurement and analysis process come from a pilot project (Coolen and

Hoekstra 2001). The purpose of this pilot was to investigate the feasibility of the means-end approach for research on housing preferences. For the pilot project ten respondents, who were considered knowledgeable with respect to housing, were interviewed. All interviews took place in the home of the respondent by one of the researchers.

4.2.2.1 Elicitation and Selection of Attributes and Attribute Levels

The first phase in measuring means-end chains concerns the elicitation of relevant attributes for the laddering interview. Often the Repertory or Kelly Grid (Kelly 1955) is used for this. In this procedure the respondents are presented with a lim-ited number of triads with constantly differing products from a particular product field. For every triad they must indicate in what way two of the three named prod-ucts are similar to each other and consequently in what way they differ from the third product. By repeating this procedure for every triad, it becomes clear what the salient characteristics of the product field are for the respondent and what the poles of each of these characteristics are. For instance, a respondent presented with three dwellings may indicate that two of the dwellings are alike because they are detached houses, and that they differ from the third dwelling because that is a non-detached house. Here the salient characteristic is the level of detachedness of a dwelling and the poles are detached and non-detached. This method is often used when the relevant attributes are unknown. In addition, the method can be readily implemented if one is dealing with a relatively homogeneous product field and/or if a product field consists of readily recognizable brands, because the procedure will then most likely result in a relatively small number of salient features.

However, this does not apply to the product field of housing. A house is an extremely heterogeneous product and brands are hardly known. Moreover, much is known about relevant housing attributes (Clark and Dieleman 1996). That is why we worked with lists of housing attributes ourselves, based on the relevant litera-ture and the goal of the research.

The second phase comprises the selection of attributes. The respondents were assigned the task of selecting from the list of attributes those that were most impor-tant for them. In addition, they had the possibility to mention attributes they con-sidered important that were not on the list. No limit was set to the number of attributes that could be chosen. If a respondent chose more than eight attributes, he/

she was then assigned the task of selecting the eight most important ones. This was done because otherwise the interview burden from the laddering interviews might become too great for the respondents.

In the third phase, the respondents were asked which level of each of the selected attributes they preferred. If for example the number of rooms was a selected attri-bute, then the respondent was asked how many rooms he/she would like. The pre-ferred level, which serves as the starting point for a laddering interview, was determined for every selected attribute.

4.2.2.2 Laddering Interviews

The key phase in measuring means-end chains is the fourth. In this phase the actual means-end chains are determined. For this purpose, a semi-structured interviewing technique known as laddering is used. It involves a tailored interviewing format using primarily a series of directed probes, typified by the ‘Why is that important to you?’ question, with the express goal of determining the links between the essential elements of a means-end chain: attributes – consequences – values. A respondent who states that he/she wants a house with five rooms would then be asked: ‘Why do you find it important that the house you want should have five rooms?’. The why question is repeated as a reaction to the answer of the respondent.

The process stops when the respondent can no longer give any more answers to these why questions. Letting the interview begin at the concrete level of the attri-butes and then continuously asking why allows the underlying consequences and values of a certain choice to be brought into the open. In this way, a means-end chain can be determined for each respondent and each attribute level; such a chain is called a ladder. A ladder shows the underlying reasons of the preference for a certain attribute level. This yields insights into the classifications employed at higher levels of abstraction and may reveal how the properties of goods are pro-cessed from a motivational perspective.

Since the respondents are asked to be introspective and to talk about their moti-vations, a non-threatening interview environment must be created. This can be facilitated during the introduction to the interview by pointing out to a respondent that in the context of this type of research there is no such thing as a correct or incorrect answer. It is primarily the respondent’s opinion that is important. Thus, the respondent is positioned as an expert and the interviewer fulfils the role of a facilita-tor, who has to keep the respondent talking. Furthermore, it is of great importance that the interviewer is able to identify the relevant elements of the respondent’s answers. This means that the interviewer needs to be fully acquainted with the means-end chain model and the content matter to which the interview refers.

The ten laddering interviews we performed in the pilot study were recorded on tape and subsequently transcribed. The researchers performed most of the transcriptions themselves. In the case that someone else performed a transcription, one of the researchers checked it thoroughly. During our interviews, respondents quite frequently gave so-called forked answers (Grunert and Grunert 1995). This means that several consequences are linked to only one attribute. According to Grunert and Grunert (1995), this occurs most often with respondents who have thought thoroughly about a certain preference or decision and consequently have an extensive meaning structure in the area concerned. This is almost certainly the case for our knowledgeable respon-dents. However, the high incidence of ‘forked answers’ in our pilot project might also be specifically related to the product field of housing. After all, a house is a good in which the consumer is seriously involved, which makes preferences and decisions in this area mostly well thought through. If respondents gave a forked answer, efforts were made to determine a separate ladder for every named consequence.

4.2.2.3 Constructing Means-End Chains: from Interviews to Ladders In the fifth phase, the means-end chains are determined on the basis of the inter-views. The raw data generated by the laddering interviews are the (transcribed) verbalizations of the respondents. First, a content analysis was carried out on these free responses. This resulted in a set of ladders for each respondent. Subsequently, the elements of these means-end chains were coded, dividing them according to topic and level in the hierarchy (attribute, consequence, value). In this process, several choices about the interpretation of the various elements of the ladders had to be made. To reach as much intersubjectivity as possible, several researchers were involved in the construction of the ladders from the interviews and the subsequent coding of these ladders. Four researchers constructed and coded the ladders of the first four interviews. After that, the ladders each researcher had constructed and coded were compared with each other in two sessions in which all four researchers participated. Possible differences were discussed until agreement was reached.

Furthermore, this consultation process resulted in a coding scheme for the remain-ing six interviews. For these interviews, ladders were first constructed and coded by two researchers separately. Subsequently, the results were compared with each other and differences were resolved. For the coding of the values that appeared in the laddering interviews, the value domains and values of Schwartz (1992, 1994) were used as a frame of reference. All the values found fitted into this framework.

Some examples of ladders that were derived from the interviews are shown in Fig. 4.2. In this figure, all the means-end chains start at the level of attributes and end at the level of values. However, this does not necessarily have to be the case (see also Fig. 4.3). Sometimes the value level is not reached and the chain stops at the level of consequences. There may be several reasons for this. Firstly, it is

five rooms

possible that the interviewees got stuck at the level of consequences. Secondly, the interviewers may not have pursued the questioning deeply enough, which, consid-ering their unfamiliarity with the laddconsid-ering method, is not inconceivable. Thirdly, it may be the case that the respondent’s motivations for a certain attribute are only formed by consequences.

The most remarkable thing about the ladders in Fig. 4.2, though, is the fact that the consequences that are mentioned differ tremendously. Some are functional (a room for every family member), while others are more psychosocial in nature (place to retire). Several consequences are rather concrete (gardening, room for every family member) and others more abstract (social contacts). We also note that some of the consequences concern everyday activities (gardening), while many others (between inside and outside, a room for every family member) do not.