THE DECLINE OF “IRRATIONALITY”
A FUNNY THING has happened to the concept of “irrationality”: Over the past four decades, thousands of experiments have suggested that by changing things like the context or framing of an offer, marketers can easily sway people to act in “irrational” ways. In the past few years, this idea has started percolating into the mainstream. Ironically, just as the concept is gaining popularity, it’s becoming less representative of reality.
It’s easier to be rational when you rely on absolute values.
No, we’re not suggesting that people will start behaving fully rationally (the way economists assume they would) anytime soon, that any basic cognitive abilities are changing, or that they will be totally immune to manipulation. (And this, of course, includes the authors of this book. We have our share of embarrassing stories where we bought lemons or useless products but made every effort to use them to “get our money’s worth”). But a shift is indeed taking place as a by-product of the new reality.
Putting it all in historical perspective, we can say this: In the beginning, economists and other scholars believed that people are generally rational utility-maximizers, know exactly the value of each product feature, and that those susceptible to decision errors will learn from their mistakes. Over the past forty years, many decision researchers have portrayed people as inherently “irrational” and error-prone (and to set the record straight, Itamar is among those responsible for the portrayal of consumers as susceptible to various seemingly irrational influences).1 They showed in numerous (often intriguing) lab experiments that with the right framing or context, an experimenter can cause people to reverse their preferences. In this book we try to go a step further: We of course recognize that people’s minds have limitations that won’t change, but we examine what happens to these limitations in a new environment that is less hospitable and conducive to such “irrationality” effects.
We need to clarify the term “irrationality” in our context. We put the word in quotation marks because, as Daniel Kahneman points out in his book Thinking, Fast and Slow, the word connotes impulsivity, emotionality, and a stubborn resistance to reasonable argument. Without quotation marks, it’s too strong a word in our context. In fact, Kahneman writes: “I often cringe when my work with Amos [Tversky] is credited with demonstrating that human choices are irrational, when in fact our research only showed that Humans are not well described by the rational-agent model.”2
So what do scholars mean when they say someone is “irrational”? Probably the least ambiguous definition is related to behavior that is inconsistent with the economic concept of value maximization. For example, a consumer who selects an option that is clearly inferior or switches from choosing A over B to choosing B over A due to seemingly irrelevant factors is said to exhibit “irrationality.” Are people who believe in UFOs irrational? Not in this context. A decision is regarded as irrational if it clearly shows that the person’s decisions are inconsistent, incoherent, or unambiguously represent an inferior choice for that person given his or her beliefs, values, and preferences. While we may be critical or even ridicule the belief in UFOs, there’s nothing “irrational” about it as the term is used in this field.
Numerous studies have demonstrated that consumers’ decisions can be influenced in different ways. These influences fall into three broad categories—framing effects, choice context effects, and task effects. “Irrationality” studies are rather intriguing and new studies keep coming, but how representative are they of today’s reality? What happens to these theories of choice manipulation and influence in the “noisy,” information-rich, and socially intensive environment that is developing? We argue that the new environment significantly changes how things work. First, the relevance of these influence tactics has diminished in a world where people can easily assess quality. On average, better decisions are being made based on the information that’s available. Second, the noise that all this information creates has a surprising effect. In the next few pages we’ll discuss each of the categories—framing, context, and task—and explain why they are less relevant in the new information environment.
FRAMING EFFECTS
On the same day in March 2012, USA Today announced that the Federal Reserve annual stress test failed four of nineteen big banks, while the New York Times’ headline read: “15 of 19 Big Banks Pass Fed’s Latest Stress Test.”3 So in this case an editor at USA Today chose to frame the news more negatively than a counterpart at the New York Times. It turns out that framing can significantly affect our perceptions, memories, and the choices we make. An intriguing study illustrates how framing can affect even the perception of taste. In an experiment conducted at the University of Iowa by Irwin Levin and Gary Gaeth, participants rated beef that was presented to them as “90 percent lean” as better tasting than beef that was presented as containing “10 percent fat.” It was the same beef and the same information. Clearly, the taste of beef should not depend on how it’s labeled. But when the researchers tweaked the message to sound more positive, those people who saw it as “lean” tended to like it better.4
Framing still works just fine in the lab, and many marketers continue to believe that they can affect people’s perceptions by framing something positively. But what happens to this effect in today’s reality? In 2012, we saw an example of how framing works today (or doesn’t): There’s a food product used in the United States that’s produced from the bits and ends left over in the butchering process. The fat is removed by heating and spinning, and then this leaner mix is treated with ammonium hydroxide to kill bacteria. The makers of the product found a nice-sounding name for it—“lean, finely textured beef”—and, for a while, framing seemed to be working like in a textbook.
