We are more likely to believe something is true, if it is framed in a negative way.
Hilbig (2009) performed three experiments to find whether there was a link between framing and the perceived truth of a statistic.
The first experiment was conducted as an online-survey. After providing consent and demographic information, participants were shown statistical information from the crime domain and instructed to provide a truth rating. As information, the success rate of crimes from the category of rape and aggravated sexual coercion (denoted ‘rape’ in what follows) was presented. The actual success rate (85%) was used. Half of the participants were told that 85% of attempted instances of rape were successful (negative frame), while the other half were told that 15% were unsuccessful (positive frame). All participants were then asked to judge the truth of the stated information on a 4-point scale.
The mean rating of truth from the negative frame group was significantly higher than that of the positive frame group. The second experiment replicated these results.
Following the logic of Experiment 1, the information frame was again manipulated. Thirty eight participants (30 female, aged M = 17.3, SD = .50, recruited from a high school course of introductory psychology) were randomly assigned to two groups. These were shown the actual clearance rate of rape (70%), either framed positively (70% of cases cleared) or negatively (30% of cases not cleared) and asked to judge, again on a 4-point scale, the truth of the provided statement.
Again, participants rate the statistic as more likely if it emphasised the negative.
The principal logic of [the third experiment] was again to manipulate the frame of the information presented (between participants) while holding the actual validity constant. In contrast to the previous experiments, the information was not from the crime domain but from demographics. Specifically, participants were shown the probability of a marriage to be divorced within the first 10 years which is, in Germany, about 20% (Federal Statistical Office, n.d.). Participants were randomly assigned to one of two conditions: in the positive frame, they were informed that 80% of marriages lasted 10 years or longer whereas their counterparts in the negative frame were informed that 20% of marriages were divorced within the first 10 years. Like in the previous experiments, participants rated the truth of this statement on a 4-point scale. The experiment was run using simple 1-page questionnaires dispersed to a community sample of 33 participants.
This third experiment also clearly demonstrated negativity bias. The exact results of each study are shown in the graph below.
Fig. 1. Mean truth ratings (original scale ranging from 1 to 4) for the negative vs. positive framing conditions in each of the experiments. Error bars represent one standard error of the mean.
However, this effect is not confined to estimations of truth.
First, it has been argued that negative instances are often more informative (Peeters & Czapinski, 1990) – parallel to the higher informativeness of disconfirming evidence (Leyens & Yzerbyt, 1992). So, there could be a simple direct association between valence and (perceived) veracity.
We also tend to dwell on the negative and discuss it with each other.
Secondly, there is evidence for increased elaboration of negative instances which has been termed ‘informational negativity effect’ (e.g. Lewicka, 1997; see also Lewicka, Czapinski, & Peeters, 1992).
Negativity bias doesn’t operate in its own, but alongside many other effects.
Finally, there is a noteworthy body of literature which confirms that more elaboration, deeper processing, and high processing motivation can increase the persuasiveness of messages (e.g. Petty and Briñol, 2008 and Shiv et al., 2004). Similarly, though investigating the realm of wishful thinking rather than negativity bias, Bar-Hillel, Budescu, and Amar (2008) showed that the causal link ‘I focus on, therefore I believe in’ (p. 283) is well-supported. Also, elaboration can increase the perceived truth of past-events, even and especially when these never happened, which has been explained as an effect of constructive processing (Kealy, Kuiper, & Klein, 2006).