I suspect that most readers not experienced in experimental psychology would still wholeheartedly accept the strongly supported principle that “bad” is stronger than “good” when it comes to human perception and evaluation. When we encounter any “bad” news, whether about a potential friend, purchase, or even ourselves, we give it more weight, credence, and consideration compared to equally “good” news.
A few years ago, Roy Baumeister, Kathleen Vohs, Ellen Bratslavsky, and Catrin Finkenauer reported one of the best literature review and synthesis articles I’ve ever read. They scanned the research looking at a wide variety of comparisons of the impact of bad versus good and found something fairly surprising in psychology: a strong main effect. In every area of application they investigated, bad always had greater impact than good. (This is a narrative review, so a general reader can easily understand it.)
Most interesting in this review was a particular absence – nothing specifically about persuasion. While the review looked at areas that focus on attitudes – if you think about it the entire idea is about attitudes, the bad versus the good is nothing but evaluative responding – there was no section that looked at how bad versus good operates with the familiar persuasion WAC trinity of WATTage, Arguments, and Cues.
So, now we reach the pivot in this story: How can we use “bad” to our persuasion benefit? It’s trickier than you think.
First, you might recall a previous post on Dan O’Keefe’s meta-analysis of message framing on health behaviors. O’Keefe and colleagues looked at the Bad is Stronger principle within a comparison of loss-framed health messages versus gain-framed health messages. You’ll recall they found a pitifully small effect size of r = .039. This is hardly a ringing confirmation of Bad Is Stronger, so what’s going on?
As I noted in that earlier post, I think the meta failed to confirm the obvious main effect that bad is stronger than good because in persuasion applications, at least, things are always more complicated than a simple main effect. Basic dual process theory, whether ELM or HSM, always argues for interactions, that persuasion rarely depends upon a simple main effect, but almost always depends upon interactions within combinations of variables. Health researchers typically do not measure or manipulate important ELM variables like WATTage (elaboration likelihood) or fail to include both strong and weak arguments within a gain or loss frame. This absence creates an empirical mashup of outcomes where sometimes the study favors gain frames, then loss frames, then neither (a classic outcome in persuasion research prior to the elucidation of dual process theory in the 1970s).
A great illustration of this complexity with frames in persuasion is found in the Smith and Petty 1996 study of framing and the ELM. In two lab experiments Smith and Petty manipulated WATTage, frames, and argument quality. Their results did not show a simple main effect for frames (bad stronger than good), but rather an interesting interaction. Study 1 demonstrated that frames operated as WATTage switches. When participants read either strong or weak arguments in favor of recycling (Green alert! Consider persuasion science rather than Mark Twain) WITHIN a gain frame, there was no difference in attitude outcomes. This is crazy, of course, because it means that people weren’t reading and thinking about the arguments since weak arguments created as much attitude change as strong arguments. By contrast, WITHIN the loss frame, argument quality mattered. Strong arguments produced more positive change while weak arguments produced negative change (r = .27, a near Medium Windowpane effect), exactly what should happen if people are reading and thinking about the information.
Thus, while bad is stronger than good in many applications, using it in persuasion requires a bit of planning.
So, now, after all this ticky-tacking, what do we know about practical persuasion with bad versus good (let’s call it, BvG)?
First, BvG clearly can function as a WATTage dimmer switch. If you want to dial up your receivers, alert them to “bad” news. As the Smith and Petty studies demonstrate, that will cause people to look for argument quality, distinguish strong from weak, then elaborate on the arguments to guide attitude change. Within this line of reasoning, one could expect that extremely high levels of “bad” news could turn objective processing into biased processing, too. Thus, if you know your receivers are already pretty much on your side, hit them with an extreme statement of the “bad” news to elicit defensive, but favorable responding. In both cases, the bad news demands that you have strong arguments. If you do this and then provide weak arguments (or elicit biased processing against your position), you will kill yourself.
Second, the large remainder of the bad is stronger than good evidence from the Baumeister review clearly demonstrates huge possiblilities for persuasion Cues. When receivers are Low WATT, bad news in all forms will typically produce negative change. Consider these quick illustrations from the CLARCCS Cues.
Comparison – if everyone else says it is bad, it is bad.
Liking – if a liked source says it is bad, it is bad.
Authority – if an expert says it is bad, it is bad.
Reciprocity – if someone gives you something bad, give back something worse.
Commit/Consist – if you support bad, you must defend bad.
Scarcity – when bad is rare, it is even worse.
Third, using BvG for argument quality seems weird and unproductive to me. I cannot easily construct “bad” strong arguments and “good” strong arguments and see how it could make a serious difference if they were both equally “strong arguments.” Both would produce the same amount of change, but, of course, in different directions (“bad” strong args would produce negative change and “good” strong args would produce positive change). The test of BvG is not direction, but intensity or amount of change.
[Research Sidebar: It would be interesting to see if "bad" strong arguments required less information than "good" strong arguments. Just hearing "ex-convict" within a string of otherwise positive attributes about a person is usually enough to produce a negative attitude. It might be easier to construct "bad" strong arguments, but as long as "bad" and "good" args are equally strong, the BvG effect does not seem to hold.]