Except that in 2002 a microbiologist at the U.S. Department of Agriculture by the name of Gerald Zirnstein offered a less flattering framing for the product. In an email message to colleagues, he referred to it as “pink slime” and added: “I do not consider the stuff to be ground beef, and I consider allowing it in ground beef to be a form of fraudulent labeling.”5 As long as Zirnstein’s framing was not made public, the makers of the product were doing fine, but in 2009, the New York Times published the term in an article. Sometime later, celebrity chef Jamie Oliver discussed the topic on his TV show and a blogger named Bettina Siegel, whose blog “The Lunch Tray” focuses on kids’ food, picked the term and started an online petition asking Secretary of Agriculture Tom Vilsack to stop using “pink slime” in school lunches. After just nine days, more than two hundred thousand people signed the petition and the USDA announced that it would offer school districts a choice of beef without the product. With additional coverage on ABC News, public perception of the product shifted, and the largest U.S. supermarket chains announced that they would no longer sell products containing “lean, finely textured beef.” The bottom line is this: Framing works in the lab where people are not exposed to any alternative frames. When consumers rely on information from multiple sources, framing effects are likely to be weaker and, in some cases, dissipate completely. Of course, marketers, such as political consultants, will continue to look for just the right terms that might help promote their causes and hurt competitors, and in some cases they will be successful. But frames offered by marketers face tougher competition and are sometime neutralized by alternative frames.6
CHOICE CONTEXT EFFECTS
Remember the camera experiment from the previous chapter? As you recall, adding an expensive camera to the choice set caused some consumers to choose the moderately priced camera instead of the cheapest one. This is a perfect example for a context (or choice set) effect (the context of the offer affected people’s decisions). Let’s look at another one: Suppose you’re shopping for a shredder and you see two options. One shredder costs $20 and can shred up to 7 sheets of paper at a time. The second shredder costs $50 and can shred up to 11 sheets of paper at a time. Which one would you choose? If you care about cost you would probably choose the first one; if you care about the number of sheets, you would choose the one with higher capacity.
Now, suppose that you’re shopping under different conditions and you actually see three shredders: one at $20 that shreds 7 sheets at a time, the second at $50 that shreds 11 sheets, and a third shredder that costs $95 and can shred up to 12 sheets of paper at a time. Which shredder would you choose now? When Taly Reich and Itamar ran this study recently, adding the $95 shredder that shreds 12 sheets increased the number of consumers who chose the $50 shredder instead of the cheapest one. While this may look at first glance like another case of the compromise effect, what’s going on here is different. This is known as asymmetric dominance, an effect that was first shown in 1982 by Joel Huber and his coauthors from Duke University. Essentially, those who chose between just two options saw a cheaper shredder that shreds fewer sheets against a more expensive shredder that shreds a higher number of sheets. But adding the $95 shredder that shreds 12 pages made the $50 shredder look like a winner: This shredder shreds almost as many pages for almost half the price! Companies can (and do) use this effect (sometimes referred to as the decoy effect) in selling hard drives, MP3 players, and other products.
But what happens to this effect when people shop online with full access to information? As with the camera experiment, Taly and Itamar also tested this effect in “Amazon” conditions. As opposed to the two groups that saw the shredders in isolated conditions, there were two groups of participants that were put in more real-life conditions. They first saw what consumers usually see when they shop for a shredder: a variety of options and prices, plus some reviews written by consumers. Once they looked at all the options, they were asked to assume that they have narrowed their choice down. Now, participants in one of the groups saw the two shredders (costing $20 and $50), while participants in a second group saw three shredders (costing $20, $50, and $95).
What was the outcome? The asymmetric dominance effect was gone. Not a shred of the effect was left.
