Page 240 - The Handbook of Persuasion and Social Marketing

Empirical Generalizations for Social Marketing 221

Comparative advertising shows the strongest average effect on behavior,
r = .22 (Grewal, Kavanoor, Fern, Costley, & Barnes, 1997). Testimonial
assertion evidence, “in which a source introduce[s] material from an out-
side source in an effort to support a position,” correlates with attitude
change at r = .25 (Reinard, 1998, p. 71). Powerful language is more per-
suasive (r = .23) and credible (r = .21) than powerless language, which
“reflects an abundance of hedges, hesitation forms, polite forms, and ques-
tioning intonations” (Burrell & Koper, 1998, p. 204). However, the studies
underlying the results mostly involve courtroom settings, which may make
them more relevant to personal presentation formats than to persuasive
mass media messages when it comes to social marketing.
Varied results across reviews may raise questions about how generaliz-
able estimated effect sizes actually are. Fear appeals and warning labels
tend to show average effects between r = .12 and r = .19, but one exception
shows r = .05 for fear and intentions (Keller & Lehmann, 2008). Two re-
views of two-sided messages show r = .14, whereas a third, larger study
shows a much lower r of .03 (Allen, 1998). The most variable results are
found in comparisons of gain framing (positive effects of compliance) with
loss framing (negative effects of noncompliance). One study shows a posi-
tive effect of gain framing at r = .15 (Kühberger, 1998), but five other stud-
ies show effects of r = .08 or less, with gain framing sometimes more and
sometimes less effective than loss framing.
Three other approaches produce effects of r = .05 or less: explicit con-
clusions, rhetorical questions, and humor. Thus, these approaches have, at
best, a small positive effect on average, but neither do they consistently
hurt the performance of persuasive messages.
It is worth noting that multiple other studies have examined the effects
of advertising executional characteristics using secondary data, sometimes
with samples of hundreds of actual ads (Mothersbaugh, Huhmann, &
Franke, 2002; Stewart & Furse, 2006; Stewart & Koslow, 1989). Although
these studies may offer useful generalizations for social marketing, they are
not meta-analyses of multiple authors’ research findings and are not in-
cluded in this review.

Advertising Elasticities

Advertising elasticities for a wide variety of products, plus separate estimates
for alcoholic beverages and cigarettes, are shown in Table 8.4. Two general
surveys showed selective (brand) elasticities declining from e = 0.22 in 1984
(Assmus, Farley, & Lehmann, 1984) to e = 0.12 in 2011 (Sethuraman, Tellis,
& Briesch, 2011). A 1994 estimate of selective demand for cigarettes had an