Reliability and the number of rating-scale categories.

Bendig, A. W.

doi:10.1037/h0055647

Cited by 148 publications

(101 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Eleven as the number of preference points to be assigned in each pair may be considered a high number 2 if one takes into account that relevant methodological studies have typically considered a scale length ranging between five and eleven (i.e., [5,11]) (see, e.g., Alwin 1997;Bech et al 2006;Bendig 1954;Böcker 1988;Churchill and Peter 1984;Friedman and Amoo 1999;Hofmans, Theuns and Mairesse 2007;Komorita and Graham 1965;Lozano, García-Cueto and Muniz 2008;McKelvie 1978;Preston and Colman 2000;Roth, Schroeder, Huang and Kristal 2008;Weng 2004). The high number eleven may be beneficial in that it allows (at least potentially) a high level of precision (when indicating one's preference), but -at the same time -it may put too high a cognitive burden on respondents, especially if (virtually) the same level of precision (accuracy) can be achieved with a lower number of scale points, such as nine, seven, or five (Moors 2008 some of which do not increase precision of measurement.…”

Section: Introductionmentioning

confidence: 99%

On the optimal number of scale points in graded paired comparisons

2012

View full text Add to dashboard Cite

In market research, it is common practice to measure individuals' brand or product preference through graded paired comparisons (GPCs). One important decision concerns the (odd) number of scale points (e.g., five, seven, nine, or eleven) that has to be assigned to either brands or products in each pair. Using data from an experiment with 122 students, we assessed the extent to which GPCs with a higher number of scale points (requiring more cognitive effort) really outperform GPCs with a smaller number of scale points (requiring less cognitive effort). Our data analyses have shown that one may reduce the (odd) number of scale points from eleven to nine or seven, depending on what minor compromises one is willing to make. The detailed psychometric results presented in this paper are useful to applied researchers as they help them in making well-informed decisions on the number of scale points in a GPC task. On the optimal number of scale points in graded paired comparisons RESEARCH NOTEOn the optimal number of scale points in graded paired comparisons Abstract In market research, it is common practice to measure individuals' brand or product preference through graded paired comparisons (GPCs). One important decision concerns the (odd) number of scale points (e.g., five, seven, nine, or eleven) that has to be assigned to either brands or products in each pair. Using data from an experiment with 122 students, we assessed the extent to which GPCs with a higher number of scale points (requiring more cognitive effort) really outperform GPCs with a smaller number of scale points (requiring less cognitive effort). Our data analyses have shown that one may reduce the (odd) number of scale points from eleven to nine or seven, depending on what minor compromises one is willing to make. The detailed psychometric results presented in this paper are useful to applied researchers as they help them in making well-informed decisions on the number of scale points in a GPC task.

show abstract

Section: Introductionmentioning

confidence: 99%

On the optimal number of scale points in graded paired comparisons

2012

View full text Add to dashboard Cite

show abstract

“…Several researchers come to the exact opposite conclusion; namely that binaryanswer formats are preferable or that, at least, how many answer options respondents are offered does not make much difference (Bendig, 1954;Dolnicar & Grün, 2007b, 2007cDolnicar, Grün & Leisch, in press;Komorita & Graham, 1965;Matell & Jacoby, 1971a;1971b;Martin, Fruchter & Mathis, 1974;Schutz & Rucker, 1975). For example, Peabody (1962) concludes that the six-point item format reflects "primarily the direction of responses" (p.73), which is captured equally well by the binaryanswer format.…”

Section: Arguments In Support Of Binary-answer Formatsmentioning

confidence: 99%

“…When measuring evaluative beliefs, the binary-answer format achieves greater concurrent validity than the seven-point format (Bendig, 1954;Dolnicar, 2003;Komorita & Graham, 1965;Matell & Jacoby, 1971a;1971b;Martin, Fruchter & Mathis, 1974;Peabody, 1962;Schutz & Rucker, 1975).…”

Section: H2mentioning

confidence: 99%

One legacy of Mazanec: binary questions are a simple, stable and valid measure of evaluative beliefs

Dolničar

Leisch

2012

International Journal of Culture, Tourism and Hospitality Research

View full text Add to dashboard Cite

One legacy of Mazanec: binary questions are a simple, stable and valid measure of evaluative beliefs AbstractPurpose -Academic researchers love multi-category answer formats, especially five-and seven-point formats.More than a decade ago Josef Mazanec concluded that these formats may not the best choice, and that simple binary-answer options are preferable in some empirical survey contexts. The purpose of the present study is to investigate empirically Mazanec's hypothesis in the context of the measurement of evaluative beliefs relating to fast-food restaurants.Design/methodology/approach -The authors conducted an online experiment that asked respondents to assess evaluative beliefs relating to fast-food brands using either a forced binary (n = 100) or a seven-point answer format (n = 100). The authors also measured preferences for each of the fast-food restaurants, user friendliness, and recorded the actual completion times for the survey.Findings -The results indicate that the full binary answer format outperforms the popular seven-point multicategory format with respect to stability, concurrent validity, and speed of completion.Practical implications -Given the demonstrated strengths of full binary measures, they should be used more by both practitioners and academics when measuring evaluative beliefs.Originality/value -This study provides empirical evidence of the strong performance of the forced binaryanswer format for the measurement of evaluative beliefs, and thus challenges current measurement practice among academics and practitioners. AbstractPurpose -Academic researchers love multi-category answer formats, especially five-and seven-point formats. More than a decade ago Josef Mazanec concluded that these formats may not the best choice, and that simple binary-answer options are preferable in some empirical survey contexts. The present study empirically investigates Mazanec's hypothesis in the context of the measurement of evaluative beliefs relating to fast-food restaurants.Design/methodology/approach -We conducted an online experiment that asked respondents to assess evaluative beliefs relating to fast-food brands using either a forced binary (n = 100) or a seven-point answer format (n = 100). We also measured preferences for each of the fast-food restaurants, user friendliness, and recorded the actual completion times for the survey.Findings -Results indicate that the full binary-answer format outperforms the popular seven-point multi-category format with respect to stability, concurrent validity, and speed of completion.Practical implications -Given the demonstrated strengths of full binary measures, they should be used more by both practitioners and academics when measuring evaluative beliefs.Originality/value -This study provides empirical evidence of the strong performance of the forced binary-answer format for the measurement of evaluative beliefs, and thus challenges current measurement practice among academics and practitioners.

show abstract

“…3 -7 Researchers have conducted empirical studies to investigate scale design issues such as the inclusion of a scale midpoint, 6,8 using labeled versus unlabeled response categories, 9,10 and so forth. Considerable extant research dating back to the works of Bendig 11 and Symonds 12 has addressed the fundamental question of how many response categories to include in a scale. However, these investigations have addressed this question in the contexts of scale reliability and validity.…”

Section: Introductionmentioning

confidence: 99%