Here’s a story we came across that can help us further discuss the diminishing relevance of context effects in today’s world: Not long ago, an Israeli real estate agent met an acquaintance who was selling his apartment in Jerusalem. When the agent inquired about the asking price, the man said he’s listed the apartment for 2.1 million shekels, which surprised the experienced agent, who pointed out that similar apartments in the area usually sell for about 1.85 to 2 million shekels. With a mischievous wink, the man shared the little trick he had devised. In addition to listing his own apartment, he placed ads for three fictitious apartments in the same neighborhood for much higher prices ranging from 2.2 to 2.35 million. His idea was that potential buyers would see these ads and therefore feel that his asking price was a bargain. This is a context effect similar to examples we discussed earlier. Putting aside the serious ethical and legal questions that are beyond the scope of our book, let’s ask a simple question: How likely is this trick going to work in two different markets? In one market, buyers have no information whatsoever about past transactions. In the other market, buyers know every single detail about all past transactions. Clearly, the man’s trick is less likely to work in the second environment.7
While we’re still far away from “knowing every single detail about past transactions,” a buyer in the United States today can use several tools that will give her a very good idea of what’s a reasonable price for a property. A quick search on Zillow.com will give her an estimate for the value of a particular house that is based on past transactions around that property. She can then view the details of many of the houses sold in the area to develop a sense of what’s a reasonable price. Zillow is far from being perfect, but it limits sellers’ ability to play such tricks.8
TASK EFFECTS
The way consumers’ preferences are expressed and formed can also significantly affect people’s choices. It turns out that if you ask people to select a product one way, they will choose product A, and if you ask them in a different way, they will choose product B. In a study that Stephen Nowlis and Itamar ran, subjects were asked to select between two toasters. When people were asked to simply pick one of the two toasters, they picked the less expensive brand (Kmart store brand). When people were asked to rate the two toasters, they preferred the better-known brand name (Black & Decker). This concept has many practical implications to marketers. For example, many retailers present their private-label brands next to the corresponding national brands. Conversely, products with a main advantage that is more qualitative (a well-known brand, for example) and harder to compare than price are likely to sell better if they are presented in a way that makes it difficult for buyers to compare (for example at an end-of-aisle display, also known as an end cap).9
This type of effect is also likely to be weaker as shopping environments are changing. A marketer of an expensive product who pays extra for an end cap display can hope to avoid direct comparison. But things can be very different when consumers are armed with smartphones and apps such as ShopSavvy that let them scan the bar code of a product and display prices in other retail stores. An app like GoodGuide displays alternative products based on attributes the consumer cares about. When assessing the quality of offers is supported by these types of technologies, swaying people’s choices using “irrationality” tactics is far from being trivial.
Up to this point, we’ve seen how framing, context, and task effects are becoming less effective and less relevant. Part of the reason is that people are able to better assess the quality of things, but there’s another, less obvious reason. . . .
THE SURPRISING POWER OF NOISE
Influence and manipulation work best when there’s full control over what people are exposed to, and “noise” is kept to a minimum. What exactly do we mean by noise? Noise refers to any information that is not under control, often consisting of diverse pieces of data from different sources, about different aspects and options.
The following experiment hints at the difference between how people decide in isolation, and how they decide in a dynamic noisy environment. In 2006, Raymond Fisman, Sheena Iyengar, Emir Kamenica, and Itamar published a paper about speed dating. A few days before the actual speed dating event, participants were asked what factors were most important to them in a mate. Not surprisingly (and consistent with prior research) women were much less likely to say that physical attractiveness would be important. That was their theory when all they saw was a well-structured survey on a computer screen. But something completely different happened once these women participated in the actual event: They appeared to have forgotten their stated criteria. Attractiveness was almost as important for women as it was for men. Part of this may be explained by the fact that when speed dating, women couldn’t accurately predict the “earning potential” of men (which has often been found to be a key driver for women). Yet a more relevant factor in our context has to do with the noise associated with speed dating—up to twenty dates, each lasting four minutes, all happening in the same (literally) noisy room. Instead of a clean survey, you’re exposed to hundreds of diverse pieces of data, with different aspects and options.
The experiments we described earlier have one thing in common: Researchers in all cases had full control over what participants saw. For example, students who participated in the original 1992 camera experiment were sitting in a quiet classroom and focused on a piece of paper that featured either two or three cameras. That’s all. They couldn’t talk to each other, surf the Web, check their iPhones (who even dreamt of an iPhone back then?), or do anything else that would distract them. One of the guiding principles in planning such studies is limiting what the participants see, how they see it, and how much information they have.
We cannot think of an environment that is farther away from a quiet lab than the World Wide Web.
When the camera experiment was repeated in 2012 in lab-like conditions (that is, people were limited in what they saw), things worked as in the original experiment. But things worked differently when people were free to see other options. A follow-up study with these participants showed that they had poorer recall for the attribute values of the options. In other words, it wasn’t necessarily that they looked around, found a better camera, and decided to stick with it. It’s probably more likely that they saw so much stuff that the noise “spoiled” the effect.
The truth is that even that experiment was more structured than what typically happens in real life. Participants were limited to Amazon and were asked to focus on Canon PowerShot cameras (or on Fellows shredders in the shredders experiment). Consumer searches in real life are often even less structured. When we recently typed the word “camera” in Google, we immediately saw ten cameras from different makers, ranging in price from $69 to $945. What’s more—hundreds of consumer reviews were accessible from that page. These reviews create two effects. The first is straightforward—even though reviews are not perfect quality indicators, their content helps consumers assess the quality of offers. But the second effect can be as powerful: This avalanche of information creates a lot of noise—a lot of distractions that create situations that are quite different from those in sanitized lab experiments and “spoil” the necessary conditions for marketers to influence.
This is the surprising power of noise—in the current information environment, influencing people is a bit like trying to hypnotize someone while riding a motorcycle at 100 mph. There are just too many distractions.
We are certainly not proposing that lab studies of consumer decision making no longer serve a useful purpose in the current environment. We argue that, by and large, our ability to generalize from tightly controlled experiments (with limited information) to consumer decision making in reality has been diminished. The gap between the lab and reality is getting wider. In particular, lab “effects” that depend on tightly controlled reference points may often not apply to a noisy environment that is characterized by unpredictable reference points and widely different contexts. This means that lab results can often lead to misleading conclusions that misrepresent what happens in reality.
Buying products in the twentieth century was an experience that was more conducive to influence by marketers. Even though you weren’t as isolated as in a lab, you were usually in some controlled environment. You stood at the store in front of a limited number of dishwashers, or you looked at a catalog that came in the mail and focused on a few items on the page. You were usually confined to a small set. Things work very differently in today’s shopping mall and certainly online.
There’s an important exception worth noting. While many of those “irrationality” demonstrations are less relevant in the new environment, there’s a key ingredient that must be present in the consumer environment in order for this “de-biasing” effect to happen. Here’s a study that looked into this in the context of a standard framing effect. It is based on a well-known experiment that showed that people tend to reject an economic policy program when they are told that it will result in 5 percent unemployment, but to prefer the (same) program when they are told that it will result in 95 percent employment. Participants in this new experiment were asked to imagine that they were faced with the decision of adopting one of two economic policies, and here, too, people tended to reject the program when it was framed negatively (5 percent unemployment) and adopt it when it was framed positively (95 percent employment). Things started to get interesting when the researcher James Druckman from Northwestern University exposed participants to heterogeneous framings or asked to discuss the problem in small groups that consisted of some participants who were exposed to positive framing and some to negative. What happened then? The framing effect was eliminated.10
This is important. A key element to ensure that people are not susceptible to relative tactics is exposure to diverse sources, perspectives, options, and considerations. It’s not enough to hear others’ opinions. Susceptibility to influence by irrelevant reference points will decline only if others see different things or see things differently. People who belong to a cult will not be protected from their leader’s influence if they never talk to others outside the cult. This means that assessing quality in a monolithic cultlike environment does not make judgments and decisions more absolute—they remain locked in the shared frame. Raving fans of a brand (or a political leader) who only listen to like-minded fans may be as susceptible to relative tactics as in the past. In short, when most others in a situation have the same information, are in the same “condition,” face the same frame, or see the same option set, “irrationality” will prevail.
EXTRAPOLATIONS ON STEROIDS
Before we go on, we need to make a couple of general points about the field of behavioral decision making. The field has offered meaningful and important insights into how people make decisions, and as we have argued, some of the effects demonstrated in the field are declining because of a changing environment. But the truth is that there are also effects that were never that strong to begin with. Some of the more prominent examples that demonstrate “irrationality” had limited relevance under normal conditions even in the old times. Some studies reached broad conclusions that were based on extrapolations from rather narrow and unrepresentative tests. The boundary conditions and the limitations of the studies that are reported in the academic papers are usually lost in the popular press. All that is left is a great story for a cocktail party that doesn’t necessarily represent what’s really going on in the world.
For example, an influential early finding of a preference reversal: When you ask people to choose between two gambles (for example, a 50 percent chance to win $10 and 50 percent chance to lose $5, or a 5 percent chance to win $100 and 95 percent chance to lose $4), they usually choose the one with the better odds of winning. But if you ask them to price the gamble (how much would you sell each gamble for?), people usually price the one that offers the higher payoff option higher. This and similar findings, which have significant theoretical implications, have been relied upon to advance the idea that people generally don’t have preferences, so they tend to “construct” preferences on the fly based on what they happen to consider at the moment. But how much can we really learn from this example about everyday preferences? Pricing gambles is not something that people normally (or ever) do, so no wonder they make mistakes.
Some of these effects get attention and make good conversation topics. They are intriguing, no doubt. Researchers, authors, and journalists all know that surprising results and counterintuitive effects make good stories and these are exactly the stories that often get more attention than they deserve and are prone to exaggerations.
Another point that has to be made in this context: People outside the field assume that these academic findings about decision errors are usually as reliable and robust as those in the natural sciences. Some indeed are robust and tend to replicate from one study to the next, but many are not. Some findings are very sensitive to a particular laboratory test methodology, and it’s unfortunately not uncommon for researchers to try three different tests of an idea and report the one that works (or works best). Moreover, in theory, different methods for testing the same principle should lead to the same results, yet many of the reported findings are notoriously method-sensitive. So sometimes that great cocktail party story is based on shaky science. When the popular press reports on such findings, they usually neglect to mention that the intriguing effects operate under very specific, rather narrow conditions, and may not apply in most situations.
At the same time, we want to make sure that the message of our book is not exaggerated or misinterpreted. We will try to describe in later chapters when our argument applies and when it does not. Furthermore, it’s important to note that not all effects that have been demonstrated in this or related fields will be affected by the new information environment. Consider, for example, the power of default choice. Fewer than 5 percent of people in Denmark choose to donate organs after they die, as opposed to 99.91 percent of the French. Is it that the French are more altruistic? No. As Eric Johnson and Daniel Goldstein showed, what determines this is the default choice. If you want to donate your organs in Denmark you have to be proactive. In France you don’t have to do anything.11 In recent years these types of effects have been used in clever ways to influence behavior, as described in the book Nudge, by Richard Thaler and Cass Sunstein. We do not expect the advantage of defaults and status quo to be significantly affected by the trends we describe here.12
IT’S ABOUT TECHNOLOGY
There are important insights that are based on robust research that simply don’t apply in the evolving consumer environment as much as they did in the past. Consider anchoring, which is one of the best-documented, most robust judgment phenomena (shown back in 1974 by Tversky and Kahneman).13 This effect can be very relevant to consumer behavior. For example, participants in an experiment that Itamar ran with Aimee Drolet, from the University of California, Los Angeles, were asked to write down the last two digits of their Social Security number. Next, they were presented with a picture of a toaster, and were asked to assume that the number they had just written down was the price in dollars of the toaster. Would they pay that amount? Some said yes, some said no. Next, participants were asked to write down the highest price they would be willing to pay for this toaster. Remarkably, people’s decisions were clearly affected by the random number: Those with Social Security numbers that ended with 50–99 were willing to pay, on average, about $10 more than people whose Social Security number ended with 00–49.14
In the past, anchoring and so-called “reference prices” were often used to show how marketers can influence consumers’ choice. There’s a story, for example, about a manager of a store that sold Brunswick pool tables who conducted a little experiment. One week he directed customers who came into his store to the least expensive table first. The following week he started with the most expensive table first. The average sale on the first week was $550. On the second week it was $1,000.15
Anchoring is still a robust effect, though “noise” and the availability of multiple anchors can decrease the effect of any one anchor that a marketer may offer. In fact, price anchors can often favor lower prices, especially when price search engines display the lowest prices first, in which case these prices are more likely to serve as anchors that determine how higher prices are perceived.
The shift from relative to absolute derives largely from the new technologies and their effect on decision making and not from some advancement of our brain. There’s a lot of talk these days about “the new consumer”—a smarter, skeptical person who’s immune to marketing. We don’t buy that view. People are fundamentally the same as they were fifty years ago and will be fifty years from now. They are becoming less susceptible to marketers’ influence not because they are smarter or more logical. It is tools like the ones we mentioned earlier (and will discuss later) that are changing things (advanced search engines, reviews from other users, unprecedented access to experts, easy access to friends and acquaintances). This is important because, as we will show later, in the absence of such tools, relative thinking will prevail.
What’s the main takeaway from this chapter? As you hear about fascinating findings about consumers’ “irrationality,” we suggest that you take them with a grain of salt. While lab experiments can demonstrate some neat effects, these experiments often depend on the researcher having full control over what participants see, which is radically different from today’s shopping reality. There is very little control over what people see when they shop online, and as smartphones are increasingly used by consumers at brick-and-mortar stores, the gap between the lab and reality gets even wider. Over-extrapolated examples that portray the consumer as an irrational and bendable Gumby will probably continue to pop up in the press. There’s always demand for the surprising and counterintuitive. But as more and more people take advantage of new tools, marketers start to realize that Gumby has a spine. Consumers are far from being as susceptible to influence as they are being portrayed.