Attribute Framing

What it is

The label on a product is a description of a feature, and a feature can be described by what is present or by what is absent. Lean and fat are two ends of the same measurement, so a shopper who reads either label has been given the same fact. Attribute framing is the finding that the two descriptions are nevertheless not treated alike. The positive wording tends to produce the more favourable judgement.

The term comes from a typology published in 1998 by Irwin Levin, Sandra Schneider and Gary Gaeth. Their starting point was that studies reported as framing effects had produced mixed results because they defined framing in different ways and so measured different underlying processes. They separated three types. In risky-choice framing, the wording changes willingness to take a gamble, as in the Asian disease problem that Amos Tversky and Daniel Kahneman published in 1981. In attribute framing, the wording changes how one characteristic of an object or event is evaluated. In goal framing, the wording changes how persuasive a message is. The authors argued that the typology resolves some of the inconsistencies in the literature, and that all three are best understood through the cognitive and motivational consequences of encoding information as positive or negative.

The distinction matters for how a finding is cited. A result about beef labels is not a result about gambles on lives, and a result about gambles says little about labels. Each type has its own stimuli, its own measures and its own average size.

In effect

The founding example is older than the typology. In 1988 Irwin Levin and Gary Gaeth published a study in the Journal of Consumer Research in which consumers rated several qualities of ground beef that was labelled either 75 per cent lean or 25 per cent fat. The lean label received the more favourable evaluations. The authors also looked at what happened after the consumers had tasted the beef. The effect of the label was smaller. They explain the pattern with an averaging model, in which an evaluation combines the framed information with the consumer's own experience of the product, so that a diagnostic experience dilutes the influence of the wording.

That result carries a useful limit. Attribute framing is strongest where the person has nothing else to go on. When the product can be tried, the label competes with the evidence of the senses and loses some of its weight.

Later work asked why the positive frame wins and when it does not. Chris Janiszewski, Tim Silk and Alan Cooke argued in 2003 that the usual preference for positively framed attribute information, which is often explained by the emotional associations of each frame, also depends on the range and level of reference values that people use to judge the attribute. In their account, when the positive frame invites a narrower range of reference values than the negative frame, values above expected performance favour the positive frame while values below expected performance favour the negative one. A wider range for the positive frame reverses that pattern. They also report that experience with a frame is one factor that narrows the range of reference values. The implication is that the direction of an attribute framing effect is not fixed by the words alone.

On size, the best available figure is a pooled analysis. Adelson Piñon and Hilda Gambara combined 51 studies, with 151 effect sizes and nearly 13,500 participants, published between 1997 and 2003. Risky-choice and goal framing had similar average effects. Attribute framing had the smallest, and results varied considerably from study to study, which the authors describe as clear heterogeneity. They judged publication bias unlikely to threaten the findings.

What it does not say

It does not say that consumers are permanently misled. The 1988 study found the gap narrowed once people tasted the product, and the 2003 account makes the direction of the effect depend on the reference values a shopper brings. A label shifts a judgement at the point of reading. It does not necessarily survive contact with the product.

It does not say the effect is large or uniform. The pooled analysis found attribute framing the weakest of the three types, with wide variation between studies. Citing the beef result as if it showed a fixed size of effect would overstate it.

It does not carry the evidence of the better-known replication. The Many Labs project that tested the Asian disease problem in 36 samples tested risky-choice framing. Its result, an effect in the original direction at about half the original size, is not a result about attribute labels.

And it does not tell anyone which description is the correct one. Seventy-five per cent lean and 25 per cent fat are both accurate. The finding concerns how people respond to equivalent descriptions, not which seller is honest. This page rests on journal abstracts and a publisher record for the 1988, 1998, 2003 and 2005 papers and has not been checked against the full texts.

---