Bipolar Scale

Survey researchers frequently employ rating scales to assess attitudes, behaviors, and other phenomena having a dimensional quality. A rating scale is a response format in which the respondent registers his or her position along a continuum of values. The bipolar scale is a particular type of rating scale characterized by a continuum between two opposite end points. A central property of the bipolar scale is that it measures both the direction (side of the scale) and intensity (distance from the center) of the respondent's position on the concept of interest.

The construction of bipolar scales involves numerous design decisions, each of which may influence how respondents interpret the question and identify their placement along the continuum. Scales typically feature equally spaced gradients between labeled end points. Data quality tends to be higher when all of the gradients are assigned verbal labels than when some or all gradients have only numeric labels or are unla-beled. Studies that scale adverbial expressions of intensity, amount, and likelihood may inform the researcher's choice of verbal labels that define relatively equidistant categories.

Both numeric and verbal labels convey information to the respondent about the meaning of the scale points. As shown in Figure 1, negative-to-positive numbering (e.g. −3 to+ 3) may indicate a bipolar conceptualization with the middle value (0) as a balance point. By contrast, low-to-high positive numbering (e.g. 0 to +7) may indicate a unipolar conceptualization, whereby the low end represents the absence of the concept of interest and the high end represents a great deal. The choice of gradient labels may either reinforce or dilute the implications of the end point labels.

Figure 1 Example of bipolar scale

While negative-to-positive numbering may seem the natural choice for a bipolar scale, this format has a potential drawback. In general, respondents are less likely to select negative values on a scale with negative-to-positive labeling than they are to select the formally equivalent values on a scale with low-to-high positive labeling. Similarly, bipolar verbal labels result in more use of the midpoint and less use of the negative values than when unipolar verbal labels are used. Systematic reluctance to select negative values shifts the distribution of the responses to the positive end of the scale, yielding a relatively high mean score. In addition, the spread of the responses attenuates, yielding a reduction in variance.

The number of gradients represents a compromise between the researcher's desire to obtain more detailed information and the limited capacity of respondents to reliably make distinctions between numerous scale values. Research suggests that 7-point scales tend to be optimal in terms of reliability (test-retest) and the percentage of undecided respondents. Thus, 7-point scales plus or minus 2 points are the most widely used in practice.

Scales featuring a large number of labeled gradients may be difficult to administer aurally, as in a telephone interview. A common solution is to decompose the scale into two parts through a process called “branching” or “unfolding.” The respondent is first asked about direction (e.g. Overall, are you satisfied or dissatisfied?) and then about degree (e.g. Are you extremely (dis)satisfied, very (dis)satisfied, somewhat (dis)satisfied, or only a little (dis)satisfied?). In certain multi-mode studies, branching may also be used to increase the comparability of responses across different modes of administration. In self-administered modes and face-to-face interviewing, respondents are often provided with a pictorial rendering of the scale, but respondents in telephone interviews usually cannot be provided with such visual aids. Administering a common branching question in each mode reduces the effect of mode on respondents' answers.